Jump to:

Women's Collection from Marketing to Counter-Marketing

Appendices to Statement of R.J. Reynolds Tobacco Company.

Date: 19 May 1997
Length: 161 pages
520864399-520864559
Jump To Images
spider_rjr 520864399_4559

Fields

Type
REPORT
Author
Rjr
Recipient
Ftc
Pitofsky, R.
Steiger, J.D.
Starek, R.B. III
Azcuenaga, M.L.
Varney, C.A.
Copied
Blynn, G.M.
Named Person
Audits & Surveys Worldwide
George Washington Univ
Weiner, R.
Beales, J.H. III
Fda
Pierce
Ftc
Nhis
Rjr
Univ, O.F. Mi
Brand
Camel
Non-RJR Brands

Document Images

Text Control

Highlight Text:

OCR Text Alignment:

Image Control

Image Rotation:

Image Size:

Page 1: ozh90d00
52086 4405 V A1'tl'd%0/ B3/tl3800001 I
Page 2: ozh90d00
a 52086 4400 -%I-
Page 3: ozh90d00
m 5aMcP 4yD3A
Page 4: ozh90d00
lOdd`"S'l0m CONFIDENTIAL APPENDICES TO STATEMENT OF R.J. REYNOLDS TOBACCO COMPANY „-: Chairman Robert P'itofsky' tommmsioner jane[ u:. aZeiger Commissioner Roscoe B' Starek~ III Commissioner Marv L."Azcueriaaa , Comrmssioner Christine A. Varney May 19, 1997
Page 5: ozh90d00
APPENDIX B THIS EJOIIBIT HAS BEEN PROVIDED TO THE COMbIISSIONERS
Page 6: ozh90d00
B5/19/1997 11:34 2126272634 AUD11S 7L SUKVtyy 1N1: rwuc uv ./ DESCRIPTION OF METHOD (CONTINIIED) • The interview continued with one age-elilpible respondent in either age group; if more than one was present, a programmed selection procedure was used to randomly select a single respondent to be interviewed. • The continuation of the interview with the selected age-eligible respondent then proceeded according to the procedure described for Stage I. • The overall incidences for the study in total are computed from the allocation of total contacts calculated for Stage I, plus the actual total contacts by age group obtained in Stage II. Total 12~2Q 12:1Z 18_2Q Total Completes 379 157 216 Age-Fli¢ible T rminates Don't Smoke 2,625 2,249 376 Smoke/Don't Buy 2(?2, 1QL 11 Total 2,827 2,440 387 Total Age-Eligibles 3,206 2,597 603 No 12-20 Year Old in Household 17,361 Total Household Contacts 20,567 Net Household Incidence 1.8% In idence % of Aqg F~{T gLbIC3 Total Smokcrs 18.1% 13.4°/. 37.6% Completes 11.8% 6.0% 35.8% 3
Page 7: ozh90d00
? t7w 1w 17JI 11.JV LILCL/LOJV NUyliJ ayi~ DESCRIPTION OF METHOD (CON1'1NUED) • Since the original age screening asked only for persons 12-20 years old - and not for 12-17 and 18-20 separately - it was decided to recontact those age-eligible respondents who did not qualify on the smoking aod/or purchase questions, i.e., age-eligible "tamninates," to ascettain the specific a8e group of those persons. This reoontact "poor consisted of: Don't Smoke 1101 Smoke But Don't Buy Own 95_ Total 1187 Recontacts were completed with 1006, or 84.8SG, of the available pool of age- eligible termiaatea. • These rocontact results were then used to estimate the proportions of 12-17 and 18-20 year oldv within the total group of ts®inated age-eligible (12-20 years) interviews; these allocations were used as the bases for calcalating incidence within each age group. For purposes of this study, incidence is calculated at two levels: 1. Total Smokers =% of Total Age-Eligibles Who Smoke, Regardless of Whether They Buy Their Own Cigarettes 2. Completed Interviews =°/. of Total Age-Eligibles Who Smoke And Buy Their Own Cigarettes intervie.rin=_StaBe Q fMav 1- 18. 19971 • At the time it was decided to comtiaue inttsvieariaB toanrd a larger samplK the household screening was:evised to determine... • The preseace of sny persona 12-17 yars of a8e, and... • The prescnce of any persons 18-20 years of age. 2
Page 8: ozh90d00
Teen Smoking - Page 2 their own best interests until they reach a certain age undergird a wide variety of restrictions on the choices teens can make, ranging from voting to driving to alcohol and tobacco consumption. Despite widespread prohibitions on the sale of cigarettes to teens below a certain age,z many teenagers experiment with smoking. In 1989, 47.5 percent of 17 year olds reported that they had smoked at least one cigarette, and 24"/e had smoked in the previous 30 days. Experimentation is strongly related to age. Only 11.1 percent of 12 year olds had ever smoked a cigarette, with 2.3 percent having done so in the previous 30 days. The laws, however, seemingly restrain smoking to some extent. At 18, 57 percent had smoked at least one cigarette, and 30.5 percent had smoked in the previous 30 days.' Moreover, studies of newly passed laws and enfo(cement campaigns for existing laws indicate that legal sanctions can reduce teenage smoking (Jason et al 1991; Feighery et al. 1991). This study examines the determinants of the decision to smoke in two large, independent cross sections of teens. One sample is national, and the other is confined to California teens. The national sample is based on the Teenage Attitudes and Practices Survey, conducted for the Centers for Disease Control in 1989. The California sample is based on the 1990 California Tobacco Survey, designed and conducted for the California Department of Health Services. Section 11 considers the various factors likely to influence teenage smoking decisions. Section III describes the data and the empirical 2Forty-four states and the District of Columbia prohibit sales of cigarettes to those under 18. Three states prohibit sales to 19 year old teeas, one state prohibits sales to 17 year olds, and 2 states have no restrictions. Tobacco Institute, State Minors Laws (October 16, 1992). 'Calculated from TAPS, the Teenage Attitudes and Practices Survey. The data are described more fully below.
Page 9: ozh90d00
Table I Perce.tage Distri6 aon of Soolti.S Status by Age National and California Sampks Age Group Smoki Status <=13' 14-15 16-17" 18-19 Total Natioeal Sample Non-smokers 86.2°/. 68.8% 55.1% 43.5% 65.6% Experimenters 10.1% 18.2% 23.5% 26.5% 18.8% Social Smokers 3.3% 9.8% 12.5% 14.4% 9.5"/0 Daily Smokers 0.4% '3.2% 8.8% 14.7% 5.8% Sample Size 2477 2558 2617 1390 9042 California Sa.pk Non-smokers 89.0% 72.7% 55.6% - 72.6% Experimente[5 7.7% 18.1% 26.2% - 17.3% Sotial Smolcets 3.1% 7.2% 12.0% - 7.4% Daily Smokers 0.2% 2.0% 6.2% - 2.8% Sample Size 1673 1678 ... 1629 - 4980 * National sample includes 81 teens age 11. No I I year olds in California data. ** California sam le includes 25 teens a 18.
Page 10: ozh90d00
Teen Smoking Page 6 direct data on such important variables as peer smoldng behavior and teen perceptions of the benefits and risks of smoking. Behavioral studies attempting to link advertising recognition and smoking behavior have suffered from reverse causality, since smokers may have more interest in cigarette advertising, and have generally not examined advertising measures within the context of other faaors that in8uence the decision to smoke (e.g., Goldstein et al. 1987). The present study seeks to fill these gaps. lll. DATA SOURCES AND EMPIRICAL METHODOLOGY A. Data Sources Two large surveys of teenagers provide the basic data sources for this study. Both surveys were household based, including teens who are not in school as well as students. The Teenage Attitudes and Practices Survey (TAPS), sponsored by the Centers fbr Disease Control, obtained data on t.eenagers' knowledge about and use of tobacco products in 1989. The nationally rept esentative sample was drawn from households that participated in the National Health Interview Survey during the second half of 1988 and the first half of 1989, and included data on 9,965 teens, 82.4 percent of teens in those households.' The sample consists primarily of 12 to 18 year olds, although about 2 percent of the observations are I I and 19 year olds. Telephone interviewing was conducted during the last quarter of 1989 (Morbidity and Mortality Weekly Report 1991). The survey used a mail questionnaire to obtain data from an additional 830 teens, but differences in the questionnaires made these obsetvatiote utnsabk for the pritnary analyses. M'issing data reduced the available sample size to 9,042. 'Participation in the original survey from which the sample was drawn was 95%.
Page 11: ozh90d00
Teen Smoking Page 7 The second data source was a survey of 5,040 California teens conducted in 1990, commissioned by the California Department of Health. Households were sampled by stratified random-digit dialing, and all teens between the ages of 12 and 17 were interviewed. The response rate was 78.4%. Telephone interviews were conducted between June 1990 and March 1991 (Pierce et al. 1991).' Missing values reduced the available sample size to 4,980. Both surveys included a variety of virtually identical questions on the perceived benefits and the perceived health risks of smoking. Moreover, both surveys obtained comparable data on peer and family use of cigarettes. In what follows, differences in question wording and the data collected are noted where they are potentially relevant. The sample designs make weighting of the data appropriate for population statistics. If the same model fits all individuals, however, weighting is unnecessary for the logistic models. The primary analysis therefore focuses on unweighted models. Weighted models were also estimated, and differences are noted where they are significant.' The high degree of comparability between the surveys made it possible to employ essentially identical models in two independent samples. All testing of relevant variables and model formulations was done using the California sample. The resulting final models were then estimated using the larger national sample from TAPS, with very slight changes in formulation necessary to accommodate the 'The California study used a different survey instrument to obtain data from 24,296 adults. Unfortunately, lack of comparable questions precludes using the same model to examine adult smoking decisions. The adult data can be used for other comparisons, however, as noted below. 'The weights in the data sets are designed for estimating population totals. They were used directly to estimate population statistics (e.g., percentages of smokers). For logistic models, the weights were normalized by dividing by the sample average weight. Thus, the avenage weight is one.
Page 12: ozh90d00
Teen Smoking Page 5 sale of tobacco to those under 19 is illegal imposes an additional cost of trial. Again, the cost of trial would be lower for teens with friends who smoke. The availability effect suggests that peer influences should be greater on the decision to try a cigarette than on decisions to smoke more intensively. The role of advertising in teenage smoking decisions has been controversial. Many have argued that advertising influences teens' preferences, and thereby encourages them to smoke when they would not otherwise do so (e.g, Surgeon General, 1994). If so, teens exposed to more advertising should be more likely to try cigarettes, and should smoke more intensively. If, however, advertising operates by providing brand-specific infonnation, such as taste, tar and nicotine content, or simply as a signal of quality (Nelson, 1974), there is no clear theoretical reason to expect that it should increase demand for the product. Teens are likely aware of cigarettes , from other sources of information, including anti-smoking programs and the behavior of those around them. If so, advertising provides no new information about the product itsel/y and should have no influence on the decision to smoke. Exposure to advertising should therefore have no influence on either trial or smoking intensity. Numerous econometric studies have examined the relationship between cigarette advertising and consumption (e.g., Schmalensee 1972; Hamilton 1972; Schneider, Klein and Murphy 1981; Bishop and Yoo 1985). Although results are mixed, these studies have generally found little or no effect of advertising on total cigarette consumption. Because changes in adult smoking behavior are likely to dominate aggregate consumption statistics, however, such studies are insensitive measures of the effect of advertising on teenage smoking decisions. Economic studies that have examined the smoking behavior of teens directly (Lewitt, Coate and Grossman 1981) have lacked
Page 13: ozh90d00
Table 3 Simple Statistics for Selected Advertising Meatares StaMard Mean Deviation Minimnm Maximum I Brand Advertieing per Honsebold Marlboro. 3 moMAs 0.27 0.07 0.10 0.36 Marlboro, 6 months 0.53 0.12 0.23 0.68 Marlboro. 3 years 4.02 0.94 1.94 5.02 Camel, 3 mon{hs 0.23 0.07 0.12 0.40 Camel. 6 months 0.47 0.13 0.25 0.74 Camel, 3 years 2.87 0.85 1.28 4.06 Industry Advertising per Honsehold 3 months 1.69 0.38 0.99 2.46 6 months 3.15 0.63 1.79 445 12 months 6.19 . 0.98 3.91 7.57 3 years 21.99 4.14 13.38 27.19 AdverBsing IAentificatioa Variabla IdentiCp Camel 0.29 0.45 0 1 I Identify Marlboro 0.43 0.50 0 1
Page 14: ozh90d00
Teen Smoking Page 4 consider the costs of quitting if they later change their minds, particularly if they think smoking is dangerous. Thus, teens who believe smoking is dangerous would be more likely to smoke if they also believe quitting is easy. Attitudes toward risk are also likely to influence smoking decisions. Given the perceived benefits and costs of smoking, teens who are more risk averse would be less likely to smoke at any given level of smoking intensity. Teens with greater tolerance of risk, however, would be more likely to smoke, and to smoke more intensively, other things equal. The choices of utility ma)omizing teenagers regarding smoking will also be affected by the behavior of others around them if teens are subject to peer pressure. Acquaintances, close friends, and family members all constitute potentially relevant peer groups that could plausibly influence teenage smoking decisions. - If individuals gain utility from behavior that is similar to the behavior of their peers, the incidence of smoking among others will influence the utility of smoking That is, the desire to be part of a group may irKxease the utility of behaviors associated with that group. Teens with many friends who smoke would therefore be more likely to be smokers, other things equal, than those with no friends who smoke.' Such interdependent preferences could also arise if it is more pleasurable to smoke with others than it is to smoke alone. Behavior of peers could also influence smoking decisions by changing the cost of smoking. If friends smoke, trial is presumably cheaper than it is for a teee who knows no smokers, since cigarettes can be obtained one at a time irtctead of in packages Moreover, the fact that in most states SFor a formal treattnent of peer pressure as a factor motivating individual effort within a firtn, see Kandel and Lazear (1992).
Page 15: ozh90d00
Teen Smoking Page 10 in the past 30 days (zero otherwise), and daily smoker is one for teens who are daily smokers. Most of the model development work with the California data was done using the dichotomous model with current smoking as the dependent variable. . A limitation of the ordered logistic model is that it takes no account of the effeyt of information acquired from trial on•decisions to smoke more intensively. To some extent, that . aWience is re9axed in the measured benefits and costs of smoking. That is, teens who have tried smoking may, because of their experience, give different answers to the questions about the benefits of smoking. There are, however, other aspects of the utility of smoking that are not measured.. If the experience gained from trial typically increases assessments of those beneSts, teens would be, more likely to progress to more intensive smoking than the ordered logistic model'would predict., , If experience reduces the perceived benefits (or increases perceived costs),_however, the ordered model will overpredict higher levels of smokirig coinpared to the dichotomous models.- -- •' Table I gives the sample distributions of smoking status. The higher incidence of smoking in the national sample, particularly daily smoking, reflects in part t6e fact that the California sample exchides 18 and 19 year olds. In the national sample, 15.9 percent of 18 and 19 year olds are daily smokers, substantially more than the 4.8 percent of 12-17 year olds. Nonethehxs, among comparable age groups, smoking, aad particularly daily smoking, is less common in California than in the national sample. In California, 2.4 percent of 12-17 year olds are daily smokers. Similarly, 74.9% of this age group in California has never smoked a cigarette, compared to 68.3°/% nationally.
Page 16: ozh90d00
Teen Smoking Page 3 approach. Section IV presents and discusses the results of the statistical analysis. A final section offers some conclusions. II. UTILITY MAXIMIZING SMOKING DECISIONS Like any other consumer behavior, utility maximizing decisions about smoking require a balancing of perceived costs and benefits. The benefits of smoking are the individual's subjesxive evaluation of the pleasures of the habit. The costs of smoking include the cost of purchasing the product and the perceived health risks that consumption entails.' Presumably the marginal benefitg of consumption increase with the quantity consumed, but at a diminishing rate. The marginal cost of consumption increases at a constant rate because of the price of the product, and at an increasing rate if the perceived risks of smoking increase with the quantity consumed. A utility maximizing teenager will choose to smoke if the benefit of smoking exceeds the cost. We can think of the difference between benefits and costs as the perceived net utility of smoking (which may be negative). A teenager's net utility from smoking will increase with his perceptions of the benefits of smoking, and decline with increases in the perceived risk of smoking. Utility nmannnaang teens who believe that smoking offers benefits are therefore more likely to smoke than those who do not perceive benefits. Moreover, they are also more likely to smoke more intensively. Conversely, utility maximizing teens who believe that smoking poses greater risks are less likely to be smokers, and are less likely to smoke intensively. Utility maximizing teens will also 'The analysis here takes consumers' perceptions of the risks as given, and does not inquire into dhair aoauscy or rationality. Whatever the actual level of risk, it is perceived risk that will influence eorsurqxion decisions. Regarding cancer risks, the evidence suggests that perceived risk is greater than actual risk, especially among younger age cohorts (Viscusi 1992, 1991).
Page 17: ozh90d00
Table 10 Effects of Brand Advertising Mwern os Probability of Smoking Alternative Dictatomons Definitiass of Smoking Coeificients and slandard ermrs - Ad Measure Ever Smoked (lrreat Smolcer Dail Smoker E Camel, 6 months 0.45 . -0.03 -0.77 (.33) (.51) (1.04) Marlboto, 6 months -0.51 0.58 0.95 (.35) (.54) (1.06) 1 Mcasun Camel, Right 0.04 -0.05 -2.27 (.15) (.26) (.72) Camel, Wrong -0.05 0.17 -0.13 (.12) (.19) (.37) Marlboro, Right 0.17 • 0.21 -0.12 (.10) (.16) (:32) ` Marlboro, Wrong 0.40 ••• 0.31 0.28 (.13) (.20) (.36) Akaike Info. Criterion 4291 4283 2067 2068 . 630 618 AIC witA no ads 4290 2064 627 •••S*ifiC".01. •• S' M0.05. • S' M0.10.
Page 18: ozh90d00
THE DETERMINANTS OF TEENAGE SMOKING BEHAVIOR J. Howard Beales III' Abstract: This paper uses both ordered logistic and dichotomous logistic models to examine the determinants of teenage smoking behavior in two independent samples of California teens and teens nationwide.. Teens' decisions are consistent with rational choice. Measures of peer smoking behavior, especially the incidence of smoking among best friends of the same sex, are the most significant determinants of the probability of smoking, followed by measures of the benefits and risks of smoking. Family influences, particularly the incidence of smoking among older siblings and adults other than parents, are also significant, but less important quantitatively. In the California sample, neither industry advertising expenditures nor advertising for the brands that are most popular among teens has any detectable influence on the probability of smoking. I. INTRODUCTION Why do teenagers smoke cigarettes? As the incidence of smoking in the population as a whole has declined, debates about smoking prevention policies have increasingly focused on these youngest smokers. Much of that debate has ignored the role of economic behavior by teens, assuming instead that teenage smokers are passive victims of social forces over which they have no control. From an economic perspective, however, teens' decisions about smoking are conceptually no different frorh any other consumption decision, involving a balancing of the costs and benefits of the activity. There are, of course, ample grounds for questioning the maturity of teenagers judgements about a wide variety of behaviors. Legislative decisions that teenagers cannot be expected to pursue 'Associate Professor of Strategic Management and Public Policy, The George Washington University. The views expressed in this article are my own. The statistical analysis was largely conducted in connection with a consulting project for R_ J. Reynolds Tobacco Company to analyze several articles concerning the relationship between Camel advertising and teenage smoking that originally appeared in the J The article was drafted ittdepadartly, and was not reviewed by Reynolds prior to submission. I wish to thank Theresa Burke and Robert Femli for their research assistance, and Timothy Muris for many helpfid suggestions. Responsibility for any remaining errors is of course my own.
Page 19: ozh90d00
Table 8 EfTects of Indust Advertisiu on Teew Smobi Variable Smoki.g Ever Smoked Clrreat Smoker Dagy Smoker Ad S per household c«fw;ai a,;k< cceffiaar Ataam coefficca nk,acc c«fi~* sk.Ac over: (Sit F..a) 4nkrMUim (Std Fi'a) infarmtiboo (s1d E~) Mtamatim Sol F~) tdam.6m No ad.a 6255.74 4289.85 2063.87 627.12 3 month -0.05 6257.47 -0.10 4290.79 0.04 2065.81 -0.004 629.12 (.09) (.10) (16) (.30) 6 month -0.02 6257.60 -0.05 4291.09 0.07 2065.38 -007 628.97 (.06) (.06) (.09) (.18) 12 montht 0.01 6257.65 -0.003 4291.84 0.07 2064.70 0.002 629.12 (.04) (.04) (.06) (.12) 3 years 0.004 6257.55 0.001 4291.83 0.016 2064.76 0.007 629.06 (.009) (.(i09) (.015) (.029) No oodFicieals are si ~ at 25 roent.
Page 20: ozh90d00
Teen Smoking - - Page 12 Table 2 presents the means and standard deviations of the independent variables employed in the analysis in both the California and national samples:" The perceived benefits of smoking are measured by four dummy variables, equal to one for teens who think that smoking helps when bored, smoking helps to relaz - smoking helps with stress, and smoking helps in social situations." All should have positive sigas. An additional variable, "enjoy risky activities," measures attitudes toward risk. Teens who like to do dangerous things should be more inclined to smoke, other things equal, so its coefficient should be positive.16 Three dummy variables measure the perceived risks of smoking. "Occasional cigarette harmful" is one for teens who believe that an occasional cigarette is hannful, and should have a negative coefficient. "Safe to smoke" is one for teens who think it is safe to smoke for a year or two; its coefficient should be positive. "Dangerous, but I can quit" is one for teens who think that smoking for a year or two is dangerous, but also think they can quit any time they want. In essence, this formulation lets attitudes toward ability to quit smoking effect decisions "Detailed definitions of these and other independent variables, including the specific survey questions and differences (if any) between the California and national surveys, are listed in Appendix Table A-1. "In developing the model, a dummy variable for teens who think that smoking helps control weight was consistently insignificant for both boys and girls, and was dropped from further analysis. 16nEnjoy risky acYivities" is equal to 1 for teens who disagree that they lihe to do dangerous things occasionally (risk averse teens), 2 for teens who neither agree nor disagree (risk neutral), and three for those who agree (risk lovers). Most teens either agree or dissgee; the variable is equal to two for only about 8% of each sample. In California models with wrrent smoking as a dichotomous dependent variable, separate dummy variables for teens who agreed and disagreed produced essentially the same results. The single variable formulation fit the data better as judged by the Akaike Information Criterion.
Page 21: ozh90d00
Table 5 CALCULATED PROBABILITY OF CURRENT SMOKING Calculated from the coefficients in Table 4 National Sam Ca6fornia Sam Relstive Relative Probabdi Odds o Probabili Odds o Sample Proportion 15.34% 10.14% At means of all variables 7.40'/0 4.36% BASE: All antismoki answers o0 1.85% 1.09"/0 Risk - Utility Variables Enjoy risky activities 2.37% 1,28 1.36% 1.24 Smoking utility measures 5.63"/0 3.04 2.22% 2.03 Occasional cigarette safe 2.91% 1.57 2.53% 2.32 Safe to smoke year or two 6.12% 3.30 2.70% 2.47 Dangerous, but I can quit 3.55% 1.91 , L64o/a 1.50 Peer Smoking Variables all same sex best frietids smoke'••• 10.89% ' 5.87 4.58% 4.20 all opposite sex best friends smoke 3.13% 1.69 2.09% 1.92 Steady who smokes 4.45% 2.40 2.26% 2.07 Most people you know smoke 3.33% 1.79 3.68% . 3.37 Family Variables Mom & dad smoke 2.23% 1.20 1.57% 1.44 Same sex older sibs smoke 4.31% 2.33 3.04% 2.78 O. sex older sibs smoke 2.66% 1.44 1.44% 1.32 Combinations of Variables . Risk-Utility Variables 20.92% 11.29 6.63% 6.08 Peer Smoking Variables 48.41% 26.11 40.23% 36.87 Famil Variables 7.28% 3.93 5.66% 5.19 NOTE: Table entries rellect the probability of social or daily smoking when a particular variable (or combination of variables) is added to the base case. Cambinations in the last section include all of the variables identified in the corresponding sections of the table. • Calculated probability relative to the base case. Base probability calculated at sample mean income, age, nights aa (and household size in California sample); Enjoy risky activities = 2(risk neutral) and Occasional cigarette dangerous = 1. All other variables = 0. ••• Smoking helps when bored, helps with sirex. helps relax, helps in social situations all = 1. **** Reflects the common eBect of best friends tor buys and ' i added effect for
Page 22: ozh90d00
r Table 2 Sim k Stati>8;ca for E Vatiahles Natio.al Sample St d d CaMor.ia Sample Standard Variable Mean ao ar Deviation Mean Deviation Risk - Utility Variables Smoking helps when bwed 0.069 0.253 0.190 0.392 Smoking helps relax 0.270 0.444 0330 0.470 Snaking helps with stress 0.191 0.393 0.246 0.431 Smoking helps, social situations 0.391 0.488 0.440 0.496 Enjoy risky activitics 1.920 0.953 2.003 0.958 Occasional cigarette hannful 0.667 0.471 0.743 0.437 Safe to smoke (a year or two). 0.070 0.255 0.069 0.254 Dangerous, but I can quit 0.149 0.356 0.116 0.320 Peer Smoking Variables Same scx best fiiends (boys and gids) 0.191 0.308 0.121 0.245 : Added effect for boys 0.095 0.238 0.061 0.185 Opposite sex best friends 0.171 0.293 0.112 0.250 People you know 1.541 0.957 1.066 0.951 Steadv who smokes 0.134 0.341 0.092 0.289. No best friends (Bo)s only) .0.001 0.028 0.005 0.072 Family Variables Adult smokers in home (number) 0.569 0.726 ., 0.446 0,696 Mom Smokes 0.260 0.439 0.177 0.382 Dad smoltes 0.256 " 0.436 '.0.203 - 0.402 No adult in household 0.003 0.057 0.032 0.175 No dad in houselwld 0.209 0.407 0.212 0.409 No mom in household 0.050 0.219 0.071 0.256 Opp. sex older sibs smoke 0.040 0.196 0.048 0.208 Same sex older sibs smoke 0.043 0.204 0.049 0.210 Household size -- 4.491 1.420 Demographic Variables Age 15.070 2.039 14.490 1.699 Nights out (per week) 2.414 1.623 2.454 1.660 Family income 33.027 25.102 42.023 33.262 No income data 0.128 0.334 0.125 0.331 Nonwhite 0.185 0.388 0.306 0.461 Hispanic boys 0.047 0.212 0.132 0.338 Not in school 0.061 0.240 0.024 0.153
Page 23: ozh90d00
Teen Smoking , Page 20 IV. RESULTS A The Decision to Smoke The estimates of the ordered logistic model for the California and national samples are presented in Table 4. The model fits both data sets quite well. Indeed, although the differences are very small, the model fits slightly better in the reserved national data sample. Virtually all of the variables of interest have the expected signs, and most variables are significant at .0001. Moreover, the resuhs are generally quite consistent in the two ssmples. Age is positive and significant in both samples. At the sample mean probability of smoking in the past 30 days, each additional year raises the probability of smoking by .03 in the national sample, or 16.4% (02 in California, 23_7%).' The notion that smoking decisions reflect simply youthfW immaturity seems to imply that younger teens shouldd bee more likely to smoke, other things equal. The data are not consistent with this hypothesis..~ Consistent with the notion that teenage smoking decisions reflect a balancing of perceived benefits and risks, three of the four measures of the benefits of smoking are positive and significant in both samples - smoking helps relax, helps when bored, and helps with stress. The coefficient for teens who think smoking helps in social situations, however, is negative and significant in the Califomia sample, but positive and not significant in the national sample. This result may reflect the fact that the pea variables also measure the social benefits of smoking. That is, the fraction of friends 24fhe derivative of the probability of being a smoker with respect to a variable i is equal to where P, is the logistic coefficient and ai is the probabifity of smoking. Because the sample means diffa, the derivative is evahiated at different points in this comparison. At the national sample mean probability of smoking in the previous 30 days, each additional year of age raises the probability of smolung by.03 in the California sample. At the California mean, the derivative for the national sample model is.02.
Page 24: ozh90d00
Teen Smoking Page 25 absence of friends who smoke would imply, but the effect is only significant in the national sample. A steady boyfriend or girlfriend who smokes also significantly increases the likelihood of smoking The importance of peers is also apparent in the calculated probabilities of smoking in Table 5. In both samples, the behavior of friends of the same sex is the single most powerful variable in raising the calculated probability of smoking. If all friends of the same sex smoke, the likelihood of smoking incrcases almost six times in the national sample, and more than four times in the California sample." In both samples, same sex Sriends are more than twice as important as opposite sex friends. Perceptions about smoking acquaintances are more important in California than in the national sample, raising the likelihood of smoking more than three times. As judged by the relative odds measure, the other peer measures are of comparable importance in the two samples. Taken together, the powerful influence of the peer variables is apparau. ~. For girls, the probability that a feen is a current smoker if all fiiends smoke; the teen thinks most people her age . smoke, and has a steady who smokes is 48 percent in the national sample, and 40 percent in the California sample. At age 16, such a teen is more likely to be a current smoker than not. Smokers in the household also have a significant effect on teenage smoking behavior. The more adult smokers in the household, the more likely a teen is to snoke. The coefficient is larger and more significant in the California sample, perhaps because the model includes the total size of the household as an additional independent variable. That information was not available in the national sample. Older siblings who smoke increase the likelihood of smoking, and, as with 8ieWs, the effect "Part of the difference between California and the national sample ntay be due to the differetce in the questions asked. The national data identify the fraction of the teen's four best friends who smoke. The California data allow for up to 99 "best" &iends. The national fonnulation may focus more dosely on the friends most likely to affect the decision.
Page 25: ozh90d00
Teen Smoking Page 24 probabilities are uniformly lower in California, but the relative importance of the variables is quite similar.Jz Measures of the perceived benefits and costs of sipoking are important. Perceived safety is most important, particularly in the national sample. Moreover, the probability of current smoking is approximately 25% greater for teens who enjoy risky activities. In combinatibn, the risk utility variables are more important in the national sample, with relative odds of just over eleveq compared to six in the California sample." In both samples, the coefficients in Table 4 reveal that peers have a powerful influence on smoking decisions. Perceptions that more teens of about the same age smoke are associated with higher probabilities of smoking ("People you know"), and the fraction of best friends who smoke , increases the likelihood as well. The effect of best frierds of the same sex is significantly greater than best friends of the opposite sex (XZ = 65.29 in the national sample; 9.71 in the California sample). Moreover, the influence of best 5iends of the same sex is significantly greater for boys than for girls, contrary to the suggestions of some that girls are more influenced by peer pressure (van Roosmalen and McDaniel 1992, 1989). Boys who have no best male friends are more likely to smoke than the 'ZThe difference between California and the national sample is largely accounted for by differences in the means of the variables, and the absence of family size data in the national model. Evaluated at the mean of the national data, and setting household size = 0(the implicit restriction in the national model) the base case smoking probability in California would be 1_66%, much closer to the national figure of 1.85%. Setting household size equal to the California mean (4.49), however, reduces the calculated base probability to 1.I9Y°. The relative odds are less affected by the differences in the models and the means, and are therefore more comparable between the samples. "When the California survey was conducted, an extensive anti-smoking advertising campaign was also underway. If the campaign influenced teens to give "socially acceptable" answers to questions about smoking's risk and utility, it may account for the lower apparent significance of these variables in the California sample.
Page 26: ozh90d00
Table 9 Effects of Brand Advertising )1ltaf.res oa Probability of Smolcing Orde.ed Lo istic Model g Coefficients and standard errors Advertising Equation Measure (1) (2) (3) (4) (5) (6) (7) Exoendinue Meawres Camel, 6 months 0.27 0.27 0.05 (.30) (.30) (.39) Marlboro. 6 months -0.22 -0.24 -0.02 (.32) (.32) (.40) Camel, 12 months 0.12 (.18) Marlboro, 12 months -0.04 (.20) Camel, 3 years 0.03 (.10) - Marlboro. 3 years -0.01 (.09) Identification Measures Identify Camel -0.04 -0.04 (.10) (.10) Identify Marlboro 0.23 .•. 0.23 ••, (-019) . ..(..09) Camel, Right -0.09 -0.08 (.16) (.15) Camel, Wrong -0.02 -0.02 (.11) (.11) Marlboro, Right 0.17 • 0.17 • (.10) (.10) Marlboro. Wrong 0.38 ••• 0.38 .13 (.12) Akailce InCo. Criterion 6259 6259 6260 6250 6251 6247 6247 Note: AIC with no advertising mrasurcs = 6256. •••s' ' K.nro.oi. •• sitwr.moato.05. • sipick~.ato.w.
Page 27: ozh90d00
Teen Smoking Page 8 differences in the two surveys.9 Thus, the reported probability levels for statistical inferences from the national sample in fact conform to the oft-violated assumption that the model is specified independently of the data it seeks to explain. The high degree of correspondence between the California and national results gives considerable confidence in the underlying model. .,. @. The Order?d I.o¢istic Model The analysis examined smoking deasions using an ordered logistic modd.10 In the model, net expected utility (or propensity to smoke) is viewed as a function of variables measuring teens' perceptions of the benefits of smoking, the risks of smoking, the behavior of their peers and other family members, and demographic variables. Although net expected utility is unobservable, it is related to observed smoking status. As net expected utility from smoking increases, and exceeds some threshold value, an individual switches from being a nonsmoker to trying a cigarette. Further increases in net expected utility beyond a second threshold lead the individual to become a current smoker, and increases beyond a third threshold lead him to become a daily smoker. The ordered logistic model uses data on ordered categorical responses to estimate the effects of independent variables on the unobservable net utility of smoking (Amemiya 1981; Kmenta 1986, Ch. 11). 9A handful of theoretically interesting variables that proved insignificant in developing the California model were added back into the national models using dichotomous dependent variables to verify the generality of the insignificance result. These results are noted where appropriate. 1° California results using an ordered probit model are extremely similar and are not reported.
Page 28: ozh90d00
Teen Smoking : Page 17 advertising for the two brands that dominate the California teenage market (Marlboro and Camel) was examined.M To control for differences in the level of expenditures that stem solely from differences in the size of the media markets, advertising expenditures were divided by the number of households in the corresponding media division. Teens were matched with the advertising data based on their county of residence (and zip code where a county was split between two different media divisions) and the month in which they were interviewed. In different formulations, advertising expenditures over periods ranging from the prior three months to the prior three years were considered." The lack of geographic information meant that no comparable analysis was possible with the national sample. zs(...continued) market advertising. Reynolds allocates national advertising (such as magaane ads) to local markets based on circulation where possible and population where local circulation figures are unavailable. Out of home advertising includes billboards and transit advertising. "Special markets" is a relatively small category including, for example, Spanish language advertising. The data do not include event sponsorships, which are more national in character, or promotional spending, which is not tracked geographically. Nationally, promotional spending is dominated by couponing and "free" items provided with cigarettes. Analytically, such spending is more akin to a price cut than to advertising. 'Artwng California teens who buy their own cigarettes, 63% buy Marlboro and 22% buy Camel. In the TAPS sample, Marlboro's share of buyers is 68% and Camel's is 8%. ='Interviews were conducted over a 10 month period_ For a teen living in media division j who was interviewed on day d of month m, the L month advertising measure was defined by 41 .x°.~- ~1 f t t~~ w~t°JJ°y), where length. is the number of days in month m, and ar is advertising expenditures in region j in month t. This formulation assigns advertising data for the current month on the assumption that ntperdiatres occur at a constant daily rate within the month. For example, in constructing the three month advertising measure, a teen in a particular region interviewed on the tenth of a 30 day month was assigned one third of that month's advertising, plus the two prior months' advertising, plus two thirds of the advertising expenditures in the third month prior to the interview.
Page 29: ozh90d00
Teen Smoking Page 26 is greater for siblings of the same sex. Opposite sex siblings are only significant at .06 in the California sample. The influence of parents operates in part through the variable measuring the number of adults who smoke and in part through the separate dummy variables for a smoking parent (mom smokes, dad smokes). The negative coefficients on the parent dummies indicate that a parent who smokes iirreages the likelihood of smoking by less than another adult who lives in the household, although the effect is only sigm&ant in the California sample. The net effect of a smoking parent, the sum of the smoking parent dummy and adult smoker coefficients, is positive and significant in either sample. Parents, however, may be less influential in smoking decisions that other adults in the household. Wang et al. (1995) also noted the limited importance of parental smoking in the TApS sample. Indeed, the absence of a parent is a more significant factor, increasing the likelihood of smoking than is the presence of a smoking parent. In both samples, teens from households with one parent absent are significantly more likely to smoke. Teens in households with no adult are less likely to be smokers in California, but more likely to be smokers in the national sample,'S a difference that may stem from the different definitions employed in the two samples. The in9uence of fanrily income is positive in both sampks but only significant in the California sample.m Apparently smoking is a normal good, at least for teenagers." In both samples smoking 'sin the weighted national sample, no dad and no adult are not significant 31.10. "This may be due to the fact that the national sample groups all incomes over $50,000 into one category, which accounts for 21 percent of the observations with income data. Although the California survey used fewer categories, the open-ended top category uses a higher income (over $75,000), and accounts for 16.8°/6 of the sample. Thus, there may be more noise in the national income data. (continued...)
Page 30: ozh90d00
Table 4 ORDERED LOGISTIC MODELS OF SMOKISYG DiTENSITY Natio.al SarFle Catifar.ia Smpk Variable Coc6cient (Std. Ermr Coefficient Std. Error Intetoept (Daily smokers) -9.02 (.25) •° -10.08 (.43) ••• lnteronpt (Social smokers) -7.24 (.24) ••• -8.08 (.42) ••• Intenzpt (Experimenters) -5.50 (_23) ••• -6.28 (.41) ••• Risk - V3tiGty Variables Smoking helps when bored 0.32 (.09) 0.24 (.09) ••• Smoking helps relax 0.39 (.06) ••• 0.37 (.09) ••• Smoking helps with stress 0.36 (.07) ••• 037 (.09) ••• Srnohing helps, social situations 0.08 (.05) -0.26 (.08) ••• Enjoy risky activities 0.25 (.03) 0.22 (.04) ••• Occasional cigarette harmfol -0.46 (.05) -0.86 (.08) ••• Safe to smoke (a year or two) , 0.78 (.09) ••• 0.06 (.13) Dangerous, but I can quit 0.67 (.06) •« 0.41 (.10) Peer Smoking Variables Same sex best friends (boys and girls) 1.97 (.11) ^• 1.47 (.18) Added eH'ect for boys 0.35 (.12) ••• 0.60 (.21) ••• Opposite sex bat friends 0.54 (.10) ••• 0.66 (.15) ••• People you know 0.20 (.03) ••• 0.41 (.04) ••• Steady who smokes 0.90 (.07) ••• 0.74 (.11) No best friends (Boys only) 1.74 (.77) •• 0.65 (_43) Family Variables Adult smokers in home (number) 0.23 (.11) •• 0.53 (.11) ••• Mom Smokes -013 (_ 12) -0.35 (.15) •• Dad smokes -0. (4 (.12) -0.34 (.14) •• No adult in household 0.75 (.38) •• -0.56 (28) •• No dad in hoosehold 0.14 (.07) •• 0.35 (.10) No mom in household 0.25 (.12) •• 0.35 (.17) •• Opp. sex older sibs smoke 0.37 (.11) ••• 0.28 (.15) • Same sex older sibs smoke 0.87 (.11) ••• 1.04 (.14) ••• Household size - -0.08 (.03) ••• Demographic Variables Age 0.19 (.01) as• 0.26 (.02) ••• Nights out (per week) 0.10 (.02) ••• 0.09 (.02) ••• Family income 0.002 (.001) • 0.006 (.001) ••• No income data 0.09 (.09) 0.30 (.14) •• Nonwhite -0.84 (.08) -0.36 (.09) ••• HLVamc boys -0.44 (.12) ••• 0.36 (.11) ••• Not in school 0.54 (.09) 0.46 (.20) •• Memra of Orera(1 Fit Chi Square Statistic 4776.595 2048.956 (pvaloe) (.0001) (.0001) McFadden's R2 0. 2699 0.2487 Concordant nse, idion 82.4"/. 80.2% •••si at0.01. *• Si ato.05. 0 S ato.1o.
Page 31: ozh90d00
96-30. "Transactional Uncertainty and Plural Forms of Governance" by Joseph P. Cannon, Ravi S. Achrol and Gregory T. Gundlach. 96-31. "Forms of Network Organization" by Ravi S. Achrol. 96-32. "The Profitability of Small Traders in Futures Markets: A Review of Research" by James V. Jordan and David I. Meiselman. 96-33. "Capital Flows and Financial Stability: The Mexican Experience" by Jiawen Yang. 96.34. "The Determinants of Teenage Smoking Behavior" by J. Howard Beals 111. 3
Page 32: ozh90d00
Teen Smoking Page 16 and for nonwhites. "Nights out" measures the number of nights.per week that the teen goes out for fim or recreation. Teens who go out often may be subject to less parental supervision, or they may be more influawed by their peers. Either explanation suggests a positive coefficient. The model does not include education. Neither the parent's highest education level achieved nor the teen's education level weno.ver very close to statistically significant, and signs were not consistent across models.x The model does not include a price variable. Within California, there are no local taxes that would generate price variation, so the exclusion of price is unlikely to influence.the results. Price obviously varies in the national sample, but the only. geographic information available on the respondent is census region. Although the national model is misspecified because of the absence of price, the high degree of correspondence with the California results suggests that the problem does not appreciably affect the results. . C Advertisiny MeAe!res The California data also allows an exploration of the influenee of advertising. R. I. Reynolds provided monthly estimates of total advertising expenditures for the leading brands and for the industry as a whole in each of its five media divisions in California." Total industry advertising and '(...continued) equal to zero and a separate dummy, "no income data," was defined to identify them This approach incorporates the information from these observations about the other,coefficients of the model without influencing the estimated income effect. "Adding family education to the model in the national sample produced essentially the same results. Both variables were positive with current smoking or ever smoked as the dependent variable, negative with daily smoking as the dependent variabk, and never significant at 10 percent. i''Phe data mchde ccpaditiues for magazines, suppleoeats, newspapers, out of home, and special (continued...)
Page 33: ozh90d00
Teen Smoking Page 42 identification of a particular brand as most advertised really measures an advertising effect, one would expect more change in the advertising expenditure coefficients. Moreover, teens are not particulady accurate in reporting which brand is in fact most advertised.r' Indeed, there appears to be no relationship between answers to the advertising questions and advertising expenditures. In chi-square tests for independence, we cannot reject the hypothesis that the•btiud named mosr advertised is independent of which brand is more advertised, whether three month, six month, or twelve month advertising is used. Even if we look separately at smokers and nonsmokers, there are no significant differences in responses within each group depending on which brand is most advertised. There are, however, significant differences between smokers and nonsmokers. Teens who have smoked in the past 30 days are more likely to answer the question than those who have not (94.7% response vs. 86.6%), and they are more likely to name Maciboro as the most advertised brand (52.6% vs. 41.6%). Separate logistic models were also estimated, with naming Marlboro as the dependent variable, and Marlboro advertising relative to Camel advertising in the previous six months as the independent variable. That model indicates there is no relationship between relative advertising and the likelihood of naming Marlboro; the chi-square statistic for the model as a whole was only .279. "The distribution of answers for smokers, nonsmokers, and all teens combined is given below. Correct Answers Incorrect Answers Don't Camel Marlboro Camel Marlboro Know Other Current smokers 7.3% 35.7% 19.5% 17.0°/s 5.3% 15.2% Nonsmokers 9.5% 29.0'/0 19.1% 12.6% 13.4% 16.4°/u Combined 9.3% 29.7% 19.1% 13.0% 12.6% 16.3% X== 39.80
Page 34: ozh90d00
Table 6 Dkhotooous Variables u the National Sas Ever Smo1tM C.rre.t Smoker Daily s.oker Variable eeeaii mtTnRim) CodiCimt(sYnFnor) Coekmt(StdFirer) tNTERCPT -5.67 (.25) 6.69 (.40) -11.44 (.79) Risk - Utility Variables Smoking helps whcn bored 0.18 (.11) 0.39 (.13) ••• 0.65 (.17) ••• Smoking helps relax 0.38 (.07) ••• 0.56 (.10) •» 0.44 (.15) ••• Smoking helps with sttess 0.12 (.08) 0.64 (.10) ••• 0.79 (.15) ••• Smoking helps, social situations 0.15 (.06) •• 0.10 (.09) -0.12 (.13) Enjoy risky activities 0.32 (.03) 0.14 (.04) ••• -0.03 (.07) Occational cigarette harmfal -0.52 (.06) ••• -0.52 (.08) ••• -0.01 (.13) Safe to smoke (a year or two) 0.80 (.11) •~ 1.07 (.13) so• 0.30 (.17) • Dangerous, but i mn quit 0.76 (.07) 1.37 (.09) -0.98 (.18) ••• Peer Smoking Variables Same sex best friends (boys and girls) 1.56 (.14) •6• 2.16 (.15) -•• 2.42 (.22) ••• Added effect for boys 0.19 (.16) 0.63 (.16) •~ 0.46 (.18) ••• O, posqe sex best frnends 0.25 (.12) •• 0.72 (.14) •-• 1.18 (.20) ••• Pcople you know 0.21 (.04) ••• 0.22 (.05) ••• 0.27 (.09) ••• Steady who smokes 0.91 (.09) ••• 0.90 (.10) •° 1.03 (.13) ••• No best friends (Boys anly) 1.17 (1.06) _. 1.92 (1.10) • 3.05 (1.16) ••• Family Variables Adult smokers in home (number) 0.18 (.13) 0.19 (.16) ,_ . ;. 0.47 (21) •• Mom Smokes -0.14 (.14) -0.14 (.18) 0.11 (.25) Dad smokes - -0.07 , (.14) ' -0.23 (.18) -0.24 .(.25) No adult in household 1.17 (.54) •• 0.57 (.53) 0.33 (.67) No dad in household 0.11 (.08) 0.24 (.11) •• 0.33 (.17) •• No mom in household 0.18 (.13) 0.30 (.18) • 0.55 (.26) •• Opp. sex older sibs smoke 0.30 (.14) •• 0.50 (.16) 0.42 (.23) • Same scr older sibs smoke 0.71 (.13) ••• 0.91 (.16) ••• 1.04 (.21) ••• Demographic Variables Age 0.21 (.02) ••• 0.13 (.02) ••• 0.29 (.04) ••• Nights out (per week) 0.09 (.02) ••• 0.09 (.02) ••• 0.11 (.03) Family income 0.002 (.001) 0.003 (.002) 0.003 (.003) No income data 0.09 (.10) 0.04 (.15) 0.07 (.22) Nonwhite -0.85 (.08) -0.80 (.13) ••• -0.77 (.25) ••• Hispanic boys -0.30 (.13) •• -0.55 (.21) ••• -1.36 (.44) ••• Not in school . 0.15 (.12) 0.41 (.14 ••• 1.08 (.16) Meaora d Overall Fit Chi Square Statistic 3454.889 3 525.458 1 876.97 (p value) (.00(n) (.0001) (.0001) McFadden's R2 0.2976 0.4549 0.5323 Concordant prediction 94.4% 92.3% 96.1% •" Si ' uant at O.01. *• Si 'ficant at 0.05. * Si ' at O.I0.
Page 35: ozh90d00
Teen Smoking Page 30 than for the extent to which it is dose related. Given willingness to incur the risk initially, decisions about intensity depend more on perceived benefits. Similarly, attitudes toward risk are rnos< important in the decision to smoke the first cigarette. The coeffitaent when ever smoked is the dependent variable is considerably larger than when current smokes are the dependent variable. Moreover, attitudes toward risk are insignificant in explaining daily smoking.•Z The effect of ineasures of the perceived benefits of smoking also diffas across the different levels of smoking intensity. i Beliefs that smoking hdps when bored and helps with stress are not significant in explaining ever smokers, but are significant in the other equations." Moreover, the coefficients increase at higher levels of smoking intensity. Belief that smoking helps relax is significant in all three equations, but the coefficient is largest in explsining current, srtmking.;. The . variable for teens who think smoking helps in social situations is positive and significant in the ever smoked equation, but insignificant with the other dependent variables." The results suggest that boredom and relaxation are more important factors in explaining continued smoking or higher levels of smoking than in the decision to try smoking initially. The peer pressure variables are significant in all three equations. The pattern of sharply increasing coefficients on the variables reflecting the behavior of close friends at higher levels of "In the California sample, attitudes toward risk are also insignificant in the current smoker equation. 'In the weighted estimates with ever smoked dependent, smoking helps whet bored is just short of significant at 5 percent (p =.0529), and smoking helps with stress is significant at .09. .; "In California, smoking helps relax is not significant in the daily smoker equation. Smoking helps in social situations is negative in all three equations, and significant except in the current smoker equations. Smoking helps with stress remains significant in explaining ever smoking.
Page 36: ozh90d00
' r 85/19/1997 11:34 2126272034 AUDITS & SIRVEYS INC PAGE 02 AUDITS & SURYEYS WORLDWIDE The Audits & Surveys Buildinq • 650 Avenue of the Ameriws New York New York 10011 • 212-627-9700 • Fax 212-627-2034 TEEN SMOKING STUDY DESCRIPITON OF 11ETHOD M 77771177 1 • The universe for this study was defined as U.S. residents 12 to 20 years of age inclusive, who (a) have smoked a cigarette in the past 30 days, aad (b) usually buy their own cigarettcs. • Interviewing for the study was conducted from a ceatcal telephone facility, with the qu,eationasiie programmed for computer assisted telephone intaviewing. The interviewing was conducted in two stages. Ietervin.iny Sta•a•„ i(April 2- 2 19971 • A random digit dial national sample of U.S. tekphone households was used for the entire study. • Household sc~reening to determine presence of any persons 12-20 years of age in household: • If one person, interview requested with that person. • If more than one, random selection of one 12-20 year old member to be interviewed. • Further screening of "age-eligfble" (12-20 year old) respondents to determiae ... • If smoked a cigarette in the past 30 days; • If so, whether that person usually buys his/her own cigarettes. • Respondents qualifying on these criteria contimad into the remainder of the interview; all other non-qualifying interviews were terminated at this point. • Completed interviews were thus obteined with a total of 226 qualified respandoats 12-20 years old: 82 in the 12-17 aga grotA md 138 in tb:e 18-20 group (6 respondents did not report their age). • At this time, it was determined that it was necessary to increase the sample to yield 200 respondents in each age group (12-17 nod 18-20); it was aLso determined necessary that measures of inadeae be obtained separately for the 12-17 and 18-20 groups. Spence forthe Mol Marlaednp
Page 37: ozh90d00
Teen Smoking Page 33 household is primarily significant in the current smoker and daily smoker equations." This is the reverse of the California pattern, where these variables are only significant in the ever smoked equation. Differences in the factors influencing decisions about different levels of smoking are also apparent in Table 7, which examines the calculated probability of smoking at different levels for combinations of variables, based on the models in Table 6. The risk utility variables, measuring perceived risks and benefits, are alone enough to lead a'teeti to try cigatettes if we adopt the : convention that trial results when the calculated probability exceeds 50 percent" They are not enough to induce current use, however, and they make a relatively small contribution to the likelihood of daily smoking. The peer variables tell a similar story. Associating with smokas is enough to induce trial, but at least for the average age teenage girl, is not enough to lead to aurent smoking." Particularly for daily smoking, something else is needed to raise the likelihood over half. The experience gained from trial would appear to reduce the likelihood of more intensive smoking rather than increase it, given the perceived risks and benefits of smoking.SO "In the weighted results, the no dad variable is insignificant in the current and daily smoker equations. No mom, however, is significant in all three equations. "The California results, consistent with the lesser importance of the risk iutihty variables, differ somewhat. The calculated probabilities from the risk utility variables alone are 28.51% (Ever smoked), 8.98% (current smoker), and 0. l 1"/0 (daily smoker). The peer and family variables are very similar to the national sample. "For boys, the greater sensitivity to peers of the same sex leads to a calculated probability of stnoking of 6 1 % for the set of peer variables. 50This is also consistent with evidence from the Califonnia adult survey. Among those 18 and over who had ever smoked a cigarette, a majority at every age except 19 were not current smokers. Among 25 to 29 year old adults who have ever smoked a cigarette, 47% had never been regular (continued ... )
Page 38: ozh90d00
P Teen Smoking Page 37 smoking decisions. Whether expenditures are measured over the previous six months, the previous year, or the previous three years, none of the brand advertising coefficients are significarit, and thegoodness of fit measures consistently indicate that the variables do not belong in the modd." Moreover, the brand advertising variables consistently have opposite signs. Although the two variables are correlated (r =.40 for six month expenditures), the estimated sum of the coefficients is no more precise. The chi-square test statistic for the sum of the coefficients never reaches 1.00. It might be objected that even three years of advertising expenditures is not sufficient to capture the influence of a lifetime of advertising exposure on teenage smoking decisions. If cumulative lifetime exposure does matter, however, we should expect to see stronger advertising results among younger teens, because the three year advertising measure captures i larger fraction of their total lifetime advertising exposure. . When the model is estimated separately for 12-14 year old teens and 15-18 year old teens, however, the results are no more supportive of an effect of advertising. Compared to the full sample (equation 3 in Table 9), both brand advertising coefficients switch signs, and their standard errors increase considerably. Although the Marlboro advertising coefficient increases considerably, to .168, it is barely larger than its standard error (.166). There are differences between younger and older teens (in particular, best 5iends of the opposite sex is not significant for the younger teens), but they do not appear to involve advertising. Before advertising expenditures can influence decisions, teens must see and recognize the ttrcasages. The inevitable errors in assigning individual teens to advertising expenditures (and, indeed, "Models were also estimated with each brand's advertising entered in six month increments over u~ the previous three years. These models also included the advertisiog recognition variables, discussed o below. In each case, the likelihood ratio test could not rejxt the hypothesis that all of the advertising m expenditure coefficients were zero. A A A W
Page 39: ozh90d00
Table 7 EFFECTS OF VARIABLE COMBINATIONS ON PROBABILITY OF DIFFERENT LEVELS OF SMOICIIVG Dichotomous and Ordered Logistic Models National Sam le Smo ' Level Ever Current Daily Smoked Smoker Smoker Sam le Pro rtion 34.13% 15.34%. 5.82•/. Dichotomous Model Predictions At means of all variables 29.84'/. 5.74% 0.63% BASE: All antismoking answers' 10.65% 0.99% 0.12°/. Combinations of Variables Risk-Utility Variables 58.48'Yo 23.03% ' 0.91% Peer Smoking Variables 76.92°/. 45.72% 21.94°/. Family Variables 27.37% 3.90'/0 ' 1.14% Ordered Logistic Model Predictions At means of all variables 31.25% 7.40% 1.34°/. BASE: All antismoking answers' 9.71% 1.85% 0.32% Combinations of Variables Risk-Utility Variables .60.09'/0 20.92%0 4.29% Peer Smoking Variables 84.23°/. 48.41'/0 ,_ 13.72% Famil Variables 30.90'/. 7.28'/0 1.31% NOTE: Table entries reflect the probability of smoking at the indicated level when a particular combination of variables is addad to the base case. Dicbotomous model predictions are calculated from the acetLcinrts in Table 6; ordered logistic model prediicrions are calculated from the coefficients in Table 4. Variables included in combinations are ideotified in Table 5. • Base probability calculated at sample nran income, age, nights out (and household size in California sample); Enjoy risky activities = 2 (risk oeoUal) and Occasional cigarette dangerous = 1. All other variables = 0.
Page 40: ozh90d00
0 52086 4464
Page 41: ozh90d00
Teen Smoking Page 22 who smoke may be a better measure of the social benefits of smoking than answers to the survey question. Similarly, the effects of the risk measures are consistent with rational choice. Teens who enjoy risky activities are more likely to smoke in both samples. Given their attitudes toward risk, teens who think that an occasional cigarette is harmful are significantly less likely to smoke in both samples. In the national sample, teens who think that it is safe to smoke a year or two are signi5cantly more likely to smoke.10 In the California sample, however, the variable identifyiog teens who think it is safe to smoke a year or two is not significant. In both samples, teens who think a year or two of smoking is dangerous but quitting is easy are significantly more likely to smoke. To clarify the relative importance of the different influences on the decision to smoke, Table 5 presents the calculated probability of current smoking for various combinations of the independent variables." As a base case, the table takes a risk-neutral teenager who has no smoking_8iends or family members, sees no utility in smoking, and thinks that smoking is dangerous. The calculated 30Even among daily smokers, 75% nationally and 86% in California think smoking a year or two is = safe, and a higher fraction of less intensive smokers think so as well. In contrast, a majority of daily and social smokers do not believe there is harm in an occasional cigarette (56% of daily smokers, 58°/* of social smokers nationally, 52% and 53%, respectively, in California). "Because the logistic model is nonlinear, there is no simple relationship between the magnitude of the effects of an independent variabb in isolation and in combination with other variables. Instead, the vector of values for the independent variables must be specified, and the resulting probability calculated from the coefficients of the models in Table 4.
Page 42: ozh90d00
SCHOOL OF BUSINESS AND PUBLIC MANAGEMENT WORKING PAPERS SERIES For more information on the School of Business and Public Management Working Paper Series, please contact Professor Robert Weiner, Phone: (202)994-5981 Fax: (202)994-7422 E-Mail: rweiner@gwis2.circ.gwu.edu. The George Washington University, Department of International Business, Washington, DC 20052. 95-01. "Informality of Interorganizational Relations and Domestic Strategies. Theory and Empirical Analysis of the Advertising Industry" by Douglas M. Sanford, Jr. 95-02. "Exchange Rate Changes and Pricing Behavior of U.S. Manufactured Exports" by Jiawen Yang. 95-03. "Exchange Rate Pass-Through in the U.S. Market; A Cross-Country and Cross- Product Investigation" by Jiawen Yang. 95-04. "Exchange Rate Pass-Through in U.S. Manufacturing Industries" by Jiawen Yang. 95-05. "Oil Futures Trading in the Gulf Crisis. Report from the Front" by Robert Weiner. 95-06. "Competition between State-Owned and Private Multinationals- Evidence from the International Petroleum Industry" by Jean-Thomas Bernard and Robert J. Weiner. 95-07. "United States Post-Cold War Economic and Trade Policies Toward the Russian Federation, Central and Eastern Europe, The People's Republic of China and The People's Republic of Vietnam" by G. Peter Lauter. 95-08. "European Logistics in Transition: Some Insights" by Prabir K. Bagchi and Tage Skjott-Larsen. 95-09. "Stability of Usage Segments, Membership Shifts Across Segments and Implications for Marketing Strategy - An Empirical Examination" by Michael Y. Hu and Pradeep A. Rau. 95-10. "Opportunities in Emerging Islamic Financial Markets" by Hossein Askari and Zamir lqbal. 95-11. "Some International Evidence on Stock Prices as Leading Indicators of Economic Activity" by Anthony Aylward and Jack Glen. 95-12. "Meeting the Transition Head on - New Ways of Thinking" by Charles N. Toftoy and Marina V. Shakalova.
Page 43: ozh90d00
Teen Smoking Page 46 dependent variable suggests that the interdependent utility explanation for the effect of peers is important. The fact that peer variables alone are sufficient to induce trial, but not regular smoking, indicates that the availability effect is present as well. Family influences on teenage decisions about smoking are considerably weaker than beGefs about risk and utility or the behavior of non-family peers. Within the faWy, parents may have less effeu; the behavior of other adults and of older siblings are more influential. Family may, however, be more important in determining daily smoking. The data are consistent with the hypothesis that advertising expenditures have no influence on teenage decisions about smoking. They cannot, of course, prove that hypothesis. Certainly, however, the data provide no evidence to support the notion that advertising has -an important or powerful effect on teenagers' decisions. The effects of total industry advertising are always insigniSrant and often negative, as are the combined effects of advertising expenditures for the two brands that account for the bulk of cigarette sales to California teais. Moreover, brand advertising variables cottsistently have opposite signs. Indeed, although Camel's advertising campaign has been heavily criticized as uniquely appealing to teens, the coefficients for its advertising are negative in models explaining current or daily smoking. The hypothesis that advertising in fact has no significant influence on teenage smoking is far more compelling.
Page 44: ozh90d00
E
Page 45: ozh90d00
. Teen Smoking Page 19 Table 3 presents summary statistics for the key advertising variables included in the analysis. Differences in advertising patterns over time and in different geographic areas cause considerable variation in estimated exposure to advertising. Total industry advertising for the prior six months, for example, ranges from $1.79 per household to 54.45, with an average of 53.15 and a standard deviation of .63. Total brand advertising over the pnevious three years also exhibits significant variability within the sample. Marlboro advertising over the prior three years, for example, has a mean of $4.02 (standard deviation 0.94), and ranges from $1.94 to $5.02. In addition, the California survey included a question about what brand teens think is most advertised. Teens were asked to "think back to the cigarette advertisements you have recently seen on billboards or in magazines. What brand of cigarettes was advertised the most?" Dummy variables were defi'ried to identify teens who named a particular brand as most advertised (Identify Camel, Identify Marlboro). Expenditures over the previous six motnhs were also used to decompose these measures based on the brand that was in fact more heavily advertised for that particular teen. Thus, "Camel Right" is one if the teen named Camel as the most advertised brand when in fact it was and. is zero otherwise; "Camel wrong" is one if the teen named Camel but Marlboro is in fact more heavily advertised.2" SimJarly, "Marlboro right" and "Marlboro wrong" are one for teens naming Marlboro most advertised when it was in fact more (less) advertised, and zero otherwise. Omitted categories are teens who don't know which brand is most advertised (about 13% of the sample) and those who name other brands (about 17%). '""Right" and 'wrong" answers were defined based on comparisons of Marlboro and Camel advertisiug only, smce data were not available at the brand level for all brands. In the ssmpk, mean ecpadituues per household for these two brands in the previous six months were approximately 32 percent of the itdustry total. It is unlikely, but not impossible, that any other brand was in fact more advertised for particular teens.
Page 46: ozh90d00
95-13. "Natural Resource Depletion and National Income Accounting: Is GNP in Kuwait and Norway Really so High?" by Shantayanan Devarajan and Robert J. Weiner. 95-14. "Are International Agreements to Regulate Global Warming Necessary?" by Shantayanan Devarajan and Robert J. Weiner. 95-15. "Japanese Governance Systems: A Critical Review" by Hicheon Kim and Robert E. Hoskisson. 95-16. "Investments Under Uncertainty and Irreversibility: Rational Indecision and a Numerical Investigation of the 2-D Free Boundary Problem" by Spiros H. Martzoukos. 95-17. "Representation and Solution of Decision Problems Using Sequential Decision Diagrams" by Zvi Covaliu and Robert M. Oliver. 95-18. "Sequential Diagrams and Influence Diagrams: A Complementary Relationship for Modeling and Solving Decision Problems" by Zvi Covaliu. 95-19. "Future Trends In Managerial Beliefs: A Comparative Study" by Salah S. Hassan and Tarek A. Hatem. 95-20. "Home Sweet Home: Cooperative Interorganizational Relations'and Domestic Strategies in the Advertising Industry" by Douglas M. Sanford, Jr. 95-21. "European Logistics Alliances: A Management Model" by Prabir K. Bagchi and Helge Virum. 95-22. "Catalog Mix Adaptation to International Markets: An Empirical Study" by Fernando Robles and Syed Akhter. 95-23. "Role of Benchmarking As A Competitive Strategy" by Prabir K. Bagchi. 95-24. "Middle East Crude Oil Pricing and Risk Management in the 1990s" by Robert J. Weiner. 95-25. "Surveys of Marketing Research Directors in Fortune 500 Firms" by James R. Krum and Pradeep A. Rau. 95-26. "The Importance of Convexity vs. Bond Theta: You May Be Surprised" by Don M. Chance and James V. Jordan. 96-27. "Management of Newly Privatized Industries: A Background Paper" by Mehmet Ozkaya and Hossein Askari. 96-28. "Firm-Theoretic Limitations on Proposition III" by Paul S. Peyser. 96-29. "The Informational Contents of Currency Options and Futures" by George M. Jabbour. 2
Page 47: ozh90d00
Teen Smoking Page 15 and zero otherwise. Its coefficient should be positive if a steady boyfriend or girlfriend is more influential than other friends of the opposite sex. t The primary variable capturing family influences on teenage smoking decisions is "adult smokers," the number of adults in the household who smoke.70 Although the variable includes parents if they are smokers, dummy variables for a smoking parent (Mom smokes, Dad smokes) allow their influence to differ from other adults. The sign on these variables depends on whether parents are more or less influential than other adults living in the household. Other dummy variables allow for different effects in households where the mother or father is not present or where there are no adults (no mom, no dad, no aduh).Z' "Same sex" and "opposite sex older sibs smoke" allow for the influence of older siblings, and allow that influence to depend on sex.'. The basic model includes several other variables. Family income measures the household's income." Dummy variables allow different effects for teens who are not in school, for Hispanic boys . 41 2OUnlike peers, the numerical formulation fit the California data better than a percentage fomnilation. The California model also includes "household size," which measures the total number of people in the household. This information was not available in TAPS. j'Largely as a result of definitional differences, there are far more teens in "no adult" households in the California sample, 3.2%, compared to only 0.3% of teens in the TAPS sample. In California, "no adult" is one if the teen Gves with another adult relative. In the TAPS sample, no adult is one only for teens living with no parent and no adult relative. pIn California, these variables are defined as the percentage of older siblings who smoke (zero if no older siblings). In the national data, which did not mclude data on the number of siblings, they are dummy variables, equal to one if the teen has an older sibling who smokes. In model development in the California data, the influence of younger siblings was consistently insignificant, a finding that was confirmed in the national sample. "Incoroe was coded to the midpoitn of the reported income ranges in the sumy instrument, using the actual Census value for the open ended top income category. In each sample, approximately 12.5% of the observations have no income information For these observations, income was set (continued...)
Page 48: ozh90d00
APPENDIX D FDA's Criticisms of the Beales Study Are Without Merit. Because Professor Beales' work directly contradicted the theoretical underpinnings of FDA's recently rejected efforts to regulate cigarette advertising, the FDA felt compelled to respond in the justification for its rule.' Rather than confront the problem Beales caused for them, the FDA raised several specious arguments, amounting to little more than hand-waving at an analysis that strikes at the heart of the war on cigarette advertising. For example, FDA contends that "there is no reason to expect to find significant changes in smoking behavior based on small regional variations ... in advertising expenditures." In fact, however, the variations are substantial. The teens exposed to the largest advertising expenditures per household over the previous six months saw three times more advertising than those who were exposed to the smallest amount. Even when advertising was examined over the previous three years, the most exposed teens saw three times more Camel advertising, and 2.6 times more Marlboro advertising, than those who were least exposed. Such variation in exposure is substantial, and would be expected to show an impact if there were one. FDA's contention that Beales' analysis did not take into account the way that cigarette manufacturers would allocate their advertising expenditures is similarly without merit. Teenage smokers are a tiny fraction of the total market, and not the target of cigarette advertising strategies. Even if advertising did attempt to reach teenage smokers, the effect of that targeting would be to induce a finding of a relationship between advertising exposure and smoking, when in fact no such relationship exists. If there were targeting, teenage smokers '61 Fed. Reg. 44486 (1996). ~ N a a m i o. A N
Page 49: ozh90d00
Teen Smoking Page 36 (I989) finds advertising depreciation rates for cigarettes in excess of eighty percent, which argues for using relatively short term advertising measiues. Given this uncertainty, expenditures were measured over a number of different time periods, ranging from three months to three years. Models were also estimated including advertising expenditures in six month increments for the previous three years. The results of adding measures of industry advertising expenditure per household to the California models based on different dependent variables are presented in Table 8. Coefficients of the other variables in the model are essentially unchanged, and are not reported. Regardless of the dependent variable, the advertising coefficients never approach statistical significance, and rarely exceed their standard error. Indeed, of the sixteen advertising coefficients, seven are negative. The Akaike Information Criterion indicates that the advertising measures do not belong in the model, although the differences in fit are slight. When advertising expenditures over the previous ttuee years were included in six month increments, none of the individual coefficients were ever significant and the sum of the coefficients over any given time period (e.g., over I year, 2 years, etc.) was never significant. The likelihood ratio test could not reject the hypothesis that all of the coefficients were zero. Of course, not all cigarette advertising is likely to be equally attractive to teenagers. Advertising for brands that teens buy may be more closely related to teenage smoking decisions than is advertising in general. This possibility is examined in Table 9, which reports the results of including in the ordered logistic model separate measures of advertising per household for Camel and Marlboro, the leading brands among California teens. The first three columns of the table examine brand advertising over different periods. Like 'udustry advertising, the brand variables provide no evidence that advertising has an effect on teenage
Page 50: ozh90d00
Teen Smoking Page 34 To facilitate comparisons betwcen the dichotomous and ordered logistic models, Table 7 also includes the calculated probabilities of different levels of smoking intensity based on the ordered logistic model. Whether evaluated at the sample means or in the base antismoking case, the ordered logistic model predicts somewhat higher levels of wrrent and daily smoking than the dichotomous models. Except for daily smoking, the same picture of the relative importance of the groups of variables emerges from the ordered logistic model. In predicting daily smoking, however, the ordered model assigns considerably greater significance to the risk and utility variables, and somewhat less significance to the peer variables. Although the differences may result from the ordered model's requirement that the effects of the independent variables are the same regardless of the level of smoking intensity, they may also stem from the fact that the dichotomous model ignores the differences between social smokers and nonsmokers. B. The Influence of Advertisin¢ Much controversy about teenage smoking, and much of the policy dispute, has revolved around the influence of advertising. Fxamining the effect of adding advertising measures to the model is therefore of considerable interest. The duration over which advertising expenditures should be measured is not clearcut, however. Although many authors have argued that advertising has persistent effects over time (e.g., Demsetz, 1979; Bloch, 1974), more recent studies have indicated that these effects may reflect retums to brand specific quality characteristics, and that the etFects of advertising are in fact short lived ('I]tomas,1989; Landes and Roserdield, 1994). Moreover, Thomas 50(...continued) smokers. 61 % were not smokers at the time of the survey.
Page 51: ozh90d00
Teen Smoking Page 27 is less likely for nonwhites than for whites, more iikely for teens who are not in school. and more likely the more often teens go out for fun and recreation (nights out). Other things equal, Hispanic boys are more likely to smoke in the California sample, but less likely to smoke in the national sample." Not auprisingly, Hispanic boys account for a substantially larger fraction of the California sample (13.2%) than the national sample (4.7%). Although the ordered logistic model is a compact way to describe changes in smoking behavior, it constrains the effects of the independent variables to be the ssme regardless of the level of smoking involved. For example, each additional year of age is constrained to have the same effect in moving a teen from nonsmoking to having tried a cigarette, and in moving a teen from smoking in the past 30 days to daily smoking. The assumption that all variables have the same effect, however, is not consistent with the data. The model was therefore reestimated using three different dichotomous independent variables based on different definitions of smoking. Although this formulation allows different effects of independent variables in crossing different thresholds, it ignores - differences in smoking intensity among those above or below the particular threshold. In the current smoker equation, for example, the model ignores differences between teens who have never smoked and experimenters, since both groups are not current smokers. Similarly, it ignores differences "( ... continued) "Teens were also asked how much money they had to spend on themselves each week. This income variable was never significant in the Califotnia sample. In the national sample, it was positive and significant at .05 with ever smoked or daily smoking as the dependent variable, but not with current smoking as the dependent variable. 'sPaR of the difference in the estimates for Hispanic boys is due to a difference in the way the two surveys treated race. The California survey allowed the choice of "IGspamc" as a race. The net effect on the likelihood of smoking for Hispanic boys who listed their race as Hispanic is thus the sum of the nonwhite and the Hispanic boys coefficients, which is -.0050.
Page 52: ozh90d00
that perhaps Beales included "irrelevant variables," in fact virtually all of the variables he employed were statistically significant, a finding that was confirmed in the independent TAPS sample. Such variables are hardly "irrelevant." Moreover, FDA suggests that the correct model of the effect of advertising might be a double hurdle model, in which teens first decide to smoke and then decide what brand to smoke. Beales' analysis, however, does not explore brand choice at all. Its focus is instead on the first decision - - whether or not to smoke. If advertising has no influence on that decision, any influence on the subsequent decision about which brand to smoke is irrelevant, to both FTC and FDA. Moreover, FDA suggests no reason to think that neglecting the influence of advertising on market share might distort estimates of the effect of advertising on the decision to smoke. Finally, FDA suggests that what matters is the number of advertising messages, not the amount spent to purchase those messages. Advertising expenditures, however, are generated by the process of purchasing advertising messages. The larger the likely audience of a particular message, the higher its price, and the greater total advertising expenditures. It was precisely to measure the number of messages available in the market that Beales deflated advertising expenditures by the number of households, thereby removing the component of expenditures that stems from different audience sizes, rather than different numbers of messages. 4
Page 53: ozh90d00
Teen Smoking Page 43 With naming Camel as the dependent variable, the modd still insignificant (Chi-square = 2.3, p=.13), and the estimated coefficient on relative advertising is positive.' These results are consistent with the notion that teens who smoke are simply reporting the most popular brand as the most advertised. The primary differenoe between smokers and nonsmokers is that smokers are more willing to answer the question, and, when they do, they name Marlboro. If the coefficients for teens who name Marlboro as the most advertised brand really reflect an effect of advertising, it is difficult to explain why the answers are unrelated to advertising expenditures. Moreover, there is no apparent reason why only Marlboro advertising would produce an effect. Reverse causality seems more consistent with the results. Reverse causality is also consistent with the results from the dichotomous models, where advertising identification is insignificant in explaining the differences between teens who are either current or daily smokers and those who are not. The model includes possible channels for an indirect effect of advertising,• if advertising • ecpendiuues influence teens' perceptions of the risks and benefits of smoking , or perceptions of how many of their acquaintances smoke, or their attitudes toward risk. The simple correlations between six month industry advertising expenditures and the risk and utility measures are consistent with this possibility. The magnitude of the correlation, however, is not particularly impressive; the largest correlation in absolute value is -.043 (with Occasional cigarette harm6tl). The correlations with brand advertising expenditures are considerably smaller, and, for any given question, of opposite sign." s'Other models also included age, nonwhite, family income, people you know who smoke, and not in school as independent variables. Although the models were significant, relative advertising over the previous 12 months was insigni5cant for Marlboro, but significant and of the expected sign for Camel. 'Irhe exception is "People you know," which nxasures perceived incidence of smoking among (continued...)
Page 54: ozh90d00
Teen Smoking Page 9 The dependent variable, smoking, included four levels of smoking intensity: never smoked, smoked at least one cigarette but not in the last 30 days (experimenters)," smoked at least one cigarette in the lau 30 days but not daily (social smokers)'z, and daily smokers. Smoking at least one cigarette in the last 30 days is the widely used definition of a current teenage smoker." Because the ordered model constrains the independent variables to have the same effect regardless of the level of smoking intensity involved, models were also estimated using dichotomous dependent variables for each level of amoldng. Essentially, the dichotomous models define smokers as those who are above a given level of smoking intensity, and estimate the effect of the independent variables on the " probability of being a smoker. Ever smoked is one for teens who have ever smoked a cigarette, and zero for those who have not. Current smoker is one for teens who have smoked either daily or less "The experimenters include some former smokers (who haye, smoked_at least 100 cigarettes in their lifetime) who have quit. This category is a relatively small fraction among teens: Forvrz snukav (weighted data) as a pacent ot Population Ever Smoked Smoked nM in past 30 days NaUonal To= 1.5 4.1 . 7.4 Calif. Teens 1.4 5.4 ' 8.4 Calif. Adults 25.7 35.5 53.5 "Some "social" smokers may be recent experimenters, who happen to have tried their first cigarette in the past 30 days. Even among younger teens, however, the fraction of teens who smoked their first cigarette at their current age is relatively small. For example, of 113 twelve year olds who have ever smoked a cigarette in the TAPS sample, 24 smoked their first cigarette at age 12. " The &action smoking their first cigarette at their current age is considerably srnaller among older teens. °Although it has the virwes of objectivity and widespread use, defining current smokers as those who have smoked in the past 30 days does not necessarily correspond to an individual's view of his own status, particularly among younger respondents. In the California adult survey, respondents were asked, "Do you smoke cigarettes now?" Among 18-24 year olds who had smoked in the previaas 30 days but less than daily, 50.6°/9 answered no (n=592). This is 22.3 percent of all those who smoked in the previous 30 days (n=1092). The percentages are based on weighted data. The differences between the objective and subjective definitions narrow considerably among older adults, but persist even among those in their 60s.
Page 55: ozh90d00
Teen Smoking Page 39 in assigning expenditures to a particular geographic region) may mean that expenditures are a relatively poor measure of the advertising teens actually see. The California survey question asking teens to identify the most advertised brand provides a way to examine this possibility. If advertising influences consumption decisions, teens who believe that one of the primary teenage brands is most heavily advertised should be more likely to be smokers. . Unfortunately, however, causality may also run the other direction: teens who smoke may be more likely to name one of the most popular teenage brands as most advertisad. The coefficients of the advertising identification variables, alone and in combination with six month expenditure measures, are also reported in Table 9. As with advertising expenditures, other coefficients do not change in meaningful ways. . Naming Marlboro as the most advertised brand is positively associated with smoking. Identifying Camel as most advertised, however, is not. The coefficient is negative, but not statistically significant. Including the identification variables does nothing to improve the significance of the ~ advertising expenditure results. That result also holds when advertising is measured over the previous three years (not reported in the table). The identification variables are the same whether expenditures are included or excluded from the model (column 4 vs. column 6). The identification variables were also decomposed, based on whether the identified brand was in fact more heavily advertised over the previous six months. Surprisingly, only for teens who name Marlboro as most advertised when in fad it is pQt is the relationship significant at 5 percent. When Marlboro is in fact most advertised, the coefficient is about half as large, and is only significant at the ten percent level. Either formulation fits the data equally well as judged by the Akaike Information Criterion. Table 10 presents selected advertising measures in models with dichotomous dependent variables. The results are basically similar to those in the ordered logistic model. The six-month
Page 56: ozh90d00
Teen Smoking Appendix B ..Page 50 . declines by about one third, and its significance weakens.'0 There are changes in the non-advertising coefficients in the other weighted models as well, but fewer coefficients change and they tend to change less. .. . It is difficult to place much confidence in the weighted results. They are driven by a single • high weight observation. They give less weight to smokers as a group, particularly heavy smokers. The weights themselves are significantly positively correlated with advertising expenditures, pahaps because both depend on geographic location. Moreover, it is unclear, exactly how the weights were determined. Although the sample design was stratified at the household level, it called for . irrterviewing all teens within each household selected. Weights reflecting only sample design should therefore be the same for teens in the same household. Our high weight smoker, however, has a younger sister in the sample who receives a slightly lower.weight (I 1.4267).61. ., , , r, 66Ihe high weight smoker is a 13 year oM.. There are 41 13 year old smokers in the sample, but on a weighted basis, this observation account for 25% of the 13 year old smokers. s`Aside from age and sex, these two teens differ about their native language.. The higher weight is assigned to the one who reports English as his native language.
Page 57: ozh90d00
Teen Smoking Page 14 for teens who think smoking a year or two is dangerous, but not for teens who think such smoking is safe." The coefficient should be positive. The influence of peers is captured by a series of four variables that measure the teen's perceptions of smoking behavior ranging from their acquaintances in general to peers that may be particularly in9uential. "People you know" measures a teen's perceptions of how many people they know about their age who smoke. It ranges from zero (none) to three (most). "Opposite sex best friends" and "same sex best friends" are continuous variabies tneasuring the fraction of a teen's best friends of the opposite sex and same sex who are smokers." These coefficients should be positive if best fiiends are more influential than other acquaintances. The model also includes an interaction term ("added effect for boys") that allows the influence of friends of the same sex to differ for boys and girls. This variable, which is equal to "same sex best friends" for boys and zero otherwise, allows for the possibility of sex differences in the influence of peers." "Steady who smokes" is a dummy variable equal to one for a teen who has ever had a steady boyfriend or girlfriend who was a smoker "In California models with current smoking as the dependent variable the full range of interaction of attitudes toward ease of quitting with "safe to snoke" was examined with a set of 4 dummy variables. Only the variable for "dangerous but I can quit" was significant. Thus, beliefs about the ability to quit were only significant for teens who think smoking for a year or two is dangerous. "In the California survey, teens were asked how many "best friends" they had of each sex, and asked separateiy how many of those best 6iends smoke. Answers ranged from 0 to 99. In the TAPS survey, teens were asked how many of their four best friends of each sex smoke. The percentage was set equal to zero for teens who reported no friatds of a particular sex. Dummy variables to allow for these teens were insignificant in the model development analysis, except for boys with no fiends of the same sex ("No best &iends"). The percentage formulation fit the California data better than models based on the number of fiiends who smoke. "In the early stages of model developrcent, dummy variables allowed virtually all coefficients to differ by sex. Most of these interactions proved insignificant.
Page 58: ozh90d00
awuN.W.1C/QW.1~,J f0%P. 484b 980Z5 on
Page 59: ozh90d00
Teen Smoking Appendix B Page 49 The substantial changes in the advertising results for current smoking appear to be driven by a small number of high weight observations. Deleting the high weight observations yields results for the advertising variables that are much closer to the unweighted resufts.x The absolute value of the ~ advertising expenditure coefficients falls by about one third, and neither variable is significant at five percent. Only one of the deleted observations is a current smoker (weight = 11.489), but that . observation appears to be the primary reason for the difference between weighted and unweighted nesults." Given the lower weight of smokers generally, this one obsdvation is weighted ahnost.13 times more heavily than the average smoker, and accounts for 2.5% of the weighted current smokers, (0.2 percent of the actual smokers). As with the advertising variables, the results for other variables move substantially toward the . unweighted results when the high weight observations are ddeted: The coefficients of "Stnoking helps with stress" and "Safe to smoke" fall substantially, and are no longer significant, even with the high weight observations deleted. The coefficients of"smoking helps when bored" and "smoking helps relax" increase substantially. The absolute value of "smoking helps, social situations" increases, and the coefficient becomes significant at .01 (.04 deleting the high weight observations). Among the peer variables, the "best friend" variables decrease slightly, while the other peer variables increase. All remain highly significant, except the coefficient allowing greater sensitivity of boys to behavior of friends of the same sex. The dummy variables for parents who smoke decline in absolute value, but remain significant. Income increases slightly, and is significant at .10. Finally, the age coefficient "Weights were not renormalized: The mean weight in this sample falls to .9843. The mean weight for smokers is.8943; for nonsmokers it is .9944. "Deleting an additional 8 observations, all nonsmokers, with weights between 7.5 and 11.0 results in a further, very slight weakening of the advertising coefficients.
Page 60: ozh90d00
Teen Smoking Page 29 between social and daily smokers, since both groups are current smokers. The estimates of the dichotomous models for the national sample are presented in Table 6." All three perceived risk variables (occasional agarette harmful, safe to smoke, and dangerous but I can quit) have the expected signs and are highly significant in the ever smoked and current smoker equations, but not in explaining daily smoking. Indeed, teens who think that smoking is dangerous but they can quit any time they want are less likely to be daily smokers, but more likely to have tried a cigarette and more likely to be current smokers.'D The most likely explanation for this resuk is that in explaining daily smokers, where experimenters and social smokers are classified as nonsmokers, the variable is picking up the difference between daily smokers and other smokers." The results suggest that risk perceptions are more important in decisions to try a cigarette and in decisions to smoke moderately than in decisions about more intense smoking. In part, this may.be ; due to the lirnitations of the variables. The risk perception that should most affect the choice between .: occasional and daily smoking is the increase in risk due to increases in the dose. Smokers who think that risk does not increase with the quantity consumed beyond sonie level would be unaffected by perceived risk once their consumption reached that threshold. The available measures, harm from an occasional cigarette and safe to smoke a year or two, are better proxies for the existence of risk "The same analysis was also conducted in the California sample. Differences between the California and national samples are noted in the following footnotes where relevant. The detailed California results are not reported, but are available on request. 'In the California sample, "occasional cigarette hanmful" remains negative and significant in the daily smoker equation. "Safe to smoke" is also negative in the daily smoker equation, but not significant at 10 percent. `Dargerous but I can quit" is one for 11% of teens who have never smoked and for 1 I% of the daily smokers. It is one for 16% of experimenters, and 41% of social smokers.
Page 61: ozh90d00
APPENDIX B Advertising Results in Weighted Models Weighting changes the estimated influence of the advertising variables substantially, as reflected in the coefficients reported in Table B-1. The effect of weighting is to reduce the influence of the observations of teens who are smokers, and particularly heavy smokers. The mean weight in the sample as a whole is.9994, but the mean weight for daily smokers is.8594, compared tb 1.0242 for nonsmokers. With srnolong or ever smoked as the dependent variable, the changes are relatively sfight, with improved significance for naming Marlboro when it is in fact most advertised. With current smoking as the dependent variable, however, (and to a lesser extent with daily smoking dependent), changes are substantial.s' The best fitting model includes both the brand identification variables and advertising expenditures. Both Camel and Marlboro;advertising expenditures are significant, but the Camel eoefficient is negative. As in the unweighted results, the sum of the advertising coefficients is positive but insignificant. The Camel identification variables are closer to statistical significance, and Marlboro identification when it is not most advertised is highly significant. Other coefficients also change appreciably in the weighted results for current smokers. Table B-2 presents the individual coefficients for this model, which is more sensitive to the weighting scheme than any other. The table also includes the results of deleting the seven observations with weights greater than 11. These observations are all ELspanics from Los Angeles county. 57The mean weight for all current smokers is .9153, compared to 1.0089 for those who are not current smokers.
Page 62: ozh90d00
Teen Smoking Page 45 coefficients. F'mally, the benefit measures were used as dependent variables in a logistic model. The model included all non-benefit measures as independent variables, along with three year and six month bmrd advertising variables. Of the 16 advertising coefticieits, only one was significant: six month Camel advertising expenditures were associated with a significantly lower probability of saying that smoking helps to relax (coefficient =-.54, s.e. = .27). In contrast, peer influence measures were frequently significant predictors of answers on the smoking benefit questions. Thus, there is no persuasive evidence that advertising operates indirectly, bg influencing teens perceptions of the benefits of smoking. V. CONCLUSIONS The data indicate that teenagers make rational decisions about smoking givea their preferences and the information they have: Whether in the California sample or in the independent national ' sample, teens' beliefs about the risks and bene6ts' of smoking are significant predictors of their decisions. Teens who believe smoking is risky are less likely to have tried a cigarette and less likely - to smoke more intensively. Similarly, teens who believe smoking helps with stress, with boredom, or to relax are more likely to smoke. Given their beliefs about smoking, teens who find risky activities fun are more likely smoke, although the effect may be more important for trial than for higher levels of smoking intensity. Younger teens are less likely to smoke at any level. The most important variables determining teenage smoking behavior are those reflecting the behavior of peers. Teens are more likely to smoke if most of their acquaintances smoke, and more likely to smoke the more of their best friends smoke. Although best friends of both sexes matter, friends of the same sex are consistently more influential, an effect that is greater for boys than for girls. The increase in the coefficients of the peer variables when higher levels of smoking are the
Page 63: ozh90d00
Nelson, Phillip (1974), "Advertising as Information," Journal of Political Economv, Vol. 82, pp. 729-54. Pierce, John P., Elizabeth Gilpin, David M. Bums, Elizabeth Whalen, Bradley Rosbrook, Donald Shopland, and Michael Johnson (1991), "Does Tobacco Advertising Target Young People to Start Stnoking? Evidence from California," Joumal of the American Medical Associatioa vol. 266 (Dec. 11), p. 3154. Schmalensee, Richard (1972), The Econbmics of Advertisma. Vol. 80, Contributions to Economic Analysis, Amsterdam: North Holland. ' Schneider, Lynn, Benjamin Klein, and Kevin M. Murphy (1981), "Government Regulation of Cigarette Health Information," Joumal of Law and Economics_ Vol. 24 (December), pp. 575-612. Surgeon General of the U. S. (1994), Preverttina Tobacco Use Amooa_ YQpg Peoole` Report of the Surgeon GeneraL Atlanta, Ga: U.S. Department of Health and Human Services,`' PubGc Health Service, Centers for Disease Control and Prevention, National Center for Chronic Disease Prevention and Health Promotion, Office on Smoking and Health. Thomas, Lacy Glenn (1989), "Advertising in Consumer Goods Industries: Durability, Economies of Scale, and Heterogeneity," Journal of Law and Economics Volume 32 (April), pp. 163-194. van Roosmalen, Erica H. and Susan A. McDaniel (1989), "Peer Group Influence as a Factoi " in Smoking Behavior of Adolescents," Adolescence_ vol. 24, p. 801. van Roosmalen, Erica H. and Susan A. McDaniel (1992), "Adolescent Smoking Intentions: Gender Differences in Peer Context," AdolP_s..^at:ce, vol. 27, p. 87. Viscusi, W. Kip (1992), SmokinQ: Making the Ri Decision (New York: Oxford University Press). Viscusi, W. Kip (1991), "Age Variations in Risk Perceptions and Smoking Decisions," Review of Economics & Statistics, Vol. 73, p. 577. Wang, Min Qi, Eugene C. Fitzhugh, R. Car1 Westerfield, and James M. Eddy, "Family and Peer Influences on Smoking Behavior Among American Adolescents: An Age Trend," Journal of Adolescent Ha& Vol. 16 (March), pp. 200-203.
Page 64: ozh90d00
Table 1 reports the results for the advertising coefficients for models explaining whether teens have ever thought seriously about quitting, and whether they have tried to quit in the past six months. Table 2 provides the advertising results for models explaining the number of quitting attempts, whether teens are former smokers, and the likelihood of having smoked at least 100 cigarettes. Results for the other variables included in the model are not reported. With some of the measures of quitting, the base model did not provide a statistically significant explanation of the measure. The 5ill model of the likelihood that teens have thought seriously about quitting, for example, was not statistically significant. Beales therefore used a pared down model, based on the individual coefficients and basic demographic variables that appeared to be significant from the full model. This "reduced base" model incorporated 10 variables, measuring age, family income (and a dummy variable for teens with missing income data), sex, dummy variables for nonwhites and Hispanics, the measure of attitudes toward risk, whether smoking helps with stress, whether an occasional cigarette is harmful, and the number of nights per week that the teen goes out. Advertising variables were then added to the reduced base model. If the reduced base model was still insignificant (as it was in explaining attempts to quit in the last six months), Beales estimated models with advertising variables only, with no other controls.' On the staffs hypothesis that advertising encourages continuation, the adve isinv 'If other variables are significant predictors of a particular measure of quitting, then models that include only advertising are misspecified, because they omit other relevant variables. The advertising coefficients would therefore be biased. When no significant predictors are identified, however, there is less reason for concern about the potential bias. 4
Page 65: ozh90d00
variable is useless because all teens "had seen and remembered the advertising." Only 29 percent of teens named Camel as the most advertised brand, and only 43 percent named Marlboro. FDA further contents that the Beales analysis does not measure the effect of advertising on teens' perceptions and beliefs. In fact, however, Beales did test for a possible influence of advertising on teens' perceptions of the benefits and risks of smoking. As previously reported to the Commission, Beales found no evidence of any indirect effect of advertising on the perceived benefits of smoking. In addition, FDA suggests that Beales' results might be due to multicollinearity, or the fact that advertising variables are closely associated with some other variables included in the analysis. The claim, however, is not consistent with the Beales results. When two measures are highly correlated, estimates of both variables tend to be quite imprecise. In the Beales analysis, however, only the advertising variables are insignificant. Moreover, Beaks tested for effects of multicollinearity with the variables most likely to be closely related to advertising, the perceived benefits of smoking. Omitting those variables did nothing to strengthen the advertising results. Beales has subsequently estimated the model without peer influence measures, and again found no effect of advertising. He has also explored various models using only Camel advertising as an independent variable, against the possibility that the high correlation between competing brands' advertising expenditures influenced the results. Again, however, he found no influence of advertising. In fact, contrary to the hypothesis that the results are due to high correlation among the variables, the results were actually weaker when only one brand's expenditures were included. Although FDA suggests 3
Page 66: ozh90d00
Table B-2 Effects of Weighting and Dekting Higb Weight Obserrations California Sam e UOweit,hted Data Weigpted Data Omit @igh Weights Coefficient (Sld. Error Coefficient (Std. Ermr) Coefficient (Std. Error) INTERCPT b.57 (.77) -6.15 (.81) ••• -6.58 (.83) ••• Risk - Utility Variables Smoking helps when bored 0.33 (.14) •• 0.40 (.14) ••• 0.40 (.14) ••• Smoking helps relax 0.62 (.14) ••• 0.88 (.15) 0.90 (.15) ••• Smoking helps with stresc 0.40 (.14) 0.05 : (.15) 0.03 (.15) Smoking helpsm social situations -0.24 (.13) • -0.36 (.13) ••• -0.27 (.14) •- Enjoy risky activities 0.07 (.07) 0.03 (.07) 0.05 (.07) Occasional cigarette hannful -0.97 (.12) ••• -0.94 (.13) ••• -1.02 (.13) ••• Safe to smoke (a year or two) 0.39 (.20) •• 0.11 (.20) 0.13 (.21) Dangerau, but I can quit 0.56 (.16) ••• 0.47 (.17) ••• 0.51 (.17) ••• Peer Smoking Variables Same sec best friendc (boys and girls) 1.92 (.26) ••• 1.76 (.26) 1.82 (.26) ••• Added e@'ect for boys 0.64 - (.29) « 0.46 (.30) 0.50 (.30) • Opposite sex best friends 0.82 (.21) 0.64 (.22) 0.72 .(.22) ••• People you know 0.43 (.07) ••• 0.58 (.07) ••• 0.46 (.07) ••• Steady who smokes 0.77 (.16) ••• 0.89 (.16) ••• 0.89 (.17) .« No best friends (Boys only) 1.58 (.53) ••• 1_29 (.58) •• 13d (.58) •• Family Variables Adult smokers in home (number) 0.64 (.16) 0.63 . (.15) ••• 0.65 (.15) iV1om Smokes -0.64 (.23) -1/.50 (.22) •• -0.45 (.22) •• Dad smokes -0.47 , • (.22) ••.. -0.51 (.21) •• -0.49 (.22) •• No adult in household -0.03 (.46) -0.05 (.44) -0.18 (.45) No dad in household 0.10 (.17) 0.00 (.17) 0.07 (.17) No mom in household 0.05 (.29) 0.38 (.29) 0.36 (.29) Opp. sex older sibs smoke 0.28 (.23) 0.26 (.24) 0.36 (.25) Same scx. okfer sibs snwke 1.16 (.20) 1.05 (.21) ••• 1.16 (.21) ••• Household size -0.07 (.05) -0.03 (.05) -0.06 (.05) llemograpbic Variables Age 0.15 (.04) »• 0.10 (.04) •• 015 (.04) ••• Nights out (per week) 0.04 (.04) 0.01 (.04) 0.03 (.04) Family income 0.003 (.002) 0.004 (.002) • 0.003 (.002) No income data 0.12 (.22) 0.00 (.22) 0.10 (.23) Nonwhite -0.36 (.14) •• -0.25 (.14) • -0.36 (.14) •• Hispanic boys 0.42 (.18) •• 0.43 (.17) •• 0.25 (.18) Not in school 0.70 (.28) •• 0.56 (.28) •• 0.53 (.28) • Advertising Variables Camel expendibrres, 6 months -0.30 (.66) -1.48 (.66) •• -1.05 (.67) Marlboro expenditures, 6 months 0.82 (.67) 1.93 (.72) ••• 1.36 (.73) • Identify Camel, right 0.04 (.27) 0.45- (.26) • 0.33 (.26) Identify Camel, wrong 0.14 (.19) 0.32 (.20) 0.33 (.20) Identify Marlboro, righl 0.18 (.17) 0.75 (.22) ••• 0.17 (.18) Iden ' Marlboro, wron 0.40 (.23) • 0.17 (.18) 0.42 (.23) • Sample Sitt 4980 3980 4973 ••' Si at 0.01. 00 Si t at 0.05. '$i ' at 0.10.
Page 67: ozh90d00
Teen Smoking Page 44 More importantly, if there is an indirect influence of advertising on consumption decisions through perceived risk and utility, introducing advertising measures into the equation should result in changes in the coefficients of these nine variables. The coefficients, however, are quite stable. When irdustry advertising measures are introduced, no coefficaent that was significant at 10 percent without the advertising variable changes by more than one percent.' Introducing brand advertising expenditures changes three coefficients (on three differem variabks in three different models) by more than one percent, but the largest change is an increase of less than 3%. In the weighted models, there is somewhat more change, particularly with daily smoking as the dependent variable. Adding 3- month or 6-month industry advertising increases the coefficients of "Occasional cigarette harmful" and "Dangerous but I can quit" by 2 to 3 percent; 12 month industry advertising reduces them 2 percent. Only one other coefficient ("Dangerous but I can quit" with current smoking dependent and 12 month industry advertising) changes as much as 2 percent. With brand advertising added, only three coefficients change more than one percent (smoking helps in social situations and helps when bored decline 2 percent; "People you know" increases 5 percent), all with daily smoking as the dependent variable. Similarly, models including the advertising identification variables and six month brand advertising expenditures were estimated without the four benefit measures. Regardless of the dependent variable, there were no significant changes in the advertising ss(...continued) a teen's acquaintances. Contrary to suggestions that advertising increases the perceived incidence of smoking (Surgeon General 1994), both Camel and Marlboro advertising are associated with lower perceived incidence of smoking. 'Across the four models, there are 36 coefficient estimates, 29 of which are significant at 10% or better when the advertising measures are not included.
Page 68: ozh90d00
i Teens who had never thought about quitting were not asked the questionj and Beales assumed they had never tried to quit. If Joe encourages continuing, then teens exposed to more advertising should have fewer attempts to quit. The responses were anal~zed using an ordered logistic model.' The third quitting measure asked teens whether they had tried to quit in the last six months. Data were available for 412 teens, of whom 231 had tried to qtjit in the last six months. This is the only question that provided any information about W{hen quit attempts had occurred. If Joe encourages continuing, then teens exposed to more vertising should be less likely to have attempted to quit in the previous six months. The ysis used a logistic regression model. The fourth measure of quitting that Beales examined was succe.ss~ul quitting attempts, I using the NHIS definition of a former smoker. According to this definitlon, a former smoker is one who has smoked at least 100 cigarettes in his or her lifetime, and has not smoked in the previous 30 days. The sample was restricted to the 309 teens who had smoked at least 100 cigarettes. Of these, 58 were former smokers. If Joe encourages continuing, then teens exposed to more advertising should be less likely to be former smokers. i The analysis used a logistic regression model. 'In addition, Beales assumed that teens who had tried to quit two or times had an average of 2.5 quit attempts, and that those who had tried to quit four o tnore times had an average of 4.0 times. The data could then be analyzed using ordinary ion methods. These results were quite similar to the ordered logistic model, and are n t reported below. 2
Page 69: ozh90d00
whether, or how, consumption levels influence attempts to quit, but consumption surely affects the likelihood of successful quitting. Regarding reverse causality, former smokers may pay less attention to cigarette advertising than current smokers. If so, they would be less likely to name either Camel or Marlboro as the most advertised brand. The fmal panel of Table 2 examines whether advertising influences the likelihood that teens have smoked at least 100 cigarettes in their lifetime. Here there are three significant coefficients, all inconsistent with the staffs hypothesis. Teens who identify either Camel or Marlboro as the most advertised brand are significantly less likely to have smoked 100 cigarettes in their lifetime. When only the identification variables are included, both coefficients are individually significant at the five percent level. The likelihood ratio test, however, cannot reject the joint hypothesis that both coefficients are zero. Camel advertising expenditures are associated with a lower probability of having smoked 100 cigarettes, but Marlboro expenditures are associated with a higher probability. The expenditure coefficients, however, are all insignificant. In sum, the California data provides no reason to believe the staffs inventive hypothesis that advertising somehow influences smoking "contin»ation." All of the coefficients for Camel advertising expenditures indicate the opposite effect, albeit statistically insignificant: The only results that are both significant and consistent with the staff s hypothesis suggest a different target for the complaint: Marlboro. 7
Page 70: ozh90d00
:. Table 5 Advertising and Smoking Trends in Michigan Subgroups Whites Blacks Advertising Measure Coefficient R F Coefficient R F (std.err.) Squared (std.err.) Squared Current Year -0.02 0.025 0.21 -0.1 0.285 3.2 (0.05) (0.06) One Year Lag 0.00 0.00 0.00 -0.10b 0.503 7.08b -0.05 -0.04 Two year lag 0.03 0.057 0.36 -0.08 0.478 5.49 (0.05) (0.04) Three year lag 0.04 0.079 0.43 -0.08 0.465 4.35 (0.06) (0.04) Two year Total -0.01 0.021 0.15 -0.06b 0.463 6.02b -0.03 -0.02 Three year Total -0.01 0.017 0.11 -0.05b 0.585 8.45b (0.03) (0.02) Four year Total -0.01 0.027 0.14 -0.05b 0.755 15.42b (0.03) (0.01) Males Females Current Year 0.01 0.001 0.01 -0.07 0.278 3.08 (0.07) (0.04) One Year Lag 0.05 0.059 0.44 -0.08 0.35 3.78 (0.07) (0.04) Two year lag 0.08 0.223 1.72 -0.02 0.028 0.17 (0.06) (0.05) Three year lag 0.07 0.175 1.06 -0.02 0.032 0.17 (0.07) (0.06) Two year Total 0.01 0.012 0.09 -0.05 0.408 4.83 (0.04) (0.02) Three year Total 0.01 0.012 0.08 -0.03 0.204 1.53 (0.04) (0.02) Four year Total 0.01 0.005 0.02 -0.04 0.298 2.13 (0.04) (0.03) Notes: a= significant at .01; b= significant at .05. 14
Page 71: ozh90d00
would see more advertising than nonsmokers, not because advertising causes smoking, but because advertisers direct their messages to smokers. The fact that Beales can detect no relationship even given that there is a bias in favor of finding a relationship strengthenss not weakens, his conclusions. Moreover, FDA contends that Beales should have looked for a difference in the effect of Camel advertising before and after the introduction of Joe Camel. In fact, Beales examined advertising expenditures over a three year period, which did include some advertising prior to the Joe Camel campaign. Moreover, he was unable to detect any effect of advertising when expenditures were entered in six month increments, which also would have allowed for the effect that FDA asserts must exist. In addition, Beales separately analyzed teens who first smoked before the Camel campaign and those who first smoked after the campaign, but found no relationship between smoking status and advertising exposure in either group. More fundamentally, however, looking for a differential effect of advertising before and after the campaign is pointless if there is no effect of advertising during the campaign. The only way there could be a differential effect would be if Camel advertising before the Joe Camel campaign somehow discouraged teenage smoking. FDA also criticizes Beales' use of teens who identified either Camel or Marlboro as the most advertised brand as a measure of advertising's effect. Although Beales acknowledged the limitations of this measure in his original paper, it was the only survey- based measure of advertising that was available, and it was devised by Dr. Pierce, who used it to conclude that Camel advertising targeted teens. Moreover, it was the only advertising measure that produced any statistically significant effects, although, as Beales argued, those effects most likely reflect reverse causation. FDA is simply wrong in its claim that the 2
Page 72: ozh90d00
G 1
Page 73: ozh90d00
Although the staff has suggested that Camel advertising might have been particularly influential among whites and among males, the data provide no evidence of any such effect. None of the advertising measures are significantly related to the incidence of smoking by white seniors or male seniors, even at the 20 percent level. In the single best regression for the staffs hypothesis, Camel advertising lagged two years accounted for only 22 percent of the variation in the incidence of smoking among males, and was not statistically significant Even omitting the data for 1996, however, the regressions indicated a statistically significant relationship between Camel advertising and Camel's share of young adult smokers. There are some statistically significant relationships between Camel advertising and the incidence of smoking among particular subgroups, but they are not as the staff's theory would predict. Smoking among blacks is negatively related to every measure of Camel advertising considered. For advertising lagged one year, or advertising totaled over two, three, or four years, the relationship is statistically significant at the 5 percent level. For females, all of the advertising coefficients are negative, but not statistically significant. In short, the data provide no evidence that the Michigan trends are attributable to Joe Camel. Although the time period is short, it is sufficient to find the effect that motivates Camel advertising: to increase Camel's share of young adult smokers. 10
Page 74: ozh90d00
There would be no reason for me to consult R&D documents over ten years old when planning a contemporary advertising campaign. 24. I was never familiar with either D. W. Tredermick, who I understand left Reynolds in 1980, or any of his work, including the memorandum addressed to F. H. Christopher Jr. and dated July 3, 1974. Subscribed and sworn to before me this IS day of May, 1997. A6M4- //. Notary Public My commission expires: 7- 2S- a D A A 8 ~O N
Page 75: ozh90d00
Table 3 Effects of Camel Advertising on Michigan Trend Dependent Variable: Cigarette Use in Last 30 Days (H.S. Seniors) University of Michigan Data Camel Advertising Intercept Ad Variable R Squared F N Expenditures Current Year 29.92a 0.00 (2.24) (0.05) 0.00 0.00 11 One Year Lag 31.82a -0.05 (2.35) (0.06) 0.07 0.6 10 Two year lag 29.77a 0.01 (2.72) (0.07) 0.003 0.02 9 Three year lag 28.08a 0.06 (2.8) (0_07) 0.107 0.72 8 Notes: a = significant at .01; b= significant at .05. Table 4 Effects of Camel Advertising on Market Share Dependent Variable: Camel Share of 18 20 Year old Smokers Tracker Data Camel Advertising Intercept Ad Variable R Squared F N Expenditures Current Year 2.52 0.14 0.2929 3.73 11 (2.92) (0.07) One Year Lag 2.38 0.16b (2.4) (0.06) 0.463 6.91a 10 Two year lag 2.77b 0.16a (1.45) (0.04) 0.74 19.94a 9 Three year lag 5.43b 0.1 (1.96) (0.05) 0.433 4.58 8 Notes: a= signific.ant at .01; b= significant at .05. 13
Page 76: ozh90d00
Table 1 Effects of Advertising on Attempts to Quit California 1990 Sample Coefficients, with standard errors in parentheses. notes: a=significant at.01; b= .05. Advertising Variables 6 Month Advertising Expenditures Advertising Identification Camel Marlboro Camel Marlboro -2 Log L Likelihood Ratio Test ChF SquareTe st for Model Seriously Thought About Quitting (Smoked regularly in last 30 days. N=415) Beales Ba0s5 Model, without Svertising (~8.~ I 0.18 I 28.36 (1.24) (1.31) 0.15 -0.06 (0.43) (0.37) 0.58 -0.27 0.16 -0.06 (1.24) (1.31) (0.43) (0.37) Reduced Base Model (10 variables) 0.5 -0.09 (1.21) (1.25) 0.06 -0.2 (0.42) (0.36) 0.58 -0.06 0.08 -0.2 (1.21) (1.25) (0.42) (0.36) Tried to Quit In Last 6 months (Have smoked regularly in last 30 days. N = 412) Beales Base Model, without advertising 0.79 -0.4 (0.92) (0.98) 0.12 -0.21 (0.33) (0.28) 0.87 -0.42 0.12 -0.22 (0.93) (0.99) (0.33) (0.28) Reduced Base Model (10 variables) 0.74 -0.59 (0.89) (0.93) 0.2 -0.25 (0.31) (0.27) 0.89 -0.52 0.21 -0.26 (0.89) (0.94) (0.31) (0.27) Models with no control variables 0.43 -0.82 (0.85) (0.88) 0.25 -0.25 (0.29) (0.26) 0.59 -0.73 0.24 -0.27 (0.86) (0.89) (0.3) (0.26) 347.92 347.7 353.98 353.79 353.27 353.01 524.3 523.55 522.64 521.74 547.07 546.25 543.53 542.48 566.98 563.15 562.32 0.35 0.57 5.7 0.19 0.71 0.97 0.75 1.66 2.56 0.82 3.54 4.59 22_66a 40.77 18 0.888 4.724 5.558 11
Page 77: ozh90d00
16. The potential ads for Camel were exposed to focus groups consisting of "prime prospect" smokers, that is, smokers of competitive brands in the age brackets of 18-24 years and 25-34 years, as well as among groups of Camel smokers 18 years and older. Smokers were - asked whether or not a particular ad was relevant to individuals their age, or to individuals older or younger than they were. Typically we asked a question such as "Is this ad for someone older than you, someone younger than you, or someone about your age?" Consistent with Reynolds' objectives, we rejected those ads that were deemed to be relevant primarily to individuals younger than the prime prospect, a point I emphasized in Paragraph 5 of the affidavit I submitted in response to the Federal Trade Commission's previous investigation of Joe Camel. That affidavit is dated June 23, 1993. 17. The critical reason for Reynolds' decision to proceed with Joe Camel as a brand "mascot" was based on market research findings that: a) Joe Camel succeeded in cutting through advertising clutter and being noticed. In my marketing experience, this is a common reason for adopting a non-traditional look. It is the same reason that celebrities are used for other products; b) Joe Camel was perceived as modem, and successfully modified perceptions that Camel was an old-fashioned brand with a harsh taste; c) The campaign successfully communicated that Camel in fact offered light and smooth taste characteristics that were desired by a large segment of smokers; 5
Page 78: ozh90d00
Paae Introduction And Methodolocv i Summarv Of Results v I. Product Advertising I Last Product For Which Advertising Was Seen 2 Awareness Of Advertising For Selected Products 3 Brand Awareness For Selected Products 4 Awareness Of Advertising Slogans For Selected Brands 6 II. Trade Characters 8 Awareness Of Selected Trade Characters 9 Awareness Of Products Advertised By Trade Characters 11 Ill. Reaction To Trade Characters 13 Reaction To Joe Camel 14 Feelings About Cigarettes 16 Reactions To Other Trade Characters 18
Page 79: ozh90d00
In addition, as a final (and somewhat different) measure of contin 'on, Beales examined whether advertising influenced the likelihood that a teen had d 100 cigarettes oke in their lifetime in the entire sample of 4,980 teens. If Joe encourages c ntinuing, then teens exposed to more advertising should be more likely to have smoked 100 c!garettes in their lifetime. The analysis used a logistic regression model. To control for other factors that might influence quitting behavior, Beales began with the full model that he developed to examine teenage smoking behavior. I's model incorporates 30 different variables, measuring attitudes toward risk, 'ons of the risks and benefits of smoking, family and peer influences, and demographic s.2 He then added variables measuring six month advertising expenditures for Camel d Marlboro to this base model.l In addition, he also added dummy variables to identify t who named either Camel or Marlboro as the most advertised brand. As noted in Beales' 'er work, the interpretation of these variables is somewhat ambiguous. Teens who are okers may pay more attention to advertising, and may therefore be more likely to identi one of the two leading brands as most advertised. Thus, smoking may cause answers to most advertised brand question, rather than naming a brand as most advertised causing sm~oking. Finally, Beales added the four advertising variables in combination. ~! =Because of the small number of observations, the full models could ot be estimated for teens who have thought seriously about quitting or who were former smo ers. The dummy variable identifying boys with no best friends of the same sex was omi to allow estimation. 'Beales also examined measures of advertising expenditures over the revious year, previous two years, and previous three years. None of these measures statistically significant, either alone or in combination with the advertising identificati in variables. Detailed results are not reported. 3
Page 80: ozh90d00
Advertising Character And Slogan Survey Conducted For R J Reynolds Tobacco Company November, 1993 TUIIIII IATA IITO IITEll10EICE •OI110111E
Page 81: ozh90d00
Camel's first appearance, an event that would occur in the fall of 1988. The brand's advertising agency of record, McCann. Erickson, which had substantial influence with top Reynolds' management, provided assistance in the search for the centerpiece of this celebration. In the summer of 1987 it was decided that a birthday party provided the needed centerpiece. 8. On July 15. 1987, we conducted two focus groups to explore alternative promotion ideas. 18-34 year old male smokers of competitive brands comprised one group and 18-50 year old Camel smokers comprised the other. Susan Nassar, a Reynolds' employee, showed one focus group an old French Camel poster, and I witnessed the focus group's immediate, very positive response. 9. 1 related the focus group's reaction to McCann, Erickson, which, despite the reactions of smokers in focus groups, did not like the French Camel idea, and opposed it vigorously. I believe the agency had this negative attitude, in part, because the idea did not originate within the agency itself, and, in part, because the agency felt I was too aggressive in my approach to Camel advertising. 10. 1 sought and gained permission from John T. Winebrenner, who then headed the brand management (or marketing) function, to bring in other advertising agencies to explore the French Camel concept, because McCann, Erickson was reluctant to do so. Several agencies (including, eventually, McCann, Erickson) provided Reynolds advertising ideas and executions based upon the French Camel. I l. During the late summer and autumn of 1987, Joe Camel emerged. The French Camel was transformed from a stylized head into a character with his own persona, expressions, and name. This initial work was mostly performed by the Trone advertising agency. The name 3
Page 82: ozh90d00
INTRODUCTION AND METHODOLOGY
Page 83: ozh90d00
State of North Carolina ) County of Forsyth ) ss: ) AFFIDAVIT OF LYNN J. BEASLEY I, LYNN J. BEASLEY, having first been duly swom, do hereby state the following: 1. I am Senior Vice President of R.J. Reynolds Tobacco Company ("Reynolds"), with responsibilities for the Winston/Camel/Salem/Vantage/MorelNow brands. I have been in my current position with responsibility for Winston/Salem/Camel since July 1995. As Senior Brand Manager for Camel, from June 1987 to April 1988, I directed, with the aid of several advertising agencies, the creation and development of the Joe Camel character that has been used by Reynolds in advertising campaigns for Camel cigarettes. 2. I first began working for Reynolds in July 1982. 3. From October 1984 to August 1985, I was one of a number of assistants who worked on the Camel brand. During the mid-1980s most of the Camel brand's advertising featured a model, Bob Beck, as a solitary adventurer, with the theme "where a man belongs." This theme had lost relevancy, and the Camel brand was on a decline. 4. As an assistant, I was responsible for developing retail promotions and had no responsibility for or involvement in the development of Camel advertising. Nonetheless, I was generally aware that focus group research was being performed in an attempt to find a campaign to supplement the so-called "Bob Beck" campaign. I was also generally aware of what we called
Page 84: ozh90d00
APPENDIX E Statistical Appendix I. Advertising and Quitting Behavior To test the staffs new theory that Camel advertising somehow causes teens to "continue to smoke," Beales used the 1990 California Tobacco Survey data. If Joe Camel does indeed cause "continuation;" then teens who are more heavily exposed to Camel advertising should be less likely to quit smoking than those who see less advertising. The California survey includes a number of different measures of quitting. Most questions were only asked of teens who had smoked in the past 30 days. Beales' analyses were confined to those teens for whom quitting was possible, and therefore generally excluded teens who had not smoked in the previous month. In addition, when first asked about quitting, approximately 25 percent of 30 day smokers responded that they had never smoked regularly. These teens were not asked subsequent questions about quitting attempts, and were also excluded from most of the analyses. The first measure of quitting asked teens whether they had ever thought seriously - about quitting. Data were available on 415 teens who had smoked regularly in the previous 30 days. Of these teens, 345 (83 percent) had thought seriously about quitting at some time. If Joe encourages continuing, then teens exposed to more Camel advertising should be less likely to have thought about quitting. Beales examined this question using logistic regression models. A second question asked those who had though about quitting how many times they had tried to quit Data were available for 408 teens. Possible responses were never (125 teens), once (91 teens), two or three times (106 teens), and four or more times (86 teens).
Page 85: ozh90d00
Teen Smoking Page 41 expenditure measures reported in the table are never significant. The brand advertising variables consistently have opposite signs, but which one is positive differs across equations. Results for 12 month and three year expenditures are similar (not reported). The strongest case for an effect of advertisug on smolong decisions comes from the model with ever smoked as the dependent variable. The individual brand advertising coefficients approach significance, but only Camel advertising has a positive e$'ect. Moreover, the sum of the coefficients is negative (-.06), and indistinguishable from zero (du squared =.03). Similarly, the advertising identification measures with ever smoked as the dependent variable are quite similar to the results in the ordered logistic model. The pattern persists , in modds explaining current smoking, but none of the coefi'icients are significant. With daily smoking as the dependent variable, teens who reported Marlboro was most advertised when it was are less likely to be daily smokers, but the coefficient is not significant. Models that include the identification variables but not advertising expenditures best fit the data when smoking or ever smoked are : dependent; with daily or current smoking as the dependent variable, the best fitting models exclude all advertising measures.sZ Interpreting the coefficients on the identification variables is difficult. introducing these variables results in little change in the advertising expenditure coeFficients, except when "correct" and "incorrect" measures are defined based on the six month advertising expenditure measures. If "Some of the brand advertising and brand identi5cation results change substantially when weighted models are estimated. In particular, with current smoking as the dependent variable, both brand advertising variables are significant at 5 percent. They still have opposite signs, however, and the sum of the coefficients is not significandy different from zero. Moreover, ideatifying Marlboro as the most advertised brand when in fact it is most advertised is also significant in the ordered logistic model. It is difficult to place much confidence in the weighted results, however. The effect of weighting is to reduce the weight given to teens who are smokers, particularly daily smokers- Moteover, the weighted results are dm en by a single, high weight observation. The weighted results and their problems are discussed in detail in Appendix B.
Page 86: ozh90d00
'01'07e !S s "S0'0x !S •s "i0'016 !Ssrs (8£') (fE') (IY (SC) 56485 Iz'0 60'0 6l'0* t(1'Z- (St') S£' (zt') (LC (64'1) (iYi Z8'SSS 19'0 £0'0- 6Z'0- •• 6S'1- it'Z 96'1- (EZ'I) (to'I 9C68S • SZ"Z iZl- 9L'68S spy ON (qq"A =00010931 .N (61') (LI') (OZ') (tZ') tZ96oZ •• 9t'0 • 6Z'0 •• Z>•'0 81'0 (ZZ') (81'1 (OZ) (9Z') (ZC9 (99') LBIt(IZ • SC0 Ll'0 ZE"0 • St'0 ••• E6'1 •• 8t'1- (09') 6t 9Z'9b0Z • t0"1 SS'o- i8'StOi spy oµ (al4s!+eA 5oomoloP!N aaPMS 7m++eJ (ZI' iI' EI' (SI' £t'661 t •• 8E'0 • 6i'0 00'0• El "0 (£1') (Il') (£i') (91') (St') (0t') 8£'£OZt •• 6E0 • 8l'0 [0'0• Sl'0 01'0 LO0- (8£') (i£) OS'90Zt SZ'o- oE'o t9'£ozt Sw aN ° (4qnumA wowio7o4a!Q) W~S JOA8 II') (0I') (ZC) , (tl') ' SYS(A9 •• 6£"0 •• 4Z'0 ii"0 Ol'0 (£i') 01' ZI') SI') (Zt' (8£') 8Z'8L09 •• St'0 •• IZ'0 80'0 91'0 8E'0 LE'0- (9E') (6Z') £o'5809 EO'o lo'o- t0' I809 spy ON pv oHt mSa t w+Owo gou nuJ umM M RI . oxqlnN( IMM'J °~ oJQI (UOQIeW lMUNJ Woow 9 ploqwncyq iad f ~11 e~IV WOR9u mPi u.MwWV smq)wftH Su.m!uy+PV WWeA NwwOWVmQ Put . MOTWI . POM WO qapoN[ W19QM 0! IMM3 WWPMS Pm MMJLw:) 8a!n!>'aepY VWU 9 I-S a19v1
Page 87: ozh90d00
d) The campaign, to the extent market research can predict, suggested that it would succeed in getting some smokers of competitive brands to consider Camel as their brand; e) Camel franchise, including both YAS and older smokers, found the new Joe Camel advertising appealing for their brand. 18. Even though the Joe Camel Campaign had a successful launch in January 1988, I believe McCann, Erickson tried to have me fired from Reynolds. Although I was not fired, in May 1988 I was reassigned to become Director of Special Markets, an unstaffed department with no budget. 19. Following the assignment to Special Markets and after a leveraged buyout changed Reynolds' management, I was made Vice President - Strategic Marketing Planning, a position I held from August 1989 to July 1991; from August 1991 to June 1993 I served as Vice President - Winston Business Unit; and from July 1993 to June 1995 I was Senior Vice President - Winston/Camel Business Unit. I started my present position, Senior Vice President - - Winston/Camel/Salem Business Unit, in July 1995. 20. At no time (prior to, during, or after my assignment to the Camel brand) did I attempt to create advertising or promotions that would appeal to those under 18, or to non- smokers of any age. And at no time did I ever hear or see any suggestion that minors should be a target of the advertising campaign for Camel, or for any other brand. Prior to being asked recently to review documents in connection with pending state litigation and in connection with the FTC's attempt to bring a case against Joe Camel, I had never seen a reference to "pre- smokers; '"learning smokers," or "first usual brand" used independently of the concept of "First 6
Page 88: ozh90d00
the "Funny" or "French Camel," featuring a stylized drawing of a Camel's head, which had been created in the mid-1970's for use in France, had been used for a tee-shirt promotion. Although I reported to Rick Caufield, the Camel Brand Manager at the time, I did not see the report Alicia Nance Mitchell prepared for him regarding the results of focus groups exposed to French Camel executions, nor was I aware that these executions had been exposed to smokers at all. 5. After working as Assistant Brand Manager for Camel, I served as Brand Manager for the Century brand from September 1985 to December 1986, and Senior Brand Manager for the CenturyBright/DoraVSterling/Magna brands from January 1987 to June 1987. 6. In June 1987, I returned to Camel as Senior Brand Manager. I viewed my assignment as Senior Brand Manager as an important opportunity to advance my career at Reynolds, in part because the decision to replace the Bob Beck campaign advertising had been made before I returned and I was confident I could find a replacement when others previously had failed. I immediately began looking for new campaign options that would attract young adult smokers ("YAS"), those between 18 and 24, who were the prime prospect for Camel. Our goal was to change smoker perceptions of the Camel brand. Camel was perceived as an old- fashioned brand with a harsh taste by those who even knew the brand had filtered styles. These perceptions were uncovered through marketing research among smokers 18 years or older. This research, as all Reynolds brand marketing research, was done only among people at least 18 years old who already were committed smokers, smoking at least several cigarettes a day. 7. Because we were having difficulty finding something to replace Bob Beck, it became clear to me that interim advertising of some sort was needed. The decision had been made, prior to my return to Camel, that Reynolds would celebrate the 75th anniversary of 2
Page 89: ozh90d00
Table 2 Effects of Advertising on Quit Attempts and Successful Quitting California 1990 Sample Coefficients, with standard errors in parentheses. notes: a=significant at .01; b= .05. Advertising Variables 6 Month Advertising Advertising Expenditures Identification amel Mariboro Camel Marlboro F2 Log L Number of Quitting Attempts (Logistic Model) (Smoked in last 30 days and didn't say never smoked regularly. N = 408) Likelihood Ratio Test Chi- SquareTe st for Model Beales Base Model, without advertising 11065.52 1 1 56.82a 0.66 0.14 1064.57 0.95 (0.79) (0.85) -0.22 -0.24 1064.47 1.05 (0.28) (0.25) 0.65 0.06 -0.2 -0.24 1063.63 1.89 (0.79) (0.85) (0.28) (0.25) Former Smoker (Smoked at least 100 cigarettes in lifetime. N = 309) Beales Base Model, without advertising 214.26 84.16a 1.55 -3.52b 210.07 4.19 (1.54) (1.77) -0.95 -0.4 211.14 3_11 (0.55) (0.43) 1.66 -3.74b -1.01 -0.44 206.57 7.69 (1.56) (1.8) (0.55) (0.44) moke 100, Full Sample Beafes Base Model, without advertising 1248.47 1067.89a -0.35 0.4 1248.04 0.44 (0.66) (0.7) -0.48b -0.35b 1243.31 5.17 (0.22) (0.19) -0.4 0.34 -0.48b -0.35 1242.86 5.61 (0.67) (0.7) (0.22) (0.19) 12
Page 90: ozh90d00
Page 91: ozh90d00
i significant. Moreover, both Camel and Marlboro advertising expenditures are associated with more quitting attempts, not fewer, as the staff s hypothesis would require. The only statistically significant coefficients consistent with the staffs hypothesis are reported in the second panel of Table 2, for the models. explaining the likelihood that a teen who has smoked 100 cigarettes in his or her lifetime is a former smokers. Unfortunately for the staff s theory, however, it is Marlboro advertising expenditures that are associated with a significantly lower probability of having successfully quit smoking, whether the advertising identification measures are included or excluded. The likelihood ratio test, however, cannot reject the joint hypothesis that all of the advertising coefficients are zero. Teens who identify either Camel or Marlboro as the most advertised brand are less likely to be former smokers, but the coefficients are not statistically significant Camel advertising expenditures are associated with an insignificantly higher probability of being a former smoker. The results for former smokers are particularly suspect, both because a highly relevant variable is omitted and because of the potential for reverse causality. The omitted relevant variable is the level of cigarette consumption, because we have no data on the level of consumption for former smokers. Quitting a pack a day habit, however, is likely to be far more difficult than quitting among those who smoke only one or two cigarettes per day. Indeed, analysis of the longitudinal data from the TAPS II sample indicates that teens who smoked fewer than three or four cigarettes per day in 1989 were more likely to have reduced their smoking status in 1993 (i.e., stopped smoking daily or quit altogether), and less likely to have become daily smokers, than those with higher levels of consumption.' It is less clear 'Reynolds comment to FDA, Figure 9, p. 65. 6
Page 92: ozh90d00
6T • -1 -Ib IV. Sources Of Awareness Of Trade Characters Ways Seen Or Heard About Joe Camel I Ways Seen Or Heard About Other Trade Characters Where Usually See Or Hear About Products In General 21 23 24 V. RespQndent CharacteHstics 26 AppgpdJx Sampling Methodology The Questionnaire SupenvisorRnteniiewer Instructions 27
Page 93: ozh90d00
"Joe" was selected for two reasons: it was the name of the circus camel upon which original Camel packaging was based, and the name also fit the image which was Camel's intended franchise -- an "ordinary Joe." To assure myself the new direction was not off the mark, I showed the new Joe to the line workers at Reynolds' manufacturing plants and was pleased with their positive reactions to it. 12. While the transformation of the French Camel into Joe Camel was taking place, McCann, Erickson lobbied Reynolds' top management, identifying the different direction it believed the brand should take. McCann, Erickson expressed the view that Joe Camel made fun of a serious Reynolds' brand that needed a "male bonding" theme rather than a humorous, light- hearted camel. 13. I knew that it would be difficult, if not impossible, to overcome McCann, Erickson's clout with top management if I worked through ordinary channels. Therefore, I decided to take the idea directly to several company executives to gain their support. I realized that this was a risky course of action and that my future at Reynolds was tied to the success or failure of the Joe Camel idea. If Joe Camel failed, I would likely have had no future at Reynolds. 14. Following my presentation of the Joe Camel idea to top management, it was approved. 15. Throughout the Campaign's development, the thrust was to treat Joe Camel as an adult over the age of 25. When the agencies developed potential advertising to achieve this goal, we conducted consumer research to assess how well it succeeded. By late 1987, it became clear to me from positive focus group reactions and other research, that Joe Camel was a good replacement for Bob Beck. Xb 4 m w
Page 94: ozh90d00
1. PRODUCT ADVERTISING
Page 95: ozh90d00
coefficients should all be negative for the quitting measures, and positive for the likelihood of smoking 100 cigarettes in a lifetime. In fact, however, for teens who have ever thought seriously about quitting or who have attempted to quit in the previous six months (Table 1), none of the advertising coefficients are statistically significant, and only one coefficient exceeds its standard error.s Moreover, in every model, the likelihood ratio test for the joint hypothesis that all of the advertising coefficients are zero cannot reject the hypothesis. Although the coefficients in Table 1 are insignificant, there is a very clear pattern in the results. All of the coefficients involving Camel advertising have the wrong sign for the staff's hypothesis. Thus, if these coefficients show anything, it is that teens exposed to more Camel advertising are more likely to have seriously thought about quitting, and more likely to have tried to quit in the last six months, as are teens who identify Camel as the most advertised brand. In contrast, all of the Marlboro coefficients are as the staff would expect: teens who see more advertising, or name Marlboro most advertised, are less likely to have seriously though about quitting or to have tried to quit in the past six months. Again, however, all of the coefficients are insignificant, whether tested individually or jointly. The results are basically similar for the number of quitting attempts, reported in the top panel of Table 2. Because the Beales base model was highly significant, only the results when advertising is added to the full base model are reported. All of the advertising coefficients are insignificant. Although naming either Camel or Marlboro as the most advertised brand is associated with a smaller number of quit attempts, the effect is not 'That coefficient is the one for identifying Marlboro as the most advertised brand in the model of quit attempts in the previous six months, with four advertising variables and no control variables. It is the last coefficient in Table 1. 5
Page 96: ozh90d00
II. Camel Advertising and Smoking Among High School Seniors The staff contends that the increase in smoking among high school seniors as measured by the University of Michigan's Monitoring the Future study is somehow evidence of a delayed effect of Joe Camel. The additional years of Michigan data, along with changes in Camel advertising since the early 1990s, make possible a test of the staffs contention. Beales examined the relationship between Camel advertising expenditures (in newspapers, magazines, and out of home advertising) and the incidence of smoking in the past 30 days as measure by the Michigan study. He also examined the relationship between Camel advertising and Camel's share of 18-20 year old smokers as measured by RJR's internal tracking data. The analysis covered the period from 1986, before the Joe Camel campaign began, through 1996. Results are similar, but statistically weaker because of the smaller sample size, if the analysis is restricted to the time period since the Camel campaign began in 1988. With only 11 observations, detailed attempts to develop statistical models are not likely to be very productive. Simple analysis of the relationship between advertising and smoking measures, however, is highly revealing. The impact of Camel advertising on smoking by high school seniors is examined in Table 3. The impact of advertising on Camel's share among young adult smokers is examined in Table 4. The data reveal no relationship between Camel advertising and the incidence of smoking. Whether current year advertising expenditures are used or advertising expenditures from previous years are used, there is simply no relationship between the Michigan trends and Camel advertising. The advertising coefficient is always insignificant, and sometimes (with one year lagged advertising) has the wrong sign, indicating that more camel advertising is A A J v
Page 97: ozh90d00
Usual Brand Young Adult Smokers" ("FUBYAS"). I had never seen, let alone read, any of the company memoranda referenced on the FTC's "new evidence" list, except the Diane Burrows piece which I do not recall reading. 21. I know that in 1984, I attended a meeting where Diane Burrows presented her research on the importance of targeting young adult smokers who were selecting a first usual brand. The meeting was well-attended by Reynolds' marketing employees. The conclusion I drew from Burrows' report was the importance of a brand's image fitting with the prime prospect, and that smokers in the 18-24 group were at an age where a first usual brand was most frequently selected. It has always been clear to me that Reynolds does not target underage smokers nor attempt, in any way, to influence a person's decision to start smoking. With the exception of material prepared for the FTC in connection with its Joe Camel proposed complaints and material shown to me in connection with litigation, I cannot recall ever having seen any marketing data concerning smokers or non-smokers under the age of 18, and I know I never relied on such data to make decisions regarding future marketing activities. Moreover, I never discussed or heard discussed marketing any Reynolds' product to underage youth. 22. When the Joe Camel Campaign was conceived and implemented, I was unaware of Claude Teague, his documents, or his views onfatty subject. Indeed, I never head of Mr. Teague until his memos (draft memos in some cases) became the subject of various news reports in the media. It was not possible that Mr. Teague, or anything he said or wrote, influenced me or my decisions during the development of the Joe Camel Campaign. 23. Likewise, I was unaware of any draft or final version of the Research and Development Department's 1976 Planning Assumptions and Forecast for the Period 1977-1986. 7
Page 98: ozh90d00
associated with a reduction in smoking. The regression models are not significant, and explain a trivial fraction of the variation in smoking. In contrast, Camel advertising is clearly related to Camel's share of 18-20 year old smokers. All four advertising measures examined are statistically significant at the 10 percent level. In the best fitting model, with advertising expenditures lagged two years, Camel advertising expenditures are significant at the one percent level. Camel's advertising clearly does influence its share among 18-20 year old smokers, but it does not influence the incidence of smoking among high school seniors. Beales also explored several different variations of the measurement of advertising expenditures, with basically similar results. Total advertising expenditures over the previous two, three, or four years significantly increase Camel's share of 18 to 20 year old smokers, but are not significantly related to the incidence of smoking among high school seniors. The published Michigan data provide only two year averages for demographic subgroups, and data are not yet available for 1996. Using the period from 1986 through 1995, Beales examined whether Camel advertising might be significantly related to smoking incidence among particular demographic groups. Four different groups were considered: whites, blacks, males, and females. The Michigan data on smoking incidence of smoking among each group of high school seniors was regressed on each of the measures of Camel advertising expenditures discussed above (current advertising, advertising lagged one to three years, and advertising totaled over two through four years).7 "As a check of whether the missing data for 1996 might influence the results, Beales reestimated the models explaining Camel's share of 18-20 year old smokers over the period from 1986 through 1995. These results were basically unchanged. 9
Page 99: ozh90d00
A Young people most often say they have become aware of Joe Camel from billboard ads (5196) and magazine ads (4596). The next most frequently claimed source of aware- ness is television (4296)--an indication of'N's pervasiveness In the Gves of young peopie n in sum, there Is no evidence from this study that the character. Joe Camel, Is more recognizable to young people than other of this advertising genre, or that he has been successful in maldng Camel the most notable brand of cigarettes among young people, or that he creates any idnd of po 's~tive image of smoidng for youth.
Page 100: ozh90d00
Teen Smoking Page 53 REFERENCES Amaniya, Takeshi (198 1), "Qualitative Response Models: A Survey," Journal of Economic LiteratuEc, Vol. 19 (December), p. 1482. Bishop, John A. and Jang H. Yoo (1985), "Health Scare, Excise Taxes and Advertising Ban in the Cigarette Demand and Supply," Southern Economic JournaL Vol. 52 (October), pp. 402-41 l. Bloch, Harry (1974), "Advertising and Profitability: A Reappraisal," Journal of Political FQmn4lay, Vol. 82, p. 267-86. Demsetz, Harold (1979), "Accounting for Advertising as a Bamer to Entry," ]ourneLQf @usiII= Vol. 52, p. 3,45-60. Feighery, Ellen, David G. Altman, and Gregory Shaffer (1991), ".The Effects of Combining Education and Enforcement to Reduce Tobacco Sales to Minors," Journal of the American Medical Associatvol. 266 (Dec. 11), p. 3168, Goldstein, A. 0., P. M. Fischer, J. W. Richards and D. Creten (1987), "Relationship between High School Student Smoking and Recognition of Cigarette Advertisements," Journal of Pediatrics_ Vol. 110, pp. 488-9L . -_ ~ ... Hamilton, James L. (1972), "The Demand for Cigarettes: Advertising, the Health Scare, arid ' the Cigarette Advertising Ban," Review of Economics and Statistics_ Vol. 54 (November), pp. 401- I1. Jason, Leonard A., Peter Y. Ji, Michael D. Anes, and Scott H. Birkhead (1991), "Active Enforcement of Cigarette Control Laws in the Prevention of Cigarette Sales to Minors," Journal o f the American Medical Association, vol. 266 (Dec. 11), p. 3159. of Kandel, Eugene, and Edward P. Lazear (1992), "Peer Pressure and Partnerships," Journal Political Economv, vol. 100, p. 801. Kmenta, Jan (1986), Elements of Econometrics (2nd Ed.). Landes, Elisabeth M. and Andrew M. Rosenfield (1994), "The Durability of Advertising Revisited," Journal of Industrial Economics_ Vol. 24 (September), p. 263-76_ Lewitt, Eugene M., Douglas Coate and Michael Grossman, "The Effects of Government Regulation on Teenage Smoking." Jounnal of Law and F.corbmics Vol. 24 (December), pp_ 545-570. M06kft and Mortaltv WWdy $qw (1991), "Cigarette Smoking Among Youth - United States, 1989", Vol. 40 (Oct. 18. 1991) p. 712.
Page 101: ozh90d00
Variable Smoking helps when bored Smoking helps relax Smoking belps with sUesa Smoking helps, social sitnations Enjoy risky activities Occasional cigarette harmfW / Safe to smoke (a year or two) Dangeran, but [ ean quit Same sex best friends (boys and girls); Opposite sex b" Bieds Added effect for boys People yiw Imrnr Steady who smdces No best Girndt (Boys only) Adult smokers in home (number) Mom Smokes; Dad smokes No adult; No dad; No ntutn in household Opp. sex older sibs smoke; Same sex older sibs slfatlCe Household size Nights out (pa week) Family income No income data Nonwhite Hispartic boys Not in school APPENDIX A Detailed Variable Definitions Ouestion/Defiai~ Aisk/Utilip• Variables Do you believe smoking can help people when they are bored? I = yes, 0 otherwise. Do you believe cigarette smoking helps people relax? 1= yes, 0 otherwise. Do you believe cigarette smoking helps tYduoe suwss? I=yes, 0 otherwise Do you believe smoking helps people feel more eomfortable at parties and in other social situations? t=yes, 0 otherwise I get a kick out of doing things every now and then that are a little risky or dangerous. 3= agttx; 2= no qpinion or dont latow; I= dtsagree Do you believe there is any harm in having an occasional cigarene? 1= yes. 0 otherwise. Do you believe its safe to smoke for only a year or two? l= yes, 0 orbawise. [lf t started to smoke regularly], I could srop smoking anytime I waored. (agme, disagree, no opinion).. = I if Safe to mwolm =0 and ag[oe cm quit any time, 0 otherwisr Pear Variables . - fraction of Giendti of same sex (opposite sex) who are smokers. Set = 0 for kids with no Giends of same (opposite) >tix. TAPS: Of yow four best (male, female) friends, how many of them smoke? Ca6fornla: About how many best frieods do you have who are (male, female)? Of your best friends that are (male, female), how many ofthem smoke? interaction tertn, = a male dummy times Satne sex best friends. How many of people you know, who are about your age, smoke cigarettes? 0 = now, - 1= a few, 2= some, 3= most Have you ever had a steady (boyfnetd/girl@tend)?' Did (hVshe) smoke cigarettes? I = Ye5 t0 bOtb, 0 Otberwlse. . . . dummy variable = 1 for boys who have tto best 6ietd4 of the same sex Family [nflnence Variables Number of adults in household who smoke, incl4slinC mom and dad. = I if mom (dad) is io household and a smoker, 0 otherwise = I if no adult (dad mwm) in the household, 0 otherwise. TAPS: No adult = I if . no parent or adult relative in the household. Catlfornia: No adult = I if no parent or guardian in the bousehold - TAPS: dummy variable = I if teen has an older sibling of the opposite sex (same sex) who smolCea, 0 otherwise. California: Fraction of older siblings of opposite sex who smoke. = 0 if no siblings of opposite sex. Family size (Califorttia only). Other Variables Including Saturdays and Stmdays, how many nights a week do you usually go out for fim or recreation? Family inoome. Coded as continuous, based on midpoints of ranges in the smvey, and oenws mean for open ended top utegoty. = 0 if income data missing. = I for teeos where thete is no household income data. = I for nonwhites, 0 otherwise. = I for Hispanic boys, 0 o0terwise. = I for teens not in school, 0 otherwise
Page 102: ozh90d00
Ill. REACTION TO TRADE CHARACTERS
Page 103: ozh90d00
2 Young people are exposed to advertising for a wide array of products. When asked to name the last product for which they recall seeing advertising (only one product was to be named), many different product categories are recalled-with cereal, toys, and beverages leading the 6st, each being cited by about one person tn ten. Cigarettes Is one of th9 least recalled product categories, being mentioned by only 2% of youth. A7 ' i o1y ;~6i 1/.  1 1iti ~ .. i1 ~ 'What was the last product that you recall seeing adveitising forY (Q2) Total Me x Cereal ` 11 Toys 10 Beverages 9 Fast food restaurants 7 Heanh a Deauty aids 6 Carsnaiacs 6 snadctood 5 AtMetic shoes/sneakerslshoes 4 Batteries 4 V'deo yameslcarUidges 4 Soap/soap products 4 Foods 3 Hair products 2 CIGARETTFr3 2 Jeans 1 Clothes (other than jeart4) I At4o/video equipment I Other 10 Don't Ivrow 10 (Number of irderviews) (1117)
Page 104: ozh90d00
7 AWARENESS (UNAIDED) OF ADYERTISING SLOGANS FOR SELECTED BRANDS 'Now I'm going to name some brands of products. For each brand I read, please tep me the slogan or phrase that is used to adysrtise that hranai Il you are not sure or donT know, just say so.' (0.4) Total youth Brand siorodtct Nice athletic shoes Sbgao neme,d Correct: (Just do R) 52% Other 10 DCnt larow 38 Kedogg's Frosted Flakes Corred: (Th" great) 33% Other 38 Don't Iviow 29 Trix Correct: (Tr6r Is for ldds) 30% Other 27 Dont iQww 43 McDonald's Correct: (What you want is what you get) 9% Other 35 Dontknow 56 Diet Pepsi Correct: (You've got the ripht one, baby-Uh Huh) 3% Other 59 Don't know 39 Miller Lite beer Corred: (Great taste, less fil&g)) 3% Other 18 Dont laiow 79 Newport cigarettes Correct: (Alive with pleasure) '9'. Other 5 Dontlmow 95 CAMEL CIGARE7TES Correct: (Smooth character) •9G ~ Other 14 Don't Iviow 86 Marboro cigarettes Correet: (Come to where the flavor Is) 0% I Other 10 ~ Don't know 89 a ~ (Numberof interviews) (1117) m ' Lete thaa 0.57G. ~An .• 0
Page 105: ozh90d00
4 Mariboro Is, by far, the cigarette brand most frequentfy recalled. For four product categories-athietic shoes, jeans, beer, and cigarettes-young people who had seen or heard of advertising for the category were asked what brands they are aware of. in each instance, one brand stands out. (See table on following page.) For cigarettes, Marlboro Is the brand most often recalled, named by almost half (47%) of those pware of cigarette advertising. Camel Is In second place, named by 26%. This reiatlve position Is the same for all age groups. White youth are much more likely to recall the Marlboro brand (53%) than are black youth (23%). Among black youth, the most frequently recalled cigarette brand is Nevw port (31 %), compared to 5% for white youth. There is no significant difference between the races for the Camel brand.
Page 106: ozh90d00
APPENDIX Sampling Methodology The Questionnaire Supervisorflnterviewer Instructions .
Page 107: ozh90d00
IV The complete questionnaire is included in the Appendix. Validation Once all questionnaires had been received by the Roper Starch field department, 20% of each interviewers assignment was validated by telephone calls to the respondents. This procedure is undertaken to insure that the Interviews were, in fact, conducted in a proper manner with the proper respondent. No problems were uncovered. Aii questionnaires were edited for completeness, responses to open-end questions were coded for quantification, and computer processing was done by in-house person- nel and facilities. Coding categories initially are established by coding a sample of the questionnaires. These categories are submitted to the project director for approval Further categories are added, if needed, as the coding progresses. Throughout the process coders check each others' work and at the end there is a bgic chedc to be sure the categories 'make sense• relative to the question asked. To account for the over-sampling of black youth and to correct for the deviations from the population sampled that occur in any survey sample, weighting adjustments were made based on sex, age, race, and region of the country. The sampling error associated with the total sample of 1,117 respondents is approxf- mateiy plus or minus 4 percentage points at a 95 precent confidence interval. For the sample of 382 black youth, the sampling error is approximately plus or minus 6 percent- age points at a 95 percent confidence intervaL The computer rounds off each percentage to the nearest whole number. As a result, the percentage for a question that allowed only one response may vary slightly from 100. Of course, where a question permits multiple responses, percentages may be much greater than 100, depending on the number of responses each respondent gives. An asterisk (') indicates a percentage less than 0.5%.
Page 108: ozh90d00
IV. SOURCES OF AWARENESS OF TRADE CHARACTERS
Page 109: ozh90d00
17 . . a• l ::a~l (Asked of a/l those respondents who identifyJoe Camel with cigarettes.) What do you persortally think about agarettes? Anyfhing eise?' (Q.8b) Youth who Identify Joe Canwl >edteslaacettes x Cigarettes are bad for your heakh 51 Hate thertVdon't Cice theMdont smoke 32 Shigross/disgus6ng/yuk 15 They cause cancedlung cancer 13 They smell bad 10 Bad for lurgs 7 Can kill you 6 qddktWe/bad habit 5 Expensivdcost too much 2 LicelhertVWre o.k 3 (Number of irdervieMrs) ON)
Page 110: ozh90d00
V. RESPONDENT CHARACTERISTICS
Page 111: ozh90d00
I I Not only Is claimed awareness of trade characters hi'gh, but correct recall of the kind of products advertised by the characters Is equally high. Those young people who claimed awareness of a trade character were shown the character' and asked what kind of product is advertised by the character. (This was asked for seven of the ac.tuai characters and the bogus character. The Trix Rabbit and Flsie the Cow were not incAided.) For aq seven of the actual characters correct unaided identificalion of the product adver- tised Is at the 92% or above level. (See table on following page.) Joe Camel Is Identified correctiy as advertising cigarettes by 95% of those youth who claim awareness of the character. This level of correct product Identification is similar across all ages, but higher for whites (96%) than for blacks (89%) ' The dapiction d the diaracta was wch as to giva no dw as b the adwrused product
Page 112: ozh90d00
26 AM 10 years 15% 11 years 12 12 years 11 13 years 14 14 years 14 15 years 10 16 years 12 17 years 12 B2X Male 51 % Female 49 0 White 77% Black 16 Asian 2 Other 6 Region Northeast 21 % Midwest 30 South ' 28 West 21
Page 113: ozh90d00
14 Reactions to Joe Camel are virtually devoid of any suggestion that the character is enticing young people to smoke. When those young people who identify Joe Camel with aigarettes are asked how they would describe Joe Camel, whaYs the character qce, only 2% of youth respond that he tries to get kids to smokeE (See table on following page.) For three other characters, significantly more young people describe the character in terms of trying to get idds to buy or use the produc~-Tony the Tiger (12%), The Jolly Green Giant (11 %), Ronald McDonald (8%). Of the top three responses-the only three voiunteered by at least a third of youth--two are simply descriptive, namely, that he smokes (37%), and that he's a camel (34%). The other aooords Joe Camel the youthfully positive attribute of'coolness' (35%). Unlike the descriptions of the other trade characters (See pp.18,19), there are consid- en3bly fewer, if any, mentions for Joe Camel of such positive attributes as funny, cute, happy, athletic.
Page 114: ozh90d00
9 Trade characters are a memorable advertising device for young people. The names of nine trade characters currently in use and one bogus character (to obtain a measure of 'noise' or guessing) were read to the respondents and they were asked to Indicate whether or not they have seen or heard of each one. If a respondent was not aware of a character, it was shown, and awareness asked about again. (See table on following page.) Trade characters register well with youth-claimed unaided awareness of seven of the nine characters Is at the 91 % or above IeveL With aided awareness Included, claimed awareness levels for these seven characters is 97% or more. Only Elsie the Cow fares poorly (27% unaided awareness; 28% aided). Joe Camel, although scoring high In awareness, falls significantly below the top seven characters. Total awareness is 86% (73% unaided; 13% aided). Guessing, using the bogus character-Adam the Android--as an indicator, registered only 7% awareness (4% unaided; 3% aided).
Page 115: ozh90d00
5 : I: ; . . : 11• : ; : . ot- r$: $11 1• ; OTIM b (Asked for athletic shoes, beer, cigarettes, and jeans, if respondent had reaendy seen orheard of any adverasing for them.) 'What brands of (rtame of mmduct) have you seen or heard of?' (Q 3a) Youth aware of adveMtislnp for Product catspory PAVIUCt (Nlrnber Ot )mends>rsl grarkft ramer Athletic shoes (1000) Nice 71% Reebok 35 LA. Gear 5 Jeans (926) Levl 58% Lee 15 Guess 10 I4aona 7 Beer (881) BudweOser 43% Coors 12 Bud Liglt 12 Miller 11 MikrLite 8 Busch 5 Cigarettes (661) Marboru 47% Camel 26 Newport 12 Kool 8 Winston 5 • eronds namoa by 5% «moro.
Page 116: ozh90d00
8 12. Sex of respondent: Male ....................1 50/ 13. Race of respondent: Female ..................2 Whitd ...................1 51/ Black ...................2 Asian ...................3 Other ...................4 14. Circle (1-9) for geographic region from samples 1 2 3 4 5 6 7 g 9 52/ NE MW S w [ GO TO NEXT PAGE ] I CERTIFY THAT THIS IS AN HONEST INTERVIEW, TAKEN IN ACCORDANCE WITH MY INSTRUCTIONS. INTERVIEWER (PRINT NANE): SIGNATURE: DATE OF INTERVIEW:
Page 117: ozh90d00
W Prior to fielding the survey, a 50 interview pretest was conducted to ensure that the questions were understandable, the flow of the interview was smooth, and the length of the inteririew was reasonable. 11.:..-'1,11= Because the subject of cigarettes In general and Joe Camel In particular is a sensitive Issue to survey among young people, the research was designed to explore the subject within the broader context of the use of trade characters In product advertising. This approach, in addition to eliminating the bias that would result from an obvious cigarette study, provides a meaningful context within which to evaluate the Joe Camel character. The'questioning proceeded as follows: 1. Unaided recall of the last product for which advertising was seen. 2. Unaided awareness of any ads for each of ten product categories, including cigarettes. • . 3. Unaided awareness of brand advertising for each of four product categories, inciuding cigarettes. 4. Unaided awareness of the slogans used to advertise nine brands, including Camel cigarettes. 5. Unaided awareness of ten trade characters, induding Joe Camel. (Also Included was a bogus character, as an indication of the level of guessing.) For each character not heard of, aided awareness was asked using a depiction of the character without product or brand identification. 6. Unaided awareness of the product advertised by each of eight of the characters, including Joe Camel. 7. Unaided description of two of the characters for which a correct product identification was made, always including Joe Camel where appropriate. 8. Unaided personal feelings about cigarettes - asked only of those who correctly identified Joe Camel with cigarettes. 9. Unaided sources of awareness of Joe Camel (or of another character if Joe Camel was not identified with cigarettes). 10. Unaided sources of awareness of products in general.
Page 118: ozh90d00
Awareness of advertising for a broad range of product categories Is very high. Asked if they have seen or heard any ads in the last few months for each of 10 kinds of products, over three-fourths of young people say'yes' for nine of them•-ranging from 90% for cereal to 77% for video games. While over half of youth-;58%--are aware of cigarette advertising, it has the lowest level of awareness of the ten product categories shidied. Awareness is considerably higher In the West (74%) than elsewhere In the oountry (54%). : .1 1 ,:: . ti ~ %! al ~ •i efl p;•ftlb Tm going to read you some ldnds ofpnoatrccs. For each one please tell me whether or not you have seen orheard any ads for thatldnd of product in the last few months. Please dDnt guess,.if you don7lmow oraren7 sure, just say so.' w) Total youth Yes, have Have not seen or seen or pon't h211ed heaed tmsm[ Cereal 90% 8 2 Cars 89% 8 2 Fast foods - 89% 8 3 Athletio shoes 89% 9 2 Batteries 86% 11 4 Jeans 82% 16 2 Cookies 79% 17 4 Beer 78% 18 4 Video pames 77% 19 4 CIGARETTES 58% 37 5 (Number ot interviews:1117)
Page 119: ozh90d00
Awareness of Joe Camel comes from a variety of sources. Leading the fst of volunteered responses as to the ways that they might have seen or heard about Joe Camel Is billboard ads (51%), followed by ads in magazines (45%). (See table on following page.) The penrasiveness of television as an advertising medium is evidenced by over four young people in ten saying that television is the source of their awareness of Joe Camel, even though there are no cigarette ads on TV. For all the other trade characters, teievision surpasses all other sources of awareness by a very large margin. (See p.23.) 21
Page 120: ozh90d00
ri 4,_: • : t- There would seem to be no question about the memon3biCity of trade characters as an advertising device to reach young people. Of those characters Included in this survey, all but one enjoy very high levels of claimed awareness and of identification with the product categories advertised. Joe Carhel is no exception-nor, however, is he the best known of the characters studied. While over seven young people in ten (73%) register unaided awareness of Joe Camel and another 13% aided awareness, he places eighth in recognition of the nine characters evaiuated-significantiy below the other seven. Even given the high level of recognition of Joe Camel and the high identification of him with cigarettes, the most often recalled cigarette brand is Marlboro-almost 2 to 1 over Camel. Slogans used to advertise products are another issue entirely from the perspective of youth. Of nine well-advertised brands, the slogan of only one--PGke athletic shoes-is recalled by half of young people and two others-Frosted Flakes and Trix-by about three youth in ten. Slogans for three cigarette brands-Camel, Marlboro, and Newport-have virtually no recaii. - The impression made by Joe Camel on those who associate him with cigarettes is by no means one of enticing young people to smoke. Only 2% of youth answer'tdes to get kids to smoke" in response to a question asking how they would describe Joe Camel, what's the character like. Most frequent responses are purely descriptive, although a third endow him with the attribute of "cooiness' Only an extremely few of the young people who identify Joe Camel with cigarettes are personally inclined to view ciga- rettes in any kind of favorable light. All responses but one to a question asking their view of cigarettes are decidedly negative-the most frequent being cigarettes are bad for your health (51 qo) and don't like them, don't smoke (32%). Only 3% of young people say they like cigarettes or they are o.k.-and this response is almost entirely from the age 16 - 17 year group. N
Page 121: ozh90d00
16 Young people have decidedly negative feelings about cigarettes. Youth who identify Joe Camel with cigarettes have nothing good to say about ciga- rettes. When asked what they personally think about cigarettes, the most frequent response (51 %) is that cigarettes are bad for your health. This is followed by the very personal response that they hate or don't 5ke them or don't smoke (32%). AII the rest of the volunteered responses, except one, are negative. (See table on following page.) Only 3% say they Gke cigarettes or that they are o.k-and this response comes primar- iy from the older age group (16 -17 years). Typical responses from those few who say they Glae cigarettes or that they are o.k are as follows: donT smoke--if They are alright f anyone does, that's their problem. ft's not good for you but smodairg's o.k !t's an ind'rvidual thing. They're o.k Maybe I'll smoke later on. I like them. They're good because I smoke them. I think it's a personal thing if people want to smoke or not I think they're o.k if not over used. I like it 'cause I smoke myself. There's nothing really wrong. !Ys up to the person. I donY mind them. They're cool for people who like to smoke. They're o.k I guess. O.k for some people. . l think smoking is o.k for those who wish to smoke. Makes you look cool.
Page 122: ozh90d00
22 WAYS SEEN OR HEARD ABOIJT JOE CAMEL (Asked of a11 those respondents who IdenbfyJoe Came1"Oh dgarettes.) 7hink again about Joe Camel. Please tep me the ways that you might have seen or heard about this character. Anything elseY (0.10) Youth who Identlfy Joe (:amel wlth claarettes ~ Billboard ad 51 Ad In magazine 45 Ad on television 42 Ad in store 32 T-shirt 22 On the prodict 4seM 21 Ad il rrevrspaper 12 From friends 6 From someone In famly 4 Racbo ' 2 Clothing 2 Ughters 1 Catalogs 1 Posters Sdtool 1 (Number of interviews) (909) ' Lase than 0.5%,
Page 123: ozh90d00
24 As would be expected, television Is, by far, the place where young people claim to see or hear about products in general. When asked where they usually see or hear about all the kinds of pnoducts they want or buy, youth cite TV (85%) with far greater frequency than any other source. i/. : •: . 14:: _:• 1 ;oTayb 'Now, Wnking about all the Rinds of products ihffi you want or that you buy, where do you usually see or hear about them? Anythung efse?' (Q.11) TOWMnh 96 Ad on television 85 Ad in magazine 36 Ad In store 31 From friends 31 Ad in newspaper 20 From someone in.fanuly 16 BflIboard ad 13 On the product Rseq 11 T-shiR 6 Racfw 6 Catalogs 2 School 2 Clothing Posters (Number of hterviews) (1117) ' Leaa than 0.5%.
Page 124: ozh90d00
6 The Camel cigarette advertising slogan, "Smooth Character," has not registered with young people. Young people were read nine brand names, including three cigarette brands, and asked for the slogan or phrase that is used to advertise each brand. (See table on foibw3ng page.) The cigarette brands fare the most poorly with respect to advertising slogan recaU. None recall the slogan for Marlboro cigarettes, and only four young people (less than 0.5%)'recall the slogans for Newport and Camel cigarettes. Only for Nike athletic shoes do a majority (52%) name the corred slogan-'Just do it.' The slogans for two other brands-Frosted Flakes ('They're great7 and Trix ("Trix is for kids'}--are named correctly by about three youth in ten. Fewer than one young person in ten is able to recall correctly the advertising slogan for McDonaid's, Diet Pepsi, and Miller Ute beer. Examples of the incorrect slogan responses for Camel cigarettes are... Joe the Camel Be lAce the Camel Cool Joe Camel Be cool man Joe Cool Great tasting, cheaper now I'd walk a mile lor a Camel Partially correct slogan responses for Camel cigarettes are... Smooth Cool smooth The smooth one Smooth Cool Joe Examples of the incorrect slogan responses for Marlboro cigarettes are... The Marlboro man !ts for outdoorperson Marlboro oountry/land Make you go back to the wild Lfght up a Marlboro The Madboro adventure team
Page 125: ozh90d00
x / 52086 4546
Page 126: ozh90d00
I INTESVIEWER: The trade cards have been sealed in plastic. After opening, please check to be sure you have received all ten cards pictured on this page and place the appropriate letter on the back of each card. TE1NE YOU....
Page 127: ozh90d00
6 9. (WRITE IN NAME OF CHARACTER): 41/ 9a. How would you describe (name ef character)7 What's the character like? 42- 43- 44-
Page 128: ozh90d00
15 (Asked of all tbose respondents who idenaly Joe Came! wih cigarettes.) "How wnuld you desaibe Joe Came11 Whaft the ourwer We?' (Q.Ba) Youth who IdenUy Joe Camel with Ogarefts He smokes/smokes Camelyhas cigarette in his moufh 37 Realy oooYacts oooUtFft he's cool 35 He's a camel 34 Wears suaqlasses 15 Advertises; seds cigarettes 11 He's smooUVsAdk0suave 5 Frfentlly/has a lot of idendslsomeone who La fuNattracts people 5 Tries to pet kids to smoke 2 (Number of interviews) (909) (Nole: Cewr than iaa rosponio, oey tlase respom.s yiwn by 5% or more aro sAowa)
Page 129: ozh90d00
52086 4552 ,
Page 130: ozh90d00
In case my supervisor wants to verify that I conducted this interview, may I please have your name, address, and phone number? RESPONDENT INFORMATION: NANE: ADDRESS: ZIP CODE: 53-57 TELEPHONE NO: (AREA CODE) (NUMBER) That concludes this survey. Thank you for your help. ( GO BACK TO PREVIOUS PAGE ]IND COMPLETE INFORMATION ] 80-2 I
Page 131: ozh90d00
Roper Starch Worldwide Sa* #343'189 New York, New York October 1993 ADVERTISING CHARACTER AND SLOGAN SURVEY_ SUPERVISOB/INIERVIEWER INSTRUCTIONS STUDY OVERVIEW This is a survey among young people between the ages of 10 and 17 about various characters and slogans that are used in product advertising. At the conclusion of the interview, you must record the name, address, and telephone number of the respondent for the purpose of validation: You must also sign the certification statement at the end of the questionnaire. TOP OF OUESTIONNAIRE Be sure to complete all of the information at the top of the questionnaire. MATERIALS In addition to the materials provided to you by our field department, you will have... The instructions The questionnaire 10 exhibit cards - one for each of the characters to be asked about. RESPONDENT ELIGIBILITY Within a selected household, the eligible respondent is the person in the household who is between the ages of 10 and 17 and has had the most recent birthdav. If this person is not at home, find out the best time to call back to speak with hinJher.
Page 132: ozh90d00
10. Think again about (name of character from o.8). Please tell me the ways that7 you might have seen or heard about this craracter. PROBE: Anything else? (ASK AS AN OPEN QUESTION. DO NOT READ ANSWERS. CIRCLE ANSWERS GIVEN. WRITE IN 'OTBER' ANSWERS) (WRITE IN NAME OF CHARACTER FR014 Q.8) 45/ 0,10. 0,11, Ad in magazine ............... 1 46/ 1 48/ Ad in newspaper ............... 2 2 Ad on television .............3 3 Ad in a store ................ 4 4 Billboard ad .................5 5 On the product itself........ 6 6 T-shirt ......................7 7 From someone in family.......8 8 From friends .................9 9 Other (specify) 47F 49/ Don't know ...................Y Y 11. Now, thinking now about all the kinds of products that you want or that you buy, where do you usually seo or hear about them? PROBE: Anything else? (RECORD ABOVE) (ASK AS AN OPEN QUESTION. DO NOT READ RESPONSES. CIRCLE ANSWERS GIVEN. WRITE IN 'OTHER' ANSWERS.)
Page 133: ozh90d00
3. I•m going to read you some kinds of products. For each one please tell me 2 whether or not you have seen or heard any ads for that kind of product in the last few months? Please don't guess, if you don't know or aren't sure, just say so. (READ EACH ITFM. ROTATE LIST TOp TO BOTPO!!, BOTfON TO TDP.) 0.3. YPS. XAVB HAVE MYP SF.FIJ OR SSF7V OR Dt7J•r HF.1RD XE1RD .Qa7N () a. Cookies ......... 1 -i, b. ATHLETIC SHOES..1 c. Cars............ 1 -i d..BEER............ 1 e. Cereal.......... 1 f. Video games..... 1 -a g. CIGARETTES...... 1 -> h. JEANS........... 1 i. Fast foods...... 1 () j. Batteries....... 1 O.3& 9aeass saEV OR EaeRD OF 2 Y 25/ 2' Y 26/ 35-36/ 37-38/ 2 Y 27/ 2 Y 28/ 39-40/ 41-42/: 2 Y 29/ 2 Y 30/ 2 . Y 31/ 43-44/ 45-46/ 2 Y 32/ 47-48/ 49-50/ 2 Y 33/ 2 Y 34/ 3a. (FOR EACH OF THE PRODUCTS IN Q.3 IN CAPITAL LETTERS AND NARKED WITH •-+' gf(Q CIRCLED •1• (HAVE SEEN OR HEARD ADVERTISED), ASK): What brands of (name o ~ prod{Hct!) have you seen or heard of? (RECORD ABOVE) : A ~ ~ A
Page 134: ozh90d00
7. (FOR EACH CHARACTER CIRCLED '1' ON EITHER Q.5 QH Q•6. STOW THE APPROPRIATE 4 RxRIarr AND ASK): As far as you know, what kind of product is advartised by (n me ef character)? (ASK AS AN OPEN QUESTION. DO 110T READ RESPONSES. EITHER CIRCLE THE CORRECT ANSWER SHOWN. OR WRITE IN •OTHER' ANSIdEtt. OR CIRCLE 'DON'T KNOW) NOTE: BLSIE THE COW AND THE CZRCLE LETTER ZF CFARAGTER ASrED AN7[A' TRIX RABBIT ARE NOT INCLUDED IN THIS QUESTION. a. The narqirar , sunny Batteries ....................................... 1 ; bther (specify) 2 b. Little 'CSasar P1zza ..........................................1 Other (specify) 2 c. Joe Camel Cigarettes ........ .............................. 1 Other (specify) 2 d. Adan the Android 1 . e. Tony the Tiger Cereal; Frosted Flakes .........................1 othar (specify) f. The Jolly Green Giant Canned/frozen vegetables .......................1 oth.r (specify) 2 ~ q. The Reebler Elvss Cookies .......................:................1 other (specify) 2 h. Ronald 1[eDonald Fast food ......................................1 Don't knov.T 26/ Don't knov.T 27/ Don't knov.Y 28/ Don't knov.T 29/ Don't knov.Y 30/ Don't know.Y 31/ Don't knov.T 32/ Don•t knov.Y 33// other (specify) 2
Page 135: ozh90d00
5 IRTERVIEWER: YOU ARE TO ASK SEVERAL QUESTIONS (QS. 8 AND 9) AHOUT TiVO OF AN~++ THE CHARACTERS FOR w 7CH THE R PONDFNP AL CORRECT ZNO-7. IF .mg _rAw_ct_. IS ONE OF THE CHARACTERS FOR WHICH A CORRECT ANSWER WAS GIVEDi IN Q.7, ASK ABOUT IT IN 0.8. THEN ASK ABOUT ANY OTHER CHARACTER IN Q.9 FOR WHICH A CORRECT ANSWER WAS GIVFN IN Q.7. IF JOE CAMEL IS b= ONE OF THE CHARACTERS ABOUT WHICH A CORRECT ANSWER'WAS GIVF11 IN Q.7. ASK ABOUT ANY TiiO OTHER CHARACTERS IN QS. 8 AND 9. DO h= ALWAYS ASK ABOUT THE SAlB3 CHARACTER(S). BE SURE TO GET A GOOD MIX OF CHARACTERS (OTEER THAN JOE CAMEL) OVER THE INTERI/IEPiS YOU COMPLETE. Now I want to ask you a little more about just two of the characters that you know about. 8. (WRITE IN NAME OF CHARACTER) : 34/ 8a. How would you describe (name of cl:aracter)? What's the character like? 35- 36- 37- ASK Q. 8b ONLY WHEN JOE CAMEL IS THE CHARACTER ASkED ABOUT IN Q.8a. 8b. What do you, personally, think about cigarettes? PROBE: Aything else? 38- 39- 40-
Page 136: ozh90d00
18 :_:0fi6. • • 1. _: G_~ a- (A sampling of respondents who identifled each of the other trade dtamcters witlt the oorrect product category were asked.) 7iow wnuld you descnbe (name of charecter)? What's the aharaderGke?' (Qs-8a,9a) Youth who Identlty character wlth c4lreCLRlogliGt The Eneralzer Bunny A bunny that keeps mnnirgrfloing & goingf lots of energy 52% He's pink/pinh 6unnyPourrry 34 Beats: plays a drum 33 Cute 13 Funny 11 (Number of inten+iews) (286) LlttleSaesa[ Shows htm wift. piua/seAtrg pizza 52% Little maMOHb man wih spear 35 Funny 21 Roman sodedandent Roman 17 (Number of interviews) (204) Tony the Tiaer He's a tigeda big; ioaring; gentle tigedtiger with stripes/roars'Ihey're great' 41% Shows him eating cereaVhe fices cereal 21 Athletlc/atrong/healthhy 17 Happy/havfng fuMenthusiastic 13 If you eat the Procfuct, you7 get strongrirnpove in spcrts 13 Trying to get kids to buy, eat pmdudJproduqt good for you 12 Funny 10 Friendly/nice 10 (Number of Interviews) (245) (Note: Only rasponsos g'rvon by 10% or more are shown.) Ln N 0 ((•i011t/IUBd) OJ Ql ~ N N
Page 137: ozh90d00
23 WAYS SEEN OR HEARD ABOUT OTHER TRADE CHARACTERS (A sampling of respondents who identified each of the other trade characters with the correct product category were asked )`Think again about ,(name of dtaracter). Please teq me the ways that you might have seen or heard about this character. Anythfng else?' (Q.10) 1: The Enerpaer Bunny 2. Latle Ceasar 3. Tony the Tiger 4. The Jolly Green Giant 5. The Keebler Elves 6. Ronald McDonald Youth who identify character with correet Prod+ct 1. 2 a 4 A S x x x x x x Ad on television 96 88 100 91 90 96 Ad in ma9azhe 22 18 23 11 18 22 Ad h store 22 27 43 32 13 30 On the prodict Rsee 19 24 49 20 15 13 BilIboard ad 12 6 9 11 6 13 Radfo 10 4 - - 6 8 From iriends 9 8' 11 3 - 5 Ad in newspeper 7 24 12 23 5 25 From some0ne in tamBy .5 _ - 7 • - 4 T-shtrt - 3 6 - - 7 (Numberofinterv(ews) (64) - (37) (29) (19) (21) (36)
Page 138: ozh90d00
12 AWARENESS (UNAIDED) OF PRODUCTS ADVERTISED BY SELECTED TRADE CHARACTERS (R85pond8lltS aware of diafad@l WBl@ SI)ON41 Cflafa[xBlBlld aSk@d): 'As far as you know, what Idnd ofpproduurx Is advefEised by 4BfIISt.QI 1~~18~tB1~?W . (Q.7) Youth aware of eMuaeter Comeet Pr'oduct J>omid Nun1dW (Of 111tarylewift Lft Ceasar (piaa) 99% (1081) The Eneipizer fiumy (naueries) 99% (1115) Tony the TiW (cereal; frosted flakes) 98% (1107) The Keebler Eties (cooides) 96% (1088) JOE CAMEL (CIGARETTES) 95% (972) Ronald McDonald (Fast food) (1114) The Jolly Green Glant (ramedrtmzen vefletaaes) 92% (1W5) Adam the Android 45'X.' ( 88) • Bogus dwadar, no oov.d anmw.r; 45% att.mpl.d a r.spons..
Page 139: ozh90d00
So as to have a sufficient number of black youth for a separate analysis, there was an oversample of them. These additional interviews were obtained in two ways. First, approximately 30 clusters were identified in areas with significant numbers of blacks, and additional black interviews were assigned there. Second, further suppiementary Interviews were obtained with blacks in and around four cities: New York, Atlanta. De- troit, and Los Angeles. Because the suppiementary black intenriews were obtained essentially by quota, rather than probability methods; their distribution was controlled by geographic region and type of place. The regions were Northeast, Midwest, South, and West and the types of place were: central city of a metropoGtan area, other cities and suburbs of a metropolitan area, and non-metropoGtan areas. In order to randomize the selection of the particular young person to be interviewed (age 10-17) in a selected household, the most necent birthday approach was used. The interviewer asked to speak with the person in the household who is age 10 to 17 and has had the most recent birthday. If that person was not available for Interview at the time of the initial contact, the interviewer ascertained the best time to call back to speak with him/her. Geographic areas reported in this survey conform to but combine U.S. Census regions. The Northeast Is New England and the Middle Atlantic states. The Midwest is the East North Central and West North Central states. South is South Atiantic, East South Cen- trai, and West South Central states. West is the Mountain and the Pacific states.
Page 140: ozh90d00
4 @ac round The use of the Joe Camel trade character by the R.J. Reynolds Tobacco Company in its advertising and promotion campaign for Camel cigarettes has received considerable criticism from a wide vadpty of groups and much negative media coverage. Essentialiy, the basis for the criticism Is the allegation that the company, by using the Joe Camel trade character, Is aiming its campaign at minors and the contention that the campaign has had a substantial influence on children's smoking behavior. The criticism of tobacco advertising and promotion reached new heights following the publication of three research studies and commentary in the December 11,1991 Issue of the Journal of the American Medical AssorjaNon. The AMA claimed, on the basis of the reported research, that'...tobacco advertising Is reaching children and teens, and should be banned from all media.. ' There bave been several reviews of the research studies, concluding that they are flawed methodologically and misrepresent the resuts - thus serving as a very dubious basis for proposing either a ban on the trade diaracter, Joe Camel, or on cigarette advertising generally. ObWahre of_the Research • Essentiaiiy, the objective of this research study was to obtain information from young people (age 10 to 17 years) regarding (1) the level of awareness of the Joe Camel trade character, (2) the mimage' portrayed by Joe Camel, and (3) the sources of awareness of Joe Camel. it should be noted that the study made no attempt to measure the extent of underage smoking orto establish a linkage between trade character recognition and product usage. Sample and Interviewing Method The survey was conducted among a projectable national sample of young persons, age 10 to 17 years, including an oversample of black youth. (See Appendix for detaiis of the sampling methodology.) in totai,1,117 interviews were conducted, 382 among black youth. The survey was conducted by way of personal in-home interviews by members of the Roper Starch field staff. All supervisors and interviewers were provided with detailed written instructions. The interviewing period was November 1-18,1993.
Page 141: ozh90d00
10 yl *1" • / . _ FHI 'Some companies advertise their pn>dUcts with some ldnd of made-up character. !'m going to name several of these charectes. For each one, please tell me whether ornot you have seen orheard of IG P/ease dont guess. H you dont know or,aren't swe. Iyst say sa' (C) 5) (if respondent had not seen or heard of clarader or did not knowAvas not sure, the character was s6own and resparxent asked.) ?h/s is (pm= of cha.acter). Now that you see Zhe diarader, do you remember seeing • or Nearing aboyt it before, or havenw you ewser seen or heard about it? Again, pfease donYguess' (a6) Total youth Have sten or heard of charaeter Unaided Alded Totai The Energizer Burny 99% • 1 100% Ronald AteConald 99% 1 100% Tony the Tiger 96% 3 99% The Trbc Rabbk 95% 4 99% The Keebler Elves 93% 5 98% The Jolly Green Giant 91% 7 98% Little Ceasar 94% 3 97% JOE CAMEL 73% 13 86% FJsie the Cow . 27% 28 55% Adam the Android' 4% 3 7% (Numbaof iMervpws: 1117) ' Bogus chancter
Page 142: ozh90d00
19 Youth who IdeMly Character wlth corFeapmduct The Jolly Green Giant A big gianNgreen gtard/gerdle gisrdlsaYs 'Ho-ho-ho' ' 75% LAceshaises vegetables 25 Atldetir/slrongAreakhy 18 Trying to get kfds to buy producNeat the producV pmdua good tor you 11 (Number of Interviews) (195) 1 ~, . ~,~,.1.-, 77' BmaN merYelves/6Ule people bakfnp cooides/ baking in trees 70% HappyRraving tuNenBasfastic • 14 Cute 14 Funny 12 (Number of interviews) (181) H4na1s! McDonald A dowNdowrrlype eharacteristks; e.g., big red nose, big shoes, etc. 65% Happy/having tuNerrtlusbtsic 26 Helps kids/ttces cMldreNgives Idds support 15 Funny 15 (Number of Uderviews) (209) (Not.: Onty rasponsss given by 10% or more aom shown.)
Page 143: ozh90d00
THE OUESTIONNAIRE Following are instructions for each question. Please read carefully and be thoroughly familiar with the questionnaire before you start interviewing. These are some specific instructions that must be followed carefully: Q_.1 Circle age of respondent. 0.2 Enter name or kind of last product respondent recalls seeing advertising for. Q_.3 Read each product (10 of them) and circle whether or not respondent has seen or heard any advertising for each one. NOTE: Question asks respondent not to guess. Also, rotate list top to bottom, bottom to top on alternate interviews. Q3a This is asked only for the 4 products in capital letters with an arrow in front and oelv . if a"1" was circled for the product in Q3. Enter brand name verbatim. 0.4 Read each brand name (9 of them) and enter the slogan verbatim. NOTE: Question asks respondent not to gttess. Also, rotate list top to bottom, bottom to top on alternate interviews. 0.5 Read name of each character (10 of them) and circle whether or not respondent has seen or heard of it. NOTE: Question asks respondent not to guess. Also, rotate list top to bottom, bottom to top on alternate interviews. 0.6 This is asked only for those characters circled "2" or "Y" in Q.5. Show the appropriate exhibit card and ask if respondent remembers seeing or hearing ; about the character before. Again, question asks respondent not to guess. 0.7 _ This question deals with 8 of the 10 characters - Elsie the Cow and The Trix Rabbit are not included in this question. For each of the 8 characters that were circled "1" on either Q.5 or Q.6, ask what kind of product is advertised by the character. Show the aoyrooriate exhibit when askin4 about the character. Either circle the correct answer - which is shown for all characters except Adam the Android - or write in answers given verbatim.
Page 144: ozh90d00
4. Now I'm going to name some brands of products. For each brand I read, please3 tell me the slogan or phrase that is used to advertise that brand. If you are not sure or don't know, just say so. (READ EACH ITFM. ROTATE LIST TOP To BO'1'TOM. HOITOM TO TOP. WRITE IN SLOGAN/PHRASE GIYP21, OR CIRCLE 'DON'T KNOW.) DaN'r MOW () a. Nike athletic shoes. Y 51-52/ b. Marlboro cigarettes. Y 53-54/ c. Kellogg's Frosted Flakes .............. Y 55-56/ d. Diet Pepsi.....::.... Y 57-58/ e. Camel cigarettes.... Y 59-60/ f..Miller Lite beer.... Y 61-62/ g. Newport cigarettes.. Y 63-641 i. McDonald's.......... Y 65-66/ ( ) j- Trix ................ Y 67-68/ 5. Soew companies advertise their products with some kind of mde-up character. I'm going to name several of these characters. For each one, please tell me whether or not you have seen or heard of it. Please don't guess, if you don't know or aren't sure, just say so. (READ EACH ITEM. ROTATE LISf TOP TO BOTTOM, BOTTOM TO TOP.) 69-79 80-1 Dupe 1-5 0.5. YE4. IIBVS DdJ'r HAVE AW IQA7/!/ SEFN QR SSffiY OR Dpr NP.1ND HFJIRD SORE 6. () a. The Energizer Bunny ....1 2 Y 6/ b. Elsie the Cow ..........1 , 2 Y 7/ c. Little Ceasar .......... 1 2 Y 8/ d. Joe Camel .............. 1 2 Y 9/ e. Adam the Android ....... 1 2 Y 10/ f. Tony the Tiger .........1 2 Y 11/ g. The Jolly Green Giant ..1 2 Y 12/ h. The Keebler Elves ...... 1 2 Y 13/ i. Ronald McDonald ........ 1 2 Y 14/ () j. The Trix Rabbit ........1 2 Y 15/ Q.6. YPS. HAVE DQN'T AAYB AVr AYOR/ SEIIP OR SSDI QR NDr IISAND RBARD SDRB 16/ 17/ 18/ 19/ 20/ 21/ 22/ 23/ 24/ 25/ m N ~ CD I FOR (FOR EACH CHARACTER CIRCLED '2' Q$ 'Y' IN Q.5. sROUr'rRE APPROPRIATE EXHIBIT F,_CHCTER AND ASK): This is (name of .h ra ). Now that you see the character, do you remember seeing or hearing about it before, or haven't you ever seen or heard about it? Again, please don't guess. (RECORD A80VE) O1 A ~ ~ ~
Page 145: ozh90d00
: i' 41 . ,iA 1.@D.a_O. A nationwide cross section of 1,117 young persons, ages 10-17 (Inclucbng an over- sample of black youth), was interviewed for this survey in face-to-face interviews in respondents' homes. , 3am ing Method The persons interviewed in this. study comprise a representative sample of the popula- tion of the Continental United States, ages 10-17, exclusive of instituiiona6zed seg- ments of the population. A multi-stage, stratified probabiGty sample of interviewing locations was employed in this research. The probabilities of selection at each stage are based on the latest U.S. Census population data, and detailed Censos maps are used to identify and locate the selected areas. At the first stage, 100 counties are selected with probabilities proportionate to popula- tion, after aN the counties In the 48 contiguous states and the District of Columbia are ordered by population size within 18 strata. The sfrata are constructed by classifying counties as metropositan and non-metropolitan wfthin each of the 9 Census Geographic Divisions. At the second stage, within each primary sampling unit, two Census block groups (or Census Enumeration Districts, when Census block statistics are unavailable) are se- lected with probabilities proportionate to'popuiatlon from a computer listing in which the block groups (ED's) are stratified by size of place in which located. At the third stage, within each sample block group (ED), two sample locations (blocks or rural equivalents of blocks) are selected. When Census block data are available, the sample blocks are selected with probability proportionate to size (population) from a cumulative computer listing. When no such block data are availabie, sample Enumera tion Districts are broken into identi6abie segments (smaii areas defined by roads, streams, railroad tracks, or other unambiguous boundaries), and the sample segments are selected with equal probability. At the block (segment) level, the interviewer Is assigned a starting point and a path to proceed around the sample area in the selection of households.
Page 146: ozh90d00
3 Q$ Be very careful with these questions. They are to be asked about 2 characters for and which the correct answer was given in 0.7. 0.9 If Joe Camel is one of the characters for which a correct answer was given in Q.7, always ask about Joe Camel in Q.8a. Then ask about any other character in Q.9 for which a correct answer was given in Q.7. If Joe Camel is not one of the characters about which a correct answer was given in Q.7, ask about any two other characters in Q.Ba and Q.9 for which a comect answer was given in Q.7. Except for Joe Camel, get a good mix of characters in these two questions. Do not always ask about the same character(s). Be sure to write in name of character asked about on the line provided for both Q.8 and Q.9. NOTE: O.8b is only asked when Joe Camel is the character asked about in Q.8a. On Q's. 8a, 8b, and 9, record responses verbatim. 0.10 This auestion is asked about the characters from 0.8. Enter name of character on the line provided. This is an open question; do not read answers. Circle a pre-coded answer and/or write in 'other" answers verbatim. Multiple answers are aeceptable. 0.11 Similar to Q.10, but asks about advertising in general. Also an open question; do not read responses. Circle a pre-coded answerr and/or write in "other' answers verbatim. Multiple answers are acceptable. 0.12 Record sex of respondent. 0.13 Record race of respondent. 0.14 Record geographic region (1-9) from sample. At the end of the questionnaire, fill in the certification on page 8. It must be the interviewer who signs. Go to the last page - page 9- and record the respondents name, address, and telephone number. Interviewing is to begin on Tuesday, November 2 and end on Tuesday. November 16. All completed interviews and all other materials ate to be returned to Lois Massel, Roper Starch, 566 East Boston Post Road, Mamaroneck, NY 10543. Ship by Federal Express Standard so they arrive at Roper Starch by Wednesday, November 17. If you have any questions or problems, check with your supervisor or call Lois Massel at (914) 698-0800.
Page 147: ozh90d00
TABLE 2 Analysis of Attitudes/Affect Rates of Liking Produets Number of Respondents Total 790 Three-yearsold 143 Four-years-old 219 Fve-years-old 224 Six-years-old 204 Ch7dren and Muh Products Attitude toward . Mkkey/Nouse, .871 .706 .840 .933 .951 Atbtude toward cereaR .873 .769 .817 .915 .961 At4hde toward hamburgersx .808 .734 .758 .853 .863 Adults-Only Products Attihrda toward cigaretlesl .152 .406 205 .045 .034 Attitude toward matehasl .175 392 .233 .085 .059 ISignieerR qu.d afe rYdioreWp of alfeU b aW 1feWy t,OMrac Y Iwcept (A -.17. P- b8j, Age (tt, -11.05. P- A001, A90 (kTi -7AZ P-AM end residual (kt, -.12. P-JM Clyrctl.c MiWapt (A -54.54. p-.0001), Age (r', -37.70. P-.0001L Ag@2(1e, -~.P-AO/janQ nptAr.l (~~ -1Zt1. P-27} Matdrs: InNiept (A -2931. P- A001). Age (A -15.l0. p..0001), Apsi (A -10.11. p-A02) end rssidusl (12, -1.t10• p-.92). 2Sipekwt in.:r 1Ns6orotlp a NfsGt b W . l.uefk YMwapt (A -111.16. P-.110C1), Age (e, -41?3. P-.0001) and msldud (6 .121. p-.35). lWraNSp.r: Nnwapt (ts, -80.67. p -A001), Age (A -13A6. p -.0003) end /esi" (A -197w p -37). trade character use and its high necognidon; and high roeog- nition of a ttade character could intensify the neativo affect. WIKO viewed acruss all ages, liking eigaeettes and ne- ognizitg Joe Camel was significantly and negatively tslsfed Wt = 24.1. P<.001).9he incteasing raeognitioa of the Joe Camel trade characterteoded to 6eassoeiated.rithdishlring, aot 1dCiag. dganettes. Howerer, eausality is nat suggested nor could it be assessed adequately with this aoss-sxtiooal data. Rr.aognition of the Marlboro Cowboy was not auoci- ated with liking dgareues. One last eontrast should be of interest because of the re- ports (e.g., Belch and Belch 1993; Wd11992) of Faaher and eolleagues' (1991) findings. Comparing the recognition of Joe Camel and the 6king of cigarettes atross each age group illuminales the strong negative relationship between the two (Figute 2). On the other hand, if recognition of the Disney (bsnnel and liking of Mickey Mouse ae plomed, a strong positive assoaatioa is evident. In this lanerease, a favorable context of affect formation for both the trade character (Dis- ney Channel) and the product (Mickey Mouse) would be expected. Discussion and Conclusions This study was developed to investigate the ability of young children to match artoon-based tr.de characters to the product with which they are associsted. Ile ediWmn's atti- tude toward those products was also tneasured. Previous ro- seach (e.g.. Wtana a as. 1991; F'isc6er et al. 1991) pro- posed that the ability of young ehiWven to recognize an as- sociation between a trade elmacter and the product was a ptedietor of developing favorable attitndes toward t6e ptod- u~ct. which would be expected to influence future 6eMavior. FIGURE 2 FeneetK Recognition and Like Product Joe cArnevc4mades LOP 4a'Reoopiwlion flace Disnay CxwiweYM~wkey Mouse 90 80 70 so 40 3~ 20 10 3 V~Reongrtion 4 s Age o/ Ctuld (M years) 6
Page 148: ozh90d00
A variant of the "smiling face scale" that provides a measure of affect and preference (Wells 1965) was used. Pretests showed that children three to six years of age had difficulty with scales that had choices intervening between a simple happy and sad face (e.g., neutral face). Separate vi- sual scales with happy and frowning faces were used for boys and girls (scc Figure Ic). Sample As was noted previously, children two to seven years of age have been viewed as most at risk to the influence of adver- tising (Raju and Lonial 1990). To provide a suitable group on which to replicate and extend Fischer and colleagues' (1991) study, a final sample of three- to six-year-old subjects was used. The regional marketing activities of the brands repre- sented by trade characters could exert an influence on the children's recognition of those trade characters and their af- fect toward the products. However, only information con- cerning Camel cigarette marketing activities was available. Nonethelcas, the potential association of the Camel market- ing offort with these dependent meatutss may be the most salient because of the extensive coverage of tfieso portions of Fischer and colleagues' (1991) findings (o.g., see Will 1992). Ten A.G Nielsen Company (1991) county markets were sampled. These markets were chosen to provide a divax sample in terms of geographic area, size, Camel's index of brand development (BDI) (i.e., Camel dollar share in the tnarket/Camel dollar share in thc United States), and Camel spending-per-person for advertising (outdoor and print media). Randomly generated telephone probability samples were developed for each market Recruiting was done from two central location WATS facilities. All households and re- spondents had to understand the English language to partic- ipate. Quotas for age (200 cach), sex (50% male, 50% fe- male), and race (based on area racial composition) were es- ublished. A target of 800 completed interviews of children three to six years of age was set A final sample of 790 ehil- dren was obtained from the 87,902 households eonlacted.t Both central location and in-home interviewing were pretested. These tests indicated that conducting the inter- view in a central location was a better approach because it provided a consistent configuration and control of the data collection environment. Presence of a Parent or Adult Caretaker Previous work with young children (e.g., Baxter 1992), as well as pretests for this study, revealed the importance of having a parent or other familiar adult present when collect- ing this type of data. This is particularly true for children three to four years of age, who were not attending a preschool or other out-of-home care (42.7% and 32%. re- spcctively). It was important, however, not to have the parent in such close proximity to the respondent that he or she interfered tA complete discussion of the mctho.:nlogy is availahle from the with and potentially biased the interview. Placing the adult in a chair to the rear of the respondent appcaied to provide adequate comfort for the child and generally precluded parental observation of the recognition task and affect judg- ment The distance of the accompanying adult from the re- spondent ranged from 14 inches (3.7% of the sample) to five feet (542% of the sample. 77% were 46 inches or more). This distance was not associated with trade character recog- nition or expressed affect toward the products. Interviewers A final group of 20 interviewers was chosen from the initial pool of 30 on the basis of their performance in pretesting in- terviews with at least two children three to siz years of age. At no time were the interviewers infomted about the true purpose of the study or the identity of the funding source. Experimentaf Procedures When the adult and child arrived for the interview, they were seated in a waiting room. The interviewer then led them into the interview room, and the parent was scated behind and out of the child's sight- The interviewer and respondent sat on the floor on identical mats with a two-foot area between them in which the product game board and likefdislike pic- tttre board were placed when needed. All interviewers were thoroughly briefed about and had a list of reasons (e.g., un- willingness of the child to eoopcratc, parental interfuencc) to terminate the interview. Terminations were highest for the throe-and four-year-old duldren. A total of 24% of the 1040 central location interviews were terminated. The roeognition task and concept of a no match C.a.. pic- turc of nothing) option were introduced in a warm-up exu- cise- Even if the child failed on thrce trials to make a correct match of the Mr. Kool-Aid trade character to a glass of bev- cnge in the warm-up, the interviewer continued with the in- terview. There wete no significant effects of deleting these few subjects in any analyses. The interviewer tttcrt put away the warm-up product board, and brought out the six-option product picture game board (see Figure la). The interviewer named the six options and then asked the respondent to point to each product pic- ture when named. This provided a basic mcasurn of com- prehension for each product and blank option that permits checking the reliability of each age group's responses. Each respondent was then handed one picture of a trade character at a time according to a randomization sequence provided, and asked: "Point to the picture on the board that this goes with. And remember, if it doesn't go with any of the ones on the board, you point to the picture of nothing" The picture of nothing option was mentioned only in the warm-up and with the first card. The interviewer i only re- sponse to each match, whether correct or incotrecl, was "okay" After all the trade character pictures were given to the respondent and the matching task was eompletcd, the af- fect/attitude task was introduccd. The respondent was asked what food they liked. The interviewer then asked: "If I showed you a picture of (food the respondent said they liked), would you point to the picture that meant that you 52086 4553
Page 149: ozh90d00
households: ztt = 249, p=.11) or the respondent's house- hold subseription to newspapers or magar3nes (X23 =5.0g, p = .17). Camel sales relative to all brands (BDI) did not sig- nificantly predict (Ft.m =2.35. p=-13) Joe Camel tecog- nition rates-nor did the relative level of Camel advertising expenditures in outdoor and ptint media (52% high versus 57% low; Ft,m = .95, p = 33). Allect/Attitude Toward Products Table 2 shows the respondents' rates of liking for each of the five product pictures. The difference between the adults- only products (a cigarette and the lit match) and the other products is striidng for several nmsoat. Fus1. both the eiga- rette and the lit match had significantly lower rates of 6king than rhe other products studied Across the total sample, 85% of the children did not like cigarettes, which was slight- ly more negative thon the affect toward matches (82S% dis- like). Seeond, except for the lit match, the children at each age ditliked cigarettes significantly mote than all the other products rated (e.g., compared to hambugcts, )0t = 945.41, p=.0001). Thitd, none of the demogtaptdc, environmental, or marketing factors studied had a signifieaat effect on lik ing cigateGes. Age and affectThe liking for eigarenes and the lit match deersases with increased age of the nespadent, whereas the liking for ce¢al, hamburgers. and Mickey Mouse inereases with increased age. The rate of liking for eigaiettes dmps significantly for children (X2j = 119.29, p = .0001) ftom tfuee years of age (40.6%) to four (20S%), five (4.5%), and six years of age (3.4%). The five- and six-year-old rcspon- dents do not differ (7C2t =.99, P= 32) from one anothec The relationship of liking to respondents' age was also tested with a weighted least squares analysis (Table 2). For both cigarettes and the lit match, liking had a negative asso- ciation with age. All of the other products had sttong posi- tive associations of liking with age. Recognizfng Joe Camel and Liking Cigaretfes In both Fischer and eolleagucs' (1991) study and my inves- tigation. the adult-only product trade character Joe Camel had relatively high recognition compared to several other trade eharxters. Recognition was particularly high among the older children. If high trade chuacter recognition alone could prompt positive affect toward a ptoduct, this trade eharacter would appear to provide a good test. However, be- nuse of the extensive and loog-lerm impkmeatatioa of an- tirnoking tampaigns, it is possible that fioquent exposure to cigarettes in this negative eonte:t could prompt negatire af- fed toward that ptoduct This could happen with or without TABLE 1 Analysis of Recognition Rates Rates of Recognition Number of Respondents Total 790 ilreeyear"d 143 Four-year"d 219 Five-yeasold 224 Six-years-old 204 Children .nd Adutt Products Uisnsy Charnel with t 6liokeyt .858 .692 .799 .929 .961 Captain Ctunch wifh cereall .719 .448 .662 .825 .643 Tony the Tiger with cersall .596 .378 .534 .688 .716 Ronald McDonald with hambutgert .510 259 .443 .634 .623 Charlie Tuna with blanl(e S16 .469 .457 .531 .598 Adults-Only Products Joe Camel with cigaratte2 .523 .252 .411 .629 .716 with Gt match .084 .084 .119 .081 .049 Marlboro Cowboy with cigarette .235 .245 274 237 .186 with It match .135 .007 .164 .152 .059 'Sipr,Yor,t quadra0c rNa6anhfp d recognidon to age: Divtay Ctlsrnst hdupq (A =.09. p. T7), llye M . u.94, p..000),11ye2 (e, .5.13, p~.02) ard residral (1a1 =134, p- 25). Captain Cnndt kMsRapt (X2, .9.41. p -.002), Age (et .2026. p ..0001), Agss (rt, .1338, p - .0003) and tasidual (*I, =0.88. P - Tony ta lipsc Mdsioept Ws - 2.80. p..08). Age (1ai =7.21. P..007). A9sa Ctal =391, p- .06) and msidual (tt1 =.69. p-A1). RW 8 d McOetqld: kMl/Oepf (i=1 - 9.60. p= Ap2). Age (t21 .1399. P=•OQQQ). Age2 (s2, =Bb1, p..003) a1W f6sidsl (o, pa sSigni6orq insar rMatlonstip of rewgaon b ape: qrerfe Tuna: Ints,eept (XI, .13.18. p=.0003). Age (x1, =8.81. p=.003). and oasidual (A2, =131. p=.47). Joe CanW: tntappl (e, .923. p - .t102). Age (y:t .11224, p=.00001), aM msidual (*2, .3.05. p=22). 66/Joumal of Marke6ng, October 1995 52086 4555
Page 150: ozh90d00
Ktumske, William (1993). "WAen Medical Doctors Conduct Mar- keting Reseatch IAMA and Old Joe, The CGmel," inAMA Wie- kr Edracatori Conference Proceedings: Marlreting Theory & Appfic,miav. Rajan Vandanjan, Bernard Jawoeski, Peter 1. Gadon, and Bert 1. Kolletman, eds. Chicago: Amaican Mar- keting Assoeiation, 405-10. lerin, Gary (1993), "Study: Some Logos Hurt Image; Adverris- ing Age. (September 13). 40. ' lipmar4 Joanne (1991), "Kools Penguin Draws Antismoking Ice," WaR Srreet Joumal, (October 8). Maetlin, M. Catolo (1983), -Do Young Children Understand the Selling Intent of Commtxcials7" The Journal ojCoemmer AJ- /airs,19 (2), 293-304. MeNeal, lamcs (1987). Children as Coasumers: Insight and Im- pGrarions. Leaington, KY: Lexington Books. Mittrslo, Ridnsd, Brenda Sontrr, and Katherine Straughn (1993), "A RoBvaluation of the Reported In/lurm of the Joe Camel Trade Charxtec on Cigarette Trial and Use by Minors; in Pro- ettdinji of the J993 American Academy ofArJvertitina. Esther Thaso4 ed Montreal. Cauuda: American Academy of Advct- tising. Moschis, George and R. Moore (1979). "Dmsioe-Making Among Ihe Young: A Socialization PetspeaivS Journaf of Consmner Reuarrh, 6 (September). 101-12. Naples, Midud (1979), BffectiMe Freqrutcy: flm Relationship Betw+een Frequency and Adverf'v-ug E,ffieaiNreess. New Yettc Associauon o( National Ad.crtisas, Inc. Normaq Naoey M. and JamtxT.Tedeschi (1989)."Self-Ptesetra- tio0. Reasoned Action, and Adolescans' Decisions to Smoke GgarWet,' Joumal of Applied Social Psycholop, 19 (1), 543-58. Oei, Tian P. S. and Anthony Burton (1990). "AtGtuda Toward Smoking in 7- to 9-Year-Old Qtitdro4' 7be International Jotorol oJthe Addictionr, 25 (1), 43-52- Ogilvy. David (1983), Daiby on Advertisin;. New Ycst V'vuage Books. Ynget. Je.n (1970). "f1.e Stages of 4+tellernul Devdopntcat of the m'14" in Raodings in Child Deve(opmant ond Psychology, P. H. Musse0. ad. New Yock: Harper and Row. Quarfan6, Joanne M. (1979). 'Ywldten's Understanding of the Nuure of Television Charactus,' Journal oJConantstianrionr, (Summa), 210-18. Raju. P. S. and Subhash C Lonial (1990), "Advenising to lail- dma: Fmdings and lmplintions; in Current Issues and Re- seateh inAdremring. Vol. 12, James l.eigh and Clade Manit4 eds. Ann AcEor, Ml: University of Michigan, 231-74. Ratdsan, Sal (1992).'7he Power of Mythology Helps Brands to Frdme," Mar4aing News, (September 2B),16. Reeves, B. and Bradley S. Gmenberg (1977),'C3ildrrn's Petoep- tions of Television puractea," Hwrmn Cornnnoticarion Re- search, 3, 113-27. Rolwtsan, Thomas and John Rossiter (1974), -Qildtrn and Com- men.ial Persuasion: An Attribution Theory Analysis,' Journal oJCowunerReiearch, 1 (June), 12-20. - -, and Terry Gleason (1979), "GTildrcn's Receptiv- ity to Proprietary Medicine Adve[tising," Journal of Consumer Research, 6 (Decembu), 247-25. Roedder. Debonh (1981). `Age Diffuenoa in QtiWtrn's Re- sponsez to Tekrision Advertising: An Informauon Processing Apptoacl4' Journd of Consumsr Research. 8 (September), 144-53. -, Brian Stemthal, and Bobby Calder (1983), "Amnde-Bo- luvior Consistency in t]iildten's Responses to Television Ad- •mising; Journal of Marketing Rrtearch, 20 (November), 337-49. Schindler, Rotiat, Mortis Holbtook and Eric Gteenleaf (1989)- "Using Connoissems to Predict Mass Tastes," Marketing Let- ters, 1(1), 47-34. Schorow, Stephanie (1992). 'Television Activists Take Atms .Against Cartoon Cu," Associated Press, (March 19). Stan, Barbara (1993), "1Tte Art of Deceptiore Poe6e License and Advatising Fiaiat,' in Praeediags oflhe Ameriom Academy ojAd.rrtising, Esther Thorsen, ed. Montreal, t.anada_ Ameci- an Acadomy of Advertising. Stewut, David and Ronald Rice (1994), "Noo-Ttaditional Media and Promotions in the Marketing of Alcoholic Bevaages," rrarking paper. School of Business, University of Southern Cal- ifomia. Sturts, Mary Am and 0. Hunnian (1987), "Can Young aildron Undastatd Disdaimris in Television Commaeialsr JonrrwJ afAdrardsal, 16 (1).4 1-6. Tessa, Abraham (1978), `Self-Generated Attitude CJunge," in AdMmtus ie F.vrri watrof Social Psydrology, L Bakowitz4 od. New York: Academic Prrss, 289-338. - and Mary Conlee (1975). "Some Hfkcts of Time and Thought on Attitude Polniaation," Jmanal of Pnsonofi(ry and Saeiaf Psyrhofogr, 31 (Febnury), 262 70." - 7hason, Psthc (1990), `~Comumer Proorssing of Advuusing," in Ciurnmt Issues aed Research in AdvrrBti+d. VoL 12, James Leigh and Clatde Maroin, eds Ann Arbor, Ml.- Utd'vasity of MikAigau,197-230. Van Aolcea, Stuart and Subbash Lonial (1985), "Otildtat's Pa- ap6oas of C6racsas: Huoaa Vessus Animate Assessing Im- plicationt for Qn-Wtm's Advatising; Journal af Adverrisiet. 14 (2),13-22. Ward, Sccts (1972), ^Ohildtat's Reactions to Caaunacials,' Jour- nd af Adverrisiea Rcsoorch, 12 (Apnl), 37-45. - (1978), `ResearcAers Look at the 'Kid-Vid' Rule: Orerview of Sasion; Adwnev in CoRSUetrRestmch, 6(Oc- tobcr), 7-11. -, D. Warirnun, and E Wastella (1977), How Childree Leam to Bu). Beverly Hills, G: Sage Publica8ont, l= Wdls, Wipiam D. (1965),'Cooawntating with CTildnuu4" Jour- +uf aJAdrertidng Research, S (2), 2t14. Will, Gearge (1992), `Gn't Be Too Tough on Tobacoo: Wash- ington Post, (January 19). Wiison, David (1990),'Tooned On: America Flips for Grtoont," TV Gride, (June 9), 22-27. 7sjonq RaEen (1968), "Anitudinal Effects of Mae Eaposue; Journal of Personality aed Sociaf Psychofogy Monogrophs. 9 (2).1 27. 70 l Journal of Marketing, October 1995
Page 151: ozh90d00
Richard Mizerski The Relationship Between Cartoon Trade Character Recognition and Attitude Toward Product Category in Young Children The author reviews and tests the effect of cartoon trade characters on product reoognition and attitude on a sam- ple of children three to six years of age. High levels of product and trade character recognition were found, indud- ing that of Joe Camel and the Marlboro Man with cigarettes. The recognition of select trade characters tended to increase with the age of the child. The leve( of reoognition and tawtab{e attitude tovratd the ptodurt were positive- ly associated with age eocept for cigarettes. The attitude for cigarettes and matches were negatively associated with age. Because the possibility of demand artifads cannot be rtded out, these fatdings must be Interpreted with caution. ~~ cffect of sdvatising for adulasonly ptoducts on ' ehildren has had limited e:paiment-baud study (Gorn and 67oubeim 1985: Robatson, Rossiter. and Gleason 1979). Most work repocts on suney responses or anecdotal accounta of how children would respond. Nonettxlesa, many would agra with McNeal (1987. p. 85) who claims "ldvettisets have the ability to oonvinoe chi!- dren to like and dxile practically any product" 7his alleged ability is of particular concern when the products nny pose healih risks (e.g.. tobacco, alcohol, nonprescription drugs). and qucstions of unfaili ness and manipulation may be ap• pliaable to all advertising that targets children (Maeklin 1985: McNeal 1987). 7be+a may be an added threat to children with the tesur- gence in the use of cartoons for tekrision and film enter- tainment. as well as an increase in their ase as trade ehacao- tas (Ftzgaald 1990) for botb adult and youth markets. Nu- merons advertising critics (r=g., ITixher et a1.1991; GarSeld 1991: Ogitvy 1983; Wilson 1990) sxm to view this dement of the creative aesenal as paRicululy effective in forming yotmg peoples' prefereocea toward product ntegoria. Many (e.g.. F'scher et aL 1991) advocate a ban on cattoon trade character use in several categories to Eoep childrai from being affected by ttaem i8dord ifmNSld's Pmfsssar of fAaAae6ig, GdBh llrwesiy. Austnfa. Ths audiaMsdtt KaRiehro StragM. Jadc Fddman Hay Phr,Jds Wa4 1ea,Tlr Fiafdr+Stal. tlmnslpMWlmtirg Depabnmkand aoormaa g/ ~aNswas br tfieir asai~tance This retemi~ rru Uaded Dy he R.t. Nepndds Tdaooo Compsrry. Edofs tiolt<.lwfY Pbicaim d Ihis aside t>ased an a n•seadt qtdy IudW by ths RJ. RepnNs Totaooo f)on"y slnuld nsiher be o0o- atrued as ae adonx/nad d mnamptim of Iobaooo-buad pioduds na as aippat lor tlie bbaeoo iidsYy. , I investigate the degree to which children corcecily asso- ciatn select pmduet categories with fiequauly used trade charxtets. In addition, I explore the relationship bctween ptoduct trade chuacterassodation and attitude toward pmd- ucts. Both recognition (Piseha et at. 1991) and atGtude ta ward a product (McNeal 1987) bave bcea vrcwed as pmdia ton of future product usG'1Lerefore. I adopt both a market- ing and public policy petspxtivc on tbe issue of trade char- acter nxognition by young chiWten. Advertising and Children's Responses MeNeat's (1987) review concluded that advertising is capa- bk of produeing duen types of behavior amoag childmn: twlehasas, purchase requests, and antisociml behavior (e.g.. requests leading to pareutthiW eooflicu). He and others (e.g., Mosdut and Moore /979; VVad 1978) argue that the purchase may not be immediate, benuse the child may do- velop and store in his or her memory many of the orienu- tions and norms reflected in advectising until a purchase op- portunity occum Apparently, tbese eBectt can result fiem advertising aimed at adults, even if youths ale not spaeifi- ally targeted. Gorn and Hlotsheim (1985) discuss sevaal studies and provide expesimental evidence suggesting that children devolop strong sterootypes of owners of adult ptod- uets. Tbcy further tr,poct that advettisiag can influenee how children view aod obtain appeoptiate models for the sdult', .radd, including concepts of appropriate products to use' ~n now and in the future (ef. Belk, Mayer, and Driscoll 1984). ~ m However, exposure to advertising does not autatnatitnl- ami ly affect behavior or tonslate into changes in the way tlwj d. child mstes dodsions about produUs.'Rie influence of ad-I A vutising may be significantly affected by the child's age and i -j JowRfl of aWheting I
Page 152: ozh90d00
Frequent repetition of the advertisements could lead to fpvonble attitudes through the mere exposure effect (Gold- berg, Gom, and Gibson 1978). Therefore, as the child ages, his or her knowledge base and favorable attitude toward the products advertised should increase. However, there is some question as to whether trade ehancter/product recognition and a favorable attitude toward the product are always posi- tively associated with each other and age. The directly ex- pressed wishes of socialization agents, such as parents, has been shown to negate the attitudinal effects of repeated ad- vertisement exposures (Goldberg. Gorn, and Gibson 1978). Also, if the young audience does not view the advertised product as acceptable for later use, repeated advertising does not appear effective in generating favorable attitudes toward the product (Goldberg and Gom 1974; Robcrtson, Rossiter, and Gleason 1979). 7ttere are cases in which a negative relationship between trade character/product recognition and attitude toward the product may be expected. Repeatedly exposing the product in a negative context (Goldberg, Gorn, and Gibson 1978) or repeated exposurc to a stimulus initially viewed as negative (Schindler. Holbrook, and Greenleaf 1989) have been shown to prompt increased negative affect toward it over time-'1be positive association of age and recognition and the negative association of recognition and affect could be explalned by the process of inen: thought (Eagly and Chaiken 1993). 'ILetefote, thern does not appear to be only one telatioaship possible between age of the child, recognition of a trade ehvacter, and affeet toward the praduct Because of the equivocal oahuo of theaetieal petspcctives and ptior find-s ings addressing the relationship between age, exposure to advenising, and affect formation, rather than presenting a formal hypothesis relating to age, trade character recogni- tion, and attitude toward the product, it is treated as an em- pirical issue. Methodology Measuring Trade Character Recognition The product picture board fotmat has long bcen used (e.g., Kobasigawa 1977) to establish recognition in young chil- dren. Using this approach also provides researchers with the abiGty to compare the trade character/product recognition rrtes obtained in this study to the recognition rates for uadc- mark logos and select trade characters fust reported by Fis- cher and colleagues (1991). However, some modifications to this procedure were necesnry to reduce the previously dis- cussed potential biases and limitations of Fischer and col- leagues' stimulus and task. Trade channcters used. Because Ihe matching of trade characters to products was a primary interest, only product categories with a bnnd currently featuring a trade character were used. In addition, ttade characters that could be asso- ciatcd with many different products (e.g.. Teenage Mutant Ninja Turtles represent ceneals, toys, television programs and movies) were eliminated from the study. The final product set consisted of five product options ar.d one no-match blank (or picture of nothing) option (Fig- ure Ia). The product pictures of a cigarette hcW in a hand and a lit match represent adults-only products that young children normally would be taught to dislike and not use. A bowl of cereal, Mickey Mouse, and a hamburger were ptods uas pictued that would be viewed as acceptable for chil- dren and adult use. Children in the pretests found pictures of these five products to be the most recognizable of those ex- andned. Except for the lit match, these products also had un- ambiguous, currently used trade character matches. Unlike cigare.ttes, no brand of matches has a recognized trade ehar- acter, nor do firms typically advertise matches to endton- sumas. All trade characters were extensively used in brand advertising. Fischer and colleagues' (1991) study also used pictures of a hand held eigarette, a bowl of cereal, Mickey Mouse, and a hamburger. The arrangement of the five color product pictures on the game board and the order of presenting the color trade character pictures were randomized for each respondent. The bade characters associated with a brand in each mixed market product ntegory ane Disney Channel with Mickey Mouse, Captain Crunch and Tony the Tigu with ceeral, and Ronald McDonald with a hamburgu Both Joe Camel and the Marlboro Cowboy are trade characters.for the adults- only product of cigarettes (see Figure lb). A picture of the tatarley'ILna trade character was also provided to the childten. However, no picture of packaged tuna was included on the product board. The ehildten should have matched the picture of nothing/blank picture to Charfey 1Lna, beeause this picture option was provided for trade characters for wfiich'the children could not find an ap- propriate product rnatch `- Product aame boani Pretests revealed that many young subjects could reliably respond to no motc than six different product pictures. The size of the board on which the product pictures and blank were placed (see Fgure la) was deter- mined on the basis of numerous pretests and was very simi- lar in size to the product picture game board used by Fisch- a and colleagues (1991). All modifications made the stim- uli easier to view for the respondents. Affect/Attitude Judgment The use of an attitudc/atTect measure provided a means for investigating the potential behavioral implications of repeat- ed exposure to trade characters, where the sum of exposures should increase with age. Norman and Tcdcschi (1989) have reported a significant eotrelation (r =.5l) betwcon atfect and intention to smoke in the future for fifth- through eighth-gradc children, though there appears to be no similar findings for other product categories. There is also a long history of using affect (like/disGke) measures in iavestigat- ing the possible impact of advertising on children (e.g. Goldberg and Corn 1974; Gom and Florsheim 1985; Oei and Burton 1990; Wells 1965). Several other responses, such as intent to use, good or bad for you, and perceived risk and harm associated with product use, which may influence future behavior toward a product, were also tuted. These were ultimately discarded because many of the children had difficulty with them. 52086 4551
Page 153: ozh90d00
Howevor, there appeats to be little evidence that this hap- pens when the children view the products as inappropriate to use (Gom and Flotsheim 1985). This has particular impor- tance in the area of adults-only products and the possible in- fluence their promotion may have on children's behavior to- ward these products before they are adults. Limitations The select group of products, trade characters, and ages sampled limit the degree to which these findings can be gen- etalized. Few children in this study would be expected to have a direct smoking experience with cigarettes. Howeva, a sample of older children would provide some insight into how the memory of a trade character functions when direct product experience is encountered. Thc cross-sectional na- ture of this study further limits interpreting the effects of aging. A longitudinal study incorporating an older sample would be a better fomtat, because peer pressure and the longer-term ambient level of product acceptance could be addressed. Changes in affect toward products could also be tracked. To the extent a lit match may constitute a correct associ- aiion for the cigarette trade characters, the recognition of Joe Camel and the Marlboro Cowboy would be higher (up 16.1% and 57.4%, respoctively). Also, the eonunon'Seouse cats" visual in the Mickey Mousr/Disney Channel match is expected to have artificially increased the recognition of that trade eharactec Recognition of the Jae Camel character could then be considered much higher relative to the other trade characters studied. It can be argued that Miekey Mouse, aot the Disney Channel logo, is the more appropriate choice of a trade char- acter. However, this choice, in addition to the visual por- tnyal of these two stimuli, was made to make a comparison with Fischer and colleagues' (1991) original study. prctests for this study also showed that using the Disney Channel as a product prompted very different matches. Apparently, ehit- dten see many trade characters in advertising on the Disney Channel. A unique correct match would have been difficult to establish. The response capabilities of children frotn thtee to six years of age vary extensively. Using ptacedutes and tttea- sutes that a three-year-old child can master also limits the response capability of older children. Socially acceptable te- sponses ara a potential demand artifaa that could have in- flucnced the statad attitude toward the ptoduUs, particularly the adult products. The presence of an adult parent or care- taker could heighten this possibility. Henke (1994) found that 96% of her sample of 83 three- to eight-year-old chil- dren reported they disliked cigarettes. Using only an inter- viewer and no parent or careuker, she found no significant effect of age for that measure. Although there was a lack of a proximity effea for the parent or caretaker in this study. the presence of any adult, including the interviewer, could increase the tendency for children to give socially desirable responses. Other formats and measures should be tried in further research, perhaps to eliminate the need for the prcs- ence of an adult by using interactive technology. 63/ Journal of Marketiny, October 1995 Finally, though market shate and level of advertising ex- penditutu were not associated with recognition of Jae Camel and affect toward the cigarette. this does not rule out the possible impac: of these factors at lower levels.'Ihe in- creasing use of nontraditional media and sales ptomotion by adults-only products (e.g., see Stewart and Rice 1994) wen; not accounted for and may provide altcnutive ways to in- duce product use. With these limitations, the Cindings re- ported must be interpreted with caution, partiwlady when inferring managerial and publie policy implications. Nonetheless, some preliminary conclusions appear warranted. Children's Recognition of Trade Characters The results support Fischer and colleagues' (1991) earlier wotk, which finds reeognition, or the ability to match trade character and product, to be positively associated with the child's age. The lone exception was the Marlboro cowboy and cigarettes. The level of product and trade character recognition was generally high, ianging from 86% for the Disney Chan- ne1/Miekey Mousa mateh to 24% for the Marlboro Cowboy and cigarette (37% if adding in the lit match). It is also clear and consistent with previous research tlut adults-oitly ptod- ua trade characters ate readily tecognizcd by ddldren as young as three years of age. Joe Cantel, the ot6u cigarette trade charactey was matched by 25% of the thra-year-old children. 52% of the total sample (60.7% if the lit match is atlded), and 72% of the six-ycaroW group. Although these levels of recognition ara lower than Fis- cher and eolleagues' (1991) pteviass study of this age gtoup, recent studies show that older cldldton have an even higher recognition of Joe Camel. Henke (1994) reports Joe Camel recognition at 86% for a subsample of eight-year-old children. A 1993 Roper Starch survey for the RJ. Reynolds Tobacco Company showed that 95% of a natiocal sample of 10- to 17-year-olds knedr Joe Camel was associated with cigarettes. These findings would strike many as alarming if recognition necessarily leads to the formation of favorable attitudes toward and ultimate purchase of these products. However, the process of mere Utought, perhaps to some extent triggered by repeated exposure to a trade charactu, may Prompt increased ntgative evaluations if the product is initially viewed in an unfavorable way. This reasoning is speculative because sevenl explanatory factors beyond ex- posute level are associated with age. For example, the child's memory, infotmation processing ability, and base of information from other socialization agents (e.g., patents. day care personnel) also usually increase with age. The spe- cific model of how these factors affect one aaotlac needs further study to understand better the way children form preferonces. These findings reinforce prior research that showcd fre- quent exposure to stimuli is likely to prompt tnota oxtreme or polarized affects (e.g., Fazio 1986; Tesser and Conlee 1975) rather than only favorable atfectc (DiFranra ct aL 1991; Fischcr ct al. 1991; Zajonc 1968). The censal, tum- butger, and Mickey Mouse products showed a positive asso- ciation of trade character recognition and attitude toward the 52086 4557
Page 154: ozh90d00
also increase affect toward the McDonald product line with which it is usually associated in advertising and sales pro- motions (Aaker 1991; Thorson 1990). The vast majority of experimental evidence for the mere exposure effect is based on neutral stimuli, such as numbers, Turkish words, or Chinese ideographs (Bornstein 1989). Trade eharacters, designed to be memorable and present a persona that can favorably accent the manufactueer's mes- sage, could produce substantially greater favorable effects on attitude and product identity formation. Ogilvy (1983, p. 108) notes that trade chancters can become the living sym- bol of the brand and tend to be particularly effective with children in developing preference. Ecperireentally induced mere exposure effects on chit- drenn. There is substantial evidence to show that repeated ex- posure of product advertising aimed at children can increase preference for the product advertised (Raju and Lonial 1990), and relatively few exposures appear to be necessary (Botnstein 1989) to generate affect toward a stimulus. Gold- berg, Corn, and Gibson (1978) repeatedly exposed first- grade students to television advertisements for sugared products (snacks and cereals) and Public ServiceAnnounce- ments (PSAs) advocating pronutrition foods (e.g., fruits,' vegetables). Both messages increased behavioral preference for their respective categories when compared to control gtoups. In addition, an anti-sugaredl"junk' food program significantly reduced the students' choice of sugared foods. Goldberg. Garn, and Gibson argue that their findings reflect dte eBeas of inete exposure and reinforce earlier eonelu- siorts that the context in which the stimulus is viewed (e.g., positive advertiaments and PSAs, negative anti-junk food programs) is important. Thus, mae expasure could also ra duee positive affect (or increase the level of negative atti- tudes). However, their study developed "bnnd' and product affect for categories for which the child respondents were users. Adults-0nly products were not used in the experiment. Goldberg and Gom (1978) also tested the influence of repeated exposure to advertising on children's preferences for material objects (e.g., a toy) over more socially-oriented alternatives (e-&, friends). According to Goldberg and Gorn, these findings support the theory that (1) advertising favor- ably alters preferences in a product category in which v-tew- ers ate present users and (2) the influence of noncommercial sourees of socialization (e.g., parents) significantly negates advertising's effect. Gom and Florsheim (1985) found only mixed support for the potential effect of adult-product promotion on young girls. The advertisement for a product most girls in a pretest thought they would use (e.g., lipstick) influenced their atti- tudes and intended preferences. However, the advertisement for a product they did not see in their future consumption (e.g., a diet drink) failed to influence product or brand pref- erence on two of three auitudinal responses. Reports ojmedia use as a surrogate for mere exposure effects. Self-reports of media use are often used to gauge the potential exposure and influence of adult-product advcrtis- ing on children. For example, several studies have invcsti- gated whether potential exposure to proprietary drug adver- tising (e.g., medicines for headache, cold, cough, stomach, sleeping) was positively associated with children's peRep- tions, attitudes, and intentions to use these dmgs. Robertson. Rossiter, and Gleason (1979) reviewed many of these stud- ies and found inconsistent and weak effects of reported ad- veniscment exposure on young children's responses to non- prucription drug products. The results from their own sur- vey of third-, fourth-, and seventh-graders show large age differences on many measures. In an attempt to establish the unique contribution of (assumed) exposure to medicine od- vatising, they applied a multiple regression analysis inelud- ing scveral independent variables (e.g., illness experience, in-home inventory) in addition to advertisement expowre. They concluded (p. 254) that "there is only limited evidence of a link between proprietary medicine advertising and ehil- dren's medicine conceptions and requests ._ [and] no evi- dence for the proposed direct relationship between medicine advertising and media usage." It should be noted that the products studied were often used by children with and with- out adult supervision. These products also have general so- cietal aooeptanee and use. . Memory as an lndicator of Future Behavior Advenising copy tests often gauge recall (e.g.. Burke DAR) or recognitian (e.g., Starch Readership) to ascertain the au- dience's memory level of advutising. Although the specific relationship of memory to decision making is far from set- tlod, memory of advertising or brand names has long been viewed as eaosally associated with favorable attitudes to- ward, inteation to purehase, and actual purchase of the brand advertised (1lwrson 1990). Pedtaps becaase of this evidcnce, many researchers (e,g., Frseher et aL 1991) have used measures of memory as an indirator for potential be- havior. The validity of this intr.rpretat:oa is compromised by evidence that neither recall nor recognition is necessary or sufficient for affect fomution, which is often a critical com- ponent in making decisions (Borattein 1989; Chattopad- hyay and Alba 1989; Hoffman 1986). Therefore, high raog- nition of a ptoduct, a trade character, or a trade character and product association may not imply anything about the direc- tion of future affect or behavior toward a product. Negative affect toward a product or trade character can also be developed by repeatedly exposing those stimuli in a negative eontext, such as the anti junlc food PSAs (Gold- bcrg, Gorn, and Gibson 1978). Furtherntore. if a product is initially viewed unfavorably, repeated exposure can lead to increased negative aSeu (Schindler. Holbrook, and Green- leaf 1989). One reason for this finding was first investigated by Tesser (1978) in his experiments on mere thougLt. and has been sferred to as the polariiarion efJect. Eagly and Chaiken (1993, pp. 604-605) note that merely thinking about attitude toward an entity often makes it more extreme. and the mote a person thinks about the entity, the greater the probability of a more extreme or polarized affect. Simply sighting the attitude object can prompt mere thought (Fazio 1986). These findings suggest that frequent exposure to a prod- uct, with or without a trade character, could prompt either a positive or negative affect toward that product. Furthennore, the number of exposures necessary for this process to occur
Page 155: ozh90d00
can be quite small, and relatively more exposure (up to some asymptote) tends to further polarize a person's level of either negative or positive affect (for a review, see Bomstein 1989). Acoessibility of Affect and Behavior The source of the stimulus evaluation is of critical impor- tance in determining whether affect will guide such future behaviors as sampling and purchasing. An attitude object such as a product may not be as strongly associated with an attitude "following an indirect experience as compared with a direct expetience" (Fazio 1986, p. 221). Attitudes based on behavior or other direct experience are (1) more readily at- tributable to internal dispositions, (2) appear to involve a stronger objeet-evaluation bond, and (3) may. thunfore, be more readily accessed from memory to guide future prefer- ence judgments and behaviors toward a product. Although mere exposure and tttete thought, driven by frequent indirect stimulus exposutes, such as advertising, can increase aceess to the trelevant evaluations that may prompt response to the ptodueY, actual experience with a product prompts more readily acocssibk attitudes.lltese expuiences may be tteg- ative as well as positive. presutnably, it trade character can act as a we to prompt retrieval of either positive or negative evaluations. For eaample, a recent study of corporate and brand Wadematk logos reports that adults' opinions of tbtn- panies can actually be damaged by their logos (Leviu 1993). Tiade Chardcfeers and Childrcn's Deoision Afaldng Tbero is some anecdotal evidence about the effectivettess of trade characters on adult consumers (e.g. Aaker 1991; Jeasen 1993). The explanations for their influeooe tend to fows on the power of mythology (Randaa.o 1992). literuy devices (Stem 1993), or soutce effeets (Callcott and Alvey 1991). Early research on characters and children looking at the effeets of /wst,ttllmg, that is, using cartoon characters in a program as spokespersons, is equirvcal regarding its effec- tiveness. However, most of the recent coverage has dis- eassad the danger these cartoon trademarks pose to young dtildren (e.g.. Dagnoli 1991; Lipman 1991; Sehorow 1992). Only recently has empirical research reported evidenee and attempted to provide an explanation for bow trade characters influena the young (DiFtanra et at 1991; Fischer et at 1991; Pierce et al. 1991). Trade character ncognition. Fiseher and colleagues (1991) studied Wce- to six-year-old children attending Kitdu-Care Learning Centers in Atlanta and Augusta. Georgia.lLc children eompleted a matching task in which they were asked to plaee pictures of select trademarks (i.e., logos and ttade t#uractets) and one cigarette warning on the cortect picaue of a product that represented the brand using that trademark (or watdng).'ittis matching procedure was labeled recognition, though a requirement of previous expo- sure is usually made when using this torm. Apparently, the authors felt that the broadly promoted nature of the twenty- thtee btands justified this assumption. In Fischer and eolkagues' (1991) results, the level of correct trademark-product matching for several adult ptod- uqs exceeded that for many child-oriented products. Most of the tndemuks were logos of brand names such as Ford Motor Co., IBM, and Coca Cola. Only two trade charxters, Joe Camcl and the Marlboro t_owboy, were ineludcd. Not only did the Joe Camel trade character get the third highest recognition for an adult brand (51.1%-sligh8y less than Chevrolet at 54.1% and Ford at 528%), but the level of matching Joe Camel to a cigarette was not significantly dif- ferent from the Disney Cfunnel/Mickey Mouse match for the six-year-old portion of the sample (91% for Joe Camel. 95% for Mickey). Fiseher and eolleagues interpreted this finding as showing that the ptomotion of adultt-only ptod. ucts could produce high levels of product recognition in an audience that was neillKt legally capable of using nor fully able to understand the potential risks associated with the product's use. 'Ihere have been a number of articles (e.g., Krumske 1993; M'vstsln, Sonner, and Sttanghn 1993) that question the mcthodology used in Fischer and oolleagues' (1991) study. Frst, praestt for this researt:lt found that snost dtib dren under five years of age were uaabk to identify many of the produets Octuted and found it di85adt to match the large number of different logos and trade characters to the correct product p-tenaes. Secad, there was no attempt to address order effects on the product board, with Mickey Mouse and, the cigarette always placed in the top and bottom eentaof a four-by-throe picture grid. Thud, the interviewers were awate of the hypolheset and several potcntial outootnet In- tetviewets "bRnd" to the purpose of the study should have been used, because young ehildtat at sensitive to even un- intended innetviewet influence (eL Gatbotino et aL 1989). YlKse 6ctots may have influeaced the rate at which the chiWten were counted as matching the Joe Camel ttade character to the cigarette. Nonethekss, rates at any signifi- cant kvel may still be botltetsotne to public poGt.ynukus, parents, and edueatarz if this form of recognition translates into latar trial and we of this potentially hamtful adultt-atly product Recognition and adect toward die produa. Pischer and colleagtKa (1991, p. 3147) argue "tlnt brand awaeeness ao- ated in childhood can be the basis for product ptnferarce later in life."'1Ley admit that it is impossible to predict the effect of this advatiizing-induoed recognition on fuuae be- havior, but cite the tnae exposure research (Goldberg and Gorn 1974; Gom and Flarsheim 1985) as the basis for a link. In addition to DiFranu and colleagues (1991). Fischer and colleagues pmpose that frequent exposure of a trada character would cause high recognition of the trade c6arac- ter and producf, lead to favotable attitudes toward the ptod- uct. and influence the tlu'ldd to use the product later in tife. However, previous literature (e.g., Gom and Flmsheim 1985) has not supported the ability of advertisement wtpo- sute to prompt favorabk attitudes in children toward ptod- ucts they do not intend to consume in the future. Nor does the previous research (cf. Eagly and Chaiken 1993; Schindler, Holbrool4 and Greenleaf 1989) support an in- evitable asaoeialion between advertisement repetition, ad- vertisemettt teeognition, and positive attitude toward the product advettised. r
Page 156: ozh90d00
liked it or the piciure that meant you did not like it?" Even if the child failed to point to the correct picture after three at- tempts, the interviewer continued with the interview. Thete were no significant effects in any analyses if these few re- spondents had been deleted. The child was then asked for the name of a food he or she did not like. Citing that food, tha intetviewet went through the same discussion about pointing to the appropriate (.e., frowning) p'ictute. Anchor- ing affect with a food example was based on previous rc- seateh with four- to six-ycar-old subjects and the pretests for this investigation, which showed that young children's con- tmptions of the product categetxa, including soroking, tend- ed to be oral and uste-aiented. The chiklreds discussion about their judgments in protests showed that they were able to generalize the affect eoncept beyond food to the Mickey Mouse "product" and the lit match. After all the product pictures (excluding the blank/pic- ture of nothing) had beea judged as liked or disliked by the ehikl, the adult accompanying the child respondent was asked the child's name and age, whcther the child was at- tending any type of school or daytste ptogtam, and what level of school the adult had completed. Fuully, the adult was asked whether someone in the household used any of the products featured on the product picture board. The child and adult were then thanked for their assistance and escort- ed out of the interview rootn_ The interviewer then paid the adult $30. Results Fiitaf Sample Composition Although the incidonee levels for duw-. four-. five-, and six-ycarold children ate similar, three-year-old respondents wete subshntially more difficult to keep until the comple- tion of the intetview. Thetefote, the final sample had fewer throe-year-old respondents (n = 143) than was originally planned, and was slightly over quota for four- (a = 219). five- (n = 224), and six-year-old (n = 204) children. Recognition of the Product The children easily identified the product pictures at the be- ginoing of the recognition task with between 99.2% (hanr burga) and 96.8% (blank/picture of nothing) of the children pointing to the correct picture when asked by the interviewer. Trade t:haracter Recognition Matching by ehance. The hypothesis proposed that (1) the respondents would correctly match trade characters to product piewres at a rate beyond ehance and (2) this tato would be positively associated with the child's age. Table 1 provides the observed rates for correct matching of trade clupctats with their appropriate product picture. The ntes of matching were higher than expected by chance (one of six, or 16.7%. per match) for all ages, which supports the first part of the hypoUKsis. Age and rccognision of tmde rlmmuen. Also as hy- pothesized, the ability to make the correct match was signi[ icantly diffenent by age ()26 = 137238, p = .0001) and in- creased with the age of the tespondent. The relationship of tF• matching and recognition rate to age was tested with a weighted least squares analysis. All trade character tocogni- tions (matches with the eonnd product). except for the Marlboro Cowboy, were significantly (p <.05) associated with age in a positive way (see Table 1).'Rtese findings an: similar to those reported by Fischer and colleagues (1991) and support the suggestion of increased learning and mem- ory through both increased exposure and more developed cognitive and infotmation processing abilities. The Marlboro Cowboy was the only trade character that did not show a significant relationship with age. Hecauso Fischer and eolleagtKs' (1991) original study did not report ehese data in an unaggregated form, it is not elear if the.par- ticuiar, trada character rendition used was a faclor. The spe- cific character shown in Marlboro cigarette advertising has not been consistent unlike the unique and consistent 6gures representing the other produrxs kmned'in this study. Also, as a child ages, he or she soes a cowboy in many non-Macl- boto settings, which could inueasingly dilute the strength of a cowboy association with eigarettes. EWiret of produet and trade dramcter. Any eondusions about the influence of product and trade chatacter on the recognition levels must be tentative bemuse of-the satall sample of seven trade ehuaetedpmduct nutcttes stud;ed. Cigarettes and eeeral had two examples each. The hambutg- et, Mickey Mouse, and the picture of nothing optioo each had one approptiate trade thataeta: Nondheless,ioolting at their relation to one another may help researchers gain some insight into the relative levels at which young children can carteedy telate a btand's trademark to ghe appropriate ptod- uct ntegory. Across Ute total sample, the Disney (haoneUMickey Mouse matc4 displayed the highest level of recognition (86%). This was followed by the two ceteal trade ehuaetets (Gptain Cttmch, 72%; Tony the Ttger, 60%), Joe Gnxl with the cigarette (52%). Charlie Tuna with the no product option (52%), Ronald McDonald with the hamburger (Sl%), and the Marlboro Cowboy with the cigarette (24%). If the lit match was also counted as a correct association with Joe Camel, the total reeognitton rate of this trade charackr would rue to 60.7%. Tbis increase has little effect on its rank or later analyses or eonelusions. If the lit nutch was also counted as a correct match for the Marlboro Cowboy. the total recognition rate fot this cbaracter would increase to 37% of the children. Even with the addition of this addi- tional possible anteh, this aado elurxter would still have the lowest reeognidon. Applying a Honferroni family etror rate and using preplanned comparisons, Joe Camel was not significantly mote (p >.0S) recognized than Ranaid Mc- Donald (7(2t = 36) or (]nrlie Tuna (x2t =.22), but was sig- nificamly (p < Ad) less recognized than Captaia Guoe6 ()zt = 101.2). Tony dte Tiget Wt = 13.14), and the Disney Channel (X2t = 348.1). There were no clear environmental factors that showed influenee. For example. the recognition of Joe Camel was not significantly associated with cigarette use in the botue- hold (56% in cigarette households, 50% in noncigateuo U N 0 m m ~ts u, tn A I
Page 157: ozh90d00
product. These products are expected to be usually viewed favorably: howevet, cigarettes ate expected to be viewed un- favonbly. lnditoct, nonexperience-based (e.g., PSAs, postess, antismoking messages from parents and teachets) recognition of cigarettes increases as the child grows older, while favonble attitude toward cigarettes appears to dc- csease. Nonetheless, the older the child, the more likely he or she is aware of the socially acceptable tesponse. Addi- tional ncscatch is neeessary to eliminate the possible influ- ence of demand artifacts. 'fTe product category of matches is an example of how this process may work without using a trade character. The favorable attitude toward matches decreased from ttwse children three to six years of ago. However, something must happen between this young age and adulthood without the aid of eonsumeroriented advertising or trade eharactess, be- cause few adults would be expeaed to have a negative affect toward matches. Fazio (1986) has shown that experience-based infotma- tion is much more influential than indirect-based infoema= tion for developing attitudes and aeeessing thesa attitudes for later behavior. In the cace of tnueha, dtildnat's eady supervised and later unsupervised use in lighting such things as outdoor grills and piles of leaves may lead to mote favor- able attitudes over time. This continued expesiettee-bued exposure as they grow older can itself reinforce the foana- tion of favorable attitudes thtough mae exposure and thoughL This may ultimately help change the polarity of the attitudes toward the product fiom unfavotabk to favorable. Further tesntett into this paradigm is ttadad to asar- tain the telative in(htenee of autlcUittg stimul't, such as saks promotions (e,g.. Stewart and Rice 1994), benuse adver- tised trade eharacrets do not appear to Inve a stroong nnpact. The activities of socialiration agents appaa to exett a very powetfitl influtatce on young childten's attitude towatd dg- aoeuea and, probably, on their attitutfea toward the other products tested, as well. As the findings regarding the lit match suggestc, neilhet compelling eottoott ttade ehacactecs nor sivble advertising budgets ate significant factots in the process of developing affect toward an adults-only product. REFERENCES A.C Nidsat Company (1991). Closrary. Nonhbtook, IL: AC N'rJsat Company. Aaket David (1991). Monoging &wd Equiry: Capitafi7ing on the Value oJa Brand Name. New Yodc The Free Press Baxtrr Pamda 1. (1992). "HowChisdn:n Use Media and Influence Pundtsus,' Jownol ojAdverti.sing Research. 31 (December). RC-2-4. Bekh, Geotge and Michael Bekh (1995), Introduction to Advcr- eisint aaJ Pnowmooe: An lntejsaad Man}etFej l.ananu.nica- tiow PerapectirG 3d ed. Chicago: Richard D. Irwin, Inc. Bdk. Russell, Roben Mayer, and Amy Driscoll (1984). '~Chil- dnvt's Recognition of Consumption Symbotisat in Children's Ptoduets," Jovmal of Consumer Research. 10 (Mareh), 386-97. Bamstein. Robert (1989). `Ecposme and Affect: Overview end Meta-Anslysis of Research. 1968-1987," Prychological Baf- k+fe. 106 (2), 265-89. 6ntcks. Menie. Marvin E Caldbecg, and Gsrt AtmWOttg (1985), 'Children's Cognitive Responsrs to Advettising," in Advances ia Cmu~rRumrd',VoL 8, Richard J. LuK. ed Pcovo. Ui: Assacirtian for CAmuma Research. 650-54. Calkan, Margaret aid Patricia Alvey (1991),'?oons Sell _ and Somuimes't1n:y Don'c An Advextising Spoka-Chuacta Ty- pology and Ecpbca[aty Sudy;' in Proceedingr of the Muri- can Amdany ofAdreniriej. Rebcan Holmun, ed San Anto- nio, T)C Amniean Academy of Advatising, 43-52 Chattop.Aiyay. Ameava and Joseph Alba (1989). "1Ue Relation- ship Baween Recd4 Cognitive Responses. and Advertising Ef- fectiveness: FJfects of Delay and Contest,' Wott;ing PaperNo. 89-103. Cambddgey MA: Marketing Science hntitute. Dagnoli, Judsim (1991). "Ilwee Faces of tCool,' Advatisiad Aae. (October 9). 54. DiFnns; Joseph D., Jom W. Ridprds, Paul M. Frecnun, Nancy Wolfr,'iBespia t7ristopher Elachu, Robert D. JaRS and David Mortay (1991), 'R1R Nabisa's Cattoon Camd Pm- motes Camd Gga¢nes to CbiWicn,' Jownol ojshe American Mediaof Asmcimion, 266 (22), 3149-53. DonduK Thomas. Tunothy P. Meyet, and Uwy L HeaJce (1978), "Blrk and White CM'Wiar Paoepaons of7V Commacials," Jownaf oJMarteting. <2 (Oaobrrr), 34-40. Eagty, Atia ana SIKBy Gaites (1993), T>tie Piychology oJAni- aMa. Na. Yod¢ Hatown Btaec lo.anovic4 College Publish- ers. Fazio. RnneR (198ti), "How Do Attituda Guiee BdnviaFf' in NaRiowt of Moarylea and Cs{adioa, Richard Sarmatino and E Higgins, ads, New Yott The Ga'Ifotd Ptess, 204-43. FtseMt. Paul M., Meyer D. Sdtwm, Joln W. Richards, Adam O. Goldacin,anA TmaH. Rojas (199q,'Bsmd i ogoRotopti(iott by Childn:n Aged 3 to 6 Yeaa,"JaonalofdieAme.im. Med- irotAaorimion, 266 (7l), 3145-a Fagaat4 K(1940),`AAn(ts CavorCtstsoe Tooosia'[hdr Ads," AlvrraaegAg4 (Septm6a 17).3. Gatb.rmo.Jsmes,FtatorsM Stat stdFaadryaftbeEnksooln srilute (1989), WiW C1y7dran t:an TaR Ui San Ftsaciseo, CA Josey-B:sr, toc, Gatfield, Bob (1991J6 "qoesaiad Spntkk Hdpa 8*iglrm Ixk- lusta Yer: Adradnin j Age, (Dobembea 23).30. Goldbag, Marvin E ana C,crald Z Cota (1974),'+Qu'Wim's Ro aUions to Tek.ision Mretdsiag: An Fifaimental Ap- ptwch,"JoundejCansrnerRamnd; I (Septeatba),69-75. aod -(1978),'Sonte Uniotnded Comequawts of 71Y Adratsing to Ot7dia4 laYnml oJCoa.nincrRermnc/s S (June).22-". --, asd Wayne Gibwe (1978), 'TV Messages for Snack and BnLthst Foodc Do7bry Ldlueaoe dhildeats Pief- emrors; Jonmal of Contano Ram.ck S(Septetnbet). 73-81. Gotdon, P. and K Holyoak (1983), "Iniplicit f.eatnittg and C+etter- alivtion of ttte Mete Sxpmre afecf,' Jawnd eJPosonoliry and Social Ps)dology. 45 (3), 49?r500. Cotn,('xraW J. and Rrnee Has6eim (1985), "1hc Effects of Cwn- soaeiais for AdultPtoductson (]dldten."Jou.nd aJCansumcr Raearch, lI (Matcti), 962-67. Hente, tycy (1994). `Young Qu'Idem's Perceptions of Cganene Bnud Adwtisiog Symbols: Awstaness. Affect. and Target Mattu Ideotifnatiott,' in Procet6njr of ihe 1994 Amenican Acodany aJAdrerrising. lCan:a W6itehill IGng, ed. Tucson, AZ• American Academy of Advanisittg, 59. Hoffman, Martin (1986).'Atfect, Cognition and Motivatian,' in Hondbook of Mapwtiae and CognFBoR, Richard Sotrentitw and E FBgg:u, ods. New Yott: The Cilfoni Press.2A4-80. Jetuet4 Je(f (1993), "Aoimuiott Renaissance Snares IGds."Advn+ +iring A;e, (Febroat)+ 8). $40. Kobasidawa. Akira (1977), 'Reeieval Strategies in the Develop- ttwd of Memory: in Penpecdver in dw Developwnr ofMan- 07 and Cognition, R. V. Kail and J. W. Hagan. eds. HiBsdale. NJ: Lawrence EAbwm As.aciaes.
Page 158: ozh90d00
agents of socialintion, such as parents, peers, and teachets. as well as government and private programs. Age and Advertising Effects For children under seven years of age, their age appears to be the most dominant factor in affecting responses to adver- tising. Piaget (1970) proposes that children go through stages of cognitive development in which they exhibit dif- kteaoes in the way they process information in their envi- Mnment It has been suggested that children between two and eleven years of age (the pteopecational and concrete operations stages) are the most vulnerabk to advertising, be- anse their eogaitive structures are beginning to form and they ate most sensitive to e:tetnal influences (Raju and Go- nial 1990). Stulls and Hunnicutt (1987) suggest focusing on childtett two to seven years of age. During this period, they become increasingly controlled by images and symbolic ptoassea, including those from advutising, and begin to maha judgroentt about products they might use in the future. Many tescatehers dtsagree with the validity of discrete stages, beptt9e the ttteasurcs used (e.g., the mone age sensi- tive verbal ttteawtes or the nonverbal measures) can signif- iatuly affect the age at which children tan be seca to man- ifest cognitive capabilities (e.g., Donohue, Meyu, and Henioe 1978; Maeklin 1985).Others such asRoeddet(198Q suggest that the age-telated ddfueaces ate due to changes in the ehild's ability to store and retrieve information. Naaedtekss, a large body of evidence has shown thu (1) age is teiated to a child's response to advettising s6muG and (2) children between the ages of two and seven tar, ar- guably most at risk. Age And fnfonnavon Prooesaing A young child's age tends to be associated with his or her capability to perform a wide range of infotuution process- ing activities that should affect the impact of advertising. Foresample, attention to advertising is generally reported to decline with age (Ward 1972), whereas a ehiW's ability to discriminate between a program and the commercials on television increases with age. Comprehending advertising oootent such as the attributes of the produtx, is also posi- tivdy associated with age (Ward, watamnn, and Wactelia l9TT). CWser to the issue of ttade character influence is ta seareh concerning whether children understand if the thar- ataers used in advertising ate tpl. Numerous studies have shown that a child's age is positively i«„wsaf..,t with his or her ability to distximinate hutttans from anitnated (Quar- fath 1979; Van Ankea and Lottial 1985) or cartoon cituao- tas (Reeves and Greenberg 1977). Raju and Lonial (1990) review this literature and warn that children under five years of age may be confused and have difficulty differentiating betwept httmans and chatacte[s. Age and Memory If then is an etlwl knowledge base for-a topic, age has weak effeefs on many memory tasks. However, a ehtld's natural knowledge base is positively related to ege (Bruda. Gold- berg, and Artttst<ong 1985). Sever+l researchers have sug- gestcd that this knowledge base can be developed through the sheer repetition or mete exposure of advertising. whether or not ehilr:rat ate the targeted eotaunta group (e.g., F'ueher et al. 1991). Such research suggests the subse- quent hypothesis. H: A young ehitd's eartieet muching (or n:cognitian) of trade charauers widi their apptoptiue product pitwas will ex- caad dmoe, and will be positively aaaciued.riTh the age of the child. Age and Attitade Formation Age has been found to be the most important dispositional factor for a child's attitude toward an advertised brand of product (toys and games). For eaample, first- and tttird- gtade childten were found to be tnote affedod by advertis- ing nhan du'Wcen in the flflh gtade (Robettton and Rossitu 1974). Age is aiso positirt;iy associatod witlt atakiag attiotde consistent daioes (Roeddor: Stanthal, and GWrt 1983). .Howaver, all these sttttGes used adna6sing _spoa5rally tar- getcd to childrett. Few s~tdies havo loolood at age effects in processing advertising far adult products (e.g.. DtFrana et al. 1991; E'isdter et aL 1991). 7he Effects of Repeafed Eigoowre Reseatdt hu hmg doctaoeated tluttGpeatod eapo:,ue to ad- vatisi ng can enhance its efftetiveoeu. Both mamory and at titude toward the produet adnatised tuually bear a pat3tive av.oriatioa to tepaiitiaa ~ 1979) aod eat:h otht 1Lor- son (1990, p. 223) 'oaieltide's tier teview of idverasing's in- nnetioe on c~tia~w.n~t~ers,amating-_-ih+yt va.~n.t,d__t o~[t its,imLp~a~et "Q1ay ~. 4{1{R IOG it~-telta ~ {N tO~tllV1R the hllndlEdf iad p~ 1601Wndt~Of ad5 l0 M~ICb MC ZitE expased' For du~dten, tltis Ivtowledae brtte beoomes laiger as they gtovr older and contaitrs infotwtion ftom sodaGrs- tion agents, as wep. HowevaG some Rseaeoh suggests that frequency eBects become weaker as the knowledge base de- velops (Raju and 1.on4a11990). Zajonc (1968, p. 1) inorodttud the ttteoe expoaue pbe- nontenon, which he deGned as eoanring when 'btete ro peated exposure of an iad'iridttal to a stimulus is a mffieirnt condition fataihanoemeat cf his suituda toaratd it' Ie other words, bmtTisity leads to lilcing (t.e„ a positive oa favorable affetx).'Ihis phenomenon appears to be robust and teTuble (Borttstdn 1989), thoagh'its rationale has boen the subjeet of some criticism. For eaaople, tbete is no single aoxpted attoory explaintng the ptocess, and it is not clear under what specific conditions mete exposure effects will oocttr (sco Gordon and Holyoak 1983). A strict iataptenuion of the ed'eet of tnere exposure im- plies that repeated atpoapte to a novel stimulus such as the Ronald McDonald trade ehuaaer leads Ute attdiaooe to in- ctease t6eir pasitive aHeet (q.uatd Ronald MeDooald- Htwr- ever, tatdoete~s gtmerallr want to sell the brand of product using the trade clutracK ttot sell the tiade dmaaer itsalf. The ability to mctease 6vonble aBect toward a stum+lut through repetition appeats to teadily genetal'me to otbet stimuli assocaaoed with that stimulut (Goldbas, Gorn, and Gibson I978; Gordon and Holyoak 1983). For e:ample, in- creased affect toward the Ranald McDonald ehanaaeould
Page 159: ozh90d00
CF I 7- 8- 9- I 10- 11- 12- 13- BLOCK N 14- COUNTY PLACE AM 1 19/ ' , AN Time started PM 2 Time finished PH Total minutes 20/21 15/16/17/18 The Rdp@r Organization, Inc. 205 East 42nd St. New York, NY 10017 Study •343-189 October 1993 ADVERTISING CHARACTSR AND SLOGAN SQRVEY My name is and I'm fron The Roper Organization, a national public opinion research firm. We're conducting a survey among young people about the use of characters and slogans in advertising. May I please speak with the person in your household who is between the ages of 10 and 17 and has had the most recent birthday? INTEMENERo IF THIS PERSON IS NOT AT HOl4E, FIND OUT THE BEST TIME TO CALL BACK TO SPEAK WITH HIN/HER. ENTER HERE: IF THERE IS NO ONE IN THE HOUSEHOLD 8EIWEEN THE AGES OF 10 AND 17, THANK AND TERMINATE. 1. To begin, please tell me how old you are. 10 years ......... 1 14 years ......... 5 22/ 11 years ......... 2 15 years ......... 6 12 years .........3 16 years ......... 7 13 years ......... 4 17 years ......... 8 2. What was the last product that you recall seeing advertising for? I (name/kind of product) 23- 24- 52086 4533
Page 160: ozh90d00
Teen Smoking - Page 31 smoking intensity suggests that the influence of peers is relatively greater in explaining more regular smoking than in explaining trial. This pattern is more consistent with the interdependent utility imupretation of the effects of peers than it is with enhanced availability, though it does not rule out the availability explanation. Moreover, the greater sensitivity of boys to the behavior of close friends of the same sex is only significant in the current and daily smoker equations." There is somewhat less consistency in the performance of the family measures across the differeat equations. The number of adults in the household who stnoke is only significant in the daily smoker equation. As noted above, the weaker performance of this variable in the national sample may stem from the lack of information on household size. Consistent with this explanation, in the California results, the household size variable is only significant in the ever smoked equation; it is significant at .15 in the current smoker equation. The influence of older siblings of the same sex closely parallels the influence of best friends of the same sex, although the coefficients are smaller. Older siblings of the opposite sex fall just short of significance in the daily smoking equation. As in the combined model, the dummy variables for a parent in the household who smokes are generally negative but not significant.'6 The greater tendency of teens to smoke if one parent is not in the 'sThe only apparent difference in the California sample is that the increase in the coefficient of same sex best fiieads at higher levels of smoking intensity is not as pronounced. In particular, there is very little difference in the coefficient between the current smoker and daily smoker equations. "In the California sampie, older opposite sex siblings are sigttificant at .06 in the ever smoked equation, and insignificant otherwise. The dummy variables for mother or father in the household who smokes are only significant in the current smoker equatioa, and the coefficient for Dad smokes is positive (but insignificant) in the daily smoker equation.
Page 161: ozh90d00
---

Text Control

Highlight Text:

OCR Text Alignment:

Image Control

Image Rotation:

Image Size: