Tobacco Documents Online

C 0 N F I D E N T I A L October 6, 1982 TO: Mr. L. W. Hall, Jr. FROM: Ms. D. S. Burrows RE: NBER MODELS OF PRICE SENSITIVITY BY AGE/SEX BACKGROUND In 1981, the National Bureau of Economic Research (NBER) published two models based on consumers' responses to government smoking surveys. Because the NBER models were cross-sectional (testing differences between people rather than over time) they were able to relate price separately to incidence and rate per day, by age and sex. The NBER elasticities may not reflect the exact effects of price increases, since their models didn't deal with changes over time. However, the effects they found for other variables (race, income, working women, etc.) are highly consistent with our understanding of market dynamics. Thus, it is likely that the NBER models have correctly identified relative price sensitivity among age/se:x groups. A detailed critique of the NBER models i Attachment B. SUMMARY OF FINDINGS According to the NBER models: • Teenagers and younger adult males are highly price sensitive. • Males over 35 have above average price sensitivity. • Women and 26-34 year old men are relatively immune to price. • Price affects incidence; rate per day is virtually unchanged. NBER PRICE ELASTICITY AMONG: TEENS AGES 20-25 AGES 35+ TOTAL 12-17 TOTAL MALES TOTAL MALES 20+ EFFECT ON: Incidence -1.19 - .74 -1.28 NA NA - .26 Consumption -1.44 - .89 -1.40 - .45 - .66 - .42 DSB/ch - 10/6/82 Code: 5.21

CONCLUSIONS In terms of immediate impact, the effect of price on males 35+ is most important. Half (50%) of the total drop in industry volume is attributable to males 35+, compared to 24% from younger adult males, and 7% from teenagers. (Calculated in Attachment A) But, the loss of younger adult males and teenagers is more important to the long term, drying up the supply of new smokers to replace the old.* This is not a fixed loss to the industry: its importance increases with time. In ten years, increased rate per day would have been expected to raise this group's consumption by more than 50%. ~ Diane S. Burrows MARKETING DEVELOPMENT DEPARTMENT DSB/ ch cc: Mr. J. R. Moore Mr. H. Ms. E. Mr. E. Mr. J. Mr. P. Mr. W. Mr. R. MDIC E. N. J. R. E.. W. A. Osmon Monahan Fackelman Hribar Galyan Doten Davis t *As discussed in my 9/21/82 memo re "Estimated Change in Industry Trend Following F.E.T. Increase". DSB/ch - 10/6/82 Code: 5.21

ATTACHMENT A IMPORTANCE OF PRICE IMPACT BY AGE/SEX TO TOTAL INDUSTRY 1. 1982 IMPORTANCE TO INDUSTRY IMPORTANCE - % OF - RATE PER X OF TOTAL SMOKERS DAY CONSUMPTION Teenagers 12-17 4.0 E 17.4 E 2.2 E Males 18-24 8.0 29.7 7.5 Males 35+ 28.7 36.9 33.6 Total 18+ 96.0 31.7 97.8 TOTAL 100.0 31.1 E 100.0 Source: Tracker (X smokers first half 1982, rate in year 1981) adjusted for estimated teenage smoking. II. IMPORTANCE TO PRICE IMPACT: Sample calculation assumes 10% price increase and flat 623.0 billion industry. Importance of group to total change is independent of total volume or price change. IMPORTANCE TO INDUSTRY NBER ELAS- LOSS FOLLOWING 10% PRICE INCREASE X BILLIONS TICITY BILLIONS IMPORTANCE Teenagers 12-17 2.2% 13.7 -1.44 2.0 7.2% Males 18-24 7.5 46.7 -1.40 6.5 23.6 Males 35+ 33.6 209.3 - .66 13.8 50.0 Total 18+ 97.8 609.3 - .42 25.6 92.8 TOTAL 100.0 623.0 - .44 E 27.6 100.0 DSB/ch - 10/6/82 Code: 5.21 - 3 -

ATTACHMENT B TECHNICAL SUPPLEMENT SUMMARY AND CRITIQUE OF THE NBER MODELS Lewit and Coate of the National Bureau of Economic Research developed two series of models relating cigarette prices (and other factors) to reported incidence, rate per day, and "total demand" (incidence times rate). Although reported rates understate true consumption, the models would still be valid if the under-reporting was constant across the other variables -- geography, age, sex, etc. We ordinarily use this assumption ourselves. Both studies were weakened by assuming that prices in all locales/outlets in a state were similar to the statewide average reported by TMA, plus local tax. This may be why both models had low adjusted R-squared values (.11 or less, but still significant). Other aspects of the NBER studies are critiqued separately below, since each used aa different data source, time period, and methodology. ` THE OVER-20 MODELS SOURCE: "The Potential for Using Excise Taxes to Reduce Smoking," Working Paper No. 764 of the National Bureau of Economic Research, by E. M. Lewit and and D. Coate, September, 1981. This ordinary least squares model used data tapes of individual responses to the Health Interview Survey of 1976. It expressed incidence, rate, and "total demand" of respondents over age 20 as a linear function of: o Retail price, defined as the TMA - reported average in the respondent's state, adjusted for any local taxes. o Family income and size o Sex o Marital Status o Health status (perceived) o Region and city size o Race o Working woman or not Using the full sample, price was not found to be a statistically significant factor in incidence, rate, or "total demand". Coefficients for most of the other variables were significant (at the 5% level) in a direction consistent with our consumer research findings. H 0 .6 %o 0 m ao DSB/ch - 10/6/82 W Code: 5.21 - 4 -

The models were respecified using a restricted sample which eliminated respondents within 20 miles of a lower priced state, on the premise that the price they actually paid might be lower than the assigned price ("border effect"). For the restricted sample, price was found to be a statistically significant factor in: . Incidence and "total demand" among 20-25 year olds. •"Total demand" among those over 35. When the regressions were done by age/sex, significant coefficients were found only for males. Though none were found in the female regressions, the higher elasticity for males/females combined than for males alone may imply some effect among females. Selected elasticities are tabulated on the next page. Only boxed values have statistical significance at normal levels. DSB/ch - 10/6/82 Code: 5.21

NBER ELASTICITIES AMONG PERSONS 20+ (1976) (Sample Restricted to Eliminate Border Effect) REPORTED REPORTED "TOTAL DEMAND" INCIDENCE RATE _ (INC. x RATE) RETAIL PRICE Total 20+ .264* - .037 I - .4 16** 20-25 .74* .20 26-35 - .44 .04 - .47 3_`i+ - .15 - .15 I Males 20+ 20-25 Fl 1.276* - .171 -1.401* 26-35 - .292 + .029 - .320 3`i+ - .246 - .204 I - .6 W-1 Females 20+ 20-25 - .136 - .026 - .302 ., . 26-35 - .388 - .134 - .577 3'i+ + .066 - .077 - .118 S INCOME Total 20+ + .03 .06** *OLS coefficient statistically significant at 5% level (2-tailed test). **OLS coefficient statistically significant at 1% level (2-tailed test). DSB/ch - 10/6/82 Code: 5.21 6

Since the NBER models are based on geographic differences in retail price at one point in time (1976), they do not translate directly to national changes over time: ., The positive income elasticity suggests that, over time, price elasticity should be applied to retail price deflated by some measure of consumer income. However, the price elasticities are strongly age/sex specific and income measures/projections are not available by age/sex. .. Since consumer prices may continue to rise faster than income (especially among young adults), using the CPI as a deflator probably yields conservative estimates of the price impact. ., Restricting the sample to eliminate "border effects" improved the model's applicability to the nation. However, it created some regional bias -- Westerners rose from 18% to 29% of the sample while Northeasterners fell from 24% to 11%. This may have made the price elasticity more negative than the "true" national value, since positive_coefficients_were_as-sosiated_with__residence in_any_regio_n_ except the West. Thus, the adult NBER elasticities are imprecise for national price changes. However, they probably give a reasonable idea of the groups most affected by price and the orders of magnitude of the effects. THE TEENAGE MODEL SOURCE: "The Effects of Government Regulation on Teenage Smoking," by E. M. Lewit,! D. Coate, and M. Grossman. Journal of Law and Economics, December, 1981. These models used personal interview•data from Cycle III of the Health Examination Survey*, conducted 1966-70. Incidence and rate findings from HES III seem reasonable, perhaps even a cut above most teenage studies. But it is not recent. A major goal of this study was pre/post assessment of the Fairness Doctrine, which loaded television with anti-smoking commercials from 7/67 through 12/70. Other variables included: *In Self-Reported Health Behavior and Attitudes of Youths 12-17 Years, United States. HEW, 1975. DSB/ch - 10/6/82 Code: 5.21 - 7 -

• TMA retail prices, adjusted to state cost-of-living and changes in the national CPI during the 4-year survey (an improvement over no adjustment, but still makeshift). • Fairness Doctrine intensity versus cigarette advertising and amount of time spent watching TV • Family income deflated by state cost-of-living • Number of other children • Parents' education • Working mother or single parent • Employment/allowance of youth respondent (detailed) • Student status • Age • Sex • Race • Region and city size. Large, negative price elasticities were found. The coefficients were stable and statistically significant in all model specifications reported. The authors suggest that these lar e elasticities ma incor orate other effects which the model did not include capture separately. They suggest that the income related variables may have failed to represent the youth's actual discretionary income. NBER PRICE ELASTICITIES AMONG YOUTH 12-17 REPORTED REPORTED "TOTAL DEMAND" INCIDENCE RATE (INC. x RATE) ~.... .~.._~. Total 12-17 - 1.19 - .25 - 1.44 The Fairness Doctrine was also found to have an important negative effect during its first year (accounting for a 3 percentage point drop in 12-17 incidence), but showed diminishing returns during its second and third year. Cigarette advertising on TV was found to partially offset the Doctrinw effects. Thus, the absence of TV advertising after 1970 was implicitly a negative effect, but the absence of the Fairness was a larger implicit positive. DSB/ch - 10/6/82 Code: 5.21 - 8 -

Results for other variablez were not reported. Weaknesses in the teenage NBER model are: • Loose estimation of price, loosely deflated. • "Impure" price elasticity (probably somewhat inflated). • Data from the late 1960's, which may not reflect today's market. National application of this model has the same problems as in the adult model, except that the problem of estimating what teenagers perceive as "real" price change is even worse. DSB/ch - 10/6/82 Code: 5.21 9