Anne Landman's Collection
1600 - SMOKER PSYCHOLOGY SMOKING AND SOCIOECONMIC STATUS
Abstract
This confidential Philip Morris internal report on "smoker psychology" explores the relationship between socioeconomic status and smoking. It finds that:
"Lower class panelists smoke more and are much more likely to be smokers than upper class panelists..."
It also found that lower class people tend to smoke nonfiltered cigarettes (tend to "avoid health filters") and that they also tend to avoid 100 millimeter-length brands.
The writers also observe that lower class people have more incidence of poor mental health, hypothesizing that people use smoking as a "strategy" to combat the stress of low class status as well as poor mental health:
"...the incidence of poor mental health is greatest among the lower class...To the extent that smoking is one of the available strategies people can adopt to combat stress, we therefore would expect greater incidence of smoking among the lower social classes."
The study also finds a correlation between lower class and poor physical health, but avoids directly confronting the possibility that smoking could account for this, preferring to attribute poor physical health status to simply to BEING a member of the lower class:
"...because the incidence of smoking differs between the social classes, we would find our research literature filled with obervations suggesting that smoking is related to poor health. The literature does show this, and it may be wrong...At least part of the reported statistical relationships between health variables and incidence of smoking can probably be accounted for in this fashion. The smoking and health relationships may be at least in part due to social class differences rather than to smoking per se."
Despite Philip Morris' internal findings of higher smoking rates among lower socioeconimic classes, as well as its findings that this group also has a higher incidence of both poor mental and physical health, it continued to promote its deadly and addictive products heavily among these groups. One must question whether this violates state charters for incorporation, which generally require that a corporation does not harm the population.
Here is what Colorado's state constitution says about revoking corporate charters:
"Section 3. Power to revoke, alter or annul charter. The general assembly shall have the power to alter, revoke or annul any charter of incorporation now existing and revocable at the adoption of this constitution, or any that may hereafter be created, whenever in their opinion it may be injurious to the citizens of the state, in such manner, however, that no injustice shall be done to the corporators."
Would it be just or unjust to Philip Morris to revoke its corporate charter if it was shown to have knowingly degraded the health of Colorado's least well-off citizens to achieve higher profits?
Fields
- Quotes
A survey of the socioeconomic status of...panelists reveals that
1) Lower class panelists smoke more and are much more likely to be smokers than upper class panelists...
...Social class influences cigarette type very little except for lower class tendencies to smoke nonfilters and avoid health filters and 100 mm brands.
...Social class of respondents cannot be ignored in studying smoking behavior.
...Kinsey reports have shown that the sexual standards and behaviors of the classes differ; even casual observation shows that the classes have different likelihood of attending specific protestant churches (the funamentalist churches appeal of the lower classes, and the Methodist and Episcopal to the upper classes); many consumer studies reveal differences among classes in preference for private brands vs. tradenames, for type of retail stores frequented, and for concern with the appearance of their homes...
...A few recent studies have suggested that (as a consequence?) there are personality test score difference between the social classes, so that lower classes appear more anxious than upper classes. As a corollary to this observation, we note that still other investigatons report that the incidence of poor mental health is greatest among the lower class...To the extent that smoking is one of the available strategies people can adopt to combat stress, we therefore would expect greater incidence of smoking among the lower social classes.
...The naive experimenter who ignores social class and measures only intelligence test scores of...[smokers and nonsmokers] will therefore conclude "that smokers have lower intelligence test scores than nonsmokers." This is true, but it is a nonsense statement. It implies that smoking is somehow related to intelligence test score. Within each social class however each smoker has exactly the same score as each nonsmoker. What the naive researcher is seeing is the confounding effects of different social class intelligence socres and different proportions of smokers...
As a second hypothetical example, suppose that some index of poor health is related to social class in such a way that the well-paid, well-educated white collar workers enjoy better health than poorly paid, poorly educated laborers. Then, because the incidence of smoking differs between the social classes, we would find our research literature filled with obervations suggesting that smoking is related to poor health. The literature does show this, and it may be wrong.
At least part of the reported statistical relationships between health variables and incidence of smoking can probably be accounted for in this fashion.
The smoking and health relationships may be at least in part due to social class differences rather than to smoking per se.
It must be cautioned that this observation does not mean smoking is independent of health problems...
- Company
- Philip Morris Cos., Inc.
- Author
- Johnson, M.E. (Philip Morris)
- Ryan, F.J. (Philip Morris)
- Recipient
- Fagan. R
- Fountaine, S
- Graham, R
- Osdene, T
- Resnick, F
- Seligman, R
- Thomson, R
- Udow, A
- Wakeham, H
- Fountaine, S
- Litigation
- United States
- Type
- Report- Scientific
- Subject
- Health effects
- Mental health
- Smoker psychology
- Mental health
Document Images
1
Accession No. 73-080
1
1
1
1
Copy No.
7
Issued To a
PHILI. P MORRIS U,S,A,
R E S E A R C H C E N T E R
CHARGE NO~, & TITLE: 1600 - Smoker Psychology
.
TYPE REPORT: OANNUAL OSEMIANNUAL © COMPLETION ®SPECIAL
za
7 DATE: June, 1973 PERIOD COVERED:
1
I
~
~
I
I
I
I
I
I
I
REPORT TITLE:
S. Fountaine
R. Seligman
A. Udow
H. Wakeham
CONFIDENiIAL
SMOKING AND
SOCIOECQNOMIC STATUS
F. Resnik.
T. Osdene
R. G raham
R. Thomson
R. Fagan
F. Daylor
L. Meyer
J. Osmalov
T. Schori
B. Hancock
KEYWORDS: Socioeconomic Status, Smoking and Health,
POL Panels, Status Consistency
I
1
1
1
1
1
~
7
i
1
,
I
I
I
I
I
I
I
j
SUMMARY
A survey of the socioeconomic status of POL panelists
reveals that
(11) Lower class panelists smoke more and are much more
likely to be smokers than upper class panelists,
(2) Status inconsistent panelists (whose income or
education are inappropriate for their occupation)
smoke more and are more likely to smoke than are
panelists with consistent incomes, education and
occupations,
(3) Social class influences cigarette type smoked very
little, except for lower class tendencies to smoke
nonfilters and avoid health filters and 100 mm
brands.
The disproportionate distributfon of smokers among the
social classes may well account for some of the reported,
differences in psychological characteristics between smokers
and nonsmokers (e.g. lower acadlemi'c achievement and intelligence
test scores of smokers). Further, the distributiomcan even
mask situations in which smokers do better than nonsmokers so
that smokers will appear worse than nonsmokers.
Social class of respondents cannot be ignored in studying
smoking behavior.
j
2
I
1
1
1
~
7
1
1
.
I
I
I
I
I
I
I
I
I
Social status* is a recognized determinant of behavior
among societies which are obviously divided into sociali classes.
That such status differences exist in the USA is well known,
but is usually ignored by psychologists. There ap~pears to: be
some egalitarian wishful thinking among social scientists
regarding the effects of social' cl'ass on behavior. They would
like all people to be equali, so they treat their behavior as if
it were unrelated to demographic characteristics, while knowing
that it is not. For example, gender is the only demographic
variable taken into account by most psychologists. As a conse-
quence there has been a taboo built up among psychological
researchers, so that they tend to ignore the potential effects
of social status on their research. Yet many of these same
researchers will, in explai'ning behavior, suggest that environ-
mental conditions, social reinforcers, and a general history
of conditioning to make specific responses in the presence of
specifi'.c cues are the contributary factors in producing many
of the differences between people which we see every day.
"Each man is different," they glibly assert to colIege classes,
"because each has a different reinforcement history."
Bu~t rei'nforcement histories are not necessarily unique
for each individual. There are many stmilarities as well as
differences, and the similarities are often common within a
given social class yet different from class to class.
*"Social Status" is a measure of prestige. When we block people
of high prestige in the community into a class or group, the ~
group is described as a high social' class, and a group: of low 0
prestige people is described as a low social class. The terms ~
"'social status" and "prestige" can be used interchangeably. p
O
~
~
w
I
1
1;;
I
I
I
I
I
I
I
.1
I
I
j
3
For example,-amomg these similarities and differences are
such conditionis as: (1) Each social class has a d1lfferent
probability of consistently rewarding, punishing, or ignori'ing
the behaviors of its children and adult members, and (2) Each
social class has a different set of mores to be punished or
rewarded, a set which differs in part from the mores olf the
other classes.
To be more specific, the Kinsey reports have shown that the
sexual standards and behaviors of the classes d1i'ffer; even
casual observation shows that the classes have different like-
lihood of attending specific protestant churches (the funda-
mental.ist churches appeal to the lower classes, and the
Methodist and Episcopal to the upper classes); many consumer
studies reveal differences among classes in preference for
private brands vs. tradenames, for types of retail stores
frequented, and for concern with the appearance of their homes;
child development studies show the lower class is more apt to
spank its children, the middle class more apt to fol!low currently
popular child training techniques; etc.
A few recent studies have suggested that (as a consequence?)
there are personality test score differences between t;he social
classes, so that lower classes appear more anxious thaniupper
classes. As a corollary to this obs-ervation, we note that still
other investigations report that the incidence of poor mental
health is greatest.among the lower class. Both observations
suggest that members of th,e lower classes are ulnder greater
personal stress than are people in other classes. To the extent
that smoking is one of the avail!able strateqies people can adopt $-
to combat stress, we therefore would expect greater incidence of ~
O
smoking among the lower social classes. O
~
I
4
I
~
1
I
1
1
1
I
I
i
I
J
Thi's is not a particul'arly daring hypothesis, for others
have already reported the greater inci'dence of smoking among
the British lower economic classes, and we have noted it before
in our study of Greenfield, Iowa. To a considerable degree it
is therefore ex post facto theorizing.
In addition to expecting general effects of social class
on smoking, we also expect that we can classify some life
situations, as described by measures of social class, as more
stressful than others. Thus, for example, we feel', that a sales
engineer with a college degree and a 1968 income of over $15,000
lives a different life than does a man in the same occupationn
who lacks either this educational background or this income
level. The fi'rst man has income, occupation, and education
consistent with high prestige in the community. We call him
status consistent. By contrast, the second man is status
inconsistent, for either his income or education or both are
inappropriate to his job. We suggest that status inconsistency
places a man under extra life stresses, and that therefore
there should be a higher proportion of smokers among the status
inconsistent than among the status consistent.
1
Ratings of Social Class
In Warner's study of social relationships in a Connecticut
community, "Yankee City," interviewers found most adults able
to rate their fellow townsmen on social status with great inter-
judge reliabi.lity, using criteria of: occupation, kind and
amount of income, moral standing, geneology, social relationships
and' organizations and kind of residential area in which the
person lived. Given information of this kind,
peoplie to classes and examine the incidence of
in the classes.
we can assign
smoking behavi,or
I
I
5
1
I
1
I
I
I
I
I
I
I
I
The Present Study
We lacked some of the social information, but did know
the occupations, approximate income, and educational back-
ground of most of the panelists on the 1968 and 1972 POL
National Roster (see Table 1) so we examined these files for
possible social class effects on smoking, using the Census
Bureau procedures to guide us in defining socioeconomic status
and status consistency. The Bureau assigns a number to each
education and income level in the population based on the
cumulative percentage of the population which had attained that
income and education level. It then averages these numbers for
each occupation on its list and raniks the occupations according,
to this average. Using the number of people in each occupation,
it then assigns a number corresponding to the cumulative
percentage of people inithe rank ordered occupations to the
occupations themselves. Therefore, for any individual of known
education, occupation, and income, three numbers can be assigned.
With numbers assigned to each measure, the Census Bureau defines
status inconsistency in terms of differences between these
numbers whi,ch exceed a certain size. There are twelve types of
inconststency possible, six in which two components are con-
sistent and' one is high or low, and six in whi'ch,all three
components are inconsistent, with one being high, one inter-
mediate and one low. Our procedure was somewhat simi'lar. We
assigned the Census Bureau's occupation score to each panelist
of known occupati'on, and assigned an education and income number
to each panelist according to the information on his panel
appl i'cati'on form.
Then we made separate subtables for each educationiclass
showing the number of people with each occupation score for
each of seven income classes.
1
1
6
TABLE 1
I
I
1
~
1
1
1
~
I
I
(Information available on the
POL Panel for three measures)
Famfily
Occupation Education Income
Professional, Technical College Graduate
(High)
Manager, Official, Prop. Attended College
(Med. High)
Clerical and Sales High School Graduate
(Med.)
Foreman, Craftsman Some High School
(Med. Low)
Operative, Unskilled, Laborer Grade School or Less
(Low)
<3 ,-000
I
J
I
7
I
I
1
1
I
3
I
I
I
1
.1
An economist then examined the education, income, and
occupation listings and then assigned each set of scores to
a status consistent or status inconsistent group according
to his knowledge of demograpihiilcs. Although assignment to
groups was somewhat arbitrary, the decilsions were made on
logical grounds. To illustrate, among the college trained,
all who were unskillled laborers with low incomes were clas-
sified as status inconsistent with education too high for
their occupation and income. Some of the college trained
reported blue collar occupations but high incomes, these were
classified as status inconsistent with occupation being low.
Similar, equally defensible cliassilfications were made for all
panelists without knowledige of their smoking habits. Classi-
fication of smokers and nonsmokers had' been made by panelists
themselves on the original questionnaires.
Although this procedure for determining status consistency
is less precise than that of the Census Bureau, it is unprej-
udiced and unbiasediwith respect to smoking behavior.
RESULTS
Among the 8,605 men aged 25-59 on our 1968 panel we
identified'5,2'69 as status consistent (61.3q) and 3,336 as
status inconsistent (38.7%). Among the status consistent
men there were 2,860 smokers (541.3q).* This average figure ~
O
is somewhat mislead'ing, for there is a strong,occupational ©
class effect (see Table 2), the proportion of smokers being p
~
l greater among working class occupations. ` ~
Gb
1
*
'
I
I
57.7% of the panel were smokers. Compare 1970 WEW
estimates
of 47.8% smokers in the general population of men aged 26-64.
At the Research Center our SEF files indicate that 48.7% of our
personnel are smokers (men and women, all ages) a figure which
is probably an overestimate, nonsmokers being less likely to
return our local questionnaires.
~
I
1
8
TABLE 2
Percentage of Smokers and Average Number of
Cigarettes per"Da yamong'Status Consistent M'aIe
Members of the Fiye,Occupation Classes
% Panel Avg. Number
% Cigarettes
N Smokers Per Day
1 29.2 Professional,-Technical -1537 43.4 25.4
15,3 Manager, Official, Prop. 806 51.0 25.0
18.0 Clerical and Sales 949 56.9 26.2
1 21.6 Foremen, Craftsmen 1140 58.2 25.7
15.9 Operatives, Unskilled
I Laborers 837 69.1 26.2
Total ~
5269 54.3
1
I
3
I
I
I
I
I
I
I
j
There appear to be no systematic differences in consumption
among,the classes, suggesting that the probability of a man
becoming an active smoker depends omhis socioeconomic class,
but that once he becomes a smoker his consumption pattern is
not affected by his occupation. (This latter statement is a
tenuous one, for we have our doubts about the measure, reported
daily consumption, which is not a very accurate estimate accord-
ing to our SEX-i study.)
Among the status consiIstent smokers there were some product
use differences (Table 3). The 70 mm nom-filter brands were
smoked by 1/4 of the lower class smokers and by only 1/16 of the
upper class smokers. Almost the reverse relationship held true
for the so called "health filter," here defined as nonmenthol
filter cigarettes with delivery under 16 mg FTC tar. About 21%
I
9'
1
I
1
1
~
1
1
I
of the upper class and 8% of the lower class smoked "health
filter" cigarettes. The frequency of use of other types of
product (including 85 mm menthol) was not markedly different
from class to class, with the possible exception of lower
class low usage of regular 100 mm brands.
TABLE 3
Percentage of Status Consistent Smokers
wi thii n the- Ffive"CTasses'Cdho Smoke Di fferent
Product Types
85 5 mm
70 mm 85 mm 100 mm 85 mm "Health
Nonfilter -Reg.. Filter Reg. Filter Menthol Filter"
Professional, Technical 6.7 31.0 7.9 16.3 21.1
~ nager, Official, Prop. 11.4 32..4 10.9 13.6 16.5
~ Clerical & Sales 11.1 30.7 8.5 17.2 16.3
Foremen, Craftsmen 12.7 32.2 7.1 17.3 11.3
~ Operatives, Unsi'lled 25.4 29.6 4.7 15.6 7.6
I
I
I
I
I
I
J
All 13.4 31.2 7.6 16.2 14.5
Status Inconsistent
Among the 3,336 status inconsistent panelists there were
2,106 smokers (63.1%), a higher percentage thanithe 54.3% among,
the 2,409 status consistent smokers. Among the 12 status incon-
sistent groups there were some subgroups with so few members
that it made sense to combine the subgroups to get more meaningful
data. Accordingly we have combined and recombined groups as shown
in Table 4. It should be noted that some smokers appear twice in
this table. For example, there were 11 smokers who had Hiigh
Education, Moderate Occupations, and Low Income - they appear
twice in the table, once as members of the High Education group,
once as members of the Low Income group.
j
I
10
TABLE 4
Percentage and Number of Status Inconsistent
Smokers withiin the Six
Combined Inconsi,stent Classes
No. of
Class N Smokers % Smokers
Income High 1,044 ~ 680
(plus Inc. Hi - 0cc. Lo 1119 82
and Inc. Hi - Ed. Lo) 99 69
Total 1,262 831 65.8
Educati on Hi'gh 331 204
(plus Ed. Hi - 0cc. Lo 27 17
and Ed. Hi - Inc. Lo) 11 5
Tota1 369 226 61.2
I
I
I
I
I
I
I
J
Occupation High 96 71
(plus 0cc. Hi - Iinc. Lo, 4 2
and 0cc. Hi - Ed~. Lo 13 1 1
Total 113 84 74.3
Income Low 289 174
(plus 0cc. Hli - Inc. Lo 4 2
and Ed. Hi - Inc. Lo) 11 5
Total 304 181 59.6
Education Low 595 365
(plus 0cc. H,i - Ed. Lo 13 11
and Inc. Hi - Ed. Lo 99 69
Total 707 445 62.9
Occupation Low 708 426
(plus Ed'. Hi - 0cc. Lo 27 17
and Inc. Hi - 0cc. Lo 119 82
Total 854 525 61.5
I
11
I
I
1
~
7
I
I
.
l
I
I
I
I
I
I
I
I
As a general rule these status inconsistent subgroups
have a higher percentage of smokers than do the top four
(status consistent) occupation classes of Table 2. The
greatest proportion of smokers in any status consistent or
inconsistent group is in the status inconsistent group with
higher occupations than their Income or education would
predict, a whopping 74.3%.
Comparing reported consumption of cigarettes, where
status consistent smokers averaged 25.7 cigarettes a day,
the status inconsistent smokers averaged 26.8 cigarettes a
day, over a cigarette more. The greatest reported daily
consumption among any of the status consistent or status
inconsistent subgroups was among those with low education.
The five groups with h1ghest consumption rates were all
.status inconsistent:
Rank Number
Cig,ts./Day
Education Low 1 29.1
Income High, Education Low 2 27.5
Income High 3 26.9
Education High, Occupation Low 4 26.6
Income Low 5 26.4
In Table 5, where c1garette type data are shown for each
of the status inconsistency subgroups we see that there are few
differences in type preferences between groups. Imgeneral all
the status inconsistent groups had preferences for nonfilter
and' health filters which more cl!osely resembled the two lower ~
status consistent groups than the hiqher status consistent ~
O
groups. No other differences are apparent. O
. ~
GD
I
I
12
1
I
1
3
~
1
1
I
ii
I
I
I
I
I
~
TABLE 5
Percentage of Status Inconsistent Smokers
(Six Typ)es)--Who Smoke Different
Product Types
85 mm
70 mm 85 mm 100 mm 85 mm Health
Class Nonfilter Reg: Filter Ref. Filter Menthol Filter
Income High 20.5 29.0 6.1 14.8 12.4
Education H'igh 17.3 25.7 11.9 15.5 11.1
Occupation High (21.4) (28.6) (6.0) (13.1) (4.8)
Income Low 17.1 35.9 8.8 14.4 7.2
Education Low 21.3 27.0 7.6 15.1 13.9
Occupation Low 14.9 31.8 7.0 16.6 13.0
Total 18.6 29.6 7.5 15.3 12.2
NOTE: Parenthesized entrfes are based on a small N.
I
13
1
1
1
~
7
1
1
I
TI
I
I
J
It is not clear why the status inconsistent* shoul'd
smoke sol much like the status consistent lower classes: They
may be under greater stress; they may have similar personalities;
they may have lower class origins; their inconsistency of status
may affect their social contacts; their social mobility may be
involved. Age may be a factor. We can only speculate.
The number of male panelists from age 25 to 59 is shown in
Table 6 for each 5-year age grouping and six occupational
classes, ranging from low to high. For conventence we will
describe these six groups as Lower Low, Upper Low, Lower Medium,
Upper Medium!, Lower High, and Upper High.
The continued inflation of incomes during the late '60's
and early '70's had its effects on our panelists and on the
criteria we had employed to define status consistency and incon-
sistency in '68.
We were unable to: send test products to some panelists
between 1968 and 1972, notably to nonsmokers and non-filter
smokers. Because of this fact, combined with the lack of a
continuous recruitment program, the anti-smokinig crusade, the
boredom of some panelists who received a number of cigarette
tests but nothing else, and the inability of the post office
to track people who had moved, many of our '68 panelists did
not respond to our '72 repolling.**
*It must be noted that the income measures are famil incomes.
When husband and wife both work, this will tenTto produce a
status inconsistent cl'assification with income high. Since
workiing wives are more common among the lower classes, some
of our status inconsistent smokers would have been classiffed
as status consistent lower class men if only their own incomes
were taken into account. This gives riise to an interesting
hypothesis: men with workiinq wi ves are more likely to smoke.
**The panel has subsequently been supplemented, and'~a continuous
recruitment proaram initiated.
14
I
I
I
1
a
1
I
I
I
I
I
I
J
TABLE 6
Some Representative Occupations
% of
% of POL Panel Occupation
U.S. Pop. Over 25 Group Scores
Occupations
5.4 16.0 Upper 90-99 Architect, Lawyer, Chemist,
Hi~gh Manager of Department Store,
Professor
7.7 20.9 Lower 80-89 Draftsman, Social Worker,
High Mortician, Postmaster,
Secretary, Teacher
11.1 25.3 Upper 70-79 Policeman, Typesetter, Bank.
Medium Teller, Manager of Auto
Repair Service, Nurse, Tool
and Diemaker
13.7 13.2 Lower 60-69 Electrician, Construction
Medium, Foreman, Bus Driver, Plumber,
Machinist
35.0 1'3.3 Upper 35-59 Brick Mason, Carpenter, Auto
Low Mechanic, House Painter,
Operatives in Steel Mill or
Brewery, Barber, Waiter
27.1' 11.2 Lower 1-34 Shoemaker, Sawyer, Coal
Low Miner, Operative in Tobacco
Industry, Laundress, Truck
Driver's Helper, Fisherman,
Sexton
15
1
I
1
1
0
I
I
I
I
I
I
I
J
Among the reduced panel, there were (a) a higher percentage
of smokers in 1972 than in 1968, 62.0% compared with 57.7%; and
(b) relatively more status inconsistent panelists, 4:7.0% compared
with 38.7%;
~
smokers
90
80
70
60
5 &
40:
.--------------~
consistent low
status inconsistent
consistent medium low
consistent medium
consistent medi'um high
consistent htgh
68 72
pane-Y panel
~
Fig. 1 - Changes in percentage smokers among the '68 and
'72 panels.
Fig. 1 shows that these trends are alike from group to group.
They are not to be interpreted as an indication of increased'
incidence in smoking in the general population over this period.
We attribute the i'ncreased incidence to the differential attrition
among smokers and nonsmokers.
Look.ing at the proportion of the panel in the different
status consistency groups (Table 7), we can see that most of
the fncrease in the proportion of the panel which is status
Inconsistent has come from decreases in the proportion of
panelists who are status consistent in the two lower classes.
j
16
TABLE 7
1
1
I
m3
7
I
I
I
I
I
Percent of Panel in Each Group in '68 and '72
.. J..,4Y~+ .
1968 1972 Change
Consistent High 17.9 18.7 + .8
Consistent High Medium 9.4 8.0 -1.4
Consistent Medium 11.0 12.3 +1.3
Consistent Medium Low 13.2 8.3 -4.9
Consistent Low 9.7 5.7 -4.0
Inconsistent 38.8 47.0 8.2
We expected therefore, to find an increase in the propor-
tion of smokers who were status inconsistent with income high,
for they would have come principally from the two lower classes.
Actually the proportion of smokers was an identical 65.8% among
the income high panelists in '68 and '72.
DISCUSSION
To illustrate the i'mportance of social class, let us
conduct an imaginary research project In which we compare the
intel'ligence test scores of smokers and nonsmokers in a hypo-
thettcal population of 100 people. Let there be no within-
class d1fferences in test sc res, but let there be between
class differences in both test scores and proportion of smokers.
Then we might find the following:
No. of Intelligence No. of No. of
Social Class PeopIe Test Score Nonsmokers Smokers
Upper 10 11120 8 2
Middle 30 105 21' 9
Lower 60 95 20 40
Total 100 49 51
J
.
17
1
1
1
3
7
I
I
I
I
I
I
J
I
J
The average test score of the 100 people is 100.5.
Ignoring Social Class, the average test score of nonsmokers
is 103.4, and the average test score of smokers is 97.7*.
The naive experimenter who ignores social class and measures
only intelligence test scores of the 51 smokers and' the 49
nonsmokers will therefore conlude "that smokers have lower
intelligence test scores than nonsmokers." This is true, but
it is a nonsense statement. It implies that smoking is some-
how related to intelligence test score. Within each social
class however each smoker has exactly the same score as each
nonsmoker. What the naive researcher is seei'ng is the
confounding effects of different soci'al class intelligence
scores and different proportions of smokers.
As a second hypothetical example, suppose that some
_index of poor health is related to social cliass in such a
way that the well-paid, well-edlucated white collar workers
enjoy better healthithan poorly paid, poorly educated'i laborers.
Then, because the incidence of smoking differs between the
social classes, we would find our research literature filled
with observations suggesting that smoking is related to poor
health. The literature does show this, and it may be wrong.
At least part of the reported statistical relationships
between health variables and iincidence of smoking can probably
be accounted for fin this fashiion.
The smoking and health relationships may be at
part due to social class differences rather than to
It must be cautioned that this observati,on does
smoking is independent of health problems. Consider
lieast i n
smoking per se $
O
O
not mean 0
for example, ~
C
*103.4 =
[(8x120)
+ (21x105) + (2'0x95)]/49 D
97.7 =
[(2x120)
+ (9x105) + (40x95)]/51 m
J
1
I
1
>
7
1
I
1
I
I
I
I
I
I
4
18
again using intelli'gence test scores as our example,
what the tabled data would'' look like if there were a two-
point difference in hypothetical intelligence test scores
for all smokers and nonsmokers in a given social cl!ass of
a population:
Social Class
No. of
Peopl.e
Intelligence
Test-Score
Nonsmokers
Smokers
Upper 10 121 8
119 2
Mi dd'1 e 30 106 21
104 9
Lower 60 96 20
94 40
100 49 51
104.3 96.7
Now the average nonsmoker has a test score of 104.3 and the
average smoker a score of 96.7,* a dAfference of 7.6 points
compared to the 5.7-point difference calculated earliier.
In practice, whenever there is an observed difference
reported in scores for "'smokers" and "nonsmokers" we can't
tell whether the difference is real unless we take out the
contributory effects of social class, or of variables related
to social class.
Incidental Note
We do have some evidence that, in the population as a
whole, there are IQ, differences related to social class. We
can combine this information with the data iin Tables 2 and the
census figures to come up with some hypothetical data for our
-$
panel: O
O
G
*104.3 = [(8x121) + (21x106) + (20x96)]/49
C1~
96.7 = [(2xll9) + (9x104) + (40x94)]/51 Cb
J
, 19
~
1 Number in
% of Our Est. % Non- 1,000 Panelists
1 Population Class _ID Smokers Smokers Nonsmokers Smokers
' 29.2 Professional 120 56.6 43.4 165 127
~ 15.3 Manager 110 49.0 51.0 75 78
18.0 Clerical Sales 105 4,3.1 56.9 78 102
21.6 Foremen,Craftsmen 105 41.8 58.2 90 126
~ 15.9 Operatives, labor 92 30.9 69.1 49 110
457 543
1
I
I
I
I
I
J
j
I
We'd expect a random~1,000 people selected from the panel to
have 457 nonsmokers and 543' smokers, distributed as shown,
which would yield estimated IQ scores of 111.5 for nonsmokers
and 108.3* for smokers if social class were ignored. Again
within each social class, there are truly no differences in IQ
related to smoking,and nonsmoking.
It must also be noted that social class differences can
serve to mask differences between smokers. Consider a popula-
tion divided into three groups, in each of which smokers score
two points hi'gher than nonsmokers, with a 5-poinit difference
between groups.
No. of Intelligence No. of No. of
Social Class People Test Score Nonsmokers Smokers
Upper 10 104 8
106 2
Middle 30 99 21
101 9
Lower 60 94 20
96 40
100 49 51
97.8 97.3
*111.5 = [(165x120) + (75x110) + (78x105) + (90x1.05) + (49x102')]/457
108.3 = [(127x120) + (78xllO) + (102'x105) + (126x105) + (110x92)]/543
J
20
1
I
I
1
.
0
I
I
I
I
I
i
I
I
I
Here the data suggest that nonsmokers have slightly higher
test scores (97.8) than smokers (97.3)* where in actual fact
the smokers have a 2-point higher score in each of the social
classes. The naive investigator would fail to see the benefits
of smoking, and instead conclude that there was either no
difference or that smokers were the poorer.
Further examples would only belabor the poi'nt unnecessarily.
We must take socioeconomic class into account in our future
research, and we must look for its confounding effects on the
research conducted by others.
k
0
©
*97.8 = [(8xT04) + (2lx99) + (20x94)]/49 O
97.3 = [(2x106) + (9x1D1 ) + (40x96)]i/51
°
