Professional Documents
Culture Documents
A Scale For Assessing Socio-Economic Status in Survey Research
A Scale For Assessing Socio-Economic Status in Survey Research
Survey Research
This article contains the description of a socio-economic scale based on the answers
to five simple questions. The scale has been found useful in urban and rural areas
and is particularly adapted to survey research.
Charles L. Vaughn is in the Marketing and Social Research Division of the Psychological Corporation.
COMMON METHOD USED in assessing socio-economic status of respondent
homes in survey research is to obtain interviewer ratings of the variable.
These ratings are generally based upon a quick inspection of the dwelling
unit in which the person is conducting the interview and upon the personal
care and speech of the respondent The interviewer may or may not have
advance knowledge of the general characteristics of the area in which he or
she is working. If the interviewer is approaching people on the street, the
personal care and speech of the respondents are almost the sole guides he or
she has for making the ratings.
Although some of the largest differences in survey results are revealed in
the analyses by socio-economic groups, ratings suffer a number of limitations,
of which the following would appear to be the principal ones: 1
B Y CHARLES L . VAUGHN
20
There are also some advantages in ratings, viz: they are simple to process
and are quickly made in the field, if interviewer procedure is not carefully
controlled. They also must have some validity, as evidenced by the large differences that are customarily revealed by analyses in terms of the ratings.
The initial stimulus for the development of the present scale was the need
for a device that would provide reliable and valid socio-economic groupings
of people interviewed on the street in small pilot research studies.8
Conclusion of this project was prompted by a shift from assigning interviewer quotas by socio-economic group to the making of interviewer assignments in terms of small pre-selectcd areas, or "clusters," and by specific routes
within those clusters. In this latter type of assignment, interviewer ratings
tended to become non-discriminative.
21
RESEARCH PLAN
of the household. These reports, however, were not substituted for ratings
because "household income" reflects less of the "social aspect of socio-economic
status" than desired. There arc also other objections to use of household income, viz: questioning on the subject tends to irritate respondents and thus
to demoralize interviewers. Family income also fluctuates widely and relatively few persons in the household can report reliably on the subject. Another
solution that was considered was to spread interviewer ratings over twelve
rather than four groups, but this procedure still resulted in poor discrimination.
22
The Criterion. The criteria against which the answers to the questions
were scaled, were interviewers' ratings of the socio-economic status of the
households in which respondents resided. These ratings were along a threepoint scale, upper, middle, and lower socio-economic group. Assignments in
a given town or city were made in such a way that 30 per cent of the ratings
were "upper"; 40 per cent "middle"; and 30 per cent "lower". The raters
were thus forced to discriminate.
was believed that Questions 2 and 3 would differentiate along the entire
range; Questions 4 and 5, at the bottom end of the scale; etc
One might assume that Questions 2 and 3 would be sufficient, and this assumption is correct, except for the fact that more than two questions were
necessary to provide relatively fine discriminations. The questions on number of bedrooms and number of people in the home (Q's 7a and 7b) were
included to obtain an index of crowding, but the questions met with so
much respondent resistance that they were later dropped. The index of
crowding also posed difficulties in processing and was not very discriminating.
The Sample. The national sample in which the questions were originally
included represented all urban and small town (rural non-farm) households
in the United States. Interviews were made in 123 localities, i.e. standard
metropolitan areas or non-metropolitan counties, and in 273 cities, towns, and
villages within those areas.
Within cities, towns, and villages, interviewers were assigned to spots on
the basis of socio-economic maps of the places, prepared by field supervisors,
or Research Associates of The Psychological Corporation. These maps showed
the respective parts of the locality where the upper, middle, and lower socioeconomic groups resided. Because of variations in socio-economic status, even
within small areas in a city, interviewers were assigned a specific quota of
interviewsi.e. a certain number of interviews with persons in the upper
group, in the middle group, and in the bottom group. The interviewer had
to use judgment in selecting households within these three groups, and rated
the homes as he interviewed.
Only one person was interviewed per household. The individuals interviewed in the households were chosen to match the distribution of individuals aged eight years and older in the United States population as shown by
the Census of 1950. If a person was not at home in a household, a substitute
was taken, either in that household or in the next household. So that persons
who are frequently away from home would be given some weight in the
sample, a small percentage of interviews were made in shopping districts near
the areas assigned. The interviewing was supervised locally by 120 psychologists serving as Research Associates of The Psychological Corporation.
23
092
341
567
1.000
The use of interviewer ratings as the criterion may, of course, raise some
questions. However, these ratings, based in part upon socio-economic maps
of the respective interviewing localities, were the basis of breakdowns used
in previous studies, and differences in answers between broad socio-economic
groups were generally quite significant. One aim of the present research was
to develop a scale that would discriminate on much the same basis as the
interviewer ratings had prior to the change in sampling methods, so that
trend comparisons would not be jeopardized by artifacts.
Also, it seemed probable that the persons best able to distinguish between
broad groups locally were the interviewers who called upon the homes and
talked to the people therein even though differentiation in the middle ranges
might not be a fine one. At least the intelligent field interviewer, supervised
by a local resident at the professional level and assisted by a socio-economic
map, would appear to be a better judge of the status of homes than would
someone basing this decision upon an a priori and theoretical conception of
what constitutes social status.
As a basis for scaling the answers to the five questions and parts thereof,
the assumption was made that socio-economic status is normally distributed.
This assumption can not be grossly in error. All it means is that some households have a very high status and others a very low status, but most fall in
a large middle group, those above average being distributed along a bellshaped curve in about the same fashion as those below average. Certainly the
normal curve approximates the underlying distribution of socio-economic
status, and the ready availability of tables of areas under the normal curve
facilitates scaling and statistical analysis.
Scaling Against Criterion.The detailed scaling procedure will be illustrated below for the occupation of the "Chief Wage Earner". The other
answers were similarly scaled.
Respondents' reports of occupation were classified into the standard Census categories: "Proprietors, managers, and officials except farm;" "professional workers, semi-professional workers;" "salesmen"; etc These were then
cross analyzed by socio-economic rating. To illustrate, details of the
procedure for "proprietors, managers, and officials except farm" will be
shown,
24
Equation 1.
Equation 2:
_c (X - Mv)
a,
(O
(X - M,)
a,
M^
<T,
a. (Z - M, , (Z - M,)
Mv
.
...
E
*- +
= -*- , in which
a*
a,
at
-
Wt,)
(X-
It was assumed that some of this occupational group would have a relatively high socio-economic status, some a relatively low one, but that the
group would tend to "peak" at a mid-point along the socio-economic continuum.
Knowing the fractions of proprietors, managers, etc who fell in the upper,
middle, and lower group respectively, one may compute the distance of the
mean of this group (Mp) from the mean of the distribution (Mt). This distance may, in turn be converted into standard deviation units of the larger
distribution (<*t = 1)> by solving these simultaneous equations:
25
EXAMPLE:
Fraction of "proprietors, managers,
officials, etc." in:
Obtained by reference
to tables:
^ = .17
. 5 2 + .52 _ 1.04 _
- . 1 7 + 1.33
1.16
= .68,
Table 1 shows the scaled, or sigma( o), scores of the five items togedier
with the field scores. The field scores, which are integers corresponding to
the scaled scores, are recorded by the interviewers. They add these scores for
the five items to obtain the total socio-economic scaled score.
26
1
m
_2_
3
_
The total socio-economic scaled score for the household is 13.
Thirteen is the maximum score that a household can make. The lowest
score would be made by a household with: no telephone (score = 0);
renters occupying it (score = 1); no car (score = 1); the chief wage earner
a laborer (score = 0 ) ; and with the chief wage earner a person who left
school after finishing grammar school (score = 0). The total score for this
household at the lowest level would be two. This range in scores from 2 to 13
facilitates IBM processing. All scores except 13 are punched as they are. Scores
of "13" are converted to "l's" during the inspection process.
In actual practice, the five basic questions are printed on one side of a
5 " x 8" card with the field scores corresponding to the response alternatives
thereon. Only the response record for numbers appears on the questionnaire
blank. Since the office coding of occupation is time consuming and tedious,
illustrative occupations and scores therefore are printed on the second side
of the card containing the questions, so that the interviewer records merely
the number corresponding to the occupational level.
DISTRIBUTION OF SCALED SCORES
In assigning field scores to the respective answers, we have grouped together the answers that were no more than 05* apart in terms of the scaled
scores. This procedure simplifies scoring in the field and IBM processing in
the office. Since in actual practice there will always be some "don't know" or
"no answer" responses we have scaled these alternatives along with the others
as shown in Table 1.
An example of how the score is obtained in the field may be helpful.
Suppose the respondent reports that: (k) the home has a private line telephone; (1) the family owns the home; (m) there are two or more cars in
the household; (n) the chief wage earner is a drug store owner; and (o) the
chief wage earner graduated from college. The interviewer records the scores
as follows:
k
_2_
u.
10
Ul
SAMPl
PER CEf
IS
Study
1 ,
9
10
II
12
13
14
a - 2.8
zo
J8
No phone
.00 to +.49
- . 5 0 to - . 9 9
Telephone
Ownership
Private line
Don't know
no ans.
Party line
.00 to +.49
+.50 to +.99
Scaled
Scores
Rent home,
no ans.
Own home
Home
Ownership
Professional
& semi-prof,
workers;
proprietors,
managers, eta
Salesmen;
clerical workers
Craftsmen,
formen, misc.;
retired, unemployed; don't
know; no answer;
protective workers,
operatives
Other svce.
workers; domestic
service wkrs.
laborers.
Occupation
(Hd. Hsehld.)
Two or more
cars or one
car bought new
Car
Ownership
TABLE 1
Some college;
Completed HS
Completed
College
Education
(Hd. Hsehld.)
Field
Scores
>
29
Interviews were made in 99 localities (Standard Metropolitan Areas and nonmetropolitan counties) and in 259 urban places and rural villages within
those localities.
All counties in the United States were stratified with respect to metropolitan vs. non-metropolitan counties, geographic area, population, per family
income, per cent engaged in agriculture, and per cent white in the South.
Interviewers were assigned to work in small areas or clustersusually a
city block or the equivalent thereofand were given definite routes to follow
in the respective "clusters". There were 400 clusters in ah1.
Only one interview was made per household. Approximately one-third of
the adults and somewhat more than a half of respondents 10-20 years of age
were interviewed in the shopping district or playground nearest the cluster.
Besides the distribution of scaled scores for all households, Illustration I
also shows the distributions for the subsamples of urban and rural village
households, respectively.
The sample of 8,000 households is essentially the same as that upon which
the various items were scaled, the only difference being that all localities,
urban places, rural villages and clusters in the present sample were randomly
selected within the respective strata. Such was not the case in the sample
used as the base for scaling.
Study # 2 . This study was based upon interviews in 10,000 households,
representing all urban households in the United States. Selection of localities,
interviewing spots, etc. was made in a similar fashion to that in Study #1,
except that there were 1,000 such spots, or clusters, rather than 400, and
interviewing was conducted in 364 urban places and fringe areas in 91
localities.
All interviews in the Barometer were made with the "man of the house"
or the "woman of the house"one interview per householdand all interviews were made in the home.
Illustration I also shows a normal curve with mean and standard deviation the same as those for the distribution of socio-economic scores from the
8,000 household sample of Study # 1 .
Discussion. The scale discriminates. The urban samples yield about the
same distributions. The distribution of scores for the sample of households
in rural villages was somewhat lower than that for the urban samples, but
not greatly so.
The distributions approximate the normal curve shown in Illustration I.
However, there is some skewness: relatively more low scores were obtained
in Study # 1 than would be expected if the distribution were normal. This
situation is most likely due to the fact that interviewing spots in this 1955
study were chosen by the use of tables of random numbers; interviewing spots
in the 1951, or original "scaling," study were chosen on a judgmental basis;
and judgment samples are customarily somewhat higher on the socioeconomic ladder than are samples from which interviewer judgment has been
largely removed. The small divergence of the distributions from normal,
however, does not seriously limit the usefulness of the scale.
TABLE 2
Socio-economic
Score
Suggested Letter
Grade Equivalent
and Description
Study fl
Study f2
Urban & Rural Village Urban Only
13
A Top
(15%)
3.6
2.7
12
5.1
4.6
11
6.5
6.7
10
8.4
9.1
9
B Upper (31%)
10.0
10.5
8
11.6
11.6
Middle
7
12.9
12.6
6
13.3
12.5
C Lower (36%)
5
11.2
10.8
Middle
4
8.6
8.8
3
6.1
D Bottom (18%)
6.0
2
2.9
3.9
100.0
100.0
(100%)
Total
Numbers in parentheses are approximate per cents achieving each suggested letter
grade.
RELIABILITY AND VALIDITY OF THE SCALE
31
Subsequent to preparation of the foregoing article the scale has been applied to a sample of 447 farm homes. This sample was a cross-section of all
farm homes in the United States and was obtained as a supplement to the
May 1957 Psychological Barometer.
No particular problems were encountered in using the scale this first time
in farm homes, and it does discriminate between homes in this group, as
shown by Table 3. On the surface, at least, the scale would appear to be as
valid with farm as with urban and rural non-farm homes.
SUMMARY AND CONCLUSIONS
This report describes a scale originally designed for assessing the socioeconomic status of homes in urban territory and rural villages in the United
States. Data obtained subsequent to construction of the scale would indicate
that it is also useful with farm homes.
The scale value assigned to a home depends upon the answers of anyone
from that home to five basic questions and parts thereof. The questions pertain to: telephone "ownership"; home ownership; automobile ownership;
occupation of chief wage earner in the home; and his (or her) education.
The numerical value assigned to each answer has been determined by
scaling the respective answers against interviewers' ratings of homes in urban
territory and rural villages in the United States. These ratings were based in
32
Achieving Bach
Score
13
12
11
10
9
8
7
6
5
4
3
2
.8
1.8
8.9
8.3
Total
16.1
9.4
14.3
13.6
S.I
10.3
6.7
4.7
100.0
Socio-Economic
Scare
33
With the relatively fine discrimination afforded by the scale, other groupings
may be used. Those above appear to be convenient, and the breaking points
loosely correspond with significant parameters of the normal curve, i.e. the
two inflection points and the mean or mode.
Use of all twelve points on the scale, however, makes it relatively simple
to derive mathematical expressions of various relationshipssuch as those
between television set ownership, incidence of Negroes, use of certain brands
of products, knowledge about various companies, etc. and socio-economic
level. Whether or not the mathematical expression of these relationships
would have any practical value is open to question, although the application
of operational research techniques to problems in marketing, public relation,
and communication might be facilitated thereby.
It would also be a comparatively simple matter to ascertain with a fairly
high degree of precision the specific channels of communication that can most
effectively and efficiently be used to reach homes in which the members are
relatively ignorant on a given topic, those who do not use a given brand of
product, etc.without relying too heavily on judgment and without repeating a media study for every brand of product and on every public relations topic.
The scale can also be used as a basis for comparing one community with
others across the nation in terms of: (a) its average cultural position and
(b) its variability in cultural status. This topic assumes significance in one
way when the attempt is made to apply a company-wide, or national, community relations program to a particular community. If the community is
definitely more homogeneous than the cross-section of communities and is
below average culturally, the media and specific appeals required there may
be quite different from those generally applicable. On the other hand, the
uniqueness of a community may be more apparent than real. The use of
objective measuring instruments is necessary to obtain a valid picture of
what the situation actually is.
34
The reader will undoubtedly see many research problems associated with
the use of a scale of t-his nature. A topic, which we are now investigating,
is the variability in cultural level within neighborhoods vs. that between
neighborhoods. Of potentially the greatest interest, however, is probably the
subject of what specific communication channels and appeals can be used the
most efficiently to reach homes at the various cultural levels.
Downloaded from http://poq.oxfordjournals.org/ at AAPOR Institutional Member on August 5, 2014