Sex Roles and Performance: an Empirical Study of Gasoline Purchasing

Kathy Pettit, University of Washington
J. K. Johansson, University of Washington
ABSTRACT - An analysis of panel data on gasoline purchases shows few significant differences between male & female buyer behaviors. A small price differential is shown to relate to differences in perceived task difficulty and effort.
[ to cite ]:
Kathy Pettit and J. K. Johansson (1980) ,"Sex Roles and Performance: an Empirical Study of Gasoline Purchasing", in NA - Advances in Consumer Research Volume 07, eds. Jerry C. Olson, Ann Abor, MI : Association for Consumer Research, Pages: 784-787.

Advances in Consumer Research Volume 7, 1980     Pages 784-787


Kathy Pettit, University of Washington

J. K. Johansson, University of Washington

[The authors want to acknowledge the contributions of Mary Lou Merrill in the early stages of this research.]


An analysis of panel data on gasoline purchases shows few significant differences between male & female buyer behaviors. A small price differential is shown to relate to differences in perceived task difficulty and effort.


An interesting question continually surfacing in the women's liberation debate is whether women placed in the same roles as men will perform as effectively as males. There are at least two sides to the argument. On the one hand, it is argued that when the basic background characteristics (education, experience, etc.) have been accounted for, there is no "residual" effect from the variable sex as such. On the other hand, it is sometimes argued that the very fact that women are placed in traditionally male roles will have a (beneficial) effect by itself, not only because of the example it provides for other women but also because women will bring something unique to these new roles. According to this latter argument, once underlying factors have been equalized, there would indeed still remain some residual differences in performance attributable to sex.

The present paper attempts to come to grips with these issues by identifying whether females and males in the same role perform similarly, or whether they do, in fact, exhibit any different characteristics. The setting is gasoline purchases for the family automobile(s) and the role is the one of principal gasoline purchaser. Although this context is decidedly of less import than, say, career choice, the basic problems of consumer role behaviors versus sex roles is in evidence here. Furthermore, there is at least one other advantageous aspect of the product in question. Gasoline is a fairly homogeneous product. It is used for the same purpose by both sexes, conspicuous quality differences are absent, and the basic information about the product needed to purchase is rather minimal. As a consequence, one would expect the product specific factors to play a relatively small role in the explanation of performance, and gasoline purchasing might perhaps provide a compact microcosm of some issues of much wider concern.

Before going further, a caveat should be expressed. This study was prompted by the availability of some interesting panel data on gasoline purchases. As a consequence, it is best seen as exploratory, with the simple aim of investigating what difference an individual's sex seems to make as evidenced in this particular set of empirical observations. There are theoretical relationships which can be--and will be--used to disentangle what is going on in the data. However, whether or not the specific explanation for the performance differences uncovered is the true one is difficult to verify without further hypothesis testing research. Fortunately, the data base is rich enough that a few competing explanations can, in fact, be ruled out, and the resulting explanations proposed are shown to be consistent with the data.

The exploratory nature of the study makes it most natural to start with a discussion of the data, followed by an empirical test to see if in fact there are differences in behaviors and performance. The discussion then formalizes a model based on some theoretical relationships between performance and underlying causal factors. The model is used to pinpoint the causal factors where in fact there are differences between the two sexes, in an attempt to answer why sex makes a difference.


The data came from a year-long panel on gasoline purchases, maintained between September 1972 and September 1973. The panel members were families selected from among the pool of subjects maintained by National Family Opinion (NFO) Inc., which also carried out the study for one of its clients. The sampling procedure was basically of a quota type. The first requirement for membership was location in one of the client's main markets, a major Midwest metropolitan area. Further, the respondents were screened for car ownership, with 50% of the sample possessing two or more cars. The total sample size was 500 families.

Respondents were paid for their participation by NFO. They did not know the identity of the sponsor. The families were given a diary to be kept under the visor in each car. Instructions were that whoever made the purchase of gasoline should record the requisite information. The diaries were collected every two weeks and new diaries issued. Information collected included quantity and brand bought, price paid, grade of gasoline, whether the station was regularly patronized, and whether there was any kind of deal offered.

In addition to the diary data, the NFO records included demographic and socio-economic characteristics of the panel families. Furthermore, during the last month of the panel a six-page questionnaire including lifestyle (AIO) measures as well as attitude measure towards brands and stations was administered to the principal gasoline buyer in the household.

The role of principal buyer of gasoline in the household was recorded in the data on the basis of self-reports. The respondent first reached was asked who was the principal buyer in the household, and the person identified was then asked to complete the rest of the AIO questionnaire.

For the present study, members with less than ten gasoline purchases, dropouts, and those with incomplete records were omitted. The analysis is restricted to the principal car in the family. Although the data were collected before the oil crisis (winter of 1973-74), related developments did occur during the panel year. While preliminary checks did not reveal any major disturbances in the recorded purchases of gasoline, this study avoids the longitudinal perspective and bases its analysis on average measures across the year.

The present, analysis focuses only on married couples, with or without children. The findings are further limited by the age distribution found in the panel. The average age of the wives was about 46, with the husband 1 or 2 years older. Because of these factors, the applicability of the results to newly formed families with the parents in their early twenties is doubtful. It is more properly viewed as a picture of what was happening in the beginning of the seventies in families where the adults were between 30 and 60 years of age.

With the imposition of these screens for the analysis, the actual working sample size was reduced from 500 to 403.


What indicators of buying performance are relevant in the gasoline buying situation? From the diary data it was possible to derive several indices which together make up a reasonably full picture of the buying behavior. The measures used include the proportion of times that a major (as opposed to an independent) brand was purchased, the mean price paid, the proportion of times that cash (as opposed to credit) was used, and the proportion of times the shopping was done at one's "regular'' station. In addition, the mean quantity bought per purchase occasion, the total monthly quantity bought and the quantity bought relative to the car's tank capacity were derived. Furthermore, quantity was used to weight each of the first four measures, to get an indication of the behavior relative to the quantity bought. Thus, measures were derived for the mean proportion of gasoline bought in major stations, the mean price paid per gallon, the mean quantity bought on cash, and the mean quantity bought at one's regular station. Since the results replicated very closely between the unweighted and weighted measures, only one of the latter--the one for price--is reported below.

Additional measures could have been developed, but the ones presented would seem to capture most of the characteristics of the gasoline buying performance. For example, the station variable can be seen as a loyalty variable, the major brand as one related to service, with cash usage interrelated with convenience. The product-specific variable identifying the amount of premium gasoline bought was eliminated since the type of car driven would often dictate a particular grade of gasoline.

How did women buyers perform relative to men on these criteria? Perhaps the most significant finding was that there was very little difference between the two sexes. Of the eight criteria, only three showed any significant differences (at the .05 level). If the previously mentioned weighted measures were included, this figure would have changed to 3 out of 11. There was clearly not that much difference in performance between households with a male buyer as opposed to a female buyer. It was particularly interesting to note that none of the convenience, loyalty, or service-related measures showed any significant differences between the sexes.

The only significant differences appeared in the cash usage and the two price variables. On the average women bought gasoline with cash 65% of the time, whereas men used cash in only 56% of their purchases (p = .038). The cash difference could be traced directly to credit card access. Checking the availability of credit cards among the panel families, it was discovered that male gasoline buyers carried an average of 2.5 gasoline cards as compared to 1.7 cards per female buyer (p < .01). Additionally, 33% of women buyers held no gasoline credit cards at all. For men this proportion was 21% (p = .005). Thus, the simplest explanation for why women, contrary to what one might expect, used credit cards less for their gasoline purchases was that they had less access to cards.

Given the cash paying propensity of females, it was even more noteworthy that they did in fact pay more per gallon for their gasoline. The mean price paid by women per gallon of gasoline was 37.84, whereas the average price paid by men was 37.44 (p = .053). This held regardless of whether average price per occasion was used or one looked at the weighted figures (p = .042). The differences were small, but consistent, across the panel members. Women buyers paid more.

One simple product-specific explanation for this result might be income differences between the two groups. It could very well be that women tended to be the principal buyers of gasoline in households with more than one car. One could argue further that in lower income households there was greater chance that there was only one car, and that, in turn, women would have had less chance to drive that car and therefore less chance to be the principal buyer. If so, the result with respect to price would be due to this underlying income differential, with higher income families less concerned about saving on gasoline prices.

To assess this possibility, the data were stratified according to whether the household owned one or more than one car. The previous analysis was then re-run for each subsample separately. The split did reduce the differences between the sexes to some extent. As for the multiple car household, the price difference was no longer significant. The mean price paid by women in this group was 37.94 per gallon, as opposed to 37.64 for men (p = .388). In the one-car households, on the other hand, the price differential still persisted. The mean price paid by women was 37.84 per gallon. This was significantly higher than the mean price paid by men of 37.14 per gallon (p = .05). By the magnitudes of the mean values tested, it was clear that the relatively higher prices paid by males in the multiple car household were responsible for the equalization of the mean prices paid.

These results were rather frustrating. On the one hand, the differences between the sexes in terms of performance were quite small. At the same time, however, there was a small but significant difference in price paid for the case of one-car households where presumably price savings would seem most welcome. As was expected, these households had a significantly lower income of approximately $11,200 per year as compared to $13,700 for the multiple car families. It is clear that the explanation for why women paid higher prices did not reside in basic income differences between male and female buying groups. It was in this context interesting to note also that there were approximately the same proportion female drivers among both one and multiple car households, namely, about one-third. The sex of the principal buyer did not seem to be related to the number of cars in the household.


What factors could be causing the small but persistent price differential between women and men? After exhausting the more obvious ones, it might be useful to deal somewhat more in depth with a few underlying psychological factors. One way to do this is to focus on an existing theoretical framework which attempts to explain performance. One such alternative is presented by the attribution model of motivation by Weiner, et al. (1972).

This model consists of two basic equations. The first relates performance (PER) to skills to perform in the given task (CAN) and effort (TRYING):

PER = f (CAN*TRYING) ,   (1)

with the multiplicative form implying that both effort and skill are needed to generate performance. The skill variable is further specified as relative to task difficulty (TASK):

CAN = h(ABILITY - TASK) ,   (2)

where ABILITY stands for the individual's ability in an absolute sense.

From the equations several reasons for differing performance are immediately suggested. Women might pay more because they lack the requisite skills (CAN) or simply because they are not motivated enough to put in the necessary effort (TRYING). Furthermore, the ability problem might arise not only because they lack the ability as such (ABILITY), but also because they might perceive the task of buying gasoline as relatively more difficult than do men (TASK).

From our viewpoint the model also provides another interesting perspective. Because of the interactive specification of skills and effort, the same performance level might show up for different reasons. Thus, high requisite skills and low motivation will lead to the same performance as low skills coupled with high effort. Since the empirical results presented so far indicate a large degree of similarity in performance between the sexes (apart from the cash and price criteria), it is of interest to identify whether or not the underlying skills-and-effort patterns might be the same.

For the purpose of the price analysis, the model was extended by a third equation. One reason why people would not try hard to obtain a lower price for gasoline relates to the perceived savings generated by the price search. Since these perceptions might vary across the sexes, an effort equation was specified:


In this equation, SAVINGS stands for the "change in price for a unit change in search," which is the usual economic measure for the perceived gains from one's price searching effort. The OBJECT factor refers to the strength of the objective to save which an individual brings to the given task. The multiplicative formulation allows for the cases where a person highly motivated to save money (as in the one-car family case, say) still might not bother to try hard because the possible price savings are not perceived as very great.

To evaluate the strengths of these various factors for two sexes, the AIO data were employed to derive measures of the constructs. The items chosen and the related Mann-Whitney test of the differences are given in Table 1. The discussion of the results will focus on the ability factors first, followed by the effort factors.

In terms of overall ABILITY, there is a significant difference between the sexes when the multiple car households are analyzed. Judging from the mean scores, this difference occurs because the males in this segment perceived themselves as having a relatively greater degree of personal ability.

On the perception of TASK difficulty, the responses seem to indicate that there are some significant differences between the sexes, but this time in the case of one-car households (and extending into the total sample). Apparently, the women who are the principal buyers in single car households find it a bit more difficult to cope with price changes than do men. Interestingly enough, these differences are not as significant in the case of multiple car households. [To check out whether this could be due to educational level, the difference between the two segments of women was assessed. No significant difference existed, with both categories of women showing an average schooling of high school completion.]

Going back to Equation (2), these results provide one possible explanation for the differences previously observed in performance. It seems that with respect to the CAN factor, women in single-car households perceive themselves to be as capable as the men, but also perceive the gasoline buying task as more difficult. Accordingly, their performance in terms of price paid tends to be weaker. In the multiple households, the women do not perceive more difficulty in the task itself, but their male counterparts consider themselves more able and this in turn might lead to the slightly lower price they pay. Since the difference in this latter case is insignificant (p = .388), it would seem that the general ability difference has less of an effect on the ultimate performance than does the perceived task difficulty.

Turning to the TRYING side, another reason for the performance discrepancy is uncovered. There tends to be a significant difference between the sexes in all three groupings, with women on the average being much more tired of trying to find a low price. From Equation (1) we see that this factor in conjunction with the CAN variable as derived from (2) could very well explain why it is that prices paid vary between the sexes.

What is it that makes women less willing to try? Adopting the specification in Equation (3), we can distinguish between a goal (OBJECT) and a belief (SAVINGS) factor. The mean scores for two items relating to these constructs are also given in Table 1. As can be seen, there is no strong difference between the sexes in terms of the objective itself. On the other hand, there appear to be some differences in beliefs about the possibility of obtaining lower prices by search. The differences, however, are really significant only for the multiple car households. It is not clear why female buyers in single car households try less.

Summing up, these data seem to say that the differences between the sexes in terms of price paid could be due to both the CAN and the TRYING component of the basic performance model. Furthermore, the differences in CAN are due to differences in perceived TASK difficulty for the single car households, and to differences in perceived ABILITY for the multiple car households. As for the TRYING differences, the discrepancy does not seem to arise from differences in underlying objectives, but rather from differences in perceived magnitude of potential SAVINGS. This latter result does not hold for one car families, however, for whom the TRYING differences have not yet been explained. [Since data were available for an evaluation of the complete model, an ordinary least squares analysis was performed. Using (2) to compute CAN and (3) to introduce the two determinants of TRYING, (1) was estimated using a logarithmic transformation of the model PER = a0CANa1SAVINGSa2OBJECTa3. As could be expected with individual level data and the unavoidable measurement errors, the explanatory power was low. For the single car households, the R-squares were .063 and .001 for males and females, respectively, while for multiple car households the corresponding R-squares were .011 and .108, respectively.]


In order to fully appreciate the situation in which the buying behavior is performed and the mental states developed, it would seem necessary to have some understanding of the specific socio-demographics of the families in the panel. It might be, for example, that the number of children or the level of gainful employment in a household provide real barriers to the achievement of a good performance level. Accordingly, the socio-demographic profiles of each of the sample segments were compared on the basis of Mann-Whitney tests of the difference in central tendency.



The most striking result was perhaps the lack of differences between the sexes. Number of young children, income levels, and the employment level of the wife all showed consistently insignificant differences. The relative educational level of the husband and wife (total sample, p = .012, and multiple car households, p = .011), the relative employment levels between the two (multiple car households, p = .008), sex of principal wage earner (multiple car households, p = .032), and difference in birth year (single car households, p = .044; multiple car households, p = .039; total sample, p = .005) did, however, show some significant results. [There was no correction made for the rise in a as a function of multiple comparisons between groups.] Interpreting the findings in terms of who will be the principal buyer in the family, these variables suggest that the more highly educated person in the household tends to be the buyer, and that, in addition, where the wife's employment level is more nearly equal that of the husband's, she is also more likely to be the buyer. These results are consistent with the idea of relative power as developed within the social exchange paradigm (see, for example, Homans 1961, Simpson 1976). It is interesting to note, however, that the spouses' relative power is not related to the number of young children at home, but rather to the income-generating potential of the individuals.


In concluding this empirical study of differences between sexes in their buying performance, the major emphasis has to be upon the similarity observed. On most counts, women and men when placed in the same buying role tend to perform equally. The credit card usage exception seems to have been due simply to access differences, and with changing company card issuing policies this difference might already have been obliterated. As for the small but significant price differential, the reasons were found to lie with perceived task difficulty and perhaps ability on one hand, and differences in effort on the other.

It should be noted that the particular operationalizations employed in this study to measure the various attributional constructs were dictated by the available data. Particularly, the ABILITY measure was not task-specific, but rather of a very general nature. Further hypothesis-testing research with more precise operationalizations of these constructs is needed to test whether the proposed explanation, shown in this paper to be consistent with the empirical facts, is ultimately correct or not.


Homans, G. C. (1961), Social Behavior: Its Elementary Forms, Harcourt, Brace and World.

Simpson, Richard L. (1976), "Theories of Social Exchange," in Contemporary Topics in Social Psychology, Thibaut, et al., General Learning Press, 79-98.

Weiner, B., et al. (1972), "Perceiving the Causes of Success and Failure," in Attribution: Perceiving the Causes of Behavior, Jones, et al., New York: General Learning Press, 95-120.