An Assessment of an Expanded Measure of Perceived Risk

ABSTRACT - A fresh look is taken at a paper and pencil test introduced by Jacoby and Kaplan (1972) designed to measure perceived risk in consumer settings. The measure is expanded by addition of a component suggested by Roselius (1971). Validity and reliability issues are given consideration.

Citation:

George Brooker (1984) ,"An Assessment of an Expanded Measure of Perceived Risk", in NA - Advances in Consumer Research Volume 11, eds. Thomas C. Kinnear, Provo, UT : Association for Consumer Research, Pages: 439-441.

Advances in Consumer Research Volume 11, 1984      Pages 439-441

AN ASSESSMENT OF AN EXPANDED MEASURE OF PERCEIVED RISK

George Brooker, University of Puget Sound

ABSTRACT -

A fresh look is taken at a paper and pencil test introduced by Jacoby and Kaplan (1972) designed to measure perceived risk in consumer settings. The measure is expanded by addition of a component suggested by Roselius (1971). Validity and reliability issues are given consideration.

INTRODUCTION

Consumers take risks in varying degrees when they buy products. In some cases, buying a brand with which they have a long use history, they are not apt to believe a great deal of risk exists; on the other hand, when buying a product or brand for the first time, the outcome is more uncertain and greater risk may be perceived. Bauer (1960) discussed numerous examples of behavior in which the consumer s perceptions of risk were apt to be important determinants of the outcomes. As Bauer hoped, his views spawned a great deal of research in the years which followed (e.g., Ross 1975). There is much interest and potential in perceived risk as an explanatory variable in consumer behavior.

Given the level of interest and potential in perceived risk, it would be desirable to have a measure of perceived risk which would be applicable in different types of consumer research settings. While there have been several approaches to the problem appearing in the literature, one of the more promising measures was the one presented by Jacoby and Kaplan (1972). It was a paper and pencil test; it seemed to be adaptable to field research data collection projects; it was brief; the instructions were straightforward and appeared to be understandable; and it seemed to contain most of the components of risk consumers would consider.

The scale had been administered, initially, to a student sample; these individuals had been instructed regarding the nature of perceived risk (Jacoby and Kaplan 1972). A cross-validation study (Kaplan, Szybillo and Jacoby 1974) was successful with a different student sample. The second sample also had been instructed regarding the nature of perceived risk. Although the scale demonstrated adequate criterion-related validity and some aspects of internal consistency were explored in those studies, respondents were not naive and were aware of the nature of the measure being administered. With such a knowledgeable population, it is possible that familiarity with the concept allowed the students to respond to scale questions more easily than would be the case with a naive population. This could pose problems in administering the scale to different populations in more common consumer research environments. It also might lead to unusually favorable validity estimates (e.g., Rosenthal and Rosnow 1969). Thus, it is appropriate to raise a question regarding the ability of naive non-students to respond appropriately to the measure. Further, the questions of validity and reliability with a naive non-student sample should be examined

This study reexamines criterion-related validity (with less emphasis than the original study by Jacoby and Kaplan). Emphasis here is given to internal consistency and factor structure of the research instrument. In the earlier reported work on this scale, there were no reliability estimates reported. Although a good deal of validity-based information has been presented on the original version of this scale and adequate validity should be sufficient, many researchers look to reliability estimates to help them assess the potential usefulness of a scale for their work. Since no scale is error-free, it is useful to have an estimate of how reliably responses to it will be ordered. An internal consistency-based reliability estimate is presented here.

In the present study the number of risk components is somewhat expanded beyond the components originally examined by Jacoby and Kaplan. In their original study, there was an overall risk measure and five components: performance (product failure) risk, physical (safety) risk, psychological (self-image) risk, social risk, and financial risk. A footnote at the end of the paper noted the existence of a potential additional component, time-loss risk (Roselius 1971). Time-loss risk has not been included as a potential risk scale element in previous work. It is included here in an attempt to improve content validity and to examine overall performance of the slightly expanded scale.

Sample respondents in the study were a naive (to the risk measure) group of adults.

METHODOLOGY

The study was conducted as part of a somewhat larger project in and around the capitol city of a Northwestern state. Interviewers were students enrolled in a marketing research class. Respondents were adult grocery shoppers interviewed in their homes as part of an experiment on the use of generic food products. They were not randomly selected, but were randomly assigned to experimental conditions

The questionnaire used contained the expanded Jacoby-Kaplan instrument to measure perceived risk. Questions for the components of the instrument were phrased as in the operational definitions used by Jacoby and Kaplan (e.g., performance risk: "What is the likelihood that there will be something wrong with an unfamiliar brand of or that it will not work properly?"). Nine-point response scales were used, with respondents shown a card to demonstrate the range and anchor points (e.g., 1 = low, 9 = high). The products involved were peaches and spaghetti, chosen for expected quality variation between branded and unbranded versions. A measure of purchase likelihood for the generic (unbranded) product was taken on a scale of 0-100. Demographics, shopping information, and a personality measure also were collected. The personality measure included, locus of control (Rotter 1966), was part of the experiment being performed. There were no reasons to expect, a priori, that the locus of control measure should be correlated with a perceived risk measure. In a theoretical sense, whether a person expects to control environmental reinforcements should have no bearing on whether the individual perceives risk in an environment.

RESULTS

Ninety-six adults completed the study. Their median age was 28. Fourteen were men, 82 were women.

Table I presents the intercorrelations between risk components for the entire sample. As with the Jacoby and Kaplan results, there is substantial evidence that the components are not independent. The time-loss risk component shows strong, positive correlations with the other scale components. The average inter-item correlation is .297.

TABLE 1

INTERCORRELATIONS AMONG RISK COMPONENTS

Scale elements were subjected to a principal components with varimax rotation factor analysis to see what the factor structure looked like. Table 2 presents the results, with and without the inclusion of the overall risk measure. In each case there are two factors. Appropriate labels would be "Non-personal Risk' (Factor 1) and "Personal Risk" (Factor 2). The two factors explain 59.6% of the variance (Table 2a) and 60.7% of the variance (Table 2b). Overall risk is the only element to load on both factors, indicating a linkage with the two general risk types. The factor structure does not change with the inclusion of overall perceived risk.

TABLE 2

VARIMAX ROTATED FACTOR-STRUCTURE OF RISK COMPONENTS

With a high (compared to recently developed consumer scales) average inter-item correlation, an internal consistency estimate of reliability is acceptable (Brooker 1979). Table 3 presents summaries of two reliability estimates (with and without the overall risk component). The two alpha coefficients are within the range of acceptability established by Nunnally (1967). The corrected item-total correlations indicate that the overall risk measure is the best available single item representation of perceived risk (from Table 3a). The components with the weakest ties to the scale were physical and social risk.

TABLE 3

RELIABILITY ESTIMATES FOR RISK SCALE

As in the original Jacoby and Kaplan (1972) study, contributions of the components to overall perceived risk in this study were examined using stepwise regression. The results are presented in Table 4. It is shown that major contributions are made (in order of importance) by performance risk, time-loss risk, and social risk. Independent contributions by psychological, financial, and physical risk were negligible. With the exception of time-loss risk, which was not included in the original scale research, this pattern of results duplicates the earlier pattern found.

TABLE 4

STEPWISE REGRESSION ANALYSIS: CONTRIBUTION TO OVERALL RISK

Table 5 presents the mean values of the individual risk measures. None of the values is large in an absolute sense. The values are comparable to those found for less expensive items by Jacoby and Kaplan (1979).

TABLE 5

MEAN RISK VALUES FOR EACH RISK TYPE

DISCUSSION

Findings of this study indicate that the expanded perceived risk scale evidences positive characteristics for further use in consumer research. One positive aspect of the scale, not shown in tables, was the facility with which the respondents in the study handled the requirements of scale responses. There were no reported difficulties in responding to any of the scale items. (This is in contrast to the personality measure, which seemed to arouse questions of ambiguity in the respondents' minds.) In terms of ability to respond by individuals not familiar with the concepts of risk perception, the scale performed well in the field. This is encouraging, also, because much consumer research is moving out of the laboratory. Having a scale which performs well with respondents in the field certainly is desirable for researchers.

The reliability and validity characteristics of the scale also seem positive. Scale items share a common fabric. The positive intercorrelations of Table 1 indicate that the items are measuring aspects of the same construct. Estimated scale reliability (based on coefficient alpha in Table 3) is favorable based on Nunnally's (1967) criterion. This is encouraging given the circumstances of data collection (in the field, respondents contacted in their homes, responses gathered under actual interview conditions, non-professional interviewers performing the data-gathering).

Table 4 provides some insight into criterion-related validity characteristics of the scale. With overall perceived risk of purchase of the generic food products in the study as the criterion, performance risk made the largest contribution to explained variance. Even ex post, this is logical; generic foods should be expected to present performance risk due to the expected variability of non-branded products. The mean values in Table 5 indicate that the greatest perceived risk was posed by financial and performance risk, and the least by physical and social risk. Again, this is logical based on the nature of the product. Variability in unbranded products should indicate performance risk is present. Risk of product failure naturally would lead to a financial risk. These two are perceived as presenting the most substantial risks. In addition, as Table 1 indicates, they are highly correlated. (This accounts for the late entry of financial risk in the stepwise regression results in Table 4.) This correlation lends support to the theoretical linkage based on the nature of the product. Physical risk perceptions should not be substantial with packaged food products; the federal government ostensibly assures the safety of food products, and the grocery chain selling an unbranded product as its "own" lends assurance here as well. The low perception of physical risk is consistent with this "reality." The low perception of social risk is more problematic; there are no obvious ties to objective reality. It may be simply that food service does not provide any substantial social risk because the server can present or withhold a food based on its appearance or other characteristics.

Overall, the criterion-related (concurrent) validity of the perceived risk scale seems positive in this study. Content validity, while difficult to assess, was considered by Jacoby and Kaplan (1972) in developing the original scale. An additional element is added here with time-loss risk. Beyond its positive correlations with other scale elements, time-loss makes a positive contribution to scale performance (Table 3 and Table 4), suggesting that content validity is improved through its addition. Construct validity cannot be assessed with this data base; different procedures are required to evaluate construct validity. Two points are appropriate to note here, however. First, the factor analysis found only two factors in the scale. This is a favorable result because it suggests the number of dimensions present in the combination of items is small and the scale probably will not correlate strongly with a large number of unrelated concepts. Second, the scale and its elements did not correlate significantly with a measure of personality (locus of control) included in the study. Thus, although the evidence is slim, there is some indication of discriminant validity for the expanded perceived risk measure.

Obviously, it would be desirable to have additional empirical information on this scale (for other products, other circumstances, construct validity assessment, etc.). Based on the results of this analysis, the measure shows a number of favorable characteristics for use in future consumer research studies.

REFERENCES

Bauer, Raymond A. (1960), "Consumer Behavior as Risk-Taking," in R. S. Hancock, ed., Dynamic Marketing for a Changing World, Chicago: American Marketing Association, 389-399.

Brooker, George (1979), "On Selecting an Appropriate Measure of Reliability," in N. Beckwith, et al., eds., 1979 Educators Conference Proceedings, Chicago: American Marketing Association, 56-59.

Jacoby, Jacob and Kaplan, Leon B. (1972), "The Components of Perceived Risk," in M. Venkatesan, ed., Proceedings, Third Annual Conference, College Park, ED: Association for Consumer Research, 382-393.

Kaplan, Leon B., Szybillo, George 3. and Jacoby, Jacob (1974), "Components of Perceived Risk in Product Purchase: A Cross Validation," Journal of Applied Psychology, 59 (3), 287-291.

Nunnally, Jum C. (1967), Psychometric Theory, New York: McGraw-Hill Book Company.

Roselius, Ted (1971), "Consumer Rankings of Risk Reduction Methods," Journal of Marketing, 35 (January), 56-61.

Rosenthal, Robert and Rosnow, Ralph L. (1969), Artifact in Behavioral Research, New York: Academic Press.

Ross, Ivan (1975), "Perceived Risk and Consumer Behavior: A Critical Review," in M. J. Schlinger, ed., Advances in Consumer Research, Volume 2, Chicago: Association for Consumer Research. 1-19.

Rotter, Julian B. (1966), "Generalized Expectancies for Internal Versus External Control of Reinforcement," Psychological Monographs, 80 (1 whole No. 609).

----------------------------------------