Mobile Edition
Print
Get the Mag
Weekly UpdatesThe participant pool consisted of twenty-five seniors and two graduate students. Approximately two-thirds of the 27 students were women. Roughly one-half had interviewed for an accounting job before, and one-half had used a placement service. The median age of the students was 27 years.
These demographic data are important for two reasons. First, given the age, prior interviewing experience, grade-point average, and motivation of these students, we feel that their responses will show high levels of self-insight, thus increasing the likelihood of not finding significant differences between the students' statistical scores and their self-insight scores. If significant differences are found between these two scores, the results will serve as strong support our argument. Second, the recruiters who volunteered to participate in the study came from firms which typically recruit from this specific participant pool. Although the students may appear to be a biased sample from the population of job-seeking students, the gender mix and age are not at all unexpected by these recruiters. Also, from experience they are more likely to have a highly-developed sense of what these students think is important in selecting a job. Their insight into the participants' weights should be good.
The recruiters were almost equally divided between public accounting firms and private industry. Recruiters had a median of nine years of experience at three (the median) different schools. Given their level of experience and knowledge of this particular group of accounting students, we expected this group to have a feel for what students think is important in a job opportunity.
Selection of Job Attributes
The job attributes were selected based on data obtained from a pilot study. Undergraduate accounting students from a different university who did not participate in the main study were used. Forty-four students were asked to consider the factors that they would find most important in the selection of an accounting job. They were given the following as examples: salary, promotion opportunities, location, type of firm, etc. These examples were based on survey results by Bundy and Norris (1992). The Bundy and Norris (1992) survey result was used only as a prompt. Participants were then asked to list those factors they felt would be important in their job choice. They were asked to list them in order of importance. The results of this exercise were consistent with prior surveys of accounting students (Bundy and Norris, 1992; Kirsch et al., 1993).
Our purpose in conducting this small-scale survey was to allow us to provide the participants in our study with variables that they believe are important, and to provide variables that have a range of perceived importance. We were particularly interested in the range and in including job attributes that were not regarded as particularly important. By including a range, we are biasing the results toward the null hypothesis of no difference in the weighting schemes. That is, some attributes seem to be clearly more important than others and, therefore, all the attributes are easier to weight using self-insight.
Based on the pilot study, the following variables were selected: (1) compensation package, (2) opportunities for advancement/job stability, (3) type of firm, (4) size of city, (5) work environment, and (6) flexible time option. Although these six variables are not the exact items reported by the pilot participants, they are our interpretation of meaningful job attributes based on the number of subjects listing an attribute and its average ranking of importance. The definitions of these variables are provided in Figure I.
Design
Each of the six job attributes was operationalized at two levels. Compensation package was either above average or below average. The opportunity/stability attribute was either a low opportunity to advance combined with high stability or a high opportunity to advance combined with low stability. Type of firm was either public accounting or private industry. Work environment was either excellent or fair. City size was either small or large. Flex-time option was operationalized by the presence or absence of a flexible time program.
The six factors were then incorporated into 32 unique job descriptions using a half-replicate factorial design (Cochran and Cox, 1957). By using one-half of the 64 possible combinations of the six variables, we hoped to lessen the probability of the students becoming tired, bored, etc. Furthermore, we are assuming that the students do not think about interactions among the attributes (Slovic, 1969). Also, to maintain comparability with the results reported by surveys only main effects are considered.
Procedure
Data From Accounting Students. Each student was supplied with a packet that contained (1) a statement of informed consent, (2) instructions for the task, (3) definitions of the six attributes of interest (see Figure 1), (4) the 32 job opportunities, (5) the attribute rating task of self-insight, and (6) a post-experimental questionnaire. An example of the task is presented in Figure II.
Data From Recruiter Accountants. Recruiter accountants were contacted using two different procedures. Most of the recruiters were contacted directly through their firms and completed the study in their offices. A few of the private recruiters completed the instrument at a regularly scheduled chapter meeting of the Institute of Management Accountants.
The study was introduced by reading a description of the task. Participants were then provided with a packet that contained (1) a copy of the description of the task, (2) definitions of the six job attributes, (3) the attribute rating task, and (4) a post-experimental questionnaire. The recruiters rated the six attributes as they perceived the students would, not what the students may say is important (i.e., "interview talk"). This distinction is subtle but important. Recruiter accountants took approximately ten minutes to complete the task.
Dependent Measures
The dependent measure of interest is the average weight assigned to each of the six job attributes (Slovic, 1969; Ashton, 1974). The average weight can be either an objectively determined weight (called statistical weight), the subjective weight assigned by the students (called self-insight weights), or the assessed weight assigned by the recruiters. The statistical weights were determined by regressing the categorical (high or low) values of the six job attributes (independent variables) against the student's rating of the job's attractiveness (dependent variable). We calculated, by student, the index [w.sup.2] for each independent variable. The [w.sup.2] is called the proportion of variance in Y accounted for by X, and it represents the strength of association between the independent and dependent variables (Hays, 1981). The six [w.sup.2]S were then rescaled to sum to 100, and an average weight was calculated for each job attribute using the observations from the 27 students.
The self-insight weights were provided by the students after they had evaluated the 32 jobs. This task is analogous to a typical survey questionnaire. Rarely, if ever, do surveys ask about tradeoffs (interactions) between job attributes. Students were asked to indicate the importance of the six job attributes by allocating 100 points among them, where higher points reflect more importance. Two different random orders of attributes were used to help mitigate any primacy or recency effect. The assigned weights were then averaged across students for each of the job attributes. The recruiter accountants were also asked to provide an allocation of 100 points based on how much weight a student they typically recruit would place on each attribute. These weights were then averaged across recruiters for each of the job attributes.
RESULTS
Students and Their Self-Insight
Table 1 presents the results from the students. A visual inspection of the data suggests that the average statistical weights for the job attributes are different from the average self-insight weights. For instance, the statistical weight for compensation package was 42% while the students' self-insight weight was 22%. Opportunities for advancement/job stability had a self-insight weight of 25% and a statistical weight of 14%. The four other variables (firm type, work environment, flex-time and city size) have about the same statistical and self-insight weights. Consistent with prior self-insight research (Mear and Firth, 1987; Ashton, 1974; Joyce, 1976), there is some evidence of "overweighting" the least important variables and "underweighting" the most important ones. Self-insight weights range from 10% to 25%, while the statistical weights range from 8% to 42%.
A Hotelling's [T.sup.2] (Morrison, 1990) was computed to detect if the difference between the vectors of means of the statistical weights and the self-insight weights was significant. There is a strong indication that the two vectors are statistically significantly different ([T.sup.2] = 23.81, p = .0059). We therefore reject the null hypothesis that there is no difference between the statistical and self-insight weights. The students are not good judges of what is important to their decision-making process.
What Recruiters Know About Students' Weights
We next compared the recruiters' assessments of the students' self-in-sight weights to the students' statistical and self-insight weights from Table 1. Table 2 provides the comparison of the recruiters' assessments to the students' average statistical weight ([[omega].sup.2]). . Three job attributes are roughly equally important according to the recruiters. They place 27%, 24% and 22% weights on firm type, compensation package, and opportunities for advancement/job stability, respectively. In comparison with the students' statistical weights, the recruiters clearly overemphasize the importance of firm type and under-emphasize the importance of compensation to the students. The test for a difference between the vectors of means is statistically significant ([T.sup.2] = 21.29, p = .0127).
FEED