Attribution theory and unemployment: Kelley's covariation model, self-esteem, and locus of control

Journal of Vocational Behavior 24, 94-109 (1984)

Attribution Theory and Unemployment: Kelley’s Covariation Model, Self-Esteem, and Locus of Control BERYL HESKETH Massey

University,

Palmerston

North, New Zealand

The validity of applying H. H. Kelley’s covariation attribution model (Nebraska on Motivation, 1967, 15, 192-238; American Psychologist, 1973, 28, 107-128) to understanding the perceived causes of success and failure of others’ job seeking activities was first tested in a laboratory study before testing the same theory on the self-attributions made by 82 unemployed in a field study. The field study also examined the relationship of self-esteem and locus of control to attributions for success and failure. In general Kelley’s theory was supported by the results from the laboratory study but only two of the twelve predicted relationships were found in the field study. Low distinctiveness (weak workrelated skills) was associated with strong attributions to lack of ability and low consistency (past job seeking activities successful) with strong attributions to bad luck. As predicted the unemployed with high self-esteem and an internal locus of control attributed failure to lack of effort and credited their success to ability. Unemployed with low self-esteem and an external locus of control attributed success to unstable factors, but failure was not attributed to lack of ability. Possible reasons offered for the lack of support for Kelley’s theory in the field study included the influence of group identity, individual differences in the perception of the stability and locus of causes, the greater realism of the field setting, and the inadequacy of the assumptions underlying the model. Symposium

Traditionally there has been full employment in New Zealand with an expectation that everyone could obtain work. Under these circumstances there was a tendency to blame the unemployed themselves for being out of work. However, as levels of unemployment rose it became unjustifiable to continue assuming that everyone could obtain a job if they wanted one. This resulted in a widespread public debate about the causes of unemployment. In the light of such debate it seemed important to find out what factors the unemployed themselves felt were responsible for their difficulty in obtaining work. Although some writers have examined the relevance of various theoretical areas in social psychology to unemployment in general The New Zealand Department of Labour is gratefully acknowledged for their help in approaching volunteers for this research. Requests for reprints should be sent to Dr. Beryl Hesketh, Department of Psychology, Massey University, Palmerston North, New Zealand. 94 OOOl-8791/84$3.00 Copyright AI1 rights

0 1984 by Academic Press, Inc. of reproduction in any form reserved

ATTRIBUTION

THEORY

AND

UNEMPLOYMENT

95

(Hayes & Nutman, 1981; Kelvin, 1980), and to explanations offered for its causes in particular (Gurney, 1981; Feather & Davenport, 1981), theoretically based empirical research is scant. In response to this need, the present study aimed to investigate empirically the validity of applying Kelley’s (1967, 1973) covariation attribution model to unemployment. In addition the study aimed to assess the influence of self-esteem and locus of control on explanations offered for being unemployed and for a previous success in job seeking. Kelley’s Covariation Attribution

Theory

Kelley’s covariation model assumes that people in general use a naive version of the statistical method of analysis of variance to combine consensus, distinctiveness, and consistency information in making attributions or assigning causes for events or behaviors. The consensus variable covers information comparing the behavior in question with that of others. Distinctiveness compares the behavior in question with other relevant behaviors displayed by the individual. Consistency relates to the similarity of the behavior across time and situations. In relation to unemployment, or failure to obtain work, high consensus would be represented by many others being unemployed, while low consensus would mean that few others were out of work. High distinctiveness would be represented by a person having strong work-related skills despite their failure to obtain a job, while low distinctiveness would be represented by the individual having weak work-related skills. A history of failure in job seeking would represent high consistency for failure, while a history of previous success in job seeking would represent low consistency for failure. For success in job seeking, high and low levels of these variables would be reversed. The theory predicts that high consensus, high distinctiveness, and high consistency are associated with stable causal factors external to the person (e.g., task difficulty), while low consensus, low distinctiveness, and high consistency are associated with stable causes within the person (e.g., lack of ability or skill). Circumstantial causes are associated with low consensus, high distinctiveness, and low consistency. Early support for Kelley’s covariation model was derived from analog laboratory-based research using a minimum information vignette and questionnaire format (e.g., Cordray & Shaw, 1978; McArthur, 1972; Grvis, Cunningham, & Kelley, 1975). Other studies testing Kelley’s theory have been widely quoted and reviewed (Kelley & Michela, 1980; Harvey & Weary, 1981). Kelley’s theory has also been tested using Weiner’s (1974) two dimensional causal model (internal-external and stable-unstable) for the dependent measures (Pruitt & Insko, 1980; Read & Stephan, 1979; Stevens & Jones, 1976). In the two-dimensional classification the earlier “circumstance” category is separated into internal unstable and

96

BERYLHESKETH

external unstable factors, a division supported by Pruitt and Inkso (1980). Given that consensus influences whether causes are assigned to internal or external factors, logically it can be assumed that low consensus, in combination with high distinctiveness and low consistency, would be associated with an internal unstable cause (effort), while high consensus in combination with high distinctiveness and low consistency would be associated with an external unstable cause (luck). Self-esteem and Locus of Control Kelley and Michela (1980) acknowledge that various biases and illusions also influence attributions. Personality variables such as self-esteem and locus of control may bias expectations for success and failure, which in turn influence attributions made for these outcomes. Expected outcomes are attributed to stable internal factors such as ability (Zuckerman, 1979), while unexpected outcomes are generally attributed to unstable factors (Weiner, Nierenberg, & Goldstein, 1976), or more specifically to luck (Valle & Frieze, 1976; Zuckerman, 1979) or effort (McMahan, 1973, Shrauger & Osberg, 1980). The work of Shrauger and Osberg (1980) provides insight into the relationship between attributions and self-esteem. They found that low self-esteem subjects with a tendency to expect failure attributed expected outcomes more to ability and unexpected outcomes more to effort. High self-esteem subjects attributed success to ability and failure to lack of effort regardless of prior expectations. This means that “since low selfesteem subjects may often be in situations in which success is unexpected there may be relatively few chances for them to enhance their self-esteem by seeing themselves as possessing additional positive attributes” (Shrauger & Osberg, 1980, p. 376). High self-esteem subjects, in turn are more likely to be in situations where success is expected. While self-esteem can help to explain specific expectancies for success and failure, these should be seen against a backdrop of Rotter’s (1966) more generalized expectancy or belief in internal versus external locus of control. It is important to differentiate Rotter’s concept of locus of control from the internal-external dimension in attribution theory. Rotter’s (1966) social learning theory is a personality theory and, as such, examines enduring traits or characteristics within the individual rather than a specific attribution about a particular behavioral outcome. Broedling (1975, p. 66) has made a similar point in discussing the relationship between locus of control and Vroom’s expectancy theory. Despite this clear conceptual difference, the distinction between causal dimensions and locus of control does blur at an operational level because dependent measures often have similar wording. Locus of control is also conceptually distinct from self-esteem although the two variables are moderately correlated (- .24 to - .30), those with

ATTRIBUTION

THEORY

AND UNEMPLOYMENT

97

high self-esteem tending to have an internal locus of control (Bhagat & Chassie, 1978; Kishor, 1981). Previous research has not examined the combined influence of self-esteem and locus of control on attributions. As the variables have unique variance, this will be done in the present study. Respondents with an internal locus of control and high self-esteem will have a generalized tendency to see internal stable causes as important, but because of their high self-esteem they will blame lack of effort for their failure (Shrauger & Osberg, 1980). Respondents with an external locus of control and low self-esteem will have a generalized tendency to see luck as important, but because they will expect failure, they will attribute failure to lack of ability. Success for the low self-esteem, external locus of control group will be unexpected and will be attributed to luck (Zuckerman, 1979) or, following Shrauger and Osberg (1980), to their own effort. Hypotheses The following hypotheses based on Kelley’s (1973) model were first tested in a laboratory study where attributions about a stimulus person’s behavior were elicited, and then tested in a field study on the selfattributions made by unemployed respondents. 1. Attributions to task difficulty/ease will be associated with high consensus, high distinctiveness, and high consistency. 2. Attributions to ability will be associated with low consensus, low distinctiveness, and high consistency. 3. Attributions to luck will be associated with high consensus, high distinctiveness, and low consistency. 4. Attributions to effort will be associated with low consensus, high distinctiveness, and low consistency. The following hypothesis was tested on the unemployed sample. 5. Respondents with high self-esteem and an internal locus of control will attribute failure to lack of effort and success to ability, while those with low self-esteem and an external locus of control will attribute failure to lack of ability and success to their own effort or good luck. The causal explanations offered by those unemployed with low selfesteem and an internal locus of control, or high self-esteem and an external locus of control, were explored empirically as there was no theoretical basis for a priori predictions. PRELIMINARY

LABORATORY

STUDY TESTING KELLEY’S MODEL

Method Subjects and design. A preliminary test of Kelley’s theory was carried out in a typical laboratory study (IV = 168 psychology students) on the

98

BERYL

HESKETH

attributions made about others’ success and failure in job seeking, using the traditional case study methodology (see Hesketh, 1982a for a full report). The design was a 2 x 2 x 2 x 3 x 2 multivariate mixed design, with a repeated measure on the last factor. The factors were consensus (high and low), distinctiveness (high and low), consistency (high and low), position of (a) consensus, (b) distinctiveness, and (c) consistency (abc, bca or cab), and a repeated measure on failure and success. Attributions to task difficulty/ease, ability, luck, and effort formed the multivariate dependent measures. Instruments and procedure. The task was presented to the subjects as a simulated personnel decision-making exercise following applications from young males for a hypothetical job. The consensus (one of many/ few applicants turned down/accepted for a job), distinctiveness (references indicated weak/strong work related skills), and consistency (previous job seeking successful/unsuccessful) information was varied under the general comments ostensibly made by the “personnel manager” on a summary sheet following an interview. One job applicant was presented as succeeding, while the other failed to get a job. Consensus, distinctiveness, and consistency information was worded appropriately for success and failure. The variation of the order of presentation of (a) consensus, (b) distinctiveness, and (c) consistency information was based on a Latin squares procedure (abc, bca, cab). This provided a measure of control aqainst order effect when collapsing across the order variable. After reading the information relating to consensus, distinctiveness, and consistency, subjects were asked to rate the importance of Weiner’s (1974) four causes (task difficulty/ease, ability, luck, and effort) in determining the outcome of the applications. A 9-point scale with verbal anchors “Very Important” and “Very Unimportant” was used. Following Elig and Frieze (1979), the dependent measures were worded separately for success and failure. Results

In the five-way analysis of variance design for the present study, the last two factors, order and the repeated measure of success/failure, were included partly as control variables. Keppel (1973) suggests that this is an acceptable approach. Planned 1 x 7 comparisons were used to test the main hypotheses (collapsing across order and success/failure). Attributions to ability were significantly higher for low consensus, low distinctiveness, and high consistency than for the remaining seven cells F(1, 144) = 5.76, p < .05, while attributions to luck were significantly higher for high consensus, high distinctiveness, and low consistency than for the remaining cells F(1, 144) = 9.65, p < .Ol. There was no significant

ATTRIBUTION

99

THEORY AND UNEMPLOYMENT

difference between the attributions to task difliculty/ease for high consensus, high distinctiveness, and high consistency and the remaining cells, nor between attributions to effort associated with low consensus, high distinctiveness, and low consistency and the remaining cells. Examination of the main effects from the multivariate and univariate analyses of variance provided further information relevant to the hypotheses. These are given in Table 1. All significant differences were in the predicted direction. Attributions to task difficulty/ease were stronger under high than low consensus, and high than low distinctiveness; attributions to ability were stronger under low than high distinctiveness; attributions to luck were stronger under high than low distinctiveness and low than high consistency; and attributions to effort were stronger for high than low distinctiveness. Main effects were not limited in any important way by interactions. Discussion

The laboratory study differed from the earlier laboratory tests of Kelley’s theory in that it controlled for order of presentation of consensus, distinctiveness, and consistency information and the success-failure bias in attributions made by others. In addition, the main manipulations for levels of consensus, distinctiveness, and consistency were completed between subjects, the use of extensive within subject manipulations in earlier research having been criticized. Kelley and Michela (1980) pointed out the need for research such as this, testing the covariation model. TABLE 1 Multivariate and Univariate F Values and Marginal Means for the Main Effects of Consensus, Distinctiveness, and Consistency on the Four Causal Categories (N = 168)

Consensus F Mean (low) Mean (high) Distinctiveness F Mean (low) Mean (high) Consistency F Mean (low) Mean (high)

Effort

Luck

Ability

Task Diff/ease

Multivariate

0.63 5.29 5.50

0.29 4.21 4.07

0.87 5.67 5.95

6.68** 4.25 4.94

2.454

27.44** 4.72 6.07

51.39** 3.18 5.10

63.89** 6.99 4.62

25.36** 3.92 5.27

34.05**

2.25 5.20 5.59

4.20* 4.41 3.86

1.70 6.04 5.57

0.05 4.83 4.36

3.08*

Nore. Degrees of freedom for multivariate F values = 4,150, and for univariate F values = 1,144. * p < .05. ** p < .Ol.

100

BERYL

HESKETH

The results from the study provided support for Kelley’s predicted combined influence of consensus, distinctiveness, and consistency on attributions to ability and luck, but not on attributions to effort. Mixed support was obtained for attributions to task difficulty/ease, a reverse influence of consistency accounting for the insignificant 1 x 7 planned comparison, although both consensus and distinctiveness had the predicted main effect. The results from the main effects for consensusand consistency were similar to those of Orvis et al. (1975) who also found a primary role for one variable on each category (task difficulty and luck, respectively, in the present study). Distinctiveness had a strong influence in the predicted direction on all attributions, a finding similar to that reported by McArthur (1972). Although the sex of the stimulus person in the case description was kept constant (male), a separate laboratory study (Hesketh, 1982b) found that the sex of the stimulus person in the case description resulted in no significant main effects or interactions important to Kelley’s theory. The field research provided an opportunity to assess whether the laboratory-based findings, applying the covariation model to success and failure in job seeking, were supported in a field setting. Without the laboratory study, using the traditional minimum information case study methodology, it would have been impossible to determine whether the outcome of the field research was due to the particular interpretation of Kelley’s model or to the greater reality of the setting. FIELD STUDY

Method Subjects. Although access to files was not allowed and staff involvement limited, the Department of Labour cooperated in approaching 184 newly registered unemployed to ask them to volunteer for the research. Of these, 128 volunteered (70%) while 82 were finally interviewed (45% of those originally approached or 64% of those volunteering). Of the 82 unemployed interviewed, 51 were followed up 1 month to 6 weeks later. The response rate is comparable to that obtained in other research on registered unemployed in New Zealand. The primary intent of the research was not to undertake a survey of the unemployed, but to test Kelley’s covariation attribution theory in relation to unemployment. Because of this it was felt that better quality data would be obtained from the unemployed if interviews were held away from their home distractions. For ethical reasons interviews were not held on the sameday as registering, as this would have added one further interview which may also have been linked to the process of obtaining welfare payments. These factors may have lowered the response rate. The sample was checked for bias. Those interviewed did have higher qualification levels than the group approached to volunteer. No other

ATTRIBUTION

THEORY

AND UNEMPLOYMENT

101

important biases were obtained. The final sample comprised 48 males and 34 females. Twenty-eight respondents were under 20, thirty-four were between 20 and 24 years, five were between 25 and 29 years, eight were in their 30s and seven were over 40. The mean age was 24 years. Interview protocol. The interview protocol involved a combination of open questions, structured questions with preceded responses on cards, and standardized instruments. The questions covered a range of variables as the test of Kelley’s attribution theory was only one part of the field research which also examined a variety of other psychological aspects of unemployment. The structured question relating to consensus simply asked the unemployed to estimate what percentage (grouped into seven intervals) of people with skills and qualifications like theirs were out of work. Distinctiveness required the respondents to rate the strength or weakness of their job related skills such as relationship skills, practical skills, number skills, etc., on a 7-point scale ranging from extremely weak to extremely strong. The skills were based on those which Renwick (1980) identified as important to New Zealand employers. The seven distinctiveness items demonstrated an alpha of .704 on the present sample. Consistency was obtained by asking the unemployed to assess their previous success and failure in job seeking, also on a 7-point scale ranging from extremely successful to extremely unsuccessful. Because of limited interview time consensus, distinctiveness, and consistency estimates were only obtained in relation to failure for the sample as a whole at the first interview. Attributions for failure were obtained by asking respondents “what do you think the reason is that you haven’t got a job yet?” and for success by asking “what do you think the reason is you obtained your last job?” Unstructured responses were first elicited, before having respondents rate the importance of Weiner’s (1974) causes (task difficulty/ ease, ability, luck and effort) using structured response options on a card with a 7-point scale. The structured response card contained verbal and numerical anchors for each point ranging from extremely unimportant (1) to extremely important (7). The unstructured responses to the question regarding the reasons for not having a job (failure), and for obtaining their last job (success), were recorded verbatim and rated “blind” and independently by two postgraduate psychology students and the author. These data were used to check the validity of the structured responses, which was found to be acceptable. At the time of the follow up interviews, 27 of the 51 respondents contacted were still unemployed while 24 had jobs. The successful respondents answered questions providing measures of their perceived consensus, distinctiveness, and consistency for success, while the same attribution related questions used at the first interview were repeated for those who remained unemployed.

102

BERYLHESKETH

Coopersmith’s (1967) 25item self-esteem scale was chosen as it has been used with adult and adolescent samples (Robinson & Shaver, 1973). It also has a sound theoretical basis provided in the form of a “nomological net,” with the emphasis on parental relationships as one of the major sources of self-esteem. Robinson and Shaver (1973) cite evidence of adequate reliability and validity for the scale. The 25item scale was administered at the first interview only and yielded an alpha of .74 on the present sample. As it was necessary to use a shorter locus of control scale than the Rotter (1966) scale, the II-item James Scale (Robinson & Shaver, 1973) was chosen. This questionnaire was administered at both the.first and the follow-up interview. The alpha for the James items obtained from the present sample was .768 (N = 82) at the first interview and 803 (N = 51) at the second interview. The test-retest correlation for this scale was .903 (N = 51), suggesting a high degree of stability. Results To achieve similar control in the field research to that afforded by the analysis of variance based design in laboratory studies, standard multiple regression was used. This procedure treats each independent variable (consensus, distinctiveness, and consistency) as though it were entered last, thereby controlling for the influence of the other two predictors (see Nie, Hull, Jenkins, Stembrenner, & Bent, 1975, p. 338). Kelley’s theory does not predict causal relationships among the predictor variables, so that a hierarchical method of multiple regression would have been inappropriate (see Nie et al., 1975, p. 339). However, as in standard analysis of variance designs, it was necessary to check for significant interactions between the predictor variables. Table 2 provides the beta weights from the standard multiple regressions where consensus, distinctiveness, and consistency were the predictor variables, and attributions for failure to task difficulty, lack of ability, TABLE 2 Beta Weights from the Multiple Regressions Testing Keliey’s Covariation Model Using the Attributions Made for Failure among the 82 Registered Unemployed Predictors Consensus Distinctiveness Consistency R RZh

Task Diff”

Ability”

Luck”

Effort”

0.162 - 0.033 0.063 0.188 0.035

0.118 -0.331** 0.040 0.388**

0.248* 0.096 0.208 0.3101

0.064 -0.025 -0.138 0.154

0.152**

0.096*

0.024

Nore. * = p < .05, ** = p < .Ol, using the appropriate F test for the beta weights or the multiple correlations (” Df 1,78; ’ Df 3,78).

ATTRIBUTION

THEORY

AND UNEMPLOYMENT

103

bad luck, and lack of effort the criterion variables. The beta weights for consensus, distinctiveness and consistency represent their relationship with each of the dependent measures, controlling for the remaining two predictors. A significant beta weight (the significance based on an F test) represents a significant incremental contribution of that predictor to the criterion variable, in addition to the contribution of the other two predictor variables. A test for interactions between the three predictor variables, and the significance of the contribution of these interactions to the criterion measures, was completed following the procedure outlined in the SPSSmanual (Nie et al., 1975, p. 373). No two- or three-way interactions added significantly to the prediction of the criterion variables, allowing for direct interpretation of the main beta weights in Table 2. Multiple correlations (I?) and the coefficient of determination (I?) between the three predictor variables and each of the attribution variables are given on the bottom row of the table. The results presented in Table 2 provided very limited support for Kelley’s covariation model. The only significant linear relationships, between the predictor variables and the causal dimensions, were found between distinctiveness and attributions to lack of ability, and between consensus and attributions to bad luck. The direction of these significant beta weights was as predicted, low distinctiveness resulting in a strong attribution to lack of ability, and high consensus resulting in a strong attribution to bad luck. Although it was anticipated that males and females would differ in the attributions they made for success and failure, Kelley’s theory does not predict sex differences in the relationship between the information variables and attributions. Significant sex differences were found for a number of key variables. Females perceived higher consensusfor their unemployment, F(1, 80) = 5.531, p < .05; lower distinctiveness for failure, F( 1, 80) = 10.61, p < .Ol; and they made stronger attributions to lack of ability for failure F(1, 80) = 5.48, p < .05. Because of this, Kelley’s theory was also tested separately for the two sexes. No predicted relationships reached the .05 significance level among the 48 males. Among the 34 females, low distinctiveness related significantly to a high attribution to lack of ability (beta weight = - .545), high consensus to a high attribution to bad luck (beta weight = .400), and low consistency to a high attribution to lack of effort (beta weight = - .448). What limited support was found for Kelley’s covariation model was mostly due to relationships found in the female group. The separate results for the males and females could be interpreted directly, as the two- and three-way interactions among the independent variables did not add significantly to the relationship with the dependent measures.

104

BERYL HESKETH

Personality variables and attributions. To test the predicted relationship between the personality variables and attributions, respondents were assigned to one of four quadrants on the basis of their falling above or below the median on the self-esteem and locus of control scales. As can be seen in Table 3, the median split did not result in equal numbers of high and low self-esteem scorers, and this could not have been achieved without arbitrarily assigning the 12 cases who fell on the median to one or the other category. Upper and lower thirds were not used as this would have made the sample size too small. Among the unemployed sample at the first interview, the females had significantly lower self-esteem, F( 1, 80) = 5.941, p < .05, than the males. This difference influenced the correlation between consensus and selfesteem markedly (- SO1 for males, and - .039 for females), but the relationships between self-esteem and the attributions made to the causal categories were not affected. Because of this, and because of the small size of the sample, hypotheses were tested only on the group as a whole. Table 3 provides the mean attributions made for success and failure to each of task difficulty/ease, ability, luck, and effort by respondents in each of the four quadrants. A series of one-way analyses of variance were completed on each set of four means, results of which are also given in Table 3. These analyses provided the appropriate error terms for the planned comparisons (Keppel, 1973) for quadrants 2 and 3. They

TABLE 3 Mean Attributions Made by Respondents with Combinations of High and Low Levels of Self-Esteem and an Internal or External Locus of Control Low self-esteem Failure

Success

Internal locus of control

TD/E A L E

5.154 4.462 3.307 3.846 F = 2.50 Quadrant 1

3.846 4.307 4.923 5.231 F = 1.17 (N = 13)

External locus of control

TD/E A L E

5.750 4.917 3.667 4.375 F = 8.43** Quadrant 3

3.458 3.875 4.417 5.083 F = 4.23** (N = 24)

_

High self-esteem - ._~. .~ Failure Success 4.733 3.667 2.533 4.767 F = 13.82** Quadrant 2 5.200 4.067 3.733 4.267 F = 1.72 Quadrant 4

Note. TD/E = Task difficulty/ease, A = ability, L = luck, E = effort, * = p < .05.

** = p < .01.

3.900 5.067 3.800 4.667 F = 3.12* (N = 30) 3.933 4.867 4.533 5.000 F = 1.02 (N = 15)

ATTRIBUTION

THEORY

AND

UNEMPLOYMENT

105

also provided the basis for deciding whether post hoc analyses on quadrants 1 and 4 were permissible. None of the overall F values were significant for quadrants 1 and 4 (low self-esteem internal, and high self-esteem external, respectively), and because no a priori predictions were made, no further exploratory analyses were completed. The nonsignificant overall F values for these two quadrants were consistent with the theoretical analysis and review of previous research which failed to allow a priori hypotheses for these two groups. Quadrants 2 and 3 are relevant to the hypotheses relating to the joint influence of self-esteem and locus of control on causal attributions. All the overall F values for these quadrants were significant. For the group with high self-esteem and an internal locus of control, as predicted, attributions for failure to lack of effort were significantly stronger than the average of attributions for failure to task difficulty, bad luck, and lack of ability, F(1, 26) = 11.71, p < .Ol. It must be noted that the attributions to task difficulty too were high. As predicted, attributions for success to ability were significantly stronger than the average of attributions for success to task ease, luck, and effort, F(1, 26) = 5.65, p < .05. The responses which the unemployed in this group made to the open question follow the same pattern. Of the 30 respondents with high self-esteem and an internal locus of control, 13 attributed failure to lack of effort, 5 attributed it to lack of ability, 8 to task difficulty with no one attributing failure to bad luck. For their past success 14 of these respondents attributed their success to ability, 7 to effort, none to task ease, 4 to luck, and 5 to contacts. Among the group with low self-esteem and an external locus of control, attributions for failure to lack of ability were not significantly stronger than to the remaining categories, F(1, 20) = .858, p > .05. However, this group did make stronger attributions for success to effort and luck than to ability and task ease, F( 1, 20) = 10.06, p < .Ol. This was tested by averaging the means for effort and luck (the unstable factors), and comparing this with the averaged mean for task ease and ability, using a planned comparison. In response to the open questions, 7 of these 24 low self-esteem external locus of control respondents attributed failure to ability, 9 to lack of effort, 8 to task difficulty, and none to luck. No one attributed success to effort, six attributed it to ability, six thought the task was easy, while two felt success was due to luck. Ten of these respondents attributed success to “contacts.” To summarize, hypothesis 5 was supported most clearly among those with an internal locus of control and high self-esteem, with failure being attributed to lack of effort and success to ability. Among the group with low self-esteem and an external locus of control, those whose well-being Gurney (1981) claimed could be at risk, the predicted attributions for failure did not occur, although success was attributed to unstable factors.

106

BERYL

HESKETH

GENERAL DISCUSSION There are a variety of possible reasons why there was limited support for Kelley’s theory in the field research. It is possible that the theory is more valid when making attributions about others’ behavior than for “self attribution,” particularly where outcomes involve success or failure. The different results obtained in the field study may also have been due to the greater reality of the setting and to the use of registered unemployed, and not university students, as subjects. Group identity may have influenced causal explanations in the field research. Additional exploratory analyses on the 27 respondents who remained unemployed at the second interview indicated that they saw the task of obtaining a job as difficult, but only for people without skills, a group with whom they identified. This identification was indicated by their high consensus and low distinctiveness estimates, and the relationship between these two variables. For this small group high, not low, consensus was associated with strong attributions to lack of ability (beta = .49.5), and low, not high, distinctiveness was associated with strong attributions to task difficulty (beta = - .457). In the full sample, high self-esteem was related to low consensus estimates for being unemployed (r (80) = - .37, p < .05), yet, despite this and contrary to Kelley’s theory, those with high selfesteem blamed their unemployment on the unavailability of jobs. Many respondents with high self-esteem did not identify with those out of work, despite their having registered as unemployed. Future research should investigate at what point the unemployed person starts identifying with those out of work. Kelvin (1980) discusses the shift which may take place at a societal level from rejecting the unemployed to accepting them. Such acceptance will happen when “people become adapted to the possibility of unemployment among ‘ordinary’ people like themselves” (Kelvin, 1980,p. 309). Based on attribution theory predictions, Kelvin suggests that those who do identify with the unemployed will attribute their difficulty to the situation. Although not directly tested, the present study suggests that the relationships are somewhat more complex. Many of the nonsignificant results may also be able to be explained by the respondents perceiving the stability or internality of task difficulty/ ease, ability, luck, and effort differently. In Weiner’s model these are assumed to be external stable, internal stable, external unstable, and internal unstable causes, respectively. Ability was the only category which was viewed consistently as internal and stable. Perhaps this partly explains why greater support was obtained for Kelley’s theory on this causal dimension. Future research should ask respondents to rate the extent to which task difficulty, ability, luck, and effort are seen as stable or unstable and internal or external.

ATTRIBUTION

THEORY AND UNEMPLOYMENT

107

The lack of support for Kelley’s covariation theory among the registered unemployed throws doubt on the assumption that all people are analyzers of variance. It could perhaps be suggestedthat the model of man underlying Kelley’s theory is more appropriate for a group of university students than for the registered unemployed. Similar criticisms have been offered of attribution theory in general by Semin (1980) and of Kelley’s covariation theory in particular by Newcombe and Rutter (1982). The results from the present research indicated that the personality variables did have some influence on causal attributions. Gurney (1981) found no relationship between self-esteem and attributions in a study on unemployed school leavers. However, Gurney’s result was not surprising as the relationship was examined by collapsing across the Weiner (1974) dimensions, thereby canceling any possible correlation with self-esteem. Conclusion

The main strength of the present research was that it tested Kelley’s theory in a field study after checking the validity of applying the model to success and failure in job seeking in a traditional laboratory study. Much stronger support was obtained for the theory in the laboratory study than in the field research. The hypotheses tested in the present study were based on theories derived from laboratory studies, and clearly more field research is needed to check their validity. The present study was partly motivated by the public debate about the causes of unemployment. Personality variables do relate to selfattributions, probably through their influence on prior expectancies for success and failure. Skill level (distinctiveness) and attributions to ability were clearly related. However, it seems that a simple estimate of the number of people out of work (consensus) on its own did little to influence attributions. While the unemployed perceive that they lack skills, they will continue to attribute their failure to lack of ability. Similarly, while the general public view the unemployed as lacking work skills they will continue to blame them for being out of work. Attempts to educate the public not to blame the unemployed for being out of work are unlikely to succeed unless the message is accompanied by a clear picture of the work skills which the unemployed do have to offer. REFERENCES Bhagat, R. S., & Chassie, M. B. The role of self-esteem and locus of control in the differential predictions of performance, program satisfaction and life satisfaction in an educational organisation. Journal of Vocational Behavior, 1978, 13, 317-326. Broedling, L. A. Relationship of internal-external control to work motivation and performance in an expectancy model. Journal of Applied Psychology, 1975, 60, 65-70. Coopersmith, S. E. The antecedents of self-esteem. San Francisco: Freeman, 1967. Cordray, D. S., & Shaw, J. I. An empirical test of the covariation analysis in causal attribution. Journal of Experimental Psychology, 1978, 14, 674-685.

BERYLHESKETH

108

Elig, T. W., & Frieze, I. H. Measuring causal attributions for success and failure. Journal of Personulity

and Social Psychology,

1979, 37, 621434.

Feather, N., & Davenport, P. Unemployment and depressive affect: A motivational and attributional analysis. Journal of Personality and Social Psychology, 1981, 41, 422436. Gurney, R. M. Leaving school, facing unemployment and making attributions about the causes of unemployment. Journal of Vocational Behavior, 1981, 18, 79-91. Harvey, J. H., & Weary, G. Perspectives on attributional processes. Dubuque, Iowa: W. C. Brown, 1981. Hayes, J., & Nutman, P. Understanding the unemployed: The psychological effects of unemployment. London: Tavistock, 1981. Hesketh, B. L. Psychological aspects of unemployment: Attributions about the causes of success and fai/ure in job seeking. Unpublished Ph.D. thesis, Massey University, 1982. (a) Hesketh, B. L. Attributions about success and failure in job seeking: a test of comparison behavior consensus, sex ofjob seeker and the successfailure bias. Unpublished report, Department of Psychology, Massey University, 1982. (b) Kelley, H. H. Attribution theory in social psychology. In D. Levine (Ed.), Nebraska Symposium on Motivation, 1967, 15, 192-238. Lincoln: Univ. of Nebraska Press, 1967. Kelley, H. H. The process of causal attribution. American Psychologist, 1973, 28 (2), 107128. Kelley, H. H., & Michela, J. L. Attribution theory and research. Annual Review of Psychology,

1980, 31, 457-501.

Kelvin, P. Social Psychology 2001: The social psychological bases and implications of structural unemployment. In R. Gilmour & S. Duck (Eds.), The development of social psychology. New York/London: Academic Press, 1980. Keppel, G. Design und ana/ysis: A researcher’s handbook. Englewood Cliffs, N.J.: PrenticeHall, 1973. Kishor, N. The effect of self-esteem and locus of control in career decision-making of adolescents in Fiji. Journal of Vocational Behavior, 1981, 19, 227-232. McArthur, L. Z. The how and what of why: Some determinants and consequences of causal attributions. Journul of Personality and Social Psychology, 1972, 22, 171-193. McMahan, 1. D. Relationships between causal attributions and expectancy of success. Journal of Personality and Social Psychology, 1973, 28, 108-l 14. Newcombe, R. D., & Rutter, D. R. Ten reasons why ANOVA theory and research fail to explain attribution processes: 1. Conceptual problems. Current Psychological Reviews, 1982, 2, 95-108. Nie, N., Hull, C. H., Jenkins, J., Sternbrenner, K., & Bent, D. SPSS, Statistical Package for the Social Sciences. New York: McGraw-Hill, 1975. 2nd ed. Orvis, B. R., Cunningham, J. D., & Kelley, H. H. A closer examination of causal inference: The roles of consensus, distinctiveness and consistency information. Journal ofpersonality and Social Psychology, 1975, 32, 605-616. Pruitt, D. J., & Insko, C. A. Extension of the Kelley attribution model: The role of comparison-object consensus, target-object consensus, distinctiveness and consistency. Journal of Personality and Sociul Psychology, 1980, 39, 39-58. Read, S. J., & Stephan, W. G. An integration of Kelley’s attribution cube and Weiner’s achievement attribution model. Personality and Social Psychology Bulletin, 1979, 5, 196-200. Renwick, W. L. Education and working life. In Is there life after school? Proceedings of the Association for the Study of Childhood Seminar on “The implications of unemployment for the education of our children.” Wellington, New Zealand, 1980. Robinson, J. P., & Shaver, P. R. Measures of social psychological attitudes (rev. ed.) Michigan: Survey Research Centre, 1973.

ATTRIBUTION

THEORY AND UNEMPLOYMENT

109

Rotter, J. B. Generalised expectancies for internal versus external control of reinforcement. Psychological Monographs, 1966, 80 l(Whole No. 609). Semin, G. R. A gloss on attribution theory. British Journal of Social and Clinical Psychology, 1980, 19, 291-300. Shaver, K. An introduction to attribution processes. Cambridge: Winthrop, 1975. Shrauger, J. S., 8~ Osberg, T. M. The relationship of time investment and task outcome to causal attributions and self-esteem. Journal of Personality, 1980, 48, 360-375. Stevens, L., & Jones, E. E. Defensive attribution and the Kelley cube. Journal of Personali@ and Social Psychology, 1976, 34, 809-820. Valle, V. A., & Frieze, I. H. Stability of causal attributions as a mediator in changing expectations for success. Journal of Personality and Social Psychology, 1976, 33, 579-587. Weiner, B. Achievement motivation and attribution theory. Morristown, N.J.: General Learning Press, 1974. Weiner, B., Nierenberg, R., & Goldstein, M. Social learning (locus of control) versus attributional (causal stability) interpretations of expectancy of success. Journal of Personality, 1976, 44, 52-68. Zuckerman, M. Attribution of success and failure revisited or: The motivational bias is alive and well in attribution theory. Journal of Personality 1979, 47, 245-287. Received: January 28, 1983.