Results Fairness of performance evaluation procedures and job satisfaction The role of outcome‐based and non‐outcome‐based effects

undertaken. The Cronbach alpha of 0.95 indicates high internal consistency. Satisfactory factor load- ing results of above 0.5 from a confirmatory factor analysis are found for the four items. Trust in the superior We use Read’s 1962 four-item instrument to measure trust in the superior. The items ask each respondent to rate: i the extent to which the re- spondent’s superior takes advantage of opportuni- ties to further the respondent’s interests; ii the extent to which the respondent feels free to discuss with the superior the problems and difficulties in the respondent’s job without jeopardising the re- spondent’s position; iii the confidence the re- spondent feels that the superior keeps the respondent fully and frankly informed; and iv the trust the respondent has that the superior’s deci- sions are justified when the superior’s decisions seem to be against the interests of the respondent. According to Read 1962: 8, these items are in- tended ‘to reflect the manager’s trust and confi- dence in the superior’s motives and intentions with respect to matters relevant to the subordinate’s ca- reer and status in the organisation.’ Several prior management accounting studies have also used this instrument e.g. Hopwood, 1972; Otley, 1978; Ross, 1994; Magner and Welker, 1994; Magner et al., 1995. The Cronbach alpha of 0.83 indicates high internal consistency for the instrument. Satisfactory factor analysis results with loading in excess of 0.5 are found for all items. Organisational commitment. The nine-item, seven-point Likert-scaled short- form version of the Organisational Commitment Questionnaire OCQ developed by Porter et al. 1974 is chosen to measure organisational com- mitment. Almost all prior management accounting studies involving organisational commitment have used this instrument e.g. Magner and Welker, 1994; Magner et al. 1995; Nouri and Parker, 1998. Cohen 1993 finds an overwhelming ma- jority of 29 out of the 36 investigations in 34 stud- ies he reviewed relied on this instrument to measure organisational commitment. It has also been tested rigorously and successfully by Mowday et al. 1979 for its psychometric proper- ties. In our study, the Cronbach alpha for the seven items is 0.89 which provides support for high in- ternal consistency. The factor loading results indi- cate that eight of the nine items have factor loadings equal to or in excess of 0.5. Only one item has a factor loading of 0.39 which is not too far below the 0.5 benchmark. Prior studies which relied on this instrument have generally used all the nine items to measure organisational commit- ment. Moreover, as the fit indices from the struc- tural modeling analysis indicate acceptable model fit to be discussed further in the Results section, we use all nine items to measure organisational commitment. Job satisfaction Job satisfaction is measured by the 20-item short form version of the Minnesota Satisfaction Questionnaire MSQ. This instrument was devel- oped by Weiss et al. 1967. A five-point scale anchored at ‘1’ for ‘Not satisfied’ to ‘5’ for ‘Extremely satisfied’ is used. The short-form ver- sion of the instrument, instead of the 100-item long-form version, is adopted to keep the ques- tionnaire to a manageable length, since senior managers with time constraints may not be willing to respond to lengthy questionnaires. As the 20 items for the job satisfaction are made up from 20 different job facets e.g. satisfaction with superior, pay, chance of promotion, working conditions, a confirmatory factor analysis is not undertaken as these items are each already a sepa- rate subdimension of overall job satisfaction. As the subscales are the subdimensions of overall job satisfaction, the sum the scores of these subdimen- sions is an appropriate measure of employee over- all job satisfaction. Hence, the overall job satisfaction score is derived by a summation of the 20 subscale scores. This summation approach is prevalent in the literature. Prior management ac- counting studies which adopt this instrument in- clude Brownell 1982, Chenhall and Brownell 1988, Frucot and Shearon 1991 and Harrison 1992. All have adopted the summation approach to derive a composite score to represent overall job satisfaction. Descriptive statistics of the variables investigated in this study are presented in Table 1.

5. Results

Given the relationships of the theoretical model in Figure 1, the structural equation modeling SEM technique is used. We select it because, apart from its ability to model relationships among multiple predictors and criterion variables and construct un- observable latent variables, it is also able to evalu- ate the relative importance of the various direct and indirect links among variables and as such helps in the understanding of the causal mecha- nism among variables. We use AMOS Version 7 for the analysis. The structural model is undertaken with all the items in the questionnaire for each construct, as discussed in the Measurement Instruments section, except for the job satisfaction construct. For the job satisfaction construct, a single score compris- ing the sum of all 20 items for this construct is used for each respondent. The results of the structural equation model pro- vide a chi-square of 349 with 223 degrees of free- 128 ACCOUNTING AND BUSINESS RESEARCH Downloaded by [Universitas Dian Nuswantoro], [Ririh Dian Pratiwi SE Msi] at 20:52 29 September 2013 dom, returning a probability value of less than 0.001. By convention, the chi-square test suggests that if the probability value p is less than 0.05, the null hypothesis that the model fits the data is rejected Bacon, 1997. Since our result has a probability smaller than 0.05, it suggests that our model does not fit the data. However, as the chi- square test is an absolute test of model fit and is sensitive to sample size, the chi-square results may be attributed to our small sample size Browne and Mels, 1994. Our sample size is only 112, which is considerably smaller than the recommended 200 to 400 cases needed for fitting model by chi-square statistics Bacon, 1997. Moreover, the modifica- tion index values from the structural model are generally quite small with only one pair of residual covariances in excess of 0.5 at 0.533. There are also no indications of errors or warnings. These re- sults suggest that it is safe for us to proceed to the relative fit statistics to ascertain the extent by which our theoretical model outperforms other baseline models. Table 2 indicates that the Tucker- Lewis Index TLI or NNFI is 0.91. The Comparative Fit Index CFI is 0.921 and the Incremental Fit Index IFI or Delta2 is 0.922. By convention, for a model to be accepted, these base- line comparison fit indices should be equal or greater than 0.90 Segars and Grover, 1993; Hartwick and Barki, 1994; Bentler, 1995. Since the indices for the theoretical model in our study are all in excess of 0.90, they indicate acceptable model fit. Table 2 also indicates that RMSEA is 0.072. By convention, there is adequate model fit if RMSEA is less than or equal to 0.08 Arbuckle, 2006. Based on these results, we use the stan- dardised direct effects path coefficients, the stan- dardised indirect effects and the standardised total effects generated by the structural equation for hy- potheses testing. 5.1. H1a, H1b, H2a, H2b, H3a, H3b and H4 Table 3 presents the zero order correlations be- tween the variables investigated in this study. These results indicate that the correlation between fairness of procedures and job satisfaction is high- ly significant r 15 =0.377; p0.01. Further analyses are needed to ascertain the nature of this signifi- cant effect. H1a, H1b, H2a, H2b, H3a, H3b and H4 each predict a significant relationship between their re- spective two variables. The results in Table 3 indi- cate that the zero order correlation coefficients for all these relationships are highly significant. Specifically, the zero order correlations are signif- icant for the relationships between i fairness of procedures and fairness of outcomes r 12 =0.418 for H1a; ii fairness of outcomes and job satisfac- tion r 25 =0.313 for H1b; iii fairness of proce- dures and trust in superior r 13 =0.259 for H2a; iv trust in superior and job satisfaction r 35 =0.533 for H2b; v fairness of procedures and organisa- tional commitment r 14 =0.407 for H3a; vi or- ganisational commitment and job satisfaction r 45 =0.543 for H3b; and vii trust in superior and organisational commitment r 34 =0.236 for H4. These hypotheses are therefore supported. Vol. 38 No. 2. 2008 129 Table 1 Descriptive statistics and Cronbach alphas Theoretical Actual range range Cronbach Variables Mean Std dev Min Max Min Max alpha Fairness of procedures 13.23 2.88 4 20 6 20 0.78 Trust in superior 13.70 3.56 4 20 5 20 0.83 Organisational commitment 42.70 9.61 9 63 16 63 0.89 Fairness of outcomes 15.17 5.07 5 25 5 25 0.95 Job satisfaction 69.83 11.24 20 100 44 98 NA Table 2 Fit indices Model TLI CFI IFI RMSEA Default model 0.910 0.921 0.922 0.072 Independence model 0.000 0.000 0.000 0.240 Downloaded by [Universitas Dian Nuswantoro], [Ririh Dian Pratiwi SE Msi] at 20:52 29 September 2013 5.2. H1c, H2c, H3c indirect effects via fairness of outcomes, trust and organisational commitment H1c states that there is an indirect relationship between fairness of evaluation procedures and job satisfaction through fairness of outcomes distrib- utive fairness. H2c states that the indirect rela- tionship is through trust. H3c states that the indirect relationship is through organisational commitment. In order to ascertain whether each of the hypoth- esised intervening variables fairness of outcomes, trust in superior and organisational commitment mediates the relationship between fairness of eval- uation procedures and job satisfaction, it is neces- sary to decompose the standardised total effect of the relationship between fairness of evaluation procedures and job satisfaction into i the stan- dardised direct effect, and ii the standardised in- direct effects. The standardised direct effects or path coefficients from the structural equation analysis by AMOS are presented in Table 4 and Figure 2. Based on these path coefficient results in Figure 2, the indirect effects of fairness of procedures FP1 on job satisfaction JS5 through fairness of outcomes FO2, trust TR3 and organisa- tional commitment OC4 are ascertained as in Table 3a below. Path 1-2-5 indicates that the indirect effect ex- clusively via fairness of outcomes is 0.062. Path 1-3-5 indicates that the indirect effect via trust is 0.165. Paths 1-4-5 and 1-3-4-5 indicate that the indirect effect through organisational commitment is 0.199 0.182 + 0.017. An indirect effect in ex- cess of an absolute amount of 0.05 may be consid- ered meaningful Bartol, 1983; Pedhazur, 1982. For this study, the indirect effects via fairness of outcomes 0.062, via trust 0.165 and via organ- isational commitment 0.199 are each in excess of 0.05. Moreover the combined indirect effect is 0.426 out of a total effect of 0.434. These results support the expectation that the effects of fairness of procedures on job satisfaction are almost entire- ly indirect rather than direct. Accordingly, H1c, H2c and H3c are supported.

6. Discussion and conclusion