Cost–quality conflict in audit firms: an empirical investigation

Abstract

Findings are reported from an empirical investigation of cost–quality conflict using a survey of audit juniors and seniors in four of the Big Five audit firms in Ireland. Prior findings, mainly from US studies, indicate the presence of a cost–quality conflict in the audit environment. Conventional accounting controls are only partially effective and, in response to such controls, auditors sometimes engage in dysfunctional behaviours. Evidence from this study suggests that dysfunctional behaviours are significantly related to time pressure (created through a combination of budgets and deadlines) and performance evaluation (incorporating both style and frequency of evaluation). The relationship between dysfunctional behaviours and other specified variables (participation and leadership style) was not statistically significant. There was evidence that major changes have taken place in the audit environment such as a flattening of organizational structures and a de-emphasis on budgets as a form of control. The findings are interpreted in the context of the changed audit environment. Implications for resolving the cost–quality conflict from the perspective of audit firm management are set out and future research opportunities are identified.

Notes

Hopwood (1972) also referred to a profit-conscious style of evaluation which emphasizes increasing the long-term performance of a unit's operations in relation to the goals of an organization. This style represents a middle ground between the other two more extreme styles of evaluation: budget-constrained and non-accounting. For the purposes of the multiple regression, the essential elements of the profit-constrained style are captured in the other two styles and this variable would not be expected to add significantly to the explanatory power of the regression model.

The linearity of the phenomenon measured was examined for the overall regression using a residual plot. The Studentized residuals were compared to the predicted values of the dependent variables for both regressions. No evidence was found of a non-linear relationship. Also, scatterplots of the residuals of each of the independent variables against the dependent variables did not reveal any non-linear relationships. Regarding the second assumption, if the variance of the error term is not constant, the residual plot discussed above will show a consistent pattern. There was no evidence of significant heteroscedasticity from an examination of the scatterplot of Studentized residuals against predicted values for each regression. Violation of the third assumption is not of concern as there is no reason to assume that a sequencing variable exists. Regarding the fourth assumption, for the overall regression a normal probability plot of Studentized residuals against the normal distribution showed that the plotted residuals followed the diagonal closely. For each of the independent and dependent variables, a normal probability plot and the Kolmogorov–Smirnov statistic were examined to test fit. Some of the variables showed a significant departure from normality (Budget style of evaluation (z = 1.557, p = 0.016), URT (z = 1.635, p = 0.010), Non-accounting style of evaluation (z = 1.443, p = 0.031) and Time deadline pressure (z = 1.743, p = 0.005)). Examination of the normal probability plots showed that the departures from normality in these plots are not extreme and in situations such as this, Hair et al. (1995) suggested that using these variables instead of transformations may be preferable for interpretation of the results.

The regression models reported in Table 11 were based on a forced entry model. With a forced entry model the researcher completely specifies all the independent variables in the regression and is a more demanding approach to stepwise regression (Hair et al., 1995). Alternative regressions were subsequently computed using backward elimination. Results showed the same significant variables included for QTB (adj. R ² = 0.115, F = 5.035, p = 0.001) and URT (adj. R ² = 0.222, F = 12.770, p = 0.000) as the forced entry models.

Bartol's (1979) measure of desired professional autonomy was included in this study but due to its poor reliability and narrow focus it has been excluded from this paper.