The risks are coded 1 for low 2 for medium or 3 for

Info iconThis preview shows page 1. Sign up to view the full content.

View Full Document Right Arrow Icon
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: ction of cash as part of the revenue cycle examined. Variable Measurement and Model Specification To test H1, we employ the following equation: RMM 0 *MI 1 2 *PUBLIC 5 *REVENUE 3 *TENURE *INDUSTRY 6 e. 4 *PYERR (1) RMM is defined as the combination of ranked auditor-assessed revenue cycle inherent and control risks (SAS No. 47). The data-granting firm measures and documents inherent and control risk assessments as ‘‘low,’’ ‘‘medium,’’ or ‘‘high.’’ The risks are coded 1 for low, 2 for medium, or 3 for high. A multiplicative combination would result in risk of material misstatement values of 1, 2, 3, 4, 6, and 9.9 For purposes of interpretation, we use a ranking mechanism to create a RMM metric that facilitates using OLS as follows:10 7 8 9 10 Of the 78 client observations included in our sample, 18 did not receive a detailed control risk assessment, and therefore did not include a management integrity (MI) evaluation. There may have been an informal evaluation, but it was not documented in the working papers we examined. Descriptive statistics suggest the 18 firms are more risky, have shorter audit tenure, and are more likely to be private firms than the 60 firms with MI assessments. In additional analysis, if we assume that all 18 firms are low-MI firms consistent with auditors treating the firms as if they are high-risk firms and not formally evaluating MI, we find even stronger support of our hypotheses. As such, we feel that removing these firms from our analysis is conservative and if anything biases against us finding results. We did not perform any tests for inter-coder reliability. Instead we relied on the firm’s help in coding the variables to insure they measure accurately the constructs we examine. The multiplicative RMM metric is not measured on a continuous scale and therefore a multinomial logit may be a more fitting statistical model than OLS. However, the marginal effects of the regressors on the probabilities are not equal to the coefficients, rendering coefficient interpret...
View Full Document

{[ snackBarMessage ]}

Ask a homework question - tutors are online