Using Non-Financial Measurements as Predictors of Fraud: A Multiple Regression Analysis

Auditors have been criticized for failing to detect fraud. The changing business environment of the past decade with its growing emphasis on the importance of fraud discovery has required auditors to consider different types of evidence necessary to improve fraud discovery. One type of evidence shown to be effective in exposing revenue fraud is nonfinancial measurements (NFMs), which are operational measurements not part of the financial statement presentation. NFMs are considered difficult to falsify since some are validated by outside sources. The purpose of this quantitative study was to determine how accurately the predictors of capacity growth (Capacity Diff) and employee headcount growth (Employee Diff) predicted the criterion of fraud in the financial accounts of cost of goods sold and salaries and wages expense using a multivariate logistic regression analysis. The study was conducted using a sample size of 526 companies, 263 indicted for fraudulent filing of 10K reports with the Securities and Exchange Commission (SEC), and 263 non-fraud competitors, for the years from 2001 through 2015. The companies were identified through the Accounting and Auditing Enforcement Releases (AAERs) filed by the SEC for both revenue and expense misstatements during the period. A multivariate logistic regression was performed to test the capacity growth (Capacity Diff) and employee headcount growth (Employee Diff) predictor variables with the control variables related to the Fraud Triangle theory controlling for incentive, opportunity, and suspicious accounting, and the indicator for fraud as the dependent variable. A Pearson's correlation coefficient was calculated between the predictor variables of capacity growth (Capacity Diff) with cost of goods sold and employee headcount growth (Employee Diff) with salaries expense resulting in .992 and .968 respectively and both (one-tailed) statistically significant p < .001. A multiple logistic regression analysis resulted in R2 = .144, F (11, 514) = 7.891, and p < .001. This provided support for the usefulness of NFMs to act as a benchmark for both cost of goods sold growth and salaries expense growth at a statistically significant level.