Audits and logistic regression, deciding what really matters in service processes: a case study of a government funding agency for research grants

Governmental agencies, the back office of private firms and nongovernmental organizations experience bureaucratic processes that are often repetitive and out-of-date. These imperfections cause resource misuse and support activities that diminish to the value of the process. An important element of these bureaucratic processes is checking whether certain projects approved by the office have actually been successful in their proposed objectives. Banks and credit card companies must evaluate whether creditors have fulfilled their supposed financial worthiness, tax authorities need to classify sectors of the economy and types of tax payers for probable defaults, and research grants approved by government funding agencies should verify the use of public funds by grant recipients. In this study, logistic regression is used to estimate the probability of conformity of research grants to the financial obligations of the researcher analyzing the correlation between certain characteristics of the grant and the grant´s final status as approved or not. The logistic equation uncovers those characteristics that are most important in judging status, and supports the analysis of results as false positives and false negatives. A ROC curve is constructed which reveals not only an optimal cutoff separating conformity from nonconformity, but also discloses weak links in the chain of activities that could be easily corrected and consequently public resources preserved.