A Conditional Reasoning Test for Risk and Incident Propensity
Type of DegreePhD Dissertation
MetadataShow full item record
Occupational injuries and fatalities persist in large numbers in the United States. Injuries and fatalities that occur at work have a high social and financial cost to employers. Moreover, human error is often to blame for these outcomes. Countless organizations could stand to benefit from the capability to predict safety performance, or alternatively risky behavior, of individuals. Traditional measures of risk propensity generally take the form of self-report test batteries, which are prone to faking and socially desirable responding. Additionally, evidence suggests that individuals are not fully aware of certain aspects of their personality and motivational tendencies. A conditional reasoning test for risk and incident propensity (CRT-RIP) was developed with hopes of tapping into implicit patterns of thought involved in risk propensity. Test items sought to identify individuals who were inclined to endorse biased answer choices in what was presented as a logic test with inductive reasoning problems. Of the 27 CRT-RIP items, 12 items were retained. With initial versions of CRTs, item retention is generally fairly low. As was expected, the CRT-RIP showed small nonsignificant correlations with explicit measures of risk propensity demonstrating that implicit personality is distinct from explicit personality. The CRT-RIP had positive and significant correlations with a risk-taking report and a report of substance use and/or abuse. In hierarchical multiple regressions, the CRT-RIP was shown to have incremental validity over the explicit measures of risk propensity when the dependent measures were risk-taking and substance use and/or abuse. This was true for both time points collected for each scale (i.e., in the past 12 months and in the lifetime). Exploratory factor analysis (EFA) and item response theory (IRT) were used to further examine the scale properties and individual item characteristics, respectively. The EFA suggested that a four-factor model is reasonable, but that a more parsimonious model is also feasible based on multiple fit statistics. A two-parameter logistic model fit both the 27-item and 12-item CRT-RIP well. A few items stood out as high-performing items.