Demographically adjusted norms generally enhance accuracy of inferences based on neuropsychological assessment. However, we hypothesized that demographic corrections diminish predictive accuracy for real-world activities with absolute cognitive demands. Driving ability was assessed with a 45-minute drive along a standardized on-road route in participants aged 65+ (24 healthy elderly, 26 probable Alzheimer's disease, 33 Parkinson's disease). Neuropsychological measures included: Trail-Making A and B, Complex Figure, Benton Visual Retention, and Block Design tests. A multiple regression model with raw neuropsychological scores was significantly predictive of driving errors (R2 = .199, p = .005); a model with demographically adjusted scores was not (R2 = .113, p = .107). Raw scores were more highly correlated with driving errors than were adjusted scores for each neuropsychological measure, and among healthy elderly and Parkinson's patients. When predicting real-world activities that depend on absolute levels of cognitive abilities regardless of demographic considerations, predictive accuracy is diminished by demographic corrections.