Thor Osborn, PhD, Principal Systems Research Analyst, Sandia National Laboratories
Most large organizations link employee salaries to an internally defined salary band structure. These structures are typically market-based, representing a simplified conception of external market salary data aggregated into a series of discrete steps. Salary band structure design involves a tradeoff between discretization error, fit to market, simplicity, and expectation setting for employees regarding salary prospects. Frequently, Human Resources fits a geometric growth model adapted from financial future value calculations to the band midpoint salaries. While this approach is an established modern practice, it does not consider personnel job frequency – so it amplifies the impact of outliers and the supra-geometric growth often seen among top-salaried positions. Moreover, mounting evidence suggests that employee perceptions and reactions to corporate salary decisions are guided by their interpretations of fair compensation and expectations regarding employment alternatives, and that these reactions are exaggerated among the lower quantiles. Consequently, asymmetric loss functions may be essential for optimizing salary structures in light of the substantial costs associated with employee turnover and disengagement. I’ll present a generalized approach that employs nonlinear modeling to accommodate acceleration of executive salaries while minimizing discretization and market fit errors, and includes asymmetric loss functions based in experimental psychology to explore optimization of net value to the organization.