The empirical literature on compensating wage differentials has a mixed history. While there have been some successes, much of this literature finds weak support for the theory of equalizing differences. We argue that it is dispersion in total job values or "job dispersion" that leads to biased compensating wage differential estimates. We begin by demonstrating how job dispersion can lead to biased hedonic estimates. Then we take a partial equilibrium on-the-job search model with utility from non-wage job characteristics, structurally estimate it and then simulate a dataset. Using our simulated dataset, we conduct a detailed analysis of the sources of bias in hedonic wage estimates. While worker heterogeneity and job dynamics are important sources of job dispersion, a significant proportion of the variation in jobs can only be explained by the inherent randomness of job offers.