Unobserved Heterogeneity and Bias in Tests of the Leviathan Hypothesis

This paper highlights a potential source of bias in empirical tests that attempt to link the level of fiscal centralization to the total amount of public spending in a given jurisdiction. This paper is the first to propose biased estimation as a possible explanation for the mixed results in this line of literature and is also the first to test for this bias by controlling for both time-variant and time-invariant population heterogeneity. We begin by establishing a link between government centralization and the level of heterogeneity of preferences among constituents. We then combine this link with results from recent studies that link heterogeneity to lower levels of public provision, which highlights the source of our proposed bias. Our model uses ethnic fractionalization as well as income inequality as measures of time-variant heterogeneity of preferences. We find that the omission of time-invariant heterogeneity has an impact on the estimated relationship between centralization and public spending. Depending on the specification, the resulting bias can be severe enough to change the estimated direction of the relationship between centralization and spending.