Association for Public Policy Analysis & Management

Building off an ongoing debate in the public finance literature about the role of revenue diversification in the larger public financial system, this study proposes a new approach to help bridge the gap between competing theoretical constructs and provides a new variety of evidence to advance the discussion.

Following decades of scholarly attention, much of the current literature continues to fall into the same two historic camps. First, there is the revenue-complexity hypothesis from political economy, which holds that revenues are diversified in an effort to mask the full tax burden and allow expenditures to increase. Second, the revenue volatility argument from public financial management holds that diversification is a tool to reduce volatility in total revenue from year to year, which can have adverse effects for budgeting and growth. While both threads in the literature continue to produce empirical evidence to conditionally support the preferred theoretical construct and occasionally explicitly refute the other, the nature of this back-and-forth discussion makes definitive statements about the purpose of revenue diversification difficult.

This paper first develops a unified theoretical model that combines the insights of both of the principal theories in the extant literature, in which both hypotheses exist as a part of a unified revenue system rather than as strict alternatives to one another. Creating a unified system solidifies the idea that feedback loops are likely prevalent and that endogeneity between the variables is a major concern. Keeping this endogeneity in mind, Granger causality methods are employed in order to test the relationships within the system. While not completely foreign to public finance, Granger causal methods in the field have been relatively rare until recent years, despite the fact that they offer a ready strategy for providing baseline evidence in a causally complicated system such as this. A panel of financial data for the U.S. States from 1977-2012 is analyzed using this strategy. The methodology exploits the intertemporal aspect of the panel data and employs a carefully-specified lag structure in order to test the interrelationships between all of the primary variables – revenue diversification, revenue volatility, and government size – in both directions.

Results indicate the presence of substantial feedback loops in the system that have often been overlooked in past research, confirming concerns over possible endogeneity and the reliability of some past results. The evidence for the revenue-complexity hypothesis is weak and susceptible to non-significance under certain specifications, as some dissenting scholars have long claimed. However, the results regarding the revenue volatility argument are more stable and surprising; on average, it appears that governments do employ greater revenue diversification in response to revenue volatility (as expected) but that the revenue volatility does not necessarily decrease in response to such diversification efforts. As such, some of the traditional assumptions in both threads of the literature addressing revenue diversification are called into question and should be more carefully considered in future research.