Essays on State Tax Revenues During the Great Recession and Beyond Open Access
Downloadable ContentDownload PDF
This dissertation is a collection of three essays exploring how U.S. state tax revenues – total, personal income, sales, and corporate – have fared during the Great Recession (GR) and beyond. In Chapter 2, we examine whether Virginia could have better forecasted the extreme revenue shortfalls during the GR and the subsequent weak economic recovery. We determine the viability of using Bayesian autoregression (BVAR) in estimating the four tax revenue series for Virginia. Although findings are somewhat mixed, it does appear that BVAR models with exogenous variables generate the lowest forecast errors for total and personal income revenue series. Hence, there is some policy merit for considering BVARs as alternatives to – or complements of – frequentist models.In Chapter 3, I analyze states’ decisions to enact tax increases in response to the GR using event history models – the multi-state Markov and repeated event history models. I show that the GR had a big impact on how states approached improving their fiscal conditions. I show that many states decided to tap simultaneously – and early in the downturn – minor and major revenue sources in response to the downturn and its aftermath rather than take more modest steps as to increase only major or minor revenue coffers as we initially hypothesized. Furthermore, it appears that solely political factors influenced states’ decisions to take positive net tax actions.In Chapter 4, I study time variation by analyzing fluctuations in the four tax revenue series. I parse out whether negative fluctuations have become more pronounced than positive fluctuations since the GR using a threshold generalized autoregressive conditional heteroskedastic (TGARCH) model. Along with the TGARCH, I fit a variety of time series models to find the best fitting model(s) for each of these revenue series. I find that the TGARCH does not always fit the series as well as other simpler renditions of time series models, whether heteroskedastic error is accounted for. Fitting TGARCH to revenue series before the GR was far more difficult than fitting models to series to include the GR and beyond, suggesting that data since the GR are driving the volatility of the revenue series.