UC Riverside

This dissertation is a collection of eclectic topics of interest in macroeconomics. The first chapter augments a medium-scale DSGE model with progressive income taxes levied on interest income, and shows how its interaction with a positive government expenditure shock may lead to increased overall tax revenues. Since with interest income tax government has an additional source of revenue, following the shock government's debt obligation is lower than the one implied by the existing models with no tax on interest income.

The second chapter re-examines the ``natural resource curse'' hypothesis popularized by the cross-sectional study of \citet{sachs1995}. This chapter provides evidences against this hypothesis using the same set of variables used in \citet{sachs1995}. Both static and dynamic panel methods are utilized to overcome omitted variable bias generally present in cross-sectional regressions. The results show that resource abundance proxied by primary commodity exports share of GDP (SXP) has no statistically significant negative impact on growth. Time fixed effects point out that the debt crisis of 1980s drives the apparent negative relation between SXP and growth in cross-sectional regressions.

The third chapter seeks to identify the point in time when inflation uncertainty actually started to decline in the US, and to examine the performance of a two-regime Markov Switching-GARCH (MS-GARCH) model forecasting inflation uncertainty in the U.S. Results indicate that the switch to the low volatility regime happened approximately between April, 1979 and mid-1983. This time frame coincides with the period of aggressive monetary policy changes implemented by the then Fed chairman Paul Volcker. In addition, for a shorter horizon, normally distributed MS-GARCH forecasts and, for a longer horizon t-distributed MS-GARCH forecasts appear superior.

The final chapter seeks to predict the US recessions over the last three decades using the Treasury term spread data by employing a novel non-parametric approach called Dynamic Time Warping (DTW). Although compared to all parametric and non-parametric methods it is computationally much simpler, it has successfully signaled recessions as early as six months before the onsets of the actual recessions of 1990-1991, 2001, and 2007-2009. Compared to other non-parametric methods, DTW raises significantly fewer false recession signals.