| Romer (1993) shows that openness and inflation are negatively related. In his empirical findings, the relationship is stronger for subsamples of countries with higher levels of inflation. In Chapter 1, we extend the literature by introducing a theoretical model justifying such differences across levels of inflation (and money growth rates). Exchange rate ‘pass-through’ plays a key role in the model by allowing for different exchange rate adjusted inflation rates. Furthermore, exchange rate ‘pass-through’ has been documented extensively as an empirically significant phenomenon. Our empirical findings lend partial support to our theoretical results—money growth and inflation are affected differently at high and low quantiles by openness and exchange rate ‘pass-through.’; Empirical work on economic growth typically uses mean regression estimation on an equation that relates average growth rate and initial income for a cross-section of countries. Although, this methodology is popular, it has been criticized. Friedman (1992) and Quah (1993) analyze the ‘regression fallacy’ first discussed by Hotelling (1933). Koenker and Machado (1999) point out that the model underlying mean regression is too restrictive in terms of distributional assumptions (pure location shift). In Chapter 2, we propose to estimate the growth equations using quantile regression. Empirically, we find evidence of unconditional convergence for the top 35% fastest growing countries. Our estimates also suggest that there is ample evidence of conditional convergence. Tests of the location-scale shift hypothesis indicate that the initial income per capita may affect the location and the scale of the distribution of the GDP growth rates.; Estimation of spatial limited dependent models in Chapter 3 allows us to overcome several of the shortcomings inherent, in particular, to the approach of Glick and Rose (1998) and, in general, to the contagious currency crises literature. Our results confirm previous findings: (i) we do support the idea of contagious currency crises through trade channels, even when other channels, e.g. political, are present; (ii) we find some evidence of political membership contagion, although the evidence is dependent upon the definition of political interactions. Monte Carlo experiments are conducted on test statistics and estimation procedures specific to spatial probit models. Since these have only recently been purposed, small sample properties were not yet well studied and understood. |
