Six Essays on Meta-Regression Analysis
|Author:||Zuzana Havránková (22.4.2015)|
|Year:||2015 - summer|
|Leaders:|| prof. Ing. Oldřich Dědek CSc.
|Work type:|| Dissertations
|Awards and prizes:|
|Abstract:||This dissertation thesis consists of six papers on macroeconomics, international economics, and energy economics. All the papers are tied together by the use of metaregression analysis, which is essential for the derivation of robust policy-relevant conclusions from often conflicting results presented in the empirical literature. I use meta-analysis to quantitatively synthesize the reported research results on a given topic, correct the literature for publication selection bias, and filter out the effect of various misspecfications present in some primary studies. My results can be summarized as follows:
1) The elasticity of intertemporal substitution in consumption, a key input to all dynamic models in finance and macroeconomics, varies significantly across countries. The differences can be explained by the level of stock market participation, when countries with higher participation exhibit larger values of the elasticity; the mean reported elasticity is 0.5.
2) The effect of borders on international trade, which most authors find to be surprisingly large, can be explained away by innovations in methodology introduced in the last decade. When these innovations are taken into account jointly, the border effect disappears for developed countries, and is relatively small for developing countries.
3) When all published estimates of the effect of foreign investment on local firms in the same industries are considered and corrected for publication bias, the literature indicates a zero effect. 4) Publication bias is present also in the literature estimating the effect of foreign investment on local firms in different industries, but here the corrected effect is positive and large. 5) The mean reported price elasticity of gasoline demand is exaggerated twofold due to publication bias. 6) Finally, I also find that publication bias distorts the literature estimating the social cost of carbon emissions, because researchers tend to preferentially report large estimates.