Dissertations, Theses, and Capstone Projects

Date of Degree

9-2018

Document Type

Dissertation

Degree Name

Ph.D.

Program

Economics

Advisor

Wim Vijverberg

Committee Members

Partha Deb

Sebastiano Manzan

Subject Categories

Econometrics | Growth and Development | International Economics

Keywords

Synthetic Control, Dynamic Panel Data, GDP

Abstract

International sanctions imposed on Iran, targeting primarily Iran’s key energy sector and its ability to access the international financial system, have harmed Iran’s economic growth, specifically from 2011 to 2014. This thesis uses this case to study and compare the applicability of two different popular approaches used in comparative case studies exploring the effect of a policy intervention.

In the Chapter 1 we study the synthetic control method. Using this method, we estimate the effect of the intensification of sanctions on Iran’s GDP during the period 2011 to 2014. The year of 2011 was Iran’s first full year under these heavy sanctions, and in 2015, the Iran nuclear deal framework was established. Prior to this time, in spite of the ongoing U.S. sanctions, Iran’s GDP had a positive trend from 1990 to 2011. However, our estimates show that the GDP suffered a hit of more than 17 percent over the period under question. We find that these effects were particularly severe in 2012 – the same year of the enforcement by the European Union of an oil embargo and added financial boycotts against Iran.

In Chapter 2, we take a different approach to the same case, and incorporate a more structural and traditional framework. We use a Difference-in-Difference model as well as a dynamic panel data model to estimate the effect of sanctions. According to the dynamic panel data estimation, the cumulative effect of sanctions on the country’s GDP is −11.40,−18.12, and −18.62 percent for 2012, 2013, and 2014. In this chapter, we also compare the synthetic control method with the dynamic panel data regression framework. First, we show that the synthetic control method provides an unbiased estimator if the underlying model of the outcome variable of interest is a dynamic panel data model. Second, we compare the prediction power of these two methods.

In Chapter 3 we design a Monte Carlo study to discuss the performance of the methods used in previous chapters over many replications. In this chapter, we examine the robustness of the method. We conclude that the dynamic panel data model seems to be performing well with the macro level aggregate data, and the assumptions are appropriate. However, for the synthetic control method we observe large standard error in the estimated values. If we translate that to a significance analysis, this means that even though we observe meaningful values reported as the effect of the intervention, they are not statistically significantly different from 0.

Download

Share

COinS