Frank McIntyre and Dr. Kerk Phillips, Economics
As the world economy becomes more integrated, questions of trade and living standards become more relevant. Developing countries see the standard of living in the industrialized nations and wonder how they can achieve the same level of prosperity. One viewpoint is that a crucial part of development for the nations of the world is access to the large consumer markets of the United States and Europe. Developing countries want open borders that will allow them to sell competitively in the industrialized nations. Many people in these industrialized countries worry about what kind of effect such open access will have on domestic jobs.
One approach to dealing with these issues is to look at how trade openness correlates to growth rates across countries and across time. We use a technique called panel regression that allows us to control for various other factor while seeing to what degree trade openness predicts growth rates.
Due to data limitations we cannot control for every possible variable. Following recent work by Sala-I-Martin, I controlled performed panel regressions where a variety of different variables were controlled for in various combinations. Each variable was drawn from a group of variables, called the conditioning set, that have been known to show some sort of explanatory power for growth rates, such as: inflation, government spending, social security expenditures, deficits, tax rates, domestic credit, literacy and others. Each panel regression included a measure of trade openness (ratio of exports plus imports to Gross Domestic Product or GDP) and several variables to be controlled for in every regression: the ratio of investment to GDP, population growth rate, secondary schooling and initial level of GDP. I then add three other variables to control for from the list of conditioning variables.
I collect the results from that regression then I run another regression using a different combination of three variables from the conditioning set. I tried all possible combinations of three from the conditioning set. This is called sensitivity analysis because it lets me see how sensitive the trade variable is to the inclusion of other variables. Since many macroeconomic variables are highly correlated to one another, a variable might appear to cause growth when, in fact, it is just highly correlated to some other variable that causes growth.
When I had run all the regressions I averaged the results and evaluated how influential trade seemed to be. I also did this procedure with a variable that was meant to measure financial market openness. The trade variable was very influential before including any variables from the conditioning set and after the sensitivity analysis it was possibly significant (p-value of .2). It was clearly sensitive what variables I included in the regression. The financial variable did not show any strong relation to country growth rates although this might be because the data I used to measure financial openness did not effectively measure financial openness.
This project was very beneficial to me as an exercise in econometrics and data manipulation. I became vary familiar with the available global data sources. I learned about the computer programs (primarily SAS) that were useful for preparing data to be processed. The programming skills will be valuable to me in graduate school and beyond.
The research combined techniques from current issues in the field of growth economics. The results highlight the importance of developing a solid theory of growth that can explain the interactions of the many variables involved in a country’s growth. Mere correlations between variables do not indicate causal relationships.