Written Assignment 5

Submit your answer through Canvas.

It is often useful to incorporate more than one independent variable in your regression analysis. One reason may be that we may want a fuller explanation of a dependent variable. A second reason is that we may be worried that some bivariate relationship we have estimated is not causal, but instead the result of some third variable — a confounder — that we have not accounted for. In this written assignment, you will engage with such questions.

1. Lately, many social scientists have focused on uncovering the causes of corruption. A common approach to answering this question is to use large-n comparative studies. Suppose you are part of a research team assigned to this task. To start with, you have a hypothesis about the variation in level of corruption across countries: Democratic governance should lead to higher levels of corruption. The reasoning behind this hypothesis is that non-democratic regimes exercise more control over economic and political institutions and this control prevents corruption.

In order to test this hypothesis, you will download the QoG Basic Data which are located under Written Assignment 5. You will use the Corruption Perception Index (CPI) as your dependent variable. The CPI is an index based on a number of experts’ assessments of how widely spread corruption is in a country. The index is measured on a scale of 0 (highest level of perceived corruption) to 100 (lowest level of perceived corruption). In the QoG data, this variable is called ti_cpi. You measure the level of economic development using GDP per capita in USD (so, the higher this value, the higher the level of economic development). In the QoG data, this variable is called mad_gdppc. Finally, you use a Freedom House’s index of democracy (called fh_pr in the QoG data) which grades countries on a scale from 1 (most free) to 7 (least free).

a) Recode the corruption variable so that higher values mean more corruption. Do this by generating a new corruption variable called cpi_re. Do a similar thing with the democracy variable, so that higher values mean more democracy, and generate a new variable called fh_re. How will these recodes make the interpretation of the regression results below easier?

b) Make a table containing the following descriptive statistics: the mean, standard deviation, and range for the three variables cpi_re, mad_gdppc, and fh_re.

c) Run a bivariate regression with democracy (fh_re) as an independent variable and with corruption (cpi_re) as the dependent variable. Present the results in a table. (1) How should the regression coefficient (b coefficient) for democracy be interpreted? (2) Is the relationship between democracy and corruption statistically different from 0? Motivate! (3) Given your answers to (1) and (2), is your hypothesis about democracy and corruption supported by the analysis? Always motivate your answers!

d) Discuss the strength of the relationship between democracy and corruption. Here, it is a good idea to think of how the two variables are coded and to make use of the descriptive statistics you calculated in part b) of this question.

e) Now run a multiple regression where you include both democracy and economic development as independent variables. Add the results to your table. Suppose that you are mostly interested in the relationship between democracy and corruption (economic development is interesting to you, but only as a control variable). Please interpret the results using all the available information! (1) How should the regression coefficient (b coefficient) for democracy be interpreted now? (2) Is the relationship between democracy and corruption statistically different from 0 now? (3) Given your answers to (1) and (2), is your hypothesis about democracy and corruption supported by the multiple regression analysis? In addition, compare the goodness of fit of this multiple regression to the bivariate regression you ran in part c) of this question. What does this comparison tell you about the bivariate and multiple regression models you have estimated?

f) Given that your interest is mainly focused on the impact of democracy on corruption, would you consider economic development a good or bad control to include in the multivariate analysis?

g) Finally, in the data there is also a variable that measures whether the chief executive is elected by the parliament (a parliamentary system) or directly elected (a presidential system). The name of the variable in your data is parliamentary and is coded 1 if the system is parliamentary and 0 if it is presidential. Run a multiple regression with corruption as the dependent variable, and democracy, economic development, and parliamentarism as independent variables. Interpret the results for parliamentarism. In the literature, some argue that parliamentarism has a negative effect on corruption, whereas others claim the opposite. Which side is more correct according to your results?