Rodden, Jonathan. 2009. “Back to the Future: Endogenous Institutions and Comparative Politics,” in Marc I. Lichbach and Alan S. Zuckerman, eds. Comparative Politics: Rationality, Culture, and Structure, 2nd edition. Cambridge: Cambridge University Press.
Summarized by Ela Rossmiller
Central Question: How do observational researchers (as opposed to experimental researchers) in comparative politics meet the standards of empirical research for demonstrating causal relationships while still grappling with big, messy, interesting questions?
After all, the world is not a scientific lab in which political institutions, for example, can be randomly assigned to countries and manipulated under conditions of experimental control to observe differences in outcomes.
Main Point: Using research on endogenous democratic political systems as an example, Rodden shows how you can do research to demonstrate causal relationships.
1. Begin with the question.
a. Countries with proportional representation seem to have larger welfare states than countries with majoritarian election rules. Why is this?
2. Define the variables.
a. In this case, the independent variable is the type of election system, and the dependent variable is the size of the welfare state.
3. Identify causal mechanisms. Examples:
a. Election rules may influence politicians’ incentives to provide benefits to the broad public.
b. Election rules may favor parties on the right or left.
c. Electoral rules may influence identities of the average voter, who then influences party platforms.
d. Majoritarian rules encourage politicians to focus on critical voters or districts while ignoring others, whereas proportional voting encourages parties to appeal to all voters by proposing nation-wide programs.
e. Proportional voting systems encourage the creation of center-left coalitions. (For an explanation as to why, see p. 337) Such coalitions are more likely to support welfare states.
f. Majoritarian voting systems favor the right, whereas proportional representation systems favor the left. (For an explanation, see p. 337-8)
g. Proportional representation systems encourage higher voter turnout, which favors a more leftist platform.
4. Consider observable implications and their measurements.
a. One example: social expenditures (Persson and Tabellini) (There are others.)
b. Note difficulty in finding measurements that will help you differentiate among causal mechanisms, especially if the measurements are correlated with each other.
i. You can try making a causal mechanism an independent variable.
ii. You can hone in on a causal mechanism by changing your level of analysis (from countries to districts within a country, for example)
5. “The Endogeneity Problem”
a. There may be an unobserved variable causing both the independent variable and the dependent variable.
b. We don’t know whether this is the case because we don’t have a control group allowing us to observe the counterfactual.
c. Nor can we assume unit homogeneity among countries, i.e. that they are the same except for different election systems.
d. Nor can we randomly assign election systems to countries.
e. In other words, our research is nothing like an experimental study. We are making many assumptions which bias our findings when arguing there is a causal relationship between political institutions and policy outcomes.
6. “Empirical Strategies to Confront Endogeneity”
a. Econometric Analysis: Use the instrumental variables strategy. “The strategy is to find a vector of instruments that predicts the endogenous regressor (electoral rules) but is orthogonal to the error term, which means that the vector of instruments has no effect on the size of the welfare state other than its impact through electoral rules.” (142) Finding good instruments is hard, and Rodden gives some examples of hopelessly flawed research using poor instruments.
b. Analytical History:
i. Examples:
1. Rokkan and Boix: Proportional representation developped because parties that knew they could never get the majority vote wanted power through participating in a coalition with majority parties.
2. Blais et al: Proportional representation voting is a natural result of democratization because it’s what the people want.
3. Alesina and Glaesner: proportional representation voting was created in response to pressure from the left. Leftist political parties are a precondition for proportional representation.
4. Ticchi and Vindigni: majoritarian vs. consensus democracies arise from relative income equality/inequality in societies.
5. Cusack, Iversen, and Soskice: PR represents cooperation between economic elites/employers and workers.
ii. Need to avoid simply recounting anecdotes that suit your theory. Have claims analyzed by disinterested parties.
c. Natural Experiments: This involves analyzing historical counterfactuals. Usually involves a large elap of faith. Still, it can be done. For example, you could look at countries where multiple election systems are used, or places undergoing electoral reform.
d. Field Experiments: Researchers partner with governments or NGOs and persuade them to randomize policies or even institutions.
7. Concluding remarks
a. The empirical strategies which have dominated the field are econometric analysis and analytical history. Rather than these strategies encouraging a retreat to small, uninteresting questions, they have reinvigorated the field.
b. We still need more research on endogenous institutions.
c. Observational research is still plagued with problems related to unobserved heterogeneity, selection, external validity and errors in variables and data mining, but we need not abandon this type of research in favor of experiments or field work exclusively. They simply have to be up-front about the biases and limitations of their work.