Parliamentarism and Economic Freedom (part 2)
Checking the empirical literature
This is the second and final post on Parliamentarism and Economic Freedom. If you haven’t read the first yet, you probably should before getting to this one.
Having looked at how parliamentary and presidential countries compare in their economic freedom rankings and talked about theory, what does the academic research on the evidence tell us?
There are (unfortunately but perhaps predictably) few papers on the subject. I will rely mostly on the literature review by Murphy who identifies only two other papers (besides his own) which deal directly with the relationship between presidentialism vs parliamentarism and economic freedom.
Mudambi, Ram, Pietro Navarra, and Chris Paul. "Institutions and market reform in emerging economies: A rent seeking perspective." Public Choice 112, no. 1 (2002): 185-202.
The authors select 29 countries they perceive as “economies in transition” at the time of writing. They run two regressions, using controls: the first examines the relationship between the level of economic freedom (as measured by the Economic Freedom of the World Report in 1995) and parliamentarism-presidentialism (as well as other independent variables), and the second, the relationship between changes in economic freedom in the period 1990-1995 and the same independent variables.
These are the results for the first:
And these the results for the second:
As long as PRES is not controlled for plurality voting, they find a positive and significant effect of presidentialism on economic freedom levels and change. My thesis says nothing about parliamentarism needing to be combined with plurality voting, and even if it is combined, the beneficial effect of parliamentarism for economic freedom would be so small it wouldn’t even be significant**. So what’s going on here?
Let’s look at the 29 countries they selected:
Argentina, Bolivia, Botswana, Brazil, Chile, Colombia, Costa Rica, Czech Republic, Dominican Republic, Ecuador, Greece, Hungary, Iran, Jordan, Malaysia, Malta, Mexico, Panama, Paraguay, Peru, Poland, Romania, Slovakia, South Korea, Syria, Tunisia, Turkey***, Uruguay and Venezuela.
Following the pattern from the book, I have used italics for presidential countries, bold for parliamentary countries and no formatting for countries which are none. Jordan was dropped because the DPI doesn’t classify it as either parliamentary or presidential. Since the authors do not specify how they classified each country (though they relied on the CIA Factbook and the WDR), I once again follow the classification of the Database of Political Institutions 2017. As Jordan is a “neither”, I don’t use it in my own analysis.
A look at the countries reveals the selected “economies in transition” have some significant differences. Half of the parliamentary countries were communist until five years prior the observation. Among presidential countries, only Poland is formerly communist (yes, Poland is presidential according to the DPI and I choose to be consistent). The other “transition economies”, however, were much less in transition - they were capitalist all the time.
When I saw this list, I immediately hypothesized that the reason most of these countries had relatively lower levels of freedom was the fact that they were communist until briefly before. However, the authors include a control for Eastern European countries, which should capture this effect.
So I had no choice but to try to replicate their findings. There are some slight differences in my data, however. Instead of using data on plurality voting and form of government from Grofman and Reynolds, 2000 (I found the text but it did not come with the data), I used the DPI. I also did not use “number of electoral districts per million registered voters”. Their source for district data is the CIA World Factbook, 1998. I couldn’t find a clear way to get the info from the Factbook, so I dropped it.
Even though my changes to the approach are due to convenience, we could see it as a robustness test. After all, I changed very little. The results shouldn’t change much. Surprisingly, they are extremely different. This is my replication of their regression:
I ran twelve models, with robust standard errors like the original. Six used the 1995 data of the study, and six used data from 2022 (when the former communist countries would have had time to improve their scores). Parliamentarism has a negative coefficient only when I run the regression with 1995 data with no controls at all (reflecting that many parliamentary countries were formerly communist early in transition). After adding some controls, we get a positive coefficient for parliamentarism in all other 1995 specifications (but not significant).
When we run the regressions on 2022 data, parliamentarism has a positive and significant association with economic freedom in all specifications.
It seems the study is probably wrong, or very sensitive to slight changes in specification. I don’t believe that “number of districts per million registered voters” can have such a real effect and that parliamentarism is merely elusive for this one piece of data. I have made my data and code available if you would like to check.
De Vanssay, Xavier, Vincent Hildebrand, and Zane A. Spindler. "Constitutional foundations of economic freedom: a time-series cross-section analysis." Constitutional Political Economy 16 (2005): 327-346.
The authors use a time-series cross-section approach, with data for the quinquennials between 1970 and 2000, and for 2001. The earliest data covers 53 countries, and the most recent cover 122. Their dependent variable is Fraser’s Economic Freddom Index. They find significant and positive results for parliamentarism, as the table below (from a replication in Murphy, 2022) shows:
We see that the positive relationship between parliamentarism and economic freedom resists to different specifications. By itself, I would not give too much weight on these regressions, since they leave out, for example, geographical controls. But the combination with my replication of the Mudambi paper above, which does include such controls, points to robustness. We have a very different specification and still we find parliamentarism to have positive results.
Murphy, Ryan H. "The constitution of ambiguity: The effects of constitutions on economic freedom." Economic Affairs 42, no. 2 (2022): 240-258.
Murphy sets out to replicate the Vanssay et al’s study above. He first follows the exact same approach as Vanssay et al., but with the updated data. The results are mostly the same. Then he drops all of the data from Vanssay and uses only data from 2001-2019. As Murphy puts it, “These results are, if anything, far stronger evidence for the claims made in the original paper”:
Murphy runs some fixed-effects regressions with the same variables, and finds little which achieves conventional significance, but note that parliamentarism still has the expected sign.
He then changes approach and runs fixed-effects regressions with only one independent variable at a time. Now parliamentarism is once again positive and significant - and the fixed effects approach will address many of our concerns with endogeneity.
We once again see that parliamentarism has a significant relationship. I am not concerned with the other variables not included. For there to be ommited variable bias it would need to be the case that those variables not included in the fixed-effects regression contribute to a country becoming parliamentary and to a country’s economic freedom, but when we don’t include them we attribute the effect of the other variable to parliamentarism. I fail to see how that could be the case. This means I would say this paper brings some strong evidence that parliamentarism has a positive effect on economic freedom.
Unfortunately, that is not how the author read it. Not only is he unconvinced by the fixed-effects model, he ran another regression in which he checked whether parliamentarism had a positive effect on economic freedom in 2019, but controlling for their economic freedom level in 2000. That is, he checked whether the change in economic freedom was more positive under parliamentarism. And he found it wasn’t:
This leads Murphy to write in the abstract: “introducing almost any degree of identification draws results strongly into question.” I cannot understand why the change in economic freedom should drive the identification. Perhaps the benefits of parliamentarism are realized early, and presidential countries only later catch up. Lots of social phenomena work like that, like economic growth or life expectancy.
If we compared gains in life expectancy between rich countries and poor countries between the same years, the gains for the poor countries would be much greater. Does this mean being poor is better than being rich for a country’s health? Certainly not.
In sum, this is the state of the evidence. The literature is thin. One paper found negative results but my replication points to them actually being positive (again, do check my data and code). The other found positive results using a different approach, but wasn’t too rich. The third expanded the second one, found positive results with completely different data, and found positive results with a fixed-effects approach. The empirical literature points to the benefits of parliamentarism.