Corruption vs components of Economic Freedom : Data

Rodolphe Topffer — Sun, 17 Feb 2013 04:09:15 GMT

Elsewhere in the mises forum, someone asked whether or not government size correlates with lower corruption. I investigated the issue. It appears not to be the case. My data for economic freedom came from the EFW "Economic Freedom of the World" (the 2012 Annual Report). Here's the pdf. There were 5 components of economic freedom, each composed of several sub-components (more information in the appendix of the pdf). The 5 components I was using are :

Area 1. Size of Government

Area 2. Legal System and Property Rights

Area 3. Sound Money

Area 4. Freedom to Trade Internationally

Area 5. Regulation

There is also a Summary Ratings, but I'm more interested here by the components of freedom index rather than the summary rating, which is less informative I guess. To note, there were some missing values in the EFW, especially for earliers years (e.g., 1980), but nothing very important. Just a little bit annoying (for me).

The variable I used for assessing the corruption rates around the world is the Corruption Perception Index also known as (CPI), for the year 2008. Here's the link for the data.

A very important point I should note. For both economic freedom index and corruption index, the higher the score, and the higher freedom is, and the less is the corruption rate. So when you see a country with a high corruption index, that's a good news.

If you want to replicate the present finding, you need the whole dataset. You can email me at rodolphe.topffer @ gmail.com if you want my .sav file. It's for SPSS however. So, you also need SPSS. A free, and official, trial version is available here. (you can also use PSPP instead of SPSS, but it's awful...)

So I collected the data for the following years (1980, 1985, 1990, 1995, 2000, 2005, 2010) in order to see whether or not the same components correlates with each other over time. It appears to be the case, as shown below. However, as I expected, the correlations diminish over time. For example, government size index for 1980 (variable labelled SIZE1980) correlates at 76% with SIZE1985, at 53% with SIZE2000, and at 50% with SIZE2010.

If the below picture is too small to be read, please go here :

(interesting to note is the correlations between the "Sound Money" variables over time. The correlations fell to almost nothing 20 or 30 years later which means that they vary considerably over time. Nothing surprising here.)

So after thinking about it, I decided to aggregate the estimates. In SPSS, I entered the following syntax. Copy/paste the syntax below, highlight it, and click on "Run", and "Selection" :

COMPUTE SUMMARY_1980_2010 = MEAN(SummaryRating1980, SummaryRating1985, SummaryRating1990, SummaryRating1995, SummaryRating2000, SummaryRating2005, SummaryRating2010).

EXECUTE.

COMPUTE SIZE_1980_2010 = MEAN(A1SizeGovernment1980, A1SizeGovernment1985, A1SizeGovernment1990, A1SizeGovernment1995, A1SizeGovernment2000, A1SizeGovernment2005, A1SizeGovernment2010).

EXECUTE.

COMPUTE RIGHTS_1980_2010 = MEAN(A2LegalSystemPropertyRights1980, A2LegalSystemPropertyRights1985, A2LegalSystemPropertyRights1990, A2LegalSystemPropertyRights1995, A2LegalSystemPropertyRights2000, A2LegalSystemPropertyRights2005, A2LegalSystemPropertyRights2010).

EXECUTE.

COMPUTE MONEY_1980_2010 = MEAN(A3SoundMoney1980, A3SoundMoney1985, A3SoundMoney1990, A3SoundMoney1995, A3SoundMoney2000, A3SoundMoney2005, A3SoundMoney2010).

EXECUTE.

COMPUTE TRADE_1980_2010 = MEAN(A4FreedomTradeInternationally1980, A4FreedomTradeInternationally1985, A4FreedomTradeInternationally1990, A4FreedomTradeInternationally1995, A4FreedomTradeInternationally2000, A4FreedomTradeInternationally2005, A4FreedomTradeInternationally2010).

EXECUTE.

COMPUTE REGUL_1980_2010 = MEAN(A5Regulation1980, A5Regulation1985, A5Regulation1990, A5Regulation1995, A5Regulation2000, A5Regulation2005, A5Regulation2010).

EXECUTE.

Now, a bivariate correlation produced the following :

Again, click here if you cannot see the picture properly. What I wanted to check is whether or not the components correlate with each other, and if so, to what extent ? The above picture shows that the components are indeed highly correlated. However, the government size index is negatively correlated with three of the other components, although these correlations were weak, except with property rights. What does it mean ? It means that ... when freedom for the other 3 components increases, freedom regarding especially the government size decreases. Finally, it seems surprising that regulation correlates very modestly with government size (only a positive 20%). I expected much more.

Now, I will perform a multiple regression analysis. I entered my corruption index variable as a dependent variable, and the five components as independent variables. All were put in Box 1 (that is, model 1). In Box 2, I put another variable : national IQs as estimated by Lynn & Vanhanen in IQ and Global Inequality (2006). This is because IQ strongly correlates with corruption (Lynn & Vanhanen, 2012). It was interesting therefore to take IQ into account when regressing corruption on freedom. Click here if you cannot see the below picture properly :

As you can see, in Model 1, "Size" and "Money" variables are not correlated with corruption, as shown by their respective standardized Betas (-0.017 and -0.068). The coefficients are not statistically significant either, and probably due to sampling error (i.e., occurred simply by chance). The better predictor was, as expected, property rights with a Beta coefficient of +0.632.

The Correlation column gives the zero-order correlations, partial correlation, and part correlations. The zero-order is just a bivariate correlation. The partial correlation coefficient, in fact, is the correlation between the dependent variable Y and the independent variable X1 when the effect of X2 (say, the covariate) on both X1 and Y has been removed. In contrast, the part correlation (or semi-partial correlation) is the correlation between the dependent Y and the independent X1 when the effect of X2 on either X1 or Y has been removed.

We can see that the zero-order correlation for government is a negative -0.258. In other words, when the influence of the other 4 components of economic freedom has been removed, government size is no longer correlated with corruption as evidenced by its standardized Beta.

I also displayed the Collinearity Statistics column, to check whether or not there is a risk of multicollinearity. Tolerance should not be too low, and the Variance Inflation Factor (VIF) not be too high. Usually, the minimum recommended for Tolerance is 0.10 or so and the maximum for VIF is 10 or so. This appears not to be the case, given the acceptable values of Tolerance and VIF displayed above.

In Model 2, I added National IQ as independent variable. The zero-order correlation is very high (+0.737). But when the influence of the components of economic freedom has been removed, the correlation fell to +0.145. Some social scientists (e.g., Lynn, Potrafke, Rindermann, Rushton, and so on...) might say that high IQ leads to lower corruption through higher economic freedom; in fact, IQ is correlated with economic freedom (a simple bivariate shows a Pearson r of +0.623 with a p-value of 0.000). But I will not discuss the issue here.

Now that the issue about the relationship between government size and corruption has been seen, a question left unresolved is why government size is higher when freedom in property rights, money and international trade is higher ? I don't think government size really causes, for example, freedom regarding property rights. I have to think more of it later.

Maybe later, I will post some results (in another thread) about the relationship between the components of freedom and economic growth as well as GDP. The hardest part is to collect good data (I'm really bad at googling - perhaps if you can help me). What I want to try is to correlate, for example, the components of economic freedom in 1980 and 1985 with growth and GDP in later years, say, 1995 or 2000. The purpose is to see whether or not earlier freedom predicts later growth/GDP.

Hope that post is not too long.

(Edit: I forgot to report the sample size = 85 countries.)

Political Theory

Corruption vs components of Economic Freedom : Data