Exploring the Impact of Democratic Capital on Prosperity Lisa L. Verdon * SUMMARY Capital accumulation has long been considered one of the driving forces behind economic growth. The idea that democratic experience accumulates and depreciates like other forms of capital is a relatively unexplored idea. Much like financial and physical capital, it has been suggested that democracy is a necessary but insufficient condition for economic growth. Building on the work of Persson and Tabellini (2006b), this paper constructs several alternative calculations of democratic capital. These values of democratic capital are then compared to GDP, as a measure of prosperity, and economic freedom for 161 countries. The results of this analysis support the idea that democracy acts indirectly through economic freedom to enhance prosperity. The causal relationships between democracy, economic freedom, and prosperity appear to be self-perpetuating. ABSTRACT Capital accumulation is one of the driving forces behind economic growth. The idea that democratic experience accumulates and depreciates like other forms of capital is a relatively unexplored idea. Like other forms of capital, it has been suggested that democracy is a necessary but insufficient condition for economic growth. This paper constructs several alternative calculations of democratic capital. These values of democratic capital are then compared to GDP per capita and economic freedom for 161 countries. The results support the idea that democracy acts indirectly through economic freedom to enhance prosperity and the relationship is self-perpetuating. JEL CLASSIFICATIONS: E02, O43 * Assistant Professor of Economics, College of Wooster, 206 Morgan Hall, Wooster, OH 44691, 1 (330) 263-2216, lverdon@wooster.edu
I. INTRODUCTION It is a standing tenant of growth theory that capital is an essential component. This capital comes in many forms including physical, human, and financial capital. For any form of capital to grow there must be continuous investment in excess of depreciation and other losses. A relatively new idea is that democracy can be thought of as a form of capital. The idea of democratic capital is that the more experience or history a country has with a democratic system, the more likely that country is to develop and maintain the institutions of democracy. Although democracy and free-markets are not synonymous, there is a very strong correlation between these two ideas. The purpose of this paper is to more clearly identify the impact of democratic capital on prosperity. I begin by recreating a basic measure of democratic capital based on the work of Persson and Tabellini (2006b). I then suggest and calculate alternative measures of democratic capital. Using these measures of democratic capital, I measure the impact of democratic capital on prosperity, measured as GDP per capita. Finally, I identify the direction of the relationship between democratic capital, economic freedom, and GDP per capita. II. LITERATURE REVIEW The positive connection between democracy and capitalism has been believed for many years. Almond (1991) provides a survey of economists views on this relationship. Among those economists is Schumpeter who suggests that democracy is supportive of capitalism. The counter argument by Marx that there will always be a tension between democracy and capitalism has proven invalid, or at the very least, inconsequential. However, understanding and quantifying this relationship is a matter that has only been addressed in the past few decades. Helliwell s (1994) analysis of the relationship between democracy and growth is the basis of most of the literature in this line if inquiry. Helliwell looked at 125 countries from 1960 to 1985 and determined that income has a positive impact on democracy and that democracy has a positive but indirect affect on growth through increases in education and investment. A multitude of papers follow which suggest different measurements and/or model specifications. These
differences result in mixed results on the nature of the relationship between democracy and growth. There are many papers that identify a positive correlation between democracy and growth. Arat (1988) identifies the positive correlation but determines that economic growth is not a sufficient condition for democracy. Alesina and Perotti (1994) find that there is a strong positive correlation between democracy, GDP per capita, and education. Alesina and Perotti then turn their focus to the stability of political regimes rather than the structure, finding that instability reduces growth regardless of the type of regime. However, Feng (1997) determines that democracy reduces the probability of a regime change and thus indirectly promotes growth through stability. Taking a different approach, Burkhart and Lewis-Beck (1994) use Granger causality to determine that growth Granger causes democracy but democracy does not Granger cause growth. This causality relationship holds up even in the poorest countries. However, Heo and Tan (2001) suggest that the relationship between democracy and growth could go either way and use Granger causality on 32 individual, developing countries. Heo and Tan find that 34 percent show growth Granger causing democracy, 31 percent show democracy Granger causing growth, 9 percent show a feedback loop, and 25 percent show no relationship. Most recently, Persson and Tabellini have produced a series of working papers looking at the relationship of democracy and growth. In their 2006 paper they look at the experience a country has with democracy, identifying that experience as democratic capital. They use this idea of democratic capital for a country and its neighbors to model the probability of a country becoming democratic and the probability of staying democratic. They find a dynamic, positive feedback loop between democratic capital and growth. They find that becoming a democracy accelerates growth by 0.75 percentage point and that this affect is still positive and significant when paired with market reforms. In their 2007 paper they suggest that previous measures of the effects of democracy on growth have been significantly underestimated. They find that the positive impact of democracy is up to 1.8 times greater than any previously reported estimates and the negative impact of autocracy is at least 1.8 times greater than previous estimates.
III. MEASURES OF DEMOCRATIC CAPITAL Quantifying the accumulation of democratic capital, or any other type of institutional capital, is full of challenges. The most obvious challenge is accounting for periods of external control. I approached this challenge as having several possibilities. The first possibility is that a nation is formed after being a colony. The second possibility is that a nation is in existence but is occupied and controlled by another country for a period of time. The third possibility is that a smaller nation splits from a larger existing nation. Finally, there is the possibility that two or more nations join together. The case where a country splits from a larger country seems the easiest to handle. A country like Bangladesh, that was part of Pakistan until 1972, shares a democratic and institutional history with Pakistan. So when Bangladesh became its own country in 1972, it started with the same democratic capital as Pakistan at that time. The opposite of this is where two or more countries join together to form one country. This is the case for Germany that was once split between East and West. For this situation, the current country shares both histories. For these cases I use the simple average of the countries democratic capital during the period(s) of separation 1. For countries that were formerly colonies, there is evidence that the legal institutions tend to stick. The same cannot be said for political institutions. For that reason, I calculate democratic capital for former colonies starting from zero at the time they become an independent country. The most interesting challenge is for countries that are occupied and controlled by another country. This situation has occurred to many countries and on several occasions. The most glaring examples are the countries of the former Soviet Union. Many of these countries were independent and democratic before they were overtaken. During the period of foreign control there are two possibilities regarding democracy. People either buy into or accept the new regime or they resist and resent the change. It seems unlikely that democratic capital can grow in this scenario but it is also not realistic to assume that the country takes on the democratic capital of its oppressors. Therefore, for these situations the democratic capital accumulated before occupation is simply depreciated while occupied. 1 It could be argued that a weighted average, based on population or some other measure of relative power, should be used. I use the simple average to avoid questions of determining relative power.
Table 1 Measures of Democratic Capital My calculations of democratic capital are based on the work of Persson and Tabellini. Persson and Tabellini (2006a) identify two different types of democratic capital, domestic and foreign.
For the purposes of this paper, I concern myself only with the domestic measure of democratic capital. To calculate democratic capital, Persson and Tabellini use the Polity II value to identify if a country is democratic or autocratic. If a country has a positive Polity II value then it is considered democratic. To calculate the democratic capital, countries receive a 1 each year they are democratic, 0 otherwise, but depreciate only in years of autocracy. They estimate a depreciation rate between 0.06 and 0.01. Like Persson and Tabellini, I will use the depreciation rate of 0.06 for δ throughout the remainder of the paper. The binary variable of democracy is represented in formula 1 as D t. The stock of democratic capital is represented as DC t-1. My calculations based on this method 2, identified as Binary, are reported for select countries in Table 1. (1) This calculation is a good place to start. A potential problem with this measure is that it only depreciates democratic capital in years of autocracy. In contrast, when we look at other types of capital accumulation, depreciation of the capital stock occurs every year. Beyond this technical argument, there is an idea in the public choice literature that suggests people become complacent in their valuation of democracy. That is, the longer they live in a democracy, the more they take it for granted. Van Den Doel and Van Velthoven (1993) support this idea with theory and evidence. In particular, they describe Olson s model of participation in democracy as supporting very low participation when policy is either very near optimal or very far from optimal. When policy is near optimal, people believe things will continue to go well so their participation is unnecessary. When policy is far from optimal, people believe their contributions will have no impact so refrain from participation. With these arguments in mind I calculate a second measure of democratic capital, identified as Binary Depreciated, based on formula 2. (2) Both of the measures produced to this point have been based on a binary value of whether a country is democratic or autocratic in a given year. The Polity II values that this binary is 2 Because they are concerned with probabilities, Persson and Tabellini adjust their calculation to stay bounded between 0 and 1. My calculation does not impose this restriction.
calculated from can provide more information. The Polity II value ranges from negative ten to positive ten. This scale provides more information (and potentially reduces measurement error) in that two countries may be considered democratic because they have positive Polity II values but one may have a score of 2 while the other has a score of 8, implying that the second is far more democratic than the first. Figure 1 Comparison of Measures (Ghana) If you believe that all autocracy has the same impact, it would be appropriate to only use the positive values of Polity II. In this case, the annual measure of democracy is equal to the Polity II value if it is positive and zero otherwise. Using these values or democracy in equation 2 provides a measure referred to as Positive Polity II. Finally, using the full range of Polity II provides a measure that allows for the varying degrees of both democracy and autocracy. Using the full range of Polity II values for democracy in formula 2 provides the final measure of democratic capital which I refer to as Polity II.
Comparing the alternative measures, it is clear that the binary measures provide very smooth time paths. There is often little difference between the Binary and Binary Depreciated values as shown in Figures 1 and 2. However the Positive Polity II and Polity II time paths show much more variation and volatility. Additionally, there are obvious breaks in trajectory that can generally be tied to major political events. This is an important distinction that the use of different measurements makes inherent assumption about government stability which must be recognized. If your view is that political institutions are generally stable over time, use of the binary measures fits this assumption. If your view is that dramatic shocks to the political system are important and far reaching, then the broader measures of Positive Polity II and Polity II are more appropriate. It should be obvious at this point that whichever assumption is made regarding political institutions will affect any subsequent analysis. Figure 2 Comparison of Measures (Chile)
IV. DATA ANALYSIS My approach to the analysis involves three parts. The first part employs a graphical analysis focusing on a few select countries. The second part uses regression analysis on the panel dataset. The third and final part looks at causality with a series of Granger type tests. The data set is based on the calculations of democratic capital described in the section three. The democratic capital measures cover the period 1800 to 2004 for approximately 160 countries 3. Addition data includes economic freedom from the Fraser Institute s Economic Freedom of the World covering the period from 1975 to 2005 and GDP per capita from the World Bank for the same period. The difference in scale of these variables makes the use of logs very appealing. However, the democratic capital value based on the full Polity II scale ranges into negative values. Taking logs of this variable cuts the sample size in half. For this reason, it is necessary to conduct some of the analysis using raw values. 1. Graphic Analysis For the graphic analysis, I compare five countries: China, Estonia, United States, Venezuela, and Zimbabwe. These five countries represent a wide range of political histories and economic prosperity. I first look at the Binary democratic capital versus the Polity II measure. This visual inspection confirms the general accuracy of the measures as we see the US at the top and China at the bottom. It is more interesting to note that Venezuela is above Estonia in both measures. However, a closer look at the Polity II measure reveals Estonia to be on a steep upward path 4 while Venezuela s path is headed downward. If we revisit this data in 50 years and both countries stay on their current trajectories, Estonia will have greater democratic capital than Venezuela. 3 The Polity IV project reports values for all independent states with populations greater than 500,000. The number of countries changes when states split, merge, etc. 4 The democratic capital for Estonia and other former Soviet block countries start (or re-start) in the mid-1990 s as they established their independence. See section III for further explanation.
Table 2- Summary Statistics
Figure 3 Binary Democratic Capital Moving on to economic freedom, we see again that the US is at the top with Estonia close behind, again showing a steep upward trend until recent years. An important difference comparing economic freedom and democratic capital is that China is in the middle of the economic freedom graph. This reflects the many steps that China has taken towards economic liberation while holding to their politically autocratic traditions. It is also not surprising to find Zimbabwe at the bottom. Zimbabwe s trend is downward sloping in the Polity II democratic capital. Zimbabwe s economic freedom shows a similar pattern, though the downturn starts later. This again supports the idea that there is a strong relationship between democratic capital and economic freedom, though that relationship may be lagged.
Figure 4 Polity II Democratic Capital Figure 5 Economic Freedom
Finally, I turn to GDP per capita. The picture here is not surprising either. The order of the countries reflects the same order as economic freedom. The US and Estonia both have strong upward sloping GDP per capita while Venezuela and Zimbabwe have little or no growth. This, once again, demonstrates the strong relationship between economic freedom and prosperity, as measured by GDP per capita. Figure 6 GDP per capita 2. Regression Analysis The next step is an attempt to quantify the relationships between democratic capital, economic freedom, and GDP per capita. I model the relationship of GDP per capita as dependent on lags of itself, lags of economic freedom, and lags of democratic capital. I choose to use a fixed effects model with in-panel AR(1) disturbances. The fixed effect absorbs the country specific characteristics, which allows our results to be a general representation of this relationship.
Including the AR(1) component provides for a consistent estimator 5. I consider up to five lags of each variable. The best fitting model is the simple model that contains one lag of each variable, represented in equation 3. (3) Although it is difficult to interpret the exact relationships between the variables, because of the large variation in units, the relationships are positive and statistically significant with the exception of democratic capital. The binary measure of democratic capital is the only version that is statistically significant. This is a first hint that how democratic capital is measured will be important to our results. One might challenge the use of OLS or the AR(1) disturbances. As robustness check, I ran the same OLS regression without the AR(1) disturbances as well as GLS with and without the AR(1) disturbances. There is little variation in the results of these alternate methods. To have a better understanding of the relationship, I rerun the regression using logged values. Using the logged values results in all of the democratic capital measures being statistically significant, though some are weakly significant and all are relatively small. From these results we can see that GDP per capita is highly dependent on previous period GDP per capita values. For economic freedom, we see that a 1 percent increase in economic freedom is associated with an approximate increase in GDP per capita of 0.1 percent. The impact of a 1 percent increase in democratic capital is less than a tenth of a percent. 5 The consistency of the fixed effects, AR(1) estimator holds in the face of heteroskedasticity under most assumptions. It is not necessarily efficient. (Wooldridge 2002)
Table 3 Estimates of Equation 3 Another approach to this type of problem is to use an Arellano-Bond estimation. In this method, past realizations of the dependent variable affect its current level and past realizations, beyond the specified lag(s), of the independent variables are used as instrumental variables. The results of the Arellano-Bond estimation are presented in Table 5 6 and are very similar to the previous results. This provides yet another robustness check. All of the coefficients are still positive and statistically significant. The primary difference is the estimated impact of democratic capital is larger while the impact of economic freedom and previous GDP per capita coefficient estimates are smaller under this method. 6 The model is not estimated using the logged values of Polity II Democratic Capital because the sample was reduced by more than half due to the negative values of the raw variable.
Table 4 Estimates of Equation 3 with Logs This set of results has at least two major implications. The first is that democratic capital may not directly affect GDP, or the direct affect is relatively small. This is not a surprising result, as it has been suggested by Feng (1997), Helliwell (1994), and Leblang (1996) that the impact of democracy is indirect. The second, and more important, implication is that how you measure democratic capital, and democracy by extension, impacts the statistical significance of your results. This supports the implications of the graphic analysis of Figures 1 and 2 in section 4.1.
Table 5 Arellano-Bond Estimates The body of research in this area often goes back and forth as to whether democracy is important. The results here suggest that those papers that are based on different measures of democracy are likely to get different results. This is particularly poignant when one considers the first two results here; Tables 3 and 4 7 are based on the exact same data and econometric approach but result in different estimates. Using different econometric methods, as presented in Table 5, simply adds to the confusion. 7 The model in Table 4 is not estimated using logged values of Polity II Democratic Capital. See the previous footnote.
3. Granger Causality Clearly, the three variables of GDP per capita, economic freedom and democratic capital are closely connected. There is a strong theoretical connection that has been demonstrated both graphically and statistically in this paper. This begs the question of which one causes the others to move. The standard approach to answering this question is Granger causality tests. Table 6 Granger Causality using VAR on Binary Democratic Capital Table 7 Granger Causality using Wald Tests on Binary Democratic Capital The challenge here is there is no exact way to translate the traditional Granger causality test to panel data. For this reason, I took several approaches to the calculations. The first approach is to
collapse the data to a measure of central tendency 8, then use the traditional Granger test. The second approach is to run a series of panel regressions with Wald tests. For these regressions I used the same fixed effect, AR(1) disturbance model employed in section 4.2. Determining the appropriate number of lags can be important. Regression results suggest that one lag is sufficient for this relationship. However, the Granger test can be biased if too few lags are included. Hsiao testing suggests that the correct number of lags is two. Since including more lags does not bias the Granger test, I include five lags. Table 8 Granger Causality using VAR on Polity II Democratic Capital Table 9 Granger Causality using Wald Tests on Polity II Democratic Capital 8 I chose to collapse to the median, instead of the mean, to reduce the impact of outliers.
The results of both approaches were generally consistent. Using the binary measure of democratic capital the results are exactly the same under both methods. Democratic capital Granger causes economic freedom and vice versa. Similarly, economic freedom Granger causes GDP per capita and vice versa. However, GDP per capita Granger causes democratic capital but democratic capital does not Granger cause GDP per capita. (4) Another observation from these results is that improvements in these factors seem to be selfperpetuating. That is, increases in democracy supports increases of economic freedom which, in turn, supports increases in democracy and GDP which then supports increases in economic freedom, and so on. The downside of this is that it also implies a negative loop for those countries that are low in any of these three areas. The results are slightly different when using the Polity II democratic capital measure. The results based on median values suggest that economic freedom Granger causes democratic capital but the relationship does not go the other direction. A more significant deviation is under the Wald tests, we find that democratic capital Granger causes GDP per capita but not the other way around. This is the only place where the results suggest a direct relationship of democratic capital causing GDP per capita. These variations in results may reflect the volatility of the Polity II measure that we see in the graphs in section 3 or it may be that there is more noise in this measure. Finally, I use simultaneous equations to see if we can better understand the complicated relationship between democratic capital and GDP per capita. It appears that the complication comes from the relative strength of relationships. This estimation uses the logged values so one can simply compare the coefficient estimates. The coefficient estimate on economic freedom, in the GDP per capita equation, is over 15 times greater than the coefficient estimate on democratic capital. This implies that even though democratic capital does contribute to GDP per capita, its contribution is often overshadowed by economic freedom.
Table 10 Simultaneous Equations
V. CONCLUSION I have put forth three original measures of democratic capital. Comparisons of these measures suggest that the most appropriate measure depends on your assumptions about political stability. I have also demonstrated that econometric results are sensitive to the measure of democratic capital employed. This sensitivity helps to explain the conflicting results in this line of research. Using a fixed effects model with in-panel AR(1) disturbances, I show that GDP per capita depends on lags of itself, economic freedom, and democratic capital. Based on these results the relationship between GDP per capita and democratic capital is either direct or indirect, depending on which measure is used. Further analysis, using Granger causality tests, support the indirect relationship of democratic capital acting through economic freedom to positively influence prosperity. VI. FURTHER RESEARCH The depreciation used to calculate the various measures of democratic capital are based on calibration of a different model. One extension would be to calibrate each of these measures separately. Given that two of the measures are based on binary values while the other two have far larger ranges suggests that the depreciation rates should be different, although the different depreciations rates may not have a significant impact. Another area of extension is to include other variables important to prosperity such as other types of capital and geographic factors. Including these requires dealing with more endogeneity issues but it will also reduce the unidentified factors that are currently being absorbed into the country fixed effects and would lend further strength to these arguments.
REFERENCES Alesina, A. & Perotti, R. (1994). The Political Economy of Growth: A Critical Survey of the Recent Literature. The World Bank Economic Review. 8: 351-371. Almond, G. (1991). Capitalism and Democracy. PS: Political Science and Politics. 24: 467-474. Arat, Z. (1988). Democracy and Economic Development: Modernization Theory Revisited. Comparative Politics. 21: 21-36. Burkhart, R. & Lewis-Beck, M. (1994). Comparative Democracy: The Economic Development Thesis. The American Political Science Review. 88: 903-910. Feng, Y. (1997). Democracy, Political Stability and Economic Growth. British Journal of Political Science, 27. 391-418. Helliwell, J. (1994). Empirical Linkages between Democracy and Economic Growth. British Journal of Political Science. 24: 225-248. Heo, U. & Tan, A. (2001). Democracy and Economic Growth: A Causal Analysis. Comparative Politics. 33: 463-473. Leblang, D. (1997). Political Democracy and Economic Growth: Pooled Cross-Sectional and Time-Series Evidence. British Journal of Political Science. 27: 453-466. Persson, T. & Tabellini, G. (2006a). Democracy and Development: Devil in the Details. The American Economic Review. 96: 319-324. Persson, T. & Tabellini, G. (2006b). Democratic Capital: The Nexus of Political and Economic Change. NBER Working Paper W12175. Persson, T. & Tabellini, G. (2007). The Growth Effect of Democracy: Is It Heterogenous and How Can It Be Estimated? NBER Working Paper W13150. StataCorp LP. (2005). Stata Longitudinal/Panel Data Reference Manual. College Station: Stata Press. Van Den Doel, H. & Van Velthoven, B. (1993). Democracy and Welfare Economics. Cambridge: Cambridge University Press. Wooldridge, J. M. (2002). Econometric Analysis of Cross Section and Panel Data. Cambridge: The MIT Press.