Tuesday, October 26, 2010

Public pension crisis in OECD countries

The central aim of my bachelor's thesis is to demonstrate the unsustainability of public pension system in OECD countries in the longer run through the lens of a rigorous theoretical and empirical analysis.

The origins of contemporary public pension schemes date back to 19th century when Bismarckian Germany in 1881 first adopted a universal old-age public pension system based on pay-as-you-go (PAYG) funding principle. The principle itself captures full advantages of high (stationary) population growth rate. In the simplest form, PAYG pension scheme is based on the notion of generational solidarity upon which current generations pay mandatory social security contribution into the public scheme. Aggregate contributions are then paid out to current retirees. The cycle is then expanded through generations. However, PAYG funding scheme is sustainable as long as the population growth is high and above the marginal productivity of the capital. Back in 19th century, public pension schemes were adopted under unrealistic assumptions about future population prospects. In 19th century, advanced countries experienced high population growth rate, high fertility rate and an extremely low share of dependent old population that was receiving universal old-age support from PAYG pension schemes. These set of assumptions was crucial to the stability of government-provided old-age support embodied in the public pension schemes.

The sustainability of PAYG pension system requires the equivalence of population growth rate and real interest rate. In the early 20th century, the advanced world shifted towards aging population, declining fertility rates and lower labor market entry rate. In broad terms, a growing old-age dependency ratio led to the pure disequilbrium effects. In a theoretical framework, I re-examined the neoclassical framework of lifecycle hypotheses embodied in Samuelson and Cass-Yaari models of life-cycle utility maximization. The lifecycle hypothesis is based upon the assumption of the three-period model where individuals maximize the consumption in the course of a lifetime. In the first period, individuals do not discount the future consumption since, in this period, individuals acquire the human capital. In the second period individuals enter the working age and discount the future consumption. Hence, in the third period, individuals retire consume the output produced in the working-age period. Since future discounting is compounded, the lifetime consumption increases geometrically. In purely analytical terms, the individuals maximize the utility of consumption through time preference rate.

Considering the abovementioned equivalence between population growth rate and real interest rate, the stability of the equilibria requires the period discount rate to equal the population growth rate. If population growth rate decreases, the stability of the equilibria requires that individuals decrease the future discount rate by the same rate to keep the PAYG pension system within the theoretical limit. The rigorous theoretical formulation of the neoclassical model of lifetime consumption, which essentially captures the necessary conditions for equilibrium stability of public pension schemes, had been put forth by Paul A. Samuelson in his seminal contribution to the theoretical foundations of stationary "PAYG" public pension scheme .

In the course of the last decades, OECD countries have experienced a significant drop in fertility rates, population growth and, under the political climate of social democracy, a widespread adoption of early retirement schemes and generous social security benefits. In addition, labor market exit age dropped significantly, initiating a trend towards the unprecendent growth of generational indebtedness.

The OECD estimated that between 2000 and 2050, old-age dependency ratio is forecast to increase to the largest extent in Japan (193 percent), Spain (136 percent), Portugal and Greece (135 percent). The astonishing increase in the estimated old-age dependency ratio directly reflects the declining fertility rate in OECD countries from 1960s onwards. I estimated the ratio of fertility rate between 1960-1970 and 2000-2006 for OECD countries at around 2, which means that average fertility rate between 1960-1970 was twice the fertility rate between 2000-2006. The highest fertility ratios were found in Spain (2.23), Italy (1.96), Ireland (2.00) while the lowest ratios were found in Denmark (1.37), Netherlands (1.72) and the United States (1.46).

High and stable effective retirement age is the main assumption underlying the stationary stability of PAYG pension system. In the 20th and 21st century, OECD countries have experienced an unprecendent decline in effective retirement age. Blöndal and Scarpetta (2002) estimated the decline in labor market exit age for OECD countries between 1960 and 1995. The female labor market exit age had declined significantly in Ireland (10.7 years), Spain (9.1 years) and Norway (8.8 years). Male labor market exit age exerted persistent decline in all developed OECD countries except for Iceland. The exit age declined significantly in the Netherlands (7.3 years) and Spain (6.5 years).

In a large part, declining labor market exit age has confluenced the rapid growth of unemployment and disability benefits and early retirement incentives from the second half of the 20th century onwards. As the OECD correctly contemplated, in a number of countries, disability pensions and unemployment benefits can be used as de facto early retirement schemes. In a large part, widespread growth of early retirement schemes and implicit incentives for moral hazard in retiring too early via unemployment and disability schemes is held responsible by generous welfare states in the aftermath of the World War II.

When I examined various features affecting early retirement choices, I came across an interesting finding. I regressed labor market exit age and marginal tax rate in a cross section of 23 OECD countries in 2007. I estimated the relationship between exit age and marginal tax rate using a classical OLS linear regression model. The estimate suggests that, holding all other factors constant, if marginal tax rate increases by 1 percentage point, average labor market exit age decreases by 1.88 months. Surprisingly, 51.74 percent of sample variation is explained by marginal tax rate alone. The sample constant is statistically significant, suggesting that if the hypothetical marginal tax rate were zero, the average labor market exit age in randomly chosen country from OECD sample would be 69.65 years. The sample constant is consistent with a prior theoretical expectations since it concurs with the "substitution effect" hypothesis that higher marginal tax rate leads to lower labor supply and fewer working hours.

The cost of early retirement in OECD countries
Source: T.T. Herbertsson & J.M. Orszag, The Cost of Early Retirement in OECD, 2001. OECD, Pensions at Glance, 2009.

Fiscal imbalances arising from unsustainable PAYG public pension systems in OECD countries cannot be assessed without a sufficient estimate of economic costs of unfunded pension liabilities. I approximated the cost of early retirement using Auerbach-Kotlikoff-Gokhale (1999) methodology that directly estimates the size of generational imbalances created by public social security systems. Large and rapidly unsustainable net pension liabilities occured in late 1980s. Van den Noord and Herd (1993) estimated the size of net pension liabilities in seven major OECD countries. The results suggest that continental European countries have had the largest net pension liabilities in terms of GDP. The size of pension liabilities in France and Italy had been about 2.5 times the size of their respective GDPs and twice the stock of the public debt.

Gokhale (2008) directly estimated fiscal imbalances arising from unfunded pension liabilities to current and prospective generations. The size of generational fiscal imbalance, as a share of the GDP, is extremely large and rapidly unsustainable in all OECD countries. In fact, the size of the imbalance is the most severe in Greece (875 percent of the GDP), France, Finland and the Netherlands (500 percent of the GDP) while it is more than twice the size of the GDP in all OECD countries except for the United States, Canada, Australia and New Zealand.

Fiscal imbalance in OECD countries
Source: J. Gokhale, Measuring Unfunded Obligations of European Countries, 2009.

I built the econometric model of public pension expenditure for a cross section of 23 OECD countries in 2007 to assess which variables might explained the cross-country variation in public pension expenditures. I've been aware of the possible drawbacks of choosing a cross-section model since it might be vulnerable to specification errors and the unbiasedness of regression coefficients. To account for possible specification bias, I conducted Kolmogorov-Smirnov, Shapiro-Wilk and Jarque-Bera normality tests. By performing normality tests, I have examined whether the normality assumption of normally distributed error terms is valid in the studied sample of 23 OECD countries considering error terms as identically and independently distributed.

In the set of explanatory variables that might yield consistent and robust estimates of regression coefficients I chose 10 various demographic, economic and institutional independent variables. Apart from demographic and economic variables, institutional variables are dichotomous since the institutional features can be captured by binary modes of choice. The dependent variable is the size of public pension expenditures in the share of the GDP.

The results suggest that public pension expenditures are positively correlated with the share of population aged 65 and older (0.746**), difference in life expectancy after age 65 between 1960 and 2005 (0.477*) and dichotomous variable for continental European countries (0.697**) where * and ** indicate the statistical significant of the sample correlation coefficient at the 5% and 1% level. The estimates suggests that the probability of higher pension expenditures in the share of the GDP is likely to occur in a continental European country known for a relatively large share of older population and a high difference in life expectancy after age 65 between 1960 and the present. On the other hand, public pension expenditures are negatively correlated with average effective retirement age (-0.475**), private pension funds as a share of GDP (-0.658**), labor market exit age (-0.523**), dichotmous variable for Anglo-Saxon countries (-0.544**) and a dichotomous variable for private pension system (-0.672**), where ** denotes the statistical significant of the sample correlation coefficient at the 1% level. Again, the estimates suggest that the probability of lower pension expenditure is likely to occur if a randomly chosen country from the OECD sample is Anglo-Saxon and has a high effective retirement age, large private pension funds as a share of the GDP, high labor market exit age and a mandatory private pension system. The coefficients suggest that in repeated sampling, the estimated sample correlation coefficient will include the true or correct population value in 99 percent of cases.

I conducted the econometric model which consisted of 8 regression specifications. I chose double-logarithmic model which yields direct elasticities as regression coefficients. However, I added two exceptions. In regression specifications 5 and 6, I chose a mixed specification mostly due to the inclusion of private pension funds (assets) variable in the regression specification. Unfortunately, but the share of private pension funds in Greece in 2007 equals 0 percent of the GDP which does not enable the researcher to apply double-logarithmic model as the basis of regression specification.

The estimates suggest that the share of population aged 65 and older is statistically singificantly positively related to the share of public pension expenditures in the GDP. Hence, the elasticity of public pension expenditures with respect to effective retirement age ranges from -1.465 to -4.935, suggesting that an increase in effective retirement age by an additional year leads to per unit increase in public pension expenditures by more than a unit increase in the share of the GDP. The coefficient of private pension funds is highly statistically significant. The elasticity of public pension expenditures with respect to private pension funds (as a share of the GDP) ranges from -0.34 to -0.38 and is statistically significant at the 1% level. The elasticity suggests that a 10 percentage point increase in the share of private pension funds reduces the share of public pension expenditures in the GDP, on impact, by 3.4-3.8 percent, holding all other factors constant. In addition, the estimates of coefficients for dichotomous variables suggest the following: the probability of higher public pension expenditures (as a share of GDP) is likely to occur in continental European countries with mandatory private pension system. Five estimates of dichotomous coefficients are statistically significant at the less than 10% level.

The significance of dichotomous (dummy) coefficients has been tested by beta coefficient analysis to rank the magnitudes of separate effects of explanatory variables on public pension expenditures as dependent variable. The results suggest that continental European countries are significantly more likely to face higher public pension spending in the share of GDP compared to Anglo-Saxon countries.

Earlier I mentioned the necessity of normality assumption in yielding robust, consistent and unbiased estimates of regression coefficients. The assumption has been questioned by conducting Kolmogorov-Smirnov test (K-S), Jarque-Bera test (J-B) and Shapiro-Wilk (S-W) normality test. The aim of the testing the normality assumption is to observe whether error terms distribute normally so that estimated test statistics, standard errors and confidence intervals are reliable. In setting test statistic, I set the normality assumption as null hypothesis. The results from K-S, J-B and S-W tests show that the null hypothesis cannot be rejected at 5% level, suggesting that the normality assumption is valid in the studied sample. Hence, test statistics, standard errors and confidence intervals are both valid and reliable.

The meaningful question to evaluate the prospects of the coming public pension crisis is how to reverse the growth of fiscal imbalances and reform public pension system as to avoid erratic generational indebtedness. Aging population and the growth of old-age dependency ratio trigger an enormous future burden on public finances in OECD countries. Lower fertility rate and population growth shall place an incurable burden on the stability of PAYG public pension systems. The estimates suggest that life-expectancy after the age of 65 is likely to increase by 2050 and gradually approach the age of 90 for both male and female. Assuming the effective retirement age is 65, the remaining life expectancy is 25 years or almost one-third of the average lifetime. As Alemayehu and Warner (2004) suggest: "Old-age health care costs thus will impose increasingly severe pressure on private finances and government coffers. Indeed, applying our age-specific estimates to the age distribution anticipated for the year 2030, we find that if nothing is done to alter current patterns of health care, per capita health care expenditures will rise by one-fifth due to population aging alone."

The long-term pension reform that aging societies of the West should undertake is a complementary measures of three key policy features of the reform.

First, the transition to fully-funded retirement savings accounts is the only viable and sound pension reform that can alleviate the damage generated by the growing fiscal imbalances. The theoretical foundation of the transition from public pension systems to fully-funded pension system has been laid down by Feldstein and Liebman (2001). The authors derived an algebraic solution which suggests that keeping a PAYG public pension system does not attenuate the persistence of a growing demographic pressure on the stability of public pension system. As I discussed earlier, PAYG system crucially depends on three key assumptions: high fertility rate, very low share of population older 65+ and high population growth. These assumptions are incompatible with actual demographic parameters and, hence, OECD countries should undertake a drastic transition towards fully-funded pension systems based on individual savings accounts. Otherwise, the growing demographic pressure will inevitably result in the exponential growth of generational debt, creating an enormous deadweight loss for current and prospective generations.

Fully-funded pension system is based on the premise of investing pension contributions into the capital market, earning a compound interest over time. The stock of individual's lifetime earnings is paid in the form of annuities upon individual's withdrawal from the labor market. In addition, there is a growing disparity between the implicit return of PAYG public pension system and real rate of return in the capital market. Under realistic assumptions, such as that the marginal product of capital (MPK) is below the welfare-maximizing level and the real rate of return exceeds the implicit return from PAYG system, fully-funded pension system would not create a deadweight consumption loss to the working-age population. In fact, Feldstein and Liebman (2001) derived an analytic solution for the transition to fully-funded pension system in which the transition induces a short-term consumption loss in the next period while, at the same time, it creates a geometrically-growing future consumption for both retired and working-age population.

The only remaining question is whether the real rate of return would compensate the consumption loss of working-age population and, hence, increase the stock of future consumption to all generations. According to Feldstein and Liebman (2001), assuming 6.5 percent inflation-adjusted rate of return, the payroll cost of fully-funded pension system would represent only 27 percent of the payroll cost incured under PAYG public pension system. Tax rate, required to bear the cost of current stock of pension liabilities is 12.4 percent respectively.

According to Congressional Budget Office, the average real rate of return for large-company stocks between 1926 and 2000 is 7.7 percent, 9.0 percent of small-company stocks and 2.2 percent for long-term Treasury bonds. Feldstein (1997) estimated that PAYG implicit rate of return is 2.6 percent.

Assume an individual wants to maximize the lifetime earnings in the capital market. An individual is offered 2.6 percent implicit return from PAYG system. The individual enters the labor market at certain age, say 25, and intends to retire upon the age of 65. Assume he invests $10.000 annually in the capital market to create retirement annuities upon labor market withdrawal. Assuming the implicit rate of return (2.6 percent), the stock of overall annuity would be 10 times the initial investment in 90 years. Assuming the average long-run real rate of return from large-company stocks (7.7 percent), the the overall annuity would be 10 times the initial stock of investment in 31 years. Therefore, the individual would reach the desired level of lifetime earnings at the age of 56 or 9 years before the targeted retirement age.

I assumed the distribution of lifetime investment portfolio is weighted average of availible asset types: large-company stocks (33 percent), small-company stocks (19 percent), long-term corporate bonds (20 percent), long-term Treasury bonds (20 percent) and 3-month Treasury bills (8 percent). According to the average annual real rates of return in the United States (1926-2000), I calculated the weighted average real rate of return (5.247 percent). Investing $10.000 annually at the age of 25 would buy $100.000 annuity at 5.247 real rate of return in 45 years (the age of 70) compared to 90 years (the age of 115) under the PAYG implicit rate of return (2.6 percent). Of course, the time to buy the annuity would shift alongside the changing composition of portfolio.

In addition, OECD countries should immediately increase the effective retirement age. I believe the solution suggested by Gary Becker is both meaningful but sustainable in reversing the growth of generational debt. Becker (2010) suggested "One simple and attractive rule would be to raise retirement age by an amount that makes the ratio of years spent in retirement to years spent working equal to the ratio that existed at the beginning of the social security system."

When President Roosevelt signed the notorious Social Security Act in 1935, the normal retirement age was 65. However, life expectancy after the age of 65 was significantly lower than is today. In 1940, average life expectancy after 65 in the U.S was 13.7 years. In 2006, it stood at 18.6 years, according to OECD. In 1935, the average life expectancy at birth in the United States was 61.7 years. We assume that individuals in 1935 worked for 35 years and spent 12 years in retirement. The ratio is thus 0.4 (12/ 35=0.34). Today, if individuals retire at the age of 65, they can expect further 18.6 years in retirement. To equalize the ratio to the 1935 level, (18.6/x=0.34), individuals should spend 54.7 years working. The estimate time is an equivalent measure of years required to spend working if PAYG public pension system is left intact. Assuming the individuals enter the labor market at the age of 25, then the expected effective retirement age is the age of 80.

In the long run, PAYG public pension system is unsustainable since demographic parameters do not suffice the assumptions under which the PAYG system is possible without distortions of labor supply incentives. The future of OECD countries will be marked by aging population, lower fertility rates and a growing demographic pressure on public finances. Without bold and decisive pension reform, OECD countries will experience increasing pension deficits and, hence, an explosive growth of generational indebtedness.

Parametric pension reforms are not a substitute for the postponement of paradigmatic pension reform. Thus, implementing the transition to fully-funded pension system essentially requires higher effective retirement age. A comprehensive pension reform cannot be made possible without these measures. At last, but not least, the major challenge in the systematic pension reform in OECD countries to address the burden of global aging, is whether political courage will withstand the pressure of interest groups to maintain the status quo of early retirement incentives. Nonetheless, eliminating early retirement incentives is the essential step towards creating retirement system without perverse incentives to retire too early. Unless political leaders encourage a transition to fully-funded pension system, OECD countries will be unable to withstand the deadly consequences of an enormous generational indebtedness.

The economic future of Ireland

The economic and financial crisis of 2008/2009 hit Ireland heavily. The asset price bubble and the subsequent deflation have added to the uncertain macroeconomic outlook. How did the country went from the times of the "Irish miracle" to the prolonged economic slowdown? Following the beginning of the 2008/2009 economic and financial crisis, Ireland was hit by an unprecedent economic slowdown. In 2008, the GDP declined by 3.0 percent on the annual basis. In 2009, the GDP further declined by 7.1 percent in real terms. The unemployment rate increased to almost 12 percent.

Prior to the outburst of the economic crisis, Ireland enjoyed stable and predictable levels of public debt. In 2007, the country was known for having stabilised the public debt at 25 percent of the GDP - the lowest level of any Western European country. In 2009, the debt-to-GDP ratio increased to 64 percent of the GDP. Once known as the sick man of Europe, Ireland's economic policymakers have implemented a set of fiscal policy measures aimed to boost the long-term economic growth and abolish the economic policy based on the state intervention, high tax rates on labor and capital and export-led growth.

Eversince the 1960, Ireland pursued a soft version of industrial policy targeted at the promotion of inward foreign direct investment and the education of highly skilled workers. In addition, Ireland reduced the corporate income tax rate to 12.5 percent and provided a thorough technical assistance and to multinational companies located in Ireland. Indeed, U.S. multinationals such as Microsoft, Dell and Intel were encouraged to locate in Ireland mainly because of its geographic proximity to key European markets, skilled English-speaking workforce, membership in the EU, relative low wage level and favorable corporate taxation.

In early 1990s, the results of a precise set of economic policies were spectacular. By the end of 2006, the unemployment rate dropped to 4.6 percent from 18 percent in early 1980s. Between 1992 and 2005, Irish GDP increased by an average of 6.9 percent while the investment grew by 8.6 percent on the annual basis. The largest contribution to GDP growth was domestic demand (5.3 percentage point). Hence, Ireland's public finance enjoyed a favorable outlook mainly due to the rapid decline of debt-to-GDP ratio from 1980s onwards, and from a relatively low demographic pressure on the budgetary entitlements.

During the Irish boom, Irish banking and financial sector were highly dependant on the wholesale funding. Due to largely positive macroeconomic outlook from 1990 onwards, Irish banking sector received high and consistent credit ratings from agencies such as Moody, S&P and Fitch. In turn, the reliance on fragile wholesale funding resulted in overleveraged balance sheets. After the failure of Lehman Brothers in September 2008, the short-term outlook on Irish banking sector signalled a significant rise in credit-default swaps which raised concerns over the ability of banks to provide the wholesale funding for a mountain of short-term debt liabilities. And since the overleveraged balance sheets downgraded the outlook on Irish banking sector, the institutional investors demanded higher risk premium to extend the funding channel to the Irish banks.

The Directorate Generale for Economic and Financial Affairs of the European Commission downgraded the macroeconomic forecast of Irish GDP growth. By the end of 2009, the economic activity plummeted by 7.1 percent. The housing market crash was largely a result of the asset price bubble channeled through the overinvestment in the construction sector which represented 12 percent of the GDP. Nothing could explain the deflationary pressures in the aftermath of the financial crisis than excessive housing prices during the pre-crisis Irish economic boom. After 2008, Ireland's household savings rate increased to the level above 10 percent which is a result of the adjustment in the household balance sheet. In fact, between 2001 and 2007, the share of household debt in the GDP nearly doubled.

Meanwhile, the mountain of liabilities in the Irish banking and financial sector raised the concern over its solvency. The Irish Government immediately facilitated a bailout plan for the troubled banking sector. Consequently, the large budget deficit resulted in excessive debt-to-GDP ratio which grew by 39 percentage points between 2007 and 2009. In the annual European Economic Forecast (Spring, 2010), the European Commission estimated that by the end of 2011, the debt-to-GDP ratio could reach as high as 87.3 percent. while the cyclically-adjusted government balance is estimated to increase up to -10.2 percent of the GDP. The contraction of domestic demand which, by all measures, is the main engine of Ireland's economic growth led to a rapid increase in the unemployment rate which increase from 6.3 percent in 2008 to 11.9 percent in 2009. By 2011, the European Commission forecast that the unemployment rate is expected to further increase by 1.5 percentage point compared to 2009. In World Economic Outlook, the IMF estimated that the unemployment rate in Ireland would increase by 1.1 percentage point by the end of 2011. In 2009, Ireland experienced net outward migration for the first time since 1960s in the wake of expected 13.8 percent unemployment rate in 2010.

The macroeconomic forecast for 2011 is favorable. The European Commission upgraded GDP growth estimate to 3 percent. Meanwhile, the investment is expected to increase for the first time since the 60 percent cumulative decline of the construction sector. The positive contribution of net exports to the gradual narrowing of the current account deficit could be an important measure to alleviate the rising pressure over debt-to-GDP ratio. On the other hand, Ireland's Department of Finance revised the macroeconomic forecasts and estimated that by the end of this year, the GDP would grow by 1 percent on the annual basis.

The essential measure of Irish economic recovery is the retrenchment of wage rates in the public sector and the adjustment of public sector wages to the cyclical dynamics of economic activity to prevent the possibility of excessive inflationary pressures in the course of economic recovery. Current measures of retrenching public sector wages successfully anchored the inflationary expectations. According to the IMF, the annual inflation rate is estimated to peak at nearly 2 percent by the end of 2015. The falling wage rates in the private sector could induce the reallocation of resources in the tradeable sector, further adding to the contribution of net external trade to the GDP growth.

The key measures to alleviate the consequences of economic and financial crisis in both real and financial sector are the immediate narrowing of Ireland's excessive budget deficit and public debt in the share of GDP. High public debt is mainly the result of government capital injection into Anglo-Irish Bank which represents about 2 percentage points of net deficit increase in 2010. The entire consolidation package represents 2.5 percent of the GDP.

Deutsche Bank recently published Public Debt in 2020 and estimated the levels of public debt by the end of that year for both advanced and emerging-market economies. The analysis by Deutsche Bank predicted the effect of a combined negative shock in real interest rate, primary government balance and real GDP growth. If the combined shock of all three variables were to change by about one-fourth standard deviation from the estimated growth rate, the public debt in 2020 would reach 154 percent of the GDP. If the combined shock of all three variables increased by one-half standard deviation from the baseline estimates, the public debt in 2020 would increase to 197 percent of the GDP. The difference in the estimated increase is due to higher intensity of the combined shock. In addition, to restore the debt-to-GDP ratio to pre-crisis level, Ireland would be required to increase the primary government balance to 6 percent of the GDP.

Given the enormous magnitude and burden of public debt and overleveraged corporate and financial sector, the immediate facilitation of measures to alleviate the public indebtedness is necessary. Ireland's economic future is constrained by the persistence of budget deficit which adds to the future burden of public debt. Prudent efforts to reduce the burden of both debt and deficit are of the essential importance. Nevertheless, Irish policymakers should not neglect the economic policies that created the Irish miracle as well as the policy errors that caused the deepest economic decline in Western Europe during the 2008/2009 economic crisis.

Friday, October 01, 2010

Economic growth and democratic institutions

Professor William Easterly recently presented (link) an intriguing empirical evidence on the relationship between nation's politics and economic growth. In particular, professor Easterly presented data on long-run economic growth and the scope of democracy for a majority of countries between 1960 and 2008. Professor Easterly identified that the highest-growing countries in the world were those with autocratic political regimes. Among ten highest-growing economies between 1960 and 2008, all of them, except for Cyprus, have been characterized by hybrid and autocratic political regimes. On the other hand, ten countries with the lowest growth rates of real GDP per capita between 1960 and 2008 were equally known for authocratic political systems or flawed democracies.

Presumably, the evidence bodes against the recent prediction by Dani Rodrik that authoritarian political regimes ultimately create economic systems vulnerable to external shocks and structural change, thus hampering the prospects of structural change as a neccessary condition for economic development.

To estimate the general pattern of the relationship between economic growth and the nature of political system, I reviewed real per capita GDP growth rates between 1970 and 2007 for a group of 134 countries across the broad spectrum of different levels of GDP per capita. Based on Summers-Heston dataset of real GDP per capita growth rates (link) between the stated time period, I estimated average rates of growth of GDP per capita and collected data from Economist Intelligence Unit on the level of democracy across the world in 2008 (link). The intuition behind this approach is the identification of endogenous and casual direction between the two variables. From the theoretical perspective, it is nonetheless difficult to establish a relationship between the form of government and long-run economic growth. There are at least two possible directions of casuality.

First, the underlying assumption of the relationship could be that systemic changes in political environment are essential to the structural change and, hence, are the main mechanism behind the enforcement of constitutional changes and public policies. The assertion of the underlying theory is that autocratic and authoritarian political system hinder structural changes and the establishment of institutions and democratic governance that is crucial for economic growth. This particular view has been asserted by Dani Rodrik (link), Andrei Shleifer, Florencio Lopez de Silanes and Rafael La Porta (link). While Dani Rodrik's perspective heavily relied on the importance of institutions for long-run economic growth, Shleifer, Lopez de Silanes & La Porta captured the essence of economic development in the legal origins of nations.

Second, the casuality in economic growth and political system could also stem in the opposite direction. The basic underlying assumption could be that higher rates of economic growth encourage systemic changes in the political system and enable the adoption of democratic institutions. The notion of economic growth as the engine of democratic changes has deserved a strong empirical support.

Robert Barro's analysis of long-run economic growth across the world (link) has examined the relationship between the level of democracy and long-run economic growth rate. The empirical evidence suggests a non-linear, inverted-U relationship between democracy and 10-year growth residuals, both coefficients in partial quadratic equation and the partial correlation coefficient being statistically significantly different from zero.

The notion would suggest that as countries depart from a low level of real GDP per capita, the adoption of democratic institutions accelerates economic growth but only up to some point. After the tipping point, the economic outcome of further democratization results in lower growth of real GDP per capita, partly because a high level of democracy tends to promote public policies that diminish growth prospects such as higher tax rates on labor and capital and the redistribution of income, all of which exert a somewhat negative effect on productivity growth and incentives for labor supply and investment.

The first table portrays the distribution of real per capita GDP growth rates across 134 countries between 1970 and 2007. The distributive pattern resembles the properties of normal distribution curves. In fact, the estimated coefficients of skewness and kurtosis suggest a rather very mild departure from the assumption of normality which is of the high importance, especially in testing hypotheses about the effects of explanatory variables on long-run growth dynamics. The normality assumption of normally distributed errors was not tested via normality tests.

Ten highest growing countries in terms of real GDP per capita between 1970 and 2007 are Equatorial Guinea (8.39 percent), Taiwan (5.98 percent), China (5.97 percent), St. Kitts & Nevis (5.49 percent), Botswana (5.45 percent), Bhutan (5.38 percent), Maldives (5.38 percent), Hong Kong (5.37 percent), Macao (5.30 percent) and Singapore (5.29 percent). In real terms, the estimated average real per capita GDP growth rates suggest that it took only 13 years for Singapore's real GDP per capita to double and 21 years to triple. In China, where the estimated average growth rate exceeded Singapore's growth rate only by 0.69 percentage point, it took roughly 11 years for real GDP per capita to double and only 19 years to triple. In the lower tail of growth distribution are mostly countries from Sub-Saharan Africa such as Democratic Republic of Congo, Liberia, Somalia, Central African Republic and Niger. Average real growth rates of GDP per capita of these countries were negative. The negative average real GDP per capita growth rate occured in 11 percent of country observations.

Distribution of economic growth across 134 countries between 1970 and 2007
Source: own estimate based on Summers-Heston dataset

The following graph illustrates the relationship between long-run average growth rates and the level of democracy in 2008 for the entire sample of 134 countries. The attempt to analyze the effect of democracy level on long-run economic growth is based on the notion that democratic institutions elevate economic growth in the longer run. The estimated slope coefficient (0.2277) suggest that a one-point increase in democracy index increases the average long-run per capita GDP growth rate by 0.2277 percentage point controlling for other factors.

Although the cross-country variation in the level of democracy explains only about 9 percent of growth rate variance, and even though the direct effect of democracy on economic growth seems minor and almost non-existent, the estimated sample regression coefficient is statistically significant at 5 percent level. It suggests that the effect of democracy on growth is persistant and evident in the particular sample.

Democracy and average long-run growth rates in a sample of 134 countries
Source: own estimates

Hence, to account for different degree of variation in average real GDP per capita growth rates, I divided the sample into quartiles. The goal of the pursued empirical strategy is to see whether the difference in variance composition between countries with similar growth rates persists. I divided the total sample into four groups: high growth performers (average growth rate higher than 3 percent) moderate growth countries (average growth rate below 3 percent and above 2.05 percent) and low growth countries (average growth rate below 1.09 percent). The next graph shows the relationship between democracy and average real GDP per capita growth rate in high-growth countries between 1970 and 2007. The parameters suggests a different relationship. The estimated slope coefficient is negative (-0.1638), suggesting that a one point increase in democracy index decreases the average real GDP per capita growth rate by about 0.1638 percentage point.

The share of variance explained by the democracy variable increased by 22.5 percent. In the statistical sense, the effect of democracy on economic growth in high-growth countries has been more powerful compared to the total sample. The estimated slope coefficient is statistically significant at 5 percent level. I also estimated beta coefficient (-0.338) to account for the effects of standard deviation increase on the average growth rate in real GDP per capita. The estimated beta coefficient suggests that a one standard deviation increase in democracy level (2.4 points) would, on impact, decrease the average real GDP per capita growth rate by 0.338 standard deviation or 0.394 percentage point in real terms.

From a theoretical perspective, the enforcement of democratic policies in high-growth countries would have a minor negative effect on economic growth, holding all other factors constant. Surprisingly, authortarian regimes previal in 44 percent of countries in the high-growth sample. Thus, the hypothetically negative effect of democracy on economic growth is evident but it is far from significantly negative.

Democracy and average long-run growth rates in high-growth countries
Source: own estimates based on Summers-Heston and EIU datasets

The next graph portrays the relationship between democracy and average real GDP per capita growth rates in low-growth countries. Contrary to the sample estimate in high-growth country group, the effect of democracy on real GDP per capita growth rate is positive and persistent. The correlation coefficient is positive and moderate (0.458) and statistically significant at 1 percent level. The beta coefficient (0.458) from the regression specification suggests that a one standard deviation increase in democracy level (cca. 1.497 points) would raise the average real GDP per capita growth rate on impact by 0.458 standard deviation or 0.382 percentage point, ceteris paribus. In fact, the variability in level of democracy explains 21.1 percent of the variance of average per capita GDP real growth rates. The estimated slope coefficient is statistically significant at 2.1 percent level and 0.6 percent level, suggesting a very low probability of rejecting the null hypothesis and a strong influence of democratic institutions on economic growth in the long run.

Democracy and average long-run growth rates in low-growth countries
Source: own estimate based on Summers-Heston and EIU datasets

In the next subsample, I jointly added high-growth and low-growth countries in the single sample and changed the casual direction. The underlying assumption is that democracy level is endogenously determined by the long-run average real GDP per capita growth rate. In real terms, I assumed that the public choice of political institutions across the world depend on the real GDP growth rate. Hence, I estimated the relationship by including the squared term in the regression equation. The estimated slope coefficients suggest a typical inverted-U relationship between real GDP per capita growth rate and the level of democracy. The real GDP per capita growth rate alone explains 30.6 percent of the cross-country variaton in the level of democracy. Intuitively, the results suggest that there exists an optimum level of real GDP per capita growth that maximizes the level of institutional democracy.

Differentiating the conditional expectation function of the level of democracy with respect to the real GDP per capita growth rate yields the partial derivate dy/dx = -(ß2/2ß3). Plugging the two coefficients in the partial derivate yields 3.65. Thus, the growth rate of real GDP per capita that maximizes the level of institutional democracy is 3.65 percent. Hence, both coefficients are statistically significant. The p-values are 0.000 suggesting a zero probability of rejecting a null hypothesis when it is, in fact, true - and a strong predictive influence of both variables on the expected level of democracy.

The effect of long-run economic growth on democratic institutions in high-growth and low-growth countries
Source: own estimates based on Summers-Heston and EIU datasets

Countries with the comparable growth rate are Iceland, Ireland, Trinidad & Tobago and Spain. Except for Trinidad & Tobago, none of these countries is either flawed democracy or an authoritarin political regime. Therefore, the expected level of democracy is low in countries where the average growth rate of GDP per capita is either very low or negative or very high.

Hypothetically, the conditional pattern of real per capita GDP growth supports the notion that the highest-growing countries in the 20th century such as Singapore, Taiwan and Botswana had a relatively low level of democracy and a significant degree of political authoritarianism. In addition, countries with the lowest growth rate of real GDP per capita such as Liberia, Sierra Leone and Somalia were also authoritarian political regimes. The predictive power of the regression equation is reasonably high since more than 30 percent of the variance of the level of democracy is explained by a non-linear shifts in the long-run average real GDP per capita growth rate.

Democracy is a controversial question of the modern theory of economic growth. Indeed, the empirical evidence suggests that the highest growth rates were achieved in those countries with a considerable degree of political dictatorship. However, the lowest long-run growth rates of real GDP per capita were achieved by countries in which political dictatorship prevails. The pattern suggest that the quality of institutions such as the rule of law, judicial independence and a constitutional democracy complement the significance of human capital which is the essential engine of long-run economic growth.

The most important growth engine of the highest growing countries such as East Asian tigers and Ireland has been the emphasize on human capital that resulted in a high level of knowledge intensity and high productivity growth rate. These countries were known for heavy doses of state interventionism aimed towards the implementation of industrial policy conducive to economic growth. However, the conclusion should be taken with caution. Political dictatorship or authoritarianism were detrimental to least-developed countries since it encouraged predatory political behavior and resulted in the political environment with a complete absence of the rule of law, judicial independence, protection of private property rights, institutional integrity and constitutional democracy.

The question which set of growth policies is essential to high long-run growth of real GDP per capita involves two answers. First, the primacy of institutional quality alongside the investment in human capital is by far the most important engine of long-run economic growth. Without first-class institutions and human capital, the vicious circle of poverty and social deprivation for less developed nations can be endless. And second, the components of constitutional democracy such as electoral rights and pluralism, good functioning of government, high level of political culture and civil liberties can deliberately increase the prospects of economic growth.

However, if the power of state is left unrestrained by the absence of the rule of law and a coherent set of checks and balances on the coercive strenght of redistributive interest groups, even a high level of democracy would not alleviate the persistence of poverty and weak structural indicators. On the contrary, it would only worsen the prospects of long-run economic growth.