Shortsightedness for aging leaders

Shortsightedness for aging leaders

The effect of an autocrat’s age on his country’s growth

Abstract

Using a cross-section approach, I test whether the age of a political leader affects their countries’ economic growth rates, with a sample of 140 autocrats from 85 countries from 5 continents in the period 1951-2012. Furthermore, I test whether a degree of personalism in an autocracy affects growth, and I compare the growth rates from the autocrats’ final three regime years to the average growth rates from their regimes. I find a negative correlation between personalism and growth, but the findings on the effect of age are inconclusive.

Student name: Jeroen Thomas Baars Student number: s2160773

Student e-mail: j.t.baars@student.rug.nl

2

1. Introduction

The question why some countries experience more GDP growth than others is one that researchers have studied for a long time. Besides the economic characteristics that are described by Solow (1956) and tested by Mankiw, Romer, and Weil (1992), it also depends on the political situation in a country. While autocracies typically experience less growth than democracies (see, for example, Faust (2007) ), we can still observe vast differences between autocracies. Jones and Olken (2005) suggest that personal characteristics of the leader may also matter, as they discovered that the replacement of leaders causes structural breaks in countries’ observed growth patterns.

As described by Olson (1993), we can roughly divide all autocrats into stationary bandits and roving bandits. Roving bandits who expect only a brief tenure might confiscate assets whose tax yield over their tenure does not exceed their total costs. A stationary bandit, who expects a long regime, has a lower time preference and is more willing to invest in the country’s capital structure. Intuitively, we see that a country with a stationary bandit will likely realize higher growth than a country with a roving bandit. However, what are the characteristics by which we can distinguish between the two? A possible answer might be the leader’s age.

In this thesis I investigate whether the age of a political leader affects the country’s economic performance in terms of growth. This will be tested on a sample of countries with autocratic regimes, i.e. where one could connect the economic policy to one single person. I will also test whether the degree of political power in the hands of one physical person harms economic growth, and whether leaders respond to their older age with more predatory behavior (i.e. lower growth rates) towards the end of their regime, e.g. because they have additional knowledge about the approaching end.

3

2. Literature review

Explaining differences between countries with respect to economic growth is an important goal in the field of economics. The foundation in this field was built by Tinbergen (1942), Tobin (1955), Solow (1956), Swan (1956), and Meade (1961) with their work on growth theory, most notably the neoclassical growth model which is commonly known as the Solow-Swan model. In this model, output growth is explained as a function of capital accumulation, labor growth and technological growth. This model was empirically tested by Mankiw, Romer, and Weil (1992). They tested the Solow-Swan model and added the extension of human capital. They concluded that the main drivers of economic growth are physical capital, human capital, and increases of the labor force. Their model was able to explain much of the variation of growth rates observed by the countries in their sample.

But, do leaders matter? That is the central question that Jones and Olken (2005) ask. They observed vast differences in growth rates within countries over time, especially in developing countries, and wondered whether the driving force behind these changes could be the political leader. To test this hypothesis, they studied sudden changes in political leaders and found that countries experience persistent growth rate changes after an unexpected leadership transition. But the interesting follow-up question is which traits of leaders’ personalities and backgrounds can cause these changes in their economic policy.

Hayo and Neumeier (2013) used panel data from 29 OECD countries in order to explain the differences in government budget deficits. They found significant evidence that political leaders who grew up in a family of lower socio-economic status are more prone to run a government deficit. As the main explanation for this causation, they mention that people of a lower socio-economic status will likely have a higher time preference, since they are not used to be able to save and rather feel the need to spend everything whenever they can.

In a similar approach, Besley et al (2011) observed that highly educated leaders obtain higher economic growth rates, Moessinger (2014) found that a finance minister’s experience decreases the countries’ debt-to-GDP ratio, and Horowitz et al. (2005) found that older leaders behave more aggressively in military conflicts.

4

3. Theoretical framework

Similar to Jong-A-Pin and Mierau (2011), I consider an executive in two periods, t and t+1, which can be referred to as the present and the future. The executive rules in the present with probability 1, and rules in the future with probability 1-q, where 0 < q < 1 is the probability that he either dies or is overthrown. The value for q will rise exponentially, as the effect of age on mortality is not linear (Dolejs and Maresova, 2017). The executive is interested in leaving an economically healthy country if he expects to still be leading the country in the future, i.e. if q is relatively low. This corresponds to the ‘stationary bandit’, described by Olsen (1993). An executive does not have perfect knowledge about the value of q, since he does not know when he will die (of natural causes) or be overthrown within the current period. He can, however, make a prediction based on his health, lifestyle, popularity, and of course mainly his age, of which he does have perfect information.

The power of the executive, i.e. the degree to which he can single-handedly determine the country’s economic policy, is given by 0 < p < 1. For an all-powerful dictator, as in Jong-A-Pin and Mierau (2011), p = 1.

The production sector of the economy is characterized by a linear Cobb-Douglas production function with a normalized labor force. This production function is given by

𝑌𝑡 = 𝐴𝐾𝑡, (1)

where 𝑌𝑡 represents the aggregate output at time t, A is the (exogenous) state of technology, and 𝐾𝑡

is the capital stock at time t.

In period t, the executive must decide the fraction of 𝑌𝑡 that he extracts from the economy for

consumption purposes. The maximum fraction that the executive may extract is limited by his power p. The remaining fraction of 𝑌𝑡 will be invested in the capital stock, i.e. 𝐾𝑡+1 will grow in order to

increase 𝑌𝑡+1. Thus,

𝑌𝑡 = 𝐶𝑡+ 𝐼𝑡. (2)

The capital level in period t+1, , is given by 𝐾𝑡+1 = (1 − 𝛿)𝐾𝑡+ 𝐼𝑡, where 0 < δ < 1 is the rate of capital

depreciation. Assuming full capital depreciation (δ = 1) in every period t implies 𝐾𝑡+1= 𝐼𝑡. If we

5 Combining (1), (2), and (3), we can see that

𝐶𝑡+1 = 𝐴(𝐴𝐾𝑡− 𝐶𝑡) − 𝐼𝑡+1 , (4)

which displays that the executive’s future consumption is negatively impacted by his present consumption.

The connection between the levels of output in respective periods is given by 𝑌𝑡+1 = (1 + 𝑔) 𝑌𝑡, where

𝑔 is the relative growth rate, measured as a percentage of 𝑌𝑡, which depends on the executive’s

decision made in (2), as follow’s from (3). When the executive chooses to give up some consumption in the current period and invest it in extra capital for the next period, there will be a higher value for 𝑔.

A high level of consumption in period t will increase the executive’s utility in period t, but it will decrease his future consumption and thus his future utility.

The discounted life-time utility function of an executive at any period t is given by: 𝛬𝑡 = p ln(𝐶𝑡) + (1-q) p ln(𝐶𝑡+1), (5)

where 𝛬𝑡 is the discounted level of utility in period t, p is the executive’s economic power, C represents

consumption, and q is the possibility that the executive will not rule in the next period.

As follows from (5), an executive’s utility depends on both current consumption and future consumption, with current consumption being restricted by only his power level and future consumption being restricted by his power level but also discounted by the possibility of losing his position in period t and diminished by the rate of inflation. Since consumption depends strongly on the level of output, the executive faces a decision problem where he must divide aggregate output as described in (2) in order to maximize his utility function (5). This decision problem will be the core of my research, specifically how the decision is influenced by the executive’s age, summarized in the following hypotheses:

6 where 𝑔𝑎𝑣𝑒𝑟𝑎𝑔𝑒 is the average annual GDP growth rate during the leader’s regime, and 𝑔𝑓𝑖𝑛𝑎𝑙 is the

average annual GDP growth rate during the final three years of the leader’s regime. The period of three years is chosen since it is relatively short, but not so short that any observed effects can be contributed solely to an ‘off-year’. By using an average value over multiple years, I attempt to reduce the impact of economic fluctuations in the observed period. However, would I take too long a period, it would not truly be considered the end of a leader’s reign, as several leaders in the sample have had a regime of no more than five years.

H.1 follows from the intuition that, as leaders age, their mortality rate increases, i.e. a higher value for

q will be entered into (5), thus future consumption will be more heavily discounted. Therefore, as a

utility-maximizing individual, the leader will increase his consumption in period t, which, according to (4), decreases his consumption for period t+1, since less capital will be accumulated. Therefore the country will achieve less economic growth than with a younger leader.

H.2 implies that more powerful leaders, i.e. those where a higher value for p enters (5), are able to extract more funds from the economy. For a p-value of 0, i.e. a perfect democracy, the leader is indifferent between current and future consumption. As the value increases, the leader has more to gain from balancing present and future consumption well, which implies that the discount factor q plays a larger role, thus the leader is less willing to invest in capital accumulation.

H.3 follows from the intuition that q rises as the executive ages, but not exponentially. As the effect of age on economic policy grows as the executive ages, I expect to see a difference between the policy enacted in the final part of his regime compared to the overall policy. This effect may be increased if the executive has more knowledge about the value for q, e.g. due to a mortal illness.

7

4. Methodology

The hypotheses will be tested in a model that is similar to the approach by Mankiw, Romer, and Weil (1992). I estimate a cross-section model in order to compare the regimes of the considered leaders. The estimated model can be written as

𝐺𝑗 = 𝑋𝑗𝛽 + 𝑍𝑗𝜃 + 𝜀𝑗 ,

where 𝐺𝑗 is the dependent variable, either the average annual GDP growth rate for H.1 and H.2, or

the difference between the average annual GDP growth rate during a regime and the average annual GDP growth rate during the final three years of the regime in H.3, achieved by executive j.

𝑍𝑗 is a vector of control variables, specific to the executive or his country during his regime. This

includes the executive’s tenure and several macroeconomic variables for the country during the executive’s regime, which are the average annual population growth, .

𝛽 and 𝜃 are vectors of regression parameters and 𝜀𝑗 is the random error term.

𝑋𝑗 is a vector of explanatory variables that correspond to the hypotheses, specifically:

- 𝑋_𝑗𝛽 = 𝑒𝑥𝑎𝑔𝑒_𝑗∗ 𝛽₁+ 𝑒𝑥𝑎𝑔𝑒2

𝑗∗ 𝛽2+ 𝑙𝑖𝑛𝑑𝑒𝑥𝑗∗ 𝛽3 ,

where exagej is executive j’s age during the final year of his regime and lindexj is the personalism index

awarded to executive j. As can be seen, and as explained in the theoretical framework, the executive’s age will be used as a method to estimate q, which implies that I mainly focus on the possibility that an executive’s term ends with his (natural) death, as this is the only universal, quantifiable indicator that will predict the end of his reign, or at least the only indicator that the executive has perfect knowledge of. Since the executive’s mortality rate does typically not increase linearly with age but rather exponentially (Dolejs and Maresova, 2017), the marginal effect of age on the dependent variables is (𝛽1+ 2𝛽2∗ 𝑒𝑥𝑎𝑔𝑒) rather than only 𝛽1, which is why I will use both the executive’s age and the square

of his age as explanatory variables, rather than just exage.

For all hypotheses, I will test a null-hypothesis of no effect1 _{at a 95% significance interval, that I will}

reject if there is sufficient evidence of a correlation. If the coefficients are not statistically significant, they can still be interpreted, but it will not be possible to draw economic conclusions from them.

1 _{E.g. for H.1, the null-hypothesis is} ∂g

8 In accordance with H.3, I will also apply the estimation strategy described above to a transformed model, where the response variable is (gdpgrowthaverage - gdpgrowthfinal) where gdpgrowthfinal is the

average annual GDP growth rate during the final three years of the regime. A similar transformation has been made with the explanatory variables population growth and capital accumulation. This is to test whether executives that may have additional knowledge about the true value of q, e.g. in the case of a disease or threats about an impending coup, will behave extra predatory.

After all, it may be possible that, towards the end of their regime, executives will expect that the end is near and respond by behaving more predatory. In terms of the theoretical framework, this implies that the executive has a higher value for q, which is the discount factor to which future consumption enters the executive’s utility function. If an executive knows that there is a high value for q, e.g. because he is of old age or because he has a terminal disease, he will receive less utility from future consumption and will therefore behave more predatory. This implies that he will extract more present consumption, which is not discounted by q, from the economy and thus realize a lower growth rate. As a White test indicated that heteroskedasticity is an issue in the datasets, the model will be estimated with robust standard errors.

5. Data

The hypotheses described in section 3 will be tested with a dataset of 140 political leaders from 85 (former) autocracies from five continents. All of the considered leaders have led their country for a minimum of five years in the period between 1961 and 2012. The base for this dataset is the data by Magaloni et al. (2013), who characterized all political leaders from the world2 _{in the years 1950-2012}

as either a democracy or an autocracy. Magaloni et al. (2013) describe an autocracy as anything that fails to meet one or more of these criteria:

1. A civilian government (as opposed to military or royal court) provides the main source of policy making.

2. Political leaders form multiple and competitive parties, and the parties interact and run the government through a legislature.

3. The executive is institutionally constrained or checked by other parts of the government.

4. Elections are used to select the political leadership, and they are largely open, competitive, and free and fair.

2 _{In fact, they only include countries that have a population greater than 500,000, which excludes a handful of}

9 Furthermore, they distinguish between military autocracies, monarchies, single-party autocracies, and multi-party autocracies. Since all of these regime types imply that there is a leader who is able to, to some degree, influence the country’s economic policy, I include all four regime types and do not differentiate between them.

In the dataset by Magaloni et al. (2013), an executive is considered responsible for that year’s policy if he has ruled for more than six consecutive months within that year. This ensures that every combination of country and year is unique.

Magaloni et al. (2013) also define the variable ‘personalism’, which they define as either ‘the degree of constraints imposed on the executive’ or ‘the degree to which a particular regime is associated with a single leader’. One of two ways to express this is with the personalism index, which ranges from 0 (perfect democracy) to 1 (perfect autocracy). Since perfect democracies are not considered in my research, the value of 0 is not observed.

Data on the leaders’ age was extracted from Lentz (2014), who listed all the heads of states and governments since 1945. In the case of unclarity about a leader’s regime, I also consulted Archigos by Goemans et al. (2009), a dataset of all known political leaders with details about their regimes. I consider the leader’s age during the final year of his regime.

Data on the GDP levels and annual GDP growth rates was extracted from the World Bank, as well as the data for population on and capital accumulation. GDP per capita and GDP growth rate are measured with constant price levels, i.e. they are real values. GDP growth, population growth, and capital accumulation are calculated as the average annual growth rate during the leader’s regime, e.g. for leader j who ruled from 1985 until 1993, the considered value for GDP growth is calculated as

𝑔𝑑𝑝𝑔𝑟𝑜𝑤𝑡ℎ𝑗 =

𝑝𝑐𝑎𝑝𝑔𝑑𝑝1993− 𝑝𝑐𝑎𝑝𝑔𝑑𝑝1985

𝑝𝑐𝑎𝑝𝑔𝑑𝑝1985

10 Descriptive statistics of the data can be found in table 1 below.

Mean Standard deviation Minimum Maximum Executive’s age 62.5714 11.2883 35 95 Personalism index 0.7472 0.3115 0.1291 1 GDP per capita 1661.991 2785.381 61.867 23573.63 Average annual GDP growth rate 2.1644 3.7939 -6.686 24.685 Difference between average growth and growth towards the end of the regime

0.1689 3.0820 -10.507 17.182 Average annual capital accumulation 7.8648 9.1124 -21.737 43.068 Average annual population growth 2.3936 2.0616 -14.286 10.99

Table 1: Descriptive statistics of the dataset

11

6. Estimation results

In this section, I will present the results of the regressions performed as described in section 4 on the data presented in section 5. The results are organized by the hypotheses described in section 3.

H.1: A country will have lower growth rates as the executive ages

In accordance with the estimation strategy from section 4, the dependent variable in this estimation is the annual GDP growth rate observed in the year that an executive first enters the office (gdpgrowth), and the main dependent variables are the executive’s age (exage) and the squared age (exage2_{). Other explanatory variables used are the tenure of a leader (extenure), the average annual}

population growth rate (popgrowth), the annual GDP per capita in the final year of the observation (pcapgdp), the average annual rate of capital accumulation (capital) and the personalism index (lindex). According to H.1, the coefficients for exage and exage2 are expected to be negative. The regression output can be found in table 3 below.

Dep var:

gdpgrowth

Coefficient Robust SE t P>|t| 95% confidence interval

exage 0.0824 0.1921 0.43 0.668 -0.2976 0.4625 exage2 -0.0004 0.0015 -0.26 0.792 -0.0033 0.0026 pcapgdp 0.0002 0.0001 2.37 0.019 0.00004 0.0004 popgrowth 0.0286 0.1906 0.15 0.881 -0.3489 0.4061 capital 0.1461 0.0524 2.79 0.006 0.0424 0.2498 lindex -2.6206 1.1584 -2.26 0.025 -4.9121 -0.3292

Table 2: regression output from testing H.1 and H.2.

Contrary to H.1, the coefficient for exage is 0.0824, which would suggest that, e.g., the growth rate realized by a country with an 52-year-old executive will be 0.0824 percentage points higher than that of a country with a 51-year-old executive, that is, if we assumed the marginal effect of age on growth to be linear. However the marginal effect of age on growth is assumed to be exponential, which is why we also computed the coefficient for exage2_{. This coefficient is -0.0004, which is in accordance with}

H.1.

From the P-values that we can observe in table 3, we see that the coefficients for neither exage nor

exage2 _{are statistically significant at any of the common levels of significance, which means that we}

fail to reject the null-hypothesis that there is no effect of exage on gdpgrowth. As can be expected,

gdpgrowth seems closely correlated with the control variables that were used in the traditional growth

12 As derived before, the total marginal effect of exage on gdpgrowth, given by 𝑑𝑔_𝑑𝑞 = 0.0824 − 0.0008 ∗ 𝑒𝑥𝑎𝑔𝑒, implies that, for all leaders in my sample, there is a positive marginal effect of age on GDP growth, although not statistically significant. If the sample included any leaders with an age of more than 103, with a continuing trend, we might see the marginal effect slide into negative values, but the oldest dictator in my sample is 95, therefore that is not the case. The total marginal effect of age on GDP growth, relative to the executive’s age is plotted in figure 1, which also displays the 95% confidence interval.

Figure 1: marginal effect and 95% confidence interval for H.1

13

H.2: A powerful executive will harm the country’s economic growth

From table 1, we can see that the coefficient for lindex is -2.6206. This implies that, within this sample, there is a negative effect of personalism on GDP growth, as predicted in H.2. Specifically, it would imply that the average annual growth rate will be 2.6206 percentage points lower for an autocracy than for a perfect democracy. This negative effect is statistically significant at the 5% level, which implies that I could reject the null hypothesis of no correlation between personalism and GDP growth.

H.3: Aging executives will behave more predatory at the end of their regime

As described in section 4, I also transformed the model to see if there are any differences between the average growth rate during the entire regime and during the final three years of the regime.

The estimation strategy is similar to the one described before, but the response variable and the explanatory variables are transformed to the difference between the average value over the entire regime and the average value over the final three years of the regime.

The regression output for testing H.3 can be found in table 3 below. Similar to before, I will comment on the sign and the magnitude of the coefficients for my explanatory variables, but keep in mind that they are not statistically significant at the 5% level. From table 3, we can see that the sign of the coefficient for exage is different compared to when this method was applied to non-transformed model.

Table 3: regression output from testing H.3.

The marginal effect, given by 𝑑𝑔_𝑑𝑞 = −0.0207 − 0.00004 ∗ 𝑒𝑥𝑎𝑔𝑒, is plotted in figure 2 below, enclosed by its 95% confidence interval.

Dep var:

diffgrowth

Coefficient Robust SE t P>|t| 95% confidence interval

14 From figure 2, we see that although the marginal effect of age on growth is negative for all values within the age range of this sample, and growing further below zero for higher ages as expected, the 95% confidence interval encloses 0 for all ages considered in this sample. We can therefore conclude that exage and exage2 _{are also not jointly significant.}

15

7. Discussion and recommendations

Looking at the estimation results, there are a few things that I have to address:

- Since none of the results on the effect of age are statistically significant, I can discuss them but not draw economic conclusions from them.

- None of the hypotheses regarding the main explanatory variable age could be proven with the data from this sample, although the signs often correspond with the signs predicted by the hypotheses.

In H.1, I found that there is no significant correlation between a leader’s age and his country’s average growth rates during his regime. Growth can be explained by the commonly-used explanatory variables from the Solow-Swan model as described by Mankiw, Romer, and Weil (1992). Especially capital accumulation and the already existing level of development have an important role in explaining growth.

In H.2, I found that, when attempting to explain economic growth in non-democracies, there is also a role for the personalism index, i.e. the degree to which an executive can single-handedly determine the economic policy. In this sample, it seems that a dictator with a high value, i.e. a lot of power, may harm his country’s growth, perhaps by pursuing his own utility and not acting in the country’s best interest.

When comparing the end of leaders’ regimes to the rest of their term I found no significant results, but the signs of the coefficients were negative, similar to what I predicted in H.3. The intuition behind this hypothesis was inspired by a recommendation from Jong-A-Pin and Mierau (2015), who suggested to consider the scenario where an executive has near-certain knowledge about his impending death, such as Hugo Chavez when he was diagnosed with cancer and died almost two years later, even after being reelected while suffering from the disease. While I expected that executives would have more knowledge about this at the end of their regimes, I found no evidence of more predatory behavior in their final year.

16 besides the 51-year time period of the data including multiple global business cycles, there will naturally be larger standard errors than with more converged data. A more converged dataset could be reached by eliminating either the time dimension, e.g. considering only observations from the 1980s, or by eliminating the space dimension, e.g. considering only observations from Latin America. This would account for some of the country- or time-fixed effects, such as the fact that GDP growth in the early 1980s was low everywhere, or that African countries fluctuate a lot between years in terms of GDP growth and capital accumulation.

8. Summary and conclusion

The main research question was whether older political leaders behave more predatory than their younger colleagues because they care less about the future and thus discount future utility at a higher rate, since there is a higher probability that they will not ‘be around’ in the future. According to the theoretical framework, the higher discount rate would cause leaders to extract more funds from the economy to boost present consumption, rather than investing it in the future capital stock to boost economic growth.

In order to investigate this, I constructed a dataset with observations from 140 non-democratic political leaders from 85 countries who ruled for at least 5 years between 1961-2012. I estimated a growth model including the age of the dictator and the political power held by the dictator. I tested whether there is a correlation between the executive’s age and the average annual GDP growth rate and between the executive’s political power and the average annual GDP growth rate. The results indicated that there is no significant support for my main hypothesis that an older leader harms the country’s economic growth. However, I did find evidence that a country’s growth might slow down if a leader is too powerful.

17

9. Sources and references

Acemoglu, D. and Robinson, J.A., (2000). Political Losers as a Barrier to Economic

Development, American Economic Review, 90: 126-130.

Besley, T., Montalvo, J.G. and Reynal-Querol, M. (2011). Do Educated Leaders Matter?, Economic Journal, 121: F205-208.

Dolejs, J. and Maresova, P. (2017). Onset of mortality increase with age and age trajectories of

mortality from all diseases in the four Nordic countries. Clin Interv Aging, 2017; 12: 161-173

Dreher, A., Lamla, M.J., Lein, S.M., and Somogyi, F. (2009). The impact of political leaders'

profession and education on reforms, Journal of Comparative Economics, 37:169-193.

Faust, J. (2007): Democracy's Dividend: Political Order and Economic Productivity. World Political Science Review, 3 (2): 1-26.

Goemans, H., Gleditsch, K., and Chiozza, G. (2009). Introducing Archigos: a dataset of

political leaders. Journal of Peace Research, 46: 269-283.

Hayo, B. and Neumeier, F. (2013). Political leaders' socioeconomic background and public deficits:

Evidence from OECD countries. Joint Discussion Paper Series in Economics, No. 08-2013

Horowitz, M., McDermott, R., and Stam, A., (2005). Leader age, regime type, and violent

international relations. Journal of Conflict Resolution, 49: 661-685.

Jones, B.F., and Olken, B.A., (2005). Do leaders matter? National leadership and growth

since World War II. Quarterly Journal of Economics 120: 835-64.

Jong-A-Pin, R. and Mierau, J.O. (2011). No Country for Old Men: Aging Dictators and Economic

Growth. Working paper for University of Groningen and NETSPAR.

18 Magaloni, Beatriz, Jonathan Chu, and Eric Min (2013). Autocracies of the World, 1950-2012 (Version

1.0). Dataset, Stanford University.

Mankiw, G. N., Romer, D., & Weil, D. N. (1992, May). A Contribution to the Empirics of Economic Growth. The Quarterly Journal of Economics, 107(2), 407-437.

Meade, J.E. (1961) A Neo-Classical Theory of Economic Growth. 1983 reprint of 1962 edition, Westport, Conn: Greenwood.

Moessinger, M. (2014). Do the personal characteristics of finance ministers affect changes in public

debt? Public Choice, Vol. 161, No. 1/2 (October 2014), pp. 183-207

Olson, M., (1993). Dictatorship, democracy, and development. American Political Science Review, 87: 567-576.

Robinson, J.A. (1997). When Is a State Predatory? Unpublished manuscript, University of California-Berkeley.

Robinson, J.A. (1997). A Political Theory of Underdevelopment. Unpublished manuscript, University of California-Berkeley, 1999.

Solow, R. M. (1956). A Contribution to The Theory of Economic Growth. Quartaly Journal of Economics, 70(1), 65-94.

Smirnov, S. (2015). Economic Fluctuations in Russia (from the late 1920s to 2015). Russian Journal of Economics, Volume 1, Issue 2, pp. 130-153

Swan, T.W. (1956). Economic Growth and Capital Accumulation, Economic Record, Vol. 32 (2), p.334-361