The study will focus on characteristics of a revolution and their effect of Economic Growth

(1)

The art of revolution

“Can characteristics of a revolution influence the future economic growth of a country?”

Allard de Ree S1630008

S1630008@student.rug.nl

University of Groningen, Faculty of Economics and Business Supervisor: Dr. R.K.J Maseland

(2)

2 Contents

Abstract 3 1. Introduction 4-5 2. Literature Review 6-11

2.1 Economic indicators of an approaching revolution 6 2.2 Political Instability and Economic Growth 6-7 2.3 Economic growth and left- or right-wing politics 7-8 2.4 Democracy and growth 8-11 3. Theory and Model 12-13

3.1 Theory summary behind the hypotheses 12 3.2 Theoretical Framework 13

4. Methods and Data 13-18 4.1 Explanation of data collection and variables 13-16

4.2 Summary statistics of variables 17 4.3 Statistical analysis 17-18 5. Analysis and results 19-31

5.1 Estimation for the Short –Run 19-22 5.2 Estimation for the Long-Run 22-25 5.3 Estimation for the Short-Run (Limited Dataset) 26-28 5.4 Estimation for the Long-Run (limited Dataset) 28-31 6. Conclusion 31-32 References 33-34 Appendix 35-41

Appendix 1A 35 Appendix 1B 36-37 Appendix 1C 38

Appendix 1D 38-40 Appendix 1E 40-41

(3)

3 Abstract

In the recent period, political revolutions in the Arab World have dominated the news. In this study, I will try to find results that answer the question if people should always follow the mass in search for more welfare, or if they should first look at the possible political outcome of a revolution. The study will focus on characteristics of a revolution and their effect of Economic Growth. The data set will consist of a number of revolutions in the time period from 1955-2002.

Key Words: Characteristics of a revolution, Economic growth, revolution from 1955-2002

(4)

4 1. Introduction

At the beginning of February 2011, the people of Tunisia began a revolution against the regime that had been in office for over twenty years. The hope for more political freedom spread trough the entire Arab World rapidly, and soon people of the Arab World opted for government reform. At the time of writing, revolutions in Tunisia and Egypt were successful in the sense that the old regime was overthrown.

One root of the unhappiness with the ruling regime can be found in the economic situation of the Arab World at the beginning of 2011. In an article taken from The Economist of February 5^thit is stated that, for the case of Egypt, “some 40 percent of the Egyptians still live on $2 a day”. Moreover, “the poor have won little relief from relentlessly rising food prices and sharper competition for secure jobs”. The most important group of protesters were young academics, who fuelled their unhappiness with the fact that, despite earning a proper degree, could not find a job. Last, the many years in office made the regime vulnerable to corruption and ineffective policies aiming at pleasing the people close to the presidents that were in office.

The question arises if an uprising or revolution will help to improve the economic situation.

However, such a study is hard to conduct, since one cannot predict what would have happened to a country if a revolution had not occurred. Despite the fact that one cannot easily make conclusion on whether or not a revolution is good for growth, one can make a study on the characteristics of a revolution and whether some characteristics are more beneficial for economic growth than others. Considering that results from the analyses are significant, the relevance of the study will lie in the fact that it gives a prediction on the magnitude of benefits or losses of a revolution for the society. On the basis of these results, the society can choose to support the revolution and it gives them a clear picture of magnitude of economic growth in the future.

A broad literature has already been written on a link between democracy and growth (Barro, 1996), a transition to democracy and growth (Hausmann et al, 2005), political instability and economic growth (Alesina, 1996). But these articles fail to get a grasp on the effects that characteristics of a political revolution have on the economic growth of a country. Therefore it will be interesting to research a revolution itself. In short, this means that I will investigate different characteristics of a revolution, and how these characteristics will influence the economic growth. Among other things, characteristics of a revolution will include the transitions a new government has been through, such a higher level of democracy or a change

(5)

5

to a more left-wing ideology, and the length of the revolution, which is linked to a higher degree of political instability.

The general research question for my thesis will thus be: Can characteristics of revolutions influence the future Economic growth of a country. One cannot instantly answer this question without breaking it up into specific research questions.

The characteristics that I will be focussing on will be the transition to democracy or autocracy, a transition to left-wing or right-wing ideology and the length of a revolution. In addition, the number of casualties during the revolution will be included, as it indicates the level of chaos in a country which will reflect on the economic performance of the country.

This leads to the following specific research questions.

- Does a transition to democracy/autocracy lead to higher short-term economic growth after a political revolution?

- Does a transition to democracy/autocracy lead to higher long-term economic growth after a political revolution?

- Will a left-wing/right-wing government change be of influence to economic growth?

- Is the length of a revolution important for the economic development of a country after a revolution?

- Is the number of casualties important for the economic development of a country after a revolution?

(6)

6 2. Literature Review

2.1 Economic indicators of an approaching revolution

Literature has been dedicated to the question what variables trigger or increase the likelihood of a political revolution. Most of these papers focus on the economic situation prior to a revolution.

Mark J. Gasiorowski (1995) and Morris Silver (1974) both discuss the economic triggers. The latter uses an economic approach in which the driving factors of the individual to take part in a revolution are listed. In the theoretical paper, Silver argues that unemployment, unemployment among adolescents and unemployment among the people with a high education are important in making a decision to either take part in or support a revolution. A more general approach is made by Gasiorowski, in which a statistical analysis results in the conclusion that both inflation and slow economic growth are positively related to the likelihood of a revolution.

Articles taken from “The Economist” focus more on the political indicators that increase the likelihood of a revolution to commence. Most of these indicators can be found in the so- called Shoe thrower’s Index, which was used to summarize the recent upheaval in the Arab World. The political indicators include press-freedom, corruption and the number of years that the head of state is in office.

From the literature it becomes apparent that the more a country’s situation is deteriorated, the greater the chance of a revolution commencing in a country is. In light of the research question, the literature from this subsection can be used to see what before situation is

“beneficial” for a positive outcome after a revolution. In theory, economic (and political) hardship before a revolution should be positively related to an acceleration in economic growth, because it is more likely that a revolution will improve a situation of a country when it is at the bottom of the economic barrel.

HYPOTHESIS I: Economic hardship is positively related to post-revolution economic growth

2.2 Political Instability and Economic Growth

In the papers by Alberto Alesina, Sule Özler, Nouriel Roubini and Phillip Swagel (1996) and Yiannis Venieris and Dipak Gupta (1986), the authors focus their arguments on the uncertainty that follows from political instability.

Alesina et al. separate political instability in regular political changes, major political changes and coups. They find that regions with the highest probability of irregular government

(7)

7

changes (coups) experience the lowest growth rates. For these irregular government changes, growth rates are reduced by fifty percent during a government change. In order to make sure that these results are robust, the authors try to find the right causality (from political instability to slow growth or vice versa). In doing so, evidence is found that political instability is harmful for growth, but the suggestion that the causality runs in the other direction cannot be rejected.

Venieris and Gupta focus on social-political instability and its effect on the savings rate.

According to the authors, some national savings are identified with investments financed by domestic means, called magnitude recorded savings, and are used to purchase precious metals, foreign currency, hoardings of currencies etc. It is suggested that “the assumed dependence of S on SPI arises from our conjecture that the effect of an increase in SPI is to increase the perceived riskiness and lower the expected values of future income as well as associated return with the risky asset S”, where S are recorded savings. This analysis of savings leads to a regression where the marginal propensity to save is the dependent variable and GDP and SPI are the main independent variables. The main results from the regression with respect to socio-political instability and saving behaviour suggest that there is evidence of a negative link between both variables. In addition, the effects of income distribution are weaker than the fluctuations in savings associated with a fluctuation in political instability.

However, the authors are cautious with these results, as “the equations clearly capture only that portion of saving that is invested (and re-corded) by each group”. Although the results from the equations have to be interpreted with some caution, it gives reasons to believe that political instability has a negative impact on the savings rate, which will diminish investment, leading to a lower economic growth.

The literature that has been discussed does not explicitly mention the length of a revolution or its severity, but I think that the general term “political instability” has the same effect as

‘length/severity of a revolution”, because both add to the uncertainty in a country.

HYPOTHESIS II: The duration and intensity of a revolution is negatively related to growth accelerations.

2.3 Economic growth and left- or right-wing politics

According to Jakob de Haan and Jan-Egbert Sturm (1994), left-wing government experience a higher share of government spending. This government spending will lead to “a negative wealth effect for the representative household” (Ramey and Shapiro, 1998). If one would

(8)

8

agree with this intuition, a transition to a left-wing regime will lead to decrease in economic growth. Moreover, left-wing governments are likely to experience high inflation, since they are more concerned with employment performance and output (Vittorio Grilli and Gian M.

Milesi-Ferreti, 1995). From the research the authors find some evidence that left-wing governments have higher real interest rates. Historically speaking, “high real interest rates stem from slow money growth, falling inflationary expectations and a drop in inflation uncertainty” (Makin, 1983).

There is not a lot of literature on the effects of right-wing politics, but De Haan and Volkerink (2001) find that, in history, right-wing governments have been somewhat more responsible when running the government budget.

From the above literature, it seems that a right-wing ideology is favoured when it comes to economic growth. Therefore, the third hypothesis will focus on the possible positive relationship between a right wing ideology that drives a revolution and economic growth.

HYPOTHESIS III: A right-wing ideology associated with a revolution will lead to higher economic growth relative to the economic growth associated with a left-wing ideology.

2.4 Democracy and growth

In the papers from Robert J. Barro (1996), Roger C. Kormundi & Philip G. Meguire (1985) and Gerald W. Scully (1988), the positive link between democracy and growth is discussed.

Barro finds that the non-linear relationship is both significant and positive, but it has to be said that the results show that democracy is less relevant for economic growth as variables like human capital, male schooling, government consumption and the investment ratio. The papers by Kormundi & Meguire and Scully back up these findings by results that show that country that are in a high civil liberty category experience greater economic growth (between 1-2 percent), have a higher investment ratio and experience a higher level of efficiency when it comes to the allocation of economic resources. Both papers warn of the explanatory power of the articles, as more sophisticated models need to be developed in which more comprehensive measures of investment, government spending, and the degree of openness to international trade are used.

Ricardo Hausmann, Lant Pritchett and Dani Rodrik (2005) also discuss the effects of democracy and economic growth, but the article differs from previous papers because it focuses on the change towards democracy. The authors also discuss external shocks that might influence growth accelerations. Among these external variables the authors include

(9)

9

political changes as a trigger of growth accelerations. The authors measure the political changes by including a variable using regime changes observed by Polity IV. The authors find that nearly half of the growth accelerations are preceded or accompanied by a political regime change. If the variable for regime change is separated into a positive (towards democracy) and a negative (towards autocracy), more striking results become apparent. The impact of a negative regime change is three times larger than the impact of a positive change (10.8 against 2.9 points). This is somewhat puzzling for the authors, as they cannot find a reason why a negative change should have a larger impact on growth accelerations. Therefore, Haufmann et al. distinguish growth episodes into unsustainable and sustainable growth, where sustained episodes have a growth of two percent, seven years after the regime change and growth beneath the two percent level is defined as an unsustainable episode. Results show that a positive regime change has a “statistically and quantitatively significant impact on the likelihood of sustained accelerations”, but not on the unsustainable growth acceleration. As argued in the previous papers, the authors believe that democratic transitions are less important than the standard growth determinants.

Rustow (1969) also finds a theoretical argument that backs up a move towards democracy.

Although the paper is not explicitly tries to find results from democratization, it does hint towards several indicators of economic growth resulting from a transition to democracy. In his conclusion, Rustow notes that mass education and social welfare services are likely to be the result of democratization. Despite the fact that economic growth is not directly mentioned in the article as being a variable which is changed after a transition to democracy, mass education and service welfare services do influence economic growth, increasing the validity of the article being in this literature review

Lorenz Blume, Jens Müller and Stefan Voigt (2009) investigate the effects of direct democracy on the economy. In the paper, direct democracy is measured by de jure and de facto variables, where de jure variables include percentages of countries that have direct democratic institutions and plebiscites (such as mandatory referendums, optional Referendum, initiatives, plebiscites and parliamentary plebiscites) at there disposal and de facto variables include the actual number of mandatory- and optional referendums and initiatives at a given period in time. These variables are used to group the 88 countries in the study into four groups, where zero is the lowest level of direct democracy and three is the highest level of direct democracy. These groups are later used to compare the effects of direct democracy on the economy. Most valuable implication of the regression model is the effects direct democracy has on government effectiveness. The number of mandatory referendums held is

(10)

10

positively correlated with the measurement of government effectiveness taken from the governance indicators of the World Bank. In addition the number of mandatory referendums and initiatives has a constraining effect on corruption. As has already been discussed in earlier parts of the literature review, the authors find that direct democracy has a stronger effect in weak democracies. Blume, Müller and Voigt indeed state: “beyond a certain level of development, the economic effects of direct democracy begin to decrease.

Adam Przeworski and Fernando Limongi (1993) also discuss government efficiency as part of a theoretical framework in which the arguments for democracy and growth are discussed.

They argue that “some productive role of the state is optimal for maximizing efficiency, growth or welfare. The active role for the state is determined by the size of the government.

Under dictatorship, in the form of either an autocracy or a bureaucracy, the state apparatus is indifferent to the size of the government, as opposed to a democracy, where perfectly informed voters will choose the size of the government for which growth is maximized. Of course, the framework is somewhat unstable, since the assumption of perfectly informed voters leaves room for discussion.

On the other hand Przeworski and Limongi also take into question arguments in favour of a positive relationship between democracy and growth that have been claimed by previous authors. First, they reject the invention that democracy protects property rights, which are highly relevant to secure economic growth. The argument behind this reasoning is found by the fact that “distributions of consumption caused by the market and those voted on by citizens must differ since democracy offers those who are poor, oppressed or otherwise miserable as a consequence of the initial distribution of endowments and opportunity to find redress via the state”. However, the authors state that this argument does not favour autocracy; it only rejects the mechanism where democracy protects property rights. Second, as opposed to Kormundi and Meguire, the authors suggest that democracy undermines investment. It is said that democracy will unleash pressure for immediate consumption, which will undermine investment and thus retard growth (Galenson and De Schweinitz, 1959;

Huntington, 1968).

The third theoretical framework that Przeworski and Limongi give is more in favour of dictatorship than it is against democracy. It involves the ability of the state to pursue developmentalist policies without any pressures originating from large firms and unions.

These pressures stem from competition for rents, maximizing their net difference from the gain from the policy and the cost of lobbying. This rent-seeking behaviour will result in an inefficient equilibrium because “lobbying is wasteful and because transfers of income that

(11)

11

result from group pressures cause deadweight losses. Moreover, when the state becomes permeated by private pressures, policies lose internal coherence”. The problem behind this reasoning is the strong assumption that autocracies will behave in the interest of the people whereas a democracies can do no such thing.

Although the above literature suggests that there is some evidence of a positive relationship between autocracy and long-term growth (or a negative relationship between democracy and investment) I have decided to use the hypothesis in which the positive relationship between democracy and long-term economic growth is central. The reason for this choice is that arguments of a negative relationship between democracy and growth are backed up with arguments that are solely based on theory, whereas the relationship between democracy and long-term economic growth is backed up with a number of empirical studies.

Because I think that there is a distinction between the effects of a period of prolonged democracy and a transition to democracy, a second hypothesis is needed for short-run effects.

HYPOTHESIS IVA: A transition towards democracy is positively related to economic growth in the long-run

HYPOTHESIS IVB: A transition towards democracy is positively related to economic growth in the short-run

(12)

12 3. Theory and Model

3.1 Theory summary behind the hypotheses

In order to demonstrate how the hypotheses listed in the previous section were found I will summarize the theories of the literature in a table, in which I list all the papers from the literature review with the views that have been taken from the papers.

Author Subject Stance towards Growth Argument

Gasiorowski (1995) Triggers of a revolution - High Inflation, Slow Economic Growth

Silver (1974) Triggers of a revolution - Unemployment

"The economist" (2011) Triggers of a revolution - Press Freedom, Years in Office, Corruption Alesina et al. (1996) Political Instability Negative relationship Uncertainty --> Slow economic growth Venieris and Gupta (1986) Political Instability Negative relationship Uncertainty--> Low Savings --> Low Investment De Haan and Sturm (1994) Political Ideology Negative for leftists Increased government spending

Grilli and Milesi-Ferreti (1995) Political Ideology Negative for leftists High real interst rates --> Slow Money Growth De Haan and Volkerink Political Ideology Positive for rightists More responsible with government budget Barro (1996) Democracy and growth Positive Relationship Non-linear relationship, little significance Kormundi and Meguire (1985) Democracy and growth Positive Relationship High civil liberty --> positive realtionship

Hausmann et al. (2005) Democracy and growth Positive Relationship Short-term --> growth acceleration, little significance Rustow (1996) Democracy and growth Positive Relationship transtion --> mass education and welfare services Blume et al. (2009) Democracy and growth Positive Relationship Direct Democracy --> Economic Efficiency Przeworksi Limongi (1993) Democracy and growth Positive Relationship Government Efficiency

Przeworksi Limongi (1993) Democracy and growth Negative relationship Arguments in Favour of democracy are false, undermines investment, developmentalist policies in autocracies

The theoretical background in the above table resulted in four hypothesis that were already mentioned once in the literature review. To conclude the summary of the theory, I will once more list the four hypotheses below.

HYPOTHESIS I: Economic hardship is positively related to a future economic growth, compared to an economic upturn.

HYPOTHESIS II: The duration and intensity of a revolution is negatively related to growth accelerations.

HYPOTHESIS III: A right-wing ideology associated with a revolution will lead to higher economic growth relative to the economic growth associated with a left-wing ideology.

HYPOTHESIS IVA: A transition towards democracy is positively related to economic growth in the long-run

HYPOTHESIS IVB: A transition towards democracy is positively related to economic growth in the short-run

(13)

13 3.2 Theoretical Framework

From the hypotheses, one can see that the research is divided into three main streams, namely

“form of government”, “political ideology”, and “length and severity of a revolution”. The literature that has been found explains how one of the previous streams might influence economic growth. This has been summarized in a theoretical framework, which can be found below.

4. Methods and Data

4.1 Explanation of data collection and variables

From the conceptual framework, one can notice that data for variables “Form of Government”, “Political Ideology” and “Length and Severity” are vital for the right-hand side (independent variables), whereas an indicator for economic performance is important for the left-hand side of the regression (dependent variables).

The pool of revolutions that is used to examine the effects of a revolution on economic growth is found in the database created by the Polity IV project. This describes 215 successful coups from 1946-2009.

For the dependent variable, some indicator for economic performance has to be generated.

The most complete measure of economic performance is GDP growth. Since I am interested in the difference in GDP growth before and after a revolution has occurred, a variant of GDP growth is needed. Simply subtracting GDP growth one year before a revolution from GDP growth one year after a revolution has ended will not work, because, as found in the literature,

(14)

14

there is a difference between long-term and short-term effects. Moreover, using GDP growth from one year before a revolution will not give a steady picture of a country’s GDP growth.

For that reason, I will use the average GDP growth of one business cycle (Kitchin Cycle), which comes down to the average growth for the four years prior to the revolution.

As said, there is a difference between short-term and long-term effects, and to capture these differences, two separate regressions need to be made, with two different dependent variables.

I have established that the dependent variables should capture the difference between the GDP growth after a revolution and (average) GDP growth before a revolution. To capture short- term and long-term effects, I will partly follow the example of Hausmann, Pritchett and Rodrik (2005) who make a distinction between unsustainable and sustainable growth. They find sustainable growth with GDP levels from one to two years after a revolution, and sustainable growth with GDP levels from 7 years after a revolution. However, I feel that I cannot simply copy this method if I want justified results, because, in doing so, I will use different variables to calculate average growth before and after a revolution (four-year average, two-year average, and GDP growth from one year). Therefore, the average of four years right after a revolution and the average of four years around the 7^th year will be used to indicate short- and long-run effects.

The latter will be a good indicator for studying the effects of the characteristics of a revolution and economic growth. However, there is one flaw to this method: if the economic growth in two countries has increased with seven percentages, but one country began at a two per cent level and the other began at a ten per cent level, no distinction has been made between the countries where I feel it could be necessary. Therefore, I will also use a dependent variable in which the ratio of GDP growth rate is incorporated. This can be done by dividing the average growth rate after the revolution (short-run and long-run) with the average growth rate before a revolution.

The Data to find GDP growth levels is found in the Penn World Tables. In the Penn World Tables, data is available on yearly GDP levels. To obtain the average GDP growth over four years, one has to calculate the yearly GDP growth level for all the researched years. This will result in economic growth rates for four years, from which one can obtain the average growth rate over four years. Unfortunately, this data goes as far back as 1950, which limits the pool of revolutions a bit more, reducing it to 191 observations. Moreover, for some countries, the data is only available for later periods, which limits the number observations to 139.

Next, the independent variable for the change in government form should be produced. This change in government form should be measured by a move to either democracy or autocracy.

(15)

15

Such a measure can be found in the dataset created by the Polity IV project, in which so called polity scores are assigned to each country for each year, dating back to 1946. These scores range from -10 (fully institutionalized autocracies) to +10 (fully institutionalized democracies). The Polity IV project has also produced graphs to illustrate trends in polity scores from 1946-2009. These trends are very useful, as one can find the changes in polity scores that originate from the coups. These changes can be used in the regression to find effects on short- and long-term GDP growth. In addition, an extension of this variable will be made In this extension, I will not use the numerical change in polity scores, but I will group the variables into three main changes in government form, namely “a move to more autocracy” “a move to more democracy” and “no change in government form” In doing so, other results might be produced, because I will only study the direction of the government form change, and the size of the change will not matter anymore.

The variable to capture any effects from government ideology has to be created by finding out what ideology drives the installed government. Is it a communist or a capitalist, a socialist or a liberal, in short: Is the government left-wing or right-wing? Because the data on what ideology a government drives is not readily available, it has to be created from scratch. The dataset from the Polity IV project is helpful in producing the variable, as it includes the name of the leader behind the revolution/coup. With this name, one can find more information about the person leading the revolution, including its ideology. It is most likely that the ideology from the leader behind a revolution will also be the ideology that drives the newly instated government. From this data, a dummy variable is created where governments with a right-wing ideology are converted to 1 and left-wing governments are converted to 0. Now, the variable is numerical and can be used in the estimation model.

In addition, I will also use an independent variable where the change in government ideology is captured. This variable is found by comparing the regime prior to the revolution with the regime installed after a revolution. The scores for these changes will include -1 (a change to a left-wing ideology), 0 (no change in government ideology) and 1 (a change to a right-wing government ideology).

The last independent variables are used to measure the severity (number of deaths) and length of a revolution. The data on the number of deaths during a revolution has been included in the Policy IV project. For some revolutions, the number of deaths are not specified or even not documented at all and I have decided to give these a value of 0, as Polity IV indicates that these “missing variables” are unlikely to include numbers of great magnitude.

(16)

16

The amount of time that a revolution is ongoing is a bit more difficult to identify, since there is no clear-cut definition to when a revolution starts and ends. In my opinion, a revolution either ends when the new government is installed, when the military rule decides to claim the power themselves, or when a new government is installed after a regent council has been in office for a period of time. This will also make sense when comparing instability with GDP growth levels, because the literature has indicated that uncertainty about the future will trigger negative effects on savings, indirectly affecting GDP growth. If one would use an earlier point to indicate that a revolution has ended, than uncertainty about the future will still be present, and its negative effects on savings will not be fully incorporated in the statistical analysis. The unit of time that will be used is a month, as most of the revolutions do not continue for more than a year.

I have controlled for a number of factors associated with the event of a political revolution. Finding control variables for which data is available for all the observations is hard. However, one can produce one general control variables, in which economic hardship is perfectly included, namely GDP per Capita. This variable is perfect as a control variable, since it is closely related to both economic growth and the probability of a revolution to occur. Not only can one use this variable as a control variable, but it can also be used as a tool to test Hypothesis I.

The variable itself is calculated by taking the average of the four years prior to a revolution.

Below, one can find a brief overview of the variables that I will use for the statistical analysis.

The other control variable that is used is found by including the growth rate of a country before the revolution.

Variable Type of Variable Data Source Explanantion

ChngGDPGrth T (1-4) Dependent Penn World Tables Average growth after -Average growth before (SR) ChngGDPGrth T (6-9) Dependent Penn World Tables Average growth after -Average growth before (LR) PropGDPGrth T(1-4) Dependent Penn World Tables Average growth after/ Average growth before (SR) PropGDPGrth T(6-9) Dependent Penn World Tables Average growth after/ Average growth before (LR) Chng PolityScr. Independent Polity IV Change in Polityscore

Chng PolityScr. Grouped Independent Polity IV Change in Polityscore, Dummy variables

RgtWingIdeology Independent Polity IV, Various Dummy Variable, assign 1 if right-wing government ChngIdeology Independent Polity IV, Various -1 for left-change, 0 for no change, 1 for right-change

Casualties Independent Polity IV Number of casualties

MonthsRevolution Independent polity IV, Various Months a revolution is ongoing

GDPCAPBefore Control/ Independent Penn World Tables Avergae GDP Per Capita before a revolution GDPGrowthBefore Control Penn World Tables Average GDP Growth before a revolution

(17)

17 4.2 Summary statistics of variables.

For the group of dependent variables, summary statistics are given in the table below, in which the number of observations, mean, standard deviation and the minimum and maximum value of the variable are given.

Variable Obs Mean Std. Dev. Min Max

Change in GDP Growth (SR) 139 1.152446 8.843125 -18.49 75.46 Change in GDP Growth (LR) 139 0.832518 8.096 -21.69 35.67 Ratio of GDP Growth (SR) 139 1.898417 7.826066 -10.062 90.179 Ratio of GDP Growth (LR) 139 3.163799 20.70004 -5.901 244

For further descriptive statistics I refer to Appendix 1A, in which the descriptive statistics for each variable is given independently.

The summary statistics for the independent variables are shown in the table below.

Chang in PolityScore 139 -2.1223 6.48796 -18 18

Change in Polity Score, No

Change 139 0.460432 0.500235 0 1

Change in Polity Score, -

Change 139 0.402878 0.49225 0 1

Change in Polity Score, +

Change 139 0.129496 0.336963 0 1

Dummy Variable, 1 if right-

wing 139 0.532374 0.500755 0 1

Change in Governemnt

Ideology 139 0.028777 0.669658 -1 1

Number of Deaths 139 941.9424 9769.858 0 115000

Months of Revolution 139 4.785252 8.37733 0 45

Further descriptive statistics, a correlation table and scatter plots for the interaction between the dependent and independent variables can be found in appendix 1B-1D, where the variables will be dealt with separately. The last table summarizes the statistics for the two control variables which can be found below. Descriptive statistics for these two control variables can be found in Appendix 1E.

GDP Growth Before Revolution 139 7.378921 6.103134 -38.84 20.01 GDP per Capita Before

Revolution 139 940.3672 2050.707 18.45 23167.78

4.3 Statistical analysis

(18)

18

The data that is going to be used will result in two cross-sectional studies, one for the short- run and on for the long-run. As said, these two studies should yield different results, as one tries to capture the effects of a transition after a revolution, and the other tries to capture the long-run performance of the new installed government.

For both studies, the statistical analysis will consist of two regressions, one with the numerical variable for polity scores, and one with the grouped polity scores (autocracy, democracy, no change).

The estimation models for the short run will thus be:

ChngGDPGrth (T(1-4)-T(-4-0)) = β1 + β2ChngPolityScr. + β3RgtWingIdeology + β₄Casualties + β5MonthsRevolution + β6GDPprevious + β7GDPCAP

ChngGDPGrth (T(6-9)-T(-4-0)) = β1 + β2ChngPolityScr.Grouped + β3RgtWingIdeology + β4Casualties + β5MonthsRevolution + β6GDPprevious + β7GDPCA

PropGDPGrth (T(1-4)-T(-4--1)) = β₁ + β₂ChngPolityScr. + β₃RgtWingIdeology + β₄Casualties + β₅MonthsRevolution + β₆GDPprevious + β₇GDPCAP

PropGDPGrth (T(1-4)-T(-4--1)) = β₁ + β₂ChngPolityScr.Grouped + β₃RgtWingIdeology + β4Casualties + β5MonthsRevolution + β6GDPprevious + β7GDPCA

The estimation models for the long-run will be:

ChngGDPGrth (T(8)-T(-4-0)) = β1 + β2ChngPolityScr. + β3RgtWingIdeology + β4Casualties + β5MonthssRevolution + β6GDPprevious + β7GDPCAP

ChngGDPGrth (T(8)-T(-4-0)) = β1 + β2ChngPolityScr.Grouped + β3RgtWingIdeology + β4Casualties + β5MonthsRevolution + β6GDPprevious + β7GDPCAP

ChngGDPGrth (T(8)-T(-4-0)) = β₁ + β₂ChngPolityScr. + β₃RgtWingIdeology + β₄Casualties + β₅MonthssRevolution + β₆GDPprevious + β₇GDPCAP

ChngGDPGrth (T(8)-T(-4-0)) = β₁ + β₂ChngPolityScr.Grouped + β₃RgtWingIdeology + β₄Casualties + β5MonthsRevolution + β6GDPprevious + β7GDPCAP

(19)

19

Change in GDP Growth (SR) 1 2 3 4 5 6 7

change in PolityScore -0.04 -0.07

(-0.35) (-0.98)

Dummy: 1 if Right-Wing -0.292 -1.584

(-0.19) (-1.3)

Change in Ideology 0.866 0.654

(-0.77) (-0.71)

Months of Revolution -0.158 -0.014

(-1.77) (-0.24)

Number of Casualties 0 0

(-1.43) (-1.09)

GDP Per Capita Before 0 0

Revolution (-0.07) (-0.19)

GDP Growth Before -1.147

Revolution (14.85)**

Constant 1.067 1.308 1.128 1.91 1.049 1.177 10.347

(-1.35) (-1.19) (-1.5) (2.23)* (-1.4) (-1.42) (10.30)**

Observations 139 139 139 139 139 139 139

R-squared 0 0 0 0.02 0.01 0 0.65

Absolute value of t statistics in parentheses

* significant at 5%; ** significant at 1%

In both models, the regression below each line will be done in three steps, where each step will be a regression with one of the dummy variables for the grouped Polity Score variable.

This has to be done as one cannot use each dummy variable in the same regression since the variables are related to each other.

5. Analysis and Results

In this section, the empirical analysis will be done on the basis of the theory that has been discussed in previous sections. The estimation results will be split up into two sections, namely a subsection for the short-run and a section for the long run.

5.1 Estimation for the Short -Run

In the first two tables below, one can see the results from the regressions with the numerical variable for the change in polity score. In the first table, the independent variables are tested against the dependent variable for the change in GDP growth rate (growth after – growth before). The second table shows the result when the independent variables are tested against the variable for the ratio of GDP growth (average growth after / average growth before).

(20)

20

Ratio of GDP Growth (SR) 1 2 3 4 5 6 7

change in PolityScore 0.025 0.039

(-0.24) (-0.37)

Dummy: 1 if Right-Wing (-1.12) -2.638

(-0.84) (-1.48)

(-0.41) (-1.12)

(-0.58) (-0.31)

(-0.74) (-0.62)

Revolution (-0.14) (-0.28)

Revolution (-1.17)

Constant 1.951 2.495 1.887 2.119 1.851 1.856 4.31

(2.78)* * (2.57)* (2.83)* * (2.76)** (2.77)* * (2.53)* (2.94)* *

Observations 139 139 139 139 139 139 139

R-squared 0 0.01 0 0 0 0 0.03

For both regressions, the results are striking, since none of the dependent variables show any sign of effecting either the change in growth rate or the ratio of the growth rate. One can say that this is a surprise result, as most of the literature discussed indirectly showed that there should be a connection between the dependent and independent variables. In the conclusion I will further elaborate on possible explanations for these unexpected results.

It is highly unlikely that by grouping the polity scores together, other results will become apparent, however, one cannot rule out any changes before actually doing the calculations, and therefore, the two tables below show the results for the regression with grouped polity scores. Each dummy variable is first tested alone, and later is tested with all other control and independent variables.

(21)

21

Change in GDP Growth (SR) 1 2 3 4 5 6

No Change in Polity Score 1.366 0.512

-0.91 -0.55

- Change in Polity Score -0.629 0.298

-0.41 -0.31

+ Change in Polity Score -1.891 -1.894

-0.85 -1.38

Dummy: 1 if Right-Wing -1.792 -1.676 -1.631

-1.47 -1.36 -1.35

Change in Ideology 0.816 0.741 0.69

-0.89 -0.8 -0.76

Months of Revolution -0.011 -0.013 -0.008

-0.2 -0.22 -0.14

Number of Casualties 0 0 0

-1.09 -1.02 -1.03

GDP Per Capita Before 0 0 0

Revolution -0.21 -0.19 -0.2

GDP Growth Before -1.141 -1.146 -1.146

Revolution (14.71)** (14.74)** (14.91)**

Constant 0.523 1.406 1.397 10.316 10.417 10.741

-0.51 -1.44 -1.74 (9.60)** (9.69)** (10.86)**

Observations 139 139 139 139 139 139

R-squared 0.01 0 0.01 0.64 0.64 0.65

Ratio of GDP Growth (SR) 1 2 3 4 5 6

No Change in Polity Score -0.963 -0.906

(-0.72) (-0.67)

- Change in Polity Score 1.543 1.359

(-1.14) (-0.98)

(-0.59) (-0.46)

(-1.38) (-1.29) (-1.41)

(-1.03) (-0.95) (-1.05)

(-0.35) (-0.3) (-0.29)

(-0.58) (-0.54) (-0.62)

Revolution (-0.3) (-0.3) (-0.28)

Revolution (-1.23) (-1.28) (-1.19)

Constant 2.342 1.277 2.05 4.612 3.589 4.289

(2.59)* (-1.49) (2.87)** (2.95)** (2.30)* (2.96)**

Observations 139 139 139 139 139 139

R-squared 0 0.01 0 0.04 0.04 0.03

(22)

22

To no surprise, the grouped polity score variables show no sign of having any effect on the dependent variables for GDP Growth. Therefore, it is of little use to comment on the signs of these variables, since commenting on insignificant results will not yield any justifiable conclusions. However, from the regression tables one can conclude that the dependent variables have no influence on GDP growth.

5.2 Estimation for the long-run

Now that the short-run has been discussed, the next step is to find possible results for the long-run. It is likely that these results will differ somewhat from the previous subsection, as I have explained that a country’s economy might react differently to a change in government policy then to a prolonged period of exposure to new government policies. The results from the regression are given in the same fashion as in the first group of regressions. First, the two regression tables for numerical polity scores are given below.

Change in GDP Growth (LR) 1 2 3 4 5 6 7

change in PolityScore -0.049 0.051

(-0.46) (-0.18)

Dummy: 1 if Right-Wing 0.514 -6.13

(-0.37) (-1.3)

(-0.06) (-0.77)

(-2.12)* (-0.26)

(2.19)* (-0.25)

Revolution (-0.17) (-0.1)

Revolution (-1.3)

Constant 0.728 0.559 0.831 1.656 0.689 0.778 9.462

(-1) (-0.55) (-1.2) (2.12)* (-1.01) (-1.03) (2.43)*

Observations 139 139 139 139 139 139 139

R-squared 0 0 0 0.03 0.03 0 0.03

(23)

23

Ratio of GDP Growth (LR) 1 2 3 4 5 6 7

change in PolityScore 0.023 0.051

(-0.08) (-0.18)

Dummy: 1 if Right-Wing -3.352 -6.13

(-0.95) (-1.3)

(-0.07) (-0.77)

(-0.54) (-0.26)

(-0.34) (-0.25)

Revolution (-0.01) (-0.1)

Revolution (-1.3)

Constant 3.212 4.948 3.158 3.707 3.105 3.173 9.462

(-1.73) (-1.93) (-1.79) (-1.83) (-1.75) (-1.64) (2.43)*

Observations 139 139 139 139 139 139 139

R-squared 0 0.01 0 0 0 0 0.03

From both results, none of the results show any sign of significance, from which one can once more conclude that the independent variables have no effect on GDP growth, and thus, characteristics of a revolution have no influence on the economic growth of a country after a revolution. Further comments on the explanation of these results will be given in the conclusion.

Last, the regression tables are given for the effects of the grouped polity scores variables on economic growth in the long run. On forehand, one can expect that no major changes will occur after the polity scores are grouped.

(24)

24

Change in GDP Growth (LR) 1 2 3 4 5 6

(-1.31) (3.02)* *

(-1.21) (2.99)* *

(-0.29) (-0.52)

Dummy: 1 if Right-Wing 1.18 1.4 0.905

(-1) (-1.17) (-0.74)

Change in Ideology -1.52 -1.679 -1.358

(-1.71) (-1.88) (-1.48)

(-1.42) (-1.22) (-1.2)

(2.25)* (2.17)* (2.38)*

Revolution (-0.08) (-0.08) (-0.02)

Revolution (12.94)** (12.98)** (12.35)* *

Constant 1.663 0.153 0.91 8.94 6.447 7.758

(-1.78) (-0.17) (-1.23) (8.59)* * (6.19)* * (7.79)* *

Observations 139 139 139 139 139 139

R-squared 0.01 0.01 0 0.6 0.6 0.57

* significant at 5%; * * significant at 1%

Ratio of GDP Growth (LR) 1 2 3 4 5 6

(-0.9) (-0.89)

(-1.3) (-1.22)

(-0.57) (-0.48)

(-1.2) (-1.09) (-1.25)

(-0.69) (-0.59) (-0.72)

(-0.31) (-0.24) (-0.23)

(-0.19) (-0.14) (-0.25)

Revolution (-0.13) (-0.13) (-0.11)

Revolution (-1.37) (-1.43) (-1.32)

Constant 4.621 1.293 3.55 10.787 7.313 9.569

-1.93 -0.57 -1.88 (2.61)* -1.77 (2.49)*

Observations 139 139 139 139 139 139

R-squared 0.01 0.01 0 0.03 0.04 0.03

(25)

25

When looking at the results of both regression tables one can see that, when tested alone, no significant effects on GDP growth become apparent. However, when using all variables in the regression with the grouped polity score, significant results are shown in the first table for both the “no change” variable and the “negative change” variable. One can see that the “no change” variable has a negative effect on the change in GDP growth rate and the negative change in polity score has a positive effect on the change in average GDP growth rate.

Especially the last result is somewhat surprising, because literature shows that a positive polity score, and thus, a prolonged period of democracy should have a positive effect on GDP growth, whereas a negative polity score change should have the opposite result. Although the two results found show significance, the robustness of the results is highly doubtful, since none of the previous regressions show the same results. Thus it is likely that the significance of the results come from the fact that both variables interact with other data in the regression.

More on the intuition behind the results will be presented in the conclusion and therefore, I will not go into it further in this section. On a sign note, one can also see that the number of deaths in the first regression gains some significance. However, the explanatory power of the variable is quite low since the beta is approaching zero.

One explanation for the lack of significance might be that there exists a flaw in the data. Until now, I have chosen to use all 140 coups for which I can find observations on the dependent and independent variables. The problem with this list of coups is that, in some countries, one revolution is followed by another revolution in a short period of time. This could contaminate the dataset somewhat, as the observations on changes in economic growth that originate from a revolution include years in which another revolution has taken place. To overcome this problem, I will delete the revolutions that are followed by another revolution within eight years.

(26)

26 5.3 Estimation Short-Run (Limited Dataset)

Deleting all revolutions that are not followed by a period of “peace” which is long enough will result in a new dataset with only 93 revolutions or coups.

From the tables below, one can see that none of the independent variables have any significant effect on both the change in GDP growth rate and the ratio of GDP growth. From this, one can conclude that the results found in the previous section did not originate from the flaw in data discussed earlier. Indeed, the results give an even stronger incentive to conclude that the characteristics of a revolution are of no influence to the GDP growth after a revolution (for a numerical value of polity score in the short-run).

Change in GDP Growth (SR) 1 2 3 4 5 6 7

change in PolityScore -0.028 -0.028

(-0.18) (-0.34)

Dummy: 1 if Right-Wing 0.51 -0.59

(-0.24) (-0.42)

(-0.78) (-0.91)

Months of Revolution -0.144 0.057

(-1.15) (-0.83)

(-1.2) (-0.58)

Revolution (-0.02) (-0.39)

Revolution (15.62)* *

Constant 1.621 1.453 1.665 2.392 1.543 1.683 11.126

(-1.45) (-1.01) (-1.59) (-1.98) (-1.47) (-1.47) (10.17)* *

Observations 92 92 92 92 92 92 92

R-squared 0 0 0.01 0.01 0.02 0 0.75