What Is the influence of democratic reform on the economic growth in European countries, allowing for leads and lags in the effect of democracy to provide insights on possible causality?

(1)

What Is the Influence of Democratic Reform on the Economic

Growth in European Countries, Allowing for Leads and Lags in the

Effect of Democracy to Provide Insights on Possible Causality?

By Hans Agtereek

Bachelor Thesis Economics

Supervisor: dhr. S. Singh

January 2017

Abstract: This paper examines the effect of democratic reform on the economic prosperity of European

countries. A study is provided on five countries over a 25- or 30-year window, where radical regime changes towards democracy occur somewhere in the middle of that time period. This enables an analysis on the economic development of before, during, and after the implementation of democratic reform. By introducing a ‘lagged’ and a ‘leading’ democracy variable, a distinction can be made between the direct and indirect effects of democracy, and the problem of possible reversed causality can be elaborated on. The findings presented in this paper are mostly in line with economic theory of previous research. The direct effect of the implementation of democratic reform appears to be negative, but economic growth seems to be positively correlated with the democracy level of the recent past. However, the negative direct effect of democracy is found to be more significant than generally believed. This may be explained by the presence of regional specific effects, i.e. the economic prosperity of European countries may be more directly susceptible for sudden regime changes compared to other regions. Due to ambiguous results, no evidence is presented in this paper that points towards the problem of reversed causality.

(2)

2

Statement of originality

This document is written by Hans Agtereek who declares to take full responsibility for the

contents of this document.

I declare that the text and the work presented in this document is original and that no

sources other than those mentioned in the text and its references have been used in

creating it.

The Faculty of Economics and Business is responsible solely for the supervision of

completion of the work, not for the contents.

(3)

3

Table of Content

1. Introduction……….4

2. Literature Review……….5

3. Political Context……….10

4. Methodology………13

5. Results and Analysis………14

5.1 The Direct Effect………..15

5.2 The Indirect Effect………..15

5.3 Reversed Causality……….17

5.4 Results in Political Context….……….19

6. Concluding Remarks and Issues for Further Research………20

7. References……….21

8. Appendices………22

Appendix A: Regressions...………22

Appendix B: Tests for Heteroskedasticity ……….………27

Appendix C: Robust Regressions……….……….………31

(4)

4

1. Introduction

In general, richer countries tend to be democratic (Persson and Tabellini, 2006), and so, often it is automatically assumed that democracy strongly promotes economic prosperity. However, when looking at historical data representing economic development and measures on political freedom, a causal relationship between democracy and economic growth does not seem to be that straight forward. In fact, a lot of research outcomes on this matter turn out to be inconclusive or even state there to be a negative relationship between the implementation of democracy and economic growth (Doucouliagos and Ulubaşoğlu, 2008). However, with the aim to provide a more comprehensive understanding of the influence of democracy, a distinction can be made between direct and indirect effects. Allowing for democracy to bear its fruits over time changes the conception on how it is that this political structure influences the economic development of a country, since the indirect effect is mostly proved to be positive. It is even argued by Gerring et al. (2005) that, instead of solely using information on the current political status, the accumulative years a country has experienced a democratic regime over the last century provide the most effective measure for democracy in this case. Also, reversed causality is often mentioned as a problem in research on this subject, since it might as well be economic prosperity that stimulates democratic reform. Therefore, in this paper, there will be some experimentation with the timing of the data sets, with the aim to provide some insights on these problems.

Sometimes radical events can cause a country to undergo a sudden regime change. And although a correlation between economic prosperity and democracy seems to be obvious due to the earlier stated reason, it is events like these that allow us to analyze growth figures prior, during, and after the implementation of democratic reform. Therefore, data on countries undergoing suchlike regime shocks will be used in this paper in order to further examine the causality of democracy on economic growth.

With the aim to elaborate on the distinction between the direct and indirect effects

produced by democratic reforms and the possible problem of reversed causality, three explanatory variables will be introduced in the model that is applied for this analysis. The direct effect will be considered by analyzing the correlation between the democracy level and the GDP growth of matching years. The indirect effect will be addressed on by linking the growth data to a democracy variable that experiences a 10-year lag, i.e. the correlation between the current economic

development and the political status of ten years ago. In order to indicate whether reversed causality might impose a problem, the same time difference in the opposite direction, i.e. a 10-year lead, will be operated for the last democracy variable.

(5)

5 democracy and growth will be discussed in detail. Also, the political context of the countries that are considered will be discussed in more detail in order to possibly enable a better interpretation of the results. Following right after is the chapter on methodology, where a comprehensive explanation is provided on the data that is collected and the model that is applied. Then, the empirical results obtained from several regressions will be analyzed and interpreted with the use of economic theory that is presented in previous studies. And the final section provides some concluding remarks and issues for further research.

2. Literature Review

For the last 60 years or so, there has been an ongoing discussion onto whether the implementation of democracy has an impact on economic prosperity, and whether this would be positive or negative (Doucouliagos and Ulubaşoğlu, 2008). Ambiguous empirical results only nurture this controversy rather than facilitate consensus, and so, both the parties in favor and against the positive causal effect of democracy on economic growth keep on pleading their case.

Introducing the literature on the subject, a good start can be made with Seymour Lipset’s (1959) paper, in which cross-sectional correlations between some measures of economic

development and the type of regime are examined. In order to control for possible complications with cultural differences in the political landscapes, the sample was divided into Latin-American countries on the one hand and European and English speaking countries on the other. For the latter group he found the average income per capita in the stable democracies to be more than twice as high as that in the dictatorships and unstable democracies. In the Latin American countries he found a less extreme yet substantial difference of about 40 per cent in average income per capita in favor of the democracies and unstable dictatorships as compared to the stable dictatorships. The research showed similar outcomes for measures of education, urbanization and industrialization with respect to economic growth.

However, as mentioned by Helliwell (1994), there are two major shortcomings in the research done by Lipset. First of all, there is a problem of causality, since the measures on economic development are only represented by data derived from a time period later than the data classifying the regime. Therefore, these correlations do not provide insightful information onto whether it is democracy that causes economic prosperity or the other way around. And secondly, the Second World War has had in most cases a substantial impact on both the degree to which the European countries were democratic and their economic development. This causes drawn conclusions based on the correlation between prewar levels of democracy and postwar levels of income to be dubious.

(6)

6 With another 30 years of data, more sophisticated measures of democracy and comparative real income, and the availability of a significantly larger sample of countries, Helliwell attempts to provide more useful insights on the subject with his paper on the empirical linkages between democracy and economic growth. In his analysis, he uses data on 125 countries, which are

considered over the time period of 1976-1985, where every year is covered separately, and so, 1250 observations were to be made. The average real income per capita is used as a measure for

economic development, where the Gastil index is used to indicate the degree to which the countries are democratic. In the latter, countries with no civil rights and no political freedom will be denoted with a zero, where a value of 1.0 will represent a country with maximum political freedom and civil rights.

Helliwell did not only run regressions where he simply tested the effect of economic prosperity on the level of democracy, but also, by introducing some regional dummy variables, attempted to control for the notion that possible cultural differences contribute largely to the explanation onto whether a country is democratic. This did diminish the correlation figure on income and democracy considerably, however, it remained to be largely significant. Also, the slope

coefficients were considered in order to test alternative quadratic and cubic income functions against the initially used log-linear model. These tests did not reveal any significant outcomes, and so, did not provide any reason to imply that there is some sort of threshold level when it comes to income above which the probability of democracy in a country sharply increases. Then, another variable that is presented by Helliwell is secondary schooling, since Dahl (1973), among others, emphasized on the possibility that factors such as education and literacy might lead to a higher demand for democracy and therefore may in part explain the apparent link between democracy and economic development. And indeed, even after controlling for regional differences, the correlation between schooling and the democracy level appears to be significant. The effect of GDP per capita on the level of

democracy, however, stays highly significant as well. Then, an alternative measure for the degree to which a country is democratic is introduced, namely the Bollen index. This index, as compared to the Gastil index, uses slightly different component measures, has a somewhat different scale, and provides data on a smaller sample of countries, all in order to quantify fairly similar concepts dealing with democracy. Not surprisingly, however supportively, using the Bollen index as the measure for democracy, the results turn out to be very much in line with that of the Gastil index.

Now, taking into consideration these outcomes, it seems apparent that a high level of economic prosperity is followed by a high degree of democracy. However, the possibility of reversed causality should be taken into account. Therefore, Helliwell ran regressions attempting to provide insights on the effect of democracy on economic development instead of the other way around. In this model, where the dependent variable represents the growth in real GDP per adult over the time

(7)

7 period of 1960 to 1985, basic explanatory measures of economic growth such as economies of scale, schooling, and investment are included. Later, representing the measures of democracy, both the Bollen index and the Gastil index are introduced as independent variables. For both indexes the correlations turn out to be statistically insignificant, where the Bollen index shows a weak negative correlation and the Gastil index a weak positive one. The Gastil correlation, however, is likely a result of reversed causality since no data previous to 1976, which is roughly halfway through the growth period, is available for this index. Consequently, whenever the incomplete Gastil data-set is complemented with data of the Bollen index representing democracy measures previous to 1976, there seems to be a fairly large, however still insignificant, negative correlation. Hence, democracy seems to have an insignificant negative direct effect on economic growth. However, the democracy level of a country does appears to have a positive indirect effect through schooling and investment, which in their turn is believed to promote economic prosperity. Adding up the direct and indirect effect of democracy on economic growth leaves a slightly positive number, but this is considered to be highly imprecise and insignificant.

In conclusion, Helliwell’s research shows that countries that are economically more prosperous tent to adopt a more democratic political system. This conclusion is not subject to the problem of reversed causality, since the apparent direct effect of democracy on economic growth is more often found to be negative than positive, and moreover, turns out to be statistically

insignificant. The indirect effect of democracy, however, through investment and education, seems to be positive, and so, the implementation of democracy appears to come at a very low economic cost, or no cost at all.

Although most research on the subject is largely in line with Helliwell’s conclusion, which finds the effect of democracy on economic growth to be either slightly negative or insignificant, from an historical perspective, this is argued to be rather shortsighted (Gerring, Bond, Barndt, & Moreno, 2005). According to this view, it is no more than logical that no significant correlation is found, let alone a positive one. It would be unrealistic to expect the present degree of democracy in a country to have a significant effect on the economic growth of the time period (10 to 20 years or so)

following right after. It is only plausible to assume that the long-term regime history of a country has a substantial influence on the economic prosperity, rather than solely considering the current or recent political status of that country. In other words, instead of treating democracy as a level variable, it should be expressed as a stock variable, which in fact accumulates the years of a country being democratic.

In their theory, Gerring et al. introduce a multiple causal pathway in order to elaborate on the democracy-growth relationship. Herein it is presumed that the more years a country has had a democratic regime, the more political, social, physical, and human capital would prosper. These

(8)

8 forms of capital in their turn would stimulate economic growth.

Several regressions are ran in their research, where the annual growth rate of GDP per capita again functions as the measure of economic growth and therefore as the dependent variable. The democracy measures are this time based on a Polity IV data set, which provides information on political regime characteristics of a very large sample of countries. They present a 21-point scale that ranks a country on the degree to which it is democratic, based on various measures of democracy. The scale ranges from -10 (hereditary monarchy) to 10 (consolidated democracy) and can be categorized into “autocracies” (from -10 to -6), “anocracies” (from -5 to 5) and “democracies” (from 6 to 10). The measures that are used contain information on the competitiveness of political

participation, constraints on competitive authority, and the competitiveness and openness of

executive recruitment. Gerring et al, however, do not solely use the Polity IV data to create a variable that represents the annual democracy level. They also introduce an additional variable that includes the accumulative stock of democracy in order to investigate the effect of long-term regime history. This stock variable is constructed by adding up all the yearly Polity scores of a country going back as far as the year 1900, including an annual depreciation rate of 1 percent. Furthermore, some

specification tests are included in order to assess the relevance and significance of the democracy stock variable by the usage of a substantial amount of control variables such as investment, instability, and life expectancy, but also the trade-weighted growth in GDP per capita, government consumption, and literacy. Lastly, there is a split-sample test provided in order to control for the possibility that particular regions, on a continental scale, might have a significant influence on the outcomes.

Their initial findings, using solely the annual representation of the Polity data on the level of democracy, are very much in line with other research, namely that its effect on economic growth does not appear to be statistically significant. The stock variable, however, does turn out to have a highly significant positive effect on economic growth, even after the implementation of the large set of control variables. Also, the statistical significance of the stock variable does not seem to be threatened by controlling for regional differences.

Gerring et al. hereby proclaim, in contrast to previous research, that there in fact is a positive significant effect of democracy on economic prosperity. However, in order to generate these

findings, the democracy measure has to contain information on the long-term regime history, rather than merely consider the current political status or recent history. In other words, the stock variable of democracy explains a positive relationship with economic growth, where the level variant does not.

In another attempt to provide useful insights on the issue of democracy and economic growth, Krieckhaus (2006) tries to make his case by refocusing the emphasis on the influence

(9)

9 different political contexts might have on this relationship by considering specific regions separately. In his theory, which is based on the expectation that in different contexts democracy might have different effects, he introduces the notion that the implementation of democratic regimes is really only desirable in regions where the political landscape is shaped a certain way. In large parts of Asia for that matter, it is believed that democratic pressures may only inhibit economic growth, since this would restrain state elites from implementing effective economic policy. Considering the African political context, on the other hand, democratic reforms are presumed to positively stimulate economic growth due to the idea that this may present the opportunity for the public to evict corrupt politicians that do not necessarily strive for economic prosperity at the national level.

Consequently, in Krieckhaus’ analysis he constructs two sub-samples, differentiating between regions that are expected to economically prosper under democratic regimes and regions that are presumed to show the opposite response. Respectively, Sub-Saharan African countries on the one hand and Asian and Latin-American countries on the other. Again the growth in GDP per capita is used as the measure for economic growth and Polity IV data represents the measure for democracy. Also, similar control variables as in the research done by Gerring et al. are introduced (i.e. measures of schooling, life expectancy, government spending, etc.). Furthermore, both simple cross-sectional and time-series cross-sectional models are implemented in this analysis. This all considers data originating from the time period of 1960-2000.

Krieckhaus starts with a regression where he doesn’t make any distinctions between regions. Similar to other research, he doesn’t find any evidence that indicates democracy to have a significant effect on economic development. More interesting results are presented after the sample is split into the two previously mentioned groups. In line with his hypotheses, he finds there to be a significant positive effect of democracy on economic growth in the Sub-Saharan Africa region, and a significant negative influence in the Latin-American and Asian regions. Even after doing a substantial amount of sensitivity analyses such as adding several control variables, dropping outliers, and implementing alternative measures, these democracy effects remain significant.

With these findings, Krieckhaus provides yet another insightful explanation onto how to approach the relationship between democracy and economic prosperity, and to a certain extent, why previous research generally failed to find a significant correlation. In a way, his research confirms the legitimacy of both sides of the argument. Democracy can be validly justified as a tool that provides economic growth, just as much as it can be argued to seriously inhibit this same development, but this is in fact strongly dependent on the region that is assessed.

The fact that plenty of research done on the subject, overall providing very ambiguous results that seem to bring about, in their words, ‘a consensus of an inconclusive relationship’, causes

(10)

10 Rather than providing yet another empirical analysis directly using data on democracy and economic growth measures, they strive to present more insightful conclusions by the means of a meta-analysis. Herein 84 published studies are presented in order to provide a largely comprehensive assessment of the relationship between democracy and economic prosperity.

In order to conclude on the magnitude of this relationship, the mean values of the democracy-growth effect, as well as their 95% confidence and credibility intervals, have been

calculated in their meta-regression analysis. Hereby, the main focus lays on the partial correlations of democracy on economic development. Furthermore, these partial correlations are weighted based on both sample size and quality, where larger sample sizes naturally induce information to be weighted more heavily. The quality is in this case measured by the number of citations obtained by a study and the importance of the journal it is published in. In their analysis, two data-sets are derived from the studies. One that represents the estimates of all the regressions, and one that picks out the best estimate of every study.

With their research, where they strived to present accumulative evidence on the democracy-growth issue, Doucouliagos and Ulubaşoğlu come to several firm conclusions. Firstly, no conclusive evidence is found on the direct effect of democracy on economic development, i.e., the relationship turns out to be statistically insignificant. However, through several channels, the indirect effect of democracy turns out to be significant. Additionally, accumulative evidence does point towards a significant region-specific effect in some cases. Hereby evidence suggests that in Latin-American countries the influence of democracy is larger than average, whereas in Asia it is the other way around. Furthermore, it is concluded that no less than a third of the differences in the reported findings is caused by the variation in the design of the researches and the differences in econometric specifications.

3. Political Context

Five European countries that have experienced rather radical regime changes towards democracy are considered for the analysis provided in this paper. Such drastic shifts in the political structure of a country can be caused by several different factors, which in their turn can result in varying outcomes when it comes to the behavior of the economic growth. Therefore, taking into account the political context of the countries that are examined is imperative to properly interpret the regression outputs. Consequently, the political landscapes during the period of change will be elaborated on further in this section.

For Albania, the fall of communism started in 1990 after several student demonstrations were held at the capital that forced the government to no longer inhibit the possibility for other

(11)

11 political parties to be formed, which gave rise to the formation of the Democratic Party. By then they had experienced a totalitarian regime, imposed by the Communistic party, for over four decades. As stated by Kajsiu et al. (2003), when considering the shift in Albania’s political landscape, several transitions should be taken into account. Firstly, after being considered Europe’s most isolated society, Albania, in a way, opened up to the world, by providing the freedom to travel abroad and abandoning its monopoly on foreign commerce. Also, economically a shift presented itself, where the centrally planned market had to make place for a free market economy. This, however, turns out to be a rather slow and difficult process. And Lastly, a dramatic shift from a totalitarian and

communistic regime towards a democratic system had to occur. Not surprisingly, these intense and complicated processes were accompanied by some challenging times for the political stability of the country. After the fall of the communistic regime, Albania experienced a legitimacy crisis, which caused the democratic system to be seriously flawed in the sense that corruption and election fraud appear to be the norm rather than the exception. The new political system did not seem to provide real progress in fixing the dysfunctional economy that was left behind by the communistic regime. And, besides the institutional weakness in the country, the level of unemployment only increased.

Similar to Albania, a communist regime completely dominated the Bulgarian political landscape for over four decades before being overthrown in 1989. The highly complicated and problematic process of democratization, however, is believed to have started way before this time (Stoper & Ianeva , 1996). This process was accompanied by the marketization of the Bulgarian economy, i.e. the shift towards a free market system that is opening up to the world, coming from a centrally planned economy. According to Stoper and Ianeva, once the decision was made to abandon the centrally planned economy and to allow for marketization, Bulgaria was left no real choice other than to start the process of democratization simultaneously, since some institutional changes that require a certain level of democracy had to be made in order for a market economy to succeed. Also, the totalitarian regime Bulgaria experiences for some decades ironically may have laid the

foundation for a democratic system by providing education for the entire population and hereby inducing a level of political awareness that was considerably high. After the communist regime was overthrown, Bulgaria seemed to have no choice other than continuing the course towards

marketization and with it democratization. Not surprisingly, also for Bulgaria these complex processes work very gradually and it takes time for the institutions to develop properly.

In Greece on the other hand, a different set of events led to the democratic system that is in place now. An autonomous military coup in 1967 led to a dictatorship that lasted for seven years. This dictatorship, however, was led by right-wing colonels instead of communists. Initially there seemed to be a positive trend in the economic prosperity under the military regime, but after three years the economy slowed down, and therefore discontent started to severely increase under the

(12)

12 Greek population, a sentiment that was widely matched with that of countries outside of Greece ever since the coup took place. Also, the military regime did not receive any political support from either the left-wing or the right-wing. With this lack of political support and with resistance coming from the popular masses, it seemed impossible to make this dictatorship last (Poulantzas & Nicos, 1976). Therefore, in 1973, they declares a so-called “presidential parliamentary republic” in order to still maintain their position of power somehow. This development caused several civil and political liberties to be reenacted. Also, a date was set for new elections, which presented the opportunity for revolt. This, alongside with the inability of the military that were in power of the dictatorship to act properly in the Cyprus crisis of 1974 in their confrontation with the Turkish, led to the transition towards democracy. Since the transition no real thread to the Greek democracy has been present, the economy has grown and structural changes have been made. These developments can also partly be attributed to their accession to the European Union.

In Spain, the transition towards democracy is believed to start after the death of Francisco Franco in 1975, who by then ruled the dictatorship for almost four decades. After 1939, when the Spanish Civil War ended and he overthrew the Second Republic that was democratically elected with his Nationalist forces, he rose to power. Franco dictated the culture, persecuted his political

opponents and even banned the Catalan and Basque languages. Six years before his death he appointed his official successor, namely Prince Juan Carlos, of which the grandfather had been a former king of Spain. It was Juan Carlos who, after the death of Franco, facilitated the pathway towards a democratic constitutional monarchy. He, alongside with his newly appointed Prime Minister Adolfo Suarez, initiated a slow process of political reforms. Democratic elections were held to form a government and the new Constitution was passed in 1978. The transition, not surprisingly, did provoke political and social tension, which, for example, motivated a military coup that held the parliament hostage. However, the radical Spanish transition towards a democracy has been believed by scholars to have been remarkably gentle (Edles, 1995).

The Portuguese transition towards democracy turned out to be far less smooth and peaceful. The country had experienced a military dictatorship that was headed by António Salazar for more than three decades before he was incapacitated in 1968. A repressive regime was imposed where political opposition was not tolerated and strict censorship was implemented. The successor of Salazar was Marcello Ceatano, who tried to improve the old Salazar system by making it more modern and liberal. The fact that he did not succeed due to strong opposition made the sentiment grow under several major groups that revolt was imperative to provide the necessary changes towards democracy. The real beginning of the transition is believed to be initiated by the military coup of 1974 (Costa Pinto, 2006). Some turbulent years followed, where several coups, revolutions and interest group conflicts occurred. In contrast to Spain, the government lacked the continuity in

(13)

13 order to provide for a stable transition of power. Tremendous economic and social problems

followed.

4. Methodology

In this paper, some empirical findings are presented with the aim to provide further insights on the relationship between the degree to which a country is democratic and the economic growth of that country. In order to provide these findings, data on five countries is collected and several regressions are ran.

The countries on which the data is collected are selected in the following way. Firstly, in order to prevent possible regional differences in the effect of democracy on economic growth to influence the research outcomes, all the sample countries had to be located in the same region, and therefore only European countries are considered. Secondly, only countries that have been through a radical regime change towards democracy are interesting for this study. Therefore, countries are picked of which data on before, during, and after the regime change can be analyzed. And lastly, there has to be enough democracy and growth data available on the country in order to at least enable a 25-year analysis, where the regime change occurs somewhere in the middle of that time period. Taking these criteria into consideration, five European countries appear to be suitable for testing, namely Albania, Bulgaria, Greece, Portugal, and Spain. For the two first mentioned countries the available data only allowed for a sample of 25 consecutive years, whereas for the other countries a 30-year analysis will be provided.

Very much in line with other research, the annual percentage growth of GDP per capita will serve as the measure for economic growth, and therefore as the dependent variable. These annual growth rates are obtained from the http://data.worldbank.org/ website. In the regressions, the variable that contains this data of a certain country will be labeled as ‘GDP’ followed by the first three lowercase letters of that same country.

The democracy measure is represented by Polity IV data, which provides, as mentioned in the literature section, a 21-point scale that ranks the democracy level of a country. Hereby, a value of -10 represents a hereditary monarchy, whereas a consolidated democracy is indicated with 10 points. The most sophisticated variable provided by Polity IV, namely the polity2 variable, will be directly used in the regressions presented in this paper.

Besides simply regressing GDP data with democracy values of the corresponding years, there will be some experimenting with the timing of the democracy variable. The first independent variable, namely dem0, will represent data that matches the time period of the GDP variable. A significant correlation would hereby imply a direct effect of democracy on economic prosperity.

(14)

14 According to most literature, this correlation is expected to be either negative or insignificant. The second independent variable, however, will experience a ten year lag as compared to the data on economic growth. This lagged variable, which is labeled as ‘demlag’, will provide information on the extent to which the relationship between democracy and growth can be explained by the regime status of the recent past. The motivation behind the implementation of this variable originates from the notion mentioned in most, if not all, scientific research done on the subject, that democracy appears to have a positive indirect effect, which will only be detectable some time after the regime change. Herein, significant results would point towards the idea that democracy is also capable of promoting economic prosperity, but only after it is given some time to bear its fruits. Lastly, a leading variable is introduced, labeled as ‘demlead’ in the regression outcomes. This measure contains data on the democracy status of a country with a 10-year lead on the growth variable. In this case, significant outcomes would raise the issue of reversed causality, since this would imply, given a positive correlation, that democracy follows from economic prosperity instead of the other way around. A negative correlation would be highly improbable and hard to interpret, in the sense that democracy is generally not believed to follow from poor economic performance.

The model used for the analysis will be as follows: A: 𝑌𝑡,𝑖 = 𝛼 + 𝛽1𝐷𝑡,𝑖+ 𝜀𝑡,𝑖 B: 𝑌𝑡,𝑖 = 𝛼 + 𝛽1𝐷𝑡,𝑖+ 𝛽2𝐷𝑡−10,𝑖+ 𝜀𝑡,𝑖 C: 𝑌𝑡,𝑖 = 𝛼 + 𝛽1𝐷𝑡,𝑖+ 𝛽2𝐷𝑡−10,𝑖 + 𝛽3𝐷𝑡+10,𝑖+ 𝜀𝑡,𝑖 Where 𝑌𝑡,𝑖: %𝐺𝐷𝑃 𝑔𝑟𝑜𝑤𝑡ℎ 𝑝𝑒𝑟 𝑐𝑎𝑝𝑖𝑡𝑎 𝑖𝑛 𝑦𝑒𝑎𝑟 𝑡 𝑓𝑜𝑟 𝑐𝑜𝑢𝑛𝑡𝑟𝑦 𝑖 𝐷𝑡,𝑖: 𝑃𝑜𝑙𝑖𝑡𝑦 𝐼𝑉 𝑟𝑎𝑛𝑘 𝑖𝑛 𝑦𝑒𝑎𝑟 𝑡 𝑓𝑜𝑟 𝑐𝑜𝑢𝑛𝑡𝑟𝑦 𝑖

5. Results and Analysis

The outcomes presented by the regressions will be analyzed and, where possible,

interpreted with the use of economic theory of previous research, provided in the literature section. In order to do this structurally, the three variables will be considered separately, one by one.

(15)

15 5.1 The Direct Effect

The first regression provides information solely on the possible direct effect of democracy on economic growth, i.e. the correlation between the democracy level of a given year and the growth in GDP per capita of that same year. For Greece, Portugal, and Spain the outcomes show there to be a negative direct effect of democracy that is highly significant (see Appendix A). Initially, the direct effect in Bulgaria appears to be far from significant. However, once the lagged variable is included in the model, a negative and strongly significant effect does show. The regressions on Albania did not provide any significant results whatsoever.

Where Helliwell, very much in line with other research, already predicted a negative,

however insignificant, direct effect of democracy on growth and Doucouliagos and Ulubaşoğlu used a meta-analysis to argue that this correlation is indeed inconclusive, the findings in this paper imply differently. The results of four out of the five countries indicate there to be a highly significant (negative) direct effect of democracy on economic prosperity. A possible explanation for these deviant findings is the presence of regional specific effects. Krieckhaus stated that democratic reforms result in higher economic prosperity in Sub-Saharan African countries, and lower growth in Asian and Latin-American regions, but he did not mention European countries, in contrary to the research presented here. Lipset did mention European and English speaking countries. However, he solely tested the correlation between average income per capita and the regime type, which resulted in a large difference between democracies and dictatorships in favor of the democracies. This is very much subject to the problem that most economically prosperous countries happen to be democratic, and so, this does not necessarily provide any insights on the effect of the implementation of

democratic reforms. Therefore, the findings presented in this paper may very well imply there to be a negative direct effect of democratic reform on economic development, considering European countries.

5.2 The Indirect Effect

Secondly, the lagged variable, and hereby the indirect effect of democracy is considered. The lagged variable showed positive significant results for Bulgaria, Portugal, and Spain. The indirect effect in Albania and Greece on the other hand turned out to be, although positive, highly insignificant. Furthermore, the implementation of the lagged variable caused the negative coefficients for the direct effect of democracy to increase by roughly half its initial value for both Portugal and Spain. For Bulgaria, as mentioned previously, the additional variable induced an even more extreme change in the coefficient for the direct effect. Moreover, where this direct effect initially turned out to be strongly insignificant, it now shows a high significance level.

(16)

16 Variable (1) (2) (3) Dem0 0.3264 0.1944 -0.2914 (0.2734) (0.3347) (0.4931) Demlag 0.2933 0.2319 (0.4196) (0.4152) Demlead 1.699 (1.284) Constant 2.191 3.918 -7.810 (1.881) (3.119) (9.377)

Regression type OLS OLS OLS

Observations 25 25 25 R-squared 0.0583 0.0788 0.1497 Variable (1) (2) (3) Dem0b -0.1284 -0.3583*** -0.7433*** (0.1340) (0.1118) (0.2449) Demlagb 0.5732*** 0.5516*** (0.1284) (0.1234) Demleadb 6.423* (3.675) Constant 2.723** 5.315*** -48.92 (1.055) (0.9736) (31.04)

Observations 25 25 25 R-squared 0.0384 0.4953 0.5594 R-squared 0.0384 Variable (1) (2) (3) Dem0g -0.3632*** -0.3403*** -0.3181*** (0.1103) (0.1122) (0.1077) Demlagg 0.1420 0.1005 (0.1341) (0.1298) Demleadg -0.2582* (0.1351) Constant 5.037*** 4.647*** (0.8166) (0.8941)

Observations 30 30 30 R-squared 0.2791 0.3078 0.3931 Variable (1) (2) (3) Dem0p -0.2209*** -0.3254*** -0.3609*** (0.07556) (0.0825) (0.0935) Demlagp 0.2586** 0.2619** (0.1083) (0.1090) Demleadp 0.101 (0.1228) Constant 4.742*** 6.232*** 5.551*** (0.6838) (0.8888) (1.218)

Observations 30 30 30

R-squared 0.2339 0.3675 0.3835

Regression analysis Portugal

% GDP Growth per capita

Regression analysis Greece

Regression analysis Bulgaria

Table 1

Regression analysis Albania

(17)

17 The fact that for Bulgaria, Portugal, and Spain the direct effect of democracy becomes increasingly positive and more significant after the lagged variable is introduced points towards the presence of omitted variable bias. For these countries, a significantly larger part of the behavior of the economic growth variable can be explained by democracy after the indirect effect is included in the model. This can also be indicated by the strong increase in the R-squared values after the

implementation of this new variable. Due to the fact that the omitted variable, i.e. the indirect effect, is shown to be positively correlated with economic growth, there appears to be an upwards bias in the first regressions, and therefore the negative direct effect of democracy was initially

underestimated.

These findings suggest that it is likely for democracy to positively influence economic development as long as it is provided some time to settle in. Where democracy does not seem to stimulate economic growth at the moment of implementation, it does appear to provide a foundation for economic prosperity in the near future (the next decade). This is very much in line with what Helliwell as well as Doucouliagos and Ulubaşoğlu concluded in their research; through various channels, such as education, investment and political stability, there appears to be a positive indirect effect of democracy on economic growth. The findings presented here also point towards the idea behind the theory provided by Gerring et al., where they argue that the democracy measure can more effectively be represented by a stock variable rather than a level variable. This stock

variable can also be interpreted as the indirect effect of democracy, only the emphasis here lies more on the long-term effects, rather than a ten-year lag. Both approaches detect a positive indirect effect of democracy.

5.3 Reversed Causality

In the last regressions, the leading variable was added in order to address on the possibility of reversed causality. The implementation of this variable produced very ambiguous results. Albania

Variable (1) (2) (3) Dem0s -0.2292*** -0.3283*** -0.2496*** (0.05556) (0.05473) (0.05679) Demlags 0.2935*** 0.2842*** (0.8363) (0.07507) Demleads -0.1817** (0.06601) Constant 4.071*** 5.523*** 6.617*** (0.4474) (0.5602) (0.6404)

Observations 30 30 30

R-squared 0.3781 0.5729 0.6693

*p < .10; ** p < .05; *** p < .01

Regression analysis Spain

(18)

18 and Portugal showed a positive leading effect that turned out to be highly insignificant, where that of Albania appears to be relatively large and that of Portugal fairly small. Bulgaria and Greece produced leading coefficients that showed a trend towards significance (significant at 0.10 level but not at 0.05 level), where that of Bulgaria turns out to be strongly positive and that of Greece slightly negative. Only on Spain the regression presented a highly significant value, which indicated a negative leading effect.

These ambiguous, and mostly insignificant, findings do not support the notion that reversed causality might be a problem here. Spain is the only country for which significant results on this variable presented itself. However, this coefficient is hard to interpret with economic theory, since it shows a negative correlation. In an attempt to interpret the negative value of the coefficient, this would imply there to be a negative relationship between the current economic prosperity and the democracy level of ten years later, i.e. poor economic performance would stimulate democracy in the near future, which is generally not believed to be the case. In Helliwell’s paper, therefore, he argues that countries that are economically more prosperous are more likely to adopt a more democratic regime type. The findings presented in this paper do not provide evidence that point towards this conclusion. However, in order to detect reversed causality, a different and more sophisticated method is used in Helliwell’s research. And therefore, although the evidence provided here does not hint on the idea that economic prosperity stimulates democracy instead of the other way around, the usage of a more sophisticated methodology might prove differently.

Furthermore, some tests were performed in order to examine whether the assumptions for the OLS model that is used for the regressions were satisfied. Firstly the Breusch-Pagan test is applied in order to provide some insights on the homoscedasticity of the error terms. These tests detected heteroskedastic results for three regressions, namely, both for Albania and Spain the second regression, and for Albania the third (see Appendix B). To solve this problem, these problematic regressions were repeated, but this time with robust standard errors (see Appendix C). The alteration barely changed the outcomes, and therefore there is no need to revise the analysis and

interpretation provided in the previous paragraphs. Also, the Jarque-Bera test was performed on all the regressions in order to test on the normality of the residuals (see Appendix D). These tests revealed some problems concerning the normality of the error term in several regressions. This potentially has consequences for the validity of the results. Furthermore, the tests on

(19)

19 5.4 Results in Political Context

By considering the political context in which the transitions towards democracy took place, an attempt is made to interpret the results in more depth and possibly discover similarities between countries in the political transition process and the behavior of economic growth.

Albania and Bulgaria, for example, both come from a communistic regime, where in Greece the right-wing dominated and in Spain a nationalistic and quasi-fascist dictator held power. Also, similarities can be seen in the smoothness of the transition of power. In Albania a high degree of corruption and election fraud followed the transition, and in Portugal very turbulent times were faced both politically and socially. Greece, Bulgaria, and especially Spain on the other hand, appeared to adopt a more democratic system with more ease. Also, Greece, Portugal, and Spain are often considered as being part of the same ‘wave of democratization’ which could cause similarities in the democratization process, and therefore possibly in the outcomes of this research.

This, however, does not appear to be the case. One would, for example, expect Portugal to show a larger negative direct effect of democracy, since the transition was met with several difficulties, and not necessarily a positive indirect effect, since the economic situation does not appear to have been improved much after the implementation of democracy. Albania is another country that has dealt with a lot of political instability after the transition. However no similarities show between these two countries. The insignificant results on Albania on the other hand, could possibly be explained by the idea that democracy has not got the chance to function properly due to the corruption and election fraud, and hereby, there is no real representation of democracy and thereby its effect on economic growth. Also, the results of Albania an Bulgaria, both coming from a communistic regime, do not appear to be more distinctively similar as compared to the other results. The outcomes of Portugal and Spain are very significant and appear to be relatively similar, which could be backed by the ‘same wave’ argument. However, in reality this would be highly improbably, due to the striking differences in the rest of the political context. They come from different regime sorts, the transition happened due to varying motivations, and differ strongly in the degree to which the processes occurred smoothly. Furthermore, the presented political context does not provide a real argument onto why Greece does not appear to show a positive indirect effect and Bulgaria, Portugal, and Spain do. The last mentioned countries show the most similarities in the results, in the sense that they all show significant effects, where the direct effect is positive and the indirect effect negative. However, again no strong link in the political context that could explain these similarities can be identified, other than that they all go through the process of democratization. Therefore, the results provided in this paper do not provide any evidence that political context strongly influences

(20)

20 the effect of a transition towards democracy. However, of course, it does not prove, or seeks to prove, the opposite.

6. Concluding Remarks and Issues for Further Research

Based on the analysis presented in this paper, several conclusions on the influence of democracy on economic development in European countries can be drawn. Firstly, the direct effect of democracy seems to be negative and highly significant in most cases. In general, economic theory presented in previous research predicts the effect to be slightly negative but mostly insignificant. Therefore, these findings imply a stronger direct effect of democracy on economic growth than is generally believed. Regional specific effects provide a possible explanation for this conclusion, in the sense that European countries might be more heavily effected by democratic reform than other regions. The provided results on the indirect effect of democracy are in line with most economic theory. This effect appears to be positive, and mostly significant. Due to ambiguous and insignificant results, no evidence that point towards the problem of reversed causality is provided in this paper. Also, no real argument is made that implies the political context of the countries before, during, and after the transition towards democracy to strongly influence the outcomes considering economic development.

Although the results on the direct and indirect effect of democracy seem to be fairly

consistent over the five countries that are used for this analysis, a more extensive research, including a larger sample of countries, should be performed in order to draw a truly firm conclusion on the dependence economic growth has on democratic reform in European countries. Also, one should be cautious with rejecting the idea of reversed causality. A more sophisticated method, as presented in Helliwell’s paper, where democracy functions as the dependent variable, and several measures of economic performance act as independent variables, would likely provide more insightful results on this matter. Additionally, in order to further investigate the possible regional specific effect

democracy might have on European countries, a more comprehensive research has to be performed, where the results of countries from strictly the European region are compared with those of other parts of the world. Lastly, if similar data and methodology as presented in this paper were to be used in future research, more reliable results might be provided by solving the problem of the

(21)

21

7. References

Costa Pinto, A. (2006). Authoritarian legacies, transitional justice and state crisis in Portugal's democratization. Democratization, 13(02), 173-204.

Dahl, R. A. (1973). Polyarchy: Participation and opposition. Yale University Press.

Doucouliagos, H., & Ulubaşoğlu, M. A. (2008). Democracy and Economic Growth: A Meta- Analysis.

American Journal of Political Science, 52(1)

Edles, L. D. (1995). Rethinking democratic transition: A culturalist critique and the Spanish case. Theory and Society, 24(3), 355-384.

Gerring, J., Bond, P., Barndt, W. T., & Moreno, C. (2005). Democracy and economic growth: A historical perspective. World Politics, 57(03), 323-364.

Helliwell, J. F. (1994). Empirical linkages between democracy and economic growth. British journal of

political science, 24(02), 225-248.

Kajsiu, B., Bumçi, A., & Rakipi, A. (2003). Albania-a weak democracy, a weak state. Albanian Institute

for International Studies

Krieckhaus, J. (2006). Democracy and Economic Growth: How Regional Context Influences Regime Effects. British Journal of Political Science, 36(02), 317.

Lipset, S. M. (1959). Some social requisites of democracy: Economic development and political l egitimacy. American political science review, 53(01), 69-105.

Persson, Torsten and Guido Tabellini. (2006). Democracy and Development: The Devil in the Details.

American Economic Review, 96(2), 319-324.

Poulantzas, N. (1976). The Crisis of Dictatorships: Portugal, Greece, Spain. Translated by David Fernbach.London: Humanities Press, 1976

Stoper, E., & Ianeva, E. (1996). Democratization and Women's Employment Policy in Post-Communist Bulgaria. Conn. J. Int'l L., 12, 9.

(22)

22

Appendix A: Regressions

Albania:

_cons -7.809701 9.376791 -0.83 0.414 -27.30981 11.6904 demlead 1.699397 1.284014 1.32 0.200 -.9708551 4.36965 demlag .2319231 .4152495 0.56 0.582 -.6316356 1.095482 dem0 -.2914251 .4930635 -0.59 0.561 -1.316807 .7339567 GDPalb Coef. Std. Err. t P>|t| [95% Conf. Interval] Total 2160.16409 24 90.006837 Root MSE = 9.3522 Adj R-squared = 0.0283 Residual 1836.7219 21 87.4629477 R-squared = 0.1497 Model 323.442186 3 107.814062 Prob > F = 0.3228 F(3, 21) = 1.23 Source SS df MS Number of obs = 25 . regress GDPalb dem0 demlag demlead

. _cons 3.918099 3.118511 1.26 0.222 -2.549296 10.3855 demlag .2933433 .4196384 0.70 0.492 -.5769335 1.16362 dem0 .1944382 .3347362 0.58 0.567 -.4997622 .8886387 GDPalb Coef. Std. Err. t P>|t| [95% Conf. Interval] Total 2160.16409 24 90.006837 Root MSE = 9.5106 Adj R-squared = -0.0049 Residual 1989.92754 22 90.4512519 R-squared = 0.0788 Model 170.236546 2 85.1182728 Prob > F = 0.4054 F(2, 22) = 0.94 Source SS df MS Number of obs = 25 . regress GDPalb dem0 demlag

.

_cons 2.19088 1.881363 1.16 0.256 -1.701015 6.082775 dem0 .3263554 .27338 1.19 0.245 -.2391741 .8918849 GDPalb Coef. Std. Err. t P>|t| [95% Conf. Interval] Total 2160.16409 24 90.006837 Root MSE = 9.4043 Adj R-squared = 0.0174 Residual 2034.12694 23 88.4403018 R-squared = 0.0583 Model 126.037146 1 126.037146 Prob > F = 0.2447 F(1, 23) = 1.43 Source SS df MS Number of obs = 25 . regress GDPalb dem0

(23)

23

Bulgaria:

_cons -48.91512 31.04052 -1.58 0.130 -113.4674 15.63717 demleadb 6.422695 3.674587 1.75 0.095 -1.219026 14.06442 demlagb .5516166 .123442 4.47 0.000 .294905 .8083283 dem0b -.7433372 .2448767 -3.04 0.006 -1.252586 -.2340881 GDPbul Coef. Std. Err. t P>|t| [95% Conf. Interval] Total 581.164201 24 24.2151751 Root MSE = 3.4918 Adj R-squared = 0.4965 Residual 256.048887 21 12.1928042 R-squared = 0.5594 Model 325.115314 3 108.371771 Prob > F = 0.0005 F(3, 21) = 8.89 Source SS df MS Number of obs = 25 . regress GDPbul dem0b demlagb demleadb

. _cons 5.315219 .9735973 5.46 0.000 3.296101 7.334336 demlagb .573185 .1284322 4.46 0.000 .306833 .839537 dem0b -.3582765 .1117991 -3.20 0.004 -.5901336 -.1264195 GDPbul Coef. Std. Err. t P>|t| [95% Conf. Interval] Total 581.164201 24 24.2151751 Root MSE = 3.6513 Adj R-squared = 0.4494 Residual 293.298456 22 13.331748 R-squared = 0.4953 Model 287.865745 2 143.932872 Prob > F = 0.0005 F(2, 22) = 10.80 Source SS df MS Number of obs = 25 . regress GDPbul dem0b demlagb

.

_cons 2.722763 1.054789 2.58 0.017 .5407659 4.90476 dem0b -.1284099 .1339583 -0.96 0.348 -.4055238 .148704 GDPbul Coef. Std. Err. t P>|t| [95% Conf. Interval] Total 581.164201 24 24.2151751 Root MSE = 4.9292 Adj R-squared = -0.0034 Residual 558.837968 23 24.297303 R-squared = 0.0384 Model 22.3262335 1 22.3262335 Prob > F = 0.3477 F(1, 23) = 0.92 Source SS df MS Number of obs = 25 . regress GDPbul dem0b

(24)

24

Greece:

_cons 6.527426 1.302403 5.01 0.000 3.850298 9.204554 demleadg -.2582428 .1351199 -1.91 0.067 -.5359857 .0195001 demlagg .100493 .1297718 0.77 0.446 -.1662567 .3672427 dem0g -.3180503 .1076665 -2.95 0.007 -.5393619 -.0967387 GDPgre Coef. Std. Err. t P>|t| [95% Conf. Interval] Total 572.220824 29 19.7317525 Root MSE = 3.6547 Adj R-squared = 0.3231 Residual 347.276859 26 13.3568023 R-squared = 0.3931 Model 224.943965 3 74.9813218 Prob > F = 0.0042 F(3, 26) = 5.61 Source SS df MS Number of obs = 30 . regress GDPgre dem0g demlagg demleadg

. _cons 4.646741 .8941276 5.20 0.000 2.812143 6.481339 demlagg .1420164 .1340779 1.06 0.299 -.1330889 .4171216 dem0g -.3402772 .1121715 -3.03 0.005 -.5704341 -.1101203 GDPgre Coef. Std. Err. t P>|t| [95% Conf. Interval] Total 572.220824 29 19.7317525 Root MSE = 3.83 Adj R-squared = 0.2566 Residual 396.065697 27 14.6690999 R-squared = 0.3078 Model 176.155127 2 88.0775637 Prob > F = 0.0070 F(2, 27) = 6.00 Source SS df MS Number of obs = 30 . regress GDPgre dem0g demlagg

.

_cons 5.036711 .816581 6.17 0.000 3.364021 6.709401 dem0g -.3631725 .1103085 -3.29 0.003 -.5891292 -.1372158 GDPgre Coef. Std. Err. t P>|t| [95% Conf. Interval] Total 572.220824 29 19.7317525 Root MSE = 3.8384 Adj R-squared = 0.2533 Residual 412.523261 28 14.7329736 R-squared = 0.2791 Model 159.697563 1 159.697563 Prob > F = 0.0027 F(1, 28) = 10.84 Source SS df MS Number of obs = 30 . regress GDPgre dem0g

(25)

25

Portugal:

_cons 5.551089 1.218276 4.56 0.000 3.046886 8.055292 demleadp .1010475 .1227564 0.82 0.418 -.1512818 .3533769 demlagp .2618583 .1090155 2.40 0.024 .0377738 .4859427 dem0p -.3608668 .0934938 -3.86 0.001 -.553046 -.1686876 GDPpor Coef. Std. Err. t P>|t| [95% Conf. Interval] Total 507.651372 29 17.5052197 Root MSE = 3.4693 Adj R-squared = 0.3124 Residual 312.941931 26 12.0362281 R-squared = 0.3835 Model 194.709441 3 64.903147 Prob > F = 0.0051 F(3, 26) = 5.39 Source SS df MS Number of obs = 30 . regress GDPpor dem0p demlagp demleadp

. _cons 6.232237 .8887747 7.01 0.000 4.408622 8.055852 demlagp .258554 .1082891 2.39 0.024 .0363631 .4807449 dem0p -.3254008 .0824774 -3.95 0.001 -.4946304 -.1561712 GDPpor Coef. Std. Err. t P>|t| [95% Conf. Interval] Total 507.651372 29 17.5052197 Root MSE = 3.4486 Adj R-squared = 0.3206 Residual 321.09749 27 11.8924996 R-squared = 0.3675 Model 186.553882 2 93.276941 Prob > F = 0.0021 F(2, 27) = 7.84 Source SS df MS Number of obs = 30 . regress GDPpor dem0p demlagp

.

_cons 4.742036 .683808 6.93 0.000 3.341319 6.142753 dem0p -.2209461 .07556 -2.92 0.007 -.3757239 -.0661684 GDPpor Coef. Std. Err. t P>|t| [95% Conf. Interval] Total 507.651372 29 17.5052197 Root MSE = 3.7268 Adj R-squared = 0.2066 Residual 388.89379 28 13.8890639 R-squared = 0.2339 Model 118.757582 1 118.757582 Prob > F = 0.0068 F(1, 28) = 8.55 Source SS df MS Number of obs = 30 . regress GDPpor dem0p

(26)

26

Spain:

_cons 6.616637 .6403978 10.33 0.000 5.300281 7.932994 demleads -.1817461 .0660122 -2.75 0.011 -.3174361 -.046056 demlags .2842168 .0750701 3.79 0.001 .1299081 .4385256 dem0s -.2496092 .0567939 -4.40 0.000 -.3663508 -.1328677 GDPspa Coef. Std. Err. t P>|t| [95% Conf. Interval] Total 265.90879 29 9.16926862 Root MSE = 1.839 Adj R-squared = 0.6312 Residual 87.9285575 26 3.3818676 R-squared = 0.6693 Model 177.980232 3 59.3267442 Prob > F = 0.0000 F(3, 26) = 17.54 Source SS df MS Number of obs = 30 . regress GDPspa dem0s demlags demleads

. _cons 5.523089 .5602241 9.86 0.000 4.373605 6.672574 demlags .2935269 .0836346 3.51 0.002 .1219229 .4651308 dem0s -.3283107 .0547302 -6.00 0.000 -.4406078 -.2160136 GDPspa Coef. Std. Err. t P>|t| [95% Conf. Interval] Total 265.90879 29 9.16926862 Root MSE = 2.0509 Adj R-squared = 0.5413 Residual 113.563846 27 4.20606836 R-squared = 0.5729 Model 152.344944 2 76.1724722 Prob > F = 0.0000 F(2, 27) = 18.11 Source SS df MS Number of obs = 30 . regress GDPspa dem0s demlags

.

_cons 4.070503 .4474019 9.10 0.000 3.154041 4.986964 dem0s -.2292494 .0555647 -4.13 0.000 -.3430684 -.1154303 GDPspa Coef. Std. Err. t P>|t| [95% Conf. Interval] Total 265.90879 29 9.16926862 Root MSE = 2.4303 Adj R-squared = 0.3559 Residual 165.372286 28 5.90615308 R-squared = 0.3781 Model 100.536504 1 100.536504 Prob > F = 0.0003 F(1, 28) = 17.02 Source SS df MS Number of obs = 30 . regress GDPspa dem0s

(27)

27

Appendix B: Tests for Heteroskedasticity

regress GDPalb dem0

regress GDPalb dem0 demlag

regress GDPalb dem0 demlag demlead

regress GDPbul dem0b

Prob > chi2 = 0.3339 chi2(1) = 0.93

Variables: fitted values of GDPalb Ho: Constant variance

Breusch-Pagan / Cook-Weisberg test for heteroskedasticity

Prob > chi2 = 0.5302 chi2(1) = 0.39

Prob > chi2 = 0.0154 chi2(1) = 5.87

Prob > chi2 = 0.1623 chi2(1) = 1.95

Variables: fitted values of GDPbul Ho: Constant variance

(28)

28 regress GDPbul dem0b demlagb

regress GDPbul dem0b demlagb demleadb

regress GDPgre dem0g

regress GDPgre dem0g demlagg

Prob > chi2 = 0.0383 chi2(1) = 4.29

Prob > chi2 = 0.2104 chi2(1) = 1.57

Prob > chi2 = 0.9452 chi2(1) = 0.00

Variables: fitted values of GDPgre Ho: Constant variance

Prob > chi2 = 0.8575 chi2(1) = 0.03

(29)

29 regress GDPgre dem0g demlagg demleadg

regress GDPpor dem0p

regress GDPpor dem0p demlagp

regress GDPpor dem0p demlagp demleadp Prob > chi2 = 0.5598 chi2(1) = 0.34

Prob > chi2 = 0.8837 chi2(1) = 0.02

Variables: fitted values of GDPpor Ho: Constant variance

Prob > chi2 = 0.5017 chi2(1) = 0.45

Prob > chi2 = 0.6067 chi2(1) = 0.27

(30)

30 regress GDPspa dem0s

regress GDPspa dem0s demlags

regress GDPspa dem0s demlags demleads

Prob > chi2 = 0.5075 chi2(1) = 0.44

Variables: fitted values of GDPspa Ho: Constant variance

Prob > chi2 = 0.0458 chi2(1) = 3.99

Prob > chi2 = 0.0630 chi2(1) = 3.46

(31)

31

Appendix C: Robust Regressions

_cons 5.523089 .4480996 12.33 0.000 4.603665 6.442514 demlags .2935269 .0412773 7.11 0.000 .2088328 .3782209 dem0s -.3283107 .0464467 -7.07 0.000 -.4236115 -.2330098 GDPspa Coef. Std. Err. t P>|t| [95% Conf. Interval] Robust Root MSE = 2.0509 R-squared = 0.5729 Prob > F = 0.0000 F(2, 27) = 36.77 Linear regression Number of obs = 30 . regress GDPspa dem0s demlags, vce(robust)

. _cons 5.315219 .6829368 7.78 0.000 3.898894 6.731543 demlagb .573185 .1084451 5.29 0.000 .3482836 .7980863 dem0b -.3582765 .1247712 -2.87 0.009 -.6170361 -.099517 GDPbul Coef. Std. Err. t P>|t| [95% Conf. Interval] Robust Root MSE = 3.6513 R-squared = 0.4953 Prob > F = 0.0001 F(2, 22) = 15.92 Linear regression Number of obs = 25 . regress GDPbul dem0b demlagb, vce(robust)

. _cons -7.809701 11.92695 -0.65 0.520 -32.61315 16.99375 demlead 1.699397 1.653644 1.03 0.316 -1.739544 5.138339 demlag .2319231 .3603659 0.64 0.527 -.5174988 .981345 dem0 -.2914251 .5633198 -0.52 0.610 -1.462913 .8800626 GDPalb Coef. Std. Err. t P>|t| [95% Conf. Interval] Robust Root MSE = 9.3522 R-squared = 0.1497 Prob > F = 0.0251 F(3, 21) = 3.81 Linear regression Number of obs = 25 . regress GDPalb dem0 demlag demlead, vce(robust)

(32)

32

Appendix D: Tests on Normality of the Residuals

Jarque-Bera test for Ho: normality:

Jarque-Bera normality test: 29.77 Chi(2) 3.4e-07 . jb ra

Jarque-Bera normality test: 21.4 Chi(2) 2.3e-05 . jb ra1

Jarque-Bera normality test: 10.4 Chi(2) .0055 . jb ra2

Jarque-Bera normality test: 1.381 Chi(2) .5013 . jb rb

Jarque-Bera normality test: .1816 Chi(2) .9132 . jb rb1

Jarque-Bera normality test: .0823 Chi(2) .9597 . jb rb2

Jarque-Bera normality test: 3.558 Chi(2) .1688 . jb rg

Jarque-Bera normality test: 5.915 Chi(2) .0519 . jb rg1

Jarque-Bera normality test: 13.37 Chi(2) .0012 . jb rg2

Jarque-Bera normality test: 11.41 Chi(2) .0033 . jb rp

Jarque-Bera normality test: 11.23 Chi(2) .0036 . jb rp1

Jarque-Bera normality test: 17.1 Chi(2) 1.9e-04 . jb rp2

Jarque-Bera normality test: .3886 Chi(2) .8234 . jb rs