Institutions and Economic Growth: Does Income Level Matter?

Munich Personal RePEc Archive Institutions and Economic Growth: Does Income Level Matter? Aziz, Nusrate and Ahmad, Ahmad H. Algoma University, Ontario, Canada;, Loughborough University, Leicestershire, United Kingdom 6 January 2018 Online at https://mpra.ub.uni-muenchen.de/83684/ MPRA Paper No. 83684, posted 11 Jan 2018 06:29 UTC Institutions and Economic Growth: Does Income Level Matter? Nusrate Aziz1 and Ahmad H Ahmad2 Abstract This paper investigates whether a country’s level of income matters to the effectiveness of institutions in fostering economic growth. The institutional variables are represented by democracy, corruption levels, and armed conflicts. The countries in the data-set are divided into high-, middle- and low-income countries based on the World Bank criteria. The overall results indicate that institutional variables have offsetting effects on economic growth. The performance of these variables appears to have been influenced by the countries’ level of income. Labour, capital and human capital are found to be positive and significant variables for economic growth, irrespective of whether the countries are in high-, middle- and low-income groups. On the contrary, corruption affects GDP negatively in high- and middle-income groups, but positive, although insignificant in low-income countries. Democracy has a mixed effect on economic growth and largely negative in high- and low-income countries, but positive in middle-income group. Armed conflicts do not appear to have any statistically significant effect on high and middle-income countries’ economic growth. However, it has a significant negative effect on low-income countries’ economic growth. Keywords: Institutions, income level, economic growth, panel study JEL Code: C23, O43, O47 1 2 Algoma University, Ontario, Canada, E-mail: [email protected] Loughborough University, UK 1 1. Introduction Identifying sources of economic growth and the factors that could hinder it has been at the forefront of academic research with relevant policy implication. Initial work in the area is based on the neo-classical theory, which can be regarded as supply-oriented (Federici and Marconi, 2002). However, these models did not explicitly recognise the role of domestic policies, including trade policies to growth. The work of Kaldor (1970) introduced the role of international trade, particularly foreign-demand in fostering and sustaining economic growth to the literature. This sub-literature became known as the demand-oriented theory. The new growth theory also known as the endogenous growth models that largely based on the work of Romer (1986) and Lucas (1988) recognised the determinant role played by skills augmented efficiency, completion and capital flows to economic growth. Coe and Helpman (1995), Arrow (1962) Uzawa (1965) and Solow (1969) have all contributed towards the development of the theory. Recent Literature (see among others Berg, et al., 2012; Hausmann et al., 2006; Easterly et al., 2006) has identified important role played by institutional factors to growth, which have been largely overlooked by both the neo-classical and the endogenous growth theories. Empirical literature has also confirm importance of such variables to a countries growth, which include political variables like democracy, government stability, economic freedom, violence, frequent armed conflicts and level of corruption. A survey of this literature has been provided by Brunneti (1997). However, what is not clear is whether level of income of a country is a factor in the importance of these sunspots variables. This paper has two objectives. First, it aims at bringing in further empirical evidence concerning the role of these variables in economic growth with a particular emphasis on the role of ‘polity’ score, existence of armed conflicts, and level of corruption, using both crosssectional and panel data for 126 and 106 countries, respectively. Secondly, the paper investigates whether the effects of these variables vary according to the income level of a country. To this end, the countries in the sample are grouped into low-, middle- and highincome countries, using the World Bank classification scheme. This study therefore examines two sets of variables – the traditional and institutional variables - in a growth model. Literature suggests that although there is a consensus about the positive effect of traditional variables such as labour and capital, on economic growth, however, the 2 influence of institutional variables are controversial. Existing studies apply either democracy or corruption or armed conflict in the growth regression. This is the first study (to the best of our knowledge) which applies these variables in the growth regression. This study will also assist in reducing the controversy about the ambiguous effect of institutional variables on economic growth. The rest of the paper is organised as follows. Section 2 provides a brief review of the relevant literature on the importance of corruption, polity and conflict on economic growth. Section 3 discusses the theoretical framework and presents the empirical model used for the analysis as well as the data-set. Section 4 presents the empirical results while Section 5 concludes. 2. Brief Overview of the Relevant Literature The modern-day world has come off a long way from the organic view of the state in which the existence of the citizens and their activities would exclusively mean for the welfare of the ‘state’ rather than the citizens themselves. The role of the government has thus shifted from maintenance of law and order, and governance to enhancing the standards of living of the citizens. This is, probably more true in democratic societies where leadership is conferred with mandates by the citizens than in other political systems. The perceived new role of the government apparently makes it to be performance-oriented. The achievement of these governments is contingent upon the presence or absence of certain parameters. The set of parameters include, among other things, governance, political violence, political volatility, corruption, and armed conflicts. In emphasising the role of the institutional variables, the contemporary growth literature brings to the forefront the institutional view on economic growth. The role of institutions in economic development was first identified by Lewis (1955). Later literature considers institutions as potential sources of differences in cross-country differences in growth (see, for example, North and Thomas (1973); Acemoglu et al. (2005); IMF (2005)). Rodik (2005) develops a four-cluster taxonomy of institutions that is vital to the study of economic growth. The taxonomy includes (a) market-creating institution that ensures the security of the property rights and enforcement of contracts; (b) marketing-regulating institution responsible for command and control; (c) market-stabilizing institution chocking out fiscal and monetary policies; and (d) market-legitimising institutions that refer to the political regime that oversees the operation of the market. Thus, a clear synergy between economic institutions as embedded in the neoclassical theory, political institutions, and political regimes is now discernible. 3 The literature on economic institution-growth nexus is still evolving. The literature in this area largely draws from the development of economic institutions in many European colonies in the past 500 years. These included the provision for private property, introduction and/or maintenance of extractive institutions, migration of the Europeans to sparsely populated regions, introduction of legal rights and the quality thereof in protecting the investors, among other issues. Empirical evidence, though not free from controversy, is indicative of a positive impact of economic institutions on economic growth. Acemoglu, Johnson and Robinson (2001, 2002) find positive effects of the development of private property and the introduction of extractive institutions in previously poor regions. Acemoglu (2001) finds that settlements of Europeans, as proxied by mortality rates 100 years ago, have no effect on per capita GDP today. However, mortality rates are likely to have contributed to the development of institutions that may affect growth. La Porta et al. (1997, 1998) show that the degree of investor protection as spelled out in the legal systems has implications for the development of equity and stock markets. Better investor protection leads to greater debt and equity markets and also to better labour-market conditions which in turn may contribute to growth (Botero et al., 2004; Mahoney, 2001). Deger, Lam and Sen (2011) find positive relationship between growth and economic institutions. The theoretical underpinnings of the role of institutional variables3 to economic growth have been brought to light by Cass and Shell (1981, 1983). Cass and Shell (1983) argue that while institutions do not matter in the static Arrow-Debru economy with complete markets, it may matter in overlapping-generations models under certain conditions. Further, in the presence of institutions, equilibrium allocations are Pareto optimal in a ‘weaker’ sense’, “which is appropriate to dynamic analysis”. Bruneti (1997) has an extensive survey of the empirical literature concerning the effects of the sunspot variables on economic growth. The survey reviewed five categories of papers respectively concentrating on democracy, political volatility, government stability, political violence, and subjective political measures. Measures of political volatility and subjective political indicators have been found to have significant effects of economic growth followed by government stability and political violence. Other literature examines and re-examined the impact of traditional determinants such as economic endowment, labour force, physical capital and human capital of economic growth. There is hardly any debate about the direct and positive impact of these variables on economic growth. 3 They are also alternatively known as “animal spirits” or “market psychology”. 4 The focus of this study is therefore not the traditional determinants rather institutional factors of economic growth. Democracy appears to have mixed results, and in most cases being unsuccessful in explaining economic growth. Fidrmuc (2003) suggests that democracy strengthens economic liberalization and effectively contributes towards growth. De Hann and Sturm (2003) find that democracy leads to greater economic freedom, which is an important ingredient in fostering economic growth of developing countries. Tavares and Wacziarg (2001) find that the net effect of democracy on growth is moderately negative. Barro (1996) and Helliwell (1994) also indicate an insignificant effect of democracy on the economic growth. Chan (2002), Dornbusch and Edwards (1991), Kohli (2004), and Leftwich (2005) document that democratic government finds it difficult to initiate painful economic reforms, which may have adversely affect the welfare of the people, even in the short-run. Deger, Lam and Sen (2011) find that political institutions including democracy do not have any conclusive effects on economic growth. A separate line of research on the association between democracy and economic growth conjectures that democracies and autocracies achieve, on average, equal economic growth, though democracies are less volatile (Doucouliago and Ulubasoglu, 2006; Mulligan, Giland Sala-i-Martin, 2004). Polity (a measure of democracy), in this study, is therefore included as one of the institutional determinants of economic growth to re-examine the impact of it on economic growth. Similarly corruption has been found to have adverse effects on the economic growth (Gyimah-Brempong, 2002; Keefer and Knack, 1997; Knack and Keefer, 1995; Li et al., 2000; Mauro, 1995; Mo, 2001; Sachs and Warner, 1997). Ades and Di (1997), Mauro (1995), and Meon and Sekkat (2005) find that the association between corruption and economic growth to be constantly negative and more dominant in countries with allegedly high levels of red tape, weak legal system and extensive government inefficiencies. Mauro (1995) also states that corruption decreases the quantity of private investments, which ultimately adversely affects the growth. Tanzi and Davoodi (1997), Ehrlich and Lui (1999), Sarte (2000), Aidt et. al (2007), Blackburna and Forgues-Puccio (2010), and Park (2012) also document negative effect of corruption on economic growth. Unlike democracy, the negative impact of corruption on economic growth is less controversial. This study investigates the effect of corruption using both cross-sectional and panel data which is missing in existing studies. A large number of studies have investigated the empirical relationship between conflict (which is our third institutional variable for economic growth) and growth. Barro and Lee (1994) examine the factors affecting economic growth from 1965 to 1985 in a large cross-section of countries and 5 find insignificant relationship between war and economic growth. Murdoch and Sandler (2002) find that civil war has an inverse relationship with short-term economic growth but does not impact long-term economic growth. Koubi (2005) finds that war severity and duration seems to contribute positively to subsequent economic growth, although war may have a negative and contemporary effect on economic growth in the short-run due to devastation of productive resources. Daria (2009) finds no direct association between war and economic growth. It can be concluded that there is little consensus about any specific effect of armed conflict on economic growth. Overall, it seems that although armed conflict negatively effects economic growth in the short-run, the long-run implication of armed conflict is ambiguous. 3.1 The Theoretical Framework, the Empirical Model and the Data The theoretical framework is based on a two-factor Cobb-Douglas production function that encompasses two basic factors of production, labour and capital, which positively affect economic growth with probably different size of contribution, represented by α and 𝛽 respectively. There are also institutional and infrastructural factors, which can be very influential determining factors to output growth. These factors can be denoted by 𝐴, represents the initial endowments of a country and therefore, capturing the differences in productivity across countries. Besides, the literature also suggests a ‘state capacity’ variable in the growth equation, which also can be captured by 𝐴. Several papers have found that human capital is also an important determinant of economic growth (see, for example, Mankiw, Romer and Weil, 1992; Mankiw, Phelps and Romer, 1995). Putting all these together provides traditional variables of the initial endowment (A), labour (L), physical capital (K) and human capital (H). As representatives of the traditional variables, the present study includes level of non-corruption, existence of armed conflicts, and level of democracy into the model. Accordingly, the model can be represented as: 𝑌 = 𝐴𝐿𝛼 𝐾𝛽 𝐻1−𝛼−𝛽 𝐸 𝛿 ; 𝛼 > 0, 𝛽 > 0, 𝛿 ⋚ 0. (1) This indicates that labour, physical capital and human capital positively contribute to production while the combination of the institutional variables as outlined here may have an offsetting effect, positive effect or insignificant influence on economic growth. The model specified in equation (1) can be re-written in logs as: 𝑙𝑛𝑌𝑖𝑡 = 𝜇𝑖 + ln(𝐴)𝑖 + 𝛼𝑙𝑛𝐿𝑖𝑡 + 𝛽𝑙𝑛𝐾𝑖𝑡 + (1 − 𝛼 − 𝛽)𝑙𝑛𝐻𝑖𝑡 + 𝛿𝑙𝑛𝐸𝑖𝑡 + 𝜀𝑖𝑡 6 (2) where, 𝑌 is output of country 𝑖 at time 𝑡; 𝜇 is the country-specific effect; 𝐴 is initial endowment of the country 𝑖; 𝐸 is the vector of institutional variables as defined above and 𝜀 is an error term. Based on equation (2) above, the empirical model is given as follows: 𝑙𝑛𝑌𝑖,𝑡 = 𝜇𝑖 + 𝛽1 𝑙𝑛𝑌𝑖,𝑡−1 + 𝛽2 𝑙𝑛𝐿𝐹𝑖,𝑡 + 𝛽3 𝑙𝑛𝐾𝑖,𝑡 + 𝛽4 𝑙𝑛𝐻𝐾𝑖,𝑡 + 𝛽5 𝑙𝑛𝑃𝑖,𝑡 + 𝛽6 𝑙𝑛𝑁𝑃𝐼𝑖,𝑡 + 𝛽7 𝐶𝐷𝑈𝑀 + 𝜀𝑖,𝑡 (3) where 𝛽1 > 0; 𝛽2 > 0; 𝛽3 > 0; 𝛽4 > 0; 𝛽5 ⋚ 0; 𝛽6 < 0; 𝛽7 ≦ 0. 𝑌𝑖𝑡 denotes PPP-adjusted GDP (constant 2005 international $) for each country 𝑖 over the period from 2000 to 20094. 𝐿 represents labour force, 𝐾 is gross fixed capital formation as percentage of GDP, 𝐻𝐾 is the country’s human capital and proxied by the percentage of population completed secondary education aged 25 and over. 𝑃 is the polity score, 𝑁𝑃𝐼 is non-corruption perception index, C_Dum is the armed conflict dummy, which takes the value of 1 if there is an incidence of conflict and 0 otherwise while 𝜀 is an error term. The cross-sectional model uses the mean values for the covered period (2000-2009) of all the variables. All the variables except armed conflict dummy are the average (2000-2009) of their respective values over cross-section. Noncorruption perception index (NPI) data are available from 1995 for few countries, but it is available for most countries from 2000. Since NPI is one of the most important variables for this study, and it is only available for the sample countries from 2000 onwards, therefore, our analysis is limited to ten years, 2000 – 2009 due to limitation of data. The economic growth equation was estimated using both cross-sectional and panel data. Crosssectional data is used for 126 countries, while balanced panel data are obtained for 106 countries. Time series data cover from 2000 to 2009 period. Some countries whose variables, such as GDP, labour force, capital, corruption and armed conflict are available, but are excluded from this study as they do not have human capital and polity variables. Extending the time series may require dropping out many countries. It is worth noting that the number of countries included in the panel analysis is smaller than that of the cross-sectional study. This is because countries without all variables have been excluded in the panel analysis. Data on PPP-adjusted GDP and PPP-adjusted GDP per capita, labour force, and gross fixed capital formation (% GDP) variables are sourced from the ‘World Bank Development Indicators’ 4 An alternative model used PPP-adjusted GDP (constant 2005 international $) per capita. 7 (WDI). Human capital variables are sourced from Barro and Lee Database (2010). Democracy level (polity), non-corruptions and armed conflicts are compiled from the Centre for Systemic Peace and the Centre for Global Policy, George Mason University (April 30, 2010); Uppsala Conflict Data Program (1 August 2011); and Corruption Perceptions Index (various issues) of the Transparency International. Both the cross-sectional regression and the panel study techniques are use for estimation. We have tested the sensitivity of data and empirical model by using sub-group analysis. Full sample, high-income, middle-income and low-income countries data are applied in both crosssectional as well as panel study. Additionally, non-oil sub-sample is used in the cross-section analysis. We apply PPP-adjusted GDP and PPP-adjusted GDP per capita in alternative regressions. The panel study applied OLS as well as fixed effects models in full-sample and sub-samples. 3.2 The Variables i) Polity The original ‘polity’ variable from the dataset consists of scores which take values between 10 (strongly autocratic) and +10 (strongly democratic). However, these were converted into a range that runs from 0 to 20 in order to facilitate the conversion of the variables into their natural logs required for the analysis. The modified ‘polity’ variable, labelled ‘polity2’ is used. The advantage of using ‘polity2’ is that it has standardized the original scores and yielded positive figures that are required for log conversion. <Figure 1> Figure 1 plots the average polity distribution for 126 countries in the world.5 The world average is 14. There is very low polity score for the socialist countries (such as China) and kingdoms (such as Saudi Arabia, Kuwait, UAE, Qatar, and Bahrain) despite the fact that they have been growing faster than many democratic countries. Bhutan (-10), Qatar (-10) and Saudi Arabia (10), Turkmenistan (-9), Uzbekistan (-9), Swaziland (-9), UAE (-8), China (-7), Vietnam (-7), 5 Some names are omitted from the figure to avoid clumsiness in appearance. 8 Laos (-7) carry very low polity score. However, average growth of Bhutan (8.5%), Qatar (12.5%), Saudi Arabia (3.4%), Turkmenistan (14.2%), Uzbekistan (6.5%), Swaziland (3.1%), UAE (5.8%), China (10.3%), Vietnam (7.3%), Laos (6.8%) indicate that less polity score does not negatively affect economic growth. In reality, the average growth rate of these countries (7.8%) was way higher than world average growth rate (4.5%). Perhaps, the stability of economy, not necessarily the level of polity, plays significant role in economic growth. Nonetheless, we have excluded countries with polity score of 0 as conversion into the logarithm generates no value. ii) Corruption Corruption is, generally, perceived as detrimental to growth. This is supported by empirical findings of, among others, Mauro (1995), Brnetti and Weder (1998) and Mo (2001) who reported negative effects of corruption on growth as it discourages investment. However, Bardham (1997) Beck and Mahar (1986) and Lien (1986) counter argue that corruption, particularly, bribery could be beneficial to growth as it can “grease the wheels” of an inefficient bureaucracy. Development of businesses that were aided by corruption are cited as examples to buttress the point6. Corruption Perceptions Index, CPI as defined by the Transparency International (TI) 7 is “poll of polls” that show the average scores which are the reflection of opinions by international businesses people and financial journalists for all the countries in the world. Countries are ranked according to the perceptions of corruption level in the public sector. Thus it is an indicator of corruption level at as perceived by businesses and how it affects their commercial activities. The higher the score of NPI, the lower the level of perceived corruption by businesses for a country. The corruption indicator is denoted by Non-corruption Perception Index (NPI). Consequently, if corruption deters economic growth, we expect a positive sign for the coefficient of NPI variable. The NPI scores for full sample are plotted against GDP and GDP per capita in Figure 2a and 2b, respectively. <Figure 2> The world average of non-corruption score is found to be approximately 4, which is below half of the total score. Scandinavian countries are the top scorers as least corrupt countries while 6 This is consistent with argument of Leff (1964), Huntinton (1968) and Leys 1965), which was later come to be known as “grease the wheels” hypothesis. 7 The Berlin-based anti-corruption non-governmental organisation, TI, defines corruption as “the abasement of entrusted power for private gain”. 9 South Asia and Africa have the lowest scores as the most corrupt countries. The trend line gives an indication of positive relationship between non-corruption and GDP. iii) Armed Conflict The UCDP/PRIO Armed Conflict Dataset Codebook defines the term ‘conflict’ has been defined as: “a contested incompatibility that concerns government and/or territory where the use of armed force between two parties, of which at least one is the government of a state, results in at least 25 battle-related deaths”. Average annual numbers of battle deaths due to both internal and external conflicts are collected and a dummy variable is constructed for the armed conflicts. A value of 1 for the dummy denotes the presence of armed conflicts and a value of 0 denotes otherwise. It may be mentioned that after the end of the ‘cold war era’ (1947 to 1991) both internal and external armed conflicts had fallen dramatically. We find 39 countries (either internally or externally involved with armed conflicts) out of the 126 selected countries were involved in armed conflicts at least once between 2000 and 2009. Among them Algeria, Burundi, Colombia, Democratic Republic of Congo, India, Israel, Nepal, Pakistan, the Philippines, Russia, Sri Lanka, Thailand, Turkey, Uganda and USA were heavily involved in conflicts. A distinctive effect of internal to external conflict is beyond the scope of this study. A third country which was not directly involved in combat, however, indirectly played a role (supported) combating countries in their internal or external conflict is not considered in this study. Any further study may find these distinctions interesting. <Figure 3> Figures 3a and 3b represent GDP and GDP per capita and armed conflicts. It is apparent from the figure that GDP per capita is more sensitive to conflict than GDP. It is, therefore, higher for a country which is not engaged in armed conflict and vice versa. The triangle shape shows the GDP of non-conflicting while star shape shows GDP of conflicting countries. Very few countries (except some developed countries such as USA, UK, Russia, Israel, who can be considered as world leader), which maintain high GDP per capita are involved in armed conflict. The UK and the US were the main players in the war against Iraq8 (30 countries were 8 Countries which were involved in the Iraq invasion are: Albania, Australia, Azerbaijan, Bulgaria, Czech Republic, Denmark, Dominican Republic, El Salvador, Estonia, Georgia, Honduras, Italy, Kazakhstan, Latvia, Lithuania, Macedonia, Moldova, Mongolia, Netherlands, Nicaragua, Norway, Philippines, Poland, Portugal, South Korea, Romania, Slovakia, Spain, Tonga, Ukraine, United Kingdom, United States of America. 10 involved in military invasion under the American-led coalition) and (Taleban) Afghanistan9 (47 countries were involved) who spent a significant amount of money for war. Although many countries were physically involved in these wars, however, the war expenditure for many countries was very insignificant amount. Subsequently, these two external armed conflicts are typically different to others. Also, including all these countries into the analysis as those engaged in armed conflict between 2003 and 2008 (which covers almost our entire sample period), could lead to some misleading results. The nature of these wars is typically different to other internal and external conflicts. Types of expenditure are also different. Many involved countries in these wars have not faced any battle deaths except Iraqis, the British and the American. Differentiating the conflicts into internal and external is beyond the scope of this paper, but it may be an interesting future extension10. 4. Discussion of the Empirical Results Table 1 presents the descriptive statistics, which indicates that the world average of growth rate is 4.5 percent with a large discrepancy between countries, which is as high as 15 percent and as low as approximately half a percent. Average polity score is about 14 out of 20 in the world. China’s average polity score was 3, while its average GDP growth rate was about 10 percent during the same period. Qatar’s average polity score was 0 (zero), while the country’s average GDP growth rate was 13.5 percent in the last decade. On the contrary, Portugal’s average polity was 20 while the country’s average GDP growth rate was less than 1 percent. A similar feature (in terms of polity and economic growth) is observed in many other countries. The average NPI score for the world as a whole is about 4 out of 10, with the highest NPI score of 9.52 for Finland is perceived as the least corrupt country in the world. Denmark (9.46), New Zealand (9.45), Singapore (9.28), Sweden (9.24) and Iceland (9.22) are the other perceived less corrupt countries. Those with the lowest scores of 1.7 point are Afghanistan and Bangladesh, perceived as the most corrupt countries in the world.11 Chad (1.73), Sudan (1.89), Democratic Republic 9 Countries which were involved in war against Taleban (in Afghanistan) are: Albania, Armenia, Australia, Austria, Azerbaijan, Belgium, Bosnia and Herzegovina, Bulgaria, Canada, Croatia, Czech Republic, Denmark, Estonia, Finland, France, Georgia, Germany, Greece, Hungary, Iceland, Ireland, Italy, Jordan, Latvia, Lithuania, Luxembourg, Macedonia (Former Yugoslav Republic of), Malaysia, Mongolia, Montenegro, Netherlands, New Zealand, Norway, Poland, Portugal, Romania, Singapore, Slovakia, Slovenia, South Korea, Spain, Sweden, Turkey, Ukraine, United Arab Emirate, United Kingdom, United States of America. 10 A third country, which is not directly involved in combat, but indirectly paly a supportive role is excluded in the analysis. 11 Somalia and Myanmar have lower NPI scores than Afghanistan and Bangladesh. Somalia, Myanmar and Afghanistan however are not included in our panel. Afghanistan and Bangladesh were included in our crosssection. 11 of Congo (1.94) are perceived as the next most corrupt countries, for the sample period. Ethiopia is the worse armed-conflict-hit country in the sample period, which was followed by Afghanistan, Pakistan, Sri Lanka, Sudan and India. The least hit country was Azerbaijan which was followed by Peru, China, Iran and Tajikistan. <Table 1> The democracy indicator, the polity, as expected, has the highest mean in the high-income countries and followed by the middle-income countries. The low-income countries group has the least average of polity. However, these are averages and can obscure some individual countries’ positions. For example, there are a number of countries that are classified as highincome countries, but are under autocratic governments. These countries include Qatar, Bahrain and the United Arab Emirates. This is evident in the difference between the minimum and the maximum. The minimum is 1, which is highly autocratic while the maximum is 20, that is, highly democratic. The scenario is similar for middle- and low-income countries, albeit, to a lesser degree in the latter. This study applies both cross-section and panel techniques. Multicollinearity is not a vital issue for cross-section. However, since, panel data includes both time-series and cross-section, multicollinearity could be a problem. Correlation between the variables in panel is tested to examine any potential multicollinearity problem. Table 2 presents the correlation matrix and it indicates that the degree of correlation between the independent variables in the empirical model and shows that the correlation is not significantly high. Therefore, multicollinearity is not likely to be a problem. The model specified in equation (3) is estimated and the results are reported in Tables 3, 5 and 6. Table 3 presents the results obtained from the cross-sectional estimation by applying mean values (2000-2009) of variables, while Tables 5 and 6 report panel study results by applying the pooled OLS and the fixed effects estimators, respectively. 4.1 Mean Cross-Section: Real GDP (PPP) per capital and real GDP (PPP) are used as dependent variables in alternative estimations. A full-sample (which includes collectively high, medium and low income countries, and does not differentiate between oil producing and non-oil producing countries) 12 and sub-samples based on different income level and oil production status are used which give us an opportunity to examine the sensitivity of the results. All variables are in real term and in logarithm form except armed conflict (which is a binary dummy). Qatar and Saudi Arabia are excluded from sample because polity score for them is zero which cannot be converted into logarithm. Total sample size in the cross-section and panel are 126 and 106 countries, respectively. The reported results from the cross-section show that, in general, the traditional variables, which are labour, capital and human capital foster GDP growth. However, the institutional determinants of growth that include the polity and the corruption variables are found to be negative and significant. A negative sign of the coefficient of the polity variable indicates that the greater the degree of democracy, the lower will be the GDP growth. <Table 3> Corruption is found to be a key player in economic growth. It is worth mentioning that high score of NPI indicates suggest less perceived corruption level. A positive sign of NPI (Noncorruption Perception Index) therefore indicates that more non-corruption leads to more GDP growth; hence, more corruption leads to less economic growth. Corruption variable is found to be significant in full-sample as well as in each sub-sample, except the low-income countries. The size of the NPI coefficient is also very high compare to other variables. In fact, the positive effect of traditional variables is fully neutralised by the negative effect of corruption variable alone. Consequently, what remains in GDP growth is actually only the autonomous growth. Specifically, in full-sample model (for example), the significant traditional variables jointly contribute 0.562 to the real GDP per capita and 1.52 to real GDP; however, these contributions have been neutralised by corruption alone (1.77 and 1.685, respectively). The coefficient of corruption is found statistically greater than one in both GDP per capita and GDP models. This finding is consistent with the literature that suggests corruption is “sand in the wheels”. That is corruption constitute an obstacle to growth (Mauro, 1995; Brutietti and Weder, 1998; and Mo, 2001). Although the estimated result from cross-section on full sample regression shows that armed conflict has a negative and significant effect at 10% significance level on GDP per capita growth, however, it does not appear to have any significant effect on GDP growth. This may be because armed conflict was not a significant phenomenon in the 2000s. The major conflicts which took place in the 2000s were some kind of unusual conflict such as group attacks on a particular country (e.g., Iraq, Afghanistan). Iraq is not included in our analysis due to non13 reliability of data. Although, Afghanistan, USA and UK are included into our sample, armed conflict has been a common phenomenon for these countries over a long-time. Moreover, in a 126 dataset, may be these three countries could not play any significant role to alter the overall results. The R-square and F-statistic values and number of observations are also presented in Table 3 which indicates overall fitness of the empirical models. 4.2 Panel Study Unit root tests are carried out on the variables in order to determine their level of integration. Panel unit root tests of Levin-Lin-Chu were used and the results are reported in Table 4. The results indicate that all the variables are stationary, I(0) on level. Consequently, model in equation (3) is estimated in panel using the pooled OLS and the fixed effects estimators. Estimated results are given in Table 5 and Table 6, respectively. <Table 5> <Table 6> The results from the pooled OLS show that overall, all traditional variables - first lag of GDP, labour, capital and human capital are positive and significant contributors to GDP growth in both empirical models (real GDP per capita and real GDP). Sunspot variable, corruption plays a negative and significant role in economic growth. In full-sample, the size of coefficient of NPI is the highest among all variables in both models (GDP per capita and GDP). This is consistent with the results reported from cross-section approach. The polity variable is found to be negative and significant in high- and low-income countries12. Overall, the effect of polity on growth is insignificant. It also found insignificant in middle income countries. This is not surprising as Bhutan (-10), Qatar (-10), Saudi Arabia (-10), Turkmenistan (-9), Uzbekistan (-9), Swaziland (-9), UAE (-8), China (-7), Vietnam (-7), Laos (-7) have very low polity scores. However, their average growth for the sample period is generally high; Bhutan (8.5%), Qatar (12.5%), Saudi Arabia (3.4%), Turkmenistan (14.2%), Uzbekistan (6.5%), Swaziland (3.1%), UAE (5.8%), China (10.3%), Vietnam (7.3%), Laos (6.8%) indicate that polity score does not positively affect economic growth. In reality, the 12 There are 32 countries in the high-income group and 17 countries in low-income group and middle-income countries group consists of 57 countries. 14 average growth rate of these countries for the sample period is 7.8%, which is a way higher than the world growth rate average of 4.5%. The results also indicate that armed conflict has a negative impact on GDP growth in low income countries. It is insignificant in the full-sample estimation as well as in the sub-sample of middle-income countries. The results are similar for high-income countries estimates that used GDP per capita as a dependent variable. Unexpectedly, conflict is found to have positive effect to GDP for the high-income countries. However, before we make a comment on this issue, we need to double check the effect of conflict on high income countries economic growth. Results from the fixed effects as reported in Table 6, which indicates that lag of GDP, labour, capital and human capital are positive and significant determinants of economic growth. The first lag of GDP can be termed as initial endowment for current year’s GDP growth. Human capital (HK) is found negative and significant for low income countries. We, then, have tested the effect of its (HK) lags on economic growth. The study finds that a one-year lag of human capital has an insignificant effect (-.018(.015)) on economic growth, however, two-year lag has appeared to be positive and significant (0.023***(0.011)) in our growth regression. This indicates that if low income countries invest in human capital, it becomes an investment (cost) for current year, however, this investment starts giving returns from year two. Institutional determinants such as corruption are found to be negative and significant in fullsample and in most of the sub-samples. Overall, polity and armed conflicts are found to be insignificant variables in the fixed effects models. Armed conflict is not found positive in high income countries data when we have applied the fixed effects estimator (see, Table 6). Number of observations in cross-section and panel are 126 and 1057 respectively. 4.3 Important Differences across the Income Groups Labour force is significant at all conventional levels for the model that uses the real GDP for all the groups. However, it has the lowest coefficient for the low-income countries. The coefficient for the labour force in the non-oil producing countries is the highest as reported in Table 3. This may be a reflection of non-existence of ‘resource curse’ as suggested by the literature on the natural resources abundance that such resources tend to have negative effect on other sectors of the countries’ economies. This is particularly, on their labour productivity. 15 Similar pattern of results has been reported in Table 5 based on pooled OLS. The only difference is that the model that uses the real GDP per capita reports higher value of coefficient than the one that uses the real GDP. Capital is significant at any conventional level for all the groups, except the middle-income countries where it is insignificant. Results for the former are consistent with general findings of the literature on growth. However, insignificance of capital to middle-income countries could be attributed to trade union activities, which may undermine productivity of capital. Human capital is insignificant in the high-income group, but is highly significant and positive in both middle- and low-income countries. The role of democracy to economic performance appears to be negative and significant to high- and low-income countries, but positive and insignificant in the middle-income countries. The former is consistent with the literature (see, for example, Chan (2002), Dornbusch and Edwards (1991), Kohli (2004), Leftwitch (2005) among others) that argues that democratic countries find it difficult to implement painful economic programmes that are beneficial to economic growth. Sometimes, even if these programmes are necessary. This is more evident in the current experience, particularly, of the Euro-Zone countries in implementing reforms after the sovereign debt crisis. This is in contrast to what was later known as the “Chilean miracles”, which referred to economic results obtained as a consequent of reforms imposed by the former dictator, Pinochet. Secondly, most of the best performing economies in the world are nondemocratic ones. For example, in its 2015 annual report, the World Economic Forum declares Qatar, an autocratic country, as the most efficient economy in the world and followed by Singapore. The results for the middle-income group suggest that for democracy to be beneficial to economic growth, a country must achieve a certain threshold of income that is sufficient enough to strengthen democratic institutions, but not too strong to hinder overall economic performance. It is also noteworthy that the results reported in Table 5 indicate that the role of corruption to economic growth differs across the groups. It is found to hamper growth in high- and middleincome group countries. However, its role in low-income countries is insignificant. The former is consistent with literature on corruption that suggests that it is ‘sand in the wheel’ while the latter is in consonant with those who argue that corruption could act as ‘grease in the wheel of the economy as it can promote productive activities. It is argued that without corruption such businesses would not have developed. But what the results indicate that level of national income important in these different roles of corruption. 16 5. Conclusion This paper investigates the effects of the institutional variables on economic growth. It is motivated by an augmented Cobb-Douglas production function. We use ten years (2000-2009) mean values of variables and construct cross-section dataset for 126 countries. Subsequently, we estimate growth model by using both full-sample and sub-samples data. The countries are classified into high-, middle- and low-income using the World Bank criteria. The study used panel of 106 countries with 1060 observations. Similar to cross-sectional study, we use both full-sample and sub-samples to estimate economic growth models using Pooled OLS and the fixed effects approach. The results from both cross-section and panel data analysis indicate that traditional variables, which are first lag of GDP (only in panel), labour, physical capital and human capital are found to be significant determinants of economic growth. These variables foster economic growth. On the contrary, corruption plays a significant negative role in fostering economic growth. The size effect of corruption is the maximum among all the traditional and institutional variables. It can be documented that most of the contributions of the traditional variables to economic growth are neutralised by corruption alone. The coefficient of polity (democracy) variable in most of the cases is insignificant which indicates that polity cannot influence economic growth. However, in few cases we found that polity plays a negative role in economic growth. One generally does not expect a negative effect of polity. However, this result is in line with the existing empirical literature (see, Bruneti, 1997). Armed conflicts do not appear to have any statistically significant effect on economic growth except in low-income countries. The results also demonstrate that influence of some certain institutional variables vary according to the income level of the countries. For example, corruption is found to have significant negative effects in high- and middle-income countries. However, it is found to be positive, although insignificant in low-income countries. This difference, therefore, sheds light on the argument that corruption could be favourable to growth. The results show that corruption could only have positive impact in poor countries with weak institutions where absence of corruption could be a barrier to investments that businesses may find not profitable. Similarly, democracy is found to be negative in high- and low-income countries, but positive in middleincome countries. This shows that for democracy to be beneficial to growth, income level of the country should be between low and high thresholds. The critical question that is a future research is to determine what these thresholds are. 17 In conclusion, our estimated results (in general) indicate that lag income, labour, physical capital and human capital are positive and statistically significant determinants of economic on growth. However, institutional variables, particularly, corruption has a significant negative effect on economic growth. Others have ambiguous effect on economic growth. Armed conflict and polity are either statistically insignificant determinants or negative determinants of economic growth. The empirical results of this study are consistent with theoretical forecasting of the paper as well as with existing literature. References Acemoglu, D., Johnson, S. and Robinson, J. A., (2005) Institutions as a Fundamental Cause of Longrun Growth, Handbook of Economic Growth, Volume IA. Acemoglu, D., Johnson, S., and Robinson, J.A., (2001) The Colonial Origins of Comparative Development: An Empirical Investigation, American Economic Review, 91: 1369-1401. Aghion, Philippe & Howitt, Peter, (1992) A Model of Growth through Creative Destruction, Econometrica, 60(2): 323-351. Arrow, Kenneth (1962) The Economic Implications of Learning by Doing, The Review of Economic Studies, 29 (3): 155–173. Botero Juan, Simeon Djankov, Rafael La Porta, Florencio Lopez-de-Silanes, Andrei Schleifer, (2004) The Regulation of Labour, The Quarterly Journal of Economics, 119(4). Brunetti, Aymo (1997) Political Variables in Cross-country Growth Analysis, Journal of Economic Surveys, 11(2): 163-190. Brunetti. A. and Weder. B. (1998). Investment and institutional uncertainty: A comparative study of different uncertainty measures. Wclrwirtschaftliches Archiv 134: 513-533 Cass, David and Shell, Karl (1983) Do Sunspots Matter Center Analytic Res. Econ. Soc. Sci. (CARESS) Working Paper no. 81-09R, Philadelphia: Univ. Pennsylvania, May 1981. Cass, David and Shell, Karl (1983) Do Sunspots Matter? Journal of Political Economy, 91(2): 193-227. Coe, D., and E. Helpman, (1995) International R&D Spillovers, European Economic Review, 39: 859–887. Deger, S., Lam, C. and Sen, S. (2011) Institutions and Economic Growth – Revisiting the Empirical Evidence from A Development Perspective, Paper prepared for the Annual Meeting of the Public Choice Society, Miami, Florida, March 2012. 18 Denison, E. F. (1980) “The Contribution of Capital to Economic Growth”, The American Economic Review, 70(2): 220-224. Federici and Marconi (2002) On Exports and Economic Growth: the Case of Italy, Journal of International Trade and Economic Development, 11(3): 323-340. Grossman G.M. and Helpman, E (1991) Quality Ladders in the Theory of Growth, Review of Economic Studies, 58: 43-61. Kaldor, N. (1970) The Case for Regional Policies, Scottish Journal of Political Economy, 17: 337-348. Koubi, Vally (2005), War and Economic Performance, Journal of Peace Research,42(1):67– 82. La Porta, R., Lopez-de-Silanes, F., Shleifer, A., Vishny, R., (1997) Legal determinants of external finance, Journal of Finance, 52: 1131-1150. La Porta, R., Lopez-de-Silanes, F., Shleifer, A., Vishny, R., (1998) Law and Finance. Journal of Political Economy, 106: 1113-1155. Lucas, Robert E. (1988) On the Mechanics of Economic Development, Journal of Monetary Economics, 22: 3-42. Mankiw, N.G., Phelps, E.S. and Romer, P.M. (1995), The Growth of Nations, Brookings Papers on Economic Activity, Vol. 1995, No. 1: 275-326. Mankiw, N.G., Romer, D. and Weil, D.N. (1992), A Contribution to the Empirics of Economic Growth, The Quarterly Journal of Economics, 107(2): 407-437. North, Douglass C., and Robert P. Thomas, (1973) The Rise of the Western World: A New Economic History, Cambridge: Cambridge University Press. Pesaran, M.H., Shin, Y., Smith, R.P. (1999) Pooled mean group estimation of dynamic heterogeneous panels. Journal of the American Statistical Association, 94, 621-634. Romer, Paul M. (1986) Increasing Returns and Long-Run Growth, Journal of Political Economy, 94 (5): 1002-1037. Romer, Paul M. (1993), Idea gaps and object gaps in economic development, Journal of Monetary Economics, 32(3): 543-573. Solow, Robert M. (1969) Growth Theory: an Exposition, Oxford University Press, New York. The World Economic Forum (2015) Global Competitiveness Report, WEF available: http://reports.weforum.org/global-competitiveness-report-2014-2015/. 19 Figure 1a Polity and Real GDP (PPP) 11 Full Sample (2000-20009) QAT ARE KWT NOR USA IRL CHE NLD DNK SWE GBR AUS FIN FRA JPN ITA ESP NZL GRC DEU ISR SVN KOR PRT TTO SVK EST HUN HRV POL LTU MEX LVA CHL RUS MYS ARG TUR BWA MUS VEN PAN URY ROM BRA ZAF CRI COL JAM DOM PER THA ECU SLV ALB UKR NAM GTM PRY BOL ARM LKA GEO HND IDN PHL MNG GUY MDANIC IND PNG SEN BEN KEN GHA LSO BGD ZMB SGP 10 BHR OMN SAU LBY GAB IRN 9 KAZ DZA TUN SWZ AZE CHN SYR EGY JOR FJI 8 MAR BTN COG LAO CMR SDN MRT 7 GMB AFG YEM PAK KGZCIV TJK TCD TZA UGA TGO RWA KHM NPL MDG CAF ETH MLI MWI MOZ NER SLE 6 LBRBDI ZAR 0 5 10 Avergate Polity Figure 2b Polity and Real GDP per Capita 20 15 20 Full Sample (2000-2009) 30 USA CHN 28 RUS TUR IDN IRN SAU 26 EGY KWT LBY SYR QAT 24 OMN THA PAK SDN COL SWE CHE GRC PHL UKR PRT NOR ROM PER CHL DNK ISR HUN FIN IRL NZL SVK ECU LKA DOM HRV GTM ETH KEN SVN LTU CRI BOL SLV URY CIV LVA PAN TTO GHA PRY NPL HND BWA EST KHM ALB JAM SEN ZAR MDG MOZ ZMBGEO MUS NIC PNG ARM MLI NAM BEN KGZ NER MWI MDA MNG TUN YEM TZA CMR UGA JOR GAB AFG BHR COG TCD TJK RWA LAO AUS NLD POL BGD KAZ MAR AZE ARG ZAF MYS VEN DZA SGP ARE JPN IND DEU FRA GBR ITA BRA MEX ESP KOR MRT TGO 22 SWZ FJI CAF BTN GMB BDI SLE GUY LSO 20 LBR 0 5 10 15 20 Average Polity The vertical axis shows both logarithm of real GDP (PPP) per capita and logarithm of real GDP (PPP) against average Polity for full sample. The mean lines of GDP and polity are also shown to indicate which country is above the mean and which country is below it. Country codes are used following World Bank coding for countries. We calculated 10 years’ average value (2000-2009) for GDP and polity score of each country. We draw the trend line which shows an upward trend in case of real GDP (PPP) as well as real GDP (PPP) per capita. Figure 2a Non-corruption Indicator and Real GDP (PPP) Full Sample (2000-2009) 11 QAT ARE GRC SAUTTO SVK HRV POL LBY GAB RUS ARG MEX TUR LVA VEN PAN ROM IRN BRA KAZ JAMCOL DZA DOM THA PER ECU ALB SLV UKR AZE EGY SYRSWZCHN FJI GTM PRY BOL ARM GEO LKA MAR COG IDNHND PHLMNG GUY NIC MDA IND YEM PAK CMR PNG LAO MRT SEN CIV SDN KGZ KHM TJK BEN KEN LSO GHA BGD GMB ZMB TCD TZA NPL UGA MLI TGOMDG RWA AFG CAF MOZ ETH MWI SLE NER USA IRL FRAJPN ESP BHR DEU ISR OMN SVN PRT ITA KOR HUN LTU MYS MUS CRI ZAF NOR SGP CHE NLD DNK GBR AUS SWE FIN NZL EST BWA URY CHL TUN NAM JOR BTN 6 7 8 9 10 KWT BDI LBR ZAR 2 4 6 8 Average Non-corruption Figure 2b Non-corruption Indicator and Real GDP (PPP) Per Capita 21 10 Full Sample (2000-2009) 30 USA CHN JPN DEU IND 28 RUS ESP TUR IDN IRN AUS NLD 24 26 POL ARG THA SAU ZAF EGY PAK COL MYS GRC VEN PHL DZA UKR PRT ROM ARE PER BGD ISR HUN KAZ MAR KWT SVK ECU LBY SYR LKA HRV SDN TUN DOM QAT GTM OMN SVN ETH KEN LTU AZE YEM TZA CRI BOL CMR URY PAN SLV LVA CIV GHA TTO NPL PRYUGA JOR HND AFG KHM ALB BHR BWA EST GAB JAM ZAR MDG GEOSEN MOZ ZMB MUS NICARM COG PNG MLI NAM TCD TJKLAO BEN KGZ NER MWI MDA RWA MNG MRT SWZ TGO FJI SLE CAF BDI LSO BTN GUY GMB LBR CHL CHE SWE NOR SGP DNK FIN IRL NZL 20 22 GBR FRA ITA KOR BRA MEX 2 4 6 8 10 Average Non-corruption Figure 3: Real GDP per capita of Countries Encaged in Conflicts Full Sample (2000-2009) 11 QAT DNK FIN FRA DEU GRC AUS 10 BHR 9 ARG HRV EST GAB DZA AZE ISR LBY LTU LVA MYS MEX MUS 8 7 6 AFG CAF RUS GHA GMB PER KEN TUN THA UKR PRY MAR PHL MNG MDA NICPAK PNG MRT LSO MDG MLI NPL MWI MOZ ETH TURURY VEN ZAF NAM EGY TCD SGP PAN ROM JAM JOR FJI GTM BOL CHN ARMBTN GEO COG HND IDN GUY IND CMR KGZ LAO CIV BEN KHM BGD ARE USA CHE SWE GBR ESP NZL KOR OMN PRTSAUSVN TTO SVK POL IRN KAZ COLDOM ECU SLV ALB KWT IRL JPN ITA HUN BWA CHL BRA CRI NOR NLD NER SWZ SYR LKA YEM SEN SDN TJK RWA SLE TZA TGOUGA ZMB LBR BDI ZAR 0 50 100 Country Mean Log GDP (PPP) Per Capita = 8.596 Non-conflict Figure 3b 22 conflict 150 Real GDP of Countries Encaged in Conflicts 30 Full Sample (2000-2009) USA 28 CHN JPN DEU FRA BRA IND 26 DZA BGD COL EGY CHL NLD MYS GRC DNK FIN RUS MEX IDN IRN AUS ARG ITA HUN ISRKAZ IRL KWT LBY GTM KEN LTU POL SAU GBR ESP KOR TUR THA ZAF PAK PHL PRT ROMSGP PER SWE CHE NOR UKRVEN ARE 22 24 MARNZL SVK ECU SYR TUN LKA HRV SDN DOM OMN QAT SVN ETH YEM TZA AZEBOLCMRCRI SLV URY CIV LVA GHA TTO UGA NPL PAN JOR PRY AFG ESTGAB HND JAM KHM ALB BHRBWA SEN MDG ZAR GEO ZMB MUSMOZNIC PNG ARMBEN TCDCOG MLI NAM LAO MWI KGZ TJK MDA RWA MNG NER MRT SWZTGO FJI SLE CAF LSO BTNBDI GMBGUY 20 LBR 0 50 100 150 Country Mean log GDP (PPP) = 24.875 Non-conflict Conflict Data Source: UCDP Battle-related death estimates (Version 5.0). Note: Countries which were directly involved in armed conflict was counted as conflicting country. Those countries which supported a third country or group however were not directly involved in combat and those countries which were not involved in any combat, were included in Non-conflicting country. Sample Table 1 Descriptive Statistics (2000-2009) – Cross-sectional Data Mean S.D. Min Max Full-sample (N =126) GDP (PPP) Per Capita 10727.04 11866.61 275.24 46432.2 GDP (PPP) (million) 418000 1310000 1270 12300000 Growth Rate 4.41 2.26 .58 14.91 Labour Force (million) 21.60 78.70 0.25 758 Capital 21.62 5.40 8.86 51.87 Human Capital 21.68 15.70 0.67 74.10 Polity 14.10 5.89 1 20 Corruption 3.99 2.07 1.70 9.52 Conflict (Battle Death) 308.34 994.54 0 8202 High income countries (N=38) 23 GDP (PPP) Per Capita 26563.36 10325.79 10181.49 46432.2 GDP (PPP) (million) 919000 2140000 19400 12300000 Growth Rate 3.24 1.74 0.58 7.01 Labour Force (million) 15.60 29.60 0.38 154 Capital 21.55 3.73 12.29 29.05 Human Capital 31.48 12.35 11.17 57.53 Polity 17.29 5.77 1 20 Corruption 6.53 2.0 2.38 9.52 Conflict (Battle Death) 294.25 1377.35 0 8202 GDP (PPP) Per Capita 5965.87 3588.98 1234.84 15999.29 GDP (PPP) (million) 315000 837000 1930 5770000 Growth Rate 4.59 2.24 0.71 14.91 Labour Force(million) 31.60 111.00 0.30 758.00 Capital 21.79 4.72 10.42 39.48 Human Capital 23.55 15.39 0.67 74.10 Polity 13.69 5.69 1 20 Corruption 3.17 0.92 2.04 5.84 Conflict (Battle Death) 194.70 604.03 0 3298.36 Middle income countries (N =60) Least Developed Countries (N=30) GDP(PPP) Per Capita 1245.8 928.28 275.24 4729.59 GDP (PPP) (million) 21200 31900 1270 168000 Growth Rate 5.48 2.29 1.12 10.66 Labour Force (million) 8.77 12.80 0.25 64.40 Capital 21.36 7.97 8.86 51.86 Human Capital 6.16 4.53 0.97 17.50 Polity 11.09 4.57 3 17.64 Corruption 2.59 0.69 1.70 5.38 Conflict (BD) 552.53 1069.05 0 4650 Note: Human capital is proxied by labour force with secondary education (% of population aged 25 and over); Polity is polity score; corruption is non-corruption score and conflict is the number of battle death. We have used World Bank (2009) information which has classified countries as per their Gross National Income Per capita (in dollar value). There are countries which moved from one group to another group due to the change in their income level. We have used average (2000-2009) income level. Hence, for example, there are some countries which are 24 shown in the middle income group; however, the average income level of those countries is higher than the lowest income level in high income group. Table 2 Correlation Matrix (Panel Data) ln(YPC) ln(Y) ln(LF) ln(K) ln(HK) ln(P) ln(YPC) 1.00 ln(Y) 0.66 1.00 ln(LF) 0.05 0.78 1.00 ln(K) 0.25 0.18 0.03 1.00 ln(HK) 0.61 0.32 -0.09 0.21 1.00 ln(P) 0.29 0.23 0.07 -0.02 0.23 1.00 ln(NPI) 0.80 0.46 -0.04 0.18 0.42 0.30 25 ln(NPI) 1.00 Table 3 Cross-sectional regressions using mean values (2000-09) of variables Variables GDP (PPP) per capita Full-Sample GDP (PPP) High- Middle- Low- Income Income Income 4.88*** 8.460*** 7.862*** 6.542*** 4.907*** (1.031) (.960) (1.186) (1.284) (1.603) (1.063) .078 .100** 1.103*** 1.076*** 1.043*** .879*** 1.088*** (.054) (.079) (.048) (.044) (.039) (.054) (.092) (.050) .393 .062 .923*** .129 .067 .115 .184 1.139*** .065 (.248) (.347) (.360) (.268) (.255) (.248) (.324) (.357) (.313) (.263) .446*** .017 .144 .074 .415*** .417*** .023 .100 .099 .385*** (.068) (.121) (.091) (.098) (.070) (.068) (.112) (.090) (.114) (.072) -.170* -.307*** -.036 -.195 -.078 -.194* -.252*** -.113 -.173 -.125 (.098) (.092) (.116) (.167) (.108) (.098) (.085) (.116) (.194) (.111) 1.770*** .877*** 1.145*** .0001 1.650*** 1.685*** .778*** 1.285*** .185 1.678*** (.158) (.164) (.291) (.450) (.239) (.158) (.153) (.289) (.525) (.247) -.289* .105 -.159 -.053 -.254 -.245 .039 -.131 .172 -.209 (.152) (.199) (.188) (.169) (.163) (.153) (.185) (.187) (.198) (.168) Adj. R2 0.74 0.56 0.32 0.51 0.63 0.88 0.97 0.89 0.85 0.86 FSTAT 58.99*** 6.17*** 4.14*** 6.07*** 27.09*** 154.35*** 175.12*** 85.47*** 29.39*** 91.90*** Countries 126 36 60 30 93 126 36 60 30 93 C lnL lnK lnHK lnP lnNPI Conflict High- Middle- Low- Non-Oil Income Income Income 3.486*** 6.930*** 5.881*** 5.659*** 3.518*** (.959) (1.270) (1.295) (1.374) .116** .073* .056 (.044) (.042) .112 Note: ***p<0.01; **p<0.05 and *p<0.10; standard errors are in parentheses. 26 Full-Sample Non-Oil 27 Table 4 Unit root tests (Levin-Lin-Chu test) Series 1st Difference Level Intercept Intercept & Trend Intercept Intercept & Trend ln(YPC) -5.54** -3.36** -5.14** -7.22** ln(Y) -3.97** -5.15** -5.33** -6.51** ln(LF) 0.95 -28.08** -17.78** -23.76** ln(K) -8.06** -8.01** -14.87** -20.84** ln(HK) -2.36** -4.75** -13.35** -23.35** ln(P) -14.00** -11.82** -11.29** -16.75** ln(NPI) -14.30** -20.95** -25.15** -26.52** **p<0.01; *p<0.05. Null hypothesis: unit root Table 5 Regression Results (Pooled OLS) Variable C GDP (PPP) Per Capita Full- High- Middle- Low- Full- High- Middle- Low - Sample Income Income Income Sample Income Income Income .619*** 1.887** .655*** 1.523** 1.819** 6.147** 4.403** 4.741** (.214) * (.258) * * * * * (.354) (.273) (.426) (.395) (.539) (.373) lnY.L1 lnL lnK lnHK lnP GDP (PPP) .605*** .645*** .812*** .758*** .367*** .118*** .336*** .332*** (.016) (.033) (.022) (.036) (.016) (.020) (.022) (.038) .039*** .042*** .018** .028 .721*** .988*** .733*** .667*** (.009) (.008) (.008) (.018) (.019) (.023) (.025) (.041) .216*** .186*** .042 .269*** .223*** .212*** .166 .290*** (.048) (.063) (.052) (.046) (.059) (.083) (.182) (.073) .197*** .028 .058*** .040** .270*** .087*** .110*** .061** (.015) (.020) (.018) (.015) (.018) (.026) (.026) (.024) .019 - .017 -.088** .004 -.256*** .016 - (.023) .146*** (.020) (.039) (.028) (.033) (.031) .190*** (.027) lnNPI (.061) .740*** .425*** .268*** -.022 1.058** .808*** .763*** -.055 (.041) (.043) (.053) (.069) * (.044) (.079) (.108) (.045) 28 Conflict -.043 .047 -.067 - -.017 .277*** -.008 - (.029) (.058) (.086) .134*** (.035) (.076) (.048) .276*** (.039) (.060) Adj. R2 0.89 0.83 0.79 0.82 0.93 0.97 0.91 0.92 FSTAT 1197.95 224.27* 306.02* 111.00* 1989.34 1701.34 1001.56 291.03* *** ** ** ** *** *** *** ** 1058 317 568 169 1058 317 568 169 106 32 57 17 106 32 57 17 Observat ion Countrie s Note: ***p<0.01; **p<0.05 and *p<0.10; standard errors are in parentheses. 29 Table 6 Regression Results (Fixed Effects) Variable GDP (PPP) Per Capita Full-Sample High-Income Middle- GDP (PPP) Low-Income Full-Sample High-Income Income C lnY.L1 lnL lnK lnHK lnP lnNPI Conflict Overall R2 Middle- Low-Income Income -.197 7.387*** -4.584*** -10.378*** 4.705*** 12.237*** 1.548*** 5.695*** (.667) (1.498) (.915) (1.655) (.658) (1.368) (.955) (.730) .036*** .075*** .047*** .123*** .015*** .006 .010** .050*** (.005) (.023) (.010) (.025) (.003) (.007) (.004) (.017) .458*** .013 .714*** 1.054*** 1.197*** .809*** 1.386*** 2.029*** (.046) (.069) (.063) (.116) (.045) (.063) (.065) (.111) .197*** .300*** .313*** .038 .194*** .285*** .318*** .088*** (.018) (.054) (.024) (.028) (.017) (.050) (.024) (.033) .265*** .373*** .196*** -.138* .246*** .331*** .198*** -.159 (.034) (.080) (.041) (.074)a (.034) (.072) (.042) (.072) -.020 -.405 -.052** .011 -.021 -.429 -.064** -.008 (.019) (.369) (.025) (.023) (.019) (.336) (.025) (.034) .073*** .403*** .050 .098* .072*** .353*** .049 .069* (.025) (.077) (.034) (.050) (.024) (.071) (.036) (.037) -.044 -.040 -.047 -.015 -.026 -.036 -.035 -.002 (.036) (.091) (.034) (.020) (.035) (.082) (.036) (.020) 0.15 0.30 0.12 0.12 0.71 0.93 0.86 0.81 30 FSTAT 99.61*** 18.29*** 106.40*** 28.80*** 272.29*** 48.88*** 196.32*** 111.40*** Observation 1057 317 568 168a 1057 317 568 169 Countries 106 32 57 17 106 32 57 17 Note: ***p<0.01; **p<0.05 and *p<0.10; standard errors are in parentheses. aThis is the effect of current human capital on current economic growth. However, one-year lag shows an insignificant effect (-.018(.015)) and two-year lag has appeared to be positive and significant (0.023***( 0.011)) in our growth regression. 31 APPENDIX A1. Countries in Cross-section Afghanistan Greece Oman Albania Guatemala Pakistan Algeria Guyana Panama Argentina Honduras Papua New Guinea Armenia Hungary Paraguay Azerbaijan India Peru Australia Indonesia Philippines Bahrain Iran Poland Bangladesh Ireland Portugal Benin Israel Romania Bhutan Italy Rwanda Bolivia Jamaica Russia Botswana Japan Senegal Brazil Jordan Sierra Leone Burundi Kazakhstan Singapore Cambodia Kenya Slovak Republic Cameroon Kuwait Slovenia Central African Republic Kyrgyz Republic South Africa Chad Laos South Korea Chile Latvia Spain China Liberia Sri Lanka Colombia Lesotho Sudan Costa Rica Libya Swaziland Cote d’Ivoire Lithuania Sweden DR Congo (Zaire) Madagascar Switzerland Congo Malawi Syria Croatia Malaysia Tajikistan Denmark Mali Tanzania Dominican Republic Mauritania Togo Egypt Mauritius Thailand El Salvador Mexico Trinidad and Tobago Ecuador Moldova Turkey Estonia Mongolia Tunisia 32 Ethiopia Morocco Uganda Fiji Mozambique Ukraine Finland Namibia UAE France Nepal UK Gabon Netherlands USA Gambia New Zealand Uruguay Georgia Nicaragua Venezuela Germany Niger Yemen Ghana Norway Zambia Albania Guyana Paraguay Algeria Honduras Peru Argentina Hungary Philippines Armenia India Poland Australia Indonesia Portugal Bangladesh Ireland Romania Benin Israel Russia Bolivia Italy Rwanda Botswana Japan Senegal Brazil Jordan Sierra Leone Burundi Kazakhstan Singapore Cambodia Kenya Slovenia Cameroon Kyrgyz Republic South Africa Central African Republic Laos South Korea Chile Latvia Spain China Lesotho Sri Lanka Colombia Lithuania Sudan Costa Rica Malawi Swaziland Cote d’Ivoire Malaysia Sweden Total: 126 countries A2 Countries in Panel 33 DR Congo (Zaire) Mauritania Switzerland Congo Mauritius Syria Croatia Mexico Tajikistan Denmark Moldova Tanzania Dominican Republic Mongolia Thailand Egypt Morocco Turkey El Salvador Mozambique Tunisia Ecuador Namibia Uganda Estonia Nepal Ukraine Finland Netherlands United Arab Emirate France New Zealand United Kingdom Gabon Nicaragua Uruguay Gambia Niger United States of America Germany Norway Venezuela Ghana Pakistan Zambia Greece Panama Guatemala Papua New Guinea Total: 106 Countries 34