EFFECTS OF PROPERTY RIGHTS AND CORRUPTION ON GENDER DEVELOPMENT

EFFECTS OF PROPERTY RIGHTS AND CORRUPTION ON GENDER DEVELOPMENT A Thesis submitted to the Graduate School of Arts and Sciences at Georgetown University in partial fulfillment of the requirements for the degree of Master of Public Policy in the Georgetown Public Policy Institute By Afroza Rahman Chowdhury, B.A Washington D.C, March 18, 2008

EFFECTS OF PROPERTY RIGHTS AND CORRUPTION ON GENDER DEVELOPMENT Afroza Rahman Chowdhury, B.A Thesis Advisor: Marcela Tarazona, Ph.D ABSTRACT This study explores the effects of property rights and control of corruption on women s socioeconomic well-being, measured by the Gender Development Index (GDI), using a macroeconomic cross-country regression analysis with available data from 2004. It finds that property rights have positive effects on GDI but control of corruption does not seem to have a consistent relationship with it. Property rights influence on GDI, however, erodes when we take into account other institutional factors such as rule of law and governance. This study concludes that political and economic institutions are important for women s socioeconomic well-being and property rights is one such institution. The absence of a strong relationship between control of corruption and gender development does not necessarily mean that there is no such link. Establishing anti corruption mechanisms and good governance practices takes time to bear fruit and effects of such efforts cannot be effectively gauged in the short run. Further research is needed to see whether its impact is substantial over a long period of time. ii

Table of Contents Abstract ii Table of Tables and Graphs... v A. Introduction 1 B. Literature Review.. 3 B.1. Findings of Previous Researchers.. 3 B.2. The Known and the Unknown... 6 C. Conceptual Model. 7 C.1. Variables of Interest 9 D. Data and Methods 12 E. Analysis Plan 13 F. Descriptive Results.. 15 F.1. Summary Statistics.... 15 F.2. Scatter Plots... 17 F.3. Mean GDI across different levels of Property Rights... 18 F.4. Mean GDI across Different Levels of Control of Corruption... 19 F.5. Mean GDI across Different Economic Structures. 20 F.6. Drawbacks of Data.... 21 F.7. Hypothesis on Property Rights and Control of Corruption... 21 G. Regression Results and Discussion...22 G.1. Regression Results 23 iii

G.2. Effects of Urbanization, Ethnic Fragmentation, being Landlocked & type of Sector 25 G.3. The role of Governance and Rule of Law. 27 G.4. What role does the structure of an economy play in influencing GDI? 29 H. Conclusion... 33 I. Appendix.. 35 Table 9: Correlation Coefficients. 35 Table 10: With Ratio of Female to Male Earned Income 35 Table 11: VIF values for regression 4 from Table 2: Regressions against GDI.. 36 J. Bibliography.. 37 iv

Table of Tables and Graphs 1. Graph 1: Graphical Representation of Conceptual Model...8 2. Table 1: Descriptive Statistics of All Variables 15 3. Graph 2: Property Rights Index (Gwartney Lawson) vs. Gender Development Index. 17 4. Graph 3: World Bank index of Control of Corruption vs. Gender Development Index 18 5. Table 2: Mean GDI across different levels of Property Rights.....19 6. Table 3: Mean GDI across different levels of Control of Corruption...19 7. Table 4: Mean GDI across Agricultural, Industrial and Service Oriented Economies... 20 8. Table 5: Regressions against GDI. 24 9. Table 6: Regression without Rule of Law... 28 10. Table 7: Regressions with structure of economy 30 11. Table 8: Regressions on countries with > 15% Agriculture as % of GDP, > 30% Industry as % of GDP and > 54% Service as % of GDP.. 32 12. Table 9: Correlation Coefficients... 35 13. Table 10: With Ratio of Female to Male Earned Income.. 35 14. Table 11: VIF values for regression 4 from Table 2: Regressions against GDI 36 v

A. Introduction In mainly agrarian economies, land is a key source of income and credit. Women can use their land for agricultural produce and earn income from their sale of crops, leave their land to work for wage income, earn income from land rent, or can use their land as collateral for obtaining credit. Bina Agarwal (1994) has done thorough research on the favorable effects of property rights on women s income and concludes that access to land is more crucial than ownership of land to ensure women s economic and social security. Research by the International Food Research Institute (1997), on the other hand, shows that unless women s differential access to education, credit and labor markets are also addressed, access to land will not increase their labor productivity or income. Corruption, on the other hand, diverts national and private funds intended for social and economic projects to increase the welfare of people. Does such diversion however, significantly and adversely affect women s welfare measured by the Gender Development Index? Research shows that corruption decreases national budgets. Where there are not enough resources, the government is less likely to spend on health and social security, which affects women and children more. Therefore apart from the fact that the allocation for programs focusing on women is only a fraction of the total national budget, a cut in public spending caused by corruption means that maternal and child health services are likely to be affected 1. On the other hand, can we be sure that corruption has 1 Idaishe Chengu, and Maud Mukwamba. "Women and the Anti Corruption Debate." Women's International League for Peace and Freedom. 18 Sept. 2007 <http://www.peacewomen.org/news/ Zimbabwe/April04/corruption.html>. 1

a significant impact on women s socioeconomic well being when holding constant other crucial factors such as rate of urbanization, level of conflict, national income and trade? This study looks at how property rights and corruption affect women s well being using the Gender Development Index (GDI) as a measure of that well being. GDI is a composite index variable that combines indicators of female life expectancy, educational attainment and income. The Human Development Index is a very similar variable that combines life expectancy, educational attainment and income, for both men and women. These variables are composed by the UNDP and are part of their database of development indicators. To understand how property rights and corruption affect GDI, in essence we are asking how institutional variables affect women s income, female literacy and female life expectancy all components of the GDI. Variables that influence an economy s ability to generate income (e.g., level of urbanization, ability to trade, structure of economy) and distribute it well (e.g., quality of governance, rule of law) should also be taken into account when conducting a study like this. The effects of corruption on women s well being is a new field of research and not much has been done so far. The effects of greater prevalence of women in bureaucracy and parliament on levels of corruption have been studied but the reverse relationship has not been explored. If in this study corruption indeed seems to have a statistically significant negative impact on GDI then a new, but rather lengthy and indirect approach, must be taken to improve women s well being. Just focusing policies on greater integration of women in the labor market and educational spheres might not 2

prove to be sufficient to enhance women s social and economic welfare. There will have to be a greater emphasis on tackling corruption. The goal of this thesis is to answer the question: Do property rights and corruption have an impact on women s well being, which is measured by the Gender Development Index? The following sections include a brief literature review of property rights and corruption, the conceptual framework of my analysis, data and econometric methods used to answer the research question, the regression results and finally a discussion of the findings. B. Literature Review and Policy Relevance This section reviews three econometric research papers that explore the effects of property rights and corruption on HDI. Through their quantitative analysis, the authors find that indeed, property rights have robust effects on human well-being. Corruption is not explored explicitly but is included as an element of political stability, which is regressed against the HDI and the GDI. All three papers conclude that property rights, urbanization and quality of governance are crucial for human development hence they should be considered in efforts to enhance economic development in poor countries. B.1. Findings of Previous Researchers Economic Freedom and Human Progress Nathan J. Ashby (2006) carries out a five year panel data econometric analysis of the effects of economic institutions and free market policies on human well-being. To 3

gauge economic freedom Ashby uses variables from the Economic Freedom of the World Index devised by Gwartney, Lawson and Block (2004), including: government size, property rights, trade, regulation and money. He also utilizes the Index of Human Progress (IHP), created by Emes and Hahn (2001), to measure the quality of life. Alternative variables such as level of urbanization, ethnic fragmentation, landlocked, tropics, governance and rule of law are held constant in his specifications. Ashby s findings highlight the importance of urbanization, as it turns out to have the greatest impact on human progress followed by economic freedom and rule of law. Ashby s theory that the degree of poverty is most serious in rural areas and that inertia and stagnation in these regions retard growth is supported by scholars such as Michael Todaro and his work on Urbanization, Unemployment and Migration in Africa: Theory and Policy (1997). The role of cities as sources of ideas and knowledge that lead to spill over effects of human capital is also emphasized by Kelly and McGreevy (1994). The argument that cities are a key to growth and development, therefore, is empirically supported and strong. In conclusion Ashby states that institutional variables such as stronger property rights and increasing economic freedom have significant positive effects on IHP. The weakest results are the estimates using the quality of government measures. The marginal benefit to increasing government involvement in the development of nations in the form of higher expenditures and tax rates is ambiguous. However, the benefits from establishing sound property rights and maintaining a legal system that protects these rights has a far greater impact than government subsidies and transfers and other policies (p. 20). 4

In Ashby s analysis urbanization is the only geographic variable that provides a significant explanation of the cross-national dispersion of human well-being. If the coefficients are any indicator of the relative importance of these variables, then Urban has the greatest impact (Ashby, 17). Property Rights, Trade and HDI Revisited A study by Seth Norton (1998) looks at the same topic of institutions and their effects on human development. Norton explores the effects of property rights alone on the Human Development Index (HDI) and the Human Poverty Index (HPI). His goal is to explore this relationship, not only in terms of how property rights lead to enhanced wellbeing for all, but in particular, how they address the socioeconomic well being of the poorest people. His study demonstrates that strong property rights can reduce deprivation and have strong influence on the welfare of impoverished people. Norton finds that property policy has robust effects in every regression and that the benefits from establishing sound property rights and maintaining a legal system that protects these rights has a great impact on human development. In general, for the poor countries of the world, about 20 percent of the children under five years of age are undernourished. However, in those poor countries of the world where property rights are weak, about 27 to 29 percent are undernourished (Norton, 243). Political Development, HDI and GDI To supplement this dialogue on economic institutions and human well being an article authored by Ana Margarida Brochado and Francis Vitorino Martins (2005), on the effects of democracy on economic and gender development, explores whether different 5

levels of political development affect HDI and GDI the way that property rights do. Their idea of political development takes into account degree of democracy, government effectiveness, rule of law, and corruption. Countries in the dataset are identified as belonging to one of the following six types of politically developed and developing states: Political Vanguard, Politically Developed, Political Effectiveness, Restricted Governmental Practices, Democratic Development and Democratic Deficit. The authors conclude that levels of political development have statistically significant impacts on variables such as GDP per capita, the Human Development Index and the Gender Development Index. The study therefore confirms that the quality of institutions has a definite and robust impact on the well being of people as measured by HDI and GDI. B.2. The Known & the Unknown Property Rights Based on the results mentioned above, we can conclude that there is a widespread consensus on the positive and significant effects of institutions on HDI. We have not, however, come across similar studies with a more gendered focus. Hence we cannot accurately conclude that the same institutional factors have similar impacts on GDI. Another question of interest is whether the security of property rights has the same level of impact on GDI in agricultural, industrial, and service-oriented economies. This question arises because in mainly agrarian economies land is a key source of income and credit. Secure property rights that guarantee titles and the ability to leverage land enables women to use their land for income (Agarwal, 1994). Industrial and service oriented economies are less agricultural and less land dependent. In these countries 6

investment in education and skills has greater dividends in terms of income and therefore human development and women s well-being might therefore be less dependent on property rights. Corruption Although the article by Brochado and Martins (2005) concludes that enhanced levels of political development (which includes measures for corruption) have favorable impacts on GDI, we have not seen an explicit and specific study of the effects of corruption on GDI. To fill this gap, the present study will explore the effects of property rights and corruption on gender development. C. Conceptual Model A cross-country multivariate analysis is conducted to estimate the effects of property rights and corruption on GDI, holding all other factors constant. The results are included in tables 5 through 8. Before going into the regression analyses in part G, this section will analyze the conceptual framework behind the relationship between Property Rights, Control of Corruption and all other relevant variables, with GDI. Graph 1 is a graphical representation of the conceptual model, where all relevant variables are included with a positive or negative sign to indicate their hypothesized relationship with GDI. Following Graph 1 is a list of those independent variables along with their description and an explanation of how they might influence gender development. 7

Graph 1. Graphical Representation of Conceptual Model Corruption _ Landlock _ Sector + Ethnic _ GDI Property Rights + Governance + Urban + Rule of Law + Since the focus of this thesis is on how the quality of institutions impacts women s well being my main independent variables are: (1) Property Rights, and (2) Control of Corruption. The key control variables are both economic and institutional variables because the strength of the economy has a substantial impact on human as well as gender development. A majority of the independent variables that were used in Ashby s (2006) research have been used in this thesis because they seem to matter for a healthy economy, human development, and hence gender development. Detailed 8

descriptions of how each variable is theoretically linked to GDI are provided in the next section. This thesis also analyzes whether the impact of property rights and corruption on GDI are the same in countries that are predominantly agricultural, industrial, and service oriented. Hence, OLS regressions are run for countries whose Agricultural, Industry and Service share of GDP is more than the average 2 contribution of their respective sectors. C.1. Variables of Interest Dependent Variable: GDI: Gender Development Index - A composite index measuring average achievement in the three basic dimensions captured in the Human Development Index a long and healthy life, knowledge and a decent standard of living adjusted to account for inequalities between men and women. Ranges are from zero to hundred. Data for this variable are taken from the UNDP Development Indicators from 2004. Independent Variables: Property Rights: Measures the extent to which a country s institutions and court systems are able to enforce contracts efficiently and quickly, and are able to punish those who unlawfully confiscate private property. This indicator has a robust positively significant effect on HDI so the hypothesis is that the same result will prevail for GDI. The values of this variable range from zero to 10. The data are from the Gwartney, Lawson and Block Economic Freedom Index dataset for 2004. 2 This refers to the average contribution of each sector of the countries included in the dataset 9

Control of Corruption: This measure ranges from 2.5 to 2.5. This indicator measures the ability of governments to control and mitigate corruption. Higher values mean better ability to control corruption and therefore lower prevalence of corruption. We expect a positive relationship of this variable with GDI. These data are from the World Bank Worldwide Governance Indicators 2004. Urbanization: This is the share of population in a country who reside in an urban setting. The more people who reside in urban areas, the more economic opportunities and access to education and health facilities are available to them. The theoretical relationship of urbanization and GDI is positive. These data are from the World Bank World Development Indicators 2004. Ethnic Fragmentation: Measures the probability that two randomly selected people from a given country will not belong to the same ethnolinguistic group. The greater the degree of ethnic fractionalization, the greater the chances of conflict and impediments in carrying out market exchanges due to language barriers. This variable ranges from 0 to 1 (capturing the probability of fractionalization). This variable has a negative and significant relationship with HDI (where lower probabilities of fractionalization led to higher values of human development); hence we expect the same relationship with GDI. These data are borrowed from the Sachs and Warner 2005 dataset. 10

Landlock: This is a dummy variable where 1 indicates a country is landlocked and 0 indicates that a country is not landlocked and has access to an ocean at its border. These data are from the Environmental Performance Index 2008 3. Sector: This is an ordinal variable that is devised for the purpose of this thesis. The values simply indicate whether a particular country is mainly agricultural, industrial, manufacturing or service oriented: Agricultural = 1; Manufacturing = 2; Industrial = 3; and Service = 4. Countries are categorized ascending in the sector accounting for the largest share of GDP. Agriculture as % of GDP: Share of GDP that is accounted for by agricultural production. These data are from the World Bank Development Indicators. Industry as % of GDP: Share of GDP that is accounted for by industrial output. These data are from the World Bank Development Indicators. Service as % of GDP: Share of GDP that is accounted for by the service sector. These data are from the World Bank Development Indicators. Although Sector already indicates the predominant sector of an economy, Agriculture, Industry and Service as share of GDP provide a greater explanatory power of illustrating the size of different sectors of an economy. Governance: This variable measures the quality of public services, the quality of the civil service and the degree of its independence from political pressures, the quality of policy formulation and implementation, and the credibility of the government s 3 Composed by Yale Center for Environmental Law and Policy (YCELP) and Center for International Earth Science Information Network (CIESIN), Columbia University, with the World Economic Forum, and Joint Research Centre (JRC) of the European Commission http://sedac.ciesin.columbia.edu/es/epi/ 11

commitment to such policies 4. The data are from the World Bank Worldwide Governance Indicators 2004. Rule of Law: This variable measures the extent to which agents have confidence in and abide by the rules of society, in particular the quality of contract enforcement, the police, and the courts, as well as the likelihood of crime and violence 5. The data is from the World Bank Worldwide Governance Indicators 2004. D. Data and Methods In the present study the data for property rights are from the Gwartney Lawson 2007 dataset. This dataset includes data on property rights from 1970 to 2007. It is a cross-country dataset of 128 countries and includes variables such as judicial independence, impartial courts, protection of property rights, military interference in rule of law and political process, and integrity of legal system. The variable of interest, Legal Systems & Property Rights, is an index of all the variables mentioned above. This variable is used for the year 2004. The gender index is from the UNDP development indicators. The datasets are cross-country including data on 187 countries. Other indicators such as Ethnic Fragmentation and Landlocked have been attained from Sachs and Warner dataset and the Environmental Sustainability Index, respectively. 4 Governance Matters 2007: Worldwide Governance Indicators 1996-2006 http://info.worldbank.org/governance/wgi2007/faq.htm#2 5 IBID 12

Data on Agriculture as share of GDP, Industry as share of GDP, Service as share of GDP, Control of Corruption, Rule of Law, Governance and Urbanization have been obtained from the World Bank Development Indicators and Governance Indicators. The data are available for approximately 183 countries for the year 2004. This study uses data from 2004 because from that year data are available for most of the variables. GDI, Urbanization and Property Rights are all converted to decimal numbers between 0 and 1. The purpose of converting the data for these variables is to make the interpretation of the results easier. If all the variables range from 0 to 1 (which all of them do with the exception of governance, rule of law, control of corruption 6 and sector 7 ) it enables the reader to easily interpret the coefficients. This method is borrowed from Ashby s (2006) work. E. Analysis Plan First variables, that have significant relationships with HDI, in alternative research papers, are regressed with GDI to see if similar results prevail. Replicating Ashby (2006) regressions, GDI (instead of HDI) is regressed against Property Rights, Control of Corruption, Urbanization, Ethnic Fragmentation, Sector, Rule of Law and Governance. Rule of law and Governance are variables that were attained from the 6 Data for governance, rule of law and control of corruption range from negative (-2.5) to positive values (2.5). Their conversion would have distorted the nature of the variable and what they try to measure. Their original values have therefore not been converted. 7 Sector was not converted since it is an ordinal variable and it needs to remain that way to effectively gauge the influence of different sectors on GDI 13

International Country Risk Guide in the original studies, but this dataset is restricted and cannot be accessed. Variables such as Rule of Law and Governance, from the World Bank Governance indicators, are used instead. Having said that, we must keep in mind that the prevailing results may not exactly mirror Ashby s findings. See Model 1. Model (1) GDI = β 0 + β 1 (Property Rights) + β 2 (Control of Corruption) + β 3 (Urban) + β 4 (Ethnic) + β 5 (Landlock) + β 6 (Sector) + β 7 (Governance) + β 8 (Rule of Law) In the next step regression 1 is replicated but Sector is excluded and Agriculture, Industry and Services as a % of GDP are added. Sector is excluded because it replicates the effects of Agriculture, Industry and Service as % of GDP. See Model 2. Model (2) GDI = β 0 + β 1 (Property Rights) + β 2 (Control of Corruption) + β 3 (Urban) + β 4 (Ethnic) + β 5 (Landlock) + β 6 (Agriculture as % of GDP) + β 7 (Industry as % of GDP) + β 8 (Service as % of GDP) Finally, a set of regressions are run separately for agricultural, industrial and service oriented countries. In order to do that, we run Model 3 for countries whose Agriculture as % of GDP is greater than 15% (since that is the average value of all observations). Model (3) GDI = β 0 + β 1 (Property Rights) + β 2 (Control of Corruption) + β 3 (Urban) + β 4 (Ethnic) + β 5 (Landlock) Where Agr % GDP > 15 The same regression is repeated for countries whose Industry as % of GDP exceeds 30 percent and Service as % of GDP exceeds 54 percent. 14

(3) GDI = β 0 + β 1 (Property Rights) + β 2 (Control of Corruption) + β 3 (Urban) + β 4 (Ethnic) + β 5 (Landlock) Where Ind % GDP > 30 & Where Serv % GDP > 54 F. Descriptive Results At the beginning of the descriptive statistics process, summary statistics are generated for all relevant variables. F.1. Summary Statistics Table 1 contains descriptive statistics for all the variables of interest. The mean values of the variables listed in the table are relevant when we conduct t tests. Table 1: Descriptive Statistics for All Variables Variable Obs Mean Std. Dev. Min Max GDI 146 0.7016 0.18961 0.29 0.96 Property Rights 138 0.582 0.177 0.176 0.941 Control of Corruption 184-0.0506 1.0130-1.72 2.51 Agriculture as % GDP 159 15.78 14.05 0 69.41 Industry as % GDP 160 30.36 11.99 9.92 92.16 Service as % GDP 159 54.18 15.24 3.75 91.55 Sector 170 3.664 0.7840 1 4 Urban 186 0.5485 0.2352 0.09 1 Ethnic 135 0.4643 0.2741 0.00 0.90 Landlocked 158 0.2341 0.4248 0 1 Governance 184 0.00 1.000-2.0484 2.2764 Rule of Law 184-0.0801 0.9943-2.3595 2.0968 15

In the preliminary findings, correlation tests and multicollinearity tests are run 8. The correlation test shows that all the variables are highly correlated to each other, which means that all of them are relevant to the specifications we want to run. Less than half of the observations (56) have below mean values for property rights and control of corruption and roughly three fourths of the observations (89) have above average values for Gender Development. Seventy out of 160 observations are predominantly agricultural economies whereas the rest are either industrial or service oriented. The hypothesis is that Property Rights might be more important, in terms of its relationship to gender development, in agricultural economies because ownership and access to land in such economies have positive effects for women s economic and social autonomy. Analysis of this variable will therefore be important. There is a mix of observations which makes this relationship a little unclear. India, for example, is a country whose industrial sector comprises of 43 percent of its GDP and property rights are above average yet gender development remains below average. The Central African Republic, however, has below average property rights and gender development and is an agricultural economy. Both countries have low gender development indices yet one is predominantly industrial and the other agricultural. 8 Correlation Matrix and Multicollinearity tables are in the Appendix 16

F.2. Scatter Plots: Below graphs 2 and 3 are scatter plots that are generated to see how Property Rights and Control of Corruption are associated with GDI. Graph 2: Property Rights Index (Gwartney Lawson) vs. Gender Development Index gdi.2.4.6.8 1 0 2 4 6 8 10 propertyg Graph 2 suggests that the Gwartney & Lawson index for property rights has a positive correlation with GDI. We still have to see whether this relationship holds when we take all other relevant factors into account. In addition, the Graph 3 scatter plot suggests a positive correlation between GDI and the World Bank index of Control of Corruption. 17

Graph 3: World Bank index of Control of Corruption vs. Gender Development Index gdi.2.4.6.8 1-2 -1 0 1 2 3 corruptwb This means that higher levels of control of corruption [low levels of corruption] have a positive relationship with higher levels of gender development. F.3. Mean GDI across different levels of Property Rights Are the mean values of property rights the same or different for different levels of GDI? To answer this question the following formula is used to conduct a t test to see if the difference in means is statistically significant or not. The following equation differs from the usual t test formula and is used only when the sample sizes are unequal. As noted in Table 1, the mean value for property rights is 5.01 hence 5 is used as the cut off value for strong and weak property rights. The results are shown in Table 2. 18

Table 2: Mean GDI across different levels of Property Rights GDI if Propertyg < 5 Variable Valid N Mean St. Dev Min Max GDI 56 0.6061071 0.1715757 0.292 0.859 GDI if Propertyg > 5 Variable Valid N Mean St. Dev Min Max Propertyg 90 0.7611556 0.1762444 0.335 0.962 Applying the t test formula to the data in Table 2, for unequal sample sizes a t statistic of 5.23 is obtained, which is statistically significant at the 1% level. Hence, overall gender development is different for countries with above average levels of property rights compared to countries with below average levels of property rights. F.4. Mean GDI across Different Levels of Control of Corruption The observations (countries) in the dataset are separated into two groups: (1) countries with above average levels of control of corruption and (2) countries with below average levels of control of corruption. As noted in Table 1 the mean value of control of corruption of all the observations is zero. Table 3: Mean GDI across different levels of Control of Corruption GDI if CorruptWb < 0 Variable Valid N Mean St. Dev Min Max GDI 56 0.6061071 0.1715757 0.292 0.859 GDI if CorruptWb > 0 Variable Valid N Mean St. Dev Min Max Propertyg 90 0.7611556 0.1762444 0.335 0.962 19

After attaining the mean GDI for both groups, from Table 3, they are used in the t test using the formula for unequal samples. The resulting t score is 10.7033 and is statistically significant at the 1% level. We have therefore established that the overall level of gender development is lower in countries with higher levels of control of corruption. But a causal relationship between these two variables yet has to be established. F.5. Mean GDI across Different Economic Structures To see how GDI differs for countries with different economic structures the t test is conducted using values from Table 4. The average contributions of the agricultural, industrial and the service sectors are used as cut off points. For example if a country s agricultural sector contribution to GDP are more than the average amount (of the observations in my dataset) then the country is included in the first category. Similarly, the average contribution of the industrial sector (30%) and the service sector (54%) are used to place countries in the Industrial and Service categories. The descriptive summaries of the resulting three groups of countries are included in Table 4. Table 4: Mean GDI across Agricultural, Industrial and Service Oriented Economies GDI if Agr > 15% Variable Valid N Mean Std. Dev Min Max GDI 70 0.5839286 0.1791592 0.292 0.947 GDI if Ind > 30% Variable Valid N Mean Std. Dev Min Max GDI 73 0.7032055 0.1960411 0.292 0.962 GDI if Serv > 54% Variable Valid N Mean Std. Dev Min Max GDI 83 0.787494 0.1661636 0.292 0.962 20

After attaining the means, three different t tests are conducted to see whether the differences in means (of GDI) across all three types of economies are statistically significant or not. It turns out that the differences in means are all statistically significant at the 1% level. Hence, we can conclude that overall gender development is different and in fact lower in agricultural economies than in industrial and service oriented economies. F.6. Drawbacks of Data After analyzing the data, it seems that using the agriculture, industry or service share of GDP might not correctly reflect whether a country is predominantly agricultural, industrial or service oriented. Agricultural products have much lower value added then industrial goods, hence even if a country is in truth agricultural, agriculture s share of GDP is not as high the share of its GDP accounted for by industrial goods. In India, service comprises of 53.74% of GDP; Industry is 43.85% of GDP; and Agriculture 18.78% of GDP. Yet a majority of the country is rural and majority of the population resides in rural areas. Perhaps by using percentage of labor employed in each sector we might be able to more accurately indicate the economic structure of an economy. Unfortunately, the UNDP does not have data on these variables, and the World Bank database has a lot of missing data for these variables. Regressions in this thesis will therefore use Agriculture, Industry and Service output as a % of GDP, but results should be drawn with caution. F.7. Hypothesis on Property Rights and Control of Corruption The t test results indicate that, where Property Rights and Control of Corruption are low gender development tends to be low as well. This is consistent with existing 21

literature in the field of institutional economics. What remains to be seen is whether a causal relationship can be established between Property Rights and GDI, and between Control of Corruption and GDI, holding all else constant. In addition based on the results, it seems that gender development also differs across different types of economies. Regression analysis is needed to explore whether the impact of Property Rights and Control of Corruption are the same across these types of economies. G. Regression Results and Discussion This section presents the main regressions of interest and analyzes the effects of Property Rights and Control of Corruption on GDI. Table 5 presents the results of six regressions that are divided into two groups. The first group, which includes regressions 1 to 3, includes all relevant variables; Property Rights and Control of Corruption are the only variables that measure quality of institutions. Regression 1 includes Property Rights and Control of Corruption along with the control variables. Initially it seemed that they might be correlated which is why Property Rights and Control of Corruption are subsequently included in separate specifications; regression 2 includes only Property Rights and other relevant variables; and regression 3 includes only Control of Corruption along with the control variables. The second group, which consists of regressions 4 through 6, includes additional institutional variables such as Governance and Rule of Law. The distinction between these two groups (regressions 1-3 and 4-6) of models is necessary because the results are 22

different and worthy of attention. Property Rights and Control of Corruption, again, are included together in some specifications (regression 4) and separately in others (regressions 5 & 6). Initially a variable called ratio of female to male earned income is included but its coefficients are statistically insignificant which is surprising. The results are nevertheless in the appendix for review 9. At first it seems as though this variable influences GDI. But how much women get paid relative to men has a greater impact on women s decision to enter and stay in the labor force than on more social variables such as GDI. The coefficients of all the variables, however, have the same sign and level of statistical significance as we see in Table 5. Hence the addition of Ratio of Female to Male Earned Income does not enhance the model nor does it increase our understanding of the variation of GDI. G.1. Regression Results Table 5 can help us analyze how property rights and control of corruption affect GDI in two ways: (1) what the statistical significance of each of the two variables is, holding the other constant; and (2) what the statistical significance of each of the two variables is without holding the other variable constant. In regression 1, when property rights and control of corruption are both included in the model, along with the remaining relevant variables, only property rights seems to have a statistically significant relationship with GDI. In regressions 2 and 3 when these two variables are regressed in separate regressions both turn out to be statistically significant, ceteris paribus. 9 Table 7 in the Appendix 23

Table 5: Regressions against GDI Dependent Variable: GDI Property Rights & Control of Corruption (1) (2) Both Property Property Rights Rights & Control Corruption Property Rights.298 (0.13)**.391 (5.43)*** (3) Control of Corruption Governance & Rule of Law included (4) Both Property Rights & Control Corruption -----.114 (0.80) (5) Property Rights.114 (0.80) (6) Control of Corruption ---- Control of Corruption.020 (0.02) ----.068 (5.35)*** -.033 (-0.78) ---- -.024 (-0.58) Urban.251 (0.06)***.265 (4.69)***.249 (4.22)***.254 (4.29) ***.239 (4.27)***.242 (4.22)*** Ethnic -.183 (0.04)*** -.192 (-4.83)*** -.178 (-4.36)*** -.193 (-4.83)*** -.191 (-4.79)*** -.193 (-4.91)*** Landlocked -.076 (0.03)*** -.073-2.88*** -.073 (-2.85)*** -.078 (-3.11)*** -.0786 (-3.13)*** -.072 (-2.94)*** Sector.045 (0.01)***.043 (3.21)***.039 (3.19)***.039 (2.86) ***.0401 (2.94)***.027 (2.18)** Rule of Law ---- ---- ----.086 (1.79) * Governance ---- ---- ---- -.689 (-1.29).054 (2.16)** -.841 (-1.70)*.099 (2.33)** -.910 (-1.83)* Constant.316 (0.09).263 (4.10).504 (8.91).568 (3.78).592 (4.05).720 (6.29) N 93 93 100 93 93 100 R Squared 0.8113 0.8095 0.7883 0.8247 0.8234 0.8144 * Significant at the 10% level, ** Significant at the 5% level, ***Significant at the 1% level Values within parentheses are the t values 24

G.2. Effects of Urbanization, Ethnic Fragmentation, being Landlocked & type of Sector Variables such as Urbanization, Ethnic Fragmentation, whether a country is landlocked and type of sector are all statistically significant in every specification in Table 5 and these results are consistent with theoretical findings. The more urbanized a country is the greater are its economic opportunities and access to public services such as schools and hospitals. In rural areas, the lack of jobs, poor infrastructure, inaccessible health and educational services limit women s ability to generate income, go to work, attain education and seek proper treatment for maladies all of which are core elements of the GDI. Hence urbanization has positive, substantial and statistically significant effects on GDI. Being landlocked mainly affects the ease at which a country can trade. The absence of a seaport increases transportation costs, increases the prices of export products and adversely affects terms of trade. Since trade is essential to economic growth, the landlocked-ness of a country hinders the ability of its economy to generate more income and therefore distribute more to its citizens. Trade is not only important in terms of how much money it brings into the economy, but also in terms of competitive low prices people pay for primary goods. If countries import corn from those who have a comparative advantage in making it, then consumers can purchase more necessary goods to sustain their livelihood. Being landlocked hinders an economy s ability to trade and generate income and insure people s ability to purchase goods at low prices. The negative 25

effects of being landlocked are also highly significant when it comes to the well being of women. The statistical results in Table 5 support this hypothesis. The level of ethnic fragmentation has a substantial influence on the level of conflict within a country. Conflict erodes the existence of economic opportunities, and levels of investment (domestic or foreign), and leads to mass immigration. Few people are left behind to work in the public sector for the delivery of social services, working as doctors in hospitals or teachers in schools. Insufficient delivery of essential social services limits people s and women s ability to lead healthy lives. Women suffer disproportionately compared to men in conflict ridden societies because they are the most economically and socially vulnerable ones. All these factors therefore suggest that ethnic fragmentation and conflict adversely affect women. And the negative and statistically significant coefficients of ethnic fragmentation, in the regressions, support this theory. The hypothesis that the level of industrialization affects the level of gender development is also strongly supported by the statistically significant coefficient (at the 1% level) for Sector. Industrialized countries, by definition, have greater resources and higher levels of income. They will, therefore, in general have more resources to distribute across their population than agricultural economies will. Industrialized and service oriented countries also provide greater economic opportunities for their female populations. As the Malaysian economy, for example, shifted from agriculture to manufacturing, the types of jobs that opened up were for women who had the skills necessary for electronic manufacturing (as electronic goods became Malaysia s main exports). Increased employment helped women raise their income and therefore afford to 26

sustain better health, and higher levels of education all of which are captured by the GDI. G.3. The role of Governance and Rule of Law In previous studies the variables Governance and Rule of Law were constructed using the International Country Risk Guide (ICRG). This dataset, however, is not available for the purpose of the present study. Alternative Governance and Rule of Law variables are attained from the World Bank Worldwide Governance Indicators. Once these two factors are taken into account we see that governance seems to have no relevant association with GDI ceteris paribus when both property rights and control of corruption are in the model. When the model includes just one of the two variables Governance seems to be statistically significant at the 10% level. Rule of law turns out to be statistically significant in every specification it is included in. Control of Corruption, however, becomes statistically insignificant when Rule of law is added to the model. Multicollinearity tests are run for regression 4 10. According to the results there is very high multicollinearity between Control of Corruption and Rule of Law. Hence we run into problems in specifications where both are included. Although multicollinearity does not bias coefficients, it tends to enlarge standard errors. If standard errors increase in value, the t statistics become small, and it becomes harder to reject the null hypothesis. As the t statistic become increasingly unreliable it is increasingly difficult to confidently say what the effect of Control of Corruption is on GDI, holding Rule of Law constant. 10 Refer to Appendix Table 2 27

In such a scenario if both of the variables are highly correlated then we run into a problem of redundancy, where each variable replicates the effects of the other. It would be interesting to see what the results are if rule of law is excluded from the model. Table 6 includes results of the regression that excludes Rule of Law. Table 6: Regression without Rule of Law Dependent Variable: GDI (7) Without Rule of Law Property Rights.231 (1.79)* Control of Corruption.0326 (1.42) Urban.233 (3.97)*** Ethnic -.194 (-4.77)*** Landlock -.0754 (-2.96)*** Sector.0409 (2.95)*** Governance -.926 (-1.77)* Constant.536 (3.55) N 93 R square 0.8180 * Significant at the 10% level, ** Significant at the 5% level, *** Significant at the 1% level, Values within parentheses are the t values In Table 6 we see that when Rule of Law is excluded the coefficient for Control of Corruption does not become anymore significant, however, Property Rights becomes statistically significant at the 10% level, ceteris paribus. 28

Property Rights also becomes insignificant when Rule of Law and Governance are added to the model. One can make the argument that Rule of Law and Governance take into account the government s ability to secure property rights and provide citizens with a legal framework that safeguards their assets. Hence when we include Governance and Rule of Law the additional effects of Property Rights on GDI become insignificant. If we include Property Rights and do not take into account Governance and Rule of Law, Property Rights have a significant impact on the Gender Development Index. This could mean that although Property Rights are important, Governance and Rule of Law take precedence over it in terms of their importance to the socioeconomic well being of women. G.4. What role does the structure of an economy play in influencing GDI? It is rather difficult to classify economies according to their dominant sectors. Agriculture as a percentage of GDP is a potential variable to gauge economic development. Another variable such as Sector is constructed as an ordinal variable to classify economies according to whether they are agricultural, industrial or service oriented. Sector is not used in the same regressions as Agriculture (Industry & Service) as % of GDP are included, because they replicate the same effect. In regressions 1 through 7 Sector always has a statistically significant impact on GDI. The Sector variable (which is an ordinal variable) is constructed as follows: Agricultural = 1; Manufacturing = 2; Industrial = 3; and Service = 4. So if a country s dominant sector is industrial then it will receive a 3. Since the coefficient on sector is 29

positive and statistically significant, holding all else constant, we can conclude that as an economy becomes more industrial or service oriented, GDI increases, probably because of the favorable economic opportunities available to women. Table 7: Regressions with structure of economy Dependent Variable: GDI (8) (9) (10) Property Rights.061 (0.49).148 (1.81)* ---- Control of Corruption.019 (0.93) ----.039 (2.77)*** Urban.145** (2.32) Ethnic -.170 (-4.15)*** Landlock -.049* (-1.98) Agriculture as -.004 % GDP (-4.42)***.162 (2.72)*** -.178 (-4.43)*** -.047 (-1.90)* -.004 (-4.36)***.147 (2.37)** -.165 (-4.11)*** -.047 (-1.91)* -.003 (-3.95)*** Industry as % GDP -.001 (-1.48) -.001 (-1.54) -.0007 (-1.03) Service as % GDP.001 (1.78)*.001 (1.88)*.001 (1.95)* Constant.688 (6.16).628 (6.93).686 (8.76) N 77 77 84 R Square 0.8492 0.8473 0.8277 * Significant at the 10% level, ** Significant at the 5% level, ***Significant at the 1% level Values within parentheses are the t values Although regressions 1 through 7 use the ordinal variable Sector, in regressions 8 through 10 in Table 7, a different approach is tried. Continuous variables (Agriculture, 30

Industry and Service as % of GDP), that measure each sector s contribution of a country s GDP, are included. It seems that by holding all else constant, the more an economy becomes agricultural, the more GDI decreases; however, the effect is not substantial. Although coefficients for service are also statistically significant at the 10% level, they are not substantially significant. Even in these regressions, we find that when Property Rights and Control of Corruption are in separate specifications both tend to be statistically significant, otherwise neither seem to have a relevant impact on GDI. Besides Property Rights and Control of Corruption, socioeconomic variables such as Urbanization, Ethnic Fragmentation, Landlocked and Sector are all statistically significant, holding all else constant. Another attempt to examine whether the effects of Control of Corruption and Property Rights are different across different economies, the same regressions are run for countries with agriculture as % of GDP greater than 15%, industry > 30% and service > 54%. The results are in Table 8. As was expected, Property Rights indeed seem to have statistically significant positive effects on GDI in agricultural economies. However it has a greater impact on GDI in industrial economies. Such is true when it is regressed with and without control of corruption (and other relevant variables) with the exception of service oriented countries. Control of Corruption, however, only seems to have statistically significant positive 31

Table 8: Regressions on countries with > 15% Agriculture as % of GDP, > 30% Industry as % of GDP and > 54% Service as % of GDP > 15% Agriculture % of GDP > 30% Industry % of GDP > 54% Service as % of GDP (11) Property Rights & Control of Corruption (12) Property Rights (13) Control of Corruption (14) Property Rights & Control of Corruption (15) Property Rights (16) Control of Corruption (17) Property Rights & Control of Corruption (18) Property Rights (19) Control of Corruption Property Rights.371 (1.89)*.474 (4.16)*** ----.430 (2.48)**.429 (5.07)*** -----.226 (1.38).413 (5.57)*** ---- Control of Corruption.025 (0.64) ----.084 (3.73)*** -.0003 (-0.01) -----.069 (4.32)***.033 (1.28) ----.068 (5.91)*** 32 Urban.369 (3.88)***.383 (4.19)***.384 (4.08)***.437 (5.10)***.436 (5.72)***.357 (4.20)***.330 (5.24)***.346 (5.58)***.322 (5.24)*** Ethnic -.141 (-2.03)** -.146 (-2.14)** -.122 (-1.88)* -.189 (-3.25)*** -.189 (-3.29)*** -.220 (-3.66)*** -.156 (-3.50)*** -.155 (-3.44)*** -.165 (-3.72)*** Landlocked -.052 (-1.43) -.051 (-1.42) -.040 (-1.13) -.062 (-1.86)* -.062 (-1.92)* -.062 (-1.79)* -.064 (-1.95)* -.056 (-1.73)* -.070 (-2.14)** Constant.342 (2.72).278 (3.64).529 (7.40).301 (.134).303 (3.98).611 (8.93).476 (4.47).359 (6.49).613 (13.26) N 46 46 50 55 55 57 59 59 60 R square 0.73303 0.7303 0.6996 0.8415 0.8415 0.8133 0.8165 0.8109 0.8081 * Significant at the 10% level, ** Significant at the 5% level, *** Significant at the 1% level, Values within parentheses are the t value