Microfinance and Poverty Alleviation: Measuring the Effectiveness of Village Banking in Haiti, a Regression Analysis

Microfinance and Poverty Alleviation: Measuring the Effectiveness of Village Banking in Haiti, a Regression Analysis Sara Thompson Presented at FINCA International's 2006 Research Symposium March 24 th, 2006

Table of Contents Executive Summary 1 Background 1 Introduction 2 Research Methods 3 Village Banking: Alleviation of Social Factors of Poverty 4 Village Banking: Alleviation of Economic Factors of Poverty 7 Conclusions 11 Recommendations 12 Appendix I 14 Appendix II 15

Executive Summary What effect does microfinance have on poverty alleviation? Recent trends in the microfinance industry have suggested that organizations assess their impact through both financial and social indicators. The main hypothesis of this paper is that microfinance, delivered through FINCA Haiti village banking, is positively effecting clients. This paper uses FINCA Haiti client assessment survey data to present evidence of the effectiveness of FINCA Haiti's village banking on both economic and social metrics. Using regression analysis, results show that village banking is increasing social wealth among clients. However the results show that village banking is not significantly effecting the economic wealth of clients. Further analysis through means testing shows that the lack of significance may be due to the control composition of non-rural bias. Crosstabulation analysis shows differing expenditure patterns between rural and non-rural clients and indicates the possibility of a lack of key indicators. The results of this paper have important research and programmatic implications for how FINCA utilizes its village banking methodology specifically and the microfinance community in general. The findings imply that a more detailed analysis could be done with survey data technologically linked to management information systems to ensure the inclusion of all key indicators. Program and operational implications would warrant increased services for rural clients such as money transfer services and home improvement loan products. Background Haiti is one of the most disadvantaged countries in the world according to social, economic, and environmental indicators. Half of Haiti's population is illiterate, 65% of the population is living under the poverty line, and half of the population has no access to potable water 1. Haiti is currently experiencing an unemployment rate greater than 66% in the formal sector 2 which contributes to a staggering number of Haitians seeking work in the informal sector. According to Alejandro Portes, the informal sector includes direct subsistence production, micro-scale commodity production and exchange, and backward capitalist production which includes both unprotected labor hired by small enterprises and laborers working under subcontracting arrangements 3. Access to credit through the microfinance industry in Haiti is critically important to the sustainability and growth of the informal sector. Affordable credit provides capital for start-up expenses, for restocking inventories of micro-scale traders and replenishing working capital for micro-scale producers. Microfinance also results in the reduction of onerous debt situations for the poor informal sector workers who borrow (as credit at more reasonable rates is not available) at the exploitative terms of money lenders. 1 World Bank country overview. Viewed March, 2006. Last Updated July, 2004.http://web.worldbank.org/WBSITE/EXTERNAL/COUNTRIES/LACEXT/HAITIEXTN 2 CIA country factbook. Viewed March, 2006. Last Updated January 10th, 2006. http://www.cia.gov/cia/publications/factbook/print/ha.html 3 Portes, A. "The Informal Sector: Definition, Controversy and Relation to National Development" Review VII, Summer 1983. 1

Microfinance is carried out in a variety of forms throughout the microfinance industry. FINCA Haiti uses village banking to deliver microfinance services to clients. Village banking is unique in that it gives borrowers the responsibility and autonomy to run credit and savings associations within their community. This paper focuses on the effectiveness of village banking as a tool to alleviate poverty. Introduction Village banking not only provides credit to micro entrepreneurs but also seeks to build social wealth among clients. Social wealth can be defined as the combination of human and social capital. Investing in human capital, such as education and health, is investing in individuals to increase the productivity and the innovation of the labor force. Investing in social capital is investing in the networks and relationships between individuals, s, and institutions. FINCA measures village banking effectiveness in terms of an increase in economic expenditure as well as social wealth. FINCA's Client Assessment Tool (FCAT) includes a metric on social wealth and a metric on Daily Per Capita Expenditure (DPCE), which serve as the basis of the analysis in this monograph. Social wealth is measured by the social wealth metric of FINCA through eight categories of questions that seek to quantify social wealth through ordinal questioning. The social wealth metric of FCAT measures indicators of both human and social capital according to: 1) the educational situation of the family; 2) the health of the family in terms of immunizations, access to potable water, and access to modern medical facilities, (which leads to less economic stress events due to health crises on the household economy) 4 ; 3) the housing condition assessed through the presence of utilities, as housing is not only for living but also often the location of business activities for women in the informal economy, 4) the literacy of the client as defined by their ability to read, write, and perform basic math 5, 5) food security as measured by the number of meals per day the family consumes, as caloric intake is correlated with health, 6) the perceived social trust in relationships in the community and market, 7) empowerment, measured by the amount of the loan that the clients manages, and 8) the lack of vulnerabilities, including large family sizes, chronically ill or recent death of family members, natural disasters, and single parent households. 4 Helzi Noponen and Paula Kantor, 2004. "Crises, Setbacks and Chronic Problems The Determinants of Economic Stress Events Among Poor Households in India Journal of International Development, 16, 529-545. 5 Coulombe, Serge, et al. International Adult Literacy Survey: Literacy scores, human capital and growth across fourteen OECD countries. Statistics Canada; Ottowa, 2004. 2

The expenditure metric of the FCAT measures the physical capital growth by determining expenditure as a proxy to income. A series of questions regarding expenditures in multiple consumption categories is asked including consumption of home grown food and expenditures on transportation, education, and fuel. Each question is asked for an appropriate recall period that the respondent is most likely to provide the highest recall accuracy--weekly, monthly or yearly. The value of expenditure is multiplied by the number of time periods in a year. The total expenditures at the end of the metric are divided by 365 and then by the number of members in the household resulting in the DPCE of the household. Income can be extremely hard to pinpoint in the informal sector as books are often not kept. Also a large portion of the informal sector is reliant on seasonal agricultural production and income therefore varies from season to season. Activities often are both income generating and subsistence based. Household and enterprise accounts are frequently intertwined. Female clients rarely know the income of their husbands or other adult males. Therefore expenditure is used as a proxy for income as the data is typically more reliable. An overall increase in expenditure indicates an increase in income. However, it is important to note that some expenditure categories may actually decrease with an increase in income as buying in bulk will cut expenditures in food and household products. The decrease in these expenditure categories should be complemented by an increase in expenditures such as health and education. Therefore it is hypothesized that DPCE will increase with village banking intervention. The social wealth and DPCE metrics are the basis of program impact analysis in the two regression models of this monograph. The first model analyzes the effectiveness of village banking on social wealth and concludes that village banking is positively impacting social wealth. The second model analyzes the effectiveness of village banking on DPCE and produces inconclusive results. Research Methods A stratified random sample of FINCA Haiti clients was selected to be interviewed by interns using FCAT. The data was collected in the field in the summer of 2005 and includes a sample size of 325 female village banking clients. A is used to analyze program impact. New clients represent the control and are defined as clients who have joined a new village banking and been with FINCA for less than one loan cycle and also clients who have recently joined an existing village banking. Current clients represent the treatment and are defined as clients who have been with FINCA for one year or more or who were previously clients and have rejoined the program. Exiting clients are not used in this analysis. The first model of this monograph uses regression analysis to determine the relationship between village banking and increased social wealth among clients. Cross-tabulations are 3

used to analyze the categories of social wealth that are most effected by village banking including vulnerabilities, social capital and educational situation of the family. The second model uses regression analysis to determine the link between village banking and the increased expenditures of clients. Means testing is used to demonstrate control composition difference and cross-tabulations are used to illustrate differences in expenditures between rural versus non-rural populations. Village Banking: Alleviation Social Poverty Factors One main goal of village banking is to give borrowers direct participation in the poverty alleviation of their own community. Village banking seeks to reduce widespread features of poverty including malnutrition, poor health care, a lack of education, unsafe housing, and social instability 6. These features of poverty are considered social factors, in that they are not quantified in monetary terms. An analysis of the effectiveness of village banking on the alleviation of social factors of poverty is crucial to understanding the effectiveness of village banking as a whole. Social wealth can be measured by a lack of, or reduction in social poverty factors. In this model social wealth is hypothesized to increase among the treatment. The treatment is composed of those clients that have been with FINCA and been participants in village banking for one year or more. There are several factors that are included in this model in addition to the treatment including the level of education as well as the DPCE. The variables are shown in Table 1 below along with the predicted signs. Holding DPCE and level of education constant there is hypothesized to be an increase in social wealth among the treatment. The dependent variable for this model is the total social score of the social metric of FCAT. The client can receive a score of 1-4 in each of the eight categories listed above. This yields a total score of 8 to 32. A higher score represents less poverty and more social wealth. The is the variable representing program impact between the control and treated clients. After treatment, or being a FINCA client for over one year, the client is hypothesized to show an increase in the social wealth. The level of education variable is a continuous scaled variable defined by the number of years of schooling attended by the client. A positive relationship was predicted between education level and social wealth of the household as women with more schooling are more likely to pursue education, immunizations, and healthcare for their households 7. 6 FINCA International History. Viewed March 2006. Updated 2005. http://www.villagebanking.org/history.htm 7 Coulombe, Serge, et al. International Adult Literacy Survey: Literacy scores, human capital and growth across fourteen OECD countries. Statistics Canada; Ottowa, 2004. 4

Table 1: Factors Associated with the Household s Social Wealth Independent Variable with Expected Sign Variable Type Coefficient Comparison Group (+) ordinal 0.818* Level of Education (+) continuous 0.445 Daily Per Capita Expenditure (+) continuous natural log 2.334 Constant 9.567 R 2 49% Adjusted R 48.5% F Statistic 98.83 Number of Cases 312 Note: All results are significant at a 99% confidence interval unless otherwise noted *indicates the results were significant at a 95% confidence interval The daily per capita expenditure variable is equal to the total of the expenditure metric divided by 365 days and by the number of household members. The expenditure metric is composed of questions regarding yearly expenditure on several categories of goods and services. As expenditures rise income is generally rising as well. The rise in income is hypothesized to lead to the rise in social wealth, especially in areas of health, education and social capital. The above results show that both daily per capita expenditure and level of education have a positive effect on the social wealth of the clients. Holding daily per capita expenditure and level of education constant, village banking is having a positive effect on the social wealth of clients. FINCA Haiti village banking is showing a positive effect on the social wealth of their clients greater than the effect of education. Further analysis shows that the most impact in social wealth is observed in the categories of vulnerabilities, social capital and educational situation of the family. Tables 2-4 show the changes in particular social wealth categories from new to current clients. Table 2: Decrease in Vulnerabilities Social Metrics Vulnerabilities 3 or more vulnerabilities Count % within Comparison Group New Current Total 25 3 28 11.1% 3.4% 8.9% 2 vulnerabilities Count % within 1 vulnerability Count % within zero vulnerabilities Count % within 65 28.9% 90 40.0% 45 20.0% Total 225 15 16.9% 44 49.4% 27 30.3% 89 80 25.5% 134 42.7% 72 22.9% 314 Note: All results are significant at a 99% confidence interval. 5

The table above shows that while 40% of new clients had 2 or more vulnerabilities, only 20.3% of current clients had 2 or more vulnerabilities. The change implies that village banking is assisting clients by enhancing their economic and social safety nets to help in times of difficulty. Members of the treatment or those who have been participating in village banking are more likely to have less vulnerabilities. Being a part of village banking is allowing the clients the ability to rely on those in their community in times of hardship more than those who are not part of village banking. Table 3: Increase in Social Trust Social Metrics Level of Trust with Others Never Count % within Comparison Group New Current Total 40 6 17.8% Seldom Count % within 30 13.3% Sometimes Count 78 % within 34.7% Always Count 77 % within 34.2% Total 225 6.7% 4 4.5% 26 29.2% 53 59.6% 89 46 14.6% 34 10.8% 104 33.1% 130 41.4% 314 Note: All results are significant at a 99% confidence interval. Table 3 shows that 34.2% of new clients as opposed to 59.6% of current clients feel that they can always trust those, with whom they associate, this 25.4% difference suggests that participating in village banking is helping clients to increasingly build social trust. Social trust here is defined as the level of trust one has in their relationships throughout the community and market place. By being a part of village banking clients are building trust with one another. 6

Table 4: Increase in Educational Situation of Household Social Metrics Educational Situation of the Family Poor Count % within Comparison Group New Current Total 75 16 33.3% Fair Count % within 92 40.9% Good Count 58 % within 25.8% Total 225 18.0% 47 52.8% 26 29.2% 89 91 29.0% 139 44.3% 84 26.8% 314 Note all results are significant at a 95% confidence interval. Table 4 shows that while 33.3% of new clients view the educational situation of their family as poor, only 18% of current clients feel the same. These results imply that FINCA Haiti village banking is providing the opportunity for more clients to pursue education for their children. The program is impacting the clients ability to invest in the future through human capital. The social impact model indicates that village banking is increasing social wealth by reducing social factors of poverty including vulnerabilities, social instability, and lack of education. Village banking is increasing the ability of members to build social safety nets and social trust. Village banking is also increasing the ability of members to educate their family members. Increases in social capital are represented by both a decrease in vulnerabilities and an increase in social trust. Increases in human capital are represented in increased education. The increases in social and human capital that are provided by village banking are crucial factors in economic growth. Village Banking: Alleviation of Economic Poverty Factors Another main goal of village banking is to provide credit for monetary investment in entrepreneurial endeavors. Village banking uses community focused credit and savings associations to provide needed credit. Through increased access to credit and savings village banking seeks to increase the monetary wealth or income of households. FINCA uses DPCE as a proxy for income. The goal is not simply to increase the DPCE but also to see a shift from expenditures on survival essentials to investments in social and human capital, such as health and education. 7

In this model it is hypothesized that DPCE will increase among the treatment. The expenditures should increase in areas of social and human capital especially. There are several other factors in this model including level of education, social wealth, location of residence, and a dependency ratio. These variables can be found in Table 5 below with the expected signs. Holding these variables constant DPCE should increase among the treatment. The dependent variable for this model is DPCE of the clients as defined by the FCAT. The variable is the natural log of DPCE. The natural log was taken to create a more linear relationship between the expenditures and independent variables. This is due to the fact that the level of expenditure rises at an exponential rate as compared to social wealth. The is the variable representing the effectiveness of village banking between the control and treated clients. It is hypothesized that village banking will have a positive effect on daily per capita expenditure. Members of village banking have access to and management of credit and savings associations that are hypothesized to increase their overall income and therefore expenditures. The years of education of the client are hypothesized to have a positive effect on the daily per capita expenditure of the client. Development theory holds that education contributes directly to the growth of GDP by improving the skills and productive capacities of the labor force 8. The increase in education is hypothesized to lead to an increase in skills and therefore productive income, leading to higher expenditures. The social wealth of the client is hypothesized to have a positive effect on the daily per capita expenditure of the client. According to development theory social wealth leads to an increase in economic growth and income or expenditures. Social wealth in the form of education and health lead to a more productive labor force, leading to increased income and expenditure. The social capital component of social wealth is also crucial. It is through the solidarity and security of social capital that institutional innovation can occur and lead to economic growth 9. The location variable is a dummy variable based on non rural residents only. A dummy variable was created by combining urban and per-urban clients for two reasons. First, peri-urban and urban areas in Haiti have similar characteristics in terms of access to infrastructure. Second, less than 30 observations each were collected in peri-urban and urban areas among current clients. Sample sizes under 30 make significance harder to obtain. By combining urban and peri-urban clients a sample size greater than 30 was obtained. It is hypothesized that the non rural clients will have higher expenditures with program impact than the rural clients. This is because non 8 Psacharopoulos, George. Education for Development: An analysis of Investment Choices; A World Bank Publication. Oxford: Oxford University Press. 1985. 9 Ruttan, Vernon W. Social Science Knowledge and Economic Development: An Institutional Design Perspective. Ann Arbor: University of Michigan Press. 2003. 10 Narayan, Deepa and Lant Pritchett. Social capital: evidence and implications. Washington, D.C.: World Bank. 1997. 8

rural clients have access to more diversified markets, including input, product, and consumer markets. The dependent ratio variable is composed of the number of household members divided by the number of income earners in the household. This ratio is hypothesized to cause a decrease in the daily per capita expenditure of the family. When there are fewer earners per dependents there is less money for expenditures per dependent. Also, as the number of dependents rise in to the number of income earners, the burden of the dependents is placed on the woman. Her time is spent more with the care of dependents than the pursuit of economic activities and therefore her income and expenditure is hypothesized to be lower. Table 5: Factors Associated with Daily Per Capita Expenditure of the Household Independent Variable with Expected Sign Variable Type Coefficient Comparison Group (+) ordinal -0.021** Level of Education (+) continuous -0.012** Social Wealth (+) continuous 0.077 Location (+) dummy -0.155* Dependent Ratio (-) continuous -0.093 Constant 2.507 R2 34.80% Adjusted R 33.80% F statistic 32.807 Number of Cases 312 Note: All results are significant at a 99% confidence interval unless otherwise noted *Indicates the results were significant at a 95% confidence interval **Indicates the results were not significant The above results show that social wealth positively effects the daily per capita expenditure of the clients. The dependent ratio has a negative effect on the daily per capita expenditure of the clients, as hypothesized. The level of education variable is not significant. However, the variable must stay in the model due to its' strong relevance in development theory. Education is a very robust indicator that is often correlated with many other variables. In this case the level of education is collinear with the, social wealth, and location. Collinearity could effect the significance of this variable. The location variable is significant but with a perverse result of rural clients exceeding non rural clients in daily per capita expenditure gains. Holding all other variables constant the variable is not significant and does not show positive village banking effects on daily per capita expenditure. A difference in the composition of the control and treatment s could be effecting the significance of the variable. 9

Table 6: Sample Size by Location and Comparison Group New Percent of Total Current Percent of Total Total Peri-Urban 61 27.1% 18 20.22% 79 Urban 57 25.3% 20 22.5% 77 Rural 107 47.6% 51 57.3% 158 Total 225 89 314 The new, representing the control is more urban and peri-urban than the treatment. The treatment is more heavily rural indicating that newer clients may be selected from urban areas more frequently than in the past. FINCA Haiti may be moving from rural to urban clients. While current village banking clients are composed of 57.3% rural, new clients are composed of only 47.6% rural. The difference in the demographics of the may be the cause for the insignificance of the variable in the above model. One additional explanation for the insignificance of the variable over non rural areas could be due to the omission of an important variable. Including savings in the expenditure metric could positively effect the daily per capita expenditure and possibly the significance of the variable. For example, if urban and peri-urban clients are saving greater amounts than rural clients, the significance of the value could be heightened by the inclusion of such data. It could be that the urban and periurban clients are choosing to save higher amounts and that rural clients are more likely to spend income in order to improve their standard of living. Examining the expense categories that have increased with the variable by location illustrates rural and urban differences in economic priorities. Table 7: Increase in Expenditures, Rural Clients Expenditures on Home Repairs, Maintenance, and Improvement in Past Year Expenditures on Remittances Sent in Past Year Comparison Group New Current New Current 107 51 107 51 N Mean Standard Deviation 1240.8879 1962.5679 3178.8235 2579.8131 5450.3922 5288.31794 6810.93557 14689.49243 Standard Error Mean 189.72861 740.51263 658.43799 2056.940 Table 7 shows that the two main areas of increased expenditure from new to current, rural clients are in home repairs and remittances. The expenditures more than doubled in each of these categories as income increased. 10

Table 8: Increase in Expenditures, Urban Clients Expenditures on Schooling for Household in Past Year Comparison Group New Current 57 20 N Mean Standard Deviation 8624.0877 13665.70696 13594.050 15493.77422 Standard Error Mean 1810.06711 3464.51324 The above table shows that the increase in expenditures for the urban population leads to the greatest increase in expenditure category of education. Current clients increased spending on education by 63.44% from new clients. By increasing expenditures on home improvement the rural clients are investing heavily in the short-term in order to improve their current standard of living and create adequate housing. It is not clear if the rural clients are sending remittances to unemployed family members who have traveled to Port-au-Prince in search of formal employment, or if the remittances are going to their children who are pursuing education in a large city. It is most likely a combination of the two and therefore a combination of short-term and longterm investments. However, the urban population overwhelmingly raised expenditures in only the category of education for their household. The investment in education is a long-term investment. If the urban clients are more inclined to invest in the long-term they are hypothesized also to be more inclined to save. The above hypothesis could be researched in more depth by including the amounts of savings in the expenditure metric, or by providing a link between the FCAT and FINCA MIS. It is important to determine the cause of the above results for the first model. FINCA Haiti may be choosing more new clients from urban areas. It is also possible that an error in sampling selection occurred and was not representative of the total population of new and current clients by location. If further research shows that FINCA village banking yields the most impact in rural areas it would be beneficial for FINCA to ensure rural clients are proportionally accepted as new clients. Conclusions The analysis of the first model shows that FINCA village banking is having a positive effect on the social wealth of the clients and their households. The positive effect is seen most clearly in the categories of vulnerabilities, social trust and education. Social wealth in general and these categories in particular are all critical to development by acting as a catalyst for economic development. The analysis of the second model is inconclusive with the current data set. There are two possibilities that arise regarding the effectiveness of village banking on the economic wealth of the clients. First, the village banking is having a positive effect in rural areas by 11

assisting clients in securing more adequate housing and supporting migrant relatives. Village banking may be having a positive effect in non rural areas by increasing the educational expenditures of a family and possibly savings. However, further research needs to be done with an appropriate sample selection to solidify the results. Secondly, it is possible that village banking is not yet showing significant impact in the expenditure metric though it may be significant in the near future. FINCA Haiti is providing access to credit and increasing social wealth at the community level through village banking. Studies have shown that increased village-level social capital raises household incomes 10. Both human capital, represented in education, and social capital, represented in vulnerabilities and social trust, show improvements due to program impact. These are all important forms of endogenous growth leading to increased efficiency in development. Recommendations Research The following are research related recommendations that have arisen from the above analysis. FINCA's commitment to program evaluation is reflected in the depth of the data collected in FCAT. The recommendations are feasible within the framework of FINCA program evaluation. The data collection of the FCAT in 2006 would benefit from a sample selection that adequately measured a sample of the population in all subcategories. Further research of program impact by location may be crucial to redefining the target of new clients. Technological links between the FCAT and FINCA Haiti MIS could assist in offering even greater depth of information to analysis of the effectiveness of village banking. Research on remittances may prove beneficial to FINCA programming as rural to urban remittances are rarely researched but common in Haiti. Program The following program recommendations have arisen from the above analysis. The program recommendations seek to improve upon the effectiveness of alleviating poverty through FINCA specifically and village banking in general. Home improvement loans may be a worthwhile addition to the financial services provided by FINCA Haiti, especially to rural clients. By providing financing for home improvement the rural clients can both pursue long-term investments in 12

education and savings, as well as improve their standard of living through adequate housing. Money transfer services may be another worthwhile addition to the financial services provided by FINCA Haiti. Providing low cost money transfers from rural clients to their family members could free up income to be used for savings. 13

Appendix I Alleviation of Social Poverty Factors Model Summary Model 1 Change Statistics Adjusted Std. Error of R Square R R Square R Square the Estimate Change F Change df1 df2 Sig. F Change.700 a.490.485 2.71296.490 98.830 3 309.000 a. Predictors: (Constant), revised, Level of Education, ln_dpce Model 1 Regression Residual Total ANOVA b Sum of Squares df Mean Square F Sig. 2182.203 3 727.401 98.830.000 a 2274.282 309 7.360 4456.486 312 a. Predictors: (Constant), revised, Level of Education, ln_dpce b. Dependent Variable: Total Social Score Model 1 (Constant) Level of Education ln_dpce revised Unstandardized Coefficients a. Dependent Variable: Total Social Score Coefficients a Standardized Coefficients Collinearity Statistics B Std. Error Beta t Sig. Tolerance VIF 9.567 1.121 8.534.000.445.038.490 11.625.000.928 1.078 2.334.269.368 8.690.000.921 1.086.818.341.098 2.395.017.992 1.008 14

Alleviation of Economic Poverty Factors Appendix II Model Summary Std. Error Model R R Square Adjusted R Square of the Estimate 1.590(a).348.338.48491 a Predictors: (Constant), Level of Education, revised, location dummy urban, household members per income earner, Total Social Score ANOVA(b) Model Sum of Square s df Mean Square F Sig. 1 Regression 38.571 5 7.714 32.807.000(a) Residual 72.188 307.235 Total 110.75 9 312 a Predictors: (Constant), Level of Education, revised, location dummy urban, household members per income earner, Total Social Score b Dependent Variable: ln_dpce Coefficients(a) Unstandardized Coefficients Standardized Coefficients Model B Std. Error Beta t Sig. 1 (Constant) 2.773.211 13.170.000 location dummy urban -.115.056 -.097-2.046.042 household members per -.093.015 -.290-6.045.000 income earner revised -.021.062 -.016 -.347.729 Total Social Score.077.009.489 8.348.000 Level of Education -.012.008 -.081-1.360.175 a Dependent Variable: ln_dpce 15