The Effects of Adult Women Education - Impact Evaluation of a Program in Chiapas

Master's Thesis 2008 77 Pages

Didactics - Business economics, Economic Pedagogy




2.1. Mexico and Chiapas: recent economic trends
2.2. Alternativa Solidaria Chiapas (AlSol): geographical and operating aspects
2.3. Literature review and discussio n of potential education program impacts

3.1. Data collection and group definitions
3.2. Control group adjustment: propensity score matching
3.3. Data and variables’ description

4.1. Women’s income and household’s poverty
4.2. Women's and children's health issues
4.3. Women’s empowerment within the household: decision making & mobility
4.4. Children’s schooling patterns
4.5. Women’s, centre’s and area characteristics’ impact on exam performance





First of all, I would like to thank AlSol and Katia for giving me the opportunity to write the thesis and thus to use the Ngo’s data and infrastructure. A huge thank you to Rosi, Luis and later Mateo for the wonderful time spent in the field, as well as to Juan for its helpfulness in translating into Spanish and his company during the long waiting times and several walking hours. I am very grateful also to Pilar, for her support in the office, and to all community organizers’ patience and understanding when I was carrying out the interviews.

I really appreciate Prof. Eliana La Ferrara acceptance to support me in this undertaking and her patience in following my work around different places in the world, as well as Prof. Martina Björkman advice when preparing the questionnaire.

I feel very lucky when thinking about my parents, who have always supported my ideas, projects and continuous restlessness, and Angela for her sisterly solidarity.

Besides my real family, I would also like to thank my little surrogate family in Junax and all the great people I met during my three months stay in Chiapas, who definitely enriched the thesis experience from a very human point of view.

Without Stefano’s immense moral support, it would have been much harder to overcome some of the temporary discouragements during the difficult and hectic moments of the thesis writing: thank you for allowing me to continue broaden my horizons.

Finally, I cannot forget Erika and Luca for their practical, but indispensable, contributions to the thesis.


illustration not visible in this excerpt


Dominga Mendez Gomez lives in the small community of Arvenza in the municipality of Chamula at approximately 20Km from San Cristóbal de las Casas. The city is located in the central highlands of the Mexican state of Chiapas and is a stronghold of the Tsotsils, one of the main ethnic groups to have descended from the Mayans. Dominga is 29 years old, married to a farmer and mother of three children, two of which are currently enrolled in primary school, whereas the youngest is still at home. She has been living in the same community all her life and when she got married she moved into her husband’s house, which she is currently sharing with her small family and her parents in law. During a typical day she works on her textile craftworks, producing blouses, skirts, shawls and pillowcases, prepares tortilla and beans, looks after her children and cleans the house, made out of dried mud with a plate roof. Even though Dominga attempted to attend primary school twice, she never managed to stay in school for longer than a few months, as her parents’ household needed her as additional labor force.

Since 2003 she is member of the microfinance program of the Mexican Ngo, “Alternativa Solidaria Chiapas” (AlSol) and since 2004 she also receives additional schooling after her regular credit meetings. Her microfinance center, called Katishtik, is composed by 14 other women, 13 of which participate in the extra-education. The loan she receives from the Ngo currently amounts to 6.000 Pesos (~ 400 Euro) and is used for the purchase of the material and threads she needs for her business activity. Thanks to AlSol's support she could expand her business and increase the textile production as she disposes of the initial capital for the raw materials purchase. Moreover, after enrolling in the first level of the education program, within three years time and after passing two examinations, she has attained the third and last level. She is very proud of her achievement as the additio nal education allows her to better calculate her expenditures and sales, as well as to improve her social and economic participation in the outlet markets of Chamula’s municipal capital and, during the tourist period, in San Cristóbal de las Casas.

Dominga Mendez Gómez in front of the microfinance center “Katishtik”- 2004

illustration not visible in this excerpt

Dominga is a typical example of the 50.1% illiterate, mainly indigenous, women in Chiapas (INI-CONAPO, 2000), that due to the extreme poverty and the patriarchal society did not have the opportunity to attend school and to become educated. The highland of Chiapas (“los Altos”) where she lives is the area where poverty and illiteracy are the greatest: in some municipalities women illiteracy reaches 70% and school absenteeism is frequent, especially among girls. Also the rates of dropouts and failure are alarmingly high as compared to the national average, which is some 30 percentage points lower (INEGI 2000). Moreover, due to the patriarchal orientation of society and families, Chiapas’ women often remain subjugated and socio-economically disadvantaged. This can imply lacking access to property right, credit and education as well as widespread acceptance of intra-household violence and harassment, especially in rural areas.

“Alternativa Solidaria Chiapas” wants to make up for this situation and besides providing credit, takes advantage of regular microfinance sessions to grant basic education to adult women. While previous literature is however available on the direct effects of adult and women education, no much information exists on education programs carried out together with microfinance projects. A natural question thus arises on the efficiency of the intervention and AlSol’s effective resource allocation. This thesis wants to provide an answer to this question, by shedding light on the programs principal impacts, and, at the same time, attempting to fill the literature gap. In particular, it aims at understanding the specific effects of the Ngo’s education program policies on the health, living, family and empowerment condition of participants. The analysis, besides clarifying the general combined effects of education and microfinance, may therefore help AlSol to learn about its own projects and to better organize future expansions.

The thesis proceeds as follows. In section 2, I provide an overview about Mexico and Chiapas' economics, AlSol’s activities, as well as a literature review on previous women and adults' education impacts studies. Section 3 is devoted to data description and the definition and adjustment of a control group, composed by women that are beneficiaries of the microfinance program, but not of the additional education. In section 4 I conduct the data analysis, trying to identify the education intervention’s effects, as well as the determinants for the programs’ success. Finally section 5 summarizes the main results.


2.1. Mexico and Chiapas: recent economic trends

Mexico is the second largest economy in Latin America and can be considered as being in an intermediate advance stage of development. Since the early 1990s the emerging economy is pursuing policies in line with the majority of all industrialized countries and thus has worked on opening up to international trade, support foreign investment and dismantle regulations. In December 1992 the neoliberal reforms lead to the signature of the North American Free Trade Agreement (NAFTA) with the United States and Canada, which first came into effect in January 1994. The intentions of the free trade agreement were to move the three countries towards a larger degree of specialization and division of labor, by also guaranteeing Mexican producers quota free access to the large US and Canadian consumer markets. Efficiency achievements from increasing the extent of the market and enhancing specialization should have produced rapid growth in Mexican productivity.

The arguments for NAFTA became stronger as a result of Mexico’s economic crisis, so called “peso crisis”, in late 1994 and throughout 1995. Indeed, in the wake of the crisis, foreign investors find Mexico less attractive; thus the guarantee of Mexican access to the U.S. market is at that time considered to be indispensable in helping the country to attract foreign capital to boost its productivity and its standard of living.

Since the beginning of the 90ties, Mexican real GDP has grown at around 3.8% per year, and exports have boomed, going from 10% of GDP in 1990 and 17% of GDP in 1999 to 32% of GDP in 2007. But this rate of growth, when combined with Mexico’s 2.2% average annual fertility rates, means that mean market income from production in Mexico is scarcely 15% above that of the pre-NAFTA situation[1]. This means that the gap between Mexicans mean income and that of the United States has widened and despite the, in the past 10 years, steadily increasing GDP and GNI per capita based on purchasing power parity, the uneven distribution of income remains an issue of concern. Indeed, while some segments of the population, such as big exporters, have benefitted from the market opening, especially Mexicans rural population, cannot compete with the US machinery, infrastructure and finally, in price. The substantial agricultural subsidies paid regularly to the US agricultural industry make Mexico’s competitive position even more difficult.

Especially negatively affected by the increasing inequalities are the southern states, where in particular the areas of Chiapas, Guerrero and Oaxaca currently lag behind the rest of the country on almost all socio-economic indicators (Figure A in the appendix illustrates the regio ns concerned). In a human-development index created by the UNDP and comparing Mexican municipalities, 95% of the places in the bottom decile are in the south. In the north, 12% of people in rural areas are extremely poor, against 47% in the south, according to the Woodrow Wilson Centre. The gap was widened especially with the American Economies’ slowdown in 2001, where Mexico, inextricably linked to its neighbors’ economy, has suffered two years of stagnation. Growth has since returned in the north, whereas the southern regio ns still struggle with several economic deficiencies. The present Gini coefficient of income inequalities amounts to 0.46, representing the highest figure within all OECD countries.

The disparities between the northern areas and the southern, mainly indigenous, regions articulates into a very low average number of years of schooling, with young workers having almost no more formal education than their older counterparts (approximately 6 years of education in the south as compared to 8,1 and 9,7 respectively in the north and Mexico City), huge segments of the population with no registered regular income or incomes below the monthly minimum wage (65% of the population in Oaxaca or Guerrero and 80% in Chiapas, as compared to 18% in Mexico City) and several southern regio ns’ households lacking access to hygienic living conditio ns[2] (INEGI 2000).

As a response to the government’s negligence and the negative effects of trade liberalization on Mexican’s southern regio ns, since the early 1990s, population uprisings have become common. In particular, in 1994 the Zapatista Army of National Liberation (EZLN) emerges in Chiapas, giving rise to the still very popular Zapatista movement. The mostly indigenous social base of the group claims the control over own resources and rights on the land on which they live and thus aims at setting an end to their secular marginalization and social exclusion, accentuated during the ‘90s as a consequences of governments’ neo-liberal policies.

The Mexican government has never accepted the movements’ autonomy claims and, in the past 15 years, consistently increased army presence in the concerned regio ns. Nevertheless, to oppose and reduce the influence of the upraising social movements, as well as to respond to the, newly reconsidered, poverty, it engages for the first time in an overall strategy of poverty alleviation in Mexico. Since August 1997, a new anti-poverty program, named “Progresa”, began its operation in conjunction with other programs aiming at developing employment and income potentials (Temporary Employment Program, PET) and at facilitating the creation of physical capital (State and Municipal Social Infrastructure Fund, FAIS). “Progresa”, through offering transfers to poor families conditional on their participation in health and nutrition programs, along with incentives to promote children’s school attendance, aimed at dise ntangling the web of poverty in Mexican’s least developed regio ns. At the end of 1999, “Progresa” covered approximately 2.6 million families or about 40% of all rural families.

The program survived the government change in 2002 from the Zedillo to the Fox administration with only a name change from “Progresa” to the current “Oportunidades” program. Since this transformation moreover, every Mexican informal-sector worker not covered by the formal sector services of the Mexican Institute for Social Security (IMSS) or the Institute of Security and Social Services for State Workers (ISSSTE), has the right to receive free health care from an array of government agencies, under the condition of requiring it at the IMSS establishment. The relatively new, and so called “IMSS- Oportunidades” social insurance system has a basic range of 151 medicaments and vaccines to attend the most frequent illnesses, whereas in more serious cases the patient may refer, without any additional costs, to a so called “Rural Hospital” (a basic range of 298 medicaments, treatments and vaccines is offered)[3].

In this political context and, although the anti-poverty program has shown positive effects in a variety of impact studies, the situation in the three Mexican States has yet not considerably changed. Further social uprisings, such as in Oaxaca in 2006, are sign of ongoing social discontent and marginalization and the land ownership controversies are far from being solved.

Since 2001 the Mexican government formally initiated a new development plan, the so called “Puebla-Panama Plan” which intends to “promote the regio nal integration and development”[4] of the nine southern states of Mexico with the Central American countries until Panama. The plan has the aim to enhance investment and trade in the region by building or improving large infrastructure project and thus to establish the foundations for a Free Trade Area of the Americas (FTAA). The plan has encountered many criticisms and fears related to the sustainable development of the concerned areas, the jeopardizing of the biodiversity, the potential poverty deepening of certain segments of the population and the national debt increase. The realization of the plan is currently ongoing, but the short and long run effects on Mexican’s southern states’ population remain controversial.

Finally the free trade policies pursued by the Mexican government since the beginning of the ‘90ties led to an increase in the number and significance of Ngo’s in the country (INEGI 2000). The fear of part of the society, not dealing with to the export business, but involved in organized labor, small businesses, agriculture or government administration led to a proliferation of new organizations. Their increasing importance can also be attributed to two other main reasons. The greater international scrutiny of Mexican affairs, due to the market opening, led the Mexican government to acknowledge the role of Ngo’s in desire to be perceived as a modem democracy. In addition, many Ngo’s were inspired by NAFTA to seek new partnerships with organizations in the United States.

The mission of the majority of these organizations is to fill the existing gap between the disadvantaged and privileged Mexican people, as well as to reduce the persistent, and even increasing, inequalities in the country.

In the wake of this proliferation also the Mexican Ngo Alternativa Solidaria Chiapas (AlSol) was founded in 1999. The subsequent section will give an overview over the Ngo’s mission, vision and main activities.

2.2. Alternativa Solidaria Chiapas (AlSol): geographical and operating aspects

In this section I will first give a brief overview about Alternativa Solidaria Chiapas, its main activities and recent trends, while then a more detailed analysis will be provided of the Ngo’s social services, this impact evaluation is on.

Ngo background and main activity

AlSol is a Mexican non-governmental organization in Chiapas that began its activities in the politically and economically troubled context of 1999 with the mission to contribute to poverty reduction in the marginalized rural and semi-urban areas. The Ngo's main activity is the granting of micro-financial services to women of scarce resources, through the organization of solidarity groups that meet weekly or fortnightly in their communities. Each group, characterized by joint liability, is composed by a minimum of 3 to a maximum of 7 women and, in order to receive a loan from AlSol, needs to be integrated into a centre, composed by a minimum number of 10 women, i.e. two solidarity groups. Whereas each group is headed by a president, each centre also has a representative, which can be considered as the main intermediary between the group and AlSol’s community organizers. The minimum age for entering AlSol is 14 and the amount of the loan offered to the clients generally begins with 1,000 or 2,000 Mexican Pesos (~65 or 130 Euro respectively) depending on the productive activity, e.g. production and sales of textile handicrafts, animal breeding projects, sales of food products, clothes, cosmetics etc. Annual interest rates depend on the type of area (rural or urban) and the duration of the loan and vary between 30 and 40%. In order to reduce the probability of uncovered debt in case of death among the clients, AlSol moreover offers a life insurance contract, which corresponds to 20% of the total loan amount. In case of a client’s death the debt is condoned and 10.000 Pesos are paid in favor of the family members or other beneficiaries designated by the client; otherwise the insurance premium is devolved when the loan is completely paid back.

In order to achieve a comprehensive development of the attended rural women AlSol believes that an integration of the financial projects with social programs is necessary. Since 2003, besides offering microfinance and insurance services, AlSol is hence involved in the realization of two non-financial programs concerning women's literacy and health education, the latter of which has experienced several changes from its first implementation. However, due to a lack of resources the social services cannot be offered in all microfinance municipalities and a selection of the “most needy” areas is made.

The below figure represents in blue the 61 municipalities in Chiapas where AlSol operated in July 2008 and in green the 3 municipalities in which, besides the microfinance, also the education program is implemented. In particular for the education program, AlSol distinguishes by intervention area and not municipality, wherefore the municipality of Zinancantan is subdivided into 3 smaller areas, for a total of 5 attended education program areas in July 2008 (N.B. only the municipal capital is named according to the municipality name).

Figure 1: Areas in Chiapas attended by AlSol’s micro finance program (in blue) and by both the microfinance and the education program (in green)

illustration not visible in this excerpt

In the past two year, i.e. within 2006-2008, AlSol has experienced a consistent growth, expanding from 2 to 6 the total number of microfinance branches. Also the number of active borrowers has doubled, reaching, at 31 December 2007, a total number of 16.047 active women for a gross loan portfolio of 3.169.488 USD[5].

Since March 2008 the Ngo is however experiencing some operative problems due to armed attacks by an organized group that is explicitly targeting the microfinance community organizers in two microfinance attended areas (one of which is also attended by the education training program) and has the aim of stealing the microfinance money amounts. AlSol is trying to cope with this problem, by developing new strategies that do not consistently affect the attended women and centers, but ensure at the same time the staffs own safety. As a first step the social projects were closed in the dangerous areas, whereas different microfinance strategies are tried out before suspending also the microfinance activities. AlSol is thus currently introducing a new payment system through bank cheques instead of cash transfers. For operative homogeneity the new system is introduced in all centers and not only in the affected areas. It entails that the community organizers, as well as the women, are not dealing with the cash amounts anymore but each transaction passes through a bank office in the closest city center. Due to this exogenous shock, and as some attended women might not be willing to adapt to the new system, it is expected that AlSol is going to lose some of its clients. The Ngo is thus currently facing an important transition phase in which the staffs safety becomes an important issue as opposed to future expansion plans.

Literacy program

The literacy program was first implemented in 2003 with the idea to increase women’s basic writing, reading and math skills by taking advantage of the fortnightly microfinance group meetings. In order to start the non-compulsory program in one centre, however at least 10 women must be interested and, when deciding to participate, are asked to stay for one more hour after the usual microfinance session. During this additional time, according to women's initial educational level, literacy classes are given and simple exercises are carried out. Depending on women’s previous schooling, AlSol classifies the women into three different levels, where the lowest level forsees the acquirement of basic math as well as simple words’ reading and writing skills, in level two the women are required to increase the complexity of both sentences and calculus and in level 3 short storie and tales are written and discussed with the literacy trainer. So as to provide an education apt to conserve women’s original culture and habits, classes are given both in Spanish and in Tsosil, their native indigeneous language. Moreover, in order to make the women feel more comfortable with the treated subjects, classes generally refer to daily living situations, so that the first words and sentences learned (both to say in Spanish and to read and spell in Spanish and Tsotsil) often concern the house, the market or their business, while the simple stories are usually on daily events and activities. Every six months the students have the possibility to proove their knowledge and to pass from one level to another as a voluntary exam is administred. On average a 60% of all women decides to take the exam, 70% of whom passes from one level to the next at each exam session.

Nutrition - Health education program

The health segment of the AlSol’s non financial service area has experienced an important mo dification since its first introduction in January 2005. The original idea of the program was to take advantage of the fortnightly microfinance sessions to deal with undernourishment and malnutrition of AlSol clients’ children below age 5. It was hence tried to improve their health conditio ns and to prevent diseases caused by poor calories’ intake through the regular granting of food provisions, destined to the preparation of a “baby pap”. The main ingredients given out to the women consisted of maize flour, milk, oil and sugar, while regular weight and height controls selected the children that needed to be attended as well as monitored the programs’ effects. Unfortunately however, the expected impacts on the approximately 300 ever involved children were never attained and it seems that the misuse of the grants was a common phenomenon. After having tried the program in several areas and municipalities, in October 2007 AlSol decided to change the project approach. The new project’s scheme since November 2007 thus aims at improving children’s, as well as women’s own health conditio ns through regular health education classes. The classes consist in schooling on basic hygiene, birth control, common disease recognition and both, herbal and modern medicine, remedies.

In this way AlSol wants to both prevent the women and their family members from contracting some illnesses and provide them with the instruments for coping with potential diseases. As regards the effect on child malnutrition, it is expected that the health education increases women’s knowledge and thus indirectly helps to improve both nourishment quantity and quality (respectively through an income effect and a better awareness of highly nutritive food intakes). In severe cases of child malnutrition the educational trainer however still refers the women to the closest health centre.

The initial nutrition program was implemented in the same municipalities as the literacy program, although not in the same areas and microfinance centers. With the transformation of the nutrition into a health education program, the project was however completely integrated to the literacy classes with regard to areas, centers and targeted women: those women that were already receiving the literacy also started getting health education, while the program was totally suspended for those women that were only receiving the nutrition grants.

In this sense, the previously separated literacy and nutrition were consolidated into a unique education program in November 2007, with the aim of providing basic schooling in math, reading, writing and health knowledge. The health and literacy classes alternatively take place after each microfinance session.

The new integrated education program operates in the same areas and centres as the previous literacy and thus attends 3 municipalities, subdivided into the 5 areas of Chamula, Larrainzar, Elambo, Nachij and Zinancantan, all depending on the San Cristobal headquarter. Like the microfinance, however, also this program was affected by the recent armed attacks and, for safety reasons, in several centres the service had to be suspended (even if the microfinance service is still going on through the “bank cheque system”). While in December 2007 the centres attended by the integrated education program were 19 for a total number of 381 women, in May of the subsequent year the number of centres was reduced to 9 for a total number of approximately 162 women. Just as the financial area, also AlSol’s social service area, is thus passing through an important transition in which, due to the external shock, the area of intervention needs to be replanified and overviewed for the own staffs safety. At present, meanwhile that the safety situation potentially improves, new centres and areas are identified so as to re-increase the number of attended women.

My own evaluation of the educational program (hereafter also defined as “training”) is based on those centres and women for which the integrated schooling service has not been suspended. Depending on the time they are in the program they first attended the literacy without the health education component or entered the program when the integration had already been done. A splitting of both services would however not be possible because of their close collaboration in the 8 months previous to the evaluation, as well as the relatively small sample size available for the analysis.

2.3. Literature review and discussion of potential education program impacts

Women's education has long been recognized having a crucial role as a development strategy and several impacts on women’s wellbeing, as well as households’ living conditio ns have been empirically proven.

This section aims at reviewing the existing evidence on the effects of women education and thus to motivate the analysis that is going to be carried out in the subsequent sections of this thesis. In particular, an impact of women’s additional schooling has been assessed in relation to four main outcomes: women’s income and poverty; women’s and family members’ health conditions, women’s empowerment and children’s educational behavior. Although also a casual linkage with fertility has been empirically assessed, no reference is here going to be provided, as the effects are of long term entity. A potential impact may therefore not yet be detectable for AlSol’s, relatively recent, education program.

Each of the four previously mentioned evidences is now going to be discussed in turn.

Women ’s income and household poverty

It has been argued that one of the most important benefits of education is that it improves individuals’ ability to obtain and process information (Rosenzweig 1995). A better educated merchant with the ability to read, write and calculate, might, for example, achieve a higher profit because he is able to find and elaborate information on market prices, demand and production process more efficiently. Several studies have established the positive impact of higher education on economic outcomes, such as wages and incomes and indirectly - through higher women’s profits and/or women’s increased bargaining power within the household - on women’s household poverty (Glewwe 1999a; Jolliffe, 1997).

Women’s average incomes are usually collected through individual’s own statements. For the measurement of household poverty the Grameen Foundation propose the use of a so called Progress out of Poverty Index (PPI). The index has been specifically developed for those countries where its microfinance partners, such as AlSol, are located and aims at enabling institutions to understand the likely poverty levels of their clients with respect to the national poverty line and how these poverty levels change over time (Schreiner, 1996). It can thus be considered as an income approximation measure, trying to capture the proportion of microfinance institution clients who are likely to be “very poor”, “moderately poor” and “above the poverty line”. AlSol started to use the index in 2006 and through an interview consistent of 10 short questions wants to measure and understand the poverty of its clients when entering and exiting the microfinance program. Table В in the appendix marks the correspondence between score and probabilities of being poor (Schreiner, 2006, from ENIGH 2002 data). The association is generated from the analysis of the living conditions of the households included in the national survey, allowing thus for the drawing of poverty lines and the linkage of each score to a poverty probability.

In a population, such as the one in which AlSol operates, where many adults and especially women are not able to read and write, an adult women education program may be an effective way of increasing personal incomes, as well as, on a longer term, households living condition. Given that one of the main risks faced by microfinance activities is the poor investment of the clients’ loan, the combination of the microfinance program together with the education program can moreover show interesting outcomes. Although women receive the credit оnly for particular investments plans in their own businesses, without basic numeracy, reading and writing skills, the initial capital is in fact hardly ever used in a productive manner. Women’s contemporaneous access to initial financial resources as well as to the tools to efficiently employ and invest such capital could thus reasonably have an important impact on returns and loan repayments.

Finally, while a potential effect on income would catch the short time impact of the education program, an effect on the PPI score would get its longer run influence.

Women’s and children's illnesses and confidence in modern health posts Researchers have long known of the link between literacy and health. Along with income, employment, working conditio ns and the social environment, literacy and education are, in fact, commonly considered key determinants of individual’s health conditions. Household survey evidences confirm these relations, by establishing important links between general education, literacy and numeracy on non economic outcomes such as health (Glewwe 1999b; Glewwe and Desai 1999). By giving people the skills needed to get, use and understand information to make good decisions for health, individuals’ own health condition are demonstrated to improve for more educated individuals. The same is true also for the children’s health status, for which a large variety of evidences from developing and non­developing countries find that mother’s education is positively correlated with better children’s health conditions and nutritional status (Glewwe P., 1997; World Development Report, World Bank, 1993).

Health status can however be difficult to measure through the household survey instrument (Me Dowell and Newell, 1996). In fact, a person’s own judgments on its own or its children’s health condition might be biased by different wellbeing perceptions of more educated or perhaps wealthier persons (Gertler, Rose and Glewwe, 2000). Following Gertler’s, Rose’s and Glewwe’s reasoning, the potential bias might originate from the respondent’s answers that tend to depend on their subjective opinion of what needs to be considered as being “healthy” as well as their personal contacts with modern health institutions. An ideal survey would try to compile information from individual health visits through the involvement of doctors and nurses, by obtaining in this way a more objective vision of the persons’ health condition. At the same time it would however require considerable financial resources as well as a consistent logistical organization, both unfortunately not available for the following impact analysis. Another way to make up for this problem is to resort to relatively “observable” illnesses, which symptoms are supposed to be less influenced by different, education-related, perceptions.

Finally, while the first part of the health-related section of this thesis is concerned with individual and children’s illnesses as well as general wellbeing, a second part tries to assess whether there potentially exists an impact on the use of health facilities. Traditionally the indigenous population in Chiapas shows quite a big mistrust towards modem medicine health post, preferring to resort to traditional healers or herbalists. AlSol’s health education part of the training program, which includes both traditio nal, and mainly herbal, medical teaching as well as basic modern medicine schooling, could bring the women closer to modem health remedial. The literacy part of the program on the other hand could have the effect of making women more confident about themselves and hence of reducing fears in modem, and often urban, institutions. Previous evidences confirming this theory can be found in Habtom & Ruys, 2006 .

Women’s empowerment within the household: decision making power & mobility There is growing evidence that the household cannot be characterized as one where individuals share the same preferences or pool their resources. Several researches show that the unitary mo del of the household can be rejected in country sceneries in both developed and developing countries (Behrman, 1997). Although the unitary model continues to stay extremely powerful and widely used, evidences in favor of a model in which individuals within the household have different preferences and maintain control over their own resources have been recently developed. According to this model, factors such as individual financial resources, as well as personal skills, knowledge and education can influence the decision making between household members and hence the, so called, “bargaining process” between individuals with different preferences. While some of these influences, such as one’s family status within a certain cultural environment or national legal rights, are external to the individual, others can be modified by the persons’ behavior or individual belongings.

One way to increase a persons’ bargaining power may be through the membership in organizations and social networks. We can hence expect that already the participation to AlSol’s microfinance program, which allows women to deal with individual financial resources and requires them to regularly leave their homes, increases women’s decisional power within the household, as well as self-esteem and confidence. An additional enhancement could be achieved through the combination of the microfinance with an improvement of the women’s education, which in developing countries is generally lower than husband’s schooling and thus represents a disadvantage in terms of intra-household bargaining power. By deliberately targeting women’s education, AlSol’s training program could therefore have the potential for affecting women’s decisional power and one of the purposes of this thesis is to ascertain this impact.

Children's school enrollment

The education of children in developing countries is widely considered to be one of the most effective ways of enhancing growth and reducing poverty. Numerous empirical studies from both developing and non developing countries show how parents’ education often has a positive and robust effect on children’s schooling behavior. It is found that more educated parents make greater educational investments both in terms of quantity and quality, and that the marginal effect of mother’s education on children’s schooling is generally larger than that of the father’s education (Sathar and Lloyd, 1993; Brown, 2006). Other studies (Herz and Khandker, 1991) also show how an increase in physical and human resources belonging respectively to one of the two parents can have different effects on the human and physical capital investments in male and female children. The unitary mo del seems hence once again rejected and mother’s and father’s child gender preferences assume a larger importance. Alsol’s women’s education program could therefore show a positive impact on children’s primary and secondary school enrollment, by potentially also highlighting mother’s gender preferences.


3.1 Data collection and group definitions

The data I analyze for the evaluation of the education program are retrieved 5 years after the first implementation of the literacy project and 9 months after the integration of the two non- financial services. I collected these data during a 3 months field work between May and July 2008.

So as to meet the education program evaluation objectives I conducted interviews of approximately 20 minutes on women's characteristics and living conditio ns. Interviews were carried out either in the local indigenous language, Tsotsil, by my external translator or the educational trainer, or by me in Spanish in approximately 30% of the cases. The questionnaire is subdivided into 5 main sections, where section 1 is on women's household and marital status characteristic and section 2 captures their children's schooling and health conditions. Section 3 and 4 concern women's health status and both, their own and their families', economic situation, while the final part tries to identify women's and husband's decision making within the household and thus only looks at women living in a stable relationship.

Before starting with the implementation of the survey I tried to adjust the questionnaire to the local context, by testing the survey on 9 “independent” women, i.e. which belong to microfinance centers that are not successively interviewed and used for evaluation. After this first test phase the control and treatment localities are identified by relying both on AlSol's staff members’ statements as well as on internal documentation on attended centers and opening procedures.

The choice of the treated centers is mainly related to the recent problems in the areas, i.e. all centers that have not been affected by the external armed attack shock are included in the evaluation. As the control groups are chosen in the same, non-affected, areas, no concerns about non-random selection should arise. Roughly 89%[6] of all women of a treated centre participate in the voluntary education classes, although they may not be all interviewed because of a time constraint. Considering thus the time limits, women were interviewed according to their microfinance group belonging and called in a descending order; although the interview was not compulsory they hardly ever refused to collaborate. As each centre was visited at least twice (generally 2-4 visits per centre, according to its size) I am moreover able to rule out big part of the potential self-selection into the interview, due to some women’s microfinance meetings’ absenteeism. Problems arising from the self-selection into the program on the other hand, will be ruled out by a matching procedure among treated and control women. The next section will outline its implementation.

The control group choice is more delicate, as it requires finding an equal amount of women in a similar area and context of that of the treatment group but that have not been exposed to the educational classes and that hence differ from the treated group only for the non participation to the program. In the absence of the education program the two groups are theoretically identical and supposed to show equal evolutional trends. The research of such a group needs to be based on AlSol’s procedure of offering the education program in a new centre. Reminding that the health education program was integrated into the already existing literacy program and established centers, I tried to reconstruct the evaluation that had been made by AlSol in 2003 and 2005, when the preponderance of all centers first received the additional literacy service. As the big majority of AlSol’s staff has not changed since then, I can both rely on the personnel’s recall, as well as some written lists and notes on potential target centers. All evidence seems to indicate that the selection of the centers was first based upon observable literacy characteristics and then on a final random choice, due to the Ngo’s lacking resources. Being at that time literacy the only component of the educational classes, it is reasonable to believe that selection has first been made on women’s initial literacy level, i.e. their highest achieved education, and their living remoteness in terms of schooling distances. Data for the municipalities’ choice derive from the “National Institute of Statistics and Geography” (INEGI). Figure В in the supplementary appendix confirms the selection process, by showing the literacy and school attendance levels of the municipalities that AlSol was attending in 2005 (Chamula and Zinancantan are placed far above the state average in terms of highest percentage of illiterate population, whereas the municipality of Larrainzar is in the average, but has a relatively larger population than the majority of other municipalities). Within each chosen municipality AlSol then selects those centers that, according to the Ngo's data collection, show the lowest literacy levels as well as least access to primary and secondary schools. As compared to the information on education the access data were less systematically retrieved and often entrusted to the individual judgment of the literacy project coordinator.

Because of the Ngo’s limited resources and the presence, at that time, of one single person, contemporaneously in charge of the literacy coordination and the literacy classes on the field, not all eligible centers can be however attended and a second, this time random, selection is carried out. Historical documentations, as well as staff members’ recall, allow me to reconstruct such final selection and hence to create a list of all theoretically eligible - and “needy” - centers, that have been excluded from the training program because of lacking resources. Within this list I make a further selection of those centers that are nearer to the treated centers in terms of geographic location and are thus supposed to be closer to the treated centers also in terms of traditions, customs and unobservable characteristics. Women in the control groups are again interviewed according to group belonging and thus called in descending order.

Table 1 graphically resumes AlSol’s and myselfs selection process procedure for the finding of the control and treatment group for the purpose of this study.

Table 1: Selection process for finding the treatment and control groups

illustration not visible in this excerpt

In order to rule out any outstanding observable differences between the treated and the control group an additional adjustment of the control group is necessary. The following section describes the adopted matching procedure adjustment.

3.2. Control group adjustment: Propensity Score Matching

AlSol’s only partly-random selection process is confirmed by a test of difference in means across control and treatment groups for all available pre-treatment variables[7], for which we must reject the null hypothesis of no difference between the treatment and the control group. Panel A of table 2 shows the results of the ttest, where the differences in the distance measures must be mainly attributed to the control group locations in the Larrainzar municipality, relatively close to the municipal capital as compared to the treated centers.

In order to reduce the existing differences between the treated and the control group, I resort to a matching approach that has become commonly used to estimate causal treatment effects. The idea of the matching is to find, in a group of non participants, those individuals who are similar to the treated individuals in all relevant pre-treatment characteristics X. That being done, differences in outcomes between the participants and the adequately selected control group are more likely to be attributed to AlSol’s education program, as a situation near to complete randomness is reproduced.

In order to successfully implement the matching procedure, Rosenbaum and Rubin (1983b) suggest the use of a so called balancing scores b(X), i.e. functions of the pertinent observed explanatory variables X such that the conditio nal distribution of X given b(X) is independent of treatment participation. For the purpose of my evaluation I will rely on a particular balancing score, commonly referred to as “propensity score”, i.e. probability of participating in a program given observed characteristics X.

For the satisfaction of the “unconfoundness assumption”[8], the matching strategy builds on, only variables that simultaneously influence the participation decision and the outcome variables and that are unaffected by participation should be included in the propensity score estimation (Caliendo and Kopeining, 2005). In my analysis I decide to enclose all those variables that plausibly satisfy these assumptions and that contemporaneously are either constant over time (distance measures) or have been retrieved by AlSol previously to the beginning of the education classes or the anticipation of the program (women’s education). The existing literature on propensity score matching uses the probit estimation method to run the relevant covariates on the probability of belonging to the treated group (Caliendo, Kopeining, 2005). By following this approach, I empirically confirm AlSol's selection process, as a positive and significant impact of lower education and higher distance to secondary school is assessed. Table 3 shows the results of such probit estimation on the probability of belonging to the treated group. For the ease of interpretation I report the marginal effects evaluated at the mean, i.e. the change in the probability for an infinitesimal change in each independent, continuous variable and the discrete change in the probability for dummy variables.

On the basis of women’s specific covariates a propensity score is associated to each interviewed woman. Then, depending of the obtained score, each of the 104 treated units, identified with i, is matched to one non-treated unit j in a one to one nearest neighbor matching procedure without replacement[9], such that:

illustration not visible in this excerpt

with DG {0,1} being an indicator of actually received treatment Of the 241 initial observations 33, belonging to the control group, remain unmatched and are thus dropped from the sample.

Quality of the matching procedure

The quality of the matching is assessed through a comparison of the situation before and after matching, by hence checking whether there remain significant differences after conditioning on the propensity score.

As a first approach I resort again to a test of difference in means across control and treatment groups for individual variables. I thus check whether, after selecting observations within the control group, there persist significant differences in covariate means for both groups, and especially for what concerns the variables of “education” and “access to education” on which AlSol performs the selection into the literacy program. Comparing panel (A) and panel (B) in Table 2 we notice an improvement in all differences across individual variables, even if not a complete disappearance. While after the matching no significant difference is perceived anymore between the education levels and women’s average distance to the municipal capital across treatment and control groups, differences in distances to the secondary school and to the clinic remain significant at a 5% level.

I therefore decide to perform another quality assessment test, which consists in re-estimating the propensity score on the matched sample of 208 observations, comparing it to the pre­matching probit regression results. As compared to the pre-program probit, we see that the significant impact of the main covariates indicating education and the distance to secondary school now disappear, even if both distance to municipal capital and distance to clinic stay significant at a 5% level. Even if, from the ttest results, the distance to secondary school appairs to be still consistently different between the treated and the control group, it does not however seem to significantly impact the probability of assignment into treatment. Finally, also the fairly low pseudo-R2 of the second post-matching probit seems to indicate that any systematic differences in the distribution of the covariates between the treatment and the control group are reduced.

The remaining differences are probably due to the difficulties in finding appropriate control groups in the area of Larrainzar, where no non-treated microfinance centers could be found outside of the municipal capital.

We can conclude that, even if the observable differences between the treatment and the control group are not completely eliminated, their determinacy for selection into treatment is consistently lessend, especially for the main “selection variables” of previous education and schooling distances. In order to however take persisting differences into account, I will control for the concerned distance measures in the analyses of the subsequent sections.

illustration not visible in this excerpt

3.3. Data and variables’ description

After the matching the main dataset contains observations on 208 individuals, subdivided equally into treatment and control groups and respectively 9 treated and 10 control centers. In each centre between 3 and 21 women are interviewed, with an average of 11 interviews per centre. As all interviewed women have on average 2.95 children I obtain a children dataset containing 429 observations, 229 of which in the treated and 200 in the control group.

A third dataset concerns a sample of women that was in AlSol’s microfinance program at the moment of one of the literacy exams in July 2007[10]. In particular, it regards basic information on women involved in the literacy classes in one of the centers in which an exam was taken, independently of their participation to the examination. Moreover, for those women that took the exam, data on passing the exams and marks are retrieved. The dataset contains informatio n on 33 variables and 288 observations, 105 of which participate in the exam.

3.3.1. Dependent variables’ description

By relying on previous evidences described in section 2.3 and by considering the available datasets, the thesis tries to provide an impact assessment of AlSol's education program on several issues that concern the women involved in the program and their children. Specifically, the aim is to understand how the program affects i) women's income and household wellbeing approximation ii) women's and children's health issues iii) household decision making and the consequent empowerment of the women iv) children's schooling behavior v) women's performance in the program depending on individual and microcredit group characteristics. Following this objective, in this section I will picture the characteristics of the dependent variables used for each evaluation.

Income and household poverty

The dataset contains both monetary income data and income-approximation measures. While the former include women’s weekly average income, other weekly family incomes, government benefits and remittances, the latter are obtained through information on dwelling materials and characteristics, household items and cooking habits. Unfortunately, no historical information on training participants’ microfinance loan defaults is available and hence, even if being of significant interest, no analysis of treatment participation impact on loan repayments can be carried out. I thus restrict the study to a comparison of individual incomes and household poverty of women involved and not involved in the education program. For this purpose woman’s individual income and household’s, income approximating, Progress out of Poverty Index (PPI) score are used as dependent variables, while the other income- related information is not supposed to be influenced by treatment participation and used as a control for potential heterogeneities between the treated and the control group.

For what concerns the PPI, the entrance poverty measure is not available for all women in the treated and control group, as several women in the dataset entered AlSol previously to the first administration of the PPI-related interview. A baseline is hence unfortunately not available and I am obliged to restrict the impact study to a cross-sectional analysis. Because of a time constraint, I moreover decide to slightly readapt the poverty score developed by the Grameen Foundation for Mexico’s rural areas. In the original PPI 10 basic questions about household living conditions and habits are addressed and different scores are assigned to each answer. In my own adaptation I decide to exclude two questions on past months shoes and lothing purchases, reducing to 8 the questions used for the PPI score calculation. A subsequent correction for the missing values allows me to obtain final scores that are comparable to the ones obtained through the 10 question measure. Table A in the appendix reports the score card of the personalized index used for the calculation of the PPI.

Figure 2 shows the adapted individual scores obtained for women in respectively the treatment and control group. The average - corrected for the missing values - PPI score for all interviewed women is 31 which, according to table В in the appendix, corresponds to an approximately 80% probability of being classified as poor

Finally, the average weekly income is of 212 Pesos (~ 14 Euro), ranging between a minimum of 25 and a maximum of 600 Pesos. Only small variations from the mean are observable for the Chamula and Elambó areas, where the weekly average income is approximately 20 Pesos lower, while for Nachij it is roughly 30 Pesos higher. Also depending on the marital status of the women it is approximately 35 Pesos higher than the sample income average for separated or divorced women and lowered of the same amount for widowed women.

Figure 2: Progress out of Poverty index scores for treated (train=l) and control groups (train=0)

illustration not visible in this excerpt

Women’s and children's illnesses and confidence in modern health posts Being aware of the previously outlined limit of a potential bias in people’s own wellbeing perception, I divide the health-related impact study into three subsections. A first part of my survey addresses questions on a woman’s own and her children’s perceived health conditio ns, i.e. the type of illness both herself and her children have had in the 8 weeks previous to the interview. A second part tries to assess the incapability of carrying out daily living activities, whereas a last part identifies women’s individual illness prevention knowledge. In the first analysis I try to include only relatively observable illnesses, where the symptoms are supposed to be rather easily identified by the respondents and the perception hence less influenced by educational or, perhaps, income characteristics. The second and third part are, on the other hand, supposed to be less subjective than other self-reported health measures as they do not require respondents to provide general opinio ns about their own health (Gertler, Rose and Glewwe, 2000). In particular, even if “illness knowledge” does not directly describe women’s or children’s health status it is supposed to have a positive repercussion on individual’s physical condition and can hence be considered as being an indicator of infirmity in the long run.

In order to assess the women’s health I therefore rely on four main dependent variables concerning women’s knowledge on how to alternatively avoid two frequent illnesses (diarrhea and skin diseases) and the amount of illnesses, as well as the number of days a woman was unable to perform her daily activities in the 8 weeks previous to the interview (e.g., dedicating herself to her own business and caring about the children). As 97% of all women are involved in the same type of textile handicraft business, daily activities should not be considerably different from one woman to another and the potential bias due to dissimilar difficulties of daily activities - and hence different required health status for performing such activities- should be relatively small. Moreover, concerning the knowledge questions, different alternatives were given to the respondents, which however, for the purpose of the evaluation are resumed into “does know” and “does not know” and hence generate the two dichotonomous “knowledge” variables. Finally, the total number of women’s illnesses range from 0 to 4 and are the result of the sum of declared individual illnesses, such as diarrhea, cough, stomach ache, back ache, bone ache, head ache, fever, skin disease and flu. Because of the small sample size, unfortunately no specific illness incidences may be evaluated.

Figure 3 shows the frequencies of illnesses and days incapable of performing daily activities, whereas figure 4 shows the trends on knowledge about illness prevention by training participation and the answers given are reclassified into “aware” and “not aware”.

Figure 3:

Frequencies in women’s illness and capability of performing daily activities

illustration not visible in this excerpt


[1] www.worldbank.org/mx

[2] For example in Chiapas, 66% of all households lack draining systems access, respectively 34% and 21% have no water and electricity and barely 68% live in earth floor houses

[3] http ://www.imss. gob. mx/programas/oportunidades

[4] www.planpuebla-panama.org

[5] http://www.mixmarket.org/en/demand/demand.show.profile.asp?ett=2709#

[6] Own calculation from “Manual de alfabetización”, AlSol; even if only 10 women are needed in order to offer the training service, approximately all women within a centre usually participate

[7] Data on age and education are retrieved from AlSol’s database that captures the information at each woman’s entrance to the microfinance project. Distance measures are part of the cross-sectional questionnaire but, because of low migration rates within communities, are supposed to remain approximately constant over the three year’s time lap between the interview and the beginning of the literacy program

[8] Rosenbaum and Rubin, 1983b: Given a set of observable covariates X which are not affected by treatment, potential outcomes are independent of treatment assignment

[9] A matching procedure with replacement has been tried but too many observations are dropped and, considering the relatively small sample size, too much information is lost.

[10] Because of the commotions in the education program area the regular December literacy exam was not administered. The last available exam date thus date back to July 2007


ISBN (eBook)
ISBN (Book)
File size
4.6 MB
Catalog Number
Institution / College
Università degli Studi di Milano
Effects Adult Women Education Impact Evaluation Program Chiapas



Title: The Effects of Adult Women Education - Impact Evaluation of a Program in Chiapas