THE IMPACT OF BUSINESS TRAINING ON BUSINESS
OUTCOMES: A RANDOMIZED CONTROL TRIAL ON
FEMALE MICROFINANCE CLIENTS IN VIETNAM
Rosine van VelzenUniversity of Groningen Faculty of Economics and Business
Supervisor: prof. dr. B. W. Lensink July 2014
Abstract
Using data of a randomized control trial I estimate the impact of a business training on business outcomes for female microfinance clients in Vietnam. Difference-in-differences estimation for a balanced panel shows that a gender and business training of the International Labour Organization improves women’s business knowledge, business practices and increases profitability of business activities. Second, inviting husbands to join training modules does not affect the impact of training on business knowledge, business practices and business results. Third, I find a significant improvement in business practices between six and twelve months, which suggest that training has a lasting impact and underlines the need of multiple surveys to fully determine the effectiveness of training. This study contributes to other literature by studying the impact of training for women with and without participating husbands. In addition, data of two follow-up surveys by 4041 women introduces the possibility to estimate the difference between short- and medium-term effects of training and husband participation.
JEL Codes: G21, O16, C21, C23, C83
I. INTRODUCTION
Human capital is besides labor and physical capital an important determinant of economic growth (Mankiw, Romer and Weil, 1992). Consequently, developing countries can trigger economic growth by either accumulating physical or human capital, because labor is often abundant. This study focuses on accumulation of human capital by providing a business training to female microfinance clients of TYM, a Microfinance Institution (MFI) in Northern Vietnam.
Typical microfinance clients are poor, unemployed and have a low education level. Most of the clients of MFI’s are women, because women have a better reputation in repaying and spending the money in the right way (Armendariz and Murdoch, 2010). Moreover, entrepreneurial activity among women is high. The employment rate for women is low, due to responsibilities for housekeeping and taking care of the children. As a consequence, entrepreneurial activity can generate income for those who are mostly out of the working force. In addition to financial services Microfinance Institutions provide nonfinancial services, for example business training. Business training to microfinance clients tries to improve management capabilities of entrepreneurs to boost entrepreneurial activity. Business training can help these women to improve their businesses and increase profits, which reduces poverty. This study does not focus on the effect of microfinance, but on the impact of a business training for those who already have access to credit.
This study is subsequent to Vu (2014). In collaboration with TYM, Vu (2014) estimates the short-term effects of a gender and business training with and without participating husbands on female microfinance clients. As a consequence, I use the same dataset as Vu (2014), but add data of a second follow-up survey. A time span of exactly two years between the first and third interview introduces the possibility to estimate the transition in effects of training and husband participation over time.
I evaluate a gender and business training with a randomized control trial (RCT) with female microfinance clients of TYM in Vietnam. The business training is based on the Gender and Entrepreneurship Together (GET) training package of the International Labour Organization (ILO). The sample of 4041 women is randomly divided into two treatment groups and a control group. Both treatment groups are offered nine training sessions provided by local loan officers from February to October 2012. In one of the treatment groups husbands are invited, and in the second treatment group women attend meetings alone. There are three interviews conducted per women: before, six months after, and twelve months after the training modules. With this data, I estimate the effect of training on women’s business knowledge, as well as the amount of implemented business practices. I use revenues, profit and profit margin of agricultural and business activities as indicators for women’s business outcomes. Examples of business activities are retail trading and manufacturing. Among other things, I estimate the effect of training on the extensive margin of business activities, i.e., business entry and business exit. Business entry is defined as the probability that a woman starts new business activities. Business failure is defined here as the probability that a woman stops her main business activity in two years. Thus, I refer to the extensive margin as the amount and range of business activities conducted, and to the intensive margin as the intensity and quality of business activities.
on women’s business outcomes. By conducting multiple follow-up interviews I can gather insights in the effect of business training over time as well as the effect of participating husbands.
In conclusion, a business training is effective for female microfinance clients in Vietnam on business knowledge as well as on business practices; training participation with and without their husbands improves business literacy and business management. The effect of training on business practices increases over time; there are significantly more business practices implemented after twelve months than after six months. The main findings show that a business training improve the profitability of females’ business activities, but does not have a significant effect on business sales. Moreover, women with training are more likely to start new business activities. The probability to stop current business activities is lower for women receiving a business training than for women in the control group. Despite an increase in sales and profits for women receiving training with their husbands after one year, the effects of training on agricultural outcomes are not significant. This might be related to the fact that the training focuses on gender equality and businesses. Although there are some differences in the effects of training between the two treatment groups, results show that the conditional effects of inviting husbands on the impact of training on female microfinance clients are not significant. Though the effect of husbands participating at training meetings is negative on business entry.
The remainder of the paper is structured as follows: Section II discusses relevant literature; Section III presents a theoretical framework and Section IV describes the estimation methods. Section V describes the content of the training; Section VI describes the data; Section VII reports the results and Section VIII concludes.
II. LITERATURE OVERVIEW Business training
starting new activities and shutting down existing businesses is not significantly affected by training (Bruhn and Zia, 2013). Drexler, Fischer and Schoar (2010) show significant improvements in business practices and revenues during bad weeks with simple rules-of-thumb training. Yet, formal accounting training to entrepreneurs from the Dominican Republic does not have any effect on business performance.
Most studies including males and females conclude that the effect of training on women is less pronounced than the effect on men. However, studies focusing on female entrepreneurs do find positive effects of training on business outcomes (Calderon, Cunha and de Giorgi, 2012; Field, Jayachandran and Pande, 2010; Valdivia, 2013). According to Calderon et al. (2012) business training increases profits, revenues and the amount of clients served due to improvements in business knowledge and practices, i.e., formal accounting techniques and tax payments. Valdivia (2013) concludes that women with general business training are more likely to stop losing business activities. Women receiving personal assistance plan and execute more innovations, which result in a more than eighteen percent increase in sales. Field et al. (2010) finds a relation between business training on women and social restrictions by traditional religious and caste institutions in India. Training does not have an effect for Muslim women, who face high restrictions and neither for lower caste Hindu women, who face the least restrictions. For Hindu women from upper caste, training has a significant effect on the probability to earn a business income. This suggests that training does not mitigate the problems faced, if the restrictions are too high.
business practices of (potential) female entrepreneurs, but the improvement does not result in more business profits, sales or capital stock. In addition, they show that training for women in combination with a cash grant improves short-term profitability, but profitability returns to pre-training levels after two years.
Participating husbands
In a field experiment in southern Mexico with Karlan and Mullainathan, information asymmetry and family tension in households decrease when husbands participate in credit meetings (Armendariz and Roome, 2008). Men tend to overestimate the amount of money women are handling and reduce their contributions to household expenditures when women receive credit. Yet, invited men are less likely to do so, and consequently, women have more money to invest. Secondly, an increase in tensions can arise when women spend less time on housekeeping due to increasing business activities (Armendariz and Roome, 2008). Time constraints can result in a decrease in quality or quantity of services provided, i.e., in meals or household chores. Men attending credit meetings help their spouses more with business activities and household chores both. Although husbands’ attendance at credit meetings is different from husbands’ attendance at a business training, this finding might be relevant for the effect of participating husbands in training meetings either.
III. THEORY OF CHANGE
The process of training and participating husbands affecting business outcomes of female microfinance clients can be explained by a theory of change. Figure 1 presents a model of the expected interaction between outcome variables. This study tests for changes in business knowledge, business practices and business results due to a business training. Even though I estimate whether husbands’ attendance influence the effect of training, I do not study the mechanism underlying this interaction. Inviting husbands may have an effect on business knowledge, business practices and business outcome by improving the gender empowerment of women, but interested readers are referred to Vu (2014).1
Figure I. Model of expected effects of training and husband participation on business outcomes P ar ti ci p at io n h u sb an d s B u si n es s tr ai n in g M ic ro fi n an ce Business outcomes Physical capital Human capital Investments Business practices Bargaining power Gender knowledge Business knowledge Financial capital
Vietnam is characterized by a patriarchal culture implying that women, in general, are responsible for housekeeping and men work outside. This limits the available time to spend on entrepreneurial activities for women. Gender empowerment seems important in this study, because most of the female clients in Vietnam are married. Moreover, as wives and husbands often make decisions together, the amount of bargaining power of a woman influences decisions. Bargaining power is defined here as the relative ability to influence the other members from the household, which will depend, i.e., on the ability to survive and succeed without the other household members. Gender knowledge can help women to increase bargaining power and make decisions more independently. Thus, I expect that the attendance of men will increase the effectiveness of training due to an increase in the quality of the discussions resulting from the work experiences of men. Moreover, inviting men can help women to strengthen their position in the household, because men’s knowledge about gender equality increases.
In conclusion, according to the expectations business knowledge of women will increase when husbands attend training meetings, because men may help to remember the training content. Second, I expect a change in business management, because men will be more involved in their spouses’ enterprise and men and women will discuss investment decisions together when they jointly participate a business training. Third, I expect an increase in business sales and profit. Improvements in business management might have positive effects on business outcomes and women may also benefit from men assisting them in household chores and business activities.
Relevant risks
Optimally, all interactions are positive and contribute to higher sales, and thus, an increase in income for poor women. However, in every step I can identify specific risks, which mitigate the positive effects of microfinance, training and husband’s attendance. This section discusses some risks for every factor, although I do not test all changes in the factors provided in the theoretical framework above.
trainees do not want to study at all. Other causes of unmotivated trainees are sickness and geographical distance to the training centre.
Business knowledge will not increase sufficiently if people forget what they have learned or if they lose the classroom material. Training will not lead to a change in business practices when the training is too difficult or too costly to implement, when administrative facilities are unavailable or due to a lack of self-control.
Training can lead to an increase in asymmetric information between the women and her spouse, which can increase conflicts and discussions between the household members. Moreover, women with more knowledge about gender issues, do not automatically receive more bargaining power. Even if women experience an increase in bargaining power, it is possible that an increase in bargaining power does not affect decisions. To decrease asymmetric information problems and to stimulate households in treating men and women equally, husbands can also participate in the business training.
Yet, improving entrepreneurial activity may still not lead to improvements in sales or profit, when demand of the product or service is low or when spillover effects are substantial; i.e., when other businesses benefit either the improved micro-enterprise loses its competitive advantage. It is also possible changes in business outcomes are not reflected in reported numbers, because people calculate sales, costs and profits different than before training.
IV. TRAINING
TYM introduces training sessions in credit centres in the regions Vĩnh Phúc and Hanoi. Loan officers provide a training of around 45 – 60 minutes during nine monthly centre meetings. Loan officers received a ‘training of trainers’ course by the TYM’s headquarters staff; the ILO taught these staff members about the GET training package. In addition to the training modules, loan officers spend 15-30 minutes on specific problems of clients during debt repayments every week. Attending training sessions is free of charge and voluntary, thus clients can make their loan repayment and leave the centre meeting before the training starts.
The Gender and Entrepreneurship Together (GET) training package of the International Labour Organization (ILO) focuses on: basics on gender and entrepreneurship; the businesswoman and her environment; the business project; people, organization and management. The training aims to improve confidence, gender equality and business management.
The topics per module are:
1. Gender and entrepreneurship, gender equality and the life cycle of people and enterprises; 2. Creating self-confidence for businesswomen
3. Businesswomen and their environment, self-development and business mapping; 4. Business projects: ideas, opportunities and challenges;
5. Marketing;
6. Financial knowledge and interest rates calculations; 7. Cash management;
8. Bookkeeping: accounts receivable and accounts payable; and 9. Calculation of production costs and cost of goods sold.
V. DATA
This study uses data of a randomized control trial on female clients of TYM Fund, the largest Microfinance Institution in Northern Vietnam. TYM operates mainly in areas with high ratios of poor households and serves 10 areas in the Northern regions of Vietnam with a total of 48,000 clients. Because the fund started as a project of the Vietnam Woman Union in 1989, all clients are female. There are 1450 lending centres in 43 branches, where TYM provides financial and non-financial services at weekly or monthly meetings. Loan officers assess loan application forms and collect repayments and savings during these centre meetings. Credit meetings are also an opportunity for clients to exchange experiences and information about production and business activities. TYM organizes social activities at credit centres either.
Figure II. Time line of the business training, interviews and focus group discussions Data collection started in October 2011 until November 2013. A baseline survey is conducted before randomization; assigned women received a business training in 2012, and follow-up surveys are conducted six as well as twelve months after the treatment. ‘F’ refers to focus group discussions with two women per group in February 2013.
Data collection started in October and November of 2011, with a baseline survey before the treatment (Figure II). During February and October 2012, the women received training and in March and April 2013 the midline was conducted. In October and November 2013 the endline survey was conducted. In February 2013, six focus group discussions (F) with two women out of each group about the relevance and satisfaction of the training were conducted. The base and endline survey were conducted in the same months of the year to reduce seasonal effects in business results.
Oct 11 Dec 13
Twenty-three experienced researchers interviewed clients within two months. They used double data entry to reduce the chance of mistakes. Demographic variables, for example age, education and marital status are collected, as well as measures for financial literacy, cognitive and non-cognitive skills. Determinants of wealth, such as income, health and social network are questioned. Questions concerning business and farming activities, such as location, types of business activities and the amount of employees are also included in the survey. Financial information is gathered by asking the type of insurance, the amount of savings and outstanding loans. Additional questions about training participation and women’s satisfaction of the quality per training module are included in the second interview for both treatment groups.
Females with an official job have permission to skip the compulsory centre meetings, because the meetings take place during working hours; consequently, they are not included in the study. Researchers interviewed 4041 women at the baseline, of which 1,859 clients are assigned to the control group. Randomization is stratified per lending branch: three in Vĩnh Phùc and one in Hanoi. Group T1 consists of 1,509 women, who are invited to participate in a business training with their husbands. There are 673 women interviewed assigned to group T2, thus the sample size of the first treatment group is higher than the second treatment group. The participation of husbands is expected to be lower on average than for female microfinance clients, because the latter are at centre meetings anyway. Consequently, the sample size of T1 has to be bigger than the sample size T2 to reach a minimum level of statistical power to detect treatment effects on women’s outcomes. A maximum of 23 members per centre are interviewed, which are randomly chosen. The amount of members per center is limited in order to maximize the amount of centres in the sample with a limited budget.
Pre-treatment analysis
A survey before the treatment can detect differences in treatment and provides a possibility to check the randomization. Table I shows descriptive statistics of variables at the baseline survey. On average women are 44 years old, Kinh (Vietnamese ethnic group) and married. They have 6.8 years of education and manage at least one farming activity, i.e., growing rice. The households consist on average of 4.7 members, hold a farming land of 1400 m2 and earn a monthly income of VND six million (about $282).2 One third of the interviewees own a business and 76% of them is interested in training.
In order to test for differences between groups, I use an ordinary least square regression of each demographic variable on dummies for assignment into group T1 and T2. Standard errors are clustered at the centre level, because randomization is not at the individual level. Cluster effects result in homogeneity among women from the same centre, consequently, I adjust the standard errors in a balancing test. Results of a balance test in Table I provide evidence for random assignment of women into treatment groups. Demographic variables per treatment group are not significantly different from the control group at the baseline. Column 7 in Table I shows that baseline variables of invited women to training do not significantly differ from the control group either. Furthermore, Table I presents several variables of business outcomes, for example, business knowledge, business practices and business results. Although there are significant differences between treatment groups in profit and profit margin of business activities in a regular month, business outcomes of the treatment groups do differ from the control group at the baseline survey.
2
Table I. Descriptive statistics baseline variables and a balance test
This table shows variables measured in a pre-treatment interview of a randomized control trial with 4041 female microfinance clients of TYM in Vietnam. Columns (1)-(4) present means, robust standard errors, minimum values and maximum values for the full sample. Columns (5) – (6) provide t-statistics estimated with an ordinary least square regression of the independent variable on T1 and T2; ‘T1’ refers to treatment group T1, in which women are invited to participate in training with their husbands together; ‘T2’ refers to women of treatment group T2, in which women are invited to participate in training alone. T-statistics in column (7) are estimated with an OLS regression of variables on Training; ‘Training’ refers to a dummy for both treatment groups. Income, sales and profit for a normal month in VND. P margin refers to profit margin: profit divided by sales. B is aggregated business activities; B1 refers to women’s main business and A is agricultural business activities. *** Denotes significance at the 1%-level, ** at the 5%-level, and * at the 10% level.
Mean
Standard error
t- statistic
Min Max T1 T2 Training
(1) (2) (3) (4) (5) (6) (7)
Age (years old) 43.77 10.33 19.00 72.00 -0.83 -0.12 -0.72
Schooling (years) 6.82 2.91 0.00 18.00 -1.11 -0.19 -0.90
Married 0.82 0.39 0.00 1.00 -0.64 0.45 -0.32
Ethnic group (Kinh) 0.94 0.23 0.00 1.00 -0.51 -0.14 -0.46
Household size 4.74 1.56 1.00 15.00 -0.40 -0.56 -0.54
City (Hanoi) 0.26 0.44 0.00 1.00 -0.14 0.11 -0.05
Credit access TYM 1.10 0.70 0.00 2.00 0.13 -1.32 -0.53
Business knowledge
Two knowledge indices measure women’s general business knowledge. The amount of correct answers given in the interview determines the value of the index for a woman. On the basis of ten general business and six financial questions women obtain a score for the first business knowledge index. The second knowledge index includes seven other financial questions, eight accounting and ten marketing questions. However, the first survey does not include the 25 questions related to the second index, consequently, index 2 is calculated in the two follow-up surveys only.
Business practices
With four practices indices I compare women’s business management. Twenty statements question whether a woman implements business practices. Based on the reported practices, a principal component analysis (PCA) predicts a score per index (see Appendix B). Each index is named after the practices, which have the highest weight in that index. The first, general business practices index consists mostly of recording, business discussion, basis marketing and a business plan. In the second index, innovation and implemented new ideas or performed any activities to increase number of customers, which are associated with ‘Innovation’ weights heavily. Marketing strategies has a high loading in a third index. Last, keeping records and business planning has a high weight in a fourth index.
Agricultural activities
Interviewers record the three main agricultural activities of each woman. The type of activities reported are growing rice, growing fruit trees, raising pigs or cows and raising chickens, ducks or doves. Subsequently, women report sales, expenses and consumption by the household of each farming activity. With the reported cycle time, I convert the amounts to revenues and costs per month.3 Sales and profit of last month as well as in a regular month are aggregated for all reported agricultural activities. Although self-reported profits are also recorded, I use sales minus costs to calculate profits. Profits divided by sales result in a profit margin on farming activities.
3
Business activities
Sales, costs and profits of women’s three main business activities are reported per business cycle and converted to outcomes per month. Outcomes of business activities are measured in two different ways. First, the business results of all existing and new business activities are aggregated to derive total sales, profits and profit margin of business activities. Business results are only reported for women with business activities. Although some of the households do not have business activities in every interview, women with at least one business activity in the first, second or third interview are included. Changes in total business outcomes give an indication of the effects on a household’s income and capture the effect of training on business entry and exit. Second, I also use business results of women’s ‘survived main business activity’. Changes in women’s main business activity provide more insight in changes within a business. The most frequent type of women’s most important business activity is retail trade, manufacturing and services, respectively. Only business outcomes of women’s most important activity are used, when the activity exists at the baseline and is still present in the last interview. Accordingly, these outcomes are named as business results of a ‘survived main business activity’.
Entry into business activities is analyzed by looking at new business activities. A start-up is defined as new business activities during the study, which were not present at the baseline. Exit of business activities is analyzed by looking at ‘business failure’. Business failure is defined as the main business activity present at the baseline not reported in the midline, endline, or both. Although the label ‘business failure’ might imply that a woman is forced to end her business activity, this may not always be the reason for a woman to stop a business activity.
Control variables
missing in the last survey and consequently, I assume that the number of household members is the same as reported in one of the interviews before. From the question whether a woman is single, married or widowed, I derive a dummy for ‘marital status’. The dummy is equal to one if a woman is married. A region dummy equals one if a woman lives in Hanoi.
Participation
Low participation rates and small sample sizes resulting in underpowered studies, are a problem in many business training studies (McKenzie and Woodruff 2014). The participation rate is defined as the percentage of women and men invited to the training who actually participated in the training. Loan officers record the attendance of every woman and their husband for every meeting. Table II shows a participation rate per module of around 80% for women. The participation rate is lower for men than for women, despite a participation fee for men. The fee diminishes, which might be one of the reasons for a decreasing participation rate for men. Furthermore, gender equality is the topic of the first two training modules and therefore the training might be experienced by men as irrelevant and/or not interesting.
Vu (2014) reports that women who participated are highly satisfied about the quality of the training and would recommend it to others. Yet, only 16% of the women are willing to pay for the business training. Women are also asked to rank the modules based on importance. The first module about gender issues is ranked as most important and module six, about interest rates, is ranked as least important. The latter module seems to be too difficult for most of the participating women.
Table II. Participation rate per module of men and women in treatment group T1 and T2.
This table shows the participation rate of women and men per training module. Women in treatment group T1 are invited to join training with their husbands together and women in treatment group T2 are invited to participate in training alone.
Women T1 Men T1 Women T2
Module Amount Percent Amount Percent Amount Percent
1 1294 84.6 1132 39.0 877 79.0 2 1294 81.8 1132 35.6 877 74.6 3 1294 84.5 1132 32.1 877 81.5 4 1294 83.8 1132 30.2 877 82.9 5 1294 82.5 1132 26.1 877 78.4 6 1294 84.6 1132 25.8 877 82.9 7 1294 84.0 1132 12.1 877 82.3 8 1294 84.2 1132 8.7 877 81.0 9 1294 80.9 1132 7.2 877 78.2 All 1294 39.8 1132 1.7 877 36.3 Attrition
Table III. Attrition analysis
This table shows coefficients of a logistic regression of attrition on treatment groups T1 and T2, age, household size, marital status and region; standard errors adjusted for cluster effects at the centre level in parentheses. The sample consists of 3940 women interviewed in the baseline; a dummy for attrition is equal to one if households are not interviewed in one of the two follow-up surveys. *** Denotes significance at the 1%-level, ** at the 5%-level, and * at the 10% level.
Attrition (1) T1 -0.17 (0.158) T2 0.26 (0.217)
Age (years old) -0.01
(0.006)* Household size -0.05 (0.039) Married -0.12 (0.151) City (Hanoi) -0.41 (0.231)* Interest in training -0.16 (0.139) Constant -1.07 (0.355)***
VI. ESTIMATION METHODS
and D provide the effects of business training on business outcomes estimated with a post-treatment analysis and ANCOVA. I estimate the effects of training and husband participation on business entry and exit with a logistic regression.
Difference-in-difference estimation of treatment effects
I estimate the effects of training and test for differences in the effectiveness of training between the midline survey and the endline survey with a difference-in-differences (DD) estimation method, in which a conditional effect of participating husbands is included:
(1)
where is an outcome variable for an individual i in centre j at time t; is a dummy equal to one if the observation belongs to the mid and endline survey; is a dummy equal to one if it is an endline observation; is a dummy equal to one if the individual is assigned to receive training; is a dummy equal to one if the woman’s husband is invited to the training; is a vector of control variables age, household size, marital status and region in time t, and is an error term. Time index t is zero, one or two for an observation belonging to the baseline, midline or endline survey. The effect of business training is estimated by beta 1 and the additional effect of business training between mid and endline survey is given by beta 2. Estimator beta 3 tests the conditional effect of inviting husbands. Lastly, the additional effect of husband participation between mid and endline survey is estimated by beta 4. Calderon, Cunha and de Giorgi (2012); Karlan and Valdivia (2011); Valdivia (2013) estimates the effect of business training on business performance with a difference-in-difference model either.
control group before and after treatment. Thus, time dummies control for time trend effects and dummies per treatment group control for differences before the treatment.
With a balanced panel, Equation (1) is equivalent to a fixed-effect model. Included households are interviewed thrice and the related variables need to be reported in all three surveys in a balanced panel. Consequently, households dropped out the sample for a difference-in-differences estimation if they missed the variable in one of the years.
In a randomized control trial treatment and control groups are similar; therefore we can expect that without intervention, individuals will behave in a similar way. Random assignment of the sample minimizes the potential of a selection bias and results in groups equal in demographics, outcome variables and trends in outcome variables over time. As a consequence, I am reasonably sure that a difference in the endline survey is only due to a difference in treatment. A balance test shows that a significant difference in profitability of business activities between treatment groups is apparent (Table I). Yet, the treatment groups are not different from the control group with respect to the other baseline variables.
To compare the estimation with a difference-in-differences method, I also estimate the effects of training and husband participation on women’s business outcomes at the endline with a post-treatment analysis, which reads as:
where refers to an outcome variable for an individual i in centre j at the endline survey; is a constant, is a dummy variable that takes the value one if the woman is selected to receive business training; is a dummy equal to one if training is offered to a woman as well as to her husband (group T1); is a vector of baseline control variables age, household size, marital status and region and is an error term. The coefficients of interest are and ; beta 1 estimates the effect of training at the endline survey and beta 2 estimates the conditional effect of participating husbands on females’ outcomes.
Post-treatment analysis is often used to estimate the effects of business training on business outcomes, for example, Berge et al. (2012); Bjorvatn and Tungodden (2010); Bruhn and Zia (2013); Field et al. (2010); Giné and Mansuri (2011) and Karlan and Valdivia (2011).
The post-treatment analysis estimates the differences between the control and treatment group after the latter receives the treatment. I estimate the outcome variables on the treatment groups (T), which are assigned to get business training. In the regression is also a dummy for T1 included, which estimates the additional effect of husband participation during the training. I include control variables, such as dummies for age, marital status and region and the number of household members, because I do not want to find differences driven by one of these demographic variables. The dependent variable of the baseline is not included in this estimation method. Therefore, treatment effects on outcomes only measured at the midline and endline, can be estimated with a post-treatment analysis, i.e., business knowledge index 2.
Third, a first-difference estimation can be used to estimate the treatment effects either. A post-treatment analysis, in which the lagged outcome variable is included in the regression as covariate (ANCOVA) reads as:
(3)
where denotes the value of the outcome variable at the baseline. The variables , , , and are defined above. As in Equation (2), estimates the effect of a business training after twelve months and the additional impact of inviting husbands on the effect of offering training to women. Although this post-treatment analysis is called ANCOVA here, I include the outcome of the baseline survey as control variable instead of including the lagged dependent variable. Drexler et al. (2010); Giné and Mansuri (2011); Karlan and Valdivia (2011); de Mel et al. (2014); Valdivia (2013) estimates the effect of business training with post-treatment analysis including the outcome variable at the baseline (ANCOVA).
experiment (Hair, Black, Babin and Anderson, 2010). Moreover, including the baseline variable accounts for systematic differences across treatment groups. Treatment groups that should be equivalent can exhibit different average levels of . Dependent baseline variables in ANCOVA eliminate differences across groups, and consequently, reduce the likelihood of finding a significant effect. Thus, including covariates in ANCOVA reduce the within-group variance and explain differences in means across different treatment groups.
Longer run effects
To estimate the effectiveness of training in the short- and medium run on the different treatment groups I use the difference-in-differences estimation in Equation (1) and reformulate it as follows:
(4)
where is a dummy equal to one if the observation belongs to the midline survey; is a dummy equal to one if men and women are invited to the training together; is a dummy equal to one if a woman is assigned to receive training, but her husbands is not allowed to join the training; ; and are defined above. The coefficients of interest are beta 1, beta 2, beta 3 and beta 4. Beta 1 estimates the short-term effect for women with participating husbands; beta 2 estimates the short-term effect of training for women without their husbands’ attendance. The long-term effect of business training is estimated for group T1 by beta 3 and for group T2 by beta 4.
Entry and exit of business activities
The logistic regression to estimate the effect of training on business entry and business exit reads as:
(5)
where is a dummy for business start-up ( ) or business failure ( ); is a dummy equal to one is the individual is assigned to receive training; is a dummy equal to one if the woman’s husband is invited to the training; is a vector of covariates age, household size, marital status and region; is an error term. The coefficients of interest are and . Beta 1 estimates the effect of training on business start-up and business failure and beta 2 estimates the effect of husband invitation to women’s business entry and business exit.
A start-up dummy for households in the endline is equal to one if the woman started at least one new business activity in two years; dummy is equal to one for households with a business activity not reported in the first interview. The sample included in the regression for start-up consists of all households present at the endline survey.
When a woman has a business activity in the first interview and she does not report this activity in the last interview, she stopped her main activity meanwhile. Therefore, a dummy business failure for women with business activities in the baseline, , is equal to one if the main activity is not present after two years. Thus, the sample of women included in the regression for business failure only those households interviewed in the endline, which had business activities in the baseline.
Cluster effects
The level of randomization affects statistical power, spillover effects and operational costs. A bigger sample size is required with randomization at the group level than randomization at the individual level to have the same statistical power. Thus, for power considerations randomization at the smallest unit is preferred. However, credit meetings and trainings are always in credit groups, which would introduce spillover effects. There is spillover when the control group is affected by the treatment, even though the control group does not receive the treatment. Contamination happens when the control group is affected by other similar treatments. Since TYM is the only Microfinance Institution in Northern Vietnam providing BDS, other business trainings are not a problem for contamination.
VII. RESULTS
This section presents the results in four parts: business knowledge, business practices, business results of agricultural activities and business results of business activities. Table IV and Table VI provide difference-in-differences (DD) estimation of the effects of offering a business training to female microfinance clients and the effects of participating husbands. I compare the results of a DD estimation with a post-treatment and first-differences estimation (ANCOVA). Appendix C provides the post-treatment analysis with and without dependent variable of the baseline included. Table V and Table VII present the total effects of training after six and twelve months for each treatment group separately. The observant reader can see that regressions in Table IV and Table V are similar as well as regressions in Table VI and Table VII, because Equation (1) can be reformulated to Equation (2).
Business knowledge
microfinance clients with and without participating husbands improve their business literacy after six as well as after twelve months.
Furthermore, column (2) in Table C.I presents a significant increase in women’s business knowledge index 2 with a business training. This second index for business knowledge measures not only financial literacy, but also accounting and marketing knowledge. Thus, after one year women who received a business training performed significantly better in a business knowledge test with 10 as well as a test with 25 questions.
The participation of husbands does not have a significant influence on the effect of a business training for female microfinance clients on business knowledge (Column 1 in Table IV). In addition, the effect of participating husbands on business knowledge is not significant with post-treatment estimation either (Table C.I and C.II).
The gender and business training offered by TYM focuses on marketing, financial literacy and calculating costs of goods sold; consequently, an increase in women’s business knowledge is conforming to the theory of change. The positive impact of training on women’s business knowledge is supported by previous studies to business training (Berge et al., 2012; Bjorvatn and Tungodden, 2010; Bruhn and Zia, 2013; Giné and Mansuri; Karlan and Valdivia, 2011).
Business practices
Although Table V shows that both treatment groups benefit from a business training, columns (2) and (3) in Table IV indicate that a significant increase in implemented practices after training is not affected by inviting husbands. Table C.I and C.II present that the effect of participating husbands is neither significant with a post-treatment analysis or ANCOVA. In conclusion, women with training do not implement more business practices with participating husbands at the training than without participating husbands.
Previous studies points out that training does not improve business outcomes of female micro-entrepreneurs, because women might have difficulties to implement new business practices (Giné and Mansuri, 2011; Valdivia, 2013). Though, there is also evidence for positive changes in business management after business training (Calderon et al.; Drexler et al, 2010; Karlan and Valdivia, 2011; deMel et al, 2014). In this study, the majority of the invited women report a change in business management due to the business training; keeping records is valued as the most important change in business practices (Vu, 2014). Thus, in alignment with the theory of change business management of female microfinance clients in Vietnam improve significantly with gender and business training.
Agricultural activities
Table IV and Table V show that there are no changes in sales, profits and profit margins of women’s agricultural activities in the short run with business training. Sales and profit increases, marginally significant, between six and twelve months afterwards (columns 4-9 in Table IV). Column (8) in Table C.I presents an increase in total agricultural sales of last month for women with training with a post-treatment analysis; the improvement is significant at a significance level of 10%. Taken together, I can conclude that a gender and business training on female microfinance clients of TYM in Vietnam do not have an impact on business outcomes of women’s agricultural activities, when husbands do not attend training meetings.
show that a business training with participating husbands increases sales and profit of women’s agricultural activities after twelve months significantly. For a post-treatment analysis and ANCOVA one can derive the total longer-term effect of offering a business training to treatment group T1 by adding the training and participating husbands estimators together; the derived coefficient in both estimation methods is significantly different from zero at a significance level of 1% (Appendix C). As a consequence, an increase in sales and profit for agricultural activities with a business training in which husbands participate holds for estimation with post-treatment and first-difference estimation techniques.
Table V shows that revenues of agricultural activities also increased for treatment group T2 after twelve months, but this effect is not significant at a significance level of 5%. In conclusion, training for women is only effective on sales and profits of agricultural activities after one year when husbands are invited to join their wives. This is in alignment with the expectation that the participation of husbands could influence women’s business outcomes, because men may have more attention for their activities when they attend training meetings.
The training does not focus on agricultural businesses, which might explain the insignificance in the treatment impact. Most of the topics might be irrelevant, although implementing business practices for agricultural activities can be as valuable for agricultural activities as for business activities. In other words, the relevant part of training for farmers can be a valuable treatment, but the intensity of the treatment is probably too low to result in significant improvements. The first module of the provided training focuses on gender equality, which is reported by women in focus groups as most valuable module for participating husbands. Thus, increases in sales and profit of agricultural activities for treatment group T1 might be due to the presence of husbands at training meetings. A suggestion for further research is to study whether participating husbands help their wives more with farming activities and whether experiences of men contribute to the effectiveness of training for women.
Business activities
changes are not significant. Aggregated profit of business activities in a normal month increases significantly with gender and business training and profit margin increases significantly in a normal month as well as in last month (Columns 3-6 in Table VI). Using post-treatment estimation with and without including the dependent base variable, I do not find significant evidence for effects of a business training on aggregated business outcomes after one year (Appendix C). In conclusion, a positive effect of a business training on the profitability of business activities estimated with a difference-in-differences method does not hold for other estimation methods. Besides, profit increases more in one year than in six months, which suggests that improvements in business knowledge and practices needs some time to result in higher profits (Table VII). The positive effects of training on the profit margins of business activities in treatment group T1 increases over time. On the other hand, the positive effect of training on profit margin of business activities in treatment group T2 decreases over time. Thus, training for women with participating husbands increases profitability of business activities; this effect is higher after twelve months than after six months. Yet, when women attend meetings without their husbands the increase in profit margin due to training is lower after twelve months than after six months.
Even though the conditional effects of participating husbands are not significant, increases in business profitability are higher for treatment group T1, than for group T2 (Columns 3–6 in Table VI and Table VII). Moreover, profitability increases over time for women with training when husbands participate, whereas, the increase in profitability fades out for women who attend training meetings alone. Thus, effects of training on business activities are different for women attending training alone than for women who join training meetings together with their husbands.
Main business activity
The short-term effect of training on the profit margin on women’s most important business activity fades out after twelve months. The additional impact between the midline and endline survey is negative for profit margin in a normal month (significant at a significance level of 10%). Thus, the impact of training on the profitability of survival main business activities of women decreases over time. An increase in profitability due to training what diminishes over time is in line with the findings of de Mel et al. (2014).
Table IV. Effects of training and participation of husbands on business knowledge, practices and agricultural outcomes with DD estimation This table shows the results of a difference-in-difference fixed-effect model with control variables age, household size, marital status and region included (Equation 1). A randomized control trial with 4041 female microfinance clients of TYM in Vietnam estimates the impact of a gender and business training with and without invited husbands. ‘Post’ refers to post-treatment surveys conducted six and twelve months after training and ‘End’ to the endline survey, which is the last follow-up survey. ‘T’ refers to post-treatment groups assigned to training and ‘T1’ to a treatment group with participating husbands. Business knowledge 1 is defined as the total amount of correct financial literacy questions (maximum 10); General and Innovation are business practices indices constructed by a PCA on 7 statements; Columns (4)-(9) reports total outcomes of women’s three main agricultural activities: sales, profit and profit margin of last month and for a regular month; Profit is sales minus costs; Profit margin is profit divided by sales. Standard errors, clustered at the centre level are shown in parentheses under the coefficients. *** Denotes significance at the 1%-level, ** at the 5%-level, and * at the 10% level.
Knowledge Business Practices Agricultural activities
Business 1 General Innovation Sales Profit Profit margin
Table V. Effects over time of training with and without husbands on business knowledge, practices and agricultural outcomes with DD estimation This table shows the results of a difference-in-difference fixed-effect model with control variables age, household size, marital status and region included (Equation 4). Data of a randomized control trial with 4041 female microfinance clients of TYM in Vietnam is used to estimate the effects of a gender and business training for women with and without participating husbands. ‘Mid’ refers to the midline survey, a first follow-up survey six months after the treatment, and ‘End’ refers to the endline survey, a second follow-up survey twelve months afterwards. ‘T1’ refers to treatment group 1 in which women and men are invited to attend training together and ‘T2’ to a second treatment group in which women attend the training alone. Business knowledge index 1 is measured by 10 financial literacy questions; General and Innovation are business practices indices constructed by a PCA on 7 statements (Appendix B); Columns (4)-(9) reports total outcomes of women’s three main agricultural activities: sales, profit and profit margin of last month and for a regular month; Profit is calculated by sales minus costs; Profit margin is profit divided by sales. Standard errors, clustered at the centre level are shown in parentheses under the coefficients. *** Denotes significance at the 1%-level, ** at the 5%-level, and * at the 10% level.
Knowledge Business Practices Agricultural activities
Business 1 General Innovation Sales Profit Profit margin
Table VI. Effects of training and participation of husbands on business outcomes with DD estimation
This table shows the results of a difference-in-difference fixed-effect model with control variables age, household size, a dummy for married women and a region dummy included (Equation 1). Data of a randomized control trial with 4041 female microfinance clients of TYM in Vietnam is used to estimate treatment effects of a gender and business training and the invitation of husbands to join their wives. ‘Post’ refers to post-treatment surveys conducted six and twelve months after training and ‘End’ to the endline survey, the last follow-up survey. ‘T’ refers to randomization groups assigned to training and ‘T1’ to a treatment group with invited husbands. Columns (1)-(6) reports total outcomes of women’s three main business activities on a sample of women who reported business activities in one of the three interviews; Columns (7) – (12) report outcomes of women’s main business activity on women who reported this activity in the baseline, midline and in the endline survey; Sales, profit and profit margin reported of last month and for a regular month; Profit is sales minus costs; Profit margin is profit divided by sales. Standard errors, clustered at the centre level are shown in parentheses under the coefficients. *** Denotes significance at the 1%-level, ** at the 5%-level, and * at the 10% level.
Aggregated business activities Survival main business activity
Sales Profit Profit margin Sales Profit Profit margin
Table VII. Short and longer-term effects of training with and without participating husbands on business outcomes with DD estimation This table shows the results of a difference-in-difference fixed-effect model with control variables with control variables age, household size, marital status and region included (Equation 4). Data of a randomized control trial with 4041 female microfinance clients of TYM in Vietnam is used to estimate the effects of a gender and business training for women with and without participating husbands. ‘Mid’ refers to the midline survey, which is a first follow-up survey six months after the treatment, and ‘End’ refers to the endline survey, which is a second follow-up survey twelve months afterwards. ‘T1’ refers to treatment group 1 in which women and men are invited to attend training together and ‘T2’ to a second treatment group in which women attend the training alone. Columns (1)-(6) reports total outcomes of women’s three main business activities on a sample of women who reported business activities in one of the three interviews; Columns (7) – (12) report outcomes of women’s main business activity on women who reported this activity in the baseline, midline as well as in the endline survey; Sales, profit and profit margin reported of last month and for a regular month; Profit is sales minus costs; Profit margin is profit divided by sales. Standard errors, clustered at the centre level are shown in parentheses under the coefficients. *** Denotes significance at the 1%-level, ** at the 5%-level, and * at the 10% level.
Aggregated business activities Survival main business activity
Sales Profit Profit margin Sales Profit Profit margin
Business entry and business exit
Entry into new business activities is reported in the second interview by 194 women. Additionally, 170 activities are started between the second and third interview, which are mostly retail trading. Table VIII shows that women with training are more likely to start new businesses (at a significance level of 1%). Although the additional effect of participating husbands is negative, the probability to start new activities is higher for treatment group T1, than for the control group ( = 0.85). The positive effect of training on business entry is in line with the expectations, because module 3 and 4 of the training focuses on self-development, business mapping and business opportunities. Moreover, start-ups are more likely when women are more confident or face less social restrictions due to gender training. However, a significant negative effect of participating husbands on business entry is not in alignment with the expectation. It might be that women discuss business plans more open with women.
Exit of business activities is defined as business activities reported at the baseline, but not present at the midline or endline. There are 1338 women with business activities interviewed in the baseline; 281 stopped their main activity in half a year and 252 stopped their main activity between the second and third interview. Table VIII states the results of estimation of business failure with a logistic regression (Equation 5). The probability that women stopped their main business activity in two years is significantly less with training ( = -0.45). Inviting husbands does not affect women’s probability in pulling the plug.
Table VIII. The entry of women into new business activities and exit of their main business activity Logistic regressions with control variables age, household size, marital status and region (Equation 5). Training refers to women assigned to a business training and ‘T1’ refers treatment group 1, in which women and men are invited to attend trainings together. Column (1) shows the treatment effects on the probability that a women starts new business activities in a year for all women interviewed in the baseline. Column (2) shows the treatment effects on the probability that a women stops her main business activity in a year for women with a business activity in the baseline. Standard errors, clustered at the centre level are shown in parentheses under the coefficients. *** Denotes significance at the 1%-level, ** at the 5%-level, and * at the 10% level.
Business entry Business exit
(1) (2) Training 1.40 -0.45 (0.267)*** (0.211)** T1 -0.55 0.29 (0.271)** (0.212) Constant -2.77 -0.92 (0.364)*** (0.373)** Sample size 4,234 1,338 VIII. CONCLUSION
might increase the effectiveness of training due to increased attention for their women’s business activities.
Participation rates in the training are high among women, but less for men; a lack of time is the most important reason for men’s non-attendance. Although men received a participation fee as incentive to participate, the opportunity costs seems to be too high for (especially employed) men to join their wives in training meetings (Vu, 2014).
I estimate the short- and longer-term effects of a business training as well as inviting husbands with a difference-in-differences fixed-effect model, because unobserved pre-existing differences or differences in time trends can confound with the treatment effect. Yet, a balance test shows that the three groups are similar before the treatment, except for the profit margin on business activities. Thus, a post-treatment and ANCOVA analysis are used to test whether the results with a difference-in-differences model hold with other estimation methods. Standard errors are adjusted for cluster effects, because randomization is at the center level and not at the individual level. Female microfinance clients, who followed a business training have more business knowledge and implemented more business practices than women in the treatment group, who did not participated in a business training. This study finds evidence for increases in women’s financial literacy and business management after six months as well as after twelve months.
The effect of training on sales, profits and profit margin on agricultural and business activities is not that uniform. Agricultural outcomes are not significantly different for women with a gender and business training than for women in the control group, which might be related to the fact that the training did not focus on agricultural businesses. Although sales of business activities do not change either, the profitability of women’s businesses improves significantly with business training. Moreover, women with business training are more likely to start new business activities, which might be related to an increase in self-confidence due to gender training modules. Discussion of business plans in training meetings can also explain a significant effect of training on business entry.
treatment group T2. Furthermore, the positive effect of training for female microfinance clients on the profitability of business activities is stronger when husbands participate. However, participating husbands diminish the positive effect of training on the entry of business activities. Supposing men influence women’s business outcomes by participating in the training, an increase in attention for their wives micro-enterprises is reasonable.
This study estimates changes in the treatment effects over time by comparing data of two follow-up surveys. Even though business practices increased in the midline as well as in the endline, the amount of business practices implemented is significantly higher after twelve months than after six months. In conclusion, training chances business management more in the longer run than in the short run, which suggest that it takes time to put training into practice. Profitability of total business activities increases over time, in contrast to, profitability of women’s most important business activity. Accordingly, different evaluation periods are necessary to determine the effectiveness of training when there are differences in treatment effects over time.
The main findings show that a gender and business training is effective for female microfinance clients of TYM by increasing women’s business knowledge, business practices and profits of business activities after one year. Although women highly appreciate the participation of husbands in training, inviting husbands does not have a major influence on the impact of training. Women in focus group discussions report that they are satisfied with the quality of the training and the majority will recommend it to others. Because the microfinance clients seems to be positive about the training and there are positive results in women’s business outcomes, a gender and business training is recommended to complement financial services. However, the majority of women are not willing to pay for this business training, consequently, conducting a cost-benefit analysis for TYM is necessary to decide whether TYM should offer training to other microfinance clients either.
Discussion
Business knowledge is measured by summing the amount of correct answers. However, the questions asked are similar in all three interviews, which might influence the results. Because both treatment groups spent more time together women with training are probably more likely to discuss questions and the correct answers. There are similar statements for requesting the amount of implemented business practices either. Karlan and Valdivia (2011) suggest that this might result in biased answers, because women aware of the relevance of business practices report differently by giving socially desirable answers. Therefore, extra statements are included in the two follow-up surveys. Besides, focus group discussions confirm the willingness to apply knowledge on business practices for treatment groups.
The participation rate at training modules is around 80% for women, which is relatively high compared to other business training studies (McKenzie and Woodruff 2014). However, the participation rate for men is relatively low, which reduces treatment variation between both treatment groups. Therefore, it would be interesting to estimate the effect of inviting husbands with a treatment-on-the-treated (TOT) estimator. Differences in participation rate will result in differences between the ITT estimator and the TOT estimator. Due to a low participation rate of men, the TOT can provide more insight in the effect of the participation of men at training meetings.
In order to reduce attrition, the interviewers tried to follow up on the women who dropped out TYM. Attrition can bias results when attrition is not random but differ between the treatment and control group. Yet, there are no differences in attrition between the treatment groups and the control group. Moreover, selection bias is not a problem here because current microfinance clients are random assigned into three groups. Age, years of schooling and interest in training of females is similar among the treatment groups. Moreover, the frequency and monthly sales of entrepreneurial activities do not differ. Thus, treatment groups are similar, which is a condition for internal validity.
multiple times results in more precise results, especially for outcomes varying over time such as sales and profits. Furthermore, multiple treatment groups will increase the validity of the results. Currently, the same business training is offered and evaluated in Uganda; this can provide more insight in the external validity of this study. Women included in the sample live in two different regions in Northern Vietnam: Hanoi and Vĩnh Phúc. The latter has a relatively high income per capita, which might affect the impact of a business training. Though, these two regions are representative for other regions in Vietnam.
ACKNOWLEDGEMENTS
I would like to thank Robert Lensink for his supervision and Nhung Vu for the set-up of the interviews and data collection in collaboration with TYM.
REFERENCES
Armendariz, B. and Murdoch, J., 2010. The economics of microfinance, second edition. The
MIT Press, Cambridge.
Armendariz, B. and Roome, N., 2008. Gender empowerment in microfinance. Munich Personal RePEc Archive. Working paper.
Berge, L. I. O., Bjorvatn, K. and Tungodden, B., 2012. Human and financial capital for microenterprise development: Short-term and mid-term evidence from a field experiment in Tanzania. Working paper.
Bjorvatn, K. and Tungodden, B., 2010. Teaching business in Tanzania: Evaluating participation and performance. Journal of the European Economic Association, 8, 2/3, 561-70.
Bruhn, M., Karlan, D. and Schoar, A., 2010. What capital is missing in developing countries? American Economic Review, 100, 2, 629-33.
Bruhn, M. and Zia, B., 2013. Stimulating managerial capital in emerging markets: The impact of business training for young entrepreneurs. Journal of Development Effectiveness, 5, 2, 232-66.
