Do Traditional Institutions Constrain Female Entrepreneurship? A Field Experiment on Business Training in

By and Seema Jayachandran and Rohini Pande∗

What constrains the entrepreneurial choices in knowledge might be rooted in social norms of poor women? Do traditional institutions pose about what females are taught. unique barriers to business growth and prof- This paper explores how traditional religious itability for female-run enterprises? and caste institutions in India that impose re- The explosion of microfinance programs, strictions on women’s behavior influence their which typically target poor female en- business activity. Our analysis makes use of trepreneurs, has drawn attention to these a field experiment in which a randomly se- questions. Indeed, one view is that inadequate lected sample of poor self-employed women were access to credit prevents women from under- trained in basic financial literacy and business taking high-return business activities. However, skills and encouraged to identify concrete finan- one recent empirical study finds low returns to cial goals. The sample is relatively homoge- capital in female-run micro-enterprises (Suresh nous in terms of socio-economic status (e.g., ed- De Mel, David McKenzie and Christopher ucation). However, differences in religion and Woodruff 2008). caste mean that they face very different tra- Thus, another view is that the primary bar- ditional restrictions on mobility and social in- rier to female entrepreneurial success is limited teractions. Muslim women face the most re- demand for rather than supply of credit, with strictions. Among Hindu women, upper castes poor women lacking high-return means of ex- (hereafter, UC) face significantly more restric- panding their businesses. For instance, due to tions than scheduled castes (hereafter, SC), the gender differences in education or business net- lowest group in the caste hierarchy. works, women might be relatively uninformed In general, the returns to entrepreneurship about investment opportunities and untrained should be highest for those least fettered by in basic cost-benefit analysis (Dean S. Karlan conservative social norms. However, this need and Martin Valdivia 2008). A second possibil- not be the case for an intervention that pri- ity is that norms governing women’s roles in so- marily influences women’s knowledge of business ciety limit women’s perceptions about what is practices and aspirations. If traditional norms achievable in the workplace.1 Even differences about gender roles can be challenged, or if they mainly work to limit women’s exposure to and ∗ Field: , efi[email protected]. knowledge of business opportunities, then re- edu. Jayachandran: , jay- turns from training may be higher for women [email protected]. Pande: Harvard University, from more restrictive social groups. rohini [email protected]. We thank Manasee Desai, Katherine Durlacher, Sarah Moshary, Natalia Our results provide some support for both Rigol, Mallika Thomas, and Divya Varma for theses: Among Hindu women, training increased excellent research assistance, the staff of SEWA borrowing and business income for those fac- for collaboration and the Exxon Mobile Educating ing more restrictions, i.e., UC women. However, Women and Girls Initiative (administered through Muslim women failed to benefit from the train- WAPP/CID at Harvard), Center for Microfinance ing program. We interpret these patterns as sug- and ICICI for logistical and financial support. 1Another possibility is that women are unable to gestive of a non-monotonic relationship between pursue high-return activities because low bargaining social restrictions and the ability to benefit from power in marriage limits their control over finances. business training. 1 2 PAPERS AND PROCEEDINGS MAY 2010

I. Gender norms in India lum, developed by Freedom from Hunger and used widely around the world, covers basic ac- The Hindu caste system developed as an ex- counting skills, interest rates and life-cycle plan- tremely hierarchical social system. Its defining ning.It emphasizes financial prudence and en- principles include the ideas of purity and pollu- courages women to avoid excess debt, save more tion. UCs maintain purity by avoiding sexual and reduce “frivolous” spending. More recently, relations, marriage and, in extreme cases, con- SEWA Bank started a second five-day course tact with lower castes. Premised on men be- that teaches business skills such as cost reduc- ing a source of pollution, restrictions are placed tion, investment, and customer service. on women to limit contact with men other than In collaboration with SEWA Bank, we de- their husbands. Requirements include that a signed a streamlined two-day training module married woman remain veiled, not remarry if that combined elements of their financial liter- widowed, not interact with older men, and have acy and business skills curricula, and added new restricted mobility outside of her house. These material focused on aspirations. The aspirations norms – particularly the latter two – signifi- component included a short film showcasing suc- cantly restrict female labor force participation. cessful SEWA members who used good financial Maintaining purity by minimizing contact practices to bring themselves out of poverty. As with lower castes is less relevant for SCs, who homework after the first day of training, partic- rank low in the hierarchy. In addition, greater ipants filled out a worksheet identifying a finan- poverty implies increased reliance on female cial goal they wanted to achieve over the next wage-earning (Karin Kapadia 1997). As a result, six months, and on the second day broke it down SC women face fewer social restrictions and, into smaller short-run steps. by virtue of being independent earners, enjoy For the experiment, 636 women were ran- greater financial autonomy and increased control domly drawn (in two phases) from the pool of over household financial decisions relative to UC SEWA Bank customers ages 18 to 50 who were women (Joan P. Mencher 1988). Notably, these both active savers within the past two years and restrictions on female autonomy among UCs are employed. Two thirds of these women were ran- not limited to the wealthy (Mukesh Eswaran, domly assigned to the treatment group.2 Bharat Ramaswami and Wilima Wadhwa 2009). Relative to Hindus, Muslims in India place Women assigned to the treatment group were more restrictions on women’s contact with peo- approached in their homes and recruited to at- ple outside but not within the sphere of kinship. tend a particular training session with seven Because Muslim women are entitled to a share in other participants at the SEWA branch nearest family real estate, controlling their relationships to them. For data collection and analysis pur- with males outside the family can be crucial to poses, women in the control group were also as- the maintenance of family property and prestige. signed but not recruited to a particular training session at their nearest SEWA branch, allowing II. Intervention and study design us to cluster standard errors by session. Our analysis sample comprises the 597 women We conducted a business training interven- who were successfully surveyed at follow-up and tion in conjunction with SEWA Bank, which could be categorized into sub-castes based on is based in the city of Ahmedabad in western surname.3 We categorized women into three India. Its 170,000 member-clients are primar- broad social groups: Muslims, Hindu SCs, ily poor women who work in the informal sec- and Hindu UCs (non-scheduled castes includ- tor (for example as incense-stick makers, tailors, ing other backward castes). In addition, we and vegetable vendors). SEWA Bank offers scored how restrictive each Hindu and Muslim these women a wide array of financial products. All clients are required to have a savings ac- 2The randomization was stratified by sampling count, and roughly a quarter of clients have ever phase and SEWA branch. taken out a loan from SEWA Bank. 3The survey attrition rate (5.3%) is similar across For several years, SEWA Bank has run a five- experimental groups. We were unable to assign caste day financial literacy training. The curricu- to seven women. VOL. 100 NO. 2 TRADITIONAL INSTITUTIONS AND FEMALE ENTREPRENEURSHIP 3 sub-caste was in regard to five norms governing Average education and family size are almost women’s behavior: ability to socialize alone, re- identical for Muslims, UCs and SCs, and SC quirements to cover the face or wear a veil, abil- women have higher household income and are ity to speak directly to elders, ability to leave slightly more likely to own a business, though the house or neighborhood alone, and ability to the differences are statistically insignificant. remarry. We created an index ranging from 0 to 5 equal to the number of norms for which the sub-caste was highly restrictive. Table 1—Baseline characteristics Figure 1 shows the value of this index across UC SC Muslim the three social groups in our sample. SC women Education 6.33 6.62 6.36 face the fewest restrictions, followed by UC and Income 4852.6 5694.6 5189.5 Muslim women. UCs are 50% more likely to Household size 5.25 5.25 5.42 face severe social restrictions as SCs, and Mus- Owns business 0.27 0.31 0.26 lims, in turn, have over twice the rate of severe N 346 70 181 restrictions as UCs. Notes: Asterisks denote significant differences in means between the group and UCs. In all tables * denotes significant at 5% level ** significant at 1% 2 1.7 level *** significant at 0.1% level. 1.6

1.2 III. Effects of business training 0.75 0.8 0.48 Our estimation strategy exploits the random 0.4 assignment to treatment, i.e., being invited to a training session. We examine the impact of 0 attending the training on economic outcomes, Muslims Upper Scheduled instrumenting for attendance with whether the participant was in the treatment group. This caste caste IV specification provides TOT estimates. We Hindus Hindus separate out the differential effect of training by social group by interacting the training dummy with indicators for being a SC and a Mus- Figure 1. Number of social restrictions lim. Outcome variables come from a survey conducted on a rolling basis four months post- training. The intervention trained 289 women in The results in Table 2 reveal that training led 57 two-day training sessions conducted from to a significant increase (13 percentage points) September 2006 to April 2007. Program take-up in the likelihood of taking out a loan within was high, with over 70% of those invited choos- four months of training (Borrowed) among UCs, ing to attend. SCs, the least restricted group, who are the omitted category. UCs who at- were nearly one third more likely to attend the tended training took out loans at nearly twice training than Muslims and UCs. The results the rate as UCs in the control group. Mean- on take-up, which are also the first stage of our while, we cannot reject that there was no effect treatment-on-the-treated (TOT) results, are re- on borrowing among SC or Muslim women. ported in the online appendix. When we estimate the training effect on Baseline characteristics are balanced across the likelihood that a woman reported problems the control and treatment groups (see online ap- managing her debt, we find no overall effect and pendix). Table 1 compares baseline characteris- no differential effects across social groups (see tics across social groups. Women in our sample appendix). Administrative bank data on loan are strikingly homogenous across social groups: default confirm this result. Thus, the training 4 PAPERS AND PROCEEDINGS MAY 2010 does not seem to have induced UC women to amount of income earned over the past week, the borrow beyond their means. point estimates suggest an increase of around 30%, but the results are too imprecise to draw conclusions. Together with the loan results, Table 2—Treatment effects on finances this pattern suggests that the training encour- Borrowed Savings aged UCs to start or expand their microenter- Trained 0.13* -315.32 prises. Further supporting these results, we find (0.07) (492.83) that training led UCs but not other women to Trained×SC -0.16 444.71 talk more frequently with family members about (0.14) (540.02) business plans (T alk Business). Trained×Muslim -0.14 317.51 (0.11) (630.07) SC 0.04 -298.12 Table 3—Treatment effects on business (0.10) (301.87) Talk Muslim 0.04 -46.50 Any Income Business (0.06) (333.99) Trained 0.19** 0.17* Mean of dep. var. 0.17 277.59 (0.09) (0.10) Trained×SC -0.37*** -0.51*** Notes: Standard errors clustered by training (0.14) (0.15) session. Regressions include SEWA branch, Trained×Muslim -0.22* -0.35*** treatment month, and sampling phase fixed effects. (0.12) (0.12) Mean of dependent variable is for UCs in the SC 0.25*** 0.35*** control group. N=597. See the appendix for further (0.08) (0.09) details. These notes also apply to Table 3. Muslim 0.09 0.18** (0.07) (0.08) Mean of dep. var. 0.79 0.69 Savings during the past month (Savings, measured in rupees) show no significant differ- ences across treatment and control, though the point estimates again go in opposite directions for UCs compared to SCs and Muslims, with IV. Discussion UCs substituting away from saving, consistent with their increased borrowing. Given the similarity in education, household To more directly measure the effect of training wealth and types of businesses across social on business activity, in Table 3 we examine the groups, the difference across groups in their re- women’s personal business activity, including an sponse to training is stark. It is made even more indicator of whether she reports any personal striking by the fact that data collected during labor income over the past week (Any Income) training reveal identical patterns of business and and an indicator of whether she reports talking financial goals across social groups. to family about business plans. Although we One possibility is that imbalances in treat- sampled on being employed, many of the women ment assignment within social groups are re- do not report regular earnings: in the control sponsible for the observed patterns. However, group, only 80% of women report any earning while there is higher baseline business ownership in the past week. among UC treated relative to control women and We observe a positive and significant effect the opposite is true for Muslims and SCs, treat- of the training on business income among UCs ment differences persist even when we control for only, suggesting that the new loans were put to- this (and/or other) observables (see appendix). ward business investments. The estimated ef- Another possibility is that differential treat- fect of training on earning any business income ment effects reflect higher program take-up is highly significant among UCs, indicating a among SCs. The training may have attracted a 25% increase in the likelihood that a woman en- selected sample of UCs who were especially re- gages in labor market activity. When we look at sponsive to training. However, observable char- VOL. 100 NO. 2 TRADITIONAL INSTITUTIONS AND FEMALE ENTREPRENEURSHIP 5 acteristics appear to be equally balanced across to easily change their aspirations or activities. caste among the subset of women who were Even with more knowledge or higher aspira- treated relative to the ITT sample, indicating tions, the most restricted women might face too that heterogeneity in take-up does not seem many social strictures to avail themselves of en- to explain the heterogeneous treatment effects trepreneurial opportunities. (though we cannot rule out unobservable differ- Our business counseling program significantly ences across groups). Furthermore, take-up can- reduced the business income gap between social not explain observed differences between UCs groups. Thus, another reading of our results is and Muslims, who had similar attendance rates. that modernization – in the absence of interven- If the explanation for UC women being espe- tions that counteract traditional norms – may cially responsive to training is that social restric- yield greater benefits for women lower in the tions caused them to have knowledge deficits caste hierarchy, a point also made by Kaivan or the training allowed them to challenge so- Munshi and Mark R Rosenzweig (2006). This cial norms that were distorting their business view, however, assumes that gender norms for practices, then an important question is why lower castes will continue to be less restrictive. Muslims, who face the highest degree of re- If, instead, modernization heightens sanskritiza- strictions, did not respond more to the train- tion – the desire of lower castes to emulate upper ing than SCs did. One possibility is that, al- castes – and SCs increasingly adopt the gender though restrictions are greater for the average norms of UCs, then economic growth may fail Muslim woman than the average SC woman in to emancipate women to the same extent. Ahmedabad, there is little difference in restric- tions across Muslim and SC members of SEWA REFERENCES Bank. Unfortunately, without individual-level data on restrictedness, we cannot test this story. De Mel, Suresh, David McKenzie, and Christopher Woodruff. 2008. “Returns to Another possibility is that Muslims in Capital in Microenterprises: Evidence from a Ahmedabad, which has a history of religious Field Experiment.” Quarterly Journal of Eco- tension, face considerable discrimination in the nomics, 123(4): 1329–1372. marketplace, which business training could not undo. Alternatively, religious restrictions on Eswaran, Mukesh, Bharat Ramaswami, interest-bearing loans might explain why de- and Wilima Wadhwa. 2009. “Status, Caste, mand for credit did not increase among Muslims. and the Time Allocation of Women in Rural In- However, this explanation is unlikely given that dia.” mimeo, University of British Columbia. the rate of borrowing is similar across Muslims Kapadia, Karin. 1997. “Mediating the Mean- and Hindus in the control group. ing of Market Opportunities: Gender, Caste A final possibility worth mentioning is that, and Class in Rural South India.” Economic and although Muslim women in this setting face a Political Weekly, 35(52): 3329–3335. high degree of social restrictions, for the norms Karlan, Dean S., and Martin Valdivia. that most directly affect business activity – abil- 2008. “Teaching Entrepreneurship: Impact Of ity to leave the home alone and talk to strangers Business Training On Microfinance Clients and – we find that the rate of being highly restricted Institutions,.” mimeo, Yale University. is in fact lower for Muslims than UCs. However, Mencher, Joan P. 1988. “Women’s Work and for these two norms, only about 5% of women Poverty: Women’s Contribution to Household in our sample were coded as highly restricted, Maintenance in South India.” In A Home Di- and average restrictedness is still considerably vided. , ed. Daisy Dwyer and Judith Bruce. higher among Muslims than UC or SC Hindus. Stanford, CA:Stanford University Press. Bearing in mind these caveats, a prima facie Munshi, Kaivan, and Mark R Rosen- explanation for our results is non-monotonicity zweig. 2006. “Traditional Institutions Meet in the effect of social restrictions: The train- the Modern World: Caste, Gender and School- ing helped women whose businesses had been ing Choice in a Globalizing Economy.” Ameri- held down by social restrictions, but women sub- can Economic Review, 96(4): 1225–1252. ject to extreme restrictions had too little agency