Gender, small firm ownership, and credit access: some insights from India

Unlike most of the previous work relying on surveys of a small sample of firms, our study is based on Census data provided by the Ministry of Small Scale Industries, Government of India. We employ unit-level data drawn from the Fourth round of the Indian Micro, Small and Medium Enterprises (MSME) Survey for the year 2006–2007.³ This rich dataset contains information pertaining to 2.24 million small firms belonging to the registered and unregistered sectors. The Ministry of Small Scale Industries defines enterprises with investment of up to INR 2.5 million in plant and machinery as micro enterprises, enterprises with investment between INR 2.5 million and INR 50 million as small firms, and enterprises with investment between INR 50 million and INR 100 million as medium enterprises. The dataset contains firms belonging to manufacturing, services and repairing and maintenance sectors. In this study, we confine the analysis to the sample of firms belonging to the manufacturing sector. The database provides information related to the gender of the owner, firm’s year of initial production, the sector of its operation, the gross output for three consecutive financial years, ownership type, export, loan status, quality certification, and net worth as well as information on reporting status on financial accounts by the firm.

Like any typical enterprise-level data from developing countries, our dataset also contains missing values and outlier observations that are likely to bias our estimates. Therefore, we clean the original dataset to take care of the missing observations and outliers. We follow certain procedures while undertaking the data cleaning process: First, we omit those firms which did not respond to one or more key questions. Second, we exclude those firms with seemingly unrealistic information such as missing, zero or negative output, missing labor, and capital stock values. These elimination norms reduced the number of firms in the dataset to 1,157,877 from 1,313,210 (about 12% of the firms in our dataset we have filtered out). We use suitable deflators (wholesale price indices) to make price corrections to the reported data on output, intermediate inputs, and value-added and fixed assets. Data related to wholesale price indices have been obtained from the report on Index Number of Wholesale Prices in India, published by the Office of the Economic Advisor, Ministry of Industry, Government of India.

Table 1 compares the mean values of our main variables of interest, by gender of the owner. It is clearly evident from the table that there exist considerable differences in firm characteristics between male-owned and female-owned firms. Compared to the male-owned firms, the average production is considerably lower in female-owned firms. The average log of output for male-owned firms is 12.63 and female-owned firms is 11.25. The female-owned firms are younger and smaller in size and mostly located in urban areas as compared to male-owned firms. As regards loan availability, male business owners are more likely to obtain credit (12%) than women (8%). Looking at the firms’ presence in the export market, male business owners are much more likely to enter export markets than women-run businesses. Thus, in terms of raw averages, female-owned firms clearly perform much worse than male-owned firms.⁴

In this section, we discuss the results regarding firms’ performance/productivity with respect to the gender of the owner/manager. Table 2 presents the baseline results. For each performance/productivity measure, we estimate three specifications based on the extent and involvement of women in the ownership and management of firms, i.e., women as owner (column 2), women as manager (column 3), and women as owner and manager (column 4). In all the specifications, we control for age and its square as well as the size of the firms along with 2-digit SIC industry dummies and state dummies to account for unobserved heterogeneity at the industry and state levels. We include age and its square term as control variables as the differences in the age of the firms owned by males and females could explain the observed gaps in the performance of male and female entrepreneurs. Women have been increasingly engaging in entrepreneurial activities recently; hence, their ventures will be much younger than the ones operated by their male counterparts. Their lower experience in managing entrepreneurial concerns might well explain their underperformance as entrepreneurs.⁶ Female-owned firms in our dataset are indeed much younger than the male-owned firms and the firms operated by male entrepreneurs are on average 4 years older than the female-owned firms.⁷

Women entrepreneurship is largely skewed towards smaller-sized firms, and this gap in firm size, at least partially, could explain the existence of a gender gap in firms’ performance (Marlow and McAdam 2013).⁸ It is argued that majority of women entrepreneurs are often in business because running a small enterprise allows them to bring in additional income with little additional effort and they are unlikely to expand or invest in their businesses. Previous research highlights that intention and motivation of women entrepreneurs are different from male-led businesses. Women-led businesses may be less interested in growth objectives but may be more motived by the flexibility and the personal fulfillment it offers (Klapper and Parker 2011; Morris et al. 2006). Some studies also show that women tend to display greater risk aversion, which leads them to restrict investment in their business concerns, thereby limiting the growth of their firms (Barber and Odean 2001, Dohmen et al. 2011). However, the differences in firm size may also be an outcome of the differences in survival rate of male and female-owned enterprises. If the survival rate of female-owned firms differs from that of the male-owned firms, we would expect the female-owned enterprises to be more skewed towards smaller-sized firms. Though our dataset does not permit us to explore these two questions, it is, however, possible to analyze whether the differences in firm size explain the gender gap in firm performance. In our dataset too, about 97% of firms owned by women entrepreneurs are micro enterprises as against 85% for male-owned firms. We therefore include firm size as a control variable in our specifications.⁹

Our result clearly suggests that there exists a significant gap in performance between firms owned by males and those owned by females. Table 2 shows that in terms of size of the firms¹⁰ (irrespective of using output or employment), there exists a significant gender differential in performance even after controlling for age, firm size (using three dummy variables for micro, small, and medium firms), caste of the owner (using three dummy variables for scheduled caste, scheduled tribe, other backward castes (general as the base category), dummies for type of organization, and industry and state effects. The annual output of an average female-owned firm is 32% points less than the annual production of an average male-owned firm. The gap is much larger for women as manager (51.9%) and women as owner and manager (56.7%). Across all specifications, all measures of performance, and measures of entrepreneurship, male entrepreneurs do perform better than female entrepreneurs. Using the number of workers as a proxy for firm size, we find an average female-owned firm to be 13.8% points smaller than that of an average male-owned firm. Our findings are robust to our alternate measures of women entrepreneurship (women as manager and women both as manager and owner) and the gaps are found to be much larger for these alternate measures.¹¹

Our results remain invariant when we consider efficiency/productivity instead of firm size. We observe that the TFP for the female-owned firms is significantly smaller than their male counterparts. Same holds true when we use labor productivity as an alternative—the results remain unaltered with respect to the role of female: as owner, as manager, or as both. Note all of our specifications include age and its square, size of the firms, and industry and state effects. Therefore, age and size are not the driving factors behind the findings that woman-led firms perform worse than those of their male counterparts.

Another explanation cited in the literature (Bardasi et al. 2011) related to the underperformance of female entrepreneurs is their predominance in low performing and less productive industries. We proceed as follows: we compute an index that measures the representation of women-led businesses in each sector and then compare the performance of firms in female-dominated sectors with those that are male dominated. Figure 1 demonstrates the representation of women businesses according to two-digit industry classifications. We observe that the ratio is greater than one in just three sectors, namely wearing apparel, textiles, and tobacco products.

We proceed to examine the relative performance of male- and female-owned businesses in female-dominated and male-dominated sectors by estimating Eq. (1) as stated above. Equation (1) helps us to differentiate between two implicit hypotheses of lower performance of the industries and lower performance of women entrepreneurs within these industries. Table 3 provides an insight into this explanation. We concentrate on two measures of performance: size (proxied by output) and productivity (proxied by TFP). The results reported in Table 3 confirm that the partial explanation for the underperformance of female entrepreneurs can be derived from the predominance of women enterprises in less efficient sectors. Across all estimations, the estimates of coefficients of Female and Female × Female Dominant Sector are negative and significant. The estimated coefficient of Female dominant sector yields a negative sign across all model specifications but is significant only in two specifications (columns 3 and 4). The statistically significant results in columns 3 and 4 indicate that firms that operate in sectors that are female-dominated are relatively smaller (− 47%) and less productive (− 21%) than those firms that operate in male-dominated sectors. This perhaps lends evidence to the fact that female entrepreneurs tend to focus more on sectors that are dominated by firms which are on average smaller and less productive. The negative and significant coefficients of the interaction term Female × Female Dominant Sector suggest that the choice of sectors (female-dominated versus male-dominated) does influence the gap between male and female entrepreneurs in size and performance of firms. Together, the negative and significant coefficients of Female and Female × Female Dominant Sector indicate that female-run firms operating in female dominant sectors are significantly smaller and less efficient than those that operating in male-dominated sectors. Considering the results in Tables 2 and 3 together, we observe that this skewed representation of women in certain sectors does not seem to fully explain the underperformance of firms owned by women entrepreneurs. This is clearly evident from the large and significant difference in the performance and size of male-owned and female-owned firms, even after controlling for the choice of the sector of operation.¹²

Our results fail to explain why female-owned firms indeed would like to be based in sectors which are poorly performing. An interesting policy question is whether women entrepreneurs are entering sectors with smaller and less efficient firms by “choice” or by “force?” If it is by choice, what are the key factors that attract them towards these sectors? If they are forced to work in these sectors, what are the key elements that exclude them from other sectors? Our dataset, however, does not permit us to undertake an analytical exercise for searching answers to these questions. However, our dataset allows us to explore whether their decision to operate in these sectors is determined by credit availability. The investigation of this issue assumes greater relevance in our present line of enquiry since various studies acknowledge severe barriers for women entrepreneurs in access to credit (Asiedu et al. 2013; Hansen and Rand 2014).

To examine the gender discrimination in accessing credit, we estimate a logit equation as given in Eq. (2). The estimation results are reported in Table 4. In all the specifications, the coefficient associated with female as owner, our main variable of interest, is negative and statistically significant. According to the estimates, the odds of getting a loan for female owners is about 10.2% lower than for male owners, for all loans. The odds continue to be lower if female is manager or both owner and manager; however, the effect is not statistically significant. Considering the smaller proportion of firms that receive loans in the MSME sector, this is a fairly large number, indicating the presence of substantial gender difference in financial constraints. The impact is lower for institutional loan: the odds of getting an institutional loan for female owners is about 6.4% lower than for male owners. The gender differential is substantially higher for non-institutional loan compared to institutional loan and this is particularly interesting given that small firms depend more on non-institutional loans rather than institutional loans. For non-institutional loan, the coefficient associated with Female (as owner, as manager and as both) is negative and statistically significant. Turning to the location of the firm, the coefficient is not significant. Exporting firms and firms that are part of a cluster have higher odds of loan approval, which implies that lenders consider it less risky to lend to those firms that market their products abroad and that are part of a larger network. Consistent with our prior expectations, we also find that entrepreneurial ability and financial literacy do matter in eliciting a positive response from the lender. Our proxies for ability and literacy (account maintenance and possessing a quality certificate) have yielded positive coefficients, showing that lenders’ satisfaction on entrepreneurial abilities increases the likelihood of loan approval.¹³ There is also a clear direction from our results that the probability of obtaining a loan is significantly higher among firms that show a better performance in terms of profitability and productivity, since the coefficients of all measures of firm performance are positive and most of them are significant across all our specifications. Age variable exerts an interesting pattern revealing a U-shaped relationship. Following Haans et al. (2016) and Lind and Mehlum (2010), we test the presence of a U-shaped relationship. In all our specifications, we notice the presence of a U-shaped relationship with respect to age and the turning point turns out to be around 39 years for all firms and those with institutional loans whereas it hovers around 29 years in case of firms with non-institutional loans.¹⁴ Furthermore, our results are also robust to alternate measures of women entrepreneurship. If we focus on the variable identifying the gender of the manager rather than female participation in ownership, the results again show evidence of gender-based discrimination in the credit market, once we condition for firms characteristics. The same result holds even when we use the variable capturing the dual role of women as owner and manager. The results reported in Table 4 remains qualitatively similar if we use output of 2 years back instead of current and lagged output. Use of better performance/productivity measures in terms of log of profit and labor productivity also did not change our results regarding the gender variable qualitatively.¹⁵

Does the reported result in Table 4 suffer from sample selection bias? Earlier studies highlight that firms often self-select not to apply for a loan (Cavalluzzo et al. 2002; Bardasi et al. 2011). Self-selection for not applying for loans may arise due to the absence of need for external finance and some of the applicants may fear a rejection. Therefore, it is essential to correct for the self-selection problem in our empirical analysis. Our empirical strategy to overcome the self-selection problem involves identifying firms as constrained and unconstrained with credit demand (Bigsten et al. 2003). We classify sample firms as “constrained” if they report a shortage of capital, while those firms receiving loans and which did not report any capital shortage were classified as “unconstrained.”¹⁶ To address the selection issue, we employ a bivariate probit model. We use lagged output growth as the exclusion restriction for identification purpose assuming that it influences demand for credit but not the final outcome.¹⁷ Lagged output growth is defined as the logarithmic difference of output between 2004/2005 and 2005/2006. We believe that lagged output growth would reduce the demand for credit in the current year as it would help firms to manage their operations with internal funds better. Results of our bivariate probit regression estimations after correcting for self-section are reported in Table 5. The Wald test of independent equations rejects the null hypothesis (H₀: ρ = 0), validating our model specification.¹⁸

The coefficient of our variable of interest Female is negative and significant suggesting that the likelihood of receiving a bank loan is higher among male entrepreneurs as compared to female entrepreneurs.¹⁹ The estimated marginal effects show that a female owner possesses a 1% lower probability of getting a loan, which indicates they faced shortage of funds as an obstacle for satisfactory performance compared to their male counterparts. The results remain almost the same when we use female acting as both owner and manager. Other controls have maintained more or less the same sign and significance as the results for our baseline econometric specification. Our findings are thus essentially robust with regard to concerns arising from self-selection, and we conclude that there exists clear evidence to support the existence of gender-based discrimination in the credit market. Further, we carry out the same analysis with respect to institutional and non-institutional loans. The results regarding gender discrimination in terms of obtaining a loan conditional on credit demand remains valid.²⁰

Given the existence of gender gap in access to finance, we next explain this gap using the modified version of the Blinder-Oaxaca decomposition technique. The decomposition technique is a kind of a matching where two groups are matched on a one-to-one basis. This decomposition technique helps in providing an answer to the extent to which differences in observable group characteristics can explain gender gap in access to credit. The extent to which one may attribute gender gap in credit access will depend on the choice of the reference group. The standard practice is to use the relatively more advantaged group as the reference (male-owned firms in our case) and show discrimination against less advantaged female-led firms. We decompose the gender differences in the probability of obtaining loan, using indicators of female involvement (ownership, management, ownership and management). The results of the decomposition (Table 6) show that female owners face 2% lower probability of getting a loan as compared to male entrepreneurs. The decomposition output, the gap in probability of getting a loan across the gender, is divided into three parts. The first part (endowment) reflects the mean increase in female’s success probability if they had the same characteristics as male, whereas the second part (coefficients) shows the change in female’s success probability when applying the men’s coefficients to the female’s characteristics. The third part is the interaction term that measures the simultaneous effect of differences in endowments and coefficients. For female manager (fourth column, Table 6), the probability of getting a loan is 3.3% lower than males and it is mainly due to the endowment effects. We observe that the probability of not getting a loan increases to 4% when female acts as both owner and manager. Although not reported, we also perform the same analysis for institutional loans.²¹ For female, the probability of getting an institutional loan is 1%, 2.9%, and 2.4% lower than males for female owner, female owner and manager, and female manager respectively, and it is again mainly due to the endowment effects.