Finance, Gender, and Entrepreneurship: India’s Informal Sector Firms

Abstract

How does informal economic activity respond to increased financial inclusion? Does it become more entrepreneurial? Does access to new financing options change the gender configuration of informal economic activity and, if so, in what ways and what directions? We take advantage of nationwide data collected in 2010/11 and 2015/16 by India’s National Sample Survey Office on unorganized (informal) enterprises. This period was one of rapid expansion of banking availability aimed particularly at the unbanked, under-banked, and women. We find strong empirical evidence supporting the crucial role of financial access in promoting entrepreneurship among informal sector firms in India. Our results are robust to alternative specifications and alternative measures of financial constraints using an approach combining propensity score matching and difference-in-differences. However, we do not find conclusive evidence that increased financial inclusion leads to a higher likelihood of women becoming entrepreneurs than men in the informal sector.

JEL classification:

• O12
• G28
• L26

Keywords:

• Entrepreneurship
• financial constraints
• gender
• informal sector
• difference-in-differences
• India

1. Introduction

One of the salient features of under-development is the existence of a large informal economy characterized by a large mass of non-entrepreneurial firms (De Vreyer & Roubaud, 2013). These firms are typically household units, which are survivalist in nature, with limited prospects for growth (Grimm, Knorringa, & Lay, 2012).¹ Many of these household units are run by women, who seem to face greater hurdles than men in making the transition to entrepreneurial firms (for a review, see Jennings & Brush, 2013). A large literature has attempted to understand why so few firms expand beyond the household unit, consequently becoming more productive, and why we see gender differences in entrepreneurship in the informal sector (see Chen, 2012; Fields, 2019). Credit constraints, a consequence of the widespread failure of credit markets in developing countries, are widely regarded as a key constraint to entrepreneurship development (Kerr & Nanda, 2011). Yet systematic evidence on the causal role of finance in determining entrepreneurship in general, and female entrepreneurship in particular, is lacking. This paper focuses on the role that finance/credit/banking access plays in expanding informal sector enterprises beyond family and household units.² We ask: (1) Is there a causal link between the availability of finance and the likelihood of informal firms making the transition from strictly family organizations to entrepreneurial ones, i.e. from those employing only family members to those hiring outside labour? (2) Does this differ between female- and male-owned enterprises?

We look at India’s unorganized sector, including manufacturing firms and unincorporated non-agricultural enterprises; we use the terms ‘unorganized’ and ‘informal’ interchangeably.³ Approximately 75 per cent of manufacturing employment and 17 per cent of manufacturing output are in the sector; about 86 per cent of these firms are family owned (see Raj & Sen, 2016 for a review of the vast literature on informality in India). Our strategy lies in taking advantage of nationwide data collected in 2010/11 and 2015/16 by India’s National Sample Survey Office (NSSO) on unorganized (informal) enterprises. These are repeated cross-sections of unincorporated non-agricultural enterprises. Informal firms are quite heterogeneous. In line with the NSSO classification, Raj and Sen (2016) categorize firms in this sector into three types: very small pure household enterprises (PHEs), mixed household enterprises (MHEs) that are somewhat larger, using both family and at least one non-family labourer, and larger non-household enterprises (NHEs) employing mostly non-family labour.⁴ We define a firm as entrepreneurial if it employs at least one hired worker, in addition to members of the proprietor’s family, i.e. we merge the MHEs and NHEs (both employ hired workers) into one, calling them entrepreneurial firms.

We find that lack of access to finance constrains the transition of firms in the informal sector in India. This effectively means that as the constraint weakens, firms are willing and able to expand hiring beyond family workers. However, we do not find clear gender differences in transition with the easing of finance constraints: female entrepreneurs are no more likely to join the entrepreneurial side of the sector than male entrepreneurs. Our findings are upheld when controlling for the endogeneity of the finance constraint using instrumental variables. We also supplement our main identification strategy with an alternate approach, combining propensity score matching (PSM) and difference-in-differences (DID), which also reinforces our main finding of the positive role of finance in promoting entrepreneurship among informal firms in India.

The paper is in six sections. Next, we provide background, including brief discussions of the literature on the roles of gender and finance in informal sector firm transition, and the relevant policy environment in India. In Section Descriptive Statistics, Empirical Strategy, and Identification, we discuss the data, including descriptive statistics, and outline our empirical strategy and approach to identification. Section Results takes up our estimation results while Section Robustness Test: PSM-DID Results outlines a critical robustness check. Section Conclusions concludes.

2. Related literature, background, and policy environment

In this section, we first examine the lessons that we draw about the firm transition from the literature on gender roles and financial constraints in development. We then discuss the background to this study and the relevant policy environment in India in subsequent subsections.

2.1. Related literature

Financial constraints are a factor limiting a firm’s growth (Oliveira & Fortunato, 2006). Investment in fixed assets is less likely for firms with financial difficulties (Ojah, Gwatidzo, & Kaniki, 2010; Winker, 1999). Smaller firms face greater financial limitations than larger ones – for example, Beck, Demirgüç-Kunt, Laeven, and Levine (2008) find that financial constraints are associated with a 10 per cent decrease in small businesses' growth.

Although studies show the clear role financial constraints play when starting a business, there is less consensus on the role of credit access in the growth and subsequent performance of small businesses. Some studies argue that access to credit helps small firms grow faster (Brown, Earle, & Lup, 2005) and is important for the development of micro-enterprises (Woodruff & Zenteno, 2007). Other studies point to a lesser role played in enterprise growth by financial constraints, finding little convincing evidence that access to formal credit is an important influencing factor (Daniels & Mead, 1998; Johnson, McMillan, & Woodruff, 2002). It seems that for enterprise growth, access to finance is needed but by itself is not enough (Nichter & Goldmark, 2009).

Gender, financing, and their interaction seem to play important roles in the development of entrepreneurship in the informal sector. The literature on gender and credit access has expanded rapidly, mostly after the appearance of the World Bank’s Enterprise Survey (WBES) of 2006. It is difficult to use these data for studying gender issues, but a series of papers have rather inventively developed ways to use its information on gender to analyse linkages between gender and credit. Unfortunately, the evidence that gender differences in access to finance make a difference in firm development is not conclusive (Presbitero, Rabellotti, & Piras, 2014).⁵ As we will soon see, our data overcome many problems faced by users of the WBES. This line of research also deals with firms larger than ours and, generally, not in the informal or unorganized economy.

2.2. Background

The incidence of entrepreneurship in India’s unorganized economy has many elements: owner’s gender, financial constraints faced by firms, owner’s caste, etc. In India, the unorganized sector has played an honoured role in development strategy. Government policy has encouraged small firms both in the unorganized and organized sectors. As part of the effort to promote industrial decentralization and increase employment, support for small firms was an important element of the Industrial Policy Resolution of 1956, reiterated in December 1977 in the Industrial Policy Statement. Such enterprises have generally been exempted from excise and other taxes, enjoyed protection from larger firms which were often restricted from producing competing products, given preferential pricing (for example, in sales to public sector firms), and so on (Gang, 1992).

2.3. Policy environment

India’s central bank, the Reserve Bank of India (RBI), follows specific policies aimed at expanding access to banking services.⁶ In 2005 the RBI began classifying districts (state subdivisions) as under-banked if their population per bank branch was greater than the national average. Various policies then encouraged bank branch expansion in these districts. In 2011, banks were instructed to open at least 25 per cent of their total branches in a year in unbanked rural centres—a 4:1 norm as against the previous 1:4 norm (Chavan, 2020). They were requested to plan for financial inclusion and to set targets for opening branches, small-sized savings deposit accounts, and debit cards, and for providing small-sized overdrafts. The period between 2010/11 and 2015/16 was one of rapid expansion of banking availability in India aimed particularly at the unbanked, under-banked, and women (see Young, 2021). This is evident from the decline in a number of under-banked districts in India, from 355 in 2010/11 to 344 in 2015/16. The population covered per branch, which was 13,027 in 2010/11, dropped significantly to 8,683 in 2015/16 (Table 1). Two main Indian government programmes were the Bharatiya Mahaila Bank (Indian Women’s Bank) started on 19 November 2013, and the Pradhan Mantri Jan Dhan Yojana (Prime Minister’s People’s Wealth Scheme) launched on 28 August 2014. The Bharatiya Mahila Bank was a public sector bank mandated to cater to the banking needs of women. Under the Pradhan Mantri Jan Dhan Yojana programme, the number of small deposit accounts, debit cards, and banking agents engaged by banks has grown significantly.

Table 1. Number of banked and under-banked districts in India, 2010/11–2015/16

Districts	2010/11	2015/16
Banked	207	218
Under-banked	355	344
National average of population per bank branch	13,027	8,683

Note: 562 districts in each year; if district population per branch > national average, the district is under-banked.

Source: Authors’ construction based on RBI (2011, 2016).

In 2013, women were included in the priority sector, which prior to this year comprised small and marginal farmers, agricultural labourers, and Scheduled Castes (SCs) and Scheduled Tribes (STs). Under priority sector lending requirements, which form 40 per cent of adjusted net bank credit (ANBC), banks must lend 10 per cent of the ANBC to groups that were economically and socially disadvantaged and also included women. Though they began late in our time period, the programmes rapidly expanded bank account ownership and are representative of numerous policies undertaken by the RBI to expand financial inclusion over the twenty-first century. With these policies, the share of adults with a bank account more than doubled from 2011 to 2017, to 80 per cent; among women, account ownership increased more than 30 per cent between 2014 and 2017.

Banking the unbanked and under-banked is a policy pushed by governments and international organizations as a multi-goaled win. Has it really been delivered? The number of financial accounts opened and by whom is often regarded as a measure of success in bringing the unbanked into the formal financial system. Does it translate into gains for the poor, more productive and efficient firms, and to greater gender equity in these outcomes and turn some of the beneficiaries into owners of entrepreneurial firms in the informal sector? In brief, how does informal economic activity respond to the extension of financial inclusion? Does it help women enter sectors from which they previously were absent?

3. Descriptive statistics, empirical strategy, and identification

In this section, we discuss descriptive statistics and outline our empirical strategy and approach to identification (see Section B1 in Supplementary Material on the sources of our data). Our intention is to establish a structure that allows an analysis of the allocation of unorganized sector firms across two parts of the unorganized sector, one part entrepreneurial, the other not. To some extent, this is a distinction between a sector that is dynamic and one is residual, allowing us to characterize an informal sector with both entrepreneurial and subsistence aspects, reflecting conflicting characterizations that we find in the literature (see, for example, Fields, 1990; Maloney, 1999).

3.1. Variables and descriptive statistics

Our primary objective is to analyse the role of financing in explaining whether firms are entrepreneurial in the Indian informal sector and whether female-owned firms are more likely to be entrepreneurial with greater access to finance. Critical, therefore, are what we mean by an entrepreneurial firm, how we capture the gendered role of financing and our measures of the availability of financing to the firm. In Table 2 we outline the key variables of interest to us and in Table A1 we present the basic descriptive statistics.

Table 2. Key explanatory variables and their construction

Variables	Description
Dependent variable
Entrepreneurial firm (E)	Binary variable for firms that employ at least one hired worker, besides family workers
Independent variables
Access to finance
FIN1	Binary variable for firms that faced any borrowing constraint in the last year
FIN2 DUM1	Dummy variable for firms with bank loan
FIN2 DUM2	Dummy variable for firms with non-government loan
FIN2 DUM3	Dummy variable for firms with government loan
FIN2	Dummy variable for firms with any type of loan
Gender
Female	Binary variable for female-run firms

Source: Authors’ construction based on Census of India (2001, 2011) and NSSO (National Sample Survey Office) (2013, NSSO, 2017).

3.1.1. Dependent variable

We classify an entrepreneurial firm as one employing at least one hired worker on a regular basis (besides family workers) and the variable ‘entrepreneurial firm’ takes the value 1 for such firms, as in, for example, Earle and Sakova (2000). Firms that are not ‘entrepreneurial’ are own-account enterprises that exclusively make use of family labour; these are mainly found on household premises.

The decision to employ a hired worker transforms an informal firm from an own-account enterprise to an employer and is seen as an indicator of entrepreneurial success (Gindling & Newhouse, 2014). The hiring of non-family workers for a household enterprise involves an implicit barrier to entry, as these employers typically need to finance the wages of hired workers by borrowing from credit markets or through the profits of the enterprise (Banerji, Raj, & Sen, 2016). Therefore, firms that have managed to make the transition from an own-account enterprise to becoming an employer have managed to overcome this barrier and can be classified as entrepreneurial firms.⁷

3.1.2. Main independent variables

Finance constraint: We construct two measures to represent the availability of finance to the firm (or the firm’s financial constraint).

1. We construct a direct measure (FIN1). The surveys ask firms if they encountered any borrowing constraints during the last year. This measure takes the form of a dummy variable with the value of 1 for firms whose owners reported non-availability or high credit costs as a major problem that they faced over the last year.⁸

2. We also construct 0–1 categorical variables for firms receiving bank loans (FIN2DUM1), non-government loans (FIN2DUM2), and government loans (FIN2DUM3), while keeping firms not taking loans as the benchmark (reference) category.

Each of these two measures of financial constraints is imperfect in itself.⁹ While the first financial constraint variable (FIN1) is an increasing measure of financial constraint, FIN2 is an inverse measure of financial constraint. Using the two measures mitigates the problem of measurement errors in any one measure. Specifically, a score of 1 may reflect either that the firm did not attempt to obtain credit or that it faced real difficulty in obtaining credit.

Female: This is a binary variable, which is equal to 1 if the firm is owned and managed by a woman and 0 otherwise. We consider only sole-proprietorship firms, that is, firms with sole owners. We do not consider partnership firms, that is, firms owned by more than one person who shares the firm’s proceeds. Conceptually, the reason for looking only at proprietorships and not partnerships is that partnerships involve joint decision-making, usually with the partners dividing responsibilities. As such, we do not know who the ‘face’ of the firm is and who may be running the firm; hence, if the partners are of different genders we do not have a clear indication of the role of gender in our question of interest. As we have such a large sample, limiting our sample to sole proprietorships does not cost in terms of losing observations and provides unambiguity.¹⁰

Year-industry fixed effects: Year and industry fixed effects are included to help capture otherwise unobserved year- and industry-specific external finance constraints. Time effects include macro shocks with possible firm productivity effects. Because some industries are more reliant on external finance than others (Rajan & Zingales, 1998), we include industry-fixed effects as controls to capture industry-specific external finance requirements that may wield an independent influence on hired labour usage over and above that exerted by the finance constraint that the firm faces.

The summary statistics for the variables used in the analysis are presented in Appendix, Table A1 with further description of firm characteristics as control variables. Appendix, Table A2 presents a simple and fascinating picture of entrepreneurship in our sample. Panel A shows the cross-tabulation of entrepreneurship with the owner’s gender; Panel B displays entrepreneurship versus whether the firm faces a financial constraint. Female-run firms are much less likely to be entrepreneurial than male-run firms. Firms not facing a financial constraint are more likely to be entrepreneurial than those with such constraints. We also see striking differences in the characteristics of entrepreneurial and non-entrepreneurial firms (Panel C in Table A2). Entrepreneurial firms are more productive than non-entrepreneurial firms, are more capital intensive, and are more likely to maintain accounts, have used a computer or internet, and be registered under any Act or Authority (the t-statistics on differences in characteristics between entrepreneurial and non-entrepreneurial firms are statistically significant in all cases). This suggests that entrepreneurial firms are significantly better performing than non-entrepreneurial firms and underscores the importance for household units to expand into non-household units for their further growth.

3.2. Estimation strategy

We estimate variations of (1) $$E_{\mathit{eit}}={\alpha }_0+{\alpha }_1{\mathit{FIN}}_{\mathit{eit}}+{\alpha }_2{\mathit{Female}}_{\mathit{eit}}+\sum _{k>0}{\gamma }_k{X}_{\mathit{ket}}+{\mu }_t+{\delta }_i+{\epsilon }_{\mathit{eit}}\mathrm{\prime }$$(1) where the dependent variable $E_{\mathit{eit}}$ is a dummy variable for an entrepreneurial firm e in the industry i at time t. We classify a firm as an entrepreneurial firm if it employs at least one hired worker (besides family workers) and the variable takes the value 1 for such firms.¹¹ $\mathit{FIN}$ measures a firm’s financial constraint, and we use two alternative measures of financial constraint.¹² $X_{\mathit{ket}}$ is a vector of firm-specific control variables. In particular, we control for differences across firms in terms of age, nature of registration, location, assistance towards training and marketing, social group of the owner, and log of labour productivity.¹³ The variables ${\mu }_t$ and ${\delta }_i$ control, respectively, for time- and industry-specific fixed effects.

3.3. Identification and estimation

Our empirical strategy aims to identify the effect of the finance constraint on entrepreneurship, and whether it differs for men and women. The validity of our analysis rests on the exogeneity of the finance variables. We employ an instrumental variable (IV) approach to allay endogeneity issues and identify causal effects. This requires one or more variables — instruments — that are strongly correlated with the endogenous regressor (financial constraint) and influence the outcome variable (entrepreneurship) through only the endogenous regressor.

We face several possible sources of omitted variable bias. The enterprise owner’s decision not to hire outside workers may be due to family environment, motivation, ability, and other unobserved characteristics. For example, the innate ability of the entrepreneur differs across firms. McKenzie and Woodruff (2014) have found this to be a significant positive factor in a firm’s success and its ability to employ hired labour (i.e. become entrepreneurial). Another source of selectivity lies in the construction of the null category in our FIN variables, as discussed above. While our control variables help ease some of the endogeneity, they are incomplete. Hence, we need to account for omitted variable bias.

The argument for our IV is strong and straightforward. We rely on RBI (central bank) policy discussed above. The policy the RBI followed was to increase the number of bank branches (or open accounts automatically, or increase liquidity, etc.) in under-banked districts, where ‘under-banked’ was defined as applying to districts with a population per branch greater than India’s nationwide mean. Since 2001, the RBI has maintained a list of under-banked districts where banks are required to open half of their new branches and are provided with incentives to do so. The policy affects entrepreneurship in the unorganized sector only via its effect on the enterprises’ financial constraints. Using this information, we construct bank branches per 100000 people (BPP) at the district level to instrument our finance variable. Our instrument varies over time, and as per our estimates, it increased from 8.9 branches per 0.1 million people to 13.09 branches per 0.1 million people. Our argument is that geographical access to banking provides better firm access to finance within the same geographical area, the district. We expect firms that are in districts with easier bank access (fewer people per bank branch, for example) to have better access to financing and, therefore, expansion (instrument relevance).¹⁴ Moreover, we believe our instrument meets the necessary exclusion criterion for an IV, as it is only through the firm’s financial constraint that it should influence the enterprise’s decision to hire outside workers. We test for the suitability of the instrument in our estimations.

We attempt to correct endogeneity by employing the two-stage residual inclusion (2SRI) method (Terza, Basu, & Rathouz, 2008). The conventional two-stage least squares (2SLS) method is the standard approach followed in these circumstances to address endogeneity when employing IV. However, as we have a categorical variable as the dependent variable, employing 2SLS will be susceptible to bias; 2SRI performs better than 2SLS and delivers consistent estimates (Wooldridge, 2010). Following Ivlevs, Nikolova, and Popova (2020) in their analysis of entrepreneurship in former Soviet economies, we first estimate a standard first-stage auxiliary regression in which our instruments and all the control variables are used to explain our potentially endogenous regressor (i.e. FIN1 and FIN2). In the second-stage equation, we include the predicted first-stage residuals, in conjunction with the endogenous regressor. More details about the 2SRI method are presented in Supplementary Material, Section B2. In this section, we have established the core of our approach. In the next sections, we discuss the results of bringing data to the equations and considerations, as well as variations in the modelling and a robustness check for our story.

4. Results

This section presents our results, including those from logit and IV estimation methods and other implementations, as discussed in the previous section. The data comprise the two repeated cross-sections for 2010/11 and 2015/16, discussed earlier.

4.1. Baseline results

Table 3 presents the marginal effects from logit estimations of our measures of finance on the probability that a firm hires non-family labour.¹⁵ Columns 1–4 are for the 2010/11 wave; Columns 5–8 for the 2015/16 wave; and Columns 9–12 pool both waves, with year controls in the even columns. In all estimations, we include firm-level variables as controls. These controls include the critical categorical variable gender of the owner, as well as location (urban or not), social group of the owner, age of the firm, whether the firm has received government assistance towards training and marketing, whether the firm is registered with some government body, and firm productivity. Even-numbered columns in the entire table include fixed effects (FE) for industries.

Table 3. Access to finance and entrepreneurship: Marginal Effects from logit estimations (dependent variable: entrepreneurial firm or not)

Variables	2010–2011				2015–2016				Pooled Data
	FIN1		FIN2		FIN1		FIN2		FIN1		FIN2
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)	(12)
FIN1	−0.035***	−0.030***			−0.017***	−0.012***			−0.022***	−0.022***
	(0.004)	(0.004)			(0.004)	(0.004)			(0.003)	(0.003)
FIN2 DUM1			0.113***	0.142***			0.128***	0.170***			0.121***	0.158***
			(0.004)	(0.004)			(0.005)	(0.005)			(0.003)	(0.003)
FIN2 DUM2			0.066***	0.064***			0.060***	0.070***			0.068***	0.068***
			(0.004)	(0.004)			(0.004)	(0.004)			(0.003)	(0.003)
FIN2 DUM3			0.107***	0.118***			0.155***	0.185***			0.125***	0.146***
			(0.013)	(0.013)			(0.018)	(0.019)			(0.011)	(0.011)
Female	−0.173***	−0.196***	−0.173***	−0.197***	−0.197***	−0.198***	−0.197***	−0.201***	−0.179***	−0.200***	−0.180***	−0.202***
	(0.003)	(0.004)	(0.003)	(0.004)	(0.004)	(0.004)	(0.004)	(0.004)	(0.002)	(0.003)	(0.002)	(0.003)
Controls	Y	Y	Y	Y	Y	Y	Y	Y	Y	Y	Y	Y
Time FE	N	N	N	N	N	N	N	N	N	Y	N	Y
Industry FE	No	Y	N	Y	N	Y	N	Y	N	Y	N	Y
N	313586	313583	313586	313583	270469	270465	270469	270465	584055	584051	584055	584051

Note: Robust standard errors are reported in parentheses; ***p < 0.01, **p < 0.05, *p < 0.1.

Source: Authors’ estimates.

FIN1 in the tables is the direct measure of financial constraints (Columns 1, 2, 5, 6, 9, 10) as discussed earlier, which is similar to the way papers using the WBES capture financial constraints. Notice that the marginal effects for FIN1 carry the expected sign and statistical significance at the 1 percent level. This implies that the financial constraint hinders firm’s transitioning from non-entrepreneurial to entrepreneurial. To state it differently, as the constraint weakens, we are more likely to find firms in the part of the informal sector we have labelled ‘entrepreneurial’. The marginal effect of FIN1 from the full model for pooled data (column 10) suggests a reduction in the likelihood of firm transition, on average, by 2.2 percent in the presence of financial constraints.

FIN2 also captures the sources of external funding, but here we use dummy variables to capture the types of loans the firm received, with FIN2DUM1, FIN2DUM2, and FIN2DUM3 as categorical variables for firms receiving bank loans, government loans, and other non-government loans, respectively. Firms not receiving any external financing are the reference group. Hence, these estimates are made with our complete sample. Again, in Table 3, relative to the base category of firms with no external financing, the marginal effects are all positive and significant at the 1 per cent level endorsing our earlier finding: more external findings, more entrepreneurial behaviour.

In all model specifications—all 12 columns—an informal firm headed by a woman is more likely to be in the family part of the informal economy: the marginal effects are negative and significant at the 1 per cent level in all specifications. We observe that the probability of a firm being non-entrepreneurial increases to 15–20 percent when a woman heads the firm.¹⁶ Our basic logit estimations in Table 3 focus on the role of financial constraints and gender in the distribution of unorganized sector firms across the entrepreneurial–non-entrepreneurial spectrum. The evidence that this offers to us is that alleviating the financial constraint promotes entrepreneurship—firms are willing and able to expand hiring beyond family workers.

We also augment Equation 1(1) $$E_{\mathit{eit}}={\alpha }_0+{\alpha }_1{\mathit{FIN}}_{\mathit{eit}}+{\alpha }_2{\mathit{Female}}_{\mathit{eit}}+\sum _{k>0}{\gamma }_k{X}_{\mathit{ket}}+{\mu }_t+{\delta }_i+{\epsilon }_{\mathit{eit}}\mathrm{\prime }$$(1) by adding the interaction term ${\alpha }_3{{\mathit{FIN}}_{\mathit{eit}}\times \mathrm{ }\mathit{Female}}_{\mathit{eit}}$ to the right-hand side, estimating several variants as in Table 3. This should allow us to see whether the negative effect of being female is weakened through better access to credit or, alternatively, whether better access to credit has a higher return for potential entrepreneurs who are female. The results, both the coefficient values and the marginal effects, are presented in Supplementary Material, Tables B1 and B2. The marginal effects on the variables for financing (FIN1and FIN2) and female firm heads continue to tell the same story as in Table 3 in terms of sign and statistical significance. If we use FIN1 as our preferred measure of financial constraints, we find that the interaction term of the owner’s gender with FIN1 is not statistically significant.¹⁷ On the other hand, the interaction terms of owner’s gender with the two FIN2 dummies are all significant at the 1 per cent level and positive (FIN2DUM1 and FIN2DUM3). As the results on the effect of the finance constraint on male versus female entrepreneurship depend on our choice of the measure to capture the finance constraint, we do not have conclusive evidence that women are at a disadvantage in joining the entrepreneurial side.¹⁸

4.2. Instrumental variable results

We discussed endogeneity concerns between entrepreneurship and financial constraints in Section Identification and Estimation; the constraints may enter as entrepreneurial firms seek investments to expand, innovate, and stay in business. Because of this, we approach the endogeneity issue using 2SRI, as discussed in Section Identification and Estimation. In Table 4, we present the marginal effects for the second-stage reduced-form estimates for one specification with two versions of our financial constraints using our instrumental variables approach.¹⁹ The results mimic, in sign and significance, the logit results discussed in the previous subsection.

Table 4. Access to finance and entrepreneurship: Marginal Effects (2SRI approach: no. of replications: 500)

Variables	FIN1	FIN2
FIN1	−0.070***
	(0.008)
FIN2		0.055***
		(0.006)
Female	−0.152***	−0.150***
	(0.002)	(0.002)
Location	0.027***	0.030***
	(0.001)	(0.001)
ST	−0.099***	−0.094***
	(0.003)	(0.003)
SC	−0.128***	−0.129***
	(0.002)	(0.002)
OBC	−0.051***	−0.052***
	(0.001)	(0.001)
Age 3–9	−0.058***	−0.055***
	(0.002)	(0.002)
Age >9	−0.059***	−0.056***
	(0.002)	(0.002)
Asst	0.126***	0.094***
	(0.005)	(0.006)
Regis	0.234***	0.232***
	(0.001)	(0.001)
lnLP	0.081***	0.080***
	(0.001)	(0.001)
XUhat	0.014***	0.010***
	(0.002)	(0.002)
Time FE	Y	Y
Industry FE	Y	Y
N	577,387	577,544

Notes: Figures in parentheses are bootstrapped standard errors; in this table we collapse FIN2DUM1, FIN2DUM2 and FIN2DUM3 into a single dummy variable, FIN2: FIN2 equals 1 if the firm has taken out any sort of formal loan, 0 otherwise; ***p < 0.01, **p < 0.05, *p < 0.1.

Source: Authors’ estimates.

However, again we see that this varies over the two alternative ways in which we capture the constraints that firms face regarding financial liquidity. Using FIN1, and referring to Equation 3 in the supplementary material, ${\mathrm{\theta }}_{\mathrm{e}}^{\mathrm{est}}$ is the predicted residual from the first-stage equation, and its coefficient, $\gamma ,$ offers a direct test for the exogeneity of the regressor of interest. $\gamma$ is estimated as 0.080 with a standard error of 0.011 suggesting that the regressor is endogenous (coefficient estimates are reported in the Supplementary Material, Table B3); in the paper, we present marginal estimates. The marginal estimate on FIN1 is −0.070 with a standard error of 0.008, indicating significance at the 1 per cent level. In comparison, in Table 3, Column 9, the marginal estimate on FIN1 is −0.022 with a standard error of 0.003.

For the two-stage procedure, we collapse FIN2DUM1, FIN2DUM2, and FIN2DUM3 into a single dummy variable, FIN2. FIN2 equals 1 if the firm has taken out any sort of formal loan. It is 0 otherwise. Using FIN2, $\gamma$ is estimated as 0.057 with a standard error of 0.010, suggesting that the regressor is endogenous. The marginal estimate on FIN2 is 0.055 with a standard error of 0.006, indicating significance at the 1 per cent level. Because in Table 3 we used three dummy variables to capture the loan sources of the firm and here we collapsed them to one, we cannot directly compare the coefficients. However, at a glance, they look similar.

The two-stage procedure we employed shows that it is necessary to account for the endogeneity of enterprise financial constraints. However, doing so does not change the sign or significance of our estimates. In other words, the 2SRI estimates reinforce the main findings arrived at using the logit estimations. The coefficients of FIN1 and FIN2 retain the same sign and significance, suggesting that lack of access to finance is a serious impediment to the transition of firms from household firms to non-household firms employing hired labour.

We also perform the 2SRI estimation using separate samples of male and female owners. The results with the marginal effects are reported in the Appendix, Table A4.²⁰ They show the same pattern observed for the pooled sample. As for our logit results, we obtain somewhat ambiguous results on whether finance constraints matter more for women than for men. For both FIN1 and FIN2, the finance constraint is binding for male and female owners alike, implying that as the finance constraint weakens, both male-owned and female-owned firms are willing and able to hire workers from outside the family. However, for FIN1 the marginal effect on the finance constraint is smaller (more negative) for women than for men; while for FIN2 the marginal effect on the finance constraint is positive and significant for men but insignificant for women, indicating that the constraint is more binding for male owners than for female owners.²¹

5. Robustness test: PSM-DID results

In this section, we outline a critical robustness test. One possible concern with our IV approach is that the instrument—BPP—may not meet the excludability condition requiring that the decision of banks to open branches in a specific district affects the likelihood of entrepreneurship only through this channel. Under the 2005 reforms, banks could choose an extensive and intensive level of entry in under-banked districts as well as the total expansion of their branch network (Young, 2021). This implies the decision to open more branches (per capita) in some districts than in others could have been influenced by (unobserved) district characteristics, which may also influence business performance and the likelihood of entrepreneurship.

In order to address the possibility that the roll-out of banking services in under-banked districts was not random, we supplement our main identification strategy with an alternate approach, combining PSM and DID. We take advantage of the fact that while banks could choose which under-banked district to open new branches in, they were constrained by the RBI’s policy to open branches in under-banked districts in order to receive licences for entry into the rich markets (Young, 2021). Over the period of our analysis, the RBI vigorously pursued its expansion policy by increasing the number of branches, especially in areas that it deemed under-banked. Further details of our PSM-DID strategy are provided in the Supplementary material, Section B3. Our PSM-DID estimates confirm the finding of our IV strategy. Our results, presented in Supplementary Material (Tables B6–B8), clearly point to the positive effect of policy change on entrepreneurship.

6. Conclusions

We began our paper by laying out two questions: (1) Does informal economic activity become more entrepreneurial in response to increased financial inclusion? (2) Does access to new financing options change the gender configuration of informal economic activity and, if so, in what ways and what directions? We picture the informal sector as composed of two firm types: family firms and entrepreneurial firms. The distinction is that entrepreneurial firms employ outside non-family labourers—that is, hired workers. We examine the impact of financial inclusion, partly captured by banking access, on proprietorship in entrepreneurial informal firms and its consequences, focusing on the gender differences in these impacts. The context of our study is India during the 2010s, a period during which banking policies greatly expanded banking access to women and to the unbanked.

We capture financial constraints using two core explanatory variables: (1) information self-reported by firms facing finance constraints (similar to the variable employed by studies making use of the WBES); (2) a set of categorical variables on whether the firm obtained bank loans, government loans, non-government loans, or no loans (the omitted variable). While one could dispute the degree to which each of these core explanatory variables faces the problem, there is an element of them that is clearly potentially endogenous. This means that our logit estimation may not produce the true causal relationship between an enterprise’s financial constraints and its entrepreneurship.

We address the endogeneity potential by implementing an IV approach, relying on the idea that an increase in the number of bank branches is correlated with the difficulty of obtaining financing: briefly, more bank branches in the district where you are living means more financial access. Under-banked districts have less access; during the 2010s, policy strove to increase the number of branches. Our principal banking access measure, i.e. our IV, is whether the district had a population per branch below (banked) or above (under-banked) the national average. This is the indicator the RBI used in implementing its policy to increase branches in under-banked districts. When the IV affects the likelihood of entrepreneurship only through our financial constraint measures, then we say that we can make causal inferences. We also supplement our IV analysis with a robustness check using PSM-DID.

Our results show a strong potential role for increased liquidity. Whichever measure of finance constraint we use, we find clear evidence that finance constraints matter for the likelihood of becoming an entrepreneur in the informal sector in India. The results are robust to concerns about reverse causality. When we use IV estimation using 2SRI or we use PSM-DID, we obtain similar results to when we use a logit estimation method. However, we find less-conclusive evidence that effective financing encourages female entrepreneurship more than male entrepreneurship. Our results here are sensitive to the choice of the financial constraint that we use as our explanatory variable, in both the logit and the 2SRI results. This does not mean that financial constraints do not matter for women entrepreneurs; instead, our results imply that financial inclusion matters for both women and men entrepreneurs.

Our findings have strong implications for policy. In the 2000s, the Indian government initiated an ambitious set of reforms with the objective of ensuring that the areas of the country which historically had not had much access to banking services would be able to able to get access to formal financial institutions. At the same time, India has an endemic problem of a large informal sector, mostly populated by micro household (own-account) enterprises that remain largely unproductive (Raj & Sen, 2016). Our findings suggest that the policy actions of the Indian government increase access to finance in India’s under-banked districts succeeded in one important dimension—it contributed to the growth of entrepreneurship in India’s informal sector, enabling many of the self-employed to become employers, hiring outside workers.

Acknowledgements

The authors gratefully acknowledge the support of UNU-WIDER. We are thankful to the National Sample Survey Office for providing us with access to the data. We also thank Olga Popova and Arnab Basu for their insights, and Olga Popova and Edward Norton for generously sharing code. We also want to thank the participants in the workshop on UNU-WIDER’s project Transforming Informal Work and Livelihoods, the two discussants who reviewed the paper, and the comments received from presentations in the Institute for Social and Economic Change, Bangalore, India, at IOS-Regensburg, Germany, and at the HSE-IZA seminar, Moscow. The usual disclaimers apply. Data availability: The data that support the findings of this study are extracted from the surveys on the unincorporated Non-agricultural enterprises conducted by the National Sample Survey Office (NSSO), Government of India. The data are publicly available in the National Data Archive at http://microdata.gov.in/nada43/index.php/catalog?&page=2. The codes used for estimation are available upon request by bona fide researchers from the authors.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 Among the poor, informal firms are important in job creation (Naudé 2010; World Bank 2013). Moreover, Ivlevs et al. (2020) highlight entrepreneurship’s contributions to bettering welfare through growth, innovation, job creation (van Praag & Versloot 2007), and improving wellbeing and health (Nikolova 2019). Most notably, Landes (1969) placed informal entrepreneurial activity at the heart of early industrialization (though he did not use the word ‘informal’).

2 For example, the development literature has spent decades addressing the many aspects of relieving financing constraints facing firms and families (Mead & Liedholm 1998; Rijkers et al. 2010).

3 Factories registered under the Factories Act of 1948 are generally referred to as the organized or formal sector in India’s data collections. Factories not using power during the preceding year and employing 20 or more workers, or those using power and employing ten or more workers, must register under this Act (Gang, 1992; Gang & Pandey 2007). Other firms fall into the unorganized sector and may be registered under other legislation and government bodies.

4 Non-family workers in MHEs and NHEs are not entitled to social security benefits or are covered by employment protection legislation, as compared to workers in registered (formal) firms in India (see NCEUS 2007).

5 These studies cover a wide geography; for example, no gender effect is found by Aterido et al. (2013) on Sub-Saharan Africa, Bruhn (2009) on Latin America, or Storey (2004) on Trinidad and Tobago. See Presbitero et al. (2014) for a succinct and informative summary of this aspect of the literature.

6 A superior source for understanding RBI policy in this period is Young (2021).

7 The existence of dynamic entrepreneurial firms along with subsistence firms in the informal sector is well documented in the literature (Grimm et al., 2012).

8 Below we discuss a possible selection issue here, in addition to an endogeneity issue for the generic FIN variable.

9 When estimating 2SRI we collapse the second measures to a single ‘received loans or not (FIN2)’. We also examined a ‘source of financing’ constraint, the proportion of outside financing firms receive from ‘formal’ sources. However, this measure faces a severe selection problem, as about 90 per cent of firms did not receive any financing from outside sources.

10 We lose 0.5% of the 619,701 observations in our dataset when we follow this condition.

11 We have also tried an alternate measure of entrepreneurship where the dependent variable takes the value 1 if the number of hired workers employed by a firm is 2 or more and 0 otherwise. Our results remain unaffected even with this alternate measure of entrepreneurship.

12 Their construction is discussed in detail in the section on data and variables in Supplementary Material, Section B1.

13 These variables are defined in Table A1 in the appendix.

14 Regional (here, district-level) data are useful and often used in constructing IVs to address reverse causality and selection for agents within the region (Dustmann & Preston 2001). This gives us instrument exogeneity, i.e. increased banking availability (etc.) is uncorrelated with the error term. It is better if there is a lag; for this, we rely on the earlier collection dates for the RBI data in comparison with the Survey data.

15 The coefficient values of these estimations are presented in the Appendix Table A3.

16 At the suggestion of a Referee, we checked for the possibility of gender impacting entrepreneurship through access to credit by estimating the model specification with and without the finance variable. Our results rule out the possibility.

17 To capture the interaction effect, we rely on the mean of the interaction effects computed using the codes shared by Professor Edward Norton.

18 We also estimated regressions using separate samples of male and female owners for each wave and for the waves pooled with one another. Our results endorse what we have observed for the pooled sample of male and female owners. One possible concern here is that the gender of the owner of the firm may be endogenous with regard to entrepreneurship, if the male members of the more successful firms take over the management or ownership of the firm when it switches from being a non-entrepreneurial firm to an entrepreneurial firm. However, we find that the industries where female owned firms are largely found are very different from the industries where male owed firms dominate – out of the 70 industries in our sample, 40 per cent of female owned firms are in only four of these industries. This suggests that there are specific attributes of these industries which make it more likely for female owned firms to be present, and the gender of the firm owner is unlikely to be endogenously determined.

19 Coefficient values and first stage results are presented in Section Variables and Descriptive Statistics.

20 Coefficient values and first stage results are presented in Supplementary Material, Table B4.

21 One concern with our IV estimates is that we are not able to assess whether unorganized sector firms in India migrate in search of finance, defining migration as crossing district borders. Therefore, the location of the firm may be endogenous to the structure of bank branch expansion. Our data is repeated cross-sections and there is no variable useful in identifying the changed location of the firm. However, it should be noted that if, as expected, the more entrepreneurial firms re-locate to districts with higher availability of finance, it should lead to a downward bias to the coefficients on the finance constraint in our main empirical specifications. Therefore, the estimated coefficients on the two measures of the finance constraint in our estimates can be seen as a lower bound of the true effect of the finance constraint on the likelihood of a firm becoming entrepreneurial.

Appendix

Table A1. Summary statistics, variables and their construction

Variable	Observations	Mean	SD	Min.	Max.
Part A: summary statistics
Entrepreneurial firm (E)	584,055	0.38631	0.48690	0	1
FIN1	584,055	0.06914	0.25369	0	1
FIN2 DUM1	584,055	0.05354	0.22511	0	1
FIN2 DUM2	584,055	0.06156	0.24035	0	1
FIN2 DUM3	584,055	0.00468	0.06828	0	1
Female	584,055	0.12908	0.33529	0	1
Location	584,055	0.51206	0.49986	0	1
ST	584,055	0.05574	0.22943	0	1
SC	584,055	0.10004	0.30005	0	1
OBC	584,055	0.46240	0.49858	0	1
Age3–9	584,055	0.44885	0.49738	0	1
Age >9	584,055	0.42633	0.49454	0	1
Asst	584,055	0.01196	0.10870	0	1
Regis	584,055	0.40060	0.49002	0	1
lnLP	584,055	10.01993	0.98134	2.10843	15.9002
BPP	577,558	10.84031	6.72247	1.39515	48.6695
Variables	Description	Source
Part B: variables and their construction
Dependent variable
Entrepreneurial firm (E)	Binary variable, firms employing at least one hired worker, besides family workers	NSSO data
Independent variables
Access to finance
FIN1	Binary variable for firms that faced any borrowing constraint in the last year	NSSO data
FIN2 DUM1	Dummy variable for firms with bank loan	NSSO data
FIN2 DUM2	Dummy variable for firms with non-government loan	NSSO data
FIN2 DUM3	Dummy variable for firms with government loan	NSSO data
FIN2	Dummy variable for firms with any type of loan	NSSO data
Gender and interaction terms
Female	Binary variable for female-run firms	NSSO data
FIN1*Female	Interaction between FIN1 and Female	NSSO data
FIN2DUM1 *Female	Interaction between FIN2DUM1 and Female	NSSO data
FIN2DUM2 *Female	Interaction between FIN2DUM2 and Female	NSSO data
FIN2DUM3 *Female	Interaction between FIN2DUM3 and Female	NSSO data
Control variables: firm characteristics
Location	Dummy variable for urban firms	NSSO data
ST	Dummy variable for Scheduled Tribe (ST)-owned firms	NSSO data
SC	Dummy variable for Scheduled Caste (SC)-owned firms	NSSO data
OBC	Dummy variable for Other Backward Communities (OBC)-owned firms	NSSO data
Age3–9	Dummy variable for firms aged between 3 and 9 years	NSSO data
Age >9	Dummy variable for firms that completed more than 9 years	NSSO data
Asst	Dummy variable for firms that received any government assistance during last three years	NSSO data
Regis	Dummy variable for firms that registered under any one of the Shops and Establishment Act, Municipal Corporation/Panchayats/Local Body, Vat/Sales Tax Act, Provident Fund Act, or Employees State Insurance Corporation Act, or with the SEBI/Stock Exchange or any other industry-specific Act/authority	NSSO data
lnLP	Log of labour productivity, where labour productivity is defined as the ratio of gross value added to employment	NSSO data
District-level variables
SHSCPOP	Proportion of SC population in total population	Census 2001
SHSTPOP	Proportion of ST population in total population	Census 2001
MIDGRADEDU	Proportion of individuals educated at secondary level and above	Census 2001
ROADVILLG	Share of villages with paved approach road in total villages	Census 2001
ELECVILLG	Share of electrified villages in total villages	Census 2001
POSTVILLG	Share of villages with post office in total villages	Census 2001
BUSVILLG	Proportion of villages situated on a bus route in total villages	Census 2001
PRIMSCHVILLG	Proportion of villages with at least a primary school in total villages	Census 2001
Instrument Used
BPP	Branches per 100000 people	RBI

Source: Authors’ construction based on Census of India (2001, 2011), NSSO (2011, 2016), RBI (2011, 2016).

Table A2. Cross-tabulations: entrepreneurship characteristics

Owner	Owner	Entrepreneurship		Total
		Non-entrepreneurial firm	Entrepreneurial firm
Panel A: entrepreneurship vs gender of owner
Male-run firms	Frequency	295,536	213,130	508,666
	Row percentage	58.1	41.9	100.0
	Column percentage	82.5	94.5	87.1
	Cell percentage	50.6	36.5	87.1
Female-run firms	Frequency	62,890	12,499	75,389
	Row percentage	83.4	16.6	100.0
	Column percentage	17.6	5.5	12.9
	Cell percentage	10.8	2.1	12.9
Total	Frequency	358,426	225,629	584,055
	Row percentage	61.4	38.6	100.0
	Column percentage	100.0	100.0	100.0
	Cell percentage	61.4	38.6	100.0
Panel B: entrepreneurship vs financial constraint
Financial constraint	Owner	Entrepreneurship		Total
		Non-entrepreneurial firm	Entrepreneurial firm
No	Frequency	331,458	212,215	543,673
	Row percentage	61.0	39.0	100.0
	Column percentage	92.5	94.1	93.1
	Cell percentage	56.8	36.3	93.1
Yes	Frequency	26,968	13,414	40,382
	Row percentage	66.8	33.2	100.0
	Column percentage	7.5	6.0	6.9
	Cell percentage	4.6	2.3	6.9
Total	Frequency	358,426	225,629	584,055
	Row percentage	61.4	38.6	100.0
	Column percentage	100.0	100.0	100.0
	Cell percentage	61.4	38.6	100.0
Panel C: firm characteristics by entrepreneurship
Characteristics		Entrepreneurship		t-test
		Non-entrepreneurial firm	Entrepreneurial firm
lnLP		9.788	10.388	0.600***
				(0.003)
Firms that maintained accounts		0.053	0.286	0.233***
				(0.001)
Firms that used computer		0.021	0.145	0.124***
				(0.001)
Firms that used internet		0.014	0.103	0.090***
				(0.001)
Regis		0.259	0.626	0.367***
				(0.001)
Mean land and building to employment (in logs)		10.607	11.233	0.626***
				(0.004)
Mean plant and machinery to employment (in logs)		7.973	8.486	0.513***
				(0.007)
Number of firms		358,426	225,629

Source: Authors’ estimates.

Table A3. Access to finance and entrepreneurship: logistic regression coefficient estimates with control variables (dependent variable: entrepreneurial firm or not)

Variables	2010–2011				2015–2016				Pooled Data
	FIN1		FIN2		FIN1		FIN2		FIN1		FIN2
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)	(12)
FIN1	−0.164***	−0.147***			−0.068***	−0.049***			−0.097***	−0.095***
	(0.018)	(0.019)			(0.017)	(0.017)			(0.012)	(0.013)
FIN2 DUM1			0.533***	0.689***			0.522***	0.699***			0.527***	0.694***
			(0.020)	(0.022)			(0.019)	(0.021)			(0.014)	(0.015)
FIN2 DUM2			0.312***	0.310***			0.245***	0.286***			0.297***	0.298***
			(0.019)	(0.020)			(0.016)	(0.017)			(0.012)	(0.013)
FIN2 DUM3			0.506***	0.571***			0.633***	0.756***			0.542***	0.645***
			(0.061)	(0.066)			(0.073)	(0.077)			(0.046)	(0.049)
Female	−0.817***	−0.951***	−0.816***	−0.958***	−0.807***	−0.813***	−0.807***	−0.825***	−0.781***	−0.882***	−0.781***	−0.891***
	(0.015)	(0.017)	(0.015)	(0.017)	(0.014)	(0.016)	(0.014)	(0.016)	(0.010)	(0.012)	(0.010)	(0.012)
Controls	Y	Y	Y	Y	Y	Y	Y	Y	Y	Y	Y	Y
Time FE	N	N	N	N	N	N	N	N	N	Y	N	Y
Industry FE	No	Y	N	Y	N	Y	N	Y	N	Y	N	Y
N	313586	313583	313586	313583	270469	270465	270469	270465	584055	584051	584055	584051
Pseudo R2	0.145	0.221	0.148	0.224	0.148	0.226	0.151	0.230	0.151	0.228	0.154	0.232

Note: Robust standard errors are reported in parentheses; ***p < 0.01, **p < 0.05, *p < 0.1.

Source: Authors’ estimates.

Table A4. Access to finance and entrepreneurship by gender: Marginal Effects (2SRI approach: no. of replications: 500)

Variables	Male-run firms		Female-run firms
	FIN1	FIN2	FIN1	FIN2
FIN1	−0.074***		−0.119***
	(0.008)		(0.020)
FIN2		0.077***		−0.002
		(0.006)		(0.009)
XUhat	0.014***	0.004**	0.026**	0.013***
	(0.002)	(0.002)	(0.004)	(0.002)
Controls	Y	Y	Y	Y
Time FE	Y	Y	Y	Y
Industry FE	Y	Y	Y	Y
Observations	503,429	503,498	73,674	73,922

Notes: Figures in parentheses are bootstrapped standard errors; in this table we collapse FIN2DUM1, FIN2DUM2 and FIN2DUM3 into a single dummy variable, FIN2: FIN2 equals 1 if the firm has taken out any sort of formal loan, 0 otherwise; ***p < 0.01, **p < 0.05, *p < 0.1.

Source: Authors’ estimates.