Aggregate Costs of a Gender Gap in the Access to Business Resources

1 Introduction

The empowerment of women in many domains of life is one of the most important transformations in the economic and social worlds in recent decades.^[1] However, a significant gender gap in entrepreneurship still persists. In OECD countries, the proportion of sole-proprietor enterprises owned by women is only between 20% and 40% (OECD 2012). Furthermore, less than 25% of businesses across the world are owned by women according to the Enterprise Surveys elaborated by the World Bank. Discrimination in the access to business resources might be blocking female participation in entrepreneurial activities, as it is implicitly suggested by the literature that studies entrepreneurship in the presence of financial frictions (Buera, Moll, and Shin 2015).

Indeed, descriptive evidence shows that women seem to have access to fewer resources than men to run their businesses (Aidis et al. 2007; OECD 2012, 2016; among others). Some studies also conclude that women entrepreneurs face worse credit conditions than men even when controlling for the firm and entrepreneur characteristics (Agier and Szafarz 2013; Aristei and Gallo 2016; Morazzoni and Sy 2021; Muravyev, Schaefer, and Talavera 2009; among others). Although other supply and demand factors could be blocking the participation of women in business activities, discrimination in access to business resources is definitely a barrier to female entrepreneurship. Furthermore, this type of distortion might be preventing an efficient allocation of talent, with aggregate consequences for the economy as suggested by Hsieh et al. (2019). However, quantifying the aggregate effects of gender barriers in entrepreneurship has received little attention in the macroeconomic literature. Our paper provides a tractable starting point by analyzing the aggregate costs of a gender-specific distortion in the access to business resources.

We model a distortion that limits the amount of capital that female entrepreneurs can employ to run their businesses. The latter distortion is implicitly motivated by the evidence suggesting the existence of a gender gap in loan size and a glass ceiling effect such that the largest female projects face the highest penalty (e.g. Agier and Szafarz 2013). We insert this type of distortion into a model with heterogeneous agents and an endogenous size distribution of production units. In the model, there is a single representative household composed by female and male members who are endowed with heterogeneous managerial skills. Agents must decide whether to run a business or be employed as a production worker in someone else’s firm. Production is carried out with a span-of-control technology that uses managerial skills, capital, female labor, and male labor. Hence, more skilled managers run bigger businesses.^[2] However, women can employ up to a given amount of capital to start their businesses. As production workers, agents are identical, although women and men are imperfect substitutes in production. Therefore, women and men are otherwise alike, except by their imperfect substitutability in the labor market as production workers, and the capital limit that women face to run their businesses.

We calibrate the model to match data of a benchmark economy that exhibits no labor market gender gaps, except regarding entrepreneurship. We choose Denmark, Norway, and Sweden to build the latter benchmark. Hence, we refer hereafter to the average data of Denmark, Norway, and Sweden as the data of the benchmark economy. As observed in Table 1, in the benchmark economy, the gender gap in education is practically null, the ratio of female-to-male labor force participation is close to one, the gender wage gap is tiny; however, a significant gender gap in entrepreneurship is still observed. Specifically, Table 1 shows that only about a third of individuals running a business are women in the benchmark economy. We calibrate the technology parameters of the model to match data on the size distribution of establishments and employment shares by large establishments in the benchmark economy. The gender-specific friction is calibrated to match the observed gender gap in entrepreneurship.

We use the calibrated model to perform counterfactual exercises that quantify the effect of the distortion on aggregate outcomes. Specifically, we build five counterfactual scenarios which gradually and progressively loosen the capital limit for female entrepreneurs. We compare different outcomes between the counterfactual scenarios and the benchmark economy. Then, our quantitative analysis responds to the following question: How costly is a capital limit on women businesses that generates a gender gap in entrepreneurship that is similar to that observed in the benchmark economy?

We find that the gender-specific capital constraint causes an output loss between 13.7% and 28.2%.^[3] Furthermore, the estimated losses in average output per efficiency unit of labor ranges from 18.2% to 33.7%, in average output per production worker ranges from 15.1% to 27.3%, and in average output per manager ranges from 24.5% to 49.3%. We also find that the distortion causes a fall in total factor productivity (TFP) between 12.2% and 19.5%. In addition, we quantify the contribution of capital intensity and TFP to the fall in output per worker. We find that the fall in productivity accounts for 65%–80% of the output per worker loss. We also show that the potential aggregate gains of releasing the distortion studied in this paper would be mostly driven by the more intensive capital utilization of existing female entrepreneurs (intensive margin) and not by the entry of more women into entrepreneurship (extensive margin).

Several articles have studied the relationship between gender inequality and economic performance (Doepke and Tertilt 2009; Esteve-Volart 2009; Fernandez 2009; Galor and Weil 1996; Greenwood, Seshadri, and Yorukoglu 2005; Lagerlof 2003; Ngai and Petrongolo 2017; among others).^[4] However, few papers quantify the macroeconomic effects of gender gaps in the labor market (Cavalcanti and Tavares 2016; Cuberes and Teigner, 2016; Cuberes and Teigner, 2018; Hsieh et al. 2019). Our goal in this paper is to contribute to this literature by analyzing the macroeconomic effects of a specific distortion that directly impacts the gender gap in entrepreneurship; namely, women’s limited access to business resources. As discussed previously, evidence suggests that this distortion is indeed relevant, although their aggregate costs have not been quantified thus far.

The rest of the paper is organized as follows. Section 2 discusses evidence of discrimination against women in the access to business resources. Section 3 develops the theoretical framework. Section 4 describes the calibration strategy. Section 5 presents and discusses the results of this paper. Finally, Section 6 concludes.

2 Literature Review

Evidence shows that gender barriers in entrepreneurship still exist. According to the World Bank Business Surveys, less than 25% of businesses worldwide are owned by women. Meunier et al. (2017) analyze data on female and male entrepreneurship from the World Bank Group Entrepreneurship Database and find that less than a third of the new owners of limited liability companies are women in most of the economies analyzed. Similarly, the OECD (2016) reports that women in the European Union were half as likely as men to be self-employed and women are also less likely than men to participate in business start-ups. Piacentini (2013) shows that, across the majority of OECD countries, the share of women-owned individual enterprises does not exceed 30%. The Global Entrepreneurship Monitor (GEM 2021) also concludes that men are more likely to start new businesses than women. In addition, the 2016 Report of the United Nations Secretary-General’s High-Level Panel on Women’s Economic Empowerment provides strong evidence that women lag behind men in terms of number of business owners, size of women-owned businesses, and their access to economic resources; specifically, women-owned businesses are smaller and disadvantaged in their access to credit, resources, and assets (UN 2016). Morazzoni and Sy (2021), using data from the U.S. Census for 2018, report that women entrepreneurs make up just over 35% of the entrepreneurial pool in the U.S. Consequently, Morazzoni and Sy conclude that gender gaps in labor market participation are more severe for employers than for salaried workers.

Non-causal evidence suggests that women face worse conditions than men to access business resources, which raises a barrier to female entrepreneurship. The OECD (2012) reports that, in the United States, 60% of female entrepreneurs start with resources less than US$5000, compared to 42% of male entrepreneurs. The same report shows that 39% of U.S. women entrepreneurs started their businesses with credit from a bank compared to 47% of male entrepreneurs; European countries exhibit similar results. OECD data also reveal that women entrepreneurs have less access to business financing than men; for example, only 36% of Danish and 41% of Swedish women business owners report having access to the resources necessary to grow their businesses, compared to 46% of Danish and 50% of Swedish male business owners (OECD 2016). The IFC of the World Bank has estimated that, worldwide, there is a financing gap of US$300 billion for formal small businesses owned by women, and more than 70% of small and medium enterprises owned by women have inadequate or no access to financial services. Along the same lines, Morazzoni and Sy (2021) report that, in 2018, women received only 2.2% of total U.S. venture capital financing. In addition, in developing countries, access to finance is commonly noted as an important barrier to business success. For example, Aidis et al. (2007), using data from Lithuania and Ukraine, show that access to funds was a more important barrier to advancement for women business owners than for men.

Therefore, non-causal evidence is relatively robust about a fact: on average, women seem to start their businesses with less resources than men. However, the latter is an equilibrium result that entails both supply and demand effects of a diverse nature. In this regard, some studies, mainly outside the U.S., provide causal evidence suggesting the existence of discrimination in access to business resources, a barrier to female participation in business activities. We now briefly discuss some of those studies.

Morazzoni and Sy (2021), using credit data comprising nearly 5000 U.S. entrepreneurs, document evidence of gender differences in business access to credit. Specifically, after controlling for individual and firm characteristics, the authors find that, compared to men, women entrepreneurs raise 35% less business debt^[5] and are 10% more likely to be rejected when applying for a bank loan. Furthermore, the authors conclude that the higher rejection rates cannot be empirically explained by higher risk from women-led businesses or lower profitability from these businesses. Therefore, Morazzoni and Sy (2021) provide empirical evidence that credit restrictions appear to penalize women entrepreneurs relatively more; not only do women entrepreneurs report lower levels of business debt than men, but among those who apply for a loan, women are also more likely to be rejected. The authors also develop a model of heterogeneous agents, business choice, and gender-specific financial frictions; relying on a counterfactual analysis, they show that eliminating gender differences in financing leads to a 16% increase in female entrepreneurship rates.

Agier and Szafarz (2013) use a database that includes 34,000 loan applications from a Brazilian microfinance institution over an eleven-year period to study whether male and female borrowers benefit from the same credit conditions.^[6] The authors then compare the attribution of loans to male and female applicants with similar creditworthiness characteristics. The authors find a gender gap in loan size; that is, in similar circumstances, the size of the loan is biased in favor of male clients and therefore women face harsher loan conditions than men. The authors also confirm the presence of a glass ceiling effect, that is, the largest female projects face the highest penalty.

Aristei and Gallo (2016) provides empirical evidence on gender differences in access to credit for firms in European transition economies. The authors find that the probability of credit rationing against female firms is not explained by observed characteristics of firms. The authors interpret the latter result as evidence of gender discrimination in credit markets.

Using a large sample of small and medium-sized European companies, Moro, Wisniewski, and Mantovani (2017) find that businesses run by women are less likely to apply for a loan, as they anticipate being rejected.^[7] As a consequence, companies run by women obtain less bank financing. However, the authors cannot rule out that women exclude themselves due to their perception that banks will discriminate against their gender. In this regard, Bellucci, Borisov, and Zazzaro (2010) find that women entrepreneurs in Italy are more likely to be denied credit and therefore may be more discouraged from applying for a loan.^[8]

Muravyev, Schaefer, and Talavera (2009) use the Business Environment and Business Performance Survey to study gender discrimination in small business lending in 26 Central and Eastern European transition economies and several Western European industrialized nations. The authors provide evidence that women-owned businesses are less likely to obtain a bank loan. Cavalluzzo, Cavalluzzo, and Wolken (2002) also find evidence of a credit access gap between firms owned by white males and white females in the U.S., with female denial rates increasing with lender concentration.

Overall, the evidence shows that women do face worse credit conditions than men. Causal evidence supports the conclusion that gender discrimination is a factor that could explain a gender gap in access to credit, although other demand-driven explanations cannot be ruled out. From a quantitative point of view, the aggregate impact of gender barriers in entrepreneurship has received little attention in the macroeconomic literature. Our paper contributes in that direction by providing a tractable model to study the impact on aggregate outcomes of removing a gender-specific capital distortion.

3 Model

We build a one-sector aggregative model in which production is carried out by heterogeneous establishments run by two types of agents: women and men. Each type of agent can also work as a production worker in the establishment run by someone else. This benchmark model is exposed to a supply-driven distortion to female entrepreneurship.

3.9 Capital Constraint for Female Business Owners

In the model economy, women face an exogenous constraint in their access to business resources. The distortion takes the form of a ceiling k ̄ on the capital amount that women can rent to run their businesses. We now discuss how this distortion impacts female entrepreneurship.

Consider first the partial equilibrium. As a direct consequence of the span-of-control production technology, more skilled women run bigger businesses, and thus, rent larger amounts of capital. Hence, the capital limit k ̄ is only binding for women with the highest managerial skills. Constrained women face a reduction in their profits as a consequence of the distortion. In contrast, the distortion is innocuous for unconstrained women. Therefore, two cases arise, which are illustrated in panels (a) and (b) of Figure 1.^[13] First, suppose low equilibrium wages (panel a). Then, the marginal women entrepreneur is relatively low-skilled, and thus, the capital limit is not binding for her. Hence, the friction is innocuous on the equilibrium number of entrepreneurs, although it certainly impacts the scale of operation of the most productive (constrained) entrepreneurs. Then, the skill threshold above which women start a business in the frictional economy, z ̄ f * , is the same as in the frictionless economy, z ̃ f * .^[14] Suppose now high equilibrium wages (panel b). In this case, the friction k ̄ constrains the amount of capital available for the marginal women entrepreneurs, who then experience a fall in their profits (dashed line). Therefore, some female entrepreneurs do not have more incentives to run a business, and thus, they become production workers. That is, z ̃ f * < z ̄ f * , and thus, the number of female entrepreneurs falls. In this partial equilibrium analysis, men’s choices are not impacted by the distortion.

Figure 1:

Partial equilibrium analysis.

We now analyze the general equilibrium. Let us start from the scenario depicted in panel (a) of Figure 1; that is, the case in which the capital limit is not binding for the marginal women entrepreneurs. Thus, the distortion does not exert a direct effect on the number of entrepreneurs. This is the partial equilibrium effect analyzed previously. However, existing female entrepreneurs are constrained in their use of capital and must run smaller businesses as a consequence of the distortion. Those firms demand less production labor which pushes wages down.^[15] Hence, profits go up and the choice of running a business becomes more attractive for both women and men. Formally, the threshold managerial skill falls; that is z ̃ f * > z ̄ f * and z ̃ m * > z ̄ m * . Therefore, both female and male entrepreneurship rises. Figure 2 illustrates these first round effects for women (panel a) and men (panel b), where we denote by w ̄ the frictional wage.

Figure 2:

General equilibrium analysis I.

Let us start now from the situation depicted in panel (b) of Figure 1. In this case, the direct effect of the capital limit is a fall in the number of female entrepreneurs, the partial equilibrium effect. Furthermore, existing female entrepreneurs are constrained by the distortion, run smaller businesses, and thus, demand less production labor. Hence, supply of female production labor rises and demand falls which unambiguously pushes wages down and profits up. For women, this is a second round effect, and thus, female entrepreneurship falls, although in a lesser magnitude compared to the partial equilibrium depicted in panel (b) of Figure 1. In the case of men, the fall in wages and rise in profits increases the number of entrepreneurs. Therefore, as first order effects, female entrepreneurship falls, whereas the opposite occurs for men. Figure 3 illustrates this case.

Figure 3:

General equilibrium analysis II.

4 Calibration

We calibrate the model to match observations of a benchmark economy, which we build using the average data for Denmark, Norway, and Sweden. As discussed in the introductory section, these economies are close to gender equality in the labor market, except regarding entrepreneurship.^[16] We now explain the specific data used to calibrate the model.

We choose a model period of a year. Along the balanced-growth path, consumption per capita and output per capita grow at a common rate x_c. To compute x_c, we consider the average annual growth rate of real GDP per capita in the benchmark economy during the period 1971–2019. Using data from the World Bank WDI dataset, we get x_c = 0.018. The U.S. Bureau of Economic Analysis (BEA) provides estimates of depreciation rates for different types of assets. This technological parameter should not be very different between the U.S. economy and our benchmark economy, so we take BEA data to calibrate this parameter. Specifically, we assume that the stock of capital is comprised of private nonresidential equipment and structures. We consider the simple average of the depreciation rates reported by the BEA for these types of assets to calibrate δ = 0.0978. In addition, in the model, ζα equals the share of capital. As we will explain later, we follow Guner, Ventura, and Xu (2008) to calibrate ζ. Then, we pick α such that ζα equals the capital income share of 30% observed in the U.S. data. The elasticity of substitution between women and men, σ, is calibrated to match the observed gender wage gap in the benchmark economy. Labor force participation is assumed to be egalitarian across gender in order to let the elasticity of substitution σ capture the whole gender wage gap that is observed in the benchmark economy. The discount factor β takes the standard value of 0.95 reported in the literature.

We now explain how we calibrate the span-of-control parameter, ζ, and the distribution of managerial skills. The span-of-control parameter, ζ, and the parameters governing the distribution of managerial abilities g(z) determine the size distribution of establishments, that is, the mean establishment size, the range of employment levels, as well as the share of total labor employed by establishments of different sizes. Following Guner, Ventura, and Xu (2008), we calibrate g(z) and ζ to match data on the fraction of establishments at different employment levels, and the share of total employment accounted for by large establishments. We use data from Eurostat Structural Business Statistics to compute the following targets: (i) mean establishment size, (ii) the fraction of establishments with 0–9 employees, 10–19 employees, 20–49 employees, 50 or more employees, and (iii) the share of total employment accounted for by large establishments, defined as those with 50 or more employees. We then select ζ and g(z) to match these statistics. As in Guner, Ventura, and Xu (2008), we assume that log-managerial ability is distributed according to a (truncated) normal distribution with mean m and variance s², and we impose that this distribution accounts for the bulk of production units, with a total mass of 1 − g ̂ . To account for the remainder of the distribution of establishments, we select a top value for managerial skill, z ̂ and its corresponding fraction, g ̂ . Hereafter we refer to managers with skills z ̂ as “superstars.” Lastly, the capital limit k ̄ is calibrated to match the fraction of female entrepreneurs in the benchmark economy.

Hence, there are seven parameters, σ, ζ, m, s², g ̂ , z ̂ , and k ̄ that we choose in order to match eight observations: gender wage gap, mean size, the fraction of establishments corresponding to four different levels of employees, the share of employment accounted for by establishments with 50 or more employees, and the fraction of female entrepreneurs. Table 2 summarizes our choices and Table 3 shows the performance of the model in terms of our targets and two additional untargeted moments.^[17]

5 Results

In this section we estimate the aggregate costs of the distortion described in Section 3.9. To do so, we build five counterfactual scenarios which gradually and progressively loosen the capital limit for female entrepreneurs. The first scenario reduces by 10% the gap between k ̄ and the maximum amount of capital used by non-superstar female entrepreneurs in a frictionless economy (S1).^[18] In the second and third scenarios, the latter gap is shortened by 20% (S2) and 50% (S3), respectively. In a fourth counterfactual scenario, all women are unconstrained, except the superstar managers (S4). Lastly, we consider a counterfactual economy where no woman is subject to a capital limit to run her business (S5). Then, we compare different outcomes in the baseline scenario to those in each of the five counterfactual economies. We consider three outcomes: aggregate output, output per worker, and total factor productivity (TFP).

Aggregate output, Y, is the total production carried out by both female and male entrepreneurs:^[19]

where z i * is the managerial skill of the marginal entrepreneur among type-i agents.

In this first analysis, we also compute the ratio of the average output produced per female manager in the counterfactual scenario to the output produced in the baseline scenario, r_f. This outcome is used to discuss the results for aggregate output losses, and we define it as follows:

where z ̄ f * and z ̃ f * are the managerial skill of the marginal female entrepreneur in the baseline and counterfactual scenarios, respectively.

Table 4 presents the results. We observe that the distortion causes a fall in aggregate output between 13.7% and 28.2%, depending on the counterfactual scenario considered. Counterfactual S1 implies that the distortion causes an output loss of 13.7%. In the case of the second and third counterfactuals (S2 and S3), we observe a fall in aggregate output by 16.3% and 19.5%, respectively, which is somewhat higher compared to the first scenario. Next, when comparing the baseline economy with one in which all women are unconstrained, except superstar managers, that is, scenario S4, the distortion causes an output loss of 21.7%. Lastly, output loss reaches 28.2% when the baseline economy is compared with a frictionless economy (S5).

Therefore, our counterfactual analysis suggests significant potential output gains derived from a small loosening in the capital limit for women entrepreneurs. However, potential gains are marginally smaller when the capital limit is further loosened. An exception to the latter occurs when the economy moves from a scenario where only the superstars are constrained to one where no women face a capital limit. The intuition behind the latter is as follows. The distortion forces women to run smaller businesses than what would be the optimal size, causing the marginal product of capital to be well above the rental price of it. Indeed, we observe in the second row of Table 4 that the ratio of the average output produced in establishments run by women in the counterfactual scenario to the output produced in the baseline scenario ranges from 4.1 (S1) to 9.9 (S5). Therefore, a small loosening of the distortion allows women to materialize the high marginal product of capital, but these benefits begin to evaporate when capital use becomes more and more intensive. However, superstar managers are highly skilled compared to the mass of managers, and thus, capital-skill complementarity in production counters somewhat the effect of capital intensity on the marginal product of capital. Hence, the marginal productivity of capital is still significant in businesses run by superstar female managers.^[20]

Next, we evaluate three measures of output per worker. Concretely, we build the average output per efficiency unit of labor, q₁, as

the average output per production worker, q₂, as

and the average output per manager, q₃, as

Table 5 presents the results using the same counterfactual scenarios as in Table 4. We observe that average output per efficiency unit of labor falls by 18.2% when the distortion considered in S1 moves to the level k ̄ . As concluded for aggregate output, this loss gradually increases when considering counterfactuals that progressively loosen the capital limit for women, reaching 33.7% when the baseline economy is compared with a frictionless one. In the case of average output per production worker the losses range from 15.1% to 27.3%, whereas for average output per manager the losses range from 24.5% to 49.3%.

Our next outcome measures aggregate productivity losses. We calculate this variable in two alternative ways. The first measure (TFP₁) considers TFP as the residual from an aggregate technology under a capital share of αζ and labor share of 1 − αζ, with no distinctions between workers and managers in the labor force. Concretely, we calculate

where

The second measure (TFP₂) considers a separation between workers and managers by their efficiency units and define aggregate labor as L + L ̃ , where L ̃ is

In this case, we have

We present in Table 6 the results for TFP. We observe that aggregate TFP falls between 12.2% and 19.5% as a consequence of the distortion. In this case, we observe moderate gains, in terms of TFP, from moving across the five counterfactual scenarios considered in Table 6.

Overall, this subsection shows that the capital limit k ̄ for women causes a fall in aggregate output between 13.7% and 28.2%, in average output per efficiency unit of labor between 18.2% and 33.7%, and in TFP between 12.2% and 19.5%. How do these effects compare to those found in the literature for other distortions? In this regard, the literature on the aggregate costs of financial frictions is indeed extensive, so we document here only a few studies that will provide an order of magnitude of the effects found in this paper.

Erosa (2001) analyzes financial frictions that arise from costly intermediation. The author estimates gains in output per capita of 40% from eliminating this type of distortion. Buera, Kaboski, and Shin (2011) analyze aggregate cost from financial frictions that take the form of imperfect enforceability contracts. The authors find that the aggregate TFP of the country with the least financial development would be almost 40% below the U.S. level. Greenwood, Sanchez, and Wang (2013) examine what they call the capital deepening and reallocation effects from financial intermediation. The analysis performed by the authors suggests that a country like Uganda could increase its output by 116% percent if it could adopt the world’s best practices in the financial sector which, however, amounts to only 29% of the gap between Uganda’s potential and actual output. Midrigan and Yi Xu (2014) analyze financial constraints arising from limits on the amount of debt and equity that agents can issue. They find that financial frictions may reduce the level of aggregate TFP by up to 40%. Hence, this brief discussion suggests that our results are somewhat smaller than those triggered by other frictions in the capital market, although still significant.^[21]

6 Conclusions

We developed a general equilibrium model to quantify the aggregate consequences of a discriminatory restriction for women in their access to business resources. Our counterfactual analysis shows that a gender-specific capital restriction causes an aggregate output loss between 14% and 28% and a total factor productivity loss between 12% and 20%. About two-thirds of the output loss is explained by the fall in aggregate productivity. Furthermore, we show that the potential aggregate gains of loosening the distortion studied in this paper would be mostly driven by the more intensive use of capital among existing female entrepreneurs and not by the entry of more women into entrepreneurship.

Our findings are policy-relevant. We show significant aggregate gains from more equal access to business resources across genders. Therefore, the evidence provided in this paper supports the need for a more vigorous discussion on mechanisms to reduce sources of discrimination against women entrepreneurs that still persist. For instance, passing and enforcing anti-discrimination laws in the financial system would be a first step. Also, the evaluation of some mechanisms to encourage banks and financial institutions to hire female loan officers or, at least, to reduce discrimination in hiring policies, could create a more favorable environment for women in the capital market. Lastly, some affirmative action policies could also be considered. For example, fiscal subsidies for women-owned businesses could be implemented as a “second best option” to mitigate the aggregate effects of any type of gender-driven misallocation. However, designing support policies for women entrepreneurs is not straightforward, since the social returns of these targeted policies are difficult to quantify and monitor. Hence, potential policies must be supported by better data and stronger institutional frameworks.

Further research should focus on assessing the effectiveness of specific policies targeting women entrepreneurs and analyzing the institutional framework in which these policies are more likely to be successful. Another avenue for future research is to explore an extension of the model developed in this paper to incorporate business risk as a determinant of business activity. One way to do this is by modeling an economy with idiosyncratic uninsurable shocks, decentralized financial markets, and gender-specific financial constraints. In that type of model, a precautionary motive for saving arises, which would interact with financial restrictions of different magnitude for women and men. It would be interesting to assess whether this latter type of framework reveals new implications regarding the aggregate effects of gender-specific constraints for entrepreneurs.