Inequality and entrepreneurial thresholds

Journal of Business Venturing xxx (xxxx) xxx–xxx

Contents lists available at ScienceDirect

Journal of Business Venturing journal homepage: www.elsevier.com/locate/jbusvent

Inequality and entrepreneurial thresholds ⁎

Soumodip Sarkara,b, , Carlos Rufínc, Jonathan Haughtond a

CEFAGE-UE and Department of Management, University of Évora, Palácio do Vimioso, Largo Marquês de Marialva, 8, 7000-809 Évora, Portugal Asia Center, Harvard University, Cambridge 02138, MA, USA c Department of Strategy and International Business, Sawyer Business School, Suffolk University, 8 Ashburton Place, Boston, MA 02108, USA d Department of Economics, Suffolk University, 8 Ashburton Place, Boston, MA 02108, USA b

AB S T R A CT We explore the relationship between inequality and entrepreneurial activity. Drawing on crosssectional data from a largescale survey of the economic conditions of individuals across India, we develop a number of dimensions of inequality to explore empirically how inequality interacts with entrepreneurship, operationalized as self-employment or as employing other people. We find compelling evidence that there are thresholds to becoming self-employed, and even more so to assembling the combinations of resources and personal attributes required to become an employer. Greater inequality leaves more people unable to make the transition to self-employment, leaving casual laboring as the occupation of necessity. At the same time, inequality increases the number of employers in a society, by concentrating resources - particularly land and finance - enough for significant numbers of people to be able to cross this higher threshold. Lastly, greater differentiation into social or religious groups curtails the ability to cross either entrepreneurial threshold, presumably by limiting the extent and benefits of social networks of value for entrepreneurship.

Executive summary One of the most important questions in the literature on business venturing is why some people become entrepreneurs while others do not. Largely missing from this research is a discussion of the links between entrepreneurship and inequality. Our key idea is that there are thresholds to becoming self-employed, and even more so to pulling together the combinations of resources and personal attributes required to become an employer. Greater inequality leaves more people unable to make the transition to self-employment, leaving casual laboring as the occupation of necessity. This same inequality increases the number of employers in a society, by concentrating resources – such as land or finance – enough for significant numbers of people to be able to cross this higher threshold. These entrepreneurs are not simply born, they are created. We test our hypotheses linking inequality with the propensity to be self-employed, or an employer, using data from the 68th round of the Socio-Economic Survey of households, conducted by the National Sample Survey Office (NSSO) of India between July 2011 and June 2012. The survey sampled 101,724 households, encompassing 456,999 individuals. Of these, 134,665 were working, and this is the subset of interest to us: 23% worked for a regular wage, 36% earned wages from casual labor, 40% were self-employed, and 1% were employers. Based largely on estimates of logistic equations relating employment status to measures of local inequality, we find that, at the

⁎ Corresponding author at: CEFAGE-UE, Department of Management, University of Évora, Palácio do Vimioso (Gab. 224), Largo Marquês de Marialva, 8, 7000-809 Évora, Portugal. E-mail addresses: [email protected] (S. Sarkar), crufin@suffolk.edu (C. Rufín), jhaughton@suffolk.edu (J. Haughton).

https://doi.org/10.1016/j.jbusvent.2017.12.009 Received 8 May 2016; Received in revised form 27 November 2017; Accepted 20 December 2017 0883-9026/ © 2017 Published by Elsevier Inc.

Please cite this article as: Sarkar, S., Journal of Business Venturing (2017), https://doi.org/10.1016/j.jbusvent.2017.12.009

Journal of Business Venturing xxx (xxxx) xxx–xxx

S. Sarkar et al.

broadest level, greater inequality appears to favor the formation of employer-owned firms, while for the most part reducing the extent of self-employment, after controlling for other relevant influences on entrepreneurship. There appear to be two thresholds, one representing a move into self-employment, and the other reflecting the jump to employer status. Crossing these thresholds—being self-employed or an employer—involves assembling combinations of resources, such as education or money, and personal attributes, such as social status, religious affiliation, gender, or family size, that can sustain entrepreneurial action. The combinations available to some individuals may simply be insufficient to cross one or both thresholds. Specifically, there is a large set of people – whom we can call the very poorest – that have difficulty crossing the first one: they lack the land (in rural areas), education, networks, and financial resources required to be sustainably self-employed, and fall back on casual wage laboring. Moreover, inequality worsens this situation by leaving more people in poverty, where self-employment is out of reach; and greater social heterogeneity (as measured by either caste membership or religious affiliation) makes it even harder to develop robust social networks that can help cross the threshold to self-employment. The second threshold, from self-employed to employer, remains a formidable barrier. The qualitative leap that is needed requires considerable personal and household resources, and in an emerging economy such as India, this is easier to achieve if there is considerable inequality, in effect concentrating enough resources and capacity in a smaller set of individuals to permit the emergence of a class of employers. The paper makes five contributions. First, at a theoretical level, we show that inequality affects entrepreneurship by altering the proportions of those who are able to cross the first threshold into self-employment, and the second and much higher threshold into becoming an employer. Second, we construct a variety of empirical measures of inequality, which allows us to test the proposition that these affect entrepreneurship, after controlling for other standard variables, in subtler and more complex ways than simply inequality in income levels. Third, we estimate the relationships between inequality and the propensity to entrepreneurship. The results are striking: greater inequality in the distribution of resources is associated with a higher propensity to be an employer, and a lower likelihood of being self-employed. This leads us to conclude, as our fourth contribution, that it is essential to retain the distinction between employers and the self-employed – they cannot simply be lumped into a single class labelled “entrepreneurs – and it also leads us to question the usefulness of the concept of ‘necessity entrepreneur’”, since it would appear that becoming an entrepreneur requires an act of volition, and is not generally the option of last resort. Fifth and last, our work offers a number of policy implications. The most obvious is that the continuing expansion of education will, for the foreseeable future in India, help increase the number of employers. Moreover, social norms about caste and gender in India appear to be holding back entrepreneurship; thus, measures that weaken caste and gender barriers would favor entrepreneurship. A land reform that would create greater equality in land holdings would have no impact on urban entrepreneurship, but it would increase the number of selfemployed in rural India. Lastly, helping the poorest will require a wider menu of interventions, although greater entrepreneurial activity would certainly help this group indirectly insofar as it would boost economic growth and create more jobs. 1. Introduction One of the most important questions in the literature on business venturing is why some people become entrepreneurs while others do not. The voluminous literature on the subject has mainly focused on personal and household factors such as education and inherited wealth, and on some institutional variables such as measures of the rule of law. Largely missing from this research is a discussion of the links between entrepreneurship and inequality. Inequality of income is rising in most rich countries (Piketty, 2013), and some believe that it is acting as a brake on economic growth by increasing financial fragility, stalling improvements in health and education, and causing “investment-reducing political and economic instability” (Ostry et al., 2014, p.4). Yet it is also possible that inequality may affect economic growth by limiting (or favoring) entrepreneurial activity. The relationship between inequality and entrepreneurial activity is the fundamental issue that we explore here. In a perfectly homogeneous world we could not predict who would become an entrepreneur or be self-employed; it would be attributable to chance, since there is no variation in the potential explanatory variables. Identifying who is more likely to be an entrepreneur, and why, thus rests on the existence of heterogeneity within a population, such as inequality in access to financial resources, education, or other valuable possessions. This heterogeneity operates at two levels, individual and collective. At the individual level, having better access to financial resources and more education (for example) appears to foster entrepreneurship in capital- and skill-intensive sectors (Lofstrom et al., 2014). Effects such as these have been already been explored in considerable depth. By contrast, of interest to us is the possibility that inequality per se also affects the propensity to be an entrepreneur or self-employed. We examine the possible reasons for this below, but the key idea is that collectively, inequality may change the ability of people to gather together the resources needed to be successful entrepreneurs. Our overarching view is that there are thresholds to becoming self-employed, and even more so to pulling together the combinations of resources and personal attributes required to become an employer. Greater inequality leaves more people unable to make the transition to self-employment, leaving casual laboring as the occupation of necessity. This same inequality increases the number of employers in a society, by concentrating resources – such as land or finance – enough for significant numbers of people to be able to cross this higher threshold. These entrepreneurs are not simply born, they are created. Working inductively, we seek to answer the question of whether inequality affects entrepreneurship through the analysis of crosssectional data from a large-scale survey of the economic conditions of individuals across India. We use a number of dimensions of inequality to explore empirically how inequality interacts with entrepreneurship, operationalized as self-employment or as employing other people. Using data from India makes good sense: only in a very large country can one observe sufficient variations in 2

Journal of Business Venturing xxx (xxxx) xxx–xxx

S. Sarkar et al.

inequality at the level of the relevant geographic area for the effects to be measurable. While this kind of dataset does constrain, as a practical matter, what we can measure – we have no information on cognitive skills, or entrepreneurial aspirations, for instance – it does provide information on the structure, at a suitably local level, of social groups (including caste and religion), educational attainment, land ownership, and consumption per capita. We make five contributions. First, we articulate the ways in which inequality may be expected to affect the number of entrepreneurs: inequality changes the proportions of those in a working population who are able to cross the threshold into selfemployment, or the more demanding threshold into becoming an employer. Second, we are able to construct operational measures of inequality along a number of dimensions – land holdings, per capita consumption, education, social group, and religion. Third, using data from India, we are able to link these measures to the propensity to be an entrepreneur (whether an employer, or self-employed), even after controlling for a large number of more traditional determinants of entrepreneurship. Broadly speaking, we find that greater inequality in the distribution of resources is associated with a higher propensity to be an employer, and a lower probability of being self-employed. Fourth, we find the distinction between the two types of entrepreneurs – employers, and self-employed – to be central to understanding the dynamics linking inequality to entrepreneurship; we also find the notion of “necessity” entrepreneurship to be unhelpful. And fifth, we are able to draw some usable conclusions for policy purposes. The next section examines current scholarship on the entrepreneurial process, and draws on this literature in order to develop a set of hypotheses regarding how inequality and heterogeneity may affect the propensity for self-employment and for being an employer. This is then followed by a description of our database, empirical approach, and the results of our statistical analyses. The last section offers a discussion of our results and overall conclusions. 2. Theory and hypotheses To explore systematically how inequality and heterogeneity may interact with entrepreneurship, we need to begin with a model of the process of entrepreneurship. Whether measured in practice by self-employment,1 or by being an employer, entrepreneurs combine their own personal attributes with resources, such as capital, land, or know-how, in order to produce goods and services and develop markets for these products. The process both creates and identifies opportunities; Sarasvathy (2004) refers to the process as one of “constructing corridors.” Being an entrepreneur calls for endless problem solving; Hirschman (1965) thinks of it as a process of making obstacles vanish. An individual's entrepreneurial propensity thus derives from the perception of opportunities that can be pursued through entrepreneurship, the availability of resources to pursue such opportunities, and the intrinsic motivation of the individual to do so (Alvarez and Barney, 2004; Alvarez and Busenitz, 2001; CBS et al., 2011; Foss et al., 2008; Parker, 2004; Shane, 2003; Shane and Venkateraman, 2000). The demands of the entrepreneurial process, however, differ according to the size of the venture envisaged by the individual. At the simplest level, a person can pursue self-employment; a larger venture will typically require the hiring of employees, i.e. becoming an employer. We can think of these as two entrepreneurial thresholds that require increasing amounts and kinds of resources relative to not being an entrepreneur (Honiga, 1998; Roy and Wheeler, 2006), even if these resources can be combined in multiple ways according to the nature of the opportunity and the entrepreneur's creativity and innovation, as emphasized for instance in the recent literature on entrepreneurial bricolage (Baker and Nelson, 2005). Given this view of the entrepreneurial process, we are interested in examining how heterogeneity or inequality in opportunity, resources, and motivation affect how these thresholds are crossed or not. For this purpose, it is essential to note, as the entrepreneurship literature has highlighted, that opportunity, resources and motivation do not exist in the abstract, but arise from a variety of individual, social, and economic conditions. Psychological factors and other individual characteristics such as cognitive abilities shape the ability of individuals to recognize or envisage opportunities, the willingness to pursue them, and the capacity to turn ideas into successful outcomes through action (Locke, 2000; Lofstrom et al., 2014; Parker, 2004; Shane, 2003; Shane et al., 2003). Since its beginnings, the field of entrepreneurship research has focused on the personal characteristics of entrepreneurs, such as a higher tolerance for risk, as an explanation for the propensity of certain individuals to become entrepreneurs (McClelland, 1967). Sridharan et al. (2014) point specifically to several cognitive or ideational characteristics that appear to foster entrepreneurship in subsistence settings: “entrepreneurial spirit and acumen,” competencies such as abstract thinking and social interaction, and attitudes toward risk. Social conditions also influence opportunity, affect the availability of resources, and impact motivation. The most immediate social force acting on the individual is the family, particularly by offering resources that an individual can tap, but also as a source of motivation (Hout and Rosen, 2000; Parker, 2004). To manage risks and make their endeavors more viable, the entrepreneur may rely on family assets such as household savings to start a microenterprise or to deal with slowdowns in business or other contingencies, as well as unpaid labor (Venugopal et al., 2015; Viswanathan et al., 2010), in a context of severe resource constraints (Webb et al., 2014). And in developing country settings, distrust of institutions and lax enforcement of contracts lead to greater reliance on family networks (Bradley et al., 2012). However, a larger family can also place greater demands on an entrepreneur, limiting his or her ability to escape the situation of necessity (Viswanathan et al., 2010). Beyond the family group, opportunity, resources, and motivation depend additionally on the broader social groupings to which an individual belongs—gender, class, caste, religion, ethnicity, culture, and community (Audretsch et al., 2013; Mueller, 2006; Noseleit, 2010; Parker, 2004). For example, individuals belonging to a social category that confers on them greater credibility or legitimacy in 1

We follow the commonly used definition of “self-employed” as someone who does not earn income through wage labor, and does not employ any salaried workers.

3

Journal of Business Venturing xxx (xxxx) xxx–xxx

S. Sarkar et al.

their community, and hence possibly a greater self-confidence, are likely to assign a lower risk to entrepreneurship than individuals of lower status. Thus, Wagner (2005) found that being a male significantly increased the probability of being self-employed relative to being unemployed or in paid employment. Another example concerns network effects. Entrepreneurship research has found such “social capital” to be an important resource for entrepreneurs (Birley, 1986; George et al., 2015; Kim and Aldrich, 2005). Bhagavatula et al. (2010) found networks to be important for small-scale handloom entrepreneurs in rural India, and the subsistence marketplaces literature also provides extensive qualitative evidence on the role of social capital for subsistence entrepreneurs, showing that “people living in subsistence conditions (…) are members of densely networked social and kinship communities (…) on which they draw to offset their lack of financial resources, access, and skills” (Viswanathan et al., 2010: 1). Networks help spread information about an entrepreneur's relationships with customers or suppliers, creating reputations that can attract additional customers, and reducing risk and uncertainty by facilitating credit and other help to address unexpected shocks, such as an illness in the entrepreneur's family (Viswanathan et al., 2008). Thus, “social capital is key to understanding the micro-level entrepreneurial process in subsistence marketplaces” (Viswanathan et al., 2014: 3; also Viswanathan et al., 2012). Additionally, the high-density settings of cities enable much greater random interaction among individuals, resulting in greater exchange of information and thus the possibility of identifying entrepreneurial opportunities (Glaeser, 2000). Lastly, economic conditions affect entrepreneurial decisions. Different industries require different sets of resources for an entrepreneur to compete in a viable manner, and offer varying opportunities according to industry conditions and trends (Shane, 2003). Entrepreneurship scholars have highlighted the importance of factors such as the presence of “feeder” industries and institutions, a skilled labor force, the accessibility of suppliers and customers, and the proximity to high-quality universities, for this form of entrepreneurship (Stenholm et al., 2013: 182). Thus, urban areas are more likely to score higher than rural areas on all of these factors, simply because of the greater physical proximity, and thus greater accessibility, of other industries, institutions, skilled labor, suppliers, customers, and educational entities. Additionally, the macroeconomic context dictates resource availability as well as opportunities and motivations for entrepreneurship (Fiess et al., 2010; Parker, 2004; Schuetze, 2000). We now examine heterogeneity or inequality in these dimensions, and ask how they are expected to affect the propensity to be self-employed, or to be an employer. It is important to emphasize that we are interested here in the effects of heterogeneity or inequality per se – i.e. the existence of differences across individuals, households, and groups within a specific geographic setting – rather than the specific attributes of an individual, household, or group on entrepreneurial propensity, which as shown above, have already been well researched. 2.1. Resources There is enormous potential heterogeneity in the degree of access to resources. Unequal distributions of wealth or income will limit the ability of some individuals, households, groups, or even entire communities to marshal enough resources to be able to become employers (Ardagna and Lusardi, 2010; de Mel et al., 2010; Ho and Wong, 2007; Parker, 2004; Webb et al., 2014; XavierOliveira et al., 2015). At the same time, some of these individuals who are deprived of opportunities to become employers may still be able to pursue self-employment, particularly if motivated by necessity (Alvarez and Barney, 2014; Collins et al., 2010; Lippmann et al., 2005; Viswanathan et al., 2010). However, lack of resources cannot be construed in a dichotomous way. Individuals can acquire resources from a variety of sources in different combinations to cross the thresholds of self-employment and employing others. Heterogeneity and inequality must be thus be seen as posing imperfect constraints on entrepreneurial action. At the limit, there may be certain resources or combinations of resources that are a necessary minimum for self-employment or for becoming an employer. We explore these possibilities in our empirical specifications and the analysis of our results. In fact, in response to heterogeneity and inequality, some individuals may form cohesive networks that facilitate entrepreneurship by pooling or exchanging resources—in other words, may accumulate social capital, substituting an intangible resource for the relative lack of tangible resources such as land or material wealth (Birley, 1986; Kim and Aldrich, 2005; Uzzi, 1997). For instance, some groups may become cohesive as a reaction to the presence of other groups, leading to stronger efforts at mutual aid, including financing business ventures; this is certainly the case of the Mourides of Senegal (Fritzon, 2016). Since this possibility is closely related to the development of entrepreneurial opportunities out of heterogeneity, we will return to it when considering the impact of heterogeneity on entrepreneurial opportunity below. For now, we consider the effects of heterogeneity and inequality in the distribution of key entrepreneurial resources—land, education, and financial resources—on the propensity for self-employed or employer status. 2.1.1. Land Inequality in the distribution of landholdings is unlikely to have much impact on the vigor of entrepreneurship in urban areas, or on the number of employers in rural areas, other than the obvious case of large landowners (who would naturally employ laborers). However, an unequal land distribution is likely to leave a substantial proportion of the rural population without enough land to support their families, obliging them to turn elsewhere for a livelihood. A common result in societies where landholdings are highly concentrated is employment as casual farm wage labor. But another major option is self-employment. Our first hypothesis is thus: Hypothesis 1. Greater inequality in the distribution of landholding (i) Has no effect on entrepreneurship in urban areas; and 4

Journal of Business Venturing xxx (xxxx) xxx–xxx

S. Sarkar et al.

(ii) Increases the amount of self-employment (but not employers) in rural areas. 2.1.2. Education At the individual level, education has an ambiguous effect on entrepreneurship. Those with more education are better equipped to function as employers and to be self-employed, but they also have a higher opportunity cost since they have more options for wage work (Van Der Sluis et al., 2008). At the collective level, greater educational inequality leaves more people behind, ill-equipped to strike out on their own as selfemployed. But it also implies the existence of a pool of more highly-educated individuals with the skills to run more complex businesses. This reasoning leads us to our second hypothesis: Hypothesis 2. Greater inequality in the distribution of educational attainment: (i) Increases the number of employers; and (ii) Reduces the number of self-employed. 2.1.3. Consumption per capita Inequality in consumption per capita is one of the more standard measures of economic inequality. A more concentrated distribution of per capita consumption may allow a subset of individuals to marshal the resources needed to start and develop an enterprise, especially as the initial capital typically comes from family and friends (Xavier-Oliveira et al., 2015). This should make it easier for relatively complex businesses to develop, increasing the number of employers. Lecuna (2014) argues that entrepreneurs accumulate more income than others, and in a disequalizing model this inequality is perpetuated in subsequent generations, as the probability of being an entrepreneur increases with inherited wealth. He finds some support for an association between greater income inequality and more entrepreneurship, based on GEM data for 54 countries over the period 2004–09. The flip side is that those at the lower end of an unequal distribution may end up self-employed out of economic necessity. Greater inequality implies that a higher proportion of the population will be deprived of resources (Carsrud and Brannback, 2011; George et al., 2015; Lippmann et al., 2005; McMullen et al., 2008). Moreover, a lower relative level of resources also implies the perception of lower opportunity costs and lower risk of self-employment (Naudé, 2010), so with a higher proportion of the population deprived of resources, self-employment will be higher (Bapuji and Neville, 2015). We thus expect the following: Hypothesis 3. Greater economic inequality increases the number of both employers and the self-employed. 2.2. Opportunity and motivation Heterogeneity can also create entrepreneurial opportunities. The most obvious ones are in the form of niche markets that can be served (Porter, 1980), such as restaurants in Malaysia that serve pork (to Chinese Malaysians) and those that do not (for Malays). More interestingly, entrepreneurial opportunity can arise as well through the deliberate and creative exploitation of heterogeneity, by means of the innovative assemblage of different, or unevenly distributed, resources to develop new products or processes (Baker and Nelson, 2005). Here, entrepreneurial action creates opportunities primarily for the “assembler,” but also for the specialized supplier of a specific input. For example, in the Indian context, a significant recycling industry has arisen in informal settlements such as Dharavi in Mumbai by exploiting the ability of microenterprises in these areas to specialize in the extraction and commercialization of different recyclable materials, such as plastics or metals (Curry, 2016). We can therefore expect greater entrepreneurial propensity in more heterogeneous communities, which leads us to our final hypothesis: Hypothesis 4. Greater individual and group heterogeneity increases the number of employers and the self-employed. It is widely observed that some groups of individuals or entire communities may favor entrepreneurial initiative more strongly than others (Clark and Drinkwater, 2000, 2010; Hart and Acs, 2011; Parker, 2004; Portes and Bach, 1985). In some cases there may be “blocked minorities”, who channel their energies into business in response to social or legal constraints on other activities, such as Quakers in nineteenth-century England (Jackson, 2010). However, there is no clear systematic effect of differentiation in motivation on overall entrepreneurial propensity, since even if some individuals or groups may be more motivated to attempt to become employers or to pursue self-employment, others may be less so, relative to a situation of homogeneity. This is a clear case where the individual effects of motivation, which do matter, are distinct from the collective effects of the distribution of this variable across the population. 3. Data and empirics 3.1. Data The data that we use to test our hypotheses come from the 68th round of the Socio-Economic Survey, conducted by the National 5

Journal of Business Venturing xxx (xxxx) xxx–xxx

S. Sarkar et al.

Sample Survey Office (NSSO) of India between July 2011 and June 2012 (NSSO, 2012). This survey collected information on employment and unemployment, and on household spending patterns. It uses a stratified cluster sampling method: households are sampled in every district in the country; within the districts, the first-stage sampling units (FSUs) are villages or urban blocks. The sampling weights are known, and have been used in the summary data set out in Table 1. The survey sampled 101,724 households, encompassing 456,999 individuals. Of these, 134,665 were working, and this is the subset of interest to us: 23% worked for a regular wage, 36% earned wages from casual labor, and the remaining 41% were selfemployed. A full listing of the variables that we use, and the relevant summary statistics, is set out in Table 1. 3.2. Dependent variables Ideally, we would like to identify those, in the sample of working people, who are truly entrepreneurs, whether out of necessity or because they seized opportunities or aspired to set up on their own. Such data are, however, extremely rare in developing countries; the sample of 156 low-income women studied by Venugopal et al. (2015) is an exception, and we are unaware of any such data being available for developing countries other than from the GEM, which, as an international survey, does not provide national data disaggregated to the district level. We too lack some of the information that would be required to identify “true” entrepreneurs, and we agree, following Parker (2004), that “at the practical level … the closest approximation to the manifestation of entrepreneurship that appears to be suitable … will usually be 'self-employment'.” Lippmann et al. (2005: 7-8) make a similar point, as do Evans and Leighton (1989). In the context of developing countries, Acs (2006: 98) confirms the close correspondence between self-employment and entrepreneurship when he states that “low-income countries like Uganda, Peru and Ecuador have very high levels of self-employment and therefore have high levels of entrepreneurial activity as measured by the GEM program.” The data allow us to distinguish between working individuals who are self-employed, and those who are wage earners, Most of the self-employed are farmers or artisans, who do not hire any labor; henceforth, we refer to this group as self-employed. But there is also an important group, representing 1.8% of the working population, who hire workers. This group includes large farmers, who hire laborers at certain points in the year. As these individuals are likely to be landowners, rather than individuals who have set up new ventures, we exclude them from our analysis. Our exclusion leaves a residual group, constituting about 1% of the working population in the sample (1600 individuals), that we consider qualitatively distinct, and that we classify in our baseline analysis as employers. The models we estimate seek to measure the propensity of a working individual to be self-employed, or to be an employer. It is tempting to equate the 58,000 self-employed individuals with “necessity entrepreneurs,” in contrast to the employers, who might be thought of as “opportunity entrepreneurs,” but in the absence of more detailed information about the motivations of the self-employed and employers any correspondence among these terms is likely to be inexact. Among employers, most (55%) declare their occupation to be “director or chief executive,” whereas for the self-employed, the most common occupation is “market gardeners and crop growers” (30%), followed by “directors or chief executives” at 16%. Employers are twice as likely as the self-employed to come from the “forward” (highest) social class, and are predominantly urban. A quarter of employers have some higher education, compared with 7% for the self-employed. More than four-fifths of the employed population in India is male – a high proportion by world standards – but among employers this figure rises to 95%. Given their higher levels of education, and urban location, it comes as no surprise that the consumption per capita of employer households is more than twice the national average (i.e. 3450 rupees per person per month); and their earnings average 283 rupees per day (about USD 5.20 in 2011–2012). 3.3. Explanatory variables and controls Summary statistics related to the explanatory variables are shown in Tables 1 and 2, and may be classified as measures of inequality, individual resources, and other controls. 3.3.1. Inequality The central aim of our paper is to examine the role played by inequality in the propensity to be self-employed (or an employer). A set of measures of inequality are set out in summary form in Table 1, and in more detail in Table 2. The first two measures measure inequality in consumption per capita, and in land ownership per household. Given that these variables are continuous, we construct Gini coefficients, which are widely used to measure inequality, and range from 0 (perfect equality) to 1 (complete inequality). When the Gini coefficient for consumption per capita is computed at the level of each of the 626 districts, and averaged, we get a value of 0.277, indicating that within any local area there is only modest inequality in living standards. However, as the unit of observation is increased – to region, state, or India overall – the measure of consumption inequality also rises, as Table 2 shows. As expected, the Gini coefficients for land ownership show considerable inequality (Table 2). These are, however, imperfect measures, because we are using land area rather than land value (which was not available): those who own a small plot of land in a large city may be more land-rich than someone who owns many hilly acres. To measure social group heterogeneity, we use the area's Herfindahl-Hirschman index (HHI) corresponding to the degree of heterogeneity in the four major social groupings in the area: scheduled tribes and classes, other “backward” classes, and others (the “forward” classes). When the data on social categories or market shares fall into a few large groups like this, it is common to use the HHI, which is given by the sum of the squared membership shares of each of these groups. It varies from about 0 (wide diversity) to 1 6

Journal of Business Venturing xxx (xxxx) xxx–xxx

S. Sarkar et al.

Table 1 Summary statistics of variables used. Full sample

Dependent variables Occupation (%) Regular wage work Casual wage labor Self-employed and employer of which: Employer of which: Non-farm employer (main job) Non-farm employer (any job) Self-employed Inequality/heterogeneity measures Gini: consumption/cap, district level Gini: land owned/hh, district level Herfindahl index: educationa Herfindahl index: social groupb Herfindahl index: religion Resource measures Land owned (%) No land Under 2 ha. 2 ha. or more Land owned (in thousandths of ha.)c Consumption per capita (Rp/mth) Earnings per day (Rp) Educational level achieved (%) Illiterate Less than primary Primary Middle school Secondary school Higher education Location Rural (%) Urban (%) Control variables Age (years) Marital status (%) Never married Currently married Widowed Divorced/separated Family size Household size (number) Personal characteristics Social group (%) Scheduled tribe Scheduled class Other “backward” class Other (“forward” class) Gender (male) (%) Religion (%) Hindu Muslim Christian Sikh Jain Buddhist Main industry where employed (%) Agriculture, forestry, fishing, mining Manufacturing Trade Construction Public administration

Self-employed

Employer

Mean

Mean

Mean

Standard devn.

22.7 35.9 41.4

41.9 48.0 49.2

1.8

13.3

0.97 1.00 39.6

9.78 9.95 48.9

100.0

0.281 0.718 0.181 0.439 0.709

0.064 0.119 0.035 0.116 0.207

0.272 0.696 0.186 0.435 0.710

0.327 0.772 0.161 0.464 0.665

13.0 80.2 6.7 536 1613.4 246.7

33.7 39.8 25.1 1244 1553.7 336.5

7.1 79.8 13.1 900 1515.1 140.5

14.7 78.2 7.1 470 3449.2 283.1

28.6 11.3 13.5 16.6 20.0 10.0

45.2 49.0 34.2 37.2 40.0 30.0

28.0 11.6 13.9 18.0 21.5 7.0

4.1 4.6 8.5 16.3 41.0 25.5

69.8 30.2

45.9 45.9

74.7 25.3

28.0 72.0

39.1

13.1

43.1

42.3

14.6 79.5 5.4 0.6

35.3 40.4 22.6 7.5

6.5 87.8 5.3 0.4

6.4 90.6 2.9 0.1

4.9

2.3

5.1

5.1

9.3 20.4 42.7 27.6 82.2

29.0 40.3 49.5 44.7 38.3

8.9 13.9 45.9 31.3 88.4

1.6 4.7 36.8 56.9 95.3

83.0 12.2 2.2 1.6 0.3 0.8

37.6 32.7 14.8 12.4 5.0 8.7

82.0 13.7 2.0 1.5 0.4 0.4

72.0 17.1 4.7 3.7 0.2 0.3

42.5 13.1 10.0 12.5 2.1

49.4 33.8 30.0 33.1 14.5

51.7 12.1 17.4 3.1 0.0

7

100.0

3.7 23.9 33.3 10.7 0.0 (continued on next page)

Journal of Business Venturing xxx (xxxx) xxx–xxx

S. Sarkar et al.

Table 1 (continued) Full sample

Services

Self-employed

Employer

Mean

Standard devn.

Mean

Mean

19.6

41.3

15.7

28.4

Notes: Sample size: 134,665 (except for earnings: 69,530; marital status: 134,646; religion: 133,253; and land holding: 118,092). All statistics reported here take account of sampling weights. a Based on 8 categories of educational achievement: illiterate, less than primary, primary, middle, secondary, higher secondary, graduate, and post-graduate. b Based on 4 social groups shown above. c For households owning land. Table 2 Measures of inequality/heterogeneity, India, 2011–2012. All India

Gini coefficient Consumption per capita Land owned per capita Wage per day Herfindahl-Hirschman index Educational attainment Social group Religion Number of units

State

Region

District Overall

Self-employed

Employer

0.360 0.771 0.471

0.321 0.741 0.443

0.305 0.731 0.424

0.277 0.710 0.383

0.272 0.696 0.383

0.327 0.772 0.417

0.162 0.298 0.570 1

0.177 0.395 0.704 35

0.179 0.411 0.680 88

0.184 0.438 0.715 626

0.186 0.435 0.710 58,514

0.161 0.464 0.665 1593

Notes: Sample size: 456,999 individuals (456,968 for District information on land; 456,931 for District information on wages). For self-employed, sample sizes are 58,515 for self-employed, and 1594 for employers. Larger values of the Gini coefficient represent greater inequality. Smaller values of the Herfindahl-Hirschman Index represent greater heterogeneity.

(complete homogeneity).2 A lower HHI would imply a greater degree of social heterogeneity within a specific territory, which is likely to limit the formation and size of social networks in a community or area. For the four social groups that we consider, there is a fairly even distribution nationally (index of 0.30) but greater concentration at the district level (index of 0.44). When religion, rather than social group, is the focus, the district-level HHI is 0.715, which reflects the predominance of Hinduism. For educational attainment, which is only available in a few categories, we again compute an HHI. Our calculations show considerable variation in years of educational attainment (i.e. a low value of the HHI index), with somewhat more inequality within districts than nationally (Table 2). 3.3.2. Individual resources These take the form of both physical and human capital. The latter is measured by a person's level of educational attainment. For other forms of capital, we lack information about the respondent's wealth, but use land ownership as a proxy for these resources. This is a better indicator of wealth than it might seem at first blush, because in countries with lower levels of financial market development, such as India, land is a preferred vehicle for the accumulation of savings and wealth: a sample drawn from India showed that 99% of very poor households living on average incomes of less than $1 a day per person owned some land apart from the land on which the household resided (Banerjee and Duflo, 2007). We also test the robustness of our results to using consumption as an indicator of financial resources (see below). 3.3.3. Other controls For social class, we rely on the survey's classification of respondents into “social groups,” which appear to be a mixture of social class and caste (closely related in India): “forward,” “other backward class,” “scheduled caste” (e.g. dalits, formerly called “untouchables”), and “scheduled tribe.” Gender is directly available from the database, as is family size, and whether the household is urban or rural. We also control for a person's age, marital status, religion, and the sector in which he or she works. Such factors have been shown to have strong, but not necessarily linear, influences on the perception of risks and potential benefits related to entrepreneurship. Controlling for the sector of activity of the venture can account for unobserved heterogeneity in opportunities and obstacles to entrepreneurship in different economic sectors (e.g. barriers to entry; Lofstrom et al., 2014). Lastly, we often include specifications with dummy variables for the state in which the respondent is located (“state-level fixed effects”), in order to control for other unspecified institutional, legal, and cultural differences across states. 2 We distinguish four social groups. If all have an equal share of the population, the HHI index would be 0.25, and this is the minimum possible value of the index in this case.

8

Journal of Business Venturing xxx (xxxx) xxx–xxx

S. Sarkar et al.

3.4. Hypothesis tests (estimation models) To explore the determinants of entrepreneurship, we estimate a set of formal models. Let Y be a binary variable that is set to 1 if the individual is an employer (Model 1) or self-employed (Model 2). Then we are interested in estimating a model of the form

Y = β0 + β1 (Resources) + β2 (Personal Characteristics) + β3 (Family Size) +

β4 (Location) + β5 (Inequality) + β6 (Controls) (1)

+ϵ

Eq. (1) cannot be estimated satisfactorily as a linear equation because the error terms (ε) will be non-normal, and the estimated values of Y could be out of bounds. Instead we estimate it in logistic form, so

F (Y ) =

1 1 + e−Y

(2)

This ensures that the probability of being an employer (or self-employed) – our dependent variable – is bounded by 0 and 1. 3.5. Results The key estimation results are shown in Table 3, for models that seek to explain who is an employer (columns 1–3), or selfemployed (columns 4–6). The reported numbers are average marginal effects that show how, on average, a unit increase in the independent variable affects the probability of being an employer. For instance, when consumption inequality rises by 0.1 (e.g. from the average of 0.277 to 0.377, Table 2), the probability of being an employer would rise from 1% to 1.8% (i.e. from 0.01 to 0.01 + 0.077 × 0.1), while the probability of being self-employed would fall from 39.5% to 34.5%. All of the underlying regressions are robust in that they correct for clustering at the level of the first-stage sampling unit. Columns 1 and 4 report the results of the basic models, while columns 2 and 4 include state-level fixed effects (coefficients not shown), and columns 3 and 6 give the results from estimating a multinomial logit equation where workers are grouped into four categories: employer, self-employed, regular wage earner, and casual wage earner. The results do not vary greatly from one specification to the next, although in most cases the inclusion of state fixed effects serves to attenuate the effects of the other variables, as expected. Table 4 provides the results for inequality and heterogeneity for additional specifications: dividing the sample into rural and urban residents for the self-employed in columns 1–2 and employers in columns 3–4, and adding heterogeneity in religious affiliation in columns 5 and 6 for the self-employed and employers, respectively. The results are clear: the factors that drive the prevalence of self-employment are, in a number of important dimensions, markedly different from those that determine who is an employer. Our hypotheses focus on the effects of ambient inequality on the propensity to be self-employed or an employer, so we report these results first, followed by a briefer discussion of the effects of the other independent and control variables. 3.5.1. Hypothesis 1: Inequality in landholding Our first hypothesis is that inequality in landholding has no effect on entrepreneurship patterns in urban areas. The evidence supports this. As the estimates in Table 4 show, the distribution of landholdings has no significant effect in the urban context, but has clear effects in rural areas. In rural areas, however, a greater concentration of landholdings is associated with less self-employment, and more employers. These results are surprising: we had expected inequality in landholdings to be associated with more (near) landless households, pushed by necessity toward self-employment. Instead it appears that inequality in landholdings in rural areas is associated with more wage labor – perhaps working on the large farms as agricultural laborers. Greater inequality in landholdings is also associated with having more employers, although the effect is small and not as strongly statistically significant. Areas with greater land inequality may be fertile grounds for recruiting employees; or members of landowning families may be able to establish themselves more easily as employers, although for now these are just surmises, and more data would be needed to test these speculations. 3.5.2. Hypothesis 2: Inequality in education Our second hypothesis is that greater inequality in educational attainment is associated with more employers, and fewer selfemployed. The estimation results reported in Table 3 do not support this hypothesis. First, there is no apparent connection between the distribution of educational qualifications and the propensity to be an employer. At the individual level, more education is clearly associated with being an employer – a common finding – but the degree of inequality in education in a district plays no role. In the regressions that include all workers, both urban and rural, there is some evidence that inequality in education appears to be associated with more self-employment (Table 3, columns 4 and 6; Table 4, column 5). However, the effect is not compelling, as it vanishes when state-level fixed effects are included (Table 3, column 5), or when separate models are estimated for the urban and rural groups (Table 4, columns 1–4). 3.5.3. Hypothesis 3: Inequality in consumption This third hypothesis postulates that greater overall economic inequality – as measured by the Gini coefficient for consumption per capita – is associated with more employers and more self-employed. Consumption is considered to be a better proxy than income 9

Journal of Business Venturing xxx (xxxx) xxx–xxx

S. Sarkar et al.

Table 3 Estimation results (marginal effects) of employment models. Dependent variable

Employer

Logit equation type

Binomial

Binomial

Multinomial

Binomial

Binomial

Multinomial

State fixed effects

No

Yes

No

No

Yes

No

(1)

(2)

(3)

(4)

(5)

(6)

0.077⁎⁎⁎ (0.000) 0.012⁎ (0.017) − 0.027 (0.145) 0.012⁎⁎ (0.005)

0.051⁎⁎⁎ (0.000) 0.008 (0.348) 0.028 (0.196) 0.010 (0.079)

0.073⁎⁎⁎ (0.000) 0.012⁎ (0.014) −0.021 (0.235) 0.011⁎⁎ (0.009)

− 0.502⁎⁎⁎ (0.000) − 0.250⁎⁎⁎ (0.000) 0.277⁎⁎⁎ (0.000) 0.008 (0.517)

−0.196⁎⁎⁎ (0.000) −0.118⁎⁎⁎ (0.000) −0.061 (0.344) 0.106⁎⁎⁎ (0.000)

−0.410⁎⁎⁎ (0.000) −0.212⁎⁎⁎ (0.000) 0.355⁎⁎⁎ (0.000) 0.007 (0.534)

0.008⁎⁎⁎ (0.000) 0.016⁎⁎⁎ (0.000)

0.6008⁎⁎⁎ (0.000) 0.016⁎⁎⁎ (0.000)

0.007⁎⁎⁎ (0.000) 0.016⁎⁎⁎ (0.000)

0.098⁎⁎⁎ (0.000) 0.260⁎⁎⁎ (0.000)

0.097⁎⁎⁎ (0.000) 0.263⁎⁎⁎ (0.000)

0.085⁎⁎⁎ (0.000) 0.268⁎⁎⁎ (0.000)

0.006⁎⁎⁎ (0.000) 0.010⁎⁎⁎ (0.000) 0.015⁎⁎⁎ (0.000) 0.018⁎⁎⁎ (0.000)

0.005⁎⁎ (0.002) 0.009⁎⁎⁎ (0.000) 0.014⁎⁎⁎ (0.000) 0.018⁎⁎⁎ (0.000)

0.004⁎⁎ (0.005) 0.008⁎⁎⁎ (0.000) 0.012⁎⁎⁎ (0.000) 0.016⁎⁎⁎ (0.000)

0.039⁎⁎⁎ (0.000) 0.048⁎⁎⁎ (0.000) 0.011⁎ (0.012) − 0.117⁎⁎⁎ (0.000)

0.040⁎⁎⁎ (0.000) 0.052⁎⁎⁎ (0.000) 0.013⁎⁎ (0.002) −0.125⁎⁎⁎ (0.000)

0.015⁎⁎⁎ (0.000) 0.019⁎⁎⁎ (0.000) 0.014⁎⁎⁎ (0.000) 0.034⁎⁎⁎ (0.000)

0.0004 (0.735) 0.008⁎⁎⁎ (0.000) 0.009⁎⁎⁎ (0.000) 0.016⁎⁎⁎ (0.000)

− 0.001 (0.658) 0.005⁎⁎ (0.002) 0.009⁎⁎⁎ (0.000) 0.017⁎⁎⁎ (0.000)

0.001 (0.525) 0.008⁎⁎⁎ (0.000) 0.009⁎⁎⁎ (0.000) 0.014⁎⁎⁎ (0.000)

− 0.039⁎⁎⁎ (0.000) 0.032⁎⁎⁎ (0.000) 0.057⁎⁎⁎ (0.000) 0.091⁎⁎⁎ (0.000)

−0.033⁎⁎⁎ (0.000) 0.051⁎⁎⁎ (0.000) 0.052⁎⁎⁎ (0.000) 0.084⁎⁎⁎ (0.000)

−0.030⁎⁎⁎ (0.000) 0.032⁎⁎⁎ (0.000) 0.068⁎⁎⁎ (0.000) 0.069⁎⁎⁎ (0.000)

0.0000 (0.743)

0.0003 (0.116)

0.0001 (0.739)

0.007⁎⁎⁎ (0.000)

0.004⁎⁎⁎ (0.000)

0.005⁎⁎⁎ (0.000)

0.003⁎⁎ (0.001)

0.004⁎⁎⁎ (0.000)

0.003⁎⁎ (0.004)

− 0.016⁎⁎⁎ (0.000)

−0.018⁎⁎⁎ (0.000)

−0.011⁎⁎ (0.006)

0.001⁎⁎⁎ (0.000) − 0.0006⁎⁎ (0.004)

0.001⁎⁎⁎ (0.000) − 0.001⁎⁎ (0.003)

0.001⁎⁎⁎ (0.000) −0.0005⁎ (0.029)

0.009⁎⁎⁎ (0.000) − 0.005⁎⁎⁎ (0.000)

0.009⁎⁎⁎ (0.000) −0.005⁎⁎⁎ (0.000)

0.007⁎⁎⁎ (0.000) −0.002⁎⁎ (0.001)

0.006⁎⁎⁎ (0.000) 0.007⁎⁎⁎ (0.000) 0.005 (0.277)

0.007⁎⁎⁎ (0.000) 0.009⁎⁎⁎ (0.000) 0.005 (0.243)

0.006⁎⁎⁎ (0.000) 0.006⁎⁎ (0.01) 0.004 (0.446)

0.096⁎⁎⁎ (0.000) 0.085⁎⁎⁎ (0.000) 0.036⁎ (0.029)

0.095⁎⁎⁎ (0.000) 0.078⁎⁎⁎ (0.000) 0.038⁎ (0.017)

0.093⁎⁎⁎ (0.000) 0.070⁎⁎⁎ (0.000) 0.026⁎ (0.089)

0.002 (0.184)

0.003⁎ (0.035)

0.002 (0.229)

0.021⁎⁎⁎ (0.000)

0.026⁎⁎⁎ 0.013⁎⁎ (0.000) (0.006) (continued on next page)

Inequality/heterogeneity measures Gini: Consumption per capita Gini: Land per household HHI: Education HHI: Social group Resource measures Land owned No land (reference) Under 2 ha. 2 ha. or more Educational level achieved Less than primary (reference) Primary Middle school Completed secondary Higher education Personal characteristics Social Group Scheduled tribe (reference) Scheduled class Other “backward” class “Forward” class Gender (Male) Family size Household size Location Urban Other control variables Age Age squared (/100) Marital status Never married (reference) Married Widowed Divorced/separated Religion Hindu (reference) Muslim

Self-employed

10

Journal of Business Venturing xxx (xxxx) xxx–xxx

S. Sarkar et al.

Table 3 (continued) Dependent variable

Employer

Logit equation type

Binomial

Binomial

Multinomial

Binomial

Binomial

Multinomial

State fixed effects

No

Yes

No

No

Yes

No

(1)

(2)

(3)

(4)

(5)

(6)

0.006⁎⁎⁎ (0.001) 0.012⁎⁎⁎ (0.000) 0.013⁎⁎ (0.001) 0.009 (0.092)

0.001 (0.420) 0.005 (0.077) 0.015⁎⁎⁎ (0.001) 0.008 (0.125)

0.006⁎⁎ (0.006) 0.012⁎⁎⁎ (0.000) 0.015⁎ (0.012) 0.008 (0.228)

0.010 (0.239) 0.004 (0.787) 0.070⁎⁎ (0.008) 0.034⁎ (0.025)

−0.018⁎ (0.044) 0.004 (0.811) 0.079⁎⁎ (0.002) 0.013 (0.401)

0.014⁎ (0.059) 0.033⁎⁎ (0.003) 0.112⁎⁎ (0.001) 0.019 (0.156)

0.026⁎⁎⁎ (0.000) 0.027⁎⁎⁎ (0.000) 0.025⁎⁎⁎ (0.000) 0.018⁎⁎⁎ (0.000) − 15.227⁎⁎⁎ (0.000)

0.026⁎⁎⁎ (0.000) 0.027⁎⁎⁎ (0.000) 0.025⁎⁎⁎ (0.000) 0.017⁎⁎⁎ (0.000) − 16.908⁎⁎⁎ (0.000)

0.019⁎⁎⁎ (0.000) 0.022⁎⁎⁎ (0.000) 0.026⁎⁎⁎ (0.000) 0.013⁎⁎⁎ (0.000)

− 0.2043⁎⁎⁎ (0.000) 0.174⁎⁎⁎ (0.000) − 0.414⁎⁎⁎ (0.000) − 0.149⁎⁎⁎ (0.000) − 1.659⁎⁎⁎ (0.000)

−0.041⁎⁎⁎ (0.000) 0.172⁎⁎⁎ (0.000) −0.416⁎⁎⁎ (0.000) −0.144⁎⁎⁎ (0.000) −2.514⁎⁎⁎ (0.000)

−0.155⁎⁎⁎ (0.000) 0.070⁎⁎⁎ (0.000) −0.364⁎⁎⁎ (0.000) −0.199⁎⁎⁎ (0.000)

0.010 127,044 0.156 14,451 14,754

0.010 126,124 0.193 13,881 14,504

0.010 133,223 0.394

0.396 127,044 0.215 137,358 137,661

0.396 127,044 0.234 134,222 134,856

0.396 133,223 0.394

Christian Sikh Jain Buddhist Main industry where employed Agriculture, forestry, fishing, mining (reference) Manufacturing Trade Construction Services Constant Memo items Mean, dependent variable Number of observations pseudo R-sq AIC BIC

Self-employed

Notes: Numbers show average marginal effects of independent variables on employment status. p-values in parentheses. Estimated by the authors using data from the 68th round of the NSS. “Employer” is someone other than a farmer who hires employees; the “self-employed” are self-employed individuals who do not hire workers. Binomial logit equations exclude public administration employees. ⁎ p < 0.1. ⁎⁎ p < 0.01. ⁎⁎⁎ p < 0.001. Table 4 Estimated marginal effects of inequality/heterogeneity variables under different specifications. Dependent variable

Self-employed

Sample

Rural only

Urban only

Rural only

Urban only

(1)

(2)

(3)

(4)

Gini: consumption/cap Gini: land per household H Index: education H Index: social group

⁎⁎⁎

− 0.165 (0.000) − 0.177⁎⁎⁎ (0.000) 0.004⁎ (0.963) 0.101⁎⁎⁎ (0.000)

Employer

⁎⁎⁎

− 0.248 (0.000) − 0.047 (0.104) − 0.095 (0.368) 0.101⁎⁎⁎ (0.000)

⁎⁎⁎

0.024 (0.001) 0.011⁎⁎ (0.006) 0.009 (0.654) − 0.001 (0.762)

⁎⁎⁎

0.084 (0.001) 0.004 (0.806) 0.033 (0.481) 0.028⁎ (0.047)

H Index: Religion Sample size, model No. entrepr. in sample

81,190 39,175

53,475 19,340

81,190 585

53,475 1009

Self-employed

Employer

(5)

(6) ⁎⁎⁎

− 0.483 (0.000) − 0.278⁎⁎⁎ (0.000) 0.209⁎⁎⁎ (0.000)

0.125⁎⁎⁎ (0.000) 0.038⁎⁎⁎ (0.000) − 0.054⁎⁎ (0.008)

0.041⁎⁎⁎ (0.000) 128,327 58,515

0.007⁎ (0.013) 128,327 1596

Notes: Numbers show average marginal effects of independent variables on employment status. p-values in parentheses. Variables included are as in Table 3, with state fixed effects in columns (1)–(4), but not in columns (5) and (6). Only the effects of the inequality/heterogeneity variables are reported here. “H Index” is HerfindahlHirschman Index. Estimated by the authors using data from the 68th round of the NSS. ⁎ p < 0.1. ⁎⁎ p < 0.01. ⁎⁎⁎ p < 0.001.

11

Journal of Business Venturing xxx (xxxx) xxx–xxx

S. Sarkar et al.

for lifetime well-being (Haughton and Khandker, 2009), but it should also be noted that the survey did not collect comprehensive data on incomes. The evidence here is mixed: there is clear support for the idea that greater overall inequality is associated with there being more employers (Table 3, columns 1–3; Table 4, columns 3, 4, and 6); on the other hand there is robust evidence that greater overall inequality is associated with fewer self-employed (Table 3, columns 4–6; Table 4, columns 1, 2, and 5). Two explanations are possible for the robust association between being an employer and inequality in consumption and land ownership. One possibility is that a greater concentration of resources, reflected in greater inequality, gives a selected number of individual's sufficient resources to develop relatively complex businesses, which employ other individuals. This explanation would be consistent with the reasonable idea that becoming an employer requires a minimum level of resources to be viable, and confirms prior research showing that financial capital becomes a stronger predictor of entrepreneurial entry as inequality rises (Xavier-Oliveira et al., 2015). Another possibility is that since our data are cross-sectional, the direction of causality is the opposite: successful entrepreneurship concentrates incomes in the hands of a few successful individuals and therefore leads to greater inequality. This is one of the explanations for the rise in income and wealth inequality in the US proposed by some economists, whereby new technologies have disproportionately enriched their creators or owners, resulting in a rise in inequality associated with the successful ventures pursued by these persons, such as Google, Facebook, Microsoft, EBay, and so forth (Gambardella and Ulph, 2002). However, our data allow us to rule out this explanation: the Gini coefficient for consumption per capita is 0.360 for the full survey sample, and 0.359 when employers are excluded. Although employers are more affluent than the average person in India, the group is too small to have anything but a negligible effect on our measures of inequality. Our result about the association of greater inequality with less self-employment contrasts with our reasoning that high inequality would imply a large pool of relatively poor people who would be driven to self-employment as a subsistence tactic (George et al., 2015). A reverse direction of causality is implausible here: the notion that self-employment might equalize consumption levels or land distribution holdings would require that the returns to self-employment somehow favor poorer individuals relative to richer ones in a systematic manner. A more plausible explanation for our findings is thus that the very poor are more likely to find themselves working as casual wage laborers, without even the resources to succeed in self-employment. Table 5 shows the levels of consumption per capita by major employment category: on average, the self-employed enjoy levels of per capita consumption that are more than a third higher than those relying on casual wage income. This could indicate that there is indeed a minimum threshold of financial resources for self-employment, although we cannot ignore the opposite direction of causality, whereby the higher consumption of the self-employed reflects higher rewards for self-employment than for casual wage labor. To shed more light on the effects of inequality on self-employment among the very poor, we consider below an alternative specification where we examine self-employment in this subset of the population alone. In any case, our results show that self-employment appears to be more widespread in communities with a more egalitarian distribution of land and consumption. Rooks et al. (2014) argue that rural regions at least have a relatively collectivistic, value-based resource-sharing ethos, which may foster the sharing of information and risk, and encourage individuals to pursue their own independent activity rather than wage labor. However, the greater limitations in terms of market size, and isolation of rural communities relative to urban ones, may make it difficult for these self-employed individuals to evolve into employers who offer wage employment to other individuals, which would also be consistent with our notion of multiple entrepreneurial thresholds. 3.5.4. Hypothesis 4: Heterogeneity in social groups Caste and religious differences remain important in India. At the individual level, Christians, Sikhs and Jains are more likely than others to be employers, while Muslims and Jains are more likely than other groups to be self-employed. Similarly, those who come from a scheduled tribe or a “scheduled class” (formerly often referred to as “untouchable”) are less likely to be self-employed, which is consistent with the interpretation that this group, as suffering the most from social exclusion, is unable to reach even the threshold of self-employment. These findings are consistent with, but more detailed, than those obtained by Audretsch et al. (2013). Our fourth hypothesis postulates that areas with greater social group heterogeneity – i.e. a lower HHI index – will, other things being equal, have more people working as employers, and self-employed. The evidence generally does not support this for measures based on caste, either for the full sample (Table 3, columns 1, 3, and 5), or separately for urban and rural areas (Table 4, columns 1, 2, and 4); similar, if weaker, effects are found when the measure if based on religion (Table 4, columns 5–6). Our interpretation is that the greater difficulty in forming social networks – which can help entrepreneurs by facilitating transactions, knowledge exchange, or Table 5 Consumption per capita by major employment category, India, 2011–12.

Employer Household employer Self-employed Works for a regular wage Works for casual wages All groups

Consumption Rp/person/month

% of those working

3449 2263 1515 2485 1103 1613

1.2 0.9 43.4 29.6 24.9 100.0

Notes: Total sample covers 134,655 employed individuals. Estimated by the authors using data from the 68th round of the NSS.

12

Journal of Business Venturing xxx (xxxx) xxx–xxx

S. Sarkar et al.

resource and risk sharing, for instance – under conditions of greater heterogeneity has a stronger discouraging effect on entrepreneurship than the greater opportunities for specialization that drove our hypothesis. 3.5.5. Other independent and control variables A few words are in order about the other independent variables and our controls, after which we consider the robustness of the results, and then return to a deeper interpretation of our findings. We find a strong direct association between access to most resources and entrepreneurship (as measured by whether one is an employer, or self-employed). Greater land ownership is significantly associated with a higher propensity for entrepreneurship. The same is true of education, until one gets to the tertiary level, at which point higher education is associated with a lower probability of being self-employed (but still a higher probability of being an employer), consistent with the higher opportunity cost of self-employment for the highly educated arising from better opportunities to become employers or for salaried employment. Household size is positively associated with higher self-employment; we interpret this result as stemming from the pooling of resources among kin. We also note that gender effects are strong: men are more likely than women to be employers, or self-employed, possibly reflecting significant cultural barriers to independent activity by women in India. The location of the two types of entrepreneurs is clearly differentiated, with self-employment being a distinctly rural phenomenon, while working as an employer is more clearly an urban one. This latter may reflect its association with the many opportunities afforded by the more modern economy of cities, where a wide variety of economic activities and resources is available to entrepreneurial individuals. To the extent possible, we take this factor into account by controlling for the sectors of activity of entrepreneurs and their ventures (see below), but the urban effect is significant even with this control. Therefore, the urban effect may also be the result of the positive externalities, extensively documented by urban economists (Glaeser, 2000), which arise in highpopulation-density settings from the exchange of information, and enhanced opportunities for finding like-minded individuals with whom to work. This may also reflect the importance of economies of scale within firms in urban areas, which favor larger businesses relative to self-employment. Age has a U-shaped effect on entrepreneurship; as we might expect, neither the young nor the old are as likely to have the resources and capacities to pursue entrepreneurial ventures. Married and widowed persons (who can access their spouse's resources) are more associated with entrepreneurship (as defined by being an employer, or self-employed) in general, relative to those who are not married. This underlines the importance of family for entrepreneurship. Being an employer is clearly associated with nonagricultural economic activities, where a much greater diversity of possibilities, but also greater barriers to entry in the form of capital, technology, and skills, create more favorable conditions for this kind of entrepreneurship. It is also more likely in manufacturing, trade, construction, and services, relative to agricultural endeavors. Once again, this contrasts with self-employment, which is positively associated with trade only, and negatively with the other sectors, relative to agriculture. Such a difference is consistent with the entry requirements for nonagricultural sectors other than trade. Trade can encompass, for instance, sidewalk or itinerant sales of simple items like food or second-hand clothing, which require little working capital and few out-of-pocket expenses, and are commonly undertaken as forms of self-employment. 3.5.6. Robustness: technical specifications Our empirical results are robust to a variety of specifications of the estimating equation. As already shown, our results in Table 3 consider both binomial and multinomial specifications, and also with and without fixed effects. Additional robustness tests are set out in Table 6 for the marginal effects of inequality on the propensity to be an employer. Column 1 reproduces the results from the baseline model (from Column 1 in Table 3). This may be compared with the effects measured using a probit specification (Column 2), the baseline model that also includes squared values of the inequality variables (Column 3), and a version that measures inequality at the level of the region rather than district (Column 4). In all of the specifications the signs remain the same, and the orders of magnitude are preserved. The strong effect of inequality in consumption per capita on the propensity to be an employer remains intact, as do the rest of our results. 3.5.7. Robustness: alternative definitions and controls Our operational measure of an employer is someone who is self-employed, hires workers, and is not a farmer; and we define the self-employed as the self-employed who do not hire workers. Nonetheless, while many farmers are self-employed, they are rarely founders of business ventures, which could skew our results for rural areas. Therefore, we run an alternative specification excluding the agricultural sector entirely, as done in Acs et al. (1994). This reduces our sample by about 35,000 workers, but still leaves us with a set of over 1500 employers and 35,000 self-employed. Our results for this alternative are shown in column 5 of Table 6 (for employers) and column 2 of Table 7 (for the self-employed). When agriculturalists are excluded, the results for consumption inequality and social heterogeneity are essentially the same. For inequality in education, there is still no effect on whether one is an employer, and a positive impact on self-employment. In contrast, the inequality of land holdings now plays no role – which is unsurprising (perhaps even reassuring), since we are now looking just at the non-agricultural sector. The entrepreneurship literature (e.g. Shane, 2003) suggests that the availability of financial resources is also important for individuals to pursue entrepreneurial ventures. We do not have detailed information about household assets or income – and certainly not at the time that the entrepreneurial activity might have begun – but we do have information on consumption per capita. We test the robustness of our results against the inclusion of this variable in two ways. First, we include this as a control variable in our estimation of the propensity to be an employer (Table 6, column 6). Higher levels of consumption per capita are associated with 13

Journal of Business Venturing xxx (xxxx) xxx–xxx

S. Sarkar et al.

Table 6 Estimated marginal effects of inequality/heterogeneity variables for individual employers under different econometric specifications. Dependent variable

Employer

Equation type

Baseline logit, full sample

Probit, full sample

Logit with squaresa

Logit, inequality by regionb

Logit, excl. agriculture

Logit, full sample

(1)

(2)

(3)

(4)

(5)

(6)

0.077⁎⁎⁎ (0.000) 0.012⁎ (0.017) −0.027 (0.145) 0.012⁎⁎ (0.005)

0.078⁎⁎⁎ (0.000) 0.012⁎ (0.023) − 0.018 (0.308) 0.012⁎⁎ (0.005)

0.188⁎⁎⁎ (0.000) 0.004 (0.924) − 0.122 (0.254) 0.030 (0.253)

0.104⁎⁎⁎ (0.000) 0.015⁎⁎ (0.007) − 0.016 (0.483) 0.018⁎⁎⁎ (0.000)

0.066⁎⁎⁎ (0.000) 0.013 (0.209) 0.044 (0.131) 0.014⁎ (0.053)

127,044 1594

127,044 1594

127,044 1594

127,044 1594

90,837 1554

0.042⁎⁎⁎ (0.001) 0.008 (0.362) 0.029 (0.177) 0.010⁎ (0.089) 0.001⁎⁎⁎ (0.000) 126,121 1594

Gini: consumption per capita Gini: land per household Herfindahl index: education Herfindahl index: social group Consumption per capita (′000) Sample size, model No. entrepreneurs in sample

Notes: Numbers show average marginal effects of independent variables on employment status. p-values in parentheses. Variables included are as in Table III, excluding state fixed effects. Only the effects of the inequality/heterogeneity variables are reported here. Estimated by the authors using data from the 68th round of the NSS. Column 1 is same as column 1 in Table 3. ⁎ p < 0.1. ⁎⁎ p < 0.01. ⁎⁎⁎ p < 0.001. a Also includes inequality variables squared. b Inequality measures are constructed for regions, not districts. Table 7 Estimated marginal effects of inequality/heterogeneity variables under different self-employment specifications. Full sample

All

Poorest 20%

Poorest 40%

Poorest 40%

Excluding agriculture

Excluding agriculture

Excluding agriculture

Non-agric., poorest 40%

(1)

(2)

(3)

(4)

(5)

−0.196⁎⁎⁎ (0.000) −0.118⁎⁎⁎ (0.000) −0.061 (0.344) 0.106⁎⁎⁎ (0.000) 127,044 58,515

− 0.300⁎⁎⁎ (0.000) − 0.002 (0.904) 0.144⁎⁎ (0.049) 0.103⁎⁎⁎ (0.000) 91,503 35,174

− 0.002 (0.891) − 0.009 (0.319) 0.090⁎⁎⁎ (0.000) 0.009 (0.327) 89,620 3572

−0.093⁎⁎⁎ (0.000) −0.017 (0.206) 0.256⁎⁎⁎ (0.000) 0.022⁎ (0.097) 90,943 9221

− 0.133⁎⁎ (0.034) 0.028 (0.436) 0.427⁎⁎⁎ (0.000) 0.160⁎⁎⁎ (0.000) 23,972 9221

Dependent variable

Gini: consumption per capita Gini: land per household HHI: education HHI: social group Sample size, model No. of entrepreneurs in sample

Notes: Numbers show average marginal effects of independent variables on employment status. p-values in parentheses. Variables included are as in Table III, including state fixed effects. Only the effects of the inequality variables are reported here. Estimated by the authors using data from the 68th round of the NSS. Column 1 is same as column 5 in Table 3. ⁎ p < 0.1. ⁎⁎ p < 0.01. ⁎⁎⁎ p < 0.001.

someone being an employer, although we cannot determine the direction of causality here. What is important is that the effects of inequality on employment status are hardly changed by the inclusion of the consumption variable. Second, for the self-employed we consider only those whose per capita consumption is in the lowest one, or two, quintiles of the full sample. This is in order to further explore the negative impacts of inequality on self-employment by the poorest population that we obtained in our larger sample (see above). Technically, we can do this because although we use consumption per capita to measure inequality, we do not use it in any other way in our estimates. The estimates, using this definition of our dependent variable, are shown in the last three columns of Table 7. Since the agricultural sector is highly dominated by large landholdings and the associated extensive casual wage labor, we confine the sample to the non-agricultural sector, while considering three different threshold levels of poverty: the poorest quintile (column 3), the poorest two quintiles (column 4), and a sample that only compares the self-employed in the poorest two quintiles with the salaried workers in the poorest two quintiles, in contrast to the previous specifications, which use all non-agricultural workers as a reference group 14

Journal of Business Venturing xxx (xxxx) xxx–xxx

S. Sarkar et al.

Table 8 Summary of estimation effects. Independent variable:

Greater land inequality (H1)

Greater educational inequality (H2) Greater economic inequality (H3)

Greater social heterogeneity (H4) More resources: - Land - Education

No. of employers

No. of selfemployed

Interpretation

Urban

No effect

No effect

Rural

Increase

Decrease

U+R

No effect

No effecta

U+R

Increase

Decrease

U+R

Decrease

Decrease

Land ownership irrelevant for urban entrepreneurship, confirming hypothesis for urban areas. Large landowners employ laborers, and such wage labor, not self-employment, absorbs poorest, contrary to the hypothesis for both self-employment and employers in rural areas. Educational heterogeneity does not imply a concentration of information and knowledge in a subgroup of individuals; hypothesis unsupported. Concentration of financial resources allows marshalling of resources to cross threshold to become employer but reduces resources available to others to become self-employed, confirming the hypothesis for employers but refuting it for the self-employed. Social fragmentation may limit the formation of networks of value for entrepreneurship, contrary to hypothesis.

Urban Rural U+R

No effect Increase Increase

No effect Increase Increase

Land ownership irrelevant for urban entrepreneurship Land still important in rural areas Education still important at individual level (but exception: more higher education cuts self-employ.)

U+R U+R U+R

Increase Increase No effect Increase Incr., then decr.

Increase Increase Increase Decrease Incr., then decr. Increase Muslim, Jain Agr, trade

Higher social status confers advantages for entrepreneurship Women face severe barriers to entrepreneurship Pooling of resources for self-employment Stronger economies of scale, networks, in towns Inverted-U effect

Controls: - Social group - Male - Household size - Urban - Age

U+R

- Ever married - Religion (relative to Hindu) - Industry

U+R U+R U+R

Increase Christian, Sikh, Jain Nonagr, trade, mfg., cconst, svc

Pooling of resources for entrepreneurship Cultural effects on motivation; intra-group solidarity Greater opportunities for larger-scale businesses in nonagricultural activities, including manufacturing, construction, and services

Italics indicate potential threshold effects, whereby threshold levels of certain resources appear to be necessary to become an employer or to be self-employed. Source: Based on estimates in Tables 3–7. a Increase for non-agricultural.

(column 5). These may be compared with the “baseline” estimates that are applied to the entire non-agricultural sector, shown in column 2 of Table 7. When the focus is on the self-employed in the poorest two quintiles, the results are similar in sign, significance, and magnitude to those found by using our original definition of self-employment, with the exception of educational heterogeneity, which seems to have a consistently positive impact for non-agricultural self-employment. Only when we confine our set of the self-employed to the poorest quintile do our original results begin to evaporate, suggesting that for this set of individuals, other factors limit the ability to attain self-employment. But this case comprises just 4% of the self-employed, so it has questionable relevance. The fact that inequality and heterogeneity have a negative relation with self-employment even for the poorer non-agricultural population, reinforces the notion that even within the same social stratum, small differences in access to financial and other resources can significantly limit entrepreneurial opportunity. This is a valuable insight into the impact of inequality on entrepreneurship, because it is consistent with the notion that entrepreneurial initiative is not dependent on a single factor, such as financial resources, but can result from the variable combination of a variety of resources. As such, focusing only on economic inequality may fail to reveal how other forms of inequality, such as differences across social groups, can subtly alter entrepreneurial opportunity even among individuals in the same income level. 4. Discussion and conclusion The empirical results are summarized schematically in Table 8, which makes clear the relatively complex nature of our findings. At the broadest level, greater inequality appears to favor the formation of employer-owned firms, while for the most part reducing the extent of self-employment, after controlling for other relevant influences on entrepreneurship. There appear to be two thresholds, one representing a move into self-employment, and the other reflecting the jump to employer status. Crossing these thresholds—being self-employed or an employer—involves assembling combinations of resources, such as education or money, and personal attributes, such as social status, religious affiliation, gender, or family size, that can sustain entrepreneurial action. The combinations available to some individuals may simply be insufficient to cross one or both thresholds. Specifically, there is a large set of people – whom we can call the very poorest – that have difficulty crossing the first one: they lack the land (in rural areas), education, networks, and financial resources required to be sustainably self-employed, and fall back on casual wage laboring. This suggests that self-employment is not “necessity entrepreneurship” but rather a rung further up the ladder, 15

Journal of Business Venturing xxx (xxxx) xxx–xxx

S. Sarkar et al.

more aspirational than an activity of last resort. Moreover, inequality worsens this situation. Greater inequality leaves more people in poverty, where self-employment is out of reach; and greater social heterogeneity (as measured by either caste membership or religious affiliation) makes it even harder to develop robust social networks that can help cross the threshold to self-employment. Another large set of individuals are better placed to marshal combinations of resources that enable them to sustain themselves in self-employment, particularly if aided by family members, a higher social status, or other personal attributes. But the second threshold, from self-employed to employer, remains a formidable barrier for them. The qualitative leap that is needed requires considerable personal and household resources, and in an emerging economy such as India, this is easier to achieve if there is considerable inequality, in effect concentrating enough resources and capacity in a smaller set of individuals to permit the emergence of a class of employers. While greater inequality can foster the rise of employers through such a concentration of resources, unfortunately it seems to reduce the resources left for others to even reach the self-employment threshold, as mentioned above. 4.1. Contributions This discussion underlines four of the contributions that we make in this paper. First, at a theoretical level, we show that inequality affects entrepreneurship by altering the proportions of those who are able to cross the first threshold into self-employment, and the second and much higher threshold into becoming an employer. Second, we construct a variety of empirical measures of inequality, which allows us to test the proposition that these affect entrepreneurship, after controlling for other standard variables, in subtler and more complex ways than simply inequality in income levels. Third, we estimate the relationships between inequality and the propensity to entrepreneurship, using data from India. The results are striking: greater inequality in the distribution of resources is associated with a higher propensity to be an employer, and a lower likelihood of being self-employed. This leads us to conclude, as our fourth contribution, that it is essential to retain the distinction between employers and the self-employed – they cannot simply be lumped into a single class labelled “entrepreneurs – and it also leads us to question the usefulness of the concept of ‘necessity entrepreneur’”, since it would appear that becoming an entrepreneur requires an act of volition, and is not generally the option of last resort. A fifth contribution is that our work has important implications for practice. We outline these implications below. 4.2. Limitations We are well aware of the limitations of our analysis. First and foremost, the use of cross-sectional data means that we can only draw conclusions about associations, not about causality. By controlling for a variety of contextual factors, such as religion, our work is certainly an improvement over simple correlations, but our results must for now remain suggestive, and will require further research to be corroborated. Second, our data do not measure entrepreneurship directly, but requires us instead to use self-employment, and non-agricultural employment of several persons, as proxies for entrepreneurship. Although the use of self-employment and of business establishments as proxies for entrepreneurship has been common in prior research (e.g. Audretsch et al., 2013), it does not detract from the desirability of using a more direct measure of each form of entrepreneurship. Third, we are missing some variables that we would like to be able to include, but are not available in our database. These include measures of psychological characteristics (such as motivation); and indexes that capture the nature of local economic conditions, regulations, and institutions, although the fixed effects control for these factors to some extent at the district and state levels. Fourth and last, our data are confined to India, in large part because of the availability of data from a sample survey sufficiently large to allow us to estimate the effects of a relatively small group (non-agricultural employers). Although India accounts for a significant proportion of the world's population, and is a country of enormous diversity in economic, social, and cultural conditions, it has its own distinctive traits, so replication of our research in other countries or even better, across countries, would be highly desirable in order to give greater external validity to our findings. 4.3. Implications There are a number of implications for policy, and for the future pattern of development in India, that emerge from our empirical work. The most obvious is that the “classical” variables, including an individual's age, education, physical assets, and marital status, have a clear impact on the probability that they will be an employer or self-employed. Not all of these variables are amenable to policy, but the continuing expansion of education will, for the foreseeable future in India, help increase the number of employers. The second observation is that social norms in India appear to be holding back entrepreneurship. If caste were irrelevant, so the HHI index for social groups were to rise to 1, our results show that there would be an increase in entrepreneurship. And the gender bias, whereby just 5% of employers are women, is virtually excluding half of the population from the potential to become an employer, which is surely a massive underutilization of talent. By implication, measures that weaken caste and gender barriers would favor entrepreneurship. Although most households own some land, the distribution of land is highly unequal. While a land reform that would create greater equality in land holdings would have no impact on urban entrepreneurship, it would increase the number of self-employed in rural India. Finally, we note that even if one could substantially increase the number of entrepreneurs, it remains unlikely that many of them would come from the poorest fifth of the population. Efforts to improve the position of the poorest will require a wider menu of interventions, although greater entrepreneurial activity would certainly help this group indirectly insofar as it would boost economic 16

Journal of Business Venturing xxx (xxxx) xxx–xxx

S. Sarkar et al.

growth and create more jobs. To conclude, for entrepreneurship to realize its potential to help address the challenge of poverty, we need to better understand how inequality shapes entrepreneurial efforts. Drawing on a large database from India—a country that represents one-sixth of the world's population and one-third of the world's extreme poor—we find that inequality in resources and social heterogeneity are associated with more people being employers and, for the most part, fewer people being self-employed. We hope that our results will encourage scholars to delve deeper into the impact of inequality on entrepreneurship, and through a deeper understanding of this relationship, make entrepreneurship a stronger instrument for the reduction of poverty around the world. References Acs, Z., 2006. How is entrepreneurship good for economic growth? Innovations 1 (1), 97–107. Acs, Z., Audretsch, D., Evans, D., 1994. Why does the self-employment rate vary across countries and over time? Discussion paper no. 871. Cent. Econ. Pol. Res. Publ. Alvarez, S.A., Barney, J.B., 2004. Organizing rent generation and appropriation: toward a theory of the entrepreneurial firm. J. Bus. Ventur. 19 (5), 621–635. Alvarez, S.A., Barney, J.B., 2014. Entrepreneurial opportunities and poverty alleviation. Enterp. Theory Pract. 159–184 (January). Alvarez, S.A., Busenitz, L.W., 2001. The entrepreneurship of resource-based theory. J. Manag. 27, 755–775. Ardagna, S., Lusardi, A., 2010. Explaining international differences in entrepreneurship: the role of individual characteristics and regulatory constraints. In: International Differences in Entrepreneurship, ed. Josh Lerner and Antoinette Schoar. University of Chicago Press, Chicago (for NBER). Audretsch, D., Bönte, W., Tamvada, J.P., 2013. Religion, social class, and entrepreneurial choice. J. Bus. Ventur. 28, 774–789. Baker, T., Nelson, R.E., 2005. Creating something from nothing: resource construction through entrepreneurial bricolage. Adm. Sci. Q. 50 (3), 329–366. Banerjee, A., Duflo, E., 2007. The economic lives of the poor. J. Econ. Perspect. 21, 141–167. Bapuji, H., Neville, L., 2015. Income inequality ignored? An agenda for business and strategic organization. Strateg. Organ. 13 (3), 233–246. Bhagavatula, S., Elfring, T., van Tilburg, A., Van De Bunt, G.G., 2010. How social and human capital influence opportunity recognition and resource mobilization in India's handloom industry. J. Bus. Ventur. 25 (3), 245–260. Birley, S., 1986. The role of networks in the entrepreneurial process. J. Bus. Ventur. 1 (1), 107–117. Bradley, S.W., McMullen, J.S., Artz, K., Simiyu, E.M., 2012. Capital is not enough: innovation in developing economies. J. Manag. Stud. 49 (4), 684–717. Carsrud, A., Brannback, M., 2011. Entrepreneurial motivations: what do we still need to know? J. Small Bus. Manag. 49 (1), 9–26. CBS, SMG, Foss, N.J., 2011. Entrepreneurship in the Context of the Resource-Based View of the Firm. In: Mole, Kevin, Ram, Monder (Eds.), Perspectives in Entrepreneurship. Palgrave, London. Clark, K., Drinkwater, S., 2000. Pushed out or pulled in? Self-employment among ethnic minorities in England and Wales. Labour Econ. 7 (5), 603–628. Clark, K., Drinkwater, S., 2010. Patterns of ethnic self-employment in time and space: evidence from British census microdata. Small Bus. Econ. 34 (3), 323–338. Collins, D., Morduch, J., Rutherford, S., Ruthven, O., 2010. Portfolios of the Poor: How the world's Poor Live on $2 a Day. Princeton University Press, Princeton. Curry, C., 2016. Amid the Dark, Narrow Alleys of India's Largest Slum, a Micro-Economy Bustles. Global Citizen (November 15). https://www.globalcitizen.org/en/ content/mumbais-biggest-slum-dharavi-bustling-micro-econom/, Accessed date: 23 November 2017. de Mel, S., McKenzie, D., Woodruff, C., 2010. Who are the microenterprise owners? Evidence from Sri Lanka on Tokman v. de Soto. In: Lerner, Josh, Schoar, Antoinette (Eds.), International Differences in Entrepreneurship. University of Chicago Press, Chicago (for NBER). Evans, D.S., Leighton, L.S., 1989. Some empirical aspects of entrepreneurship. Am. Econ. Rev. 79 (3), 519–535. Fiess, N.M., Fugazza, M., Maloney, W.F., 2010. Informal self-employment and macroeconomic fluctuations. J. Dev. Econ. 91 (2), 211–226. Foss, N.J., Klein, P.G., Kor, Y.Y., Mahoney, J.T., 2008. Entrepreneurship, subjectivism, and the resource-based view: toward a new synthesis. Strateg. Entrep. J. 2 (1), 73–94. Fritzon, V., 2016. Religion and Business: The Mourides of Senegal. The Perspective (February 25). http://www.theperspective.se/?p=1742, Accessed date: 8 September 2017. Gambardella, A., Ulph, D., 2002. Technology, entrepreneurship, and inequality. http://ssrn.com/abstract=273644 (or). https://doi.org/10.2139/ssrn.273644 (Available at SSRN. Accessed April 8, 2016). George, G., Kotha, R., Parikh, P., Alnuaimi, T., Bahaj, A.S., 2015. Social structure, reasonable gain, and entrepreneurship in Africa. Strateg. Manag. J. 37 (6), 1118–1131. Glaeser, E., 2000. The future of urban research: nonmarket interactions. In: Brookings-Wharton Papers on Urban Affairs, pp. 101–149. Hart, D.M., Acs, Z., 2011. High-tech immigrant entrepreneurship in the United States. Econ. Dev. Q. 25 (2), 116–129. Haughton, J., Khandker, S., 2009. Handbook on Poverty and Inequality. World Bank, Washington DC. Hirschman, A.O., 1965. Obstacles to development: a classification and a quasi-vanishing act. Econ. Dev. Cult. Chang. 13 (4), 385–393. Ho, Y., Wong, P., 2007. Financing, regulatory costs and entrepreneurial propensity. Small Bus. Econ. 28, 187–204. Honiga, B., 1998. What determines success? Examining the human, financial, and social capital of Jamaican microentrepreneurs. J. Bus. Ventur. 13 (5), 371–394. Hout, M., Rosen, H., 2000. Self-employment, family background, and race. J. Hum. Resour. 35 (4), 670–692. Jackson, P., 2010. How Did Quakers Conquer the British Sweet Shop? BBC News Magazine (Jan 20). http://news.bbc.co.uk/1/hi/magazine/8467833.stm, Accessed date: August 2017. Kim, P., Aldrich, H., 2005. Social capital and entrepreneurship. Found. Trends Enterp. 1 (2), 55–104. Lecuna, A., 2014. High income inequality as a structural factor in entrepreneurial activity. J. Technol. Manag. Innov. 9 (1), 13–26. Lippmann, S., Davis, A., Aldrich, H.E., 2005. Entrepreneurship and inequality. In: Keister, Lisa A. (Ed.), Research in the Sociology of Work. vol. 15. pp. 3–32. Locke, E., 2000. Motivation, cognition, and action: an analysis of studies of task goals and knowledge. Appl. Psychol. 49 (3), 408–429. Lofstrom, M., Bates, T., Parker, S.C., 2014. Why are some people more likely to become small-business owners than others: entrepreneurship entry and industryspecific barriers. J. Bus. Ventur. 29 (20), 232–251. McClelland, D.C., 1967. Achieving Society. Simon and Schuster, New York. McMullen, J.S., Bagby, D., Palich, L.E., 2008. Economic freedom and the motivation to engage in entrepreneurial action. Enterp. Theory Pract. 32 (5), 875–895. Mueller, P., 2006. Entrepreneurship in the region: breeding ground for nascent entrepreneurs? Small Bus. Econ. 27 (1), 41–58. National Sample Survey Office (NSSO), 2012. Note on Sample Design and Estimation Procedure of NSS 68th Round. NSSO, Kolkata. Naudé, W., 2010. Entrepreneurship, developing countries, and development economics: new approaches and insights. Small Bus. Econ. 34, 1–12. Noseleit, F., 2010. The entrepreneurial culture: Guiding principles of the self-employed. In: Freytag, A., Thurik, R. (Eds.), Entrepreneurship and Culture. Springer, Berlin, Heidelberg. Ostry, J.D., Berg, A., Tsangarides, C.G., 2014. Redistribution, Inequality, and Growth. IMF Staff Discussion Note, Washington DC. Parker, S., 2004. The Economics of Entrepreneurship and Self-Employment. Cambridge University Press, Cambridge. Piketty, T., 2013. Capital in the Twenty-First Century. Belknap Press, Cambridge MA. Porter, M., 1980. Competitive Strategy: Techniques for Analyzing Industries and Competitors. Free Press, New York. Portes, A., Bach, R.L., 1985. Latin Journey: Cuban and Mexican Immigrants in the United States. University of California Press, Berkeley. Rooks, G., Klyver, K., Sserwanga, A., 2014. The context of social capital: a comparison of rural and urban entrepreneurs in Uganda. Enterp. Theory Pract. 40 (1), 111–130. Roy, M.A., Wheeler, D., 2006. A survey of micro-enterprise in urban West Africa: drivers shaping the sector. Dev. Pract. 16 (5), 452–464. Sarasvathy, S.D., 2004. Constructing Corridors to Economic Primitives: Entrepreneurial Opportunities as Demand-Side Artifacts. University of MD, College Park.

17

Journal of Business Venturing xxx (xxxx) xxx–xxx

S. Sarkar et al.

Schuetze, H.J., 2000. Taxes, economic conditions and recent trends in male self-employment: a Canada–US comparison. Labour Econ. 7 (5), 507–544. Shane, S., 2003. A General Theory of Entrepreneurship: The Individual-Opportunity Nexus. Edward Elgar, Northampton. Shane, S., Venkateraman, S., 2000. The promise of entrepreneurship as a field of research. Acad. Manag. Rev. 25 (1), 217–226. Shane, S., Locke, E.A., Collins, C.J., 2003. Entrepreneurial motivation. Hum. Resour. Manag. Rev. 13 (2), 257–279. Sridharan, S., Maltz, E., Viswanathan, M., Gupta, S., 2014. Transformative subsistence entrepreneurship: a study in India. J. Macromarket. 34 (4), 486–504. Stenholm, P., Acs, Z.J., Wuebker, R., 2013. Exploring country-level institutional arrangements on the rate and type of entrepreneurial activity. J. Bus. Ventur. 28 (1), 176–193. Uzzi, B., 1997. Social structure and competition in interfirm networks: the paradox of embeddedness. Adm. Sci. Q. 42 (1), 35–67. Van Der Sluis, J., Van Praag, M., Vijverberg, W., 2008. Education and entrepreneurship selection and performance: a review of the empirical literature. J. Econ. Surv. 22 (5), 795–841. Venugopal, S., Viswanathan, M., Jung, K., 2015. Consumption constraints and entrepreneurial intentions in subsistence marketplaces. J. Pub. Policy Market. 34 (2), 235–251. Viswanathan, M., Gajendiran, S., Venkatesan, R., 2008. Enabling Consumer and Entrepreneurial Literacy in Subsistence Marketplaces. Springer, Dordrecht. Viswanathan, M., Rosa, J.A., Ruth, J., 2010. Exchanges in marketing systems: the case of subsistence consumer merchants in Chennai, India. J. Mark. 74 (May), 1–18. Viswanathan, M., Sridharan, S., Ritchie, R., Venugopal, S., Jung, K., 2012. Marketing interactions in subsistence marketplaces: a bottom-up approach to designing public policy. J. Pub. Policy Market. 31 (2), 159–177. Viswanathan, M., Echambadi, R., Venugopal, S., Sridharan, S., 2014. Subsistence entrepreneurship, value creation, and community exchange systems: a social capital explanation. J. Macromarket. 34 (2), 213–226. Wagner, J., 2005. Nascent necessity and opportunity entrepreneurs in Germany: evidence from the regional entrepreneurship monitor (REM). In: University of Luneburg Working Paper Series in Economics. vol. 10. Webb, J.W., Ireland, R.D., Ketchen Jr., D.J., 2014. Toward a greater understanding of entrepreneurship and strategy in the informal economy. Strateg. Entrep. J. 8, 1–15. Xavier-Oliveira, E., Laplume, A.O., Pathak, S., 2015. What motivates entrepreneurial entry under economic inequality? The role of human and financial capital. Hum. Relat. 68 (7), 1183–1207.

18