Social Capital and Entrepreneurial Outcomes: Evidence from Informal Sector Entrepreneurs in Ethiopia

Informal Entrepreneurs’ Social Networks in Ethiopia

In the informal sector, ethnic and family identification is one way of forming SNs (Kristiansen, 2004). In Africa city, case studies show that extended family networks span the rural–urban divide and become important sources of livelihood security and support for entrepreneurs (Beall, 2001). In terms of poverty alleviation, it is known that the urban poor call on close relations with family and friends as a form of social security. The urban poor typically have greater proportion of strong and dense networks and these networks are used to ensure them against urban shocks and stresses. In African countries, the collectivist culture of the society and strong family ties are sources of social protection for the poor (Meagher, 2005). In Ethiopia too, the collective cultural outlook in general and the conditions of informal entrepreneurs in particular favour strong family and friendship ties for continued entrepreneurship (Getahun, 2015). Hofstede’s (1984) study on national cultures shows that Ethiopia has a collectivistic tradition characterised by an intense and close pledge to members of a group and high level of trust within and between groups. Such collectivist cultures focus on group obligation and interpersonal harmony and foster strong relations where everybody takes responsibility for members of their group (Hofstede, 1984). Based on strong connections, individuals are more likely to be treated and perceived as in-group members and more strongly identified within the in-group. In such types of cultures, individuals foresee in-group members to display a higher level of trust and are under moral and social pressures to act in the interests of the group (Ma, Huang, & Shenkar, 2011). Reflecting on the Ethiopian collectivist culture, Poluha (2004) talks of mahiberawi nuro, in which one must participate in community and in-group tasks to be a meaningful member of society. She states that:

Life in Ethiopia is dependent on your neighbors and friends; you must live social life or mahiberawi nuro (literally meaning collective life) and do so through membership in various social associations (e.g. solidarity groups, redistributive groups, work groups, religious groups, funeral associations). When you yourself or a member of your family becomes sick or die your association will help you manage socially, financially and morally … they are your social security.

The aforementioned axiom shows that the central functions of family, friendship and neighbourhood collectives in Ethiopia are care and social security, and obligations are not imposed upon members but rather depend on love, reciprocity and trust. In Ethiopia, opportunities for own saving are limited partly due to low per capita income. With dismal institutional support to help entrepreneurs enter into the labour market, migrants to cities in Ethiopia thus often turn to their family, friendship and ethnic networks to access physical capital, information on market opportunities, suppliers and customers (Getahun, 2015). Indeed, migration to Ethiopian cities is not only based on individual decision-making but also is a collective responsibility of both rural and rural households. Thus, most migrants up on arrival to cities do not suffer that much from problems of disorientation. Instead, family and friendship networks operate in such a way that new comers are provided with accommodation and other supports in finding employment (Baker, 1992). Though new migrants are provided with the needed assistance, they are expected to contribute economically to support urban households they live in and/or their rural families. The quick and easy means is self-employment in informal street activities such as selling of lottery tickets, electric equipment, kitchenware, used clothes, shoes, cigarettes, newspapers, etc., (Getahun, 2015).

Newly arrived migrants are further assisted in the search for urban employment by the already established friendship or kin networks. This directly or indirectly determines the type of future employment conditions that migrants will have. It is also clear that rural skills have little bearing on neither the type of urban employment nor the income level achieved (Baker, 1992; Getahun, 2015; Mano, Yamano, Suzuki, & Matsumoto, 2011). Thus, urban contacts are vital in determining the ultimate employment and economic position that a person attains (Baker, 1992). In Ethiopia, notwithstanding the presence of several microfinance institutions providing credit for the urban poor, informal entrepreneurs are excluded from such services as they are working outside the purview of government regulations (Getahun, 2015). For financing enterprises, therefore, entrepreneurs have to depend largely on family and co-ethnic resources rather than loans from government institutions. For example, a survey made by Haftu, Tseahye, Teklu and Tassew (2009) indicates that about 89 per cent of the start-up capital for micro-entrepreneurs comes from friends and relatives. A study by Getahun (2015) on informal activities in Addis Ababa also indicates that about 93 per cent of entrepreneurs obtained start-up capital from their close relatives and friends. Indeed, those who do not have kin members in the city rely on others of their own ethnic group as a substitute.

In Ethiopia, there are also several local institutions which meet the social and economic needs of many groups, including labour-exchange societies which are rural-based and religious associations which include both rural and urban sectors (Arega & Wubliker, 2015; Baker, 1992; Caudell, Rotolo, & Grima, 2015; Hoddinott, Dercon, & Krishnan, 2009; Kebede & Butterfied, 2009). For migrants in Addis Ababa, there are two important types of associations which fulfil important socio-economic functions: the equb (saving association) and the iddir (a funeral society). The equb is a rotating savings association where members pay a fixed amount of money each week or month into a fund and then cast lots to decide who should take the collected money (Baker, 1992; Caudell et al., 2015; Hoddinott et al., 2009; Kebede & Butterfied, 2009). A person who has won the lottery once cannot obtain again until all members receive their turn. Equb provides an ideal network for collecting cash for entrepreneurs requiring capital for business start-up and thereby expansion (Bisrat, Kostas, & Feng, 2012). The equb system indeed operates on the basis of mutual trust and members are mostly of family and co-ethnic who have confidence to each other. Equb thus provides an excellent platform for saving money within the security of family and ethnic networks (Kedir & Ibrahim, 2011). Iddir is a voluntary association that serves as economic and social insurance at times of death and other calamities (Kebede & Butterfield, 2009). Members of iddir contribute a fixed sum of money, weekly or monthly, and in return obtain financial and other support for funerals. Members are expected to attend the interment of diseased families and to participate in mourning. The iddir is a vital part of urban life and allows members the chance to affirm and endorse their commitment to kinship solidarity (Arega & Wubliker, 2015; Léonard, 2013).

Informal entrepreneurs work as independent self-employed persons with no employees. When needed, they employ members of their extended family without any wage. They do so not only because of lack of market intermediaries that channel information but also because they consider family labour to be more reliable and offer flexibility that are difficult to find on the labour market (Nordman, 2016). In reality, the returns from microenterprises are also very low and hence owners are unable to hire employees outside of their family (Getahun, 2015). Entrepreneurs also use family labour in their businesses since the family expects to be given jobs in microenterprises as the extended family originally helped to start the business and want to be rewarded for that effort once the enterprise is operational (Nordman, 2016). Moreover, informal entrepreneurs depend largely on ethnic networks to establish relations of mutual aid, partnership and cooperation with fellow entrepreneurs in the process of production and sharing of markets or the exchange of customers (Getahun, 2015). They also maintain regular suppliers for the purchase of goods, raw materials and equipment through their family and ethnic networks. In case of relations with municipality and police regarding inspection for access to public markets and/or working shades, they directly or indirectly draw on their ethnic ties (Getahun, 2015).

Such family and kinship networks are able to enforce social norms of behaviour, which in turn lower transaction costs in business relationships and reduce risks when systems such as efficient contract enforcement are not given publicly (Nordman, 2016). Dense and strong ties are helpful to penalise opportunistic behaviour and encourage cooperative ones (Nordman, 2016). Family and friendship networks also reduce uncertainties about market opportunities and increase the productivity of employees particularly of family labour which needs less supervision by the owner (Nordman, 2016). Reliance on dense family and friendship networks indeed shows the presence of trust among actors in a network. Trust is important for entrepreneurs as it speeds up decision-making and conserves ‘cognitive resources’ (Uzzi, 1997). Successful entrepreneurs are those who can build networks of trust, and this trust assist them in creating legitimacy within the market (Welter & Smallbone, 2006). Likewise, Lyon (2000) states that trust is a necessary condition for the expansion of private enterprises in conditions where actors cannot rely on formal institutions.

Overall, in the context of the Ethiopian informal sector, entrepreneurs greatly depend on the presence of densely knitted family and kinship ties. Indeed, the entrepreneurial effects of these family and kinship networks on entrepreneurs differ for enterprises in the formal and informal economies (Nordman, 2016). The effects of dense family and kinship networks on entrepreneurs also differ for enterprises in the formal and informal economies. These networks may be more critical in the informal economy, where it substitutes for limited formal support mechanisms in access to and management of factor inputs of entrepreneurship, than the formal sector (Nordman, 2016). The question here is thus ‘Do these dense family and kinship ties result in positive entrepreneurial outcomes for informal entrepreneurs?’ By examining the effect of such strong family and kinship ties on entrepreneurial outcomes, the results of this study will thus help to underscore the need for policymakers to identify and try efficient informal networks and develop support policies for vulnerable informal sector entrepreneurs. A better understanding of the effect of informal entrepreneurs’ networks on enterprise performance can also lead to the identification of policies that complement existing networks that already serve the needs of informal entrepreneurs and to policies that can substitute for networks that simply are not reaching informal entrepreneurs.

Theory and Hypothesis

An important insight of the SC theory is that entrepreneurs’ actions and outcomes can be predicted by the position individuals occupy in a network of relationships (Xiao & Tsui, 2007). The way in which networks bear influence depends on the structural characteristics of the network (Jackson, 2014). There are several ways by which features of network structure have been shown to produce tangible benefits (Barnes, Kalberg, Pan, & Leung, 2016). One variant of the SC approach is the structural holes theory of Burt (1992, 1997, 2000). According to Burt (1992), the effects of networks depend on the extent to which relationships are transitive, that is, the extent to which if node i is linked to node j, and j is linked to k, then i is linked to k. The frequency with which transitivity is present affects the extent to which connections reach out to new nodes and shapes information transmission (Jackson, 2014).

Burt (1992, 2000) distinguishes between two types of network structures that can create SC: closure and structural holes. The former is characterised by a network of densely connected people with the aim to enhance collective action. The argument is that well-connected entrepreneurs centrally embedded in networks benefit from increased access to information and resources (Borgatti, Jones, & Everett, 1998). When enclosed in cohesive ties, entrepreneurs also benefit from a normative environment that facilitates trust and cooperation between actors in a network (Coleman, 1988). Well-connected and centrally located entrepreneurs in networks can have increased opportunities to mobilise these benefits in pursuing their enterprise goals. Such type of network composition has been positively linked to economic productivity (Abbasi, Wigand, & Hossain, 2014; Greve, Benassi, & Sti, 2010). The dark side of network closure, however, is that it potentially limits the diversity of information that enters a group, as well as an entrepreneur’s freedom to pursue ideas outside the norms of the group (Burt, 1992).

An alternative network mechanism associated with SC is the notion of structural holes (Burt, 1992). Structural holes are gaps in the network structure across which there are no connections between individuals or groups. Entrepreneurs whose networks bridge the holes are brokers and they have the potential to connect the structural gaps (Burt, 2005). The SC is thus created by a network in which people can ‘broker’ the linkages between disjointed pieces. Stovel, Golub and Milgrom’s stabilising brokerage (2011) define brokers as intermediary links in social, economic or political relations who facilitate the diffusion of valued resources that would otherwise be difficult to obtain. Entrepreneurs who exhibit the brokerage position can get creative ideas and are more likely to look for solutions for business problems (Burt, 2000). In other words, entrepreneurs with less dense networks are hypothesised to have a greater diversity of information and a greater freedom to act. Thus, brokerage is a source of alternative ways of thinking and behaving (Burt, 1992, 2000, 2001).

Figure 1 shows a hypothetical network structure of two entrepreneurs. In this hypothetical network structure, entrepreneur A and entrepreneur B occupy different types of network positions with different ways for SC generation. Entrepreneur A occupies a closed or denser network position (lacking structural holes) than entrepreneur B. Entrepreneur A’s norm-enforcing social structure may lead to a social context with high trust and support. However, entrepreneur A receives more redundant information than entrepreneur B. In contrast, entrepreneur B’s ‘horizon-expanding network’, which span structural holes (Burt, 1992), leads to access to a greater diversity of information and greater freedom from unwanted social pressure than entrepreneur A.

OPEN IN VIEWER

Figure 1:

Source: Adapted from Burt (1992, 1997, 2000).

At the individual level, network brokerage is argued to positively affect enterprise performance by providing access and control benefits over information and resources (Burt, 1992). This was supported by a large body of empirical evidences (e.g., Burt, 2002; Burton, Wu, & Prybutok, 2010; Greve et al., 2010; Janhonen & Johanson, 2011; Klyver & Terjesen, 2007). Yet, other studies point to a dark side of brokerage and underscore the advantage of closure in networks (e.g., Barnes et al., 2016; Batjargal, 2010; Bizzi, 2013; Gargiulo & Benassi, 2000; Krackhardt, 1999; Stovel & Shaw, 2012; Xiao & Tsui, 2007), emphasising that the pressures associated with brokerage can place constraints on actors and potentially having a negative effect on enterprise performance. This controversy suggests that context is important in network studies. However, the majority of available empirical evidences come from socially homogenous populations in corporate and organisational settings, limiting a broader understanding of how context matters in the relationship between SC and entrepreneurial outcomes (Barnes et al., 2016; Batjargal, 2010; Burt & Burzynska, 2017; Xiao & Tsui, 2007).

The argument goes in a way that while brokers succeed in the market and in societies with market-like cultures, a collectivistic environment should reward those whose behaviours are ‘consonant’ with the context’s values (Xiao & Tsui, 2007). Brokerage is individualistic in nature to the extent that it starts from the principle of an independent self and puts priority on individual targets rather than collective goals (Xiao & Tsui, 2007). Collectivistic cultures, like in Ethiopia, are thus not only intolerant of brokering behaviours but also punish them if they act in such fashion (Ma et al., 2011). As discussed in second section, informal entrepreneurs in Ethiopia are heavily embedded in a collectivistic and dense family and kinship networks. The basic social fabric of collectivistic societies is that dense networks are usually the permanent groups against to the temporary and flexible groups of less dense ties, which have structural holes. Thus, the benefits of structural holes would be less likely realised in the Ethiopian informal entrepreneurs’ context. In other words, the dense family network of informal entrepreneurs may contribute more positively to entrepreneurial outcomes than structural holes. I, therefore, propose that informal entrepreneurs’ dense networks, which lack structural holes, has a significant positive effect on entrepreneurial outcomes in the informal economy of Ethiopia.

Methodology

Empirical Setting and Sampling

This study was conducted in Addis Ababa which is the capital and the primate city of Ethiopia. With a population of about 3 million, Addis Ababa accounts for about 25 per cent of the entire urban population of the country (CSA, 2013). Data were collected from street entrepreneurs as they represent the most visible form of informal workers. Street entrepreneurs were not, however, recorded in government statistics. This has created a problem in getting a sampling frame for the study. Since there was no list from which samples can be drawn using conventional sampling techniques, a combination of ‘time–space’ (Muhib et al., 2001) and ‘random-walk’ (Singh, 2007) procedures were used.

Multi-stage sampling procedures were applied to draw samples. First, the author consulted informants in the city to gather information about street entrepreneurs and their spatial distribution. Based on consultation, a 2-day tour was organised in the 10 sub-cities of Addis Ababa to identify street-vending cluster sites. Following the tour, 10 cluster sites (one from each sub-city) were identified purposively. Once cluster sites were identified, the author observed the date and time when vendors concentrated in these areas. Street entrepreneurs were thus found working in all days of a week and from 5 pm to 8 pm is the time for their immense gathering. These specific place and time arrangements form the primary sampling units and were used to construct the sampling frame. Since the space–time arrangements are sampled with a known probability, it is possible to make statistical inference about the street entrepreneurs (Muhib et al., 2001).

After space–time arrangements were set, samples were drawn using systematic random-walk ² technique. In so doing, specific random procedures were followed depending on the nature of the cluster site. The procedure involves taking the first road right or left of the sample site and then interviewing every 2nd, 5th, 7th or 10th street entrepreneur or interviewing every vendor after a 2, 5, 7 or 10 feet walk right or left of the road. The application of this procedure and the number of samples drawn have varied from one cluster site to another depending on the pattern, size and density of vendors in each location. As ethnicity and gender matter in networks, the sampling procedure has addressed ethnic and gender variables. Samples were then drawn from the three dominant ethnic groups—Amhara, Oromo and Gurage, as they account for about 84 per cent of the population of the city (CSA, 2013). Sampling was conducted in such a way that first vendors from one ethnic group comprising men and women were sampled followed by other ethnic groups using the same procedure. At the time of sampling, if every 2nd, 5th, 7th or 10th vendor was not from the ethnic group or gender that was intended to be sampled, the next person was selected. I have used the following formula to determine the sample size:

n = p ( 100 − p ) ( S E ) 2

S E = C o n f i d e n c e I n t e r v a l C o n f i d e n c e L e v e l

Where n is the sample size, p is the estimated percentage of the target population and SE is the standard error. As the number of street entrepreneurs is unknown, I have used the results of previous studies to determine representative sample size. A study on informality in Addis Ababa by Fransen and van Dijk (2008) showed that employment in informal enterprises accounts for about 26 per cent of total employment. I assumed that from 26 per cent of informal enterprise operators, 20 per cent might be accounted by street entrepreneurs. I also presumed a confidence level of 95 per cent and a confidence interval of 5 per cent of the target population figure. With this assumption, the minimum representative sample size was found to be 246. While sampling, eight samples were added and the total sample sizes for the study became 254 (refer to Appendix I for the distribution of samples).

Variables and Measures

Structural Holes: I used Burt’s (1992) constraint (c) to measure structural holes. In an ego-centred network, the extent to which an alter, j, constrains the ego, i, is a multiplication of (a) i’s investment in j, and (b) the lack of structural holes around j, that is,

C i j = ( P i j + ∑ q P i q P q j ) 2 , f o r q ≠ i , j

Where P_ij is the proportion of i’s relations invested in contact j and ∑ q P i q P q j is the portion of i’s relations invested in contact q who are in turn invested in contact j. Summed over all alters, ∑_jC_ij is the constraint measure. It is a function of network size and density. It is designed to measure the extent to which entrepreneurs’ network lacks structural holes. Constraint index ranges between 0 and 1, the higher an entrepreneur’s constraint score, the denser the network and the fewer structural holes that exist in his or her network. I used E-NET ³ to compute this measure (Halgin & Borgatti, 2012). Characterised by the presence of strong family and friendship ties, informal entrepreneurs in Addis Ababa have a dense and highly constrained network structure. The average network constraint was found to be 0.86. This high constraint index indicates the presence of fewer opportunities that enable informal entrepreneurs to bridge different connections.

OPEN IN VIEWER

Table 1.

Distribution of Samples Who Named at Least One Alter per Item and the Total Number, Mean and Standard Deviation of Alters Mentioned, N = 254

No.	Is /are there people who	% Yes	Number of Alters Mentioned
			Total	Mean	Std. Dev.	Min	Max
1	Gave you accommodation, food and other necessities while you came to AA	89.6	445	1.75	1.20	0	5
2	Provided you business ideas and advice while planning your business	98.1	635	2.50	1.17	0	5
3	Gave you credit or financial and material support while starting your business	92.8	338	1.33	0.69	0	4
4	Helped you to expand your business and improve business performance	88.9	559	2.20	1.25	0	5
5	You create a contact for recruitment of employees, if any	0.01	1	1	0.08	0	1
6	You establish business cooperation or partnership	96.1	635	2.50	1.30	0	5
7	Is/are regular supplier/s of your products or services	81.8	297	1.17	0.96	0	5
8	Assisted you while you were involved in improving the delivery of your products	73.3	559	2.20	1.04	0	5
9	Helped you to solve any administrative problems related to your business	7.14	30	0.12	0.45	0	3
10	Are very important to your business and help you in various aspects of your enterprise performance	100	785	3.09	1.15	1	5

Source: Field Survey (2014, April–November).

Enterprise Profit: there are various indicators of entrepreneurial outcome such as growth in sales, business turnover, profit, assets, enterprise survival and employment growth (Garoma, 2012). Employment growth is one of the most frequently used measures (Bigsten & Gebreeyesus, 2007). An enterprise is successful if it increases the number of its employees. The assumption is that growth in employment size is related to enterprise profit (McPherson, 1996). In this article, the use of employment growth is not feasible because street businesses are mostly managed only by the owner. Enterprise age is another measure of enterprise performance (McPherson, 1995). The logic is that the longer the age in business, the more successful the enterprise is. Enterprise survival, however, requires panel data and hence not applicable in cross-sectional studies. Profit is the most widely used measure of enterprise performance (Garoma, 2012). In this article, thus, profit is used as an indicator of enterprise’s performance. To establish reliable profit, respondents were asked a series of questions related to average weekly business expenses and sales volume. Profit was, therefore, calculated by reducing enterprise expenses including wages, rent, repairs, supplies, parts and maintenance. Profit is expressed in Ethiopian currency (Birr) and the average monthly enterprise profit was found to be Birr 764.00 (SD = Birr 365.00). In the regression model, the logarithm of profit is used to pull outliers from a skewed distribution closer to the bulk of data to be normally distributed.

Control Variables: I controlled for a number of factors that are associated with enterprise performance. Age of entrepreneurs is controlled for assuming that it does have inverse relation with profit (Rooks, Szirmai, & Sserwanga, 2009). Gender dummy (women = 1) is included in the model to control for various characteristics such as trouble to manage business, household duties, networks, restricted mobility, physical strength and discrimination (Khalife & Chalouhi, 2013). Human capital is positively related to profitability. Aspects of human capital that are related to enterprise profit including vocational training, formal education and previous business experience (Kantor, 2005) are controlled in the model. Age of an enterprise is another factor that is accounted for. The proposition is that with additional age in business profit increases. The presence of a spouse may positively influence profit (Bula, 2012) and thus I added a dummy variable (married = 1). Indigenous ethnicity is included in the analysis as different ethnic groups may have different business behaviour (Taye, 2001). Ethnicity is a dummy variable equal to 1 if an Amhara or 1 if a Gurage in reference to an Oromo. Factors of production such as working capital and working age family members have a positive effect on profit and hence included in the model. Location dummy (1 = centre) is added to control for differences in business environment across space. Furthermore, I tested interaction terms by adding a term to the model in which two independent variables are multiplied. Refer to Table 2 for the description of variables and expected signs in the model.

OPEN IN VIEWER

Table 2.

Descriptive Statistics and Expected Signs of Variables Used in the Regression Model

Variables	Variable description	Exp. Sign	Mean	SD
Enterprise profit	Enterprise profit in Birr	NA	764.00	315.00
Network constraint	Measures the degree of connection among alters ranging from 0 to 1	+	0.86	0.26
Religious services attendance	Respondents’ attendance of religious services (dummy), 1 = yes	NA	0.95	0.38
Frequency of religious services attendance	Frequency of religious service attendance per week (in days)	NA	3.57	1.06
Gender	The gender of the street entrepreneur (dummy), 1 = women	–	0.50	0.50
Marital status	Marital status of respondents (dummy), 1 = married	+	0.31	0.46
Age of the respondent	Age of respondents in years	–	29.94	8.15
Educational level	Respondents’ years of schooling	+	4.94	4.21
Vocational training	Respondents’ training before or after starting a business (dummy), 1 = yes	+	0.19	0.39
Previous business experience	Respondents’ business experience before joining street vending (dummy), 1 = yes	+	0.57	0.49
Working capital	Enterprise current working capital in Birr	+	1,120.95	1,398.67
Working age family size	Respondents’ working age family size	+	2.10	2.01
Location dummy	The location site of street enterprises (dummy), 1 = centre	+	0.44	0.49
Age of business	Duration of the enterprise in years	+	6.90	4.93
Ethnic Gurage	A respondent who is ethnically Gurage = 1	+/–	0.36	0.48
Ethnic Amhara	A respondent who is ethnically Amhara = 1	+/–	0.35	0.47
Gender*marital status	Interaction term of respondents’ gender and marital status	+/–	0.21	0.41

Source: Field Survey (2014, April–November).

Econometric Model

The OLS regression model was used to determine the causal relationship between network constraint and enterprise profit. The reliability of the OLS regression parameters may be, however, challenged by potential endogeneity bias with regard to the direction of causality (Fafchamps & Minten, 2002). I, therefore, tested whether network constraint can be regarded as exogenous using econometric procedure of instrumental variables (IV) estimation (Wooldridge, 2013). However, obtaining an appropriate IV is problematic due to the absence of precise modelling of IV selection (Durlauf, 2002; Durlauf & Fafchamps, 2004). The selection of IVs has varied from study to study depending on the dimensions used to measure SNs and the nature of the outcome variables. For example, in a study of kinship networks and economic behaviour in rural Ethiopia, Werger (2009) instrumented SN measures with religion and ethnic dummies. He recommends that religion dummy instruments network measures better than ethnic dummies. Getahun and Odella (2014) also used religion dummy to instrument resources embedded in networks.

Religion is indeed a good proxy for SC accumulation. However, I believe that it is whether and how frequently an entrepreneur goes to houses of worship rather than whether he/she attends a Pentay, Islamic or other service that determines his/her level of SC accumulation. In this article, therefore, network constraint is instrumented with participation in religious services and the frequency of participation. This is because most people in Ethiopia attend religious services either daily or weekly depending on the respective sect of each religion. Churches, mosques and other houses of worship, thus, provide an institutional base for ‘civic good works’ and a learning opportunity for entrepreneurs (Hopkins, 2011). Indeed, involvement in religious affairs offers opportunities for social interaction among likeminded people, nurtures friendships and fosters social ties (Lim & Putnam, 2010). Religious services through worshiping places and membership in spiritual institutions offer an opportunity to entrepreneurs access business resources (Miguel, 2004). The proposal in this article is that both religious service attendance and frequency of participation are neither directly affected by enterprise profit nor do they affect enterprise profit directly. However, these two variables affect SC accumulation and mobilisation (refer to Appendix II for correlation of variables). The model fit by IV regression is given by:

Y i = y i β 1 + x 1 i β 2 + u i ​

(1)

y i = x 1 i Π 1 + x 2 i Π 2 + v i

(2)

Where Y_i = the dependent variable for i^th observation, y_i represents endogenous regressors (SN measures), x_1i represents included exogenous regressors (endowment of human capital and a vector of individual characteristics and enterprise characteristics) and x_2i represents the excluded exogenous regressors (religious service attendance and the frequency of participation). Variables x_1i and x_2i are collectively called the instruments. u_i and v _i are zero mean error terms and the correlations between u_i and the elements of v _i are apparently non-zero.

Results and Discussion

The estimation results of enterprise profit by OLS and 2SLS estimators are presented in Table 3. The models significantly predict enterprise profit. An assessment of variance inflation factor also showed no problematic signs of collinearity with inflation factor of 1.46. After fitting 2SLS, tests of endogeneity were performed to check whether network constraint supposed to be endogenous in the OLS model can be treated as exogenous or not. The result of Hausman test indicates the exogeneity of network constraint. The Sargan test of over-identifying restrictions is also presented in Table 3. The Sargan test result confirms the validity of the two instruments.

Network constrain is used as a SC measure to analyse the relationship between SNs and enterprise performance. The core networks of entrepreneurs were made up of densely connected family and friendship ties. Despite the fact that dense family networks are sources of emotional and material support for street entrepreneurs, network constraint has a significant negative effect on enterprise profit (Table 3). The higher the density of networks, the higher the network constraint and the lower the enterprise profit and vice versa. This result leads to reject the hypothesis that dense networks, which lack structural holes, have a significant positive effect on entrepreneurial outcomes. The result thus corroborates Burt’s structural holes argument.

OPEN IN VIEWER

Table 3.

Regression Estimation Results of Network Constraint Predicting Enterprise Profit

Variables	OLS	2SLS
Network constraint	–0.01108 (4.06)***	–0.01159 (2.99)***
Education in years	0.00658 (0.69)	0.00754 (0.74)
Vocational training (1 = yes)	0.05488 (0.64)	0 .01965 (0.18)
Business experience (1 = yes)	0.11360 (1.73)	0.09057 (1.14)
Gender (women = 1)	–0.24585 (2.83)***	–0.20396 (2.48)**
Marital status (married = 1)	0.02814 (0.22)	0.00671 (0.07)
Age of respondents	0.00441 (0.76)	0.00326 (0.50)
Working age family size	0.04311 (2.13)**	0.05657 (1.82)*
Working capital	0.00014 (2.79)***	0.00010 (2.33)**
Business location (1 = centre)	0.18015 (2.53)**	0.18143 (1.39)
Ethnic Gurage	0.39230 (4.83)***	0.38131 (4.34)***
Ethnic Amhara	0.13554 (1.58)	0.03411 (0.19)
Age of business	0.00607 (0.72)	0.02414 (0.00)
Gender*marital status	–0.26903 (2.00)**	–0.24721 (2.32)**
Constant	6.14361 (28.29)***	6.85731 (7.05)***
F statistic/Wald Chi2	7.14***	95.06***
R2	0.4403	0.3446
VIF	1.42	–
Hausman test (p value)	–	15.32 (0.57)
Sargan test (p value)	–	3.50 (0.31)

Source: Field Survey (2014, April–November).

Notes: Figures in parenthesis are t values for OLS and z-values for 2SLS. *** P<0.01, ** P<0.05 and * P<0.10.

Given that street entrepreneurs depend on dense networks and considering the high uncertainty markets in the informal sector, it is unforeseen that informal entrepreneurs’ networks appear to be inefficient for the performance of enterprises. This can be attributed to a number of causes. The first is related to the limitations inbuilt on dense networks in that entrepreneurs may depend on dense business support networks and become satisfied with what they have. This may damage entrepreneurs’ effort to be successful in their businesses as much as they would if they had no support from dense networks. Second, dense family networks just put entrepreneurs under pressure to sell them products at reasonable prices which reduce profitability. Third, as argued by Nordman (2016), entrepreneurs who are part of dense networks and who achieve enterprise success are often called on to share their achievement with less successful network members. In such cases, adverse incentives arise if concerns about moral pressure for jobs, credit or free business services by family and kinship ties discourage entrepreneurs from dealing with people outside dense networks.

Dense networks do not only facilitate but also hinder the behaviour of individuals due to pressure and normative expectations from family and kinship members (Krackhardt, 1999). In fact, the length and regularity of relationships did not enable entrepreneurs to obtain efficient circulation of resources (access to information, financial support, business partnership, etc.). Previous studies (e.g., Alesina & Giuliano, 2009; Chai & Rhee, 2010) have shown that high level of family embeddedness for entrepreneurs is associated with greater trust in their families and lower levels of generalised trust. This perceived labelling between family and non-family ties can generate suspect, under-appreciation or dismissal of advice given by non-family ties (Nordqvist, 2005). Such dense networks can provide a limited set of information that may not match future business needs thus hindering enterprise performance.

Elfring and Hulsink (2003) note that dense networks have the ‘risk of over-embeddedness’. Studies (e.g., Aral & Van Alstyne, 2007; Brailly, Favre, Chatellet, & Lazega, 2016; Ferriani, Cattani, & Baden-Fuller, 2009; Molina-Morales & Martínez-Fernández, 2009) underscore to the problem of being over-embedded or too central in networks. These studies show that excessive group cohesion hinders knowledge and information accessibility among actors in a network. They showed evidences of diminishing returns associated with being well connected in information-sharing networks. This is because the group’s closure may reduce the opportunities to find new and diverse information outside a given cluster. With more dense social relations, there comes increased coordination costs as more time and energy are devoted to maintain them (Ferriani et al., 2009). Since time and energy are exhaustible resources, increasing the strength of ties beyond a certain limit is likely to result in diminishing returns (McFadyen & Cannella, 2004). Moreover, as entrepreneurs become more central in networks, they are faced with high inflows of information and bear greater ‘cognitive pressure’ associated with processing it (Dodds, Watts, & Sabel, 2003). This potentially result in ‘information overload’ and difficulties in sorting relevant and irrelevant information (Ferriani et al., 2009) leading to poor business decision-making (Schneider, 1987).

As shown in Table 3, control variables, such as gender, working capital, ethnicity and the interaction term between gender and marital status, become significant predictors of enterprise profit. Gender has the expected negative sign and is significant at 1 per cent. This might be attributed to many probable causes. To start with, women do have double responsibility in taking care of household duties and managing street businesses. Most women entrepreneurs are also working in permanent places and their businesses are located near their homes where they cannot get many customers. In addition, women are engaged in very small businesses which are extensions of their reproductive role including cooked and non-cooked food stuffs. These businesses demand low capital and thereby generate low profit.

Ethnicity is presumed to be a factor that influences enterprise outcomes. As indicated in Table 3, being a Gurage is positively correlated with profit and the correlation is significant at 1 per cent level. But being an Amhara does not have a significant effect on profit compared to being an Oromo. This finding is consistent with other studies. For example, a study by Taye (2001) has shown that enterprises owned by the Gurage, despite their low educational level, perform better than members of other ethnic groups, such as the Amharas who have relatively higher level of education. This is probably because the Gurages are known for their hard-work and business skills. The coefficient of the interaction term between gender and marital status is negative and significant at 5 per cent level. This reflects that having a husband makes women to be in an undesirable position because usually women are not the sole decision makers of their businesses. According to the norm in the Ethiopian society, it is usually husbands who make decisions in the household. Even when women are involved in household discussions, the final say is made by husbands. The variable working capital has a positive sign and is significant at 1 per cent level. On the human capital side, contrary to what is expected, the estimation results show that education, vocational training and previous business experience are not significant predictors of profit. This is probably because those engaged in street activities have limited skills, low education levels and low occupational status. They are also engaged in petty businesses that do not need any training and advanced education.

Conclusion and Implications

Social capital constitutes information and resource channels that can reduce the amount of time and investment necessary to gather and process information. It can facilitate learning and the diffusion of innovations. Indeed, the majority of empirical evidences linking SC to economic advantage come largely from large-scale business environments and in developed country settings. Here, I extended the SC and entrepreneurship research to account for the urban informal sector in Ethiopia. Informal entrepreneurs in Addis Ababa are supported by their family and kinship ties to start and carry out micro-businesses. The value of being well-connected in networks is that it provides increased opportunities to capitalise resources to enterprise start-up and obtains information on employment opportunities. Family and kinship networks may reduce transaction costs in business relationships as sanctions may be used to punish uncooperative behaviour and thus to encourage cooperative ones. A central question in this study was ‘Do the dense network structures of informal entrepreneurs positively contribute to entrepreneurial outcomes?’ The findings suggest that being over-embedded in family and kinship networks has a negative effect on enterprise performance in the informal sector, which I argue is due to lack of structural holes in entrepreneurs’ networks. This is indeed against the empirical findings in other developing countries (e.g., Berrou & Combarnous, 2011; Xiao & Tsui, 2007) but corroborates the theoretical construct of Burt (1992, 2000, 2001, 2002), which reflects the condition of developed countries. This study, thus, contributes to the SC and entrepreneurship literature in which the ability to realise the benefits of networks are likely dependent on context under study either in developed or developing economies. The advantages of either being over-embedded in dense family and kinship networks or brokering structural holes depends on context. Thus, it is impractical to prescribe blanket recommendations about the role of SC in entrepreneurial outcomes that work for all situations.

As social relationships are important mechanisms in designing poor targeted programmes, policies aiming at improving the economic situation of vulnerable entrepreneurs should consider the fact that entrepreneurial behaviour is often influenced by the dense network of family ties not only through learning and complementarities but also through social norms and pressure to redistribute earnings. Dense networks provide benefits through risk-sharing and lower transaction costs, but excessive dependence on dense networks can worsen enterprise outcomes as entrepreneurs lack structural holes that can help them to obtain diverse information. Thus, policies should consider the many benefits of social contacts while taking into account the downside of being excessively embedded in dense networks. By creating proper channels, informal entrepreneurs can be linked to peoples of diverse socio-demographic groups to gain the benefits of diversity in networks.

Limitations and Areas of Future Research

Though my analysis is based on data from a single informal economy of street entrepreneurs, I believe that the results of the study are applicable to broader range of settings in the informal sectors of Ethiopia. Yet, this study has limitations. First, measuring and explaining the effects networks on enterprise outcomes are difficult. The name generator and interpreter instruments are newly developed and applied in the Ethiopian informal sector context and needs further examination. As argued by recent studies (e.g., Green, Hoover, Wagner, Ryan, & Ssegujja, 2014; Herz & Petermann, 2017), network research interviewers have an influence on network size and the nature of relationship among actors in a network. Homophilous interviewer-respondent relationships have also an effect on network characteristics. There are also evidences of training and fatigue effects on network size. However, much of the variation in network size caused by the interviewer still remains unexplained and may have an effect on the characteristics of entrepreneurs’ personal networks and thereby their effect on enterprise performance (Herz & Petermann, 2017). Second, analysis is hampered by methodological challenges related to sampling of the hidden population. The random-walk sampling strategy employed in this study is not widely applied in other urban-based studies in Ethiopia. Instead of heavily relying on the commonly used snow-ball sampling techniques, studies that involve hidden and hard to reach populations can experiment the random-walk procedure and improve its validity and application in the Ethiopian context. Third, I have used cross-sectional data to characterise the personal networks of entrepreneurs. But as entrepreneurs’ networks are dynamic in nature, it would have been to observe the dynamics of personal networks across time, using longitudinal information. Thus, other researchers can further conduct a research in the same area by studying the personal networks of entrepreneurs’ overtime. Generally, to establish the importance of networks in the urban informal sector, issues related to sampling, techniques of SN data collection and the dynamic nature of personal relationships should be dealt with. It is thus imperative to conduct further research in a broader scope including other urban areas and other informal activities. In spite of these limitations, collecting a unique set of data in a collectivistic culture of Ethiopia and considering informal entrepreneurs as a case study, this study fills a gap in the SC and economic outcomes literature by providing evidence of the dark side of over-embeddedness in networks.

Declaration of Conflicting Interests

The author declares that there is no conflict of interest with respect to the research, authorship and/or publication of this article.

Funding

The author received no specific grant from any funding agency in the public, commercial or not-for-profit sectors for the research, authorship and/or publication of this article.