The Personal Networks of Entrepreneurs in an Informal African Urban Economy: Does the ‘Strength of Ties’ Matter?

Abstract

This paper investigates Granovetter's “strength of weak ties” hypothesis in an informal African urban economy. It outlines an approach articulated around the reticular embeddedness conceptual framework associated with the notion of “ego-centred network.” The content of ties in an entrepreneur's network is described by three salient dimensions: strength, social role and exchanged resources. We use an original dataset collected in the informal economy of Bobo-Dioulasso (Burkina Faso) to evaluate how the content and strength of ties influence entrepreneurs' economic outcomes. The instrument of multiple name generators provides a vast amount of information that can be used to compute quantitative measures of the composition of networks. We show that both strength of ties and proportion of business ties have a significant positive impact on economic outcomes. It reveals the importance for small urban informal entrepreneurs to draw on both embedded social relations and more autonomous ones.

Keywords:

• informal economy
• social networks
• names generators
• MSE's performances
• Burkina Faso

Notes

In the social capital literature it is referred to as “bridging social capital” as opposed to “bonding social capital” (Woolcock 1998). Whereas the former refer to generalized social relationships across groups (such as those that are developed with more distant acquaintances and colleagues), the second refers to relations between community members and close friends (such as those developed in family or ethnic groups).

The cut from 317 to 278 entrepreneurs is the result of unavoidable field constraints. The representativeness of this sub-sample, regarding branches of activity and geographical areas, has been controlled by chi-square tests for expected values.

These generators are the most discussed in the literature, but most of the time from a merely one-dimensional standpoint. We here propose a more integrated approach.

No entrepreneur has quoted fewer than two names and less than 5% of them have a network made of two or three names (the maximum size is 15 names). The average network size is 7.1 names, and 50% of entrepreneurs recall a network made of five to nine names (80% recall a network made of 4 to 11 names).

The sub-sample is made up of the first quoted names at each generator, as Fischer (1982) proposed in his survey of personal network support in San Francisco.

We have followed the classic four levels of Burt's (1992) scale of intensity (especially close, close, less than close, distant). However, the preliminary tests of the questionnaire have shown that respondents made no significant difference between close and especially close alternatives. Then, in order to reduce the burden for respondents, we have cut the variable to three alternatives: close or especially close, less than close, distant. Reciprocity is a dichotomous variable: respondents were asked if they have already (during the last 12 months) helped (financial, material or informational support) the person they have quoted. For more detailed information about how strength of ties variables are constructed, see Table A.1 in the Appendix.

From an international comparative standpoint, this result may express a surprisingly small proportion of kinship ties compared with what might have been a priori expected in an African city. Indeed, most studies in industrialized countries show a proportion of kinships ties contained between 15 and 25% on average (Drakopoulou Dodd and Patra 2002). This is quite the same as our result, which remains far below the 36% found, for example, for South African entrepreneurs (Mitchell and Co 2004). However, as stated above, these studies focus on business discussion network. If we also focus on our first generator, we find that kinship ties represent 39% of ties conveying advice, information and ideas (46% for administrative support), which constitutes a less surprising result.

More precisely, this variable ranges from zero to one according to the median strength of ties in the network of each entrepreneur. The use of median rather than average value helps to avoid the impact on average of extreme values, particularly as a result of the obvious considerable strength of some kinship ties.

The assumptions of linearity, normally distributed errors and uncorrelated errors were checked and met. The VIF tests show that no serious multicollinearity problem occurs.

The fact that the proportion of tangible resources is never significant in the model is in line with this idea. It suggests that the combination of the two types of resources may be useful.

The control variable of network size is never significant in our model.