«Paper No. 365 Determinants of self-employment among commuters and non-commuters Mikaela Backman Charlie Karlsson May, 2014 The Royal Institute of ...»
CESIS Electronic Working Paper Series
Paper No. 365
Determinants of self-employment among commuters and
The Royal Institute of technology
Centre of Excellence for Science and Innovation Studies (CESIS)
Determinants of self-employment among
commuters and non-commuters
Centre for Entrepreneurship and Spatial Economics (CEnSE) Center for Science and Innovation Studies (CESIS), Jönköping International Business School, Jönköping, Sweden Abstract: In this paper, we analyse the determinants of the decision to become self-employed among commuters and non-commuters. In the entrepreneurship literature it is claimed that the richness and quality of an individual’s business, professional and social networks play an important role for the decision to become self-employed. People that commute between localities in the same region or between localities in different regions will most probably be able to develop richer personal networks than non-commuters, since they can develop network links both in the locality where they live and in the locality where they work. In this paper, we test this hypothesis using micro-data for around three million individuals in Sweden. As far as we know, this is the first time this hypothesis is tested. In our empirical analysis, we make a distinction between three groups of individuals: non-commuters, intraregions commuter and inter-region commuters. For each of this groups we test how the probability of becoming self-employed is influenced by a number of characteristics of individuals, characteristics of home and work localities and regions. Our results indicate a significant difference between non-commuters and commuters in terms of the role of networks for becoming self-employed. On the one hand, we find for non-commuters that living and working in a locality with rich business networks reduce the probability of becoming self-employed. For commuters, on the other hand we find that working in a locality with rich business networks increase the probability to become self-employed. In this latter case, living in a municipality with rich business networks has a non-significant effect on the probability of becoming self-employed. Our results indicate that it is the business networks where people work, rather than where they live that exerts a positive influence on the probability of becoming self-employed.Keywords: entrepreneurship, individual attributes, regional attributes, networks, micro-level data JEL Codes: C21, J24, L26, R12 *Corresponding author
1. Introduction In this paper we analyse how the labour market behaviour of individuals influence the probability that they will turn from employment to self-employment. In an earlier paper (Backman and Karlsson 2013), we among other things showed that the probability of commuters becoming self-employed is positively influenced by the accessibility to entrepreneurs in their work locality, while the same accessibility in their home locality was insignificant. We interpreted this as an indication of the importance of the business, professional and social networks that individuals build up in their work localities that are critical for becoming self-employed while the networks they build up in their home municipality play an insignificant role.
In this paper, we broaden and deepen our earlier analysis in several ways. Firstly, we check for, if being a commuter earlier, influence the probability of non-commuters to become selfemployed, i.e. if old business, social and professional networks may play a role for the decision among current non-commuters to become self-employed later in life. Secondly, we analyse if the duration of the commuting career among current commuters influence the decision to become self-employed. The idea here is that it probably take several years to build up the necessary networks in the work locality to have got all the different contacts that might be critical for becoming self-employed. Fourthly, we make separate estimations for four aggregated occupational categories and for different educational levels.
Our empirical results show, for non-commuters, that living and working in a locality (municipality) with rich “business networks” in terms of accessibility to self-employed reduces the probability of becoming self-employed but if, on the other hand the locality is located in a labour market region (commuting region) with rich “business networks” increases the probability of becoming self-employed. Interestingly, if a non-commuter earlier has been a commuter that decreases the probability of becoming self-employed.
For commuters that commute within a labour market region our results show that working in a locality with rich “business networks” increases the probability of becoming self-employed for all categories analysed except those with an occupation within management and administration. For these commuters it is only for one of seven categories – cognitive occupations – that rich “business networks” in the region has a positive effect on the probability of becoming self-employed and for one category – social occupations – the effect is even significantly negative. Years of commuting have a significant positive effect on the probability of becoming self-employed for six of the seven categories.
Turning now to long-distance commuters, i.e., people that commute between different labour market regions the effect of “business networks” on the probability of becoming selfemployed varies substantially. For people in cognitive occupations we get significant positive effects from rich “business networks” in both the home and the work locality as well as from the home region. People in management and administrative occupations rich “business networks “ in the home locality has a positive effect on the probability of becoming selfemployed, while rich “business networks” in the work locality has a negative effect. For social occupations we find a positive effect from rich “business networks” in the home locality and a negative effect from rich “business networks” in the work region. For people with standardised occupations we get no significant effects. People with a low education get a positive effect from rich “business networks in the home locality but the opposite from rich “business networks” in the work locality. For those with a medium education we found a significant positive effect from rich “business networks” in the home locality. For people with higher education all the network effects were insignificant. For all groups of long-distance commuters we found that the probability of becoming self-employed increased significantly with the number of commuting years.
This paper is organized as follows: In Section 2, we discuss the foundations and formation of “entrepreneurship” networks and their role individuals’ decisions to become entrepreneurs, i.e. to become self-employed. Section 3 present our empirical design with chosen variables and empirical results. The last section, 4, concludes the paper.
2. “Entrepreneurship” networks: their foundations, their formation and their role for new firm formation In this section, we develop the theoretical foundations for our main hypotheses that the density of the “entrepreneurship” networks in the locality where people work has a significant positive influence on their probability of becoming self-employed, i.e. to become entrepreneurs. We do so by first discussing the foundations for and the formation of general economic networks in Sub-section 2.1. In Sub-section 2.2, we concentrate the focus on the foundations for and formation of general social networks, since the entrepreneurship literature today contains numerous contributions illustrating the importance of private, business and professional networks for the start-up, launching and running of a new business. In the last sub-section (Sub-section 2.3), we identify a condensed form of social networks, that we name “entrepreneurship” networks and highlight their critical role in several dimensions for peoples’ willingness to become entrepreneurs. We focus on the mechanisms behind the formation of these networks and in which spatial milieus these networks are formed. However, the critical point here is if there are systematic differences in terms of which “entrepreneurship” networks that are most important for entrepreneurship, namely those in the locality where a person live or those in the locality where a person works.
2.1 Economic models of networks Economic models for the emergence of networks – ‘connections’ models’ (Jackson and Wolinsky 1996) – cope with circumstances where actors have the possibility to set up bilateral links with other actors in order to get access to something that they value. When all the actual links are added together they form a network. In the connection model, each actor initially possess some information. When a link is established between actor i and actor j that allows actor i to access not only the information possessed by actor j but also to other information, which actor j has access to via his other links (and vice versa for actor i).
Furthermore, the larger the distance in the network between actor i and actor j, the lesser the degree to which actor i will have access to the information possessed by actor j.
In our case the nodes, i.e., the actors, are represented by individuals. This implies that the networks we are interested in can be called social networks. The existing links in a network should be interpreted as capital objects, since they represent sunk costs. This implies that networks bring rigidity and structure into the interaction patterns at all spatial scales, since they tend to reduce interaction costs between the nodes in the network. Network links achieve capital properties since their establishment is the result of a link-specific investment that has to be carried by both nodes, i.e. individuals, but often to varying degree. When, for example, two individuals decide (explicitly or implicitly) to establish a joint link and thus a network, it is possible to think of this as the outcome of an evolutionary, gradual search and trial process.
Thus, we may regard the outcome as a Nash equilibrium of a non-cooperative game, where each part would lose by leaving the network.
It is possible to introduce a spatial dimension into the connections’ model by making the costs of establishing a link dependent upon actors spatial location. In a model presented by Johnson and Gilles (2003), geographic space is represented by a unit-length line along which the actors are located at fixed intervals. It is assumed that the cost of creating a link between two actors increases linearly with the distance on the line between the two actors. In a similar model developed by Carayol and Roux (2007), actors are located on a circle at equidistant intervals, and therefore do not initially occupy asymmetrical positions.
Within a network equilibrium both these models for reasonable large parameter intervals generate ‘small-world’ networks, i.e. networks characterized by very high rates of proximity links and relatively few long-distance links. ‘Small world’ networks satisfy two conditions: i) a small diameter of the network as well as a low average distance between the actors inside the network, and ii) a high clustering coefficient relative networks where links are created through a random process. The two models predict that spatial proximity is a central determining factor for collaborative choices, i.e. for the establishment of links. The rationale behind this is quite simple. It is more rational, ceteris paribus, for an actor to establish a link to another actor who is close by, rather than to an actor that is far away when the initial information endowment of each actor is equal and when the cost of creating a link to a nearby actor is lower. However, when actors in a network only have formed links to nearby actors, it might be worthwhile for an actor to substitute a nearby link with a more distant link due to the extra information he can gain indirectly from the distant actor’s other links despite the higher costs for establishing a distant link.
It is possible in this kind of models that in network equilibrium the number of distant links is too low in a network. This implies an efficiency problem: when one actor establishes a long distance link, this generates a positive externality for the nearby actors to which he has a link and it may now be the case that no other actor is motivated to establish a second distant link even if that would gain the network as such, since for the individual actor the cost of establishing a distant link might be higher than the gain for this individual actor.