Self-employment and job generation in metropolitan areas, 1969–2009

Abstract

Many regional development policy initiatives assume that entrepreneurial activities promote economic growth. Empirical research has presented rationale for this argument showing that small firms create proportionally more new jobs than large firms. However, little research has been performed on the issue of net job generation at the urban level, particularly when self-employment is considered as an indicator of entrepreneurial activities. This paper investigates to what extent US metropolitan areas in the 1969–2009 period characterized by relatively high rates of self-employment also have shown relatively high rates of subsequent total employment growth. The analysis corrects for the influence of sectoral composition, wage level, educational attainment, presence of research universities and size of the metropolitan area to measure the extent to which the number and quality of self-employed in a region contribute to total employment growth. It finds the relationship between self-employment rates and subsequent total employment growth to be positive on average during the 40-year period but to weaken over time.

Keywords::

• self-employment
• entrepreneurship
• metropolitan areas
• regional development

Jel Classification::

• R11
• O18

Acknowledgements

We are grateful to two anonymous referees for providing us with helpful comments on an earlier draft. Emilio Congregado and Antonio Golpe acknowledge the financial support from the MICINN (Ministerio de Economía y Competitividad, Project ECO2013-43526-R).

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

^1. Similar arguments can be found in Shane (2009) or in Congregado, Golpe, and Carmona (2010) among others.

^2. We are aware of the recent stream of literature investigating the impact of regional start-up rates on employment growth, as summarized by Acs and Storey (2004), Fritsch (2008) and Dejardin and Fritsch (2011). However, since the self-employment rate considered in the current paper is a static indicator of entrepreneurship (instead of a dynamic indicator such as the start-up rate), it is far from obvious that the relationship with macro-economic performance is similar for these two indicators of entrepreneurial activities. For example, incumbent self-employed do not suffer from liabilities of newness but may be less innovative compared to recent entrants.

^3. A recent and exhaustive analysis of business ownership developments in the USA is carried out by Hipple (2010).

^4. One could expect that a metropolitan area that had a high concentration in traditional manufacturing in the past to have lower employment growth than areas which were well positioned in expanding business sectors. This hypothesis is consistent with the empirical evidence provided by Glaeser et al. (1995) or more recently by Blumenthal, Wolman, and Hill (2009).

^5. See Carlino and Chatterjee (2002) for a study explaining the decline in the share of urban employment accounted for by the relatively dense metropolitan areas from congestion costs.

^6. Equation (2) showing how the unobservable effects are assumed to change over time is similar to the state (or transition) equation in state space models (estimated by the Kalman filter).

^7. Fölster (2000) claims a significant positive effect of self-employment on overall employment for Swedish counties in the 1976–1995 period.

^8. Papers that provide empirical evidence for industry composition affecting regional employment growth include Glaeser, Scheinkman and Schleifer (1995), Partridge and Rickman (1996), Clark (1998) and Rosenbloom and Sundstrom (1999). Recently, Blumenthal, Wolman, and Hill (2009) provide evidence of this effect for MSAs in the USA.

^9. Jackson (1984) presents evidence for the number of different product categories bought to be a positive function of income.

^10. See Eaton and Eckstein (1997) and Black and Henderson (1999) for models of the interrelationship between urbanization, accumulation of human capital and economic growth. Carlino and Chatterjee (2002) and Chatterjee and Carlino (2001) stress congestion costs to explain the trend of employment deconcentration. This deconcentration process involves the decline in the share of urban employment accounted for by the relatively dense metropolitan areas.

^11. Note that we do not incorporate fixed effects into the model. This would lead to a large number of additional parameters and to problems concerning estimation of a fixed effects dynamic panel data model with (two) lagged dependent variables (see, e.g. Hsiao, Hashem pesaran, and Tahmiscioglu 2002). More importantly, it is unlikely to have unobserved effects to remain constant over a 40-year time period.

^12. See Table A1 in the Appendix for variable definitions and data sources.

^13. For the manufacturing share variable, there have been a small number of missing observations in the data source that have been filled up. For Albuquerque, NM, (1998, 1999, 2000), Albany-Schenectady-Troy, NY (2005), Amarillo, TX (from 1998 till 2009 except 2008), Charlottesville, VA (from 1999 till 2003), Columbus, GA-AL (2000), Decatur, AL (1998, 1999, 2000), Dothan, AL (2005 and 2006), Dover (2002, 2007, 2009), Jefferson City, MO (2006, 2007), Monroe, LA (from 2002 to 2007), Portland, South Portland-Biddeford, ME (from 1998 to 2005), Providence-New Bedford-Fall River, RI-MA (2005), Rapid City, SD (2003), Sioux City, IA-NE (1998, 1999, 2000) and Washington, DC-MD-VA-WV (1998 to 2007) the MAN-variable is interpolated. The interpolation is the simple average of manufacturing employment in the period before and after the missing(s) observation.

^14. The city of Flint was dominated by the automotive industry (General Motors) during the twentieth century. The demise of this industry has contributed to Flint's population being halved during the 40 years under investigation.

^15. The first two sub-periods 1969–1974 and 1974–1979 were periods of strong total employment growth, on average. The third sub-period 1979–1984 was one of economic downturn and relatively small total employment growth. The fourth sub-period 1984–1989 was one of recovery and relatively high total employment growth. The next two sub-periods, 1989–1994 and 1994–1999, are comparable to the third and fourth sub-periods, respectively, but the downturn and recovery were both less strong. Finally, the last two sub-periods show a small total employment growth, especially the last one, affected by the financial crisis.

^16. This is in line with the findings by Birch (1981), Kirchhoff and Phillips (1988) and Neumark, Wall, and Zhang (2011) that there is empirical evidence of small firms to be at the core of new job creation in the USA, at least for the first part of our sample.

^17. This would be in line with the model introduced by Carree et al. (2002). They present empirical evidence that countries with very low self-employment rates (like the Scandinavian countries) and countries with very high self-employment rates (like Italy) have suffered in terms of forgone growth of GDP.

^18.http://www.nber.org/cycles.html.

^19. These results are available on request from the authors.

^20. See Klepper (1996) and Carree and Thurik (2000) for economic models explaining the shakeout of small producers in later stages of the industry life cycle. See Audretsch (2001) for an overview of the role of small firms in the US biotechnology industry.

^21. Audretsch and Fritsch (2002) also indicate that there are different growth regimes across time. Regional economic development policy targeted towards either fostering new start-ups or nurturing large, incumbent enterprises may be appropriate in some time periods while counterproductive in others.