**4. Empirical Results**

The empirical exercise is conducted by testing the impact of the local labor market portfolio, based on wage and salary employment data, on the new business survival rate in 2010 in county*<sup>i</sup>* . Following the convention in the employment portfolio literature, employment-based measures of risk and return trade-off are used [1,4]. We begin by looking at the role of employment-based portfolios across the spectrum of the chosen regional unit: counties in the U.S., categorized as all, metro, nonmetro, towns, and micro.

WSE portfolios from 1996 to 2005 and from 1997 to 2006 are found to have a significant impact on the survival rate of new businesses in 2010 in metro counties. These are establishments that were born in 2005–2006. All three measures, WSE-Risk, WSE-Growth, and WSE-Growth Squared, were found to be significant, though to varying degrees (Table 2).


**Table 2.** Summary results of the WSE portfolio for new business survival rates in 2010.

\* *p* < 0.1, \*\* *p* < 0.05, \*\*\* *p* < 0.01; Standard errors are in parenthesis.

We validate these results by using another potential measure of risk and return tradeoff in the local labor market, wage and salary income data. A likely criticism of using this measure is the possible influence on the results because of differences in the cost of living in the regional units. Though challenging, the use of this measure is expected to facilitate the validation of the results found earlier using employment-based measures.

Conducting the empirical exercise with the income-based risk and return measures for different timeframes yields interesting results. Similar to the employment-based portfolio, this portfolio comprises wage and salary income growth (WSI-Growth; defined as the average annual wage and salary income growth in the county during the selected timeframe(s)), wage and salary income growth squared (WSI-Growth Squared; defined as the average annual wage and salary income growth squared in the county during the selected timeframe(s)), and wage and salary income risk (WSI-Risk; the standard deviation of annual wage and salary income growth during the selected timeframe). The WSI portfolios from 1996–2005, 1995–2005, 1991–2005, 1991–2004, 1992–2005, 1993–2005, and 1997–2006 were found to be significant for new business survival in 2010. Similar to before, the observed impact was consistently visible in the metro counties, and all three measures, WSI-Risk, WSI-Growth, and WSI-Growth Squared, were found significant.

The results indicate that risk and return trade-off in local labor markets, identified through employment-based and income-based measures, is instrumental in the survival of new businesses. This impact was found to be strong, especially in metro counties, which encouraged us to focus the investigation in Metropolitan Statistical Areas (MSAs; identified using the crosswalk file from NBER). The existence of a non-linear relationship between risk and growth variables was also verified in MSAs. Similar to counties, the existence of a U-shaped relationship was found in MSAs also (See Table A1 in Appendix A, with results of the MLE tests in MSAs).

The validity of these results could potentially be challenged due to the impact of the Great Recession on the local economy. The influence of this macroeconomic shock cannot be ignored, though considerable heterogeneity was observed in its impact on regions across the country.

A widely used practice in econometric analysis is to replicate the empirical exercise in a different period. To validate the results, the exercise was replicated to determine the impact of local labor market dynamics on the new business survival rate in 2005. A similarity that encouraged the selection of new business survival rate in 2005 was found in the fact that the new businesses that operated during this period also had to go through the recession in 2001, just like those that survived in 2010.

To test the robustness of the results, the empirical exercise was replicated in MSAs only, using both employment-based and income-based risk and return portfolios separately to gauge their impact on new business survival rates in 2005 (three regional control variables (Bartik shock, county/MSA income growth 2000–2007, county/MSA employment growth 2000–2007) were dropped when running regressions, as historical data were not available for them for this timeframe) and 2010.

Similar to the results found for counties, income-based risk and return (WSI) measures were consistently found to have a significant impact on new business survival in 2010 in the MSAs as well. WSI-Risk and WSI-Growth from 1991–2001, 1991–2005, 1995–2005, 1996–2005, 1996–2006, 1997–2006, 2001–2010, 2000–2009, 1997–2007, 2000–2008, 2002–2007, and 1998–2007 were found to be significant (Table A3 in the Appendix A). WSI-Growth Squared was found to be significant for all these timeframes, except in 2000–2009 and 2000–2008.

Next, we replicate this exercise in order to specifically investigate the impacts of employment-based and income-based measures on new business survival rates in MSAs in 2005. The regression results reveal that the WSE-Risk related to employment-based measures in 1991–1999, 1991–2000, 1992–2000, 1993–2000, 1991–2001, 1991–2004, 1991–2005, 1993–2005, 1993–2003, and 1993–2004 had a consistent and significant impact on the new business survival rates in 2005 (Table A4 in the Appendix A). WSE-Growth and WSE-Growth Squared were not found to have any impact during these periods. Income-based risk and return measures were also found to have no impact on the new business survival in 2005.
