Age and entrepreneurship: nuances from entrepreneur types and generation effects

Using Health and Retirement Study data

The new entrepreneur rate by Fairlie et al. (2016) captures all new business owners, including those who own incorporated or unincorporated businesses and those who are employers or non-employers, using the US Current Population Survey data.

See https://​cps.​ipums.​org/​cps/​

See https://​www.​bls.​gov/​lau/​

For example, individuals who are interviewed in January, February, March, and April of one year are interviewed again in the next January, February, March, and April.

Considering the fact the local economic condition might have spatial influence or autocorrelation from contiguous local areas’ economic conditions, as addressed in Santarelli et al. (2009), spatial modeling was initially considered. However, for three reasons, we did not think it necessary in this study: (1) our cluster-level unit is in metropolitan areas, which are not mostly contiguous geographically. Without contiguity, the spatial interdependence is limited. (2) A metropolitan area is a commuting circle in which residents and commuters share the urban centers and socioeconomic atmosphere, rather than sharing those in another metropolitan area some distance away. This differs from other geographic units that are arbitrarily determined by political (such as county) or population size boundaries (such as census blocks). (3) When facing a noncontiguous geographic unit, one needs to use a distance matrix to measure spatial associations that typically assume Euclidian distance between centroids of metropolitan areas. This hypothetical centroid approximation is not a good representation of the urban core, and the distance-based measure of influence from another metropolitan area is further compromised by size of the metropolitan areas. (4) Our basic unit of analysis is fixed at individual level and the majority of variation across our observations is at the individual level, not at a geographic area level; thus, spatial interdependence is less of a concern.

This linear relationship makes it impossible to separately estimate those three variables using conventional linear regression models without adding constraints to identify the model. The generation construct is thus often plagued by methodological problems with inextricably intertwined constructs of age, period (specific historical time periods), and cohort (group with shared experiences) (Parry and Urwin 2011; Costanza and Finkelstein 2015).

We do not include the quadratic age term in the model because we estimate Model (2) with only 8 years of age for each pair of neighboring generations. This is explained later under.

Among others, including Evans and Leighton (1989), Kautonen et al. (2014), and Zissimopoulos and Karoly (2007)

According to Fairlie et al. (2016), the U.S. Census Bureau notes that the definitions of non-employers and self-employed business owners are not the same; although most self-employed business owners are non-employers, about a million self-employed business owners are classified as employer businesses.

Florida’s (2004) “creative class” occupations include sectors of management, business and financial operations, computer and mathematical, architecture and engineering, science, law, education, arts and media, healthcare practitioners, and high-level sales management.

CPS does not offer detailed information on an individual’s health status. Although Health and Retirement Studies offer detailed information on health, this dataset lacks information on monthly employment that is key to this study.

Those who did not report their educational attainment information or attained less than high school degrees were the omitted category.

The better economic years including the pre-recession 2006 and 2007 and growth year 2016 are the omitted.

Note we only study knowledge-based non-agricultural sector workers, including entrepreneurs.

62, 61, and 54 are industry sector codes according to the North American Industry Classification System (see https://​www.​census.​gov/​eos/​www/​naics/​).