How does the age structure affect regional productivity?

Abstract

The article studies the effects of a changing age structure on regional productivity using lagged population structure to purge estimates of the influence of endogenous migration. It is shown that workers aged 50–60 years have a positive effect on regional productivity. Together with evidence from previous research showing that the effect of the same age group on unemployment is positive, the results support the notion that well-matched workers gives rise to both high productivity and high unemployment.

Acknowledgements

I am grateful to Max Halvarsson for excellent research assistance and for comments on previous versions by Thomas Lindh and participants at the ‘Productivity Workshop’ in May 2005 at the Institute for Future Studies. The data on productivity were provided by the Institute for Future Studies who also funded the research presented in this article through the project ‘The Impact of Population Ageing on Innovation and Productivity Growth in Europe’.

Notes

¹See e.g. Haltiwanger et al. (1999), Crepon et al. (2003), Hellerstein and Neumark (1995), Hellerstein and Neumark (2004), and Lindh et al. (2005) for studies on firm or plant level data.

² It should be noted that the model fail to account for the possibility of different effects of the age distribution of males and females. The reason is that this issue cannot be studied using the current identification strategy: our instrument is very close to birth rates, and since the number of male and female births covary almost perfectly, we cannot identify their effects separately.

³ See Fair and Dominguez (1991) for the transformations. The SEs are calculated directly from Equation 5(5) after estimation of the parameters a to f using the covariance matrix of these estimates; this is possible since the j:s of the polynomial restriction are nonstochastic. Thus: .

⁴ The data on productivity comes from ‘Industristatistiken’, a data source that stopped being produced after 1996. Productivity data for earlier years do exist, although they are of somewhat lower quality. However, it is not possible to generate the instrument for earlier years.

⁵ The ‘stable plants’ data do not include any observations for the regions of Åsele and Arjeplog, which reduces the sample size somewhat. However, given that we use weighted regressions and these regions are tiny, this should not be a major concern.