Sectoral differences in the use of information technology and matching efficiency in the US labour market

Abstract

The present study examines how the heterogeneity of use of information technology in production affects the probability that an unemployed worker will be matched with a vacancy. Using US time series from 1967 to 2007, I construct measures of dispersion of the stocks of software and hardware per worker across 13 industries. The measures exhibit three waves whose timing roughly corresponds to the diffusion of mainframe computers in the 1960s and 1970s, personal computers in the 1980s and the Internet in the late 1990s. After controlling for other influences, I find that the probability of transitioning from unemployment to employment responds negatively to an increase in either measure. The results imply that by enhancing technical heterogeneity, the diffusion of a new technology may suppress the job finding rate.

Keywords:

• matching function
• information technology
• diffusion
• unemployment

JEL Classification:

• E24
• O33
• O51

Notes

¹ For examples, see Griliches (1957), Mansfield (1961), Gort and Klepper (1982), Caselli and Coleman (2001) and Manuelli and Seshadri (2003). For a discussion see Nelson (1981).

²  For an example on how the incomplete adoption of IT enhanced heterogeneity in production technology in the Chilean retail sector, see Devries and Koetter (2011).

³ The question of whether the overall use of IT, and particularly of the Internet, has improved the efficiency of matching is addressed by Kuhn and Skuterud (2004) and Kroft and Pope (2014). Both studies find no positive impact.

⁴ For a nice survey of the relevant literature, see Vivarelli (2014).

⁵ Examples include, but are not limited to, Elsby et al. (2010) and Shiyan et al. (2012).

⁶ This structure is consistent with the ARMA(3,3) in their article, which uses monthly data.

⁷ Details are also available on Shimer’s webpage: http://home.uchicago.edu/~shimer/data/flows/

⁸ To extract the trend, I use the Hodrick and Prescott (1997) filter.

⁹ For instance, suppose that N sectors start with one unit of capital each and that to adopt the technology, they must invest in a second unit. Let s be the share of sectors that have undertaken the investment. The numerator becomes . When the share is zero or one, this expression is zero. Moreover, standard calculus shows that the expression is maximized when .

¹⁰ I also tried including a dummy for the first oil shock, but it was not statistically significant.