26
also included. The basic message is that a rise in unemployment and a decline in on-the-job
search are associated with a rise in the relative number of temporary new hires.
Increased volatility in productivity or product demand conditions is another possible reason
why firms may have become more prone to use fixed-term contracts. Hiring labour on a fixed-
term contract can be seen as means to accommodate a more volatile market environment by
shifting some of the increased risk to the worker. “Lean production”, probably in part fuelled by
the IT revolution, may carry with it a stronger desire to economize on inventories of both goods
and labour. This possibility seems plausible, but it is difficult to find hard evidence on increased
volatility. Data on job turnover in Sweden do not suggest marked changes during the first half of
the 1990s, except for sharp increases in job destruction during the trough of the recession in
1992-93.17
All things considered, we argue that a compelling explanation of the rise in temporary work
in Sweden should focus on how depressed product and labour market conditions impact on
firms’ incentives to offer temporary and permanent contracts as well as workers’ incentives to
accept these offers. Although this “business cycle explanation” may not be the whole story, it is
most likely an important part of the story. Indeed, the evolutions of labour market conditions and
temporary work in the other Nordic countries add further support for this claim.
During the 1990s, unemployment remained relatively stable in Denmark and Norway while it
skyrocketed in Finland. The Finnish unemployment rate stood at around 3 percent in 1990 and
had risen to 17 percent in 1994. During the second half of the decade it declined and had reached
10 percent by the end of the decade. (OECD, 2000.) What happened, then, to temporary work in
these countries? The share of temporary work has been relatively stable in Denmark and
Norway, hovering around 10-12 percent of total employment in Denmark and around 10 percent
force surveys); we interpolated by means of a three-period moving average. The estimation period is 1989.2-1998.3.
Adjusted R-sq.=0.93 and DW=2.1. The Breusch-Godfrey LM test for serial correlation has a p-value of 0.5.