Temporary Work in Turbulent Times: The Swedish Experience

16

seasonal dummies. The period chosen for the estimation of the trend is 1975.1 - 1990.1. The
equation is used to predict trend GDP values for the period 1987.2 - 2000.2. (See Holmlund and
Vejsiu (2001) for more details.) The estimated transition rate models are parsimoniously
specified. We always include a constant and three seasonal dummies. Other variables considered
were lagged YDEV and changes in YDEV (∆YDEV), one lag on the dependent variable (to
capture sluggish adjustment), a linear time trend and a dummy for the first quarter of 1993
(capturing effects of changes in measurement methods). Insignificant variables among the group
of “others” were dropped.

An appendix available on request presents estimation results for transitions between four
states: P, T, U and O (thus excluding self-employment). A second set of estimations includes all
states but with P and S aggregated into one state, denoted P* and (also) referred to as
‘permanent’ employment. We have already noted that P and S are fairly similar in terms of the
duration of the spells. Moreover, it turns out that transitions involving self-employment are more
difficult to predict than other transitions. The four states are thus P*, T, U and O, the sum of
which is the total population aged 16-64. The estimated equations are displayed in Table 4.1.
Significant cyclical effects are always present except for transitions from permanent employment
to nonparticipation. Positive and significant trend coefficients are estimated for transitions to
temporary work from permanent employment and nonparticipation. ¹⁰ Negative and significant
trend coefficients are estimated for transitions from both states of nonemployment to permanent
employment.

¹⁰ The estimated trend coefficients will obviously be sensitive to how trend GDP is measured. Holmlund and Vejsiu
(2001) report some results of checks for robustness.

<<   15   16   17   18   19   20   21   >>

More intriguing information