Social Assistance dynamics in Sweden: Duration dependence and heterogeneity

This article uses data on all persons who ever received Social Assistance (SA) in Sweden 1991–2007 (N = 2,638,681 observations; 882,416 individuals) to study whether there is duration dependence in SA, i.e., whether the probability to remain in SA increases over the individual duration of SA recipiency. The risk of remaining in SA is higher at longer durations, but around half of this risk difference is caused by selection (those with favourable characteristics exit first, while those with higher likelihood of SA remain). This is captured by control variables and by conditioning on SA sequences as a method to control for unobserved heterogeneity. The probability to remain in SA increases with 2–5 percentage points per year during the first five calendar years, implying that duration dependence is substantively but not dramatically important: Nearly 8% stay in SA the fifth year after entry, but only 4% would do so in the absence of duration dependence.

► Question: Is there duration dependence in welfare in Sweden 1991–2007?
► The probability to remain in welfare increases with time on welfare.
► Around half of this increase is caused by selection.
► Duration dependence is substantively but not dramatically important.