Köyhyyden dynamiikka Suomessa 1995–2008
A large register based Finnish panel data set in 1995–2008 is used to examine the dynamics of income poverty. Poverty entry and exit rates are estimated using a Markov chain which is based on a dynamic Probit model. To study the evolution of poverty dynamics the observation period has been divided into three five-year long panels. In addition, the study examines how the results change when various age groups are used. The main interest lies in (structural) state dependence in poverty and the effects of dynamic changes in family size, status and structure. In addition, the paper reports correlations between individual heterogeneity and fixed effects due to age, gender, educational level and sosioeconomic status. There are two different explanations for persistence of poverty spells, first, true (structural) state dependence and second, spurious correlation due to unobservable individual heterogeneity. The results help to distinguish between these explanations, which is important for the design of policy measures. The study showed interesting differences across the range of dynamic estimators considered. The random effects estimator had a particularly poor performance in uncovering poverty dynamics.