Measuring the effect of spell recurrence on poverty dynamics – evidence from Spain

Arranz, J.M. and Cantó, O. (2012), «Measuring the effect of spell recurrence on poverty dynamics – evidence from Spain«, The Journal of Economic Inequality,10 (2), 191-217 (DOI: 10.1007/s10888-011-9191-2).

This paper has analysed the effect of spell recurrence on poverty dynamics taking into account multiple poverty and non-poverty spells (the complete poverty history) by spell order while controlling for initial conditions, household characteristics and individual unobserved heterogeneity. Using ECHP data we have estimated a mixed proportional hazard model with multiple states and multiple spells in order to provide empirical evidence on different poverty exit and re-entry hazards when spells accumulate, challenging previous studies based on poverty persistence that estimate one exit and one re-entry hazard rate independent of the number and duration of individual poverty experiences.

In general, our findings highlight the importance of considering spell order in the analysis of poverty dynamics given that lagged poverty and non-poverty durations have a significant effect on both exit and re-entry probabilities even when controlling for relevant covariates, initial conditions and unobserved heterogeneity. In particular, lagged poverty duration reduces the probability of leaving a poverty spell and lagged non-poverty duration decreases the probability of re-entering poverty in a second non-poverty spell. These results are robust to dropping left-censored poverty spells and estimating transition probabilities for a sample of entrants to poverty within the observation window, a key comparison in the analysis of transition rates at different poverty spell durations.

Also, the highest poverty exit rates are associated to individuals with shorter durations in poverty, who have a large number of earners in their household. Therefore, in a heterogeneous context controlling for relevant household characteristics and lagged poverty durations we only find some negative duration dependence after the poverty or non-poverty spell has evolved for two or 3 years. In the particular case of second poverty spells results show a strong positive duration dependence during the first 2 years of duration, which becomes largely negative thereafter. These results indicate that policymakers in Spain should be particularly aware of the large weight of recurrent transitory poverty in this country. Also, it is important to note the significantly higher chances to leave poverty of those individuals who did not experience any previous poverty spell in comparison with those that already had one. Short-term labour market policies promoting employment stability (as long as the wage level allows individuals to avoid poverty and that employment stability does not reduce the poverty exit rate by promoting persistent in-work poverty) together with short-period monetary transfers to avoid a first fall into poverty, appear to be particularly adequate interventions in Spain to reduce poverty incidence.

Interestingly, our multivariate regressions suggest that the effect of a covariate on poverty exit and re-entry is often the opposite. Thus, characteristics that help individuals in leaving poverty also help them in avoiding recurrence. Additionally, the estimated coefficients capturing the effect of covariates on exit and re-entry hazard rates change in magnitude and significance when we separate spells by their order. Household composition turns out to have a particularly strong effect on reentries, especially in second non-poverty spells. Households that are worst off, are couples with three or more children. They have a lower probability of stepping out of poverty and a higher probability of re-entering it after exit. Consequently, they are more likely to suffer long-term poverty. In contrast, individuals in one person households, single parents and couples with no children or just one child have a significantly lower probability of re-entering poverty once they have managed to step out of it.

In any case, we claim that more research is needed in order to check if our results on left-censored spells and on the effect of lagged poverty duration on poverty transitions for Spain can be generalized to other countries. Also, it would be interesting to know if they still hold in the case that individual unobserved heterogeneity is not assumed to be constant over time and exogenous to the process of a transition into or out of poverty. Indeed, some of our results may depend on our particular data given that Spain is a country with a relatively large percentage of short-term poor households in the population relative to European countries.

El lector interesado puede encontrar el artículo completo aquí:

http://link.springer.com/article/10.1007%2Fs10888-011-9191-2?LI=true