The Dynamics of Unemployment, Temporary and Permanent Employment in Italy

Abstract

In recent years, temporary employment has risen in almost all European countries, and atypical contracts have been used to obtain labour market flexibility and to respond to the high level of European unemployment. The standard arrangement of Italian employees has traditionally been full-time, permanent, and characterized by a high degree of employment protection. However, recent labour market reforms1 have changed the institutional set-up, and atypical employment forms, among them temporary contracts, have been growing in importance.

As part of my PhD thesis, this study has been financed by the Marche Polytechnic University doctoral fellowship.

Appendices

Appendix

Estimation of the Average Partial Effects

In the analysis reported, the APEs, or marginal effects, were estimated following Stewart (2007): I estimated the counter-factual outcome probabilities evaluated at the sample means, $\bar{x}_{1} .$

Let us call $p_{U} ,p_{P} ,$ and $p_{T}$ the probability of permanent employment given, respectively, unemployment, permanent employment, and temporary employment at time $t - 1.$ Hence, on performing the Wooldridge’s analysis, I had that

$$ \hat{p}_{P} = \frac{1}{N}\sum\limits_{i = 1}^{N} \Upphi \left[\bar{x}_{1} \hat{\beta }_{1} + \hat{\rho }_{11} + \hat{c}_{1i} \right ],\quad \hat{p}_{U} = \frac{1}{N}\sum\limits_{i = 1}^{N} \Upphi \left[\bar{x}_{1} \hat{\beta }_{1} + \hat{\rho }_{12} + \hat{c}_{1i}\right ],$$

and

$$\hat{p}_{T} = \frac{1}{N}\sum\limits_{i = 1}^{N} {\Upphi [\bar{x}_{1} \hat{\beta }_{1} + \hat{c}_{1i} ]} ,$$

where $\hat{c}_{1i}$ is the estimated linear approximation (7.5) of unobserved heterogeneity. When I adopted the Heckman’s estimator, the outcome probabilities had to be rescaled because of a different normalisation, so that the arguments of the standard normal c.d.f were multiplied by $(1 - \hat{\lambda })^{1/2} .$ Moreover $\hat{c}_{1i}$ is, in this case, the estimated correlated random effect according to specification (7.3).

Hence, the estimated APEs were defined as $A\hat{P}E_{H}^{J} = \hat{p}_{J} - \hat{p}_{H}$ with J,H = U,P,T. For example, $A\hat{P}E_{U}^{T}$ was the effect of a temporary job at time $t - 1,$ rather than unemployment, on the probability of being a permanent worker at time t.