In lieu of an abstract, here is a brief excerpt of the content:

Economia 4.1 (2003) 208-220

[Access article in PDF]


[Return to Article]

William F. Maloney: Gustavo Gonzaga's paper on the impact of reforms on labor turnover addresses a topic that is receiving increasing attention in the literature. Turnover in Latin America is high--Gonzaga cites a recent World Bank report that went so far as to term the Brazilian labor market hyperactive. Such high turnover, if somehow exogenously induced, may have adverse impacts on the accumulation of human capital. And if labor market institutions are partially responsible, it is important to determine how large an impact reform can have.

The conceptual development of the paper suggests the potentially ambiguous effect of the increased penalty of the FGTS. Gonzaga tells a good story, which I've had confirmed from other observers of the Brazilian labor market, that there are incentives for workers to get themselves fired and for employers to agree to do so. The incentive to get "fired" arises because it's the only way, short of buying a house or retiring, to get access to the FGTS account. A worker may want access to the funds, for example, to start a microenterprise or simply because the rate of return in the government program is low. As Paes de Barros, Corseuil, and Gonzaga note, some 62 percent of those who declare that they quit get FGTS, which legally should not be the case. 1 This suggests that they quit in such a way as to get fired. Increasing the penalty clearly increases the firing cost, and it should reduce the firms' desire to engage in such a negotiated firing. But it should also increase the desire of workers to get fired and hence to take actions that will ensure it. The net effect on turnover and labor market rigidity is theoretically ambiguous. The contribution of this paper is thus identifying empirically whether there is an impact and, if so, which effect dominates.

My first query concerns the reliability of the proxy for turnover. Gonzaga uses the same measure as Paes de Barros, Corseuil, and Gonzaga, which is based on asking people who are presently unemployed about the [End Page 208] tenure of their last job. The analysis implicitly assumes that duration of employment is independent of being observed quitting or getting fired, but this may merit a closer look. Might these unemployed be intrinsically more prone to quit or be fired than the average worker? Will their response to a change in legislation be representative? We would like to know more about the motivations of this group, how they are special, and what they go on to do later. For instance, suppose that those workers who are predisposed to open a microenterprise, but are credit constrained, are more likely to get themselves fired. Their response to an increase in the premium to getting fired may be higher than for average workers who are more risk averse and less credit constrained.

To generate an alternative measure of turnover, Mariano Bosch and I tried to determine duration based on labor market transitions. We assumed a time-homogeneous Markov process Xtof transitions among K employment states, the matrix P(t) is a discrete time-transition matrix where


and where Q is a KxK matrix of instantaneous transition intensities. Duration in the sector can be measured as

where qii is instantaneous probability of not moving from the current sector. The intuition is that the mean duration in a sector is inversely related to the probability of a worker leaving the sector. 2

Figure 6 presents three estimates calculated as the implicit duration of the previous job using periods 1 and 2, when the worker is observed in the formal sector in period 1 and in either the formal sector, unemployment (the closest analogy to the survey questions used in the paper), or any sector at all in period 3. The estimates conditional on being unemployed show the highest turnover, and they are more or less consistent with Gonzaga's estimates. However, the mean tenure of those found in the...


Additional Information

Print ISSN
pp. 208-220
Launched on MUSE
Open Access
Back To Top

This website uses cookies to ensure you get the best experience on our website. Without cookies your experience may not be seamless.