For post-communist economies, this model is not appealing for three reasons. First, there is a great deal of anecdotal and survey evidence indicating that firms can vary greatly the intensity with which they use labor. Standing (1994) documents what he calls "hidden unemployment" in several forms. Hours worked can be reduced through closing the factory for some period, through sending people on forced vacation (paid or unpaid), and through offering extended maternity leave. A person can also be paid a lower wage, either relative to others or in absolute terms, with it being understood on both sides that this implies a lower level of time commitment to the firm. Second, individuals can also respond to changes in their formal employment by engaging in various "survival strategies," or alternative ways of earning income, primarily through informal self-employment. In principle, an individual can vary his or her intensity of work in these survival activities over an almost continuous scale. Third, it seems inappropriate to assume that firms make one formal employment decision per person, and each individual then decides how to respond. Rather, the interaction between most individuals and their original employer is repeated, with both sides taking into account both the previous and expected future actions of the other. For example, if the managers of a firm see an individual heavily engaged in survival strategies, they may consider cutting back on that individual's work in the firm, although not necessarily firing him or her. The individual may respond by putting more effort into other income-earning opportunities, but not necessarily quitting.

Our empirical work is guided by a theoretical framework that modifies the standard model to take these three considerations into account. We model the individual's decision as a stochastic dynamic supermodular optimization problem. This problem has two critical assumptions. First, we require that managers choose the intensity with which each individual works in the firm, while the individual chooses the intensity with which he or she works outside the firm. As the manager reduces the intensity of firm work, there is a higher payoff to the individual from increasing the intensity of outside work. The converse may also be true: if the individual works more outside the firm, the manager derives a higher payoff from reducing the intensity of work for this person. The second critical assumption is that larger adjustments are more costly than smaller ones for both firms and individuals. This slows down the optimal speed of adjustment on both sides and means that optimal intensities of work adjust over time (although the optimal speed can vary considerably across individuals).

These two assumptions describe a supermodular optimization problem. Static versions of supermodular problems can be analyzed using the tools provided by Milgrom and Roberts (1990), but this approach does not capture important aspects of transitional economies. Instead, we use the results established by Friedman and Johnson (1996), which show that in the presence of complementarities and convex-type adjustment costs, the dynamic adjustment