Our experiments revealed several remarkable features of markets with incomplete con- tracts that experience productivity shocks. First, markets are extremely stable in terms of employment despite the productivity shocks and despite the prediction of full unem- ployment based on standard assumptions. Wages of workers go down considerably, but effort levels fall only slightly, on average, when productivity plummets. There are small effects of productivity shocks on the functioning of the markets such as shorter average firm-worker relationships and slightly lower wages and effort levels than in the control conditions, but the effects are much less severe than predicted by the standard model. Two main features of our design – entirely flexible wages and the possibility to build up reputation over time – could be the driving force behind the stability of the markets. Im- plementing two control treatments that systematically drop these features in the design allows us to conclude that it is definitely not the wage flexibility that is responsible for the functioning of the markets. Fixing the wage in the contract offer on a comparably high level does neither lead to higher levels of unemployment, nor does it create any
other frictions. On the contrary, it leads to even higher levels of efficiency on the market through higher levels of effort provided than in our main treatment. Wage rigidity should, therefore, not be viewed as negative per se; on markets with incomplete contracts and instability of productivity or profits, it could even be beneficial for certain ranges of wage levels.
In contrast to wage rigidity, taking away the possibility to form relational contracts over time makes it harder for firms and workers to overcome the negative effects of productivity shocks on the markets. Even though unemployment is still by far lower than predicted by the standard model, it is significantly higher than in all other conditions that we have analyzed. The negative effect becomes particularly apparent in phases with low produc- tivity. Furthermore, efficiency is lowest in the treatment without reputation formation. One interesting finding from our analysis is that workers exert more effort in phases of low productivity than in phases of high productivity per unit of wage. Remarkably, effort levels are highest in our control treatment with constant low productivity. Such a behav- ior is completely in line with models of other-regarding preferences, but it is obviously not efficiency-maximizing. From an efficiency perspective it would be optimal to shift effort provision to high productivity phases. Apparently, relational contracts are not strong or stable enough to allow for such an intertemporal rationale.
Nevertheless, there are important intertemporal aspects in our data that are able to ex- plain the well-functioning of the market with productivity shocks. First, relational capital that is accumulated through reciprocal behavior between firms and workers over time al- lows for a smoother transition from GT to BT. Indeed, firms are able to pay lower wages to workers in the first period of the BT-phase that have already been employed in GT. Hence, relational contracts seem to allow for a higher degree of wage flexibility within the firm compared to the market outside the firm. Second, while there is no rigorous test available because of the endogeneity inherent to the interaction, our result indicates that (i) the mere fact of receiving a (private) offer is already considered to be a kind signal that is reciprocated by many workers, and (ii) that it is not the workers who are initiating the sustaining of the relationship in the transition from GT to BT through signaling with high levels of effort, but the firms who keep on making attractive offers when entering BT. Third, we find supporting evidence for intertemporal reciprocity over regime switches
from GT- to BT-periods by looking at the relationship between the wage-effort ratio and the effort response in the transition. Highly reciprocal relationships during GT display smaller decreases or even increases in effort levels from the last GT-period to the first BT-period. Hence, it is safe to conclude that workers take intertemporal considerations into account, but they usually react on contract offers and do not signal intentions in the last GT-period.
The experimental evidence that we obtained so far appears not fully consistent with the results by Bewley (1999) based on interviews of executives about their wage and employ- ment policy during the 1991-recession. Bewley (1999) stresses the importance of ”morale“ to keep workers’ propensity to exert costly effort high, especially during recessions. For this reason, he concludes that executives consider a wage cut to be a severe hazard to a worker’s morale and thus refrain from cutting wages during a recession, but rather lay the least productive workers off ”to get the problem out of the door all at once“ when productivity of effort is low.
In our setup so far, we restricted firms to hire a maximum of one worker such that it is not possible to disentangle these two motives of firms from each other. If firms are able to hire more than one worker and when workers are differently productive, they obviously have two ways of reacting to a productivity shock: reducing wages of the existing workforce or cutting employment through lay-offs. In order to investigate Bewley’s main hypothesis in our setup and to assess the wage policy of firms vis-á-vis its employment policy when hit by productivity shocks, we extend our setup to firms that can hire up to two workers.