Bargaining Structure and Bargaining Outcomes

Abstract

<p>A recent development in the analysis of strikes and contract negotiations strategic bargaining models with asymmetric information -- allows us to study bargaining structure and outcomes via its effects on information transfer and learning among parties to bargaining. This thesis continues this new approach and attempts to add to our knowledge of bargaining structure, both theoretically and empirically.</p> <p>The whole thesis can be viewed as three main essays. In the first main essay(Chapter 2), we study learning and information transfer among unions when negotiations are sequential and there is no collusion among either unions or firms. That chapter attempts to further our understanding of relative rewards, imitation and learning. The model, which considers two union-firm bargaining pairs, generates an interest by workers in each other's wages which is based on learning their own firm's ability to pay by observing the preceding negotiations. But rather than being socially harmful, as it can be in the "informational cascade" literature, learning from actions of the others is, in a number of cases, socially beneficial. This is because learning reduces the costly mistakes made in bargaining due to asymmetric information. Using a large sample of Canadian contract negotiations for the period from 1965-1988, we find strong evidence that the more negotiations which have been concluded in the recent past in a union's industry, the less likely is a strike to occur. This can be seen as relatively convincing evidence that some social learning, with beneficial social consequences, does occur among unions negotiating wages within an industry.</p> <p>In the second main essay (Chapter 3), we use a model of learning among unions to compare bargaining outcomes in various bargaining structures and examine the effects of centralization when negotiations are simultaneous. Existing formal models of bargaining structure and outcomes typically ignore one or both of two key issues: the issue of asymmetric information and the nature of bargaining process simultaneous versus sequential negotiation). Among other things, this means that they cannot capture the implicit coordination, or' social learning, in decentralized bargaining structures. Neither can they examine the wage leapfrogging phenomenon that has been suggested as a potential important disadvantage of decentralized bargaining structures. The current model allows us to examine these key issues. We found that when negotiations are simultaneous, collusion by firms or by both firms and unions reduces expected wage settlements and raises strike incidence since they reduce learning and information transfer among unions in contract negotiations.</p> <p>In the model of learning among unions examined in the second main essay, there are clear first mover disadvantages for both unions and firms. Early negotiations generate valuable information about firms' ability to pay which unions in later negotiations can use to improve their wage settlements. Unions have an incentive to free ride and delay their wage settlements and let other unions conclude their negotiations first. In the third main essay (Chapter 4), we examine this information externality and interpret the delaying of wage settlements without strikes as holdouts. As in Cramton and Tracy's model of holdouts (1992), the model predicts that holdouts should be shorter and less frequent when the wage settlement in the existing contract is lower, and when the unions are more optimistic about the firm's ability to pay. But the model also has a number of predictions about some issues on which Cramton and Tracy's model is silent, one of which is the following: as the number of unions in the model expands, the above information externality is exacerbated, generating longer holdouts in equilibrium. This implication is tested using the large sample of Canadian contract negotiations used in the first essay, yielding strong evidence that the larger the number of negotiations taking place at the same time, the greater are both holdout incidence and duration.</p>