Jack Knight on bargaining and institutions

I just finished reading Jack Knight’s 1992 book ‘Institutions and social conflict’, which I really loved.  I think I’ll be returning to this book a lot in the future, so this is just a preliminary first summary.  Basically the book has three components: first, a very astute survey of the literature in institutional economics; second, an outline of Knight’s own theory of institutions; third, an application of Knight’s approach to a few illustrative cases.  In this post I’m going to ignore the third of these, and very briefly summarise the first two.

Start with Knight’s review of the literature.  (I’ve already discussed some elements of this in a previous post, but I’m going to repeat myself.)  Knight distinguishes two key traditions within institutional economics.  

First, there is the approach that primarily sees institutions in terms of coordination for collective benefit.  This tradition in turn can be divided into analyses that see coordination as intentional – paradigmatically, contractual – and analyses that see coordination as evolved rather than planned.  This latter, evolutionary, subtradition can again be subdivided into three different kinds of evolutionary account: those that emphasise “spontaneous emergence, exchange coordinated by the market, and social selection.”

Second, there is the approach that primarily sees institutions as the result of conflicts between rival interests.  This is the tradition that Knight aspires to extend.

For the purposes of this very brief summary I’m going to ignore the important differences – which Knight very clearly and usefully outlines – between the different ‘cooperative’ subtraditions.  The core point which animates Knight’s critique is that these traditions typically explain the creation of institutions in terms of the collective benefits the institutions generate for their members.  For these traditions, institutions are the results of cooperative actions which benefit the cooperating members of the institutions.  Often this account is further elaborated in terms of efficiency, or ‘Pareto optimality’.  For example, the transaction costs tradition exemplified by Oliver Williamson argues that economic institutions are governance structures which are created in order to minimise transaction costs, thereby increasing economic efficiency.

Knight’s core critical point is that these kinds of explanations are fundamentally functionalist ones, which typically do not provide an adequate mechanism to explain why economic efficiency, or collective welfare, or whatever it might be, is selected for in the creation and reproduction of the institution.  More specifically, such functionalist accounts tacitly transgress a methodologically individualist rational choice approach, because they explain institution formation, stability, and transformation in terms of the collective outcomes associated with the institution, rather than in terms of individual social actors’ rational maximisation of their own individual outcomes.

Knight’s alternative approach, is to explain institution formation – in classical game-theoretic, rational choice terms – as a result of individuals attempting to maximise their individual utility.  More specifically, he aims to do this using bargaining theory. An economic institution may be constructed to (say) minimise transaction costs – but it may also be constructed to serve the interests of those with the power to implement and enforce the institution’s rules. Moreover, if we want to explain why an institution is structured in the way it is, Knight argues, we have a methodological obligation to route our explanation via the interests of the various actors involved in the institution’s creation and reproduction, rather than via a ‘collective interest’ that may not be embodied in any individual actor’s actual interests. In a sense, this is a project that shares many similarities with Buchanan’s ‘politics without romance’ (although there are very important differences between Knight’s and Buchanan’s projects, which maybe I’ll one day speak to in a future post!)

To make this slightly more concrete, consider the classic ‘prisoner’s dilemma’ game.  As is well known, in a one-shot version of the prisoner’s dilemma, cooperation is a strictly dominated strategy, and mutual defection is the only equilibrium outcome.  In an iterated version of the game, however, this is not necessarily the case.  In the right conditions, mutual cooperation can be a stable equilibrium within an iterated prisoner’s dilemma, because players can adopt conditionally cooperative strategies which result in actual cooperative play.  More strongly, the so-called ‘folk theorem’ shows that a very wide variety of possible outcomes are potential equilibria.  If you have a correctly set up iterated game then countless payoff outcomes are in principle possible, because countless different equilibria can in principle be constructed.

Iterated game theory in this sense ‘proves too much’.  If the appeal of classical game theory is meant to be that it shows what strategies rational actors are likely to adopt, then iterated game theory removes a huge amount of that apparent predictive power by showing that rational actors could in principle adopt countless different strategies.  The ‘cost’ of showing how cooperation is possible within scenarios like the prisoner’s dilemma is the loss of nearly all the predictive power of the model.

Assuming we’re willing to roll with this, and are still interested in pursuing this kind of modelling approach at all, how can we narrow down what strategies we would expect our agents to actually adopt, without resorting to completely ad hoc reasoning?  One way to tackle this problem is by using an evolutionary approach, pioneered by Axelrod.  In his 1984 ‘the evolution of cooperation’, Axelrod ran a ‘tournament’ in which different strategies played against each other, and those strategies which received the highest payoffs overall ‘won’.  This broad approach can be extended in quite complex ways.  For example, in my own PhD research in the economics of science (drawing very very heavily on previous researchers!) I ran a set of evolutionary simulations of agents playing an n-player prisoner’s dilemma game within a structured graph (network), to look at the impact the graph structure could have on which strategies were selected for.  There is a lot of work in this space – some of it very interesting indeed.  The problem, however – or one problem – is that evolutionary analyses only make sense if you are dealing with a system that has evolutionary attributes: selection among variation, some kind of learning process for the dissemination of successful strategies, and (quite likely) the possibility of ‘mutation’ of strategies.  Most objects of social scientific analysis simply do not exist within such systems, or at least not to an extent that can justify the adoption of this kind of evolutionary explanatory framework as our solution to the underdetermination of possible cooperative equilibria.

Knight takes a different approach.  Knight’s argument (and here again we’re using the prisoner’s dilemma purely for illustrative purposes) is that because countless different payoffs are possible within an appropriately iterated prisoner’s dilemma, agents who are attempting to determine which strategies to adopt within the dilemma are in practice engaged in a bargaining process.  That is to say, the prisoner’s dilemma can in principle provide both players with zero payoff, or whatever the maximum payoff is, or (if appropriately iterated) anywhere in between, for either player.  In selecting their strategies, then, the players are in effect bargaining over the allocation of the potential surplus associated with consistently cooperative play.  This is a bargaining problem.  And bargaining problems are explored in their own very rich branch of classical game theory.  If we adopt a bargaining theory approach to the analysis of the selection of strategies within a cooperative problem, we are therefore potentially in a position to explain why specific strategies are selected without resorting to either evolutionary models or ad hoc assumptions.

Knight calls this a ‘distributive’ approach, because its primary focus is bargaining between social actors for the distribution of the surplus associated with cooperative play.  I think that Knight’s framing of this idea in terms of a distinction between ‘cooperative’ and ‘distributive’ approaches risks misleading readers.  After all, players bargaining over the allocation of a cooperative game’s surplus are also bargaining over whether or not to create the surplus in the first place.  I think it’s better to simply think of Knight’s approach as a ‘bargaining’ one.

But this is a semantic quibble.  At base I think that Knight’s suggestion here – that the problem of the underdetermination of equilibria in iterated games can be partially resolved by seeing the selection between equilibria as a bargaining problem, to be analysed using the tools of bargaining theory – is a major contribution.  I also think that Knight’s approach is correct to emphasise the importance of bargaining theory more broadly in formal political-economic modelling – hopefully I’ll speak to that more in future blog posts.

In any case, there is an irony here, concerning the connection between the dominant methodological and substantive commitments of economics as a discipline.  It is common on the left to criticise mainstream economics for two apparently related failings: on the one hand, its use of rational choice, methodologically individualist approaches is seen as methodologically reactionary; on the other hand, its conclusions about the efficiency and social optimality of capitalist economic institutions is seen as apologistic.  What Knight’s book does, in my view, is show that these two dimensions of the economic tradition are – at least on plausible assumptions – in fact in tension.  If we take rational choice methodological individualism seriously at the level of the analysis of political-economic institutions, using the resources of bargaining theory, it is not hard to produce analysis that shows domination of the relatively powerless by the relatively powerful, highly sub-optimal outcomes in terms of economic efficiency enacted and enforced by powerful economic actors because of their distributive benefits for narrow powerful interests, and so on.  In other words, Knight’s rational choice methodologically individualist approach sends us barrelling very quickly into an at least Marxist-adjacent world of domination, exploitation, inequality, etc.

I think this is all correct, and important.  There is, of course, much more to say about Knight’s approach – as I say, I expect to come back to this broad framework a lot in the coming years and months.  But this will do for now, I think, as a quick summary of the broad outlines of Knight’s very valuable project.

Advertisement