Professor Ostrom has put her neophyte reader in a difficult situation. Hers is a mature research program of enormous scope and intellectual yield. However, because of that, it is far too complex to be neatly summarized in an introductory section, and if fully summarized, there would be no further space for a positive contribution to that program. The neophyte has to know a great deal in order to know how inferential issues have been addressed in actual practice. It is in the spirit of ignorance (and awe at the project’s intellectual and field-based range) that I offer some queries that can probably be answered with reference to work already accomplished in the program. But without these queries addressed, some of the force of the enterprise appears weakened.
The big worry for me has to do with selection. Ostrom writes: “We and other scholars have consistently found…that rules developed with considerable input of the resource users themselves [i.e. FMIS] achieve a higher performance rate than systems where the rules are entirely determined by external authorities [i.e. AMIS].” The data in Tables 1 & 2 offer confirmation. But I wonder: on FMIS outperforming AMIS, don’t you need a selection model? Could it be that only those communities that were weak were captured by the state, and would have done even worse if not incorporated by state agencies? In this sense, FMIS could be improving the performance of that set of systems that they have captured.
Consider the research project described in the paper by Yoder and Pradhan. They “took the farmers from a district in Nepal where the farmers had not yet designed very effective rules to a district where the farmers were quite successful in building and maintaining systems with high crop yields. They then had the farmers from the unsuccessful systems spend a day attending the annual meeting of several of the successful systems and listening to the other farmers (who were obviously getting better crop yields) tell how they managed their systems and coped with different kinds of problems.” This is an example that the least effective farmers are getting state attention and intervention. If general, it would undermine the findings reported in Tables 1 and 2. See the reference to Baker’s research for further evidence on this point: “Some of the irrigation systems were unable to cope with these threats [floods, landslides, and in-migration] and stopped operating or were taken over by a governmental agency.”
AMIS institutions may also be sustaining a community that would have collapsed without state intervention, and probably been appropriated by a neighboring community that had developed efficient rules. In the efficient communities, the quasi-parameter of population would have thereby been expanding, putting pressure on members of the community to migrate. In this sense, the state can be seen as the barrier to effective evolutionary learning processes, but not by creating bad rules but by sustaining inefficient communities. These selection issues lead this reader to lose confidence in the results and interpretation of these results in the Ostrom paper.
A second worry I have is in the way game theory is presented. Ostrom writes: “Most game-theoretical analyses rely on a highly simplified theory of human behavior that has proved itself useful in predicting behavior in competitive situations.” It is better to say that game theory, by capturing the nub of a class of transactions (i.e. taking out as much texture that is assumed to be irrelevant for behavior in those transactions), allows for a clearer understanding of the calculations rational actors need to make in order to maximize their utility, thereby giving a benchmark for rational behavior, with deviations requiring some (external to the game) account. Game theory itself, despite generations of Nash refinements, cannot in most cases solve for a unique equilibrium. And the folk theorem teaches us that equilibrium selection requires theory outside of the model. It is therefore beyond the evidence to claim that game theory is “useful in predicting behavior.”
Furthermore, game theory is presented rather unproblematically as a generator of rule configurations. We are first told that “changing rules within collective-choice arenas” is one of eight evolutionary processes to be analyzed (p. 2). Later on, we get guidance on the role of game theory in analyzing this process. We read (p. 6): “To model a human-interaction as a game, the theorist must decide which components to use from a set of seven working parts of an interaction as well as how the individuals who are interacting will be modeled.” While the list of these seven rule-types that constitute a rule configuration appears to be exhaustive and mutually exclusive, the translation into a game theoretic analysis of the generation of these rules is hardly straightforward. We need to know what sort of game is being played. Is it a principal/agent game, in which an expert is recognized by the community, but the incentives of the expert and the community are not fully aligned? For example, the agent can fulfill the principal’s mandate, or instead can self-appropriate the rents of office. Or is it a pure coordination game, in which there is more than one equilibrium solution that is on the Pareto frontier, but there is no natural way to decide among them? Or is it a game of commitment, where cooperation fails because players cannot be sure of the promises of their counterparts, who have an incentive to defect? And the choice of players is equally problematic. For example, there is no mention in this paper of competing villages whose populations are growing and the threat they pose to the village under study to upgrade their rules. Adding this new actor permits a standard range of conflict games in which, if the likely costs of conflict are high, bargaining theory yields a range of possibilities, from anschluss to colonization. And if bargaining breaks down, war and extermination of the inefficient population are possible outcomes, ones that would not be considered if these competing villages were not in the model.
My point here is this: game theory itself has no rules to tell us the arena of collective choice we are in or who the relevant players are; yet without a principled justification for setting up the game structure, there is no way to analyze the “moves” by farmers as they work to enforce, abide by, or change the rules. All this depends on what the researcher believes to be the nub of the transaction under study. Thus theorizing must precede the modeling exercise.
A final worry I have is in the application of evolutionary theory. I’m unclear throughout whether there is any value added with the “evolutionary” language beyond standard models of institutional learning. The potential value added was not clearly delineated. Suppose evolutionary theory tells us e.g. that the more there is population pressure (and therefore rational fear of conquest) researchers would observe more efficient rules. Or: where there is more variety in weather patterns there will be greater flexibility in rules leading to better institutional performance. These could be tested. Have they been? Perhaps these tests would help distinguish between a rationalist learning model (e.g. the predictions of Bayes’ rule), a stochastic model (e.g. the predictions of a Markov-chain) and one of a set of evolutionary processes. In any event, more clarity is needed as to what is evolutionary about the changes that are observed and recorded in this research program.
I have a few other marginal queries:
1. “Farmers have survived over the centuries in much of Asia due to their knowledge of how to engineer complex irrigation systems including dams, tunnels, and water diversion structures of varying size and complexity.” The counter-factual, that they wouldn’t have survived without them, needs to be demonstrated. Isn’t it possible that farmers would have survived, but populations would not have expanded? Throughout Africa farmers have survived without engineering complex irrigation systems; but they did so under conditions of much lower population density.
2. “Rules are linguistic statements similar to norms but rules carry an additional, assigned sanction if forbidden actions are taken and observed by a monitor.“ Also: breaking rules need not raise community eyebrows if revealed publicly; breaking norms would. Missing is the normative element in norms.
3. “In the field, strategies are observable activities unlike rules and norms.” Not necessarily; strategies entail expectations of what others would do off the path of play; these expectations are not directly observable. In this sense, the complete strategy for any game is never (directly) observed.
4. “Whenever one is interested in understanding processes of structural change of a particular situation itself, however, one has to open up and overtly include one or more of the underlying “exogenous” sets of variables” (p. 7). In this set there are “the biophysical world”, “the broader community of participants”, and “the rules-in-use.” Maybe this is a vocabulary issue, but in no sense is the community of participants exogenous to the particular situation, as past rules influence who is a member of that community. Similarly with the rules-in-use, which are endogenous to the features of the past rules. As for the biophysical world, we read on p. 13 that it too evolves due to environmental factors. You cannot get rid of endogeneity by assumption.
Despite my queries, the important conclusion about institutional monocropping by international aid agencies as an explanation for their less than stellar success in developing high ROI in irrigation projects seems well-founded and important. And the entire research program, as I note in my opening paragraph, is awesome.