the emergence and evolution of collective action: lessons from watershed management in haiti

Pergamon

World Developmenr, Vol. 23, No. 10, 1683-1698, 1995 pp. Copyright 0 1995 Elsevier Science Ltd

Printed in Great Britain. All rights reserved 0305-750x/95 $9.50 + 0.00

The Emergence and Evolution of Collective Action:

Lessons from Watershed Management in Haiti

T. ANDERSON WHITE Inter-American Foundation, Washington, DC, U.S.A.

and C. FORD RUNGE”

University of Minnesota, St. Paul, U.S.A.

Summary. - The practice and theory of collective action is constrained by a dearth of rigorous empirical tests of why and how such institutions emerge and evolve, and under what conditions they can be successful. Empirical analyses of cooperative watershed management in Haiti reveal that, given a conducive environment and political leadership, groups will emerge and survive where a “critical mass” of individuals have practical knowledge of the potential gains from action. Emergence can be constrained in the short run by: (a) landscape factors that affect the potential net economic gain, and (b) sociocultural factors that affect the cost of constructing the new institution.

Normally, observation is the tug that tows the ship of theory; but in an extreme case the theory pulls so hard that the observation yields. (Quine and Ullian, 1970, p. 17)

1. INTRODUCTION

Garret Hardin’s (1968) gloomy prediction - that all commonly managed resources would inevitably end in tragedy - fueled trends encouraging privatiza- tion and discouraging collaborative resource management. His view, and its wide impact, is a prime example of how an a priori theory can overwhelm common observation, and of the potentially immense consequences for policy.

Many studies have now contested the universal applicability of Hardin’s prediction. Policy makers and social scientists show increasing interest in cooperative means to manage common resources (see Blomquist, 1992; Bromley, 1992; Wade, 1988; Tang, 1992; Stevenson, 1991; Uphoff, 1992; McCay and Acheson, 1987; Jodha, 1992; Ostrom, 1990). This interest is reflected in new attempts to strengthen common property systems, to develop voluntary institutions to manage transboundary resources (such as watersheds, ecosystems and global resources), and to promote community-based conservation and the role of civil society in economic development (de Janvry, Sadoulet and Thorbecke, 1993).

Indeed, it is now widely accepted that rather than being an anomaly, voluntary collaboration to manage

common resource problems is not only possible, but widespread (see Ciriacy-Wantrup and Bishop, 1975; Dion, 1986). indeed, surprising levels of cooperation have been found in some empirical examinations that were specifically designed to discourage it (see Bromley and Chapagain, 1984; Marwell and Armes, 1981). Twenty-five years after publication of “The Tragedy of the Commons,” the question is no longer whether decentralized collective action can be effective, but under what circumstances it is appropriate, and how positive synergy between the state, market and civil organizations can most efficiently and fairly supply public goods (Uphoff, 1993).

Researchers have recently made substantial progress in identifying the types of rules and resource conditions correlated with successful action (Ostrom, 1990; Schlager et al., 1994; Wade, 1988). A growing theoretical literature on institutional change is also

* We would first like to express our gratitude to the peasants of Maissade for their cooperation with the study and to the field staff of Save the Children who assisted in dara collection. We thank Hans M. Gregersen and Allen L. Lundgren of the University of Minnesota for their contributions to the research and for their comments on this paper. We also appreciate the financial support of the Inter- American Foundation, which provided funding for the field research, and the USAID, Environment and Natural Resources Policy and Training Project. We also appreciate the comments of Norman Uphoff and three anonymous reviewers. Final rev,- sion accepted: April 27, 1995.

1683

1684 WORLD DEVELOPMENT

evident (see Bardhan, 1989; Hayek, 1960; Hayami and Ruttan, 1985; North, 1990, 1994; Schotter, 1981; Sugden, 1986, 1989; Ostrom, Gardner and Walker, 1994). But despite the strong theoretical base, and advances in understanding the correlates of success, we still do not fully understand why and how collective action institutions emerge in the first place-and survive in the long run (Young, 1993; Ostrom, 1990).’

Much previous research on institutional emergence and evolution has been based on a priori suppositions of atomistic behavior, zero-sum interactions and nar- row human rationality, conducted within socially artificial boundaries (Ostrom, 1990; Uphoff, 1993). These conditions nearly always result in predictions of insufficient cooperation and hence, “tragedy.” Rigorous empirical analyses can help to improve our understanding of institutional emergence and evolution. Progress in this area also hinges on better understanding the role of ideology and beliefs on individual choice, and the actual process of personal and social learning (North, 1994). This article is part of this empirical agenda. In addition to well-defined empirical tests, we also offer some conjectures concerning further refinements of collective action theory.

Water resources management is an especially enlightening illustration of the practical and theoretical problems of collective action. At least since over irrigation and sedimentation helped bring down the Sumarian civilization some 3,000 years ago, water and watershed use have been major sources of social debate and political conflict (Gleick, 1993). Wittfogel (1957) predicted that - because of the nature of the resource - all hydraulic-based societies inevitably culminated in despotic governments. Maass and Anderson (1986), and numerous studies since, coun- tered this prediction with thorough descriptions of successful, democratic, self-governing water management institutions.

The article has three purposes: (a) to present a conceptual approach to understanding and explaining the emergence and evolution of collection action; (b) to present the results of a rigorously analyzed, long-term study of collective action to manage watersheds in Haiti; and (c) to present its implications for land use policy in Haiti and collective action theory in general. The research described in this article addresses what Dawes and Thaler (1988) describe as the two key questions in collective action research: what explains “sensible” cooperators? (i.e., what factors are associated with individual choice to either cooperate or defect); and what determines the rate of cooperation? (i.e., what factors are associated with the emergence and evolution of collective action). It builds on the concepts and preliminary tindings presented in White and Runge (1994) and White (1994). Several preliminary points concerning terminology and the research site are in order. We use the term “collective action” to describe the process and consequences of individ-

ual decisions to voluntarily coordinate behavior. Although in particular cases the attributes of the resource, patterns of use and ownership, and type of technology used to manage the resource will differ, all cases of voluntary collaborative decision making can be understood as collective action. Haiti provides a useful laboratory for the study of collective action and watershed management for two primary reasons. First, development experts have frequently attributed land degradation directly to the Haitian peasantry, often characterized as fiercely independent and uncoopera- tive, a group that, “except under extreme duress, is incapable of group action to defend its interests” (Schaedel, 1962, p. iii). Similar claims have long been lodged against indigenous peoples and peasantries around the world. Second, the peasantry has largely been denied the basic freedoms of speech, press, and association for a majority of the last 40 years, and the Haitian state has regularly opposed and violently repressed the formation of politically independent, community-based groups (Trouillot, 1990). Because of this unpropitious environment, local organizations tend to be small in number and relatively simple in structure (Maguire, 1991). The combination of a lack of positive state intervention and simple social structure simplifies the study of self-governance and collective action. According to many, Haiti presents some of the most difficult conditions for both cooperation and conservation in the developing world. If collective action can emerge and manage natural resources in Haiti, then there is reason to believe that it can succeed in other areas as well.

2. CONCEPTUAL FRAMEWORK AND RESEARCH SETTING

(a) Introduction

We investigate the emergence and evolution of collective action groups to control transboundary erosion in 22, small, multiowner watersheds in Maissade, Haiti. The collective action consisted of voluntary labor contributions to build checkdams in ravines that crossed private landholdings. Checkdams were the technical innovation that controlled the erosion externality. These checkdams were demanded by landholders because high-valued crops could be planted in the trapped sediment. Cooperative management was the institutional innovation that allowed these gains to be realized. The research focused on understanding why individuals chose either to participate (or not to participate) in the activity, and why new cooperative groups formed in some watersheds and not in others. See White (1994) for an in-depth description of the site and research methods.

Before examining the central questions of why, we first need to understand what happened. Figure 1

COLLECTIVE ACTION 1685

Socio-physical context

? status quo:

Existing interdependencies

II

Factors Factors

affecting challenge affecting choice

Change in status quo

status quo accepted

_ 2. Individual _ 3. Emergence and

choices to either evolution of

collective action _ cooperate or

proposed defect

1 Defect \

Failure or

termination

Factors

affecting emergenge

Figure 1. Conceptualframework: Assessing the emergence and evolution of collective action

shows a framework for assessing the emergence of the collaborative institutions. The approach is simple and generic, and is an extension and combination of previous works on institutional change.2 It can be used to assess the interrelationships between the physical attributes of resource systems and the overlying social systems, and the process by which individual choices lead to actions and then to new rules concerning the use of those resources. Although applicable to the emergence of collaborative institutions of all types, here it is tailored to reflect research on land and water use in watersheds.

What happened in Haiti, and what happens in general in cases of collective action can be understood to occur in three conceptually distinct phases: (i) a challenge to the status quo and proposal for collective action; (ii) choices by individual agents between cooperating and not cooperating in this action; and (iii) the emergence and evolution of the action (i.e. the summary consequences of these individual choices). In reality, these phases are recursive and iterative. Individuals would probably not initiate an action unless they believed that there was some chance that others would cooperate. Similarly, sustained action requires continued cooperation, so that individuals continually face new choices over whether to continue to cooperate or to defect.

The challenge to the status quo is a normative assertion that the present situation is either inefficient, unfair, or both (Dahlman, 1979). Because new collective action frequently involves either creating or redis- tributing rights and duties and such change inevitably involves power and authority, this challenge is an act of “political leadership” (Frolich, Oppenheimer and

Young, 1971; Guttman, 1982). It may be endogenous to the group, or supplied by an external source (e.g., a representative of the state) and could be stimulated by shifts in ideology, knowledge, power or resources (Hayami and Ruttan, 1985).

The second and third phases involve individual choices either to cooperate or to defect, and the subsequent emergence and evolution of collective action. In the second phase, individual resource users, or stakeholders, in effect cast votes either to cooperate or to defect. In the third phase, the final tally of votes determines the collective outcome. As in any voting situation, individual choices will be contingent on the expected behavior of the other participants and the probability attached to certain outcomes.

The choices made in each of these three phases are determined and explained by two preexisting conditions: the sociophysical context within which the particular action occurs (i.e. the patterns and attributes of the overall sociocultural and physical systems that influence the action); and the status quo land use and welfare interdependency (i.e., the particular contro- versies, dilemmas or externalities that establish the rationale for collective action). The overall sociophysical context constrains choice in the short run yet is the product of these choices in the long run. In this sense, institutions both shape and are shaped by individual preferences and choice. The dilemmas and externalities, or welfare interdependencies, that give rise to calls for collective action exist (and may per- sist) even when they are clearly inefficient, because the status quo often acquires a moral “legitimacy” over time, because of the transactions costs of constructing new institutions, and because of imperfect


information concerning the impacts of resource use and the likely behavior of other stakeholders (Bardhan, 1989; Dahlman, 1979).

The sum of individual choices to cooperate or defect may result in a change in status quo (i.e., new rules, relationships and physical impacts resulting from collective action). To continue with our voting analogy, if the final tally of an election is sufficient to provide a majority to a particular political party, then this collective choice may result in a new political regime. The outcomes of these collective choices become part of the prevailing sociophysical endowment. Such changes in the status quo are ultimately determined by prevailing systems of authority, which define who has the right to accrue benefits, and who is liable to bear costs (Dragun, 1983). A full description of the conceptual basis for the approach is given in White (1994). A brief description of the Maissade cooperative activity, in accordance with the conceptual framework fol1ows.j Here we begin by describing the sociophysical context within which the activity took place and then proceed to the subsequent steps in the process.

(b) Sociophysical context

Maissade is located in the Central Plateau region of Haiti, where the vast majority of adults are peasants. Maissade’s landscape is hilly and soil erosion is a prime factor contributing to steadily declining agricultural yields. The physical situation in Maissade is representative of much of the rest of Haiti and many “fragile land” areas in the developing world. The sociocultural context is richly complex.4 As in the rest of rural Haiti, governments and military officials have discouraged local peasant organizations and endogenous responses to local dilemmas. This is at least partly due to the fact that local police and justice officials have historically derived a substantial portion of their personal income by arbitrating peasant disagree- ments (Trouillot, 1990). The 22 watersheds studied averaged nine hectares in size and ranged from two to 34 hectares. These watersheds differ from other local watersheds - in terms of physical and social attributes - only in that they suffer a higher degree of soil erosion.5 At the time of survey in 1990, there was an average of nine agricultural parcels and nine different landholders per watershed.

(c) Status quo: land use and weCfare interdependency in Maissade

In Maissade, as in other similar areas where peasants rely upon rainfall for agricultural production, soil moisture retention is a critical factor affecting productivity. Whether or not rain water is effectively man-

aged affects crop composition, and can mean the difference between degraded or productive lands, and low or high crop yields. In Maissade, this problem of soil and water erosion has two components: the “private” component, in which erosion on one’s own land results in declining productivity and private benefits; and the “public” component, in which erosion dam- ages cross private property boundaries and negatively affect downstream neighbors.

Transboundary erosion is common and practically inevitable in Maissade where the soils are such that infiltration is slow, and no soil conservation structure (save an impervious dam) can effectively retain over- land flow. This physical limitation means that although soil conservation measures can drastically reduce erosion, no structures can stop it. Population density land subdivision and landscape result in a situation in which almost all peasants receive spillover from one peasant and contribute spillover to others.

For this reason, each peasant’s land use decisions are affected by the choices made by upstream actors, and those decisions are influenced by uncertainty over both the physical impacts of erosion and the expected behavior of those upstream. For example, if down- streamers independently install checkdams to control erosion without similar action upstream, then the downstream checkdams will be destroyed by torren- tial flows. For this reason, downstream peasants cannot make worthwhile investments to control erosion unless they are assured that upstreamers will also install erosion control structures. Individuals cannot optimally define their own rewards to effort without considering the expected behavior of others. Optimal production for all landholders thus involves coordinating land used to reduce and control transboundary externalities (see Runge, 1981).

This transboundary erosion problem could be solved - and the public good of a protected and productive watershed supplied-in three different ways: (i) if a sufficient percentage of peasants voluntarily and independently chose to absorb (internalize) the social costs of erosion from their parcels thus reduc- ing transboundary flow to an “efficient” level; (ii) if a critical mass of peasants coordinated land use and cooperatively reduced the transboundary erosion problem; or (iii) if state action to internalize the social costs of erosion were undertaken and enforced from outside the local group. It is in each peasant’s interest to secure expectations over the behavior of upstream peasants through rules of behavior, either endogenous to the group or imposed and enforced from outside.

In the Maissade area rules to secure expectations and coordinate behavior are not generally present, and transboundary erosion is thus generally uncontrolled. Projects by nongovernmental organizations (NGOs) have been successful in sponsoring peasant adoption of soil conservation measures on private agricultural


fields, but not in the transboundary ravines. Even the level of independent adoption on private lands is insufficient to solve transboundary problems. State actions (both policies and projects) to address “public” erosion have proven ineffective (Murray, 1979; White and Jickling, 1992). Yet our analysis of field data suggests that in some recent cases, a critical mass of cooperation has nonetheless emerged. Why?

(d) Challenge to the status quo: the cooperative watershed management program

Save the Children Federation (SCF), a nongovem- mental organization, began an integrated rural development program in the Maissade region in 1986 by: (i) encouraging formation of groupman (self-organized pre-cooperatives averaging eight members); and then (ii) encouraging individual peasant adoption of agro- forestry and soil conservation techniques. In January 1989 SCF initiated a cooperative watershed management program to encourage the complete treatment of ravines and hillside agricultural parcels in the watersheds, and the creation of new watershed management groups composed of individuals who either owned or worked lands within degraded microwatersheds (SCF, 1990). These new institutions were predicated on the members’ common interest in managing transboundary erosion and increasing agricultural yields; and it was assumed that individuals would voluntarily regroup to construct new seasonal checkdams and maintain those previously constructed (Gladdis and Smucker, 1988).

The program strategy entailed promoting the following claims: existing levels of transboundary erosion were excessive; easily adoptable and economically beneficial techniques were available; private, uncoordinated action was not sufficient since checkdams installed independently of upstream pro- tection would be destroyed with rainfall; and finally that peasants thus could and should organize to coordinate their actions. It was this last assertion that con- stituted the challenge to the status quo. The previous claims had already been widely accepted as legiti- mate. Peasants had spoken of the value of such coordination, and had expressed interest to the extension staff, but had not acted upon those beliefs.

Save the Children staff began on-site meetings with all landowners in each watershed in January of 1989 to develop consensual knowledge concerning the problem - its causes, effects and the relative rights and claims of the landholders - as well as alternative solutions to the problem. Watershed groups were self- organized and directed, and no external incentives were provided to encourage individual cooperation or group formation.

(e) Individual choice to either cooperate or defect: participation in collective action

Individual peasants in each watershed chose whether to join the group efforts. All peasants who held land in the watersheds - regardless of tenure - were invited to the initial meetings and were subse- quently encouraged by active peers to participate in the work events. Groupman members were not specifically targeted and encouraged to participate. Participation took the form of labor contributions to those specific work events. Participants were almost completely male, as women typically did not engage in the heavy agricultural labor of checkdam construction. This is not to imply that women did not influence male choice. It is widely recognized that in Haiti women play a strong role in decisions concerning the investment of household labor.

Of all landholders, 54% cooperated (i.e., joined the work events on two or more occasions) and an average of five landholders participated per watershed. An average of four individuals who did not own land in the watersheds participated per watershed. These individuals are referred to as “nonwatershed participants” in the following text. These participants contributed an average of 36% of all of the labor constituting the collective action. The findings concerning nonwatershed participation were unexpected and suggested a form of voluntary association which inspired the need to think of social interdependency beyond the physical watershed itself.

Checkdams were constructed on the lands of 49% of all landholders and an average of three checkdams were built per parcel. Only 34% of all participants had checkdams constructed on their land; in fact, 22% of all checkdams were built on nonparticipant land.

(f) Emergence and evolution of collective action: pe$ormance of the watershed groups

Within two years of activity, principal ravines were completely treated in 10 watersheds, partial treatment was achieved in seven and scant treatment was achieved in five. Groups were usually active for two to three months a year, between the late dry season and early rainy season. The number of work events averaged nine per year in 1989 and six in 1990, and ranged from zero to 28.

Although the project had initially encouraged groups to achieve complete watershed treatment, groups focused almost completely on the primary ravine. Few groups were active on the slopes and even then on few occasions. Cooperation was needed to treat the ravine - the prime source of transboundary erosion and economic damage - yet treatment of parcels was left to individual owners.

Project support of the watershed management activity ceased after two years in the summer of 1990


due to political instability and government repression of peasant groups. Despite this climate, a follow-up survey conducted in 1992 found that groups remained active in 12 of the 22 original watersheds (55%). Groups remained active in 50% of the watersheds that were categorized as “completely” treated in 1990, while 71% of the “partially” treated watershed groups remained active and 40% of the “scantly” treated watershed groups were active.

(g) Change in status quo

How the collective action altered physical and socioeconomic status quo of Maissade is inherently difficult to ascertain. The simplest tangible indicators of change relate to the physical and economic impacts of the action. Of course, the process produced a wide set of - perhaps immeasurable - behaviors and relations that could form the context for successive social changes.

After two years of activity a total of 590 checkdams were constructed by the groups. Treatment averaged 27 checkdams per watershed and ranged from 2 to 92 per watershed. These checkdams held an average of 3.2 cubic meters of sediment and retained an average of 39 tons of soil per hectare per year (White and Jickling, 1992). This sediment produced financial benefits to peasants, allowing them to both diversify and increase crop production6

3. FACTORS ASSOCIATED WITH INDIVIDUAL CHOICE TO COOPERATE

In order to understand particular cases and devise unifying theories we must explain each of the three phases of collective action: what factors affect the ini- tiation of collective action? What factors affect the choice of individual, yet interdependent, stakeholders to cooperate, when all face the option to defect? Finally, what factors affect the further evolution of the group? Ethnographic methods were used to answer the first question and statistical methods were used to answer the last two questions. The research described in this and the subsequent section respond to the last two questions. We discuss all findings in the final, concluding section and refine the conceptual approaches already presented.

In order to better understand the motivations for cooperation we tested a series of LOGIT regressions to identify the variables associated with cooperation.’ The first LOGIT model was constructed to determine the variables associated with individual landholder choice to either cooperate (or defect).& Other models were developed to better understand motivations.

(a) Predicting watershed landholder cooperation

“Membership in groupman,” “prior adoption of contour soil conservation techniques” and “number of checkdams acquired’ were the variables most strongly and significantly correlated with landholder choice to cooperate.Y Positive responses in these attributes are positive predictors cooperation. The other variables: “age, ” “religious preference,” “voodoo practice,” “wealth status,” “parcel position,” “labor acquisition type,” and “land tenure type,” were not significantly correlated with cooperation.

Motivations associated with membership in groupman and practical knowledge of the benefits of soil conservation thus overrode other motivations. When considering these results it is important to keep in mind that only those individuals with land in the watersheds were included in this model, and that 36% of the effort was contributed by individuals who were nonwatershed participants.

(b) Predicting nonwatershed cooperation

The phenomenon of nonwatershed participation was one of the more curious aspects of the activity for analysts, implying a richer sociological interdependency than defined by the watershed itself. Who were these people and why did they come into the watershed to help their neighbors? The best fitting LOGIT model indicated that their participation was positively correlated with groupman membership and membership in reciprocating labor-exchange groups, and negatively correlated with age.iO Some 90% of the nonwatershed individuals were groupman members.

(c) Predicting membership in labor exchange and groupman

Because “groupman membership,” and “membership in labor exchange groups” were strongly associated with the choice to cooperate for both watershed landholders and for nonwatershed participants, we tried to identify variables that predicted individual choices to engage in groupman and labor exchanges. The strongest LOGIT model found that groupman membership was best predicted by prior membership in labor exchange groups and individual affiliation with Protestant religions. ‘I Groupman membership was not significantly associated with age or any indi- cator of wealth.

Labor exchange membership thus appeared to be a facilitating condition for joining groupman. We also found that labor exchange membership was the strongest predictor of individual choice to independently adopt soil conservation structures on one’s own land. The finding that groupman members are dispro-


portionately Protestant correlates with several earlier findings. Cooperators in collective action were dis- proportionally Protestant and did not regularly engage in voodoo (White and Runge, 1994). Perhaps Protestant willingness to contribute labor to a common cause was due to “missionary zeal” or a sense of obligation.

Membership in labor exchanges is apparently a key variable underlying higher forms of risk-taking (e.g., technique adoption) and cooperative behavior. What are the determinants of membership in labor exchange? Of nine models tested, none provided a good fit. No variables were significant. This is intrigu- ing. Previous tests suggested that labor exchange members could tend to be poorer and might tend to exchange labor because they lacked capital to pur- chase labor. They were not poorer, however, at significant levels. Previous tests also suggested that labor exchanges might be associated with Protestantism and youth. They were not. Understanding the determinants of participation in this fundamental social organization is an area meriting further investigation.

4. FACTORS ASSOCIATED WITH EMERGENCE AND EVOLUTION OF

COLLECTIVE ACTION

The level of watershed treatment ranged from high in some of the 22 watersheds to low in others. Why this difference? What explains why groups emerged and survived in some watersheds and not in others? This question was addressed with two sets of analyses: identifying and ranking the variables significantly associated with different levels of cooperation; and identifying the variables significantly associated with group choice to continue action over the four-year test period. From these analyses models were constructed that could predict the emergence and survival of collective watershed management groups.

To conduct the analysis, the 22 watersheds were divided into three categories of treatment (complete, partial and scant) representing three levels of cooperative activity. Completely treated watersheds averaged 37 checkdams, partially treated averaged 25 and scantly treated averaged eight.‘* A series of Poisson models was tested to identify explanatory variables. A parallel series of least squares linear regressions was tested to check the findings of the Poisson models. In all of the cases discussed below, the least squares regressions verified the findings of the Poisson models.

(a) Predicting the emergence of collective action

A preliminary series of statistical tests identified the variables that were statistically different across the different categories of collective action. The likeli-

hood of collective action emerging in watersheds significantly increased with: the proportion of watershed parcels that were in the mid-stream position; the proportion of watershed landholders who had adopted soil conservation techniques; the proportion of landholders who were members of labor exchange groups; the proportion of landholders who were members of groupman; the percentage of landholders who directly benefited from the collective action (with checkdams constructed on their land); and the mean number of landholders who participated per work event. The first four of these six variables can be used to predict collective action ex ante (i.e., before the action has emerged), while the latter two are correlates of action ex post (i.e., once the action had occurred).

The following variables were not statistically significant: mean ages of both landholders and nonwatershed participants; mean levels of landholder wealth (as indicated by quantity of pigs, cows, land parcels and total land hectarage owned); proportion of formal religious preference (Catholic or Protestant); mean watershed size; mean number of years of project activity; mean number of landholders; distribution of land tenure types; and mean parcel size (White and Runge, 1994).

Poisson models were fitted to identify which of the significant variables were the best predictors ex ante of cooperation, and which of the variables were the most strongly associated with action expost. The best fitting predictive model included the “proportion of landholders who had adopted soil conservation techniques” variable only.13 The other three variables (proportion of lands in mid-stream position, proportion of landholders who were labor exchange members, and proportion of landholders who were groupman members) were not statistically significant.

The model that best identified the strongest correlates of action (ex post) included the “proportion of landholders who benefited” variable only.14 This finding strongly suggests that the most important correlate of collective action is a high ratio of benefits per capita.

Interestingly, aggregate levels of potential to gain (indicated by the number of parcels in the midstream position), and aggregate levels of landholder participation appeared to be relatively unimportant for the emergence of collective action. Such action was most strongly correlated with the ratio of gain for all landholders, the ratio of individuals who had practical knowledge of that potential gain, and the ratio of individuals who were members of social groups that facil- itated its capture. It is also notable that the ratio of gain for all landholders was significant, not the ratio of gain for each participating landholder. This suggests so-called enlightened and forward-looking behavior, and confronts predictions that a “limited good” men- tality (e.g., Foster, 1965) constrains rural development in Haiti.


Because indicators of the ratio of benefits per all potential beneficiaries proved the strongest correlate of collective action, another model was formulated to identify the variables that best predicted this outcome. Of nine variables tested, the most important, single variable predicting the ratio of gain to all landholders was the “number of nonwatershed participants per hectare of watershed.” This finding is quite striking and suggests that - holding all other variables con- stant - the level of external assistance (from nonwatershed participants) was the single strongest correlate of the rate of action. But this does not necessarily suggest causation. Perhaps nonwatershed participants were drawn into watersheds where there was a high amount of action. Apparently, not only did external assistance contribute to the gain of cooperating landholders, but also (and especially) to the gain of the noncooperating landholders.

(b) Predicting continued action

Another series of tests was prepared to identify the variables associated with group decision to continue collective action to treat the same watershed work was initiated four years previously. Two sets of models were prepared, one including variables that would permit the ex ante prediction of group choice to continue and the other the expost situation faced by the group.

The model that best predicted continued action ex ante included only the “proportion of landholders who had adopted techniques” variable.15 The “proportion of landholders who exchanged labor” and the “proportion of landholders who were groupman members” variables were the next most important followed by the “proportion of parcels in the midstream position.” The two variables indicating watershed size were the least important. This last finding again suggests the weakness of physical scale variables in predicting collective action. It also corroborates previous findings and indicates that prior practical knowledge of a potential gain is the best predictor of participation in a collective action that produces similar gains.

The model that best explained continued action ex post included only the “proportion of landholders who had adopted soil conservation” variable. This was the same model found with the ex ante variables only. Interestingly, the “proportion of landholders who benefited”, “proportion of landholders who were groupman members” and the “proportion of landholders who engaged in exchange labor” were not significant. Neither was the “number of nonwatershed participants” variable significant. Apparently the attributes required for the emergence of action are not necessarily the same ones required to sustain it. In this case, it appears that the practical knowledge of potential gain was the only variable that was both necessary and sufficient for the emergence and continuation of collective action in the long run.

5. CONCLUSIONS AND DISCUSSION

This case supports the possibility of encouraging voluntary organizations to provide technical conservation assistance and then encouraging voluntary collective action to manage watersheds. The Maissade collective management activity was successful in a majority of watersheds where action was proposed, and even continued despite unpropitious conditions for the four-year study period. In these active watersheds, once peasants had assessed their land-use problems, the alternative solutions, and demanded changes in the status quo, no external incentives were necessary to spur action, although external incentives might have been required to achieve treatment in all of the watersheds.

In essence, SCF provided a local political climate conducive for collective action, problem identification and analysis, and leadership to address both individual and collective problems. SCF essentially acted as a political entrepreneur by: facilitating dialogue over watershed interdependencies and the alternative solutions; making moral claims that watersheds could and should be treated, and setting standards of performance; and paying the transaction costs of encouraging coordination (via hiring and training extension agents).

Individual choice to cooperate in collective actions that yield public goods is conditional, and best predicted by: prior membership in reciprocating social organizations; practical knowledge of potential gains from cooperation; and some potential to gain from that action. In other words, noncooperative free-riding, or defection is not a dominant strategy or an innate trait of rural Haitian culture. Contrary to Schaedel’s claim (1962) peasants will cooperate to further their own interests. Volunteers within and beyond the watershed flocked to cooperate in the new collective activity. Choice to cooperate appeared strongly influenced by the degree to which individuals were assured of the overall feasibility of the cooperative action and of reciprocal behavior of others. In Maissade, individuals appeared to be strategically interdependent members of a multilayered social system. The motivations and concerns related to this social interdependence appeared strong and formed the basis of various expressions of collective action.

Focus group interviews also revealed that peasants were generally expected to participate and had to provide excuses when they choose not to participate. This evidence, in addition to the statistical analyses, suggests that: a social norm existed in which peasants were expected to cooperate in actions that provided certain public goods; and peasants chose to act on this predisposition when the conditions listed above were present. This is certainly not to say that rural Haitians are categorically cooperative. As with any social norm, there is a distinct difference between social


competence and social performance. But the existence of these expectations suggests that this rural culture embodied a moral economy entailing norms of solidarity and the duty of reciprocity.

These findings support Fafchamp’s (1992) formal argument that there is no contradiction between rational peasant behavior and social ethics emphasizing solidarity as a moral obligation and subsistence as a right. These findings also lend support to Sugden’s (1984) claim that a “principle of reciprocity” (that people are not always expected to contribute [or cooperate], but must do so when others do) is a pervasive social ethic.

It is also interesting that there was no discernable “spite” factor against defectors. The “spite” factor (built into some formal models) holds that even in situations of probable net gain, individuals will choose not to cooperate because they do not “want to be taken advantage of by free-riders” (Uphoff, 1992). It appears that Haitian peasants have a zone of toleration and, at least to some degree let others get away with blatant shirking. These findings suggest that the behavioral basis exists for higher order, freely composed and self- governed collective actions. This research cannot answer the question of whether Haitian peasants are especially predisposed to cooperative behavior when compared to other groups or societies.

Individuals are not motivated solely (or perhaps even principally) by the intended outputs of the collective action, but also by the benefits of participating in the process of action. In Maissade, although transboundary erosion was the inspiration for coordinated action, the resolution of that externality was not the primary basis for individual participation. In other words, the existence of an externality does not necessarily imply its resolution. The watershed management program planners expected that exposure to the costs of transboundary erosion would serve as strong basis for participation in resolving that problem. For this reason planners expected that membership in management institutions would be reflected by the watershed boundaries. They assumed that individuals with land (i.e. economic interests) within the watersheds would have the natural incentive to coordinate action with others in the watershed, and that participation would be limited to those coordinating action within the watershed. They were wrong. This finding indicates that external agents seeking to improve public participation in the provision of public goods should seek to assure that the process of action adheres to the logic and norms guiding existing, local voluntary institutions.

Short-term gain (via the output ofthe action: checkdams) was not a primary motivator for participation. The majority of cooperators did not gain (at least as measured by the study) within the two years of study. This does not suggest irrationality, but that a larger set of arguments (e.g., labor exchange or social obliga-

tions), and a longer period of time over which reciprocity is likely to occur. The fact that over one-third of all labor was contributed by individuals outside of the watershed clearly questions the wisdom of encouraging ownership in the watershed as a precondition for participation in the collective action arrangement.

We do not know precisely how much participation was motivated by social norms (i.e. duties to contribute) or by expectations of reciprocity. It appeared that participants, at least to some degree, were “bank- ing favors,” or “building debts,” in hopes that they would someday be reciprocated. It was also apparent that these debts were fungible, and that a labor contribution to build a checkdam on an individual’s land might be repaid in other types of favor at a later date. These “tit-for-tat” reciprocal relationships functioned successfully, at least in part, because they were both fungible and redeemable over longer periods of time.

Indigenous, self-organized cooperative groups are the font of higher order cooperative institutions and the voluntary supply of public goods. Participation in the cooperative action was, by far, most strongly predicted by membership in self-organized groupman. Membership in groupman was most strongly predicted by membership in labor exchange groups. Labor exchange appear to be the most fundamental kernel of self-organized and self-governing institutions. These findings provide statistical evidence for Hirschman’s (1984) theory of the “conservation and mutation of social energy,” where democratic social change is led by individuals and cooperative groups who continually transform themselves to address new social and economic concerns.

Participation in these groups apparently provides the necessary experience of sharing information, building trust, constructing rules, monitoring relations and sanctioning necessary for higher order structures. This experience, and the assurance of solidarity and reciprocity, allow members to share risk, leverage resources, extend pay-back periods, test innovations, and make leading contributions to collective actions, even when they know that these ventures might not be rewarded.

This conclusion suggests that any force that (directly or indirectly) weakens such indigenous situations weakens the potential for voluntary supply of public goods and democratic social change. Similarly, actions that increase social insecurity and uncertainty directly challenge social institutions and thus threaten self-governance. It is in this broad sense, as well as a more technical or game-theoretic level, that economic development is an “assurance problem” (see Runge, 1981).

Given the right conditions, the poor will voluntarily contribute to collective actions that provide public goods. In the watershed activity, the poorest of all social categories - the nonwatershed landholders - provided over one-third of all labor for watershed


management. These findings suggest that the strategic building of obligations is a response to poverty and a strategy for survival. This strategy appears to be used especially by young men as a means to invest in their future security. Poorer, younger males may contribute to public goods (in part) to acquire good will and labor debts so that they can eventually buy land. If community members are expected to contribute to an activity that positively impacts the community, then labor is the only resource that poor peasants can contribute. This finding again illuminates the importance of traditional labor organizations as a resource mobilization mechanism of the poor. It also affirms the importance of understanding the logic of local institutions prior to attempting aid interventions.

“Secure” land tenure is not a necessary condition for either investments in land, owner participations in collective watershed management, or the emergence of an effective management regime. The Maissade activities do not suggest that land tenure is not important, but that the costs of “insecurity” associated with uncodified and short-term tenures was overwhelmed by the assurance of potential benefit from action. Land tenure type was thus not explanatory. These findings support the conclusion of Otsuka, Chuma and Hayami (1992) that no tenure type is necessarily optimal.

These conclusions suggest that a preoccupation with tenure types is often misplaced. Focus on tenure type deflects attention from the institutions (social, moral, cultural) that give tenure arrangements mean- ing. It is not the title, or the specific definitions of property arrangements that are important, but the degree to which these arrangements assure individuals that they are part of a stable, equitable and well- adapted set of rights and duties. This finding illustrates the usefulness of the assurance problem in understanding human interdependence and potential for collective action.

The emergence of collective action is explained by a “critical mass” of practical knowledge of potential benefit, and not by physical scale and social heterogeneity variables. In Maissade, the emergence of collective action was predicted by variables indicating the proportion of individuals who had practical knowledge of the potential benefit from acting collectively, and was correlated with high levels of prior membership in collective action groups, and a sufficient level of overall net gain. None of the variables, indicating physical scale or social heterogeneity were associated with the emergence (or lack of) collective action.

These findings suggest that observations, or assumptions, that increases in scale and heterogeneity impede action wrongly assign cause. Although a lack of action might be correlated with scale or heterogeneity variables, a lack of collective action is not due to excessive scale or heterogeneity, but to insufficient levels of the explanatory variables: level of practical knowledge, existence of reciprocal institutions, and a

sufficient level of potential gain. Collective action units serve both efficiency and

distributional functions (productive in terms of private gains from internalizing the externality, and dis- tributive in terms of the diverse benefits generated by the process of acting collectively). Group performance was positively and most strongly correlated with the distribution of the outputs of cooperation to all potential beneficiaries. The proportion of participants who benefited was not significant. Apparently, if action does not equitably benefit the potentia1 beneficiaries, then it is frustrated. Fairness is acritical fea- ture of successful collective action (see Runge, 1984) and apparently “e&iency-improving institutional change cannot really be separated from that redistrib- utive institutional change” (Bardhan, 1989, p. 11, italics in original).

The fact that the ratio of gain was positively, and most strongly, associated with the level of nonwatershed assistance further supports the public goods nature of watershed management, and that geographi- cal boundaries are not useful in understanding possible solutions to degradation problems. Rather than thinking of collective action institutions as existing to solve a single public goods problem, it may be more appropriate to think of such institutions as a “bundle of opportunities,” one that solves different problems for different individuals.

Group decision to continue action is most strongly predicted by proportion of potential beneficiaries who have practical knowledge of the benefits of action. The only significant predictor of choice to continue was the proportion of landholders who had previously adopted soil conservation techniques. This evidence argues strongly that, given a conducive environment and political leadership, practical knowledge is the preeminent determinant of institutional change. This finding provides statistical evidence for North’s (1994, p. 362) claim that “the most fundamental long- run source of change is learning.” We found that in the short term, this change can be limited by two types of constraints: landscape factors that affect the potential net economic gain (e.g., in Maissade, the distribution of agricultural parcels); and sociocultural factors that affect the cost of constructing the new institution (e.g., in Maissade, the endowment of indigenous reciprocal institutions). It should also be noted that knowledge of potential gain also embodies an acceptance of a normative claim that the status quo merits changing.

External agents can more efficiently promote collective action to engage in publicly beneficial projects by indirectly facilitating self-organization rather than transferring or imposing arrangements from the outside. In Maissade, rather than following the recom- mendations of project advisors and constructing new, watershed-based management organizations, peasants transformed their existing sociocultural institutions to meet their new needs. Using Hirschman’s terms, insti-


tutional innovation was more a “renewal” of existing social energy than an “outbreak” (1984).

The implication of this result is that rather than promoting preconceived organizational forms, external agents should facilitate the adaptation of existing institutions. An advantage of endogenously created institution is that they reflect the variety of conditions from which they emerge, conditions which might be unknowable to external agents. Although institutions for collective action may be created or sustained with external subsidy or coercion, externally imposed institutions are inherently weaker and stand a greater chance of failure than those that are endogenously created. For these and other reasons already cited, collective action institutions are best encouraged by policies and programs that facilitate bottom-up design.

In summary, the Haitian government (or NGOs) can facilitate voluntary supply of public goods in Haiti by: encouraging the formation and strengthen- ing of indigenous organizations and assuring legal and political recognition of their positive role in society, then; disseminating knowledge and concern over public “bads” and the alternatives to solve them; setting minimum standards for individual contribution and group performance; encouraging civic leadership to initiate action; and subsidizing the transaction costs of coordination.

6. EXTENSIONS AND IMPLICATIONS

When reviewing the implications advanced here, the reader should keep in mind that although they derived from rigorous statistical analyses, the cases studied were relatively small scale and particular to one region in Haiti. Care should thus be taken in extending them to other situations. Much more empirical work is needed in order to build a strong body of collective action evidence and before specific institutional responses can be advocated with much confi- dence. But a few general observations for the theory of collective action are possible.

First, the study results challenge the usefulness of the prisoner’s dilemma and tragedy of the commons metaphors as conceptual bases for analysis. At least in Maissade, free-riding does not dominate. The study illustrates an inherent weakness of reductionist methods and models to assess individual behavior and social interaction. Isolating a particular group within boundaries that are socially artificial (such as ecosystems, watersheds, or hypothetical prisons) impedes useful analysis because it eliminates the very factors - moral claims, cultural ideology, long-term social interdependence, external relations, and political leadership - that can explain a large portion of behavior. Eliminating these linkages for the sake of experimen- tal efficiency automatically favors findings of free-rid- mg, shirking, and an undersupply of public goods.

This study supports the usefulness of the “assurance problem” as a conceptual basis for the study of public goods dilemmas, and the need for greater research focus of assurance, expectations and the nature of reciprocal behavior as a basis of institutional innovation.

Second, the study approach and results reinforced preliminary hypotheses explaining the emergence of collective action. A unified theory of the emergence and evolution of collective action must explain why and how collective action emerges (or is supplied), and why and how it evolves. What makes a particular problem a collective action problem? What explains why, and how collective action is initiated? What explains individual choice to cooperate or defect? What explains the coalescence and evolution of a collective action group? These four questions can (at least partially) be answered with four concepts, each corresponding to a different stage in the process of collective action emergence. A collective action problem results fundamentally from welfare interdependency. Collective action is initiated by political leadership, challenging the status quo. Individual choices to cooperate or defect are contingent on expectations of the behavior of others. Collective action evolves due to the collective consequences of individual choices to cooperate or defect. Our ethnographic and statistical analyses here allowed us to predict and analyze these choices.

We conclude that the basic rationale driving voluntary collective action is that individuals or groups are interdependent, and that their welfare - however defined by the individual actors - can somehow be improved with cooperation. Initiating and sustaining collective action involves paying the transactions costs of coordinating the expectations and efforts of others, an important entrepreneurial function. Successful entrepreneurs must voice credible claims that externalities or public “bads” exist and that the current rights structure is either unjust or socially inappropriate, and then build expectations that the issue should be resolved and the assurance that it can be resolved. It is then possible to mobilize people and resources to act on those claims and expectations.

Because interdependence leads each individual’s calculus to be contingent on expectations they hold about the likely behavior of others, individuals are more likely to join a collective action group if they expect others to do so. On the other hand, individuals are less likely to contribute if they do not expect (or are not sufficiently assured) that others will recipro- cate. Individuals are more likely to cooperate when they are already a member of a cooperative institution and have practical knowledge of the potential gain produced by the new collective action. The proposed action will be abetted if it corresponds positively to widely held cultural norms or morals.

Finally, given a conducive political environment and political entrepreneurship, collective action


groups will emerge where a critical mass of individuals have practical knowledge of the potential gains from the intended output of the collective action. Once initiated, the evolution of collective action will be based on the distribution of benefits to potential beneficiaries (not just to the participants). Collective action will continue as long as a critical mass of stakeholders have practical knowledge of the benefits of action. In a dynamic sense, practical knowledge of potential gain from action may be the only condition that is both

necessary and sufficient. Human society is an intricate web of social and

economic interdependencies, in which strands are woven for ideological, economic and political purposes. Cooperation is conditional. Additional

rigorously analyzed empirical cases can help us

better understand why collective action does or does

not work, when it is an efficient means of supplying public goods, and how states either facilitate, or hinder, this action.

NOTES

1. Singleton and Taylor (1992) and Taylor and Singleton (1993) have proposed that the theory of transactions costs and the notion of “community” combine to form the basis for a theory of collective action. Although attractive, the key fea- tures of “community” are, as yet, not defined and thus the proposal lacks explanatory and predictive power. As stated by Ostrom (1992), although a sense of “community” may be a necessary condition for the emergence of collective action, it is not sufficient.

2. The approach draws on frameworks for assessing common property regimes by Oakerson (1986) and Blaikie and Brookfield (1987) and builds on previous theoretical work by the authors (Runge 1981, 1984, 1986; White and Runge, 1994). It also roughly corresponds to the process of regime formation independently proposed by Young (1993): the prenegotiation, negotiation and postnegotiation stages. It follows in the tradition of Hayami and Ruttan (1985) by treat- ing institutional innovation as endogenous and induced by shifts in the relative abundance and scarcity of certain factors, and the relative costs and benefits of alternative strategies. It builds on the theories of ‘*spontaneous order” (Hayek, 1960; Schotter, 1981; Sugden, 1986,1989). It also follows in the tradition of cultural ecology that culture and institutions are strongly influenced by the interactions between humans and the physical landscapes where they exist, and further, that ecosystems are useful units of analysis (Steward, 1955; Geertz, 1963). In this approach new institutions are an outcome of strategic interactions between individuals (or groups of individuals), and are driven (or influenced) by preexisting individual, cultural or political norms and rules, and the distribution of power at both local and meta levels. Focus on individual choice facilitates analysis and is not a denial, or subjugation, of ideological or cultural models of social change (North, 1990). The usefulness of this-and any other rational-choice based - approach will be determined by the degree to which the analyst interprets behavior and choices in the context of the particular sociocultural structure (Peters, 1993). See White (1994) for an in-depth description of the approach and the underlying rationale.

3. Field work was conducted by the first author with the assistance of Save the Children staff during August, September and December 1990, and again in May 1992. Various survey instruments were used, including formal questionnaires, focus-group interviews with watershed groups, and aerial photo interpretation. All information was

crossreferenced among the staff extension agents, local informants and the author.

4. Of the 186 individuals who held land in the watersheds: 57% were members of preexisting social organizations called groupmnn (self-organized and governing farmer cooperatives averaging eight members); 56% had previously adopted contour soil conservation techniques in their private fields; 74% expressed a Catholic religious affiliation (and the com- plementary percent expressed a Protestant affiliation); and 70% regularly participated in indigenous religious practices. Some 70% of landholders either worked alone or purchased labor for agriculture while the remaining 30% primarily acquired agricultural labor via cooperative exchange arrangements. Of landholders: 41% claimed to conduct major agricultural tasks during the agricultural season individually @ou kont yo); 13% worked in pairs (boukanre mart-n); 16% were members of reciprocal exchange groups (asosye); 10% hired day labor (buy djob); and 20% equally hired outside labor and worked individually. By all accounts the vast majority of Haitians believe in some aspects of the so-called voodoo folk religion. The people of Maissade dis- tinguish between those who regularly w-vi loa by donating plot manje to ancestral spirits, and those who have ceased to continue this practice. In terms of wealth, landholders owned an average of three separate land parcels, totaling approxi- mately two and one-half hectares. This degree of wealth is probably above average for rural Haiti. It was this distinction that was used to categorize the individuals surveyed. For more information on Haitian folk religious practices see Metraux (1972). Herskovits (1971) Murray (1980). and Smucker (1983).

5. Hillslopes average 12% and the primary ravines that drain the watersheds, and were the focus of remedial conservation action, carry seasonal storm flow. An average of 52% of parcels within the watersheds were owned (achre, erirye, tit); 33% of parcels were inherited lands that had not been formally divided between offspring (indivize); 9% were rented (fern, pretansyon); and 5% were under crop-share arrangements (demwafye). Although the term “owned” is used here to describe the most unattenuated forms of tenure, it does not necessarily confer that the holder has an official or legally binding title. On only rare occasions do peasants have legally enforceable titles to property or written con- tracts specifying rights of use. Terms in italics represent the Haitian Creole designations that were combined for each category. The term ochte means that the land was purchased and


the holder probably has some formal description of the sale, erirye means “inherited” and suggests that the holder does not have legal title, and fit means “title” often implying that the land has been surveyed and that some type of title is in hand. For in-depth descriptions of rural land tenure see McLain and Steinbarger (1988). Murray (1977). Smucker (1983). and Zuvekas (1978). An average of 27% of parcels were located on the sideslopes of the watersheds (i.e. did not contain a major ravine); 16% were upstream (i.e. were located at the head of a ravine); 37% were midstream (i.e. located between up- and downstream positions); and 20% were downstream (i.e. either located in the most downstream position, or bordered a jointly owned ravine). About two- thirds of all parcels were contiguous with the principal ravine. Fourteen percent of landholders work lands in which the principal ravine was jointly owned. In this situation, the centerline of the principal ravine forms the property boundary between adjoining parcels.

6. On 60% of the checkdam sites, crop-types changed following construction and peasants increased the types of crops in almost 40% of the cases. Accurate estimation of the economic impacts is difficult, because of the inherent complex- ities in estimating the value of increased crop diversity, and the value of the increased certainty of production due to the greater access to moisture and the treatment of the ravine upstream. A minimal estimate of the financial returns from each checkdam (considering only production increases and not labor) was 44.4 Haitian Gourdes. The benefit-cost ratio from the peasant perspective was 2.3: 1 (White and Jickling, 1992).

7. LOGIT regressions were used because they permit the use of both continuous and categorical data, and because they predict the probability of a binary response (in our case, predicting either cooperation or defection). The following variables were tested to determine correlation with individual choice to cooperate or defect: (a) potential to directly benefit financially from the output of the collective action: the instal- lation of checkdams. This variable is indicated by landholding position in the watershed (sideslope, upstream, midstream, downstream) and length of ravine on individual’s landholding; (b) number of checkdams constructed on individual’s landholding (indicating the individual’s actual real- ization of direct gain from the collective activity; (c) land tenure of the agricultural parcel held in the watershed (rented, owned, share-cropped, or undivided inheritance); (d) individual’s official religious affiliation (Catholic or Protestant); (e) whether the individual regularly participated in voodoo ceremonies; (f) individual’s wealth. This variable is indicated by total number and aggregate size of lands held, and the number of cows an pigs owned; (g) whether individual is a member of a groupman; (h) the manner in which the individual acquires labor for major agricultural tasks - individually @ou konr yo). in pairs (boukanfe maren), via labor exchange groups (asosye), hires day labor (bay djob); (i) whether individual had previously adopted soil conservation techniques on their own land; and (j) age of the individual.

8. The backward elimination method was used to choose between candidate models. Variables were dropped from the equation if they did not contribute significantly to explaining their deviance (i.e. if the p-value of the difference in model deviance was > .OS). Models were chosen on the basis of low- est AK (Akaike Information Criterion) for models with sig-

nificant variables. Parallel linear regression models were prepared in order to test the findings of the Poisson technique. In order to run the linear regressions, all categorical variables were transformed into representative parametric statistics, and a new continuous variable was devised to represent the level of treatment. All sociological parameters were put on a per hectare (rather than proportional) basis. All of the new variables were highly correlated with the original, (proportional) variables. Candidate models were screened with the D&in-Watson test for autocorrelation and the Wilk-Shapiro test to assure that the standardized residuals were normally distributed. Standardized residuals were also plotted against the dependent variable to assure that the assumption of ran- domness was not violated. In all cases the linear regressions verified the chosen LOGIT and Poisson models.

9. Of 13 models tested to predict landholder choice to either cooperate or defect, the final model chosen was: y (log odds of cooperation) = - 1.66 + 1.85 (groupman membership) + .66 (technique adoption) + 1.09 (number of checkdams acquired). Thus, for example, for an individual who is a groupman member, and a technique adopter, yet did not benefit with the placement of checkdams on his/her land, the odds of choosing to cooperate in the collective action are as follows: y (log odds of cooperation) = - 1.66 + l.85( I) + .I l(0) = .84. Log odds (i.e., logits) are converted to proba- bilities with the following formula: probability = &“/I + el”@. Using this equation, our logit of .84 converts to a probability of .7. The coefficients of the final model can be used to indicate the odds of an individual choosing to cooperate or defect: (a) groupmnn membership alone increased the odds of choosing to cooperate by 6.4 times; (b) prior technique adoption alone increased the odds of cooperation by 1.9 times; (c) the odds of cooperation increased by 1 .I. times for each checkdam constructed on peasant land; and (d) positive grolcpman membership and technique adoption together increased the odds of peasant participation by 12.3 times.

10. Of the 12 models tested, the final model chosen was: y (log odds of nonwatershed cooperation) = - .05 (age) + 1.39 (whether individual was a member of a ,qroupman) + 59 (whether individual was a member of a labor exchange group). For example, for a nonwatershed individual who is 30 years of age, a groupman member and a member of a labor exchange group, the odds of choosing to participate are as follows: y (log odds of participation) = - .05(30) + 1.39( 1) + .59 (1) = .34. This logit of .34 converts to a probability of 59. The coefficients of the logit regression indicated that: (a) groupman membership alone increased the odds of participation by 4.0 times; (b) membership in labor-exchange groups alone increased the odds of participation by I .8 times; (c) the odds of participation decreased by I. I times for each year of age; and (d) groupman membership and labor exchange together increased the odds of nonwatershed par ticipation by 7.2 times.

1 I. Of 12 models tested to predict groupmcrn membership, the final model chosen was: y (log odds of groupman membership) = 2.36 (whether the individual was a member of a labor-exchange group) + 1.02 (whether the individual was Protestant or Catholic). For example, for individuals who are member of labor exchange and refer to themselves as Protestant, the odds of being a member of a ~r~,uprnu,i are >’ (log odds) = 2.36(l) + 1.02( 1) = 3.38 A logit of 3 38 con-


verts to a probability of .55. The coefficients indicated the odds of an individual being a groupman member: (a) labor- exchange membership alone increased the odds of membership by 10.6 times; and (b) Protestant affiliation alone increased the odds of membership by 2.8 times.

servation, the predicted level of treatment would be: In (level of treatment) = ,847. The inverse of the natural log of ,847 equals 2.3.

12. The “completely” treated category includes watersheds in which checkdams were constructed in the principal ravine from the uppermost to the most downstream parcel. The “partial” category includes watersheds in which I I or more checkdams were constructed but treatment was not achieved from up- to downstream. The “scant” category includes watersheds in which IO or fewer checkdams were constructed. Statistical analysis affirmed that the level of action between these different categories was statistically signiti- cant. The mean number of checkdams per hectare was statistically different in each of the categories (p I ,075). as was the number of checkdams per linear meter of the principal ravine (p < ,062).

14. Of seven models tested to identify exposr correlates of action, the final model chosen was: In (level of treatment: l-3) = I .4753 (proportion of landholders who benefited with the construction of checkdams on their land). For example, if 70% of all watershed landholders received checkdams from the collective action, then the correlated level of treatment would be: In (level of treatment) = 1.03. The inverse natural log of 1.03 equals 2.8.

13. The final model chosen was: In (level oftreatment: l-3) = I .2l (proportion of landholders who have adopted contour soil conservation techniques on their own land). For example, if 70% of landholders had previously adopted soil con-

15. Of I I ex ante predictive models tested, the final model chosen was: y (log odds of continued action) = -2.24 + 4.20 (proportion of landholders who had previously adopted soil conservation). Thus, for example, if 50% of the landholders in a watershed were adopters, the probability that its group would choose to continue action would be: y (log odds of continued action): -2.24 + 4.20 (.5) = -.14. A logit of - .I4 converts to a probability of .47. This model predicts that the probability of sustained action closely follows the proportion of landholders who have practical knowledge with the intended output of the collective action.

REFERENCES

Bardhan. P.. “Alternative approaches to the theory of institutions in economic development,” in P. Bardhan (Ed.), The

Economic Theory of Agrarian lnsritutions (Oxford: Clarendon Press, 1989).

Blaikie, P. and H. Brookfield, Lund Degradation and Socieo

(London: Methuen, 1987). Blomquist, W., Dividing the Wuters: Governing

Groundvvater in Southern California (San Francisco: Institute for Contemporary Studies Press, 1992).

Bromley, D. (Ed.), Making the Commons Work: Theory, Practice and Policy (San Francisco: International Center for Self-Governance, 1992).

Bromley, D., “Property relations and economic development: The other land reform,” World Developmenr, Vol. 17, No. 6 (1989), pp. 867-877.

Bromley, D. and D. Chapagain, “The village against the center: Resource depletion in South Asia,” American Journal

of Agriculturcrl Economics, Vol. 64, No. 5 (1984). pp. 868-873.

Ciriacy-Wantrup, S. V. and R. C. Bishop, ‘Common property as a concept in natural resources policy,” Natural

Resources Journal, Vol. IS (l975), pp. 713-727. Dahlman, C., “The problem of externality,” The Journal of km und Economics, Vol. 22, No. I (1979) pp. 141-162.

Dawes, R. and R. Thaler, “Anomalies: Cooperation,” Journnl c!fEconomic Perspecrives, Vol. 2. No. 3 (1988). pp. 187-197.

de Janvry. A., E. Sadoulet and E. Thorbecke, “State, market and civil organizations: New theories, new practices and their implications for rural development: Introduction,” War-/d Developmenf. Vol. 2 I, No. 4 (1993). pp. 565-575.

Dion. G.. “Collective action as conventional behavior,” Masters thesis (St. Paul. MN: University of Minnesota, 1986).

Dragun. A.. “Externalities. property rights and power.”

Journal ofEconomic Issues. Vol. XVII, No. 3 (1983). pp. 667-680.

Fafchamps, M., “Solidarity networks in preindustrial societies: Rational peasants with a moral economy,” Economic Development and Cultural Change, Vol. 41, No. 1 (1992) pp. 147-174.

Foster, G., “Peasant society and the image of the limited good,” American Anthropologist, Vol. 67, No. 2 (1965). pp. 293-315.

Frolich, N., J. Oppenheimer and 0. Young, Polifical

Leadership and Collective Goods (Princeton, NJ: Princeton University Press, 1971).

Gaddis, R. and G. Smucker, “Peasant groups and conservation farming: Evaluation of the Save the Children Project,” Maissade, Haiti, Save the Children Federation (SCF). (Port-au-Prince: SCF, 1988.)

Geertz, C., Agricultural Involution (Berkeley, CA: University of California Press, 1963).

Gleick, P., “Water in the 2lst century,” in P. H. Gleick (Ed.), Water in Crisis: A Guide to the World’s Fresh Water

Resources (New York: Oxford University Press, 1993). Guttman, J., “Can political entrepreneurs solve the free-rider

problem?,” Journal of Economic Behavior and

Organization, Vol. 3 (1982), pp. 357-366. Hardin, G., “The tragedy of the commons,” Science, Vol. 162

(1968). pp. 1243-1248. Hayami, Y. and V. Ruttan, Agricultural Development: An

fnternarional Perspective. Revised Edition (Baltimore, MD: Johns Hopkins University Press, 1985).

Hayek, F., The Constitution of Liberty (London: Routledge and Kegan Paul, 1960).

Herskovits, M., Life in a Hairian Valley (Garden City, NY: Doubleday & Company, 197 1).

Hirschman, A., Getting Ahead Collectively: Grassroots &per- iences in Latin America (New York: Pergamon Press, 1984).


Jodha, N., “Common property resources: A missing dimen- sion of development strategies,” World Bank Discussion Paper No. 169 (Washington, DC: The World Bank, 1992).

Korten, D., Getting to the 2lsr Century: Voluntary Action and the Global Agenda (West Hartford, CT: Kumarian Press, 1990).

Maass, A. and R. Anderson, And the Desert Shall Rejoice: Conjlicr, Growth and Justice in Arid Environments [ 19781 (Malabar, FL: Robert Krieger Publishing Company, 1986).

Maguire, R., “The peasantry and political change in Haiti,” Caribbean Affairs, Vol. 2. No. 4 (1991). pp. l-8.

Marwell, G. and R. Ames, “Economists free ride, does any- one else?,” Journal of Public Economics, Vol. 15 (198 1). pp. 295-3 10.

McCay, B. and J. Acheson (Eds.), The Question of ihe Commons: The Culture and Ecology of Communal Resources (Tucson, AZ: The University of Arizona Press, 1987).

McLain, R. and D. Steinbarger, “Land tenure and land use in Southern Haiti,” Land Tenure Center Research Paper No. 95 (Madison, WI: University of Wisconsin-Madison, 1988).

Metraux, A., Voodoo in Hairi [1959] (New York: Schocken Books, 1972).

Murray, G., “Population pressure, land tenure, and voodoo: The economics of Haitian peasant ritual,” in Beyond the Myths of Culture: Essays in Culrural Materialism (New York: Academic Press, Inc., 1980).

Murray, G., “Terraces, trees, and the Haitian peasant: Twenty-five years of erosion control in rural Haiti,” Report for USAID/Haiti (Port-au-Prince: USAID, 1979).

Murray, G., “The evolution of Haitian peasant land tenure: A case study in the agrarian adaptation to population growth,” Vols. 1 and 2. Ph.D. dissertation (New York: Columbia University, 1977).

North, D., “Economic performance through time,” The American Economic Review, Vol. 84, No. 3 (1994). pp. 359-368.

North, D., lnstiturions, Instirutional Change and Economical Performance (New York: Cambridge U-.iversity Press, 1990).

Oakerson, R., “Analyzing the commons: A framework,” in Proceedings of the Conference of Common Properry Resource Management, National Research Council (NRC) (Washington, DC: National Academy Press, 1986).

Ostrom, E., “Community and the endogenous solution of commons problems,” Journal of Theoretical Polirics, Vol. 4, No. 3 (1992), pp. 343-352.

Ostrom, E., Governing the Commons: The Evolurion of Inslitulions for Collective Action (Cambridge: Cambridge University Press, 1990).

Ostrom, E., R. Gardner and J. Walker, Rules, Games, and Common-Pool Resources (Ann Arbor, Ml: University of Michigan Press, 1994).

Otsuka, K., H. Chuma and Y. Hayami, “Land and labor con- tracts in agrarian economies: Theories and facts,” Journal ofEconomic Literature, Vol. 30 (1992), pp. 1965-2018.

Peters, P., “Is ‘rational-choice’ the best choice for Robert Bates? An anthropologist’s reading of Bates’s work,” World Developmenr, Vol. 21, No. 6 (1993). pp. 1063-1075.

Quine, W. and J. UIIian, The Web of Belief (New York:

Random House, 1970). Runge, C. F., “Common property and collective action in

economic development,” World Development, Vol. 14. No. 5 (1986), pp. 623-635.

Runge, C. G., “Institutions and the free rider: the assurance problem in collective action,” Journal of Politics, Vol. 46 (1984), pp. 154-181.

Runge, C. F., “Common property externalities: isolation, assurance and resource depletion in a traditional grazing context,” American Journal of Agricultural Economics, Vol. 63, No. 4 (1981). pp. 595-606.

SCF, “Rapport Couvrant la Periode D’Aout a Octobre 1989 et de Janvier a Mars 1990,” Save the Children Federation (SCF) (Port-au-Prince: SCF, 1990).

Schaedel, R., “An essay on the human resources of Haiti,” Report to USAID/Haiti (Port-au-Prince: U.S. Agency for International Development [USAID], 1962).

Schlager, E., W. Blomquist and S. Tang, “Mobile flows, storage and self-organized institutions for governing common-pool resources,” Lund Economics, Vol. 70, No. 3 (1994). pp. 294-317.

Schotter, A., The Economic Theory of Social lnsriturions (Cambridge: Cambridge University Press, 1981).

Singleton, S. and M. Taylor, “Common property, collective action and community,” Journal of Theoretical Polirics, Vol. 4, No. 3 (1992). pp. 309-324.

Smucker, G., “Peasants and development politics: A study in Haitian class and culture,” Ph.D. dissertation (Ann Arbor, MI: New School for Social Research, 1983).

Steward, J., Theory of Cultural Change (Urbana, IL: University of Illinois Press, 1955).

Stevenson, G., Common Property Economics (Cambridge: Cambridge University Press, 1991).

Sugden, R., “Spontaneous order,” Journal of Economic Perspecrives, Vol. 3, No. 4 (1989). pp. 85-97.

Sugden, R., The Economics of Rights, Cooperalion and Welfare (Oxford: Blackwell, 1986).

Sugden, R., “Reciprocity: The supply of public goods through voluntary contributions,” The Economic Journal, Vol. 94 (1984). pp. 772-787.

Tang, S., lnstirutions and Collecrive Action: Self- Governance in Irrigation (San Francisco: Institute for Contemporary Studies Press, 1992).

Taylor, M. and S. Singleton, “The communal resource: Transaction costs and the solution of collective action problems,” Politics andsociety, Vol. 2 1, No. 2 (1993), pp. 195-214.

Trouillot, M., Haiti: State Against ihe Nation; The Origins and Legacy of Duvalierism (New York: Monthly Review Press, 1990).

Uphoff, N., “Grassroots organizations and NGOs in rural development: Opportunities with diminishing states and expanding markets,” World Development, Vol. 21, No. 4 (1993), pp. 607422.

Uphoff, N., Learning from Gal Oya: Possibilities for Purriciparory Development and Post-Newtonian Social Science (Ithaca, NY: Cornell University Press, 1992).

Wade, R., Village Republics: Economic Conditions for Collecrive Action in South India (Cambridge: Cambridge University Press, 1988).

White, T. A., “Collective action for watershed management: Lessons from Haiti,” Ph.D. dissertation (Minneapolis, MN: University of Minnesota, 1994).

White, T. and J. Jickling, “An economic and institutional


analysis of soil conservation in Haiti,” Environment Department Divisional Working Paper No. 1992-33 (Washington, DC: The World Bank, 1992).

White, T. A. and C. F. Runge, “Common property and collective action: lessons from cooperative watershed management in Haiti,” Economic Development and Cultural Change, Vol. 43, No. 1 (1994). pp. 1111.

Wittfogel, K., Oriental Despotism: A Comparative Study of Total Power (New Haven, CT: Yale Univer-

sity Press, 1957). Young, 0.. “The ‘new institutionalism’ in international rela-

tions,” Beijer Discussion Paper Series No. 43 (Stockholm: Beijer International Institute of Ecological Economics, 1993).

Zuvekas, C., “Land tenure, income, and employment in rural Haiti: A survey,” Working Document Series: Haiti General Working Document No. 2 (Washington, DC: USAID, 1978).

the emergence and evolution of collective action: lessons from watershed management in haiti

Documents