Introduction. Economic development is often accompanied by environmental degradation，partly because of the reduced time the now pro-urban farmers spend on resource management. Villagers and officials equally being in favour of development，this phenomenon is caused by less capital，money and labour being dedicated to rural resource management.

To date，the commons literature has attributed this“lost interest”problem in the urban fringe areas to increasing opportunity cost，and this hypothesis enjoys a wide support by empirical studies（Baland and Platteau，1999；Lam，2001）.

Yet，the villages’ development paths in real world are more complicated than such a uni-dimensional interpretation，particularly in terms of income distribution within a village. For example，a“rich”village，measured by its average household income，can be comprised of one hundred equally rich households or of one extremely rich entrepreneur and ninety-nine poor peasants. Clearly，the villagers’ overall attitude towards resource management would be different between these two villages.

In all likelihood，the fact that such complexities have been little discussed is attributable to lack of appropriate methodologies. As an attempt to overcome this shortage，this study proposes a framework with which one can test more detailed hypotheses regarding the relationship between economic development and people’s attitudes towards collective actions.

In advance to the main analysis，however，the conventional theory of opportunity cost-other things being equal，a village’s economic development will result in less success in communal resource management-is tested. This is done with a new and original dataset from South China.

Data. The data used in this study were collected by fieldwork at Xishan District，a municiPál within the jurisdiction of Kunming City，in the Southwestern Chinese province of Yunnan. The District has 330000 permanent residents in its 1050 square metre area. Amongst them，130000 people are engaged in agriculture，making the district a typical economy in transition，where conflicts of interest tend to arise frequently between farmers with different philosophies or economic status.

Questionnaire-based interviews were conducted with cadres of 111 natural villages. Almost always coinciding with the quasi-official unit of village members group，natural villages are the smallest unit of decision making on communal activities such as water management，forest management and festivals，and therefore the most suitable sample unit for the present research on resource management. Sample villages were carefully selected from 33 of the District’s 35 village committees so that they would become a good replica of the District’s entire agricultural population.

Basic characteristics of the sample are summarised in Table 1. With 45% of the sample villages predominantly inhabited by one of the non-Han minorities，Yunnan’s racially diverse demographics makes the sample suitable for this research，as communities from different backgrounds are likely to be regulated by different types of social norms. Villages with greater water availability typically grow rice as their primary product，whereas those with less water tend to farm corn instead. This variety of irrigation methods reflects the District’s undulatory geography，with the immediate source of irrigational water ranging widely from river and lake to natural spring. Amongst all sources，reservoirs are the commonest immediate source of water. According to the interviews，about half（55%）of the sample villages do not have“sufficient”water for farming every year，an answer that strongly suggests the need for a strict water management scheme.

Results-preliminary test. The purpose of this empirical analysis is to identify biogeographical and socioeconomic factors that affect the success or failure of villagers’ cooperative resource management and to test that the opportunity cost hypothesis sustains in the research area.

The regressand of the estimated equation is a binary variable representing villagers，participation in communal water management. Here，the village’s communal water management，namely cleaning，weeding and other maintenance activities of water canals，is defined as being successful if participation in such activities is mandatory for at least one person per household and if that participation rule is strictly observed. Conversely，if a village has no rule dictating villagers mandatory participation in collective activities，or even if it does such rules are not strictly being observed，that village’s communal resource management is defined as being a failure. Preceding studies with a similar approach，which are motivated by the porposition by Agrawal（2001）and Bardhan（1993b）that community driven environment management is successful only under a set of necessary conditions，include Bardhan（2000），Dayton-Johnson（2000），Fujiie，Hay ami and Kikuchi（2005），and Lam（1996a）.

Five independent variables were chosen for the present estimation. All of them represent factors generally considered to be imperative for the success in communal resource management：relative importance of the resource in question（Baland and Platteau，1999，Lam，2001），economic heterogeneity within the community（Aggarwal and Narayan，2004；Baland and Platteau，1997；Baland and Platteau，1998；Banerjee，Mookherjee，Munshi and Ray，2001；Dayton-Johnson and Bardhan，2002；Jones，2004；La Ferrara，2002；Molinas，1998；Varughese and Ostrom，2001），certainty about water availability（Gaspart，Jabbar，Melard and Platteau，1998；Nugent and Sanchez，1998；White and Runge，1995；Wilson and Thompson，1993），culturally specific social norms（Sugden，1984），and cultural heterogeneity within the community（Alesina and La Ferrara，2000；Ostrom，2000）. The variable of interest for the present purpose is the relative importance of the resource；the hypothesis to be tested is that，as the opportunity increases，villages become less committed to communal resource management. More detailed definitions of these variables are given in Table 2，together with their basic statistics.

The results of the logit and probit estimations are given in Table 3. Both estimation methods produce almost identical results，widely accepting（not rejecting）the hypothesis at 95% significance level in one-tailed test. The four other independent variables，serving as controls for the present purpose，also show the results in line with theoretical predictions；the only exception is the estimator for economic heterogeneity，which shows the predicted sign but with a relatively weak statistical significance.

Results-main test. While the opportunity cost hypothesis seems to capture well the general tendency of villages，developmental path，more detailed information is necessary to establish economic policies that suit each village separately. The unidimensional nature of the hypothesis limits its policy implications，and hence arises the necessity to create a research framework that are inclusive of other aspects of the village.

To see this point in a clearer perspective，imagine a village that comprises of two groups of people，namely P₁ and P₂. Each player（group）is either willing or reluctant to spend time in communal resource management，depending on various other interests in life. The former case is depicted as t_h，the latter t_l.

When P₁=P₂=t_h，both parties happily participate in the collective action program. When P₁=P₂=t_l，they happily agree to discontinue the program and hire graziers，water managers or other resource professionals to replace their management duties，or possibly，abandon the management programs altogether. The trouble arises，however，when P₁=t_h and P₂=t_l or vice versa-the case in which only one of the two players wants the program to continue.

Naturally these three scenarios will result in a different outcome of this non-cooperative game，as is summarised in Table 4. The purpose of the empirical test is to econometrically test these predictions so that future theoretical works can utilise this useful hypothesis. This is made possible on the explicit recognition that the two equilibria for economically heterogeneous societies，（P₁=P₂=t_h and P₁=P₂=t_l）have considerably different dispositions-even though few previous studies reflect this fact and rather measure economic heterogeneity with a single index number.

The working hypothesis for this empirical test is that the course of actions the two players eventually take in this conflicting situation depends on three factors：P₁’s despair for the collective action program，P₂’s reluctance about it，and the relationship between P₁ and P₂. The latter represents the tacit rules that govern appropriate behaviours in group settings，are and often called social norms in the economic literature. More often than not，social norms contradict one’s economic sense.

To test the hypothesis，the multinominal probit method was employed to the same data that were used for the preliminary test. The regressand takes the value of either 1，2 or 3，depending on the village’s form of water management hypothesised（Table 4）；at Case 1 villages everyone participates in mandatory canal cleaning，at Case 2 villages some households do not participates in it，and at Case 3 villages there is no such system for water channel management that requires the participation of every household. As for regressors，the same five variables from the preliminary analysis were used to make the results comparable.

Table 5 summarises the results of the estimation in two parts：the estimators for the occurrence probability functions and the estimated marginal effects on them. The table suggests that all determinants are equally influential on people’s decision to participate in collective actions. As predicted by the theory of opportunity cost，the industrialised villages tend to have either people shirking or the participation rules amended. More importantly，intra-village discrepancy in wealth will more likely see the discontinuation of the mandatory scheme than rich people simply ignoring the rule，possibly implying that the cost of system change is lower than intra-village frictions.

In addition，the finding from the preliminary test that there is a significant degree of difference in villagers’ attitude towards collective actions owing to their ethnic characteristics as well as ethnic mix，has been maintained in the present analysis. People in multinational villages show the tendency here to simply skip mandatory programs，saving time and energy to discuss the modification of the rules.

Overall，it seems reasonable to conclude that the working hypothesis receives a sufficient statistical support to justify further theoretical investigations. Of course，further quantitative investigations are also essential to substantiate the hypothesis.

Conclusions. Needless to say，these results should be regarded as preliminary，given that the estimation does not have a structural model（beyond Table 4）behind the reduced-form equation. In this sense，the investigation has been of a heuristic nature；one can only hope that the accumulation of such results will eventually lead to construction of more general economic theories. Then again，theoretical modelling can be a less meaningful exercise when the model has to deal extensively with immeasurable concepts such as social norms and local culture. It remains to be seen whether such theoretical development will occur in the near future.

Table 1 Basic characteristics of sample villages

Table 2 Basic statistics for independent variables

Notes：

（1）The details the variables are as follows. Agriculture’s GDP：the estimated share in percent. Wealth inequality：see the next note. Certainty：whether villagers have certainty about water availability. Han people：whether the village is primarily habituated by Han people. Multinational：whether the village comprises of more than one ethnic group.

（2）Wealth inequality is measured by standard deviation of the following statistics：H₀+10H₁+15H₂+20H₃，where H₀ is the number of mud-walled houses in the village and（H_i（i=1，2，3）is the number of i-storey concrete-walled houses. The coefficients 1，10，15，20 roughly equate the proportion of capital required to build each type of the house.

Table 3 Determinants of successful communal resource management

Table 4 Participation in mandatory collective actions

Table 5 Determinants of collective actions

Notes to the table：

（1）The dependent variable is either 1（everyone participates in mandatory canal cleaning），2（not everyone participates in mandatory canal cleaning）or 3（there is no mandatory canal cleaning）. Also refer to Table 4.

（2）For the details of independent variables，see notes for Table 2.

（3）Numbers in parentheses are the standard errors of the estimators.