significant causal factors, and that the investigator should ensure that the selected cases contain at least some variation in terms of observed outcomes.50 Therefore two factors are likely to be critical in research design: awareness of the variables that are theoretically relevant, and particular knowledge of the case(s) to be researched so that the theoretically relevant variables can be operationalized. For example, when constructing a research design where the variables of interest have to do with mechanisms of monitoring and sanctioning, it would be important for the researcher to be familiar with the different forms of monitoring that groups can use. The presence or absence of a guard may only be indicative of the presence or absence of third-party monitoring, and may reveal nothing about whether the group being studied has adopted monitoring mechanisms. Other forms of monitoring could include mutual monitoring, and rotational monitoring where households in a group jointly share the tasks related to monitoring and enforcement.

The information presented in Box 2-5, organized into four major categories, can therefore be useful in the creation of a research design and for case selection. Given a particular context, it can help in the selection of the variables that need closest attention in the selection of cases. For example, if the cases to be selected lie in the same ecological zone and represent the same resource type, then variables related to resource characteristics may not be very important for case selection. The obvious tradeoff for this reduction in the number of variables is that the research will provide little or no insight into the effect of differing levels of predictability on institutional sustainability. If the research objective were to understand the effects of unpredictability, then it would be imperative to select cases where resource output varied from highly predictable to unpredictable.

However, a large-N study of commons institutions that incorporated more than 30 independent variables and their interactions would require impossibly large samples and entail astronomically high costs. Researchers conducting such studies are likely to face complex problems in interpreting the data and stating their results, even if they could collect information on thousands of cases. Even if it were possible to create purposive samples of cases that accommodated variation on more than 30 causal factors and their interactions, the problems related to contingent and multiple causation will not fade away. The problems of contingent and multiple causation make it necessary for researchers of the commons to also postulate causal relationships among the critical theoretical variables they have identified, and then conduct structured studies that examine the postulated causal links among variables.

Larger sample sizes and statistical analyses also do not constitute a global answer to the problem of many independent variables for another reason. As argued earlier in the chapter, the set of variables that constitutes the context is potentially infinite. Multiplying the number of cases may simultaneously imply an increase in the number of contextual variables that affect outcomes in a specific selected case. Because conclusions from empirical analysis cannot conceiv-

The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement