HOUSEHOLD
The headman in each village held a copy of the official Community Development Dept enumeration of households in the village. From this list 15 households were selected at random. Occasionally a houshold was found to have moved entirely or disbanded, for example, due to death of head. Interviews were voluntary and could be ended at any time, but after an explanation of the purpose of the survey few households refused to participate. Still, don't know or refused to answer is a valid code ?see the code book for details. All household replacements were selected at random from the same initial village list. All together, 2880 households were interviewed in this initial cross sectional survey .

KEY INFORMANT
The headman from each village was interviewed.

BAAC GROUP
Up to two joint liability groups of the BAAC were interviewed in each village. Some villages had only one group, and some had none. A total of 262 groups were interviewed in this initial cross-sectional survey. The head of the group answered the questions.

FINANCIAL INSTITUTIONS
The responsible member of the village team interviewed the head of the village and other officials to get an enumeration of all financial, community funds operating in village?hese included production credit groups, local rice banks, poverty eradification funds, womens' groups, rotating credit associations, buffalo banks, and other local funds. Occasionally the fund operated at a more aggregated, tambon level, and these were also included. All interviews were with the head of the local institution and occasionally with other committee members. All local records and account books were made available at the time of the interview, to answer questions about number of members since the institution's founding, for example. These records were all Xeroxed in the field. Not all villages were found to have an institution. Altogether there are 161 institutional interviews in the initial, cross sectional survey.

SOIL
One soil sample per household was collected from the second plot named by the head of the household when asked to name all of his or her plots. We did not select the plot randomly because of the difficulty in the use of random number tables by field enumerators, and we used the "second-named" method as a substitute. The plot was visited and located with a Garmin GPS 12. The top 20cm of soils were sampled with a t-handled probe at the five nodes of a W-shaped transect across the field, and the five samples were combined in a sample bag. The enumerator took photographs from the center of the field in the four cardinal directions, and one photograph diagonally across the field. The enumerator also described the general characteristics of the surrounding landscape, including vegetation present, topography, proximity to water bodies, etc. A special soil questionnaire was administered to the farmer on site.