8.3 Study Area and Data Set and Study Area
The dataset used in this study was collected by Thammasat University for the
province of Chiang Mai.
17 The data were collected for the National Economic and
Social Development Board (NESDB). Data for the study were collected for six
rounds, once every 2 years (biennially) starting in 1986 and then 1988, 1990, 1992
and 1996, for the province of Chiang Mai. Villages included in the study dataset all
responded that they lay within Forest Reserves at least once during this 11 year
study period. All villages in the dataset are registered with the Village Directory of
the Department of Local Administration (DOLA). However because forest reserve
boundaries changed a lot, all villages did not lie within Forest reserves during the
study period. Inhabitants of Forest Reserve villages are mostly hill tribe people who
are poor, and live in villages that are remote and have poor infrastructure.
Forest Reserve residents grew mainly three crops during this period – paddy rice,
upland rice and soybeans. Thailand is among the largest growers of paddy rice and
its biggest exporter. But rice is also a staple. Most villages in the study sample grew
paddy rice. On average upland rice and soybean were grown by 25 % and 26 % of
villages respectively.
The resulting panel dataset is unbalanced. Of the 670 villages that appear at least
once in the dataset, 255 (38 %) are present for all six rounds in the panel; in contrast,
124 villages are present for only one round (Tables 8.2 and 8.3). Attrition in panel
data is common: villages may choose to not participate in certain rounds or may not
be asked to participate in certain rounds for several reasons (e.g. lack of resources
with the survey agency). It is important to understand the cause of attrition or
selection.
18 Villages that are surveyed and respond in all six rounds are the single
largest group in the dataset (38 %). The second largest group is the villages that
occur only once. These constitute 18.5 % of the villages.
In the survey conducted by Thammasat University, village communities were
asked in every round of survey (there are a total of six rounds) if they had secure
property rights (‘What land title did you have?’) Using this information and the
mapping above, from the type of land titles to the security of these land titles, I
examine if these perceptions change over the different rounds. In Tables 8.2 and
8.3, I examine these responses for each round in the panel dataset. Table 8.2 shows
that 62 % or 413 villages in the dataset never believed they possessed secure rights
to village land. In contrast, only 36 villages claim to have secure rights during the
entire study period (for all six rounds). For the remaining villages, the status of their
land titles ‘flips’ from year to year. So for example Table 8.3 shows that 33 villages
in the dataset were surveyed for all six rounds of data collection, in five of six of
those rounds believed that their land title was insecure. They only report a secure
17 The larger dataset consists of 784 villages.
18 Missing observations in a panel data may not be randomly missing and, if so, estimators may be
inconsistent. Ignoring attrition and using a balanced dataset, as is common practice, may lead to
inconsistent estimates (See Heckman 1976; Nijman and Verbeek 1992).
8 Using Mixed Methods to Assess Trade-Offs Between Agricultural Decisions. . .
137
The dataset used in this study was collected by Thammasat University for the
province of Chiang Mai.
17 The data were collected for the National Economic and
Social Development Board (NESDB). Data for the study were collected for six
rounds, once every 2 years (biennially) starting in 1986 and then 1988, 1990, 1992
and 1996, for the province of Chiang Mai. Villages included in the study dataset all
responded that they lay within Forest Reserves at least once during this 11 year
study period. All villages in the dataset are registered with the Village Directory of
the Department of Local Administration (DOLA). However because forest reserve
boundaries changed a lot, all villages did not lie within Forest reserves during the
study period. Inhabitants of Forest Reserve villages are mostly hill tribe people who
are poor, and live in villages that are remote and have poor infrastructure.
Forest Reserve residents grew mainly three crops during this period – paddy rice,
upland rice and soybeans. Thailand is among the largest growers of paddy rice and
its biggest exporter. But rice is also a staple. Most villages in the study sample grew
paddy rice. On average upland rice and soybean were grown by 25 % and 26 % of
villages respectively.
The resulting panel dataset is unbalanced. Of the 670 villages that appear at least
once in the dataset, 255 (38 %) are present for all six rounds in the panel; in contrast,
124 villages are present for only one round (Tables 8.2 and 8.3). Attrition in panel
data is common: villages may choose to not participate in certain rounds or may not
be asked to participate in certain rounds for several reasons (e.g. lack of resources
with the survey agency). It is important to understand the cause of attrition or
selection.
18 Villages that are surveyed and respond in all six rounds are the single
largest group in the dataset (38 %). The second largest group is the villages that
occur only once. These constitute 18.5 % of the villages.
In the survey conducted by Thammasat University, village communities were
asked in every round of survey (there are a total of six rounds) if they had secure
property rights (‘What land title did you have?’) Using this information and the
mapping above, from the type of land titles to the security of these land titles, I
examine if these perceptions change over the different rounds. In Tables 8.2 and
8.3, I examine these responses for each round in the panel dataset. Table 8.2 shows
that 62 % or 413 villages in the dataset never believed they possessed secure rights
to village land. In contrast, only 36 villages claim to have secure rights during the
entire study period (for all six rounds). For the remaining villages, the status of their
land titles ‘flips’ from year to year. So for example Table 8.3 shows that 33 villages
in the dataset were surveyed for all six rounds of data collection, in five of six of
those rounds believed that their land title was insecure. They only report a secure
17 The larger dataset consists of 784 villages.
18 Missing observations in a panel data may not be randomly missing and, if so, estimators may be
inconsistent. Ignoring attrition and using a balanced dataset, as is common practice, may lead to
inconsistent estimates (See Heckman 1976; Nijman and Verbeek 1992).
8 Using Mixed Methods to Assess Trade-Offs Between Agricultural Decisions. . .
137