2. RESULTADOS 2021
2.5. Eje: Agua Potable y Saneamiento Cloacal
19 The agricultural household survey, PATANAS might be an exception. However, the second round of the panel survey would only have been available only in late 2010, and moreover, the geographic coverage of this survey means that it would most likely only cover a few households engaged in palm oil production.
20 Area data include immature, mature and damaged plantations.
21 I consider that the quality of the data is largely satisfactory. However, I dropped two observations from the analysis that were likely outliers, i.e. had inconsistent growth in the production level compared to the previous year.
plantations. However, I primarily focus on smallholder production, as this may have a stronger relationship with household expenditure than production on large plantations.
Another advantage of using smallholder data is that it covers a longer time period (2003-2006) than do other available production data. Moreover, the smallholder data covers the whole area of Indonesia, including over 350 districts (using the 2002 definitions of districts). Following a decentralization process, a number of new districts were created in the 2000s. In this study, 2002 is therefore taken as a base year.
According to the national district-level data, the average area of smallholder plantations is 5,800 hectares and the average production is around 11,500 tonnes. Between 2003 and 2006, the average district smallholder oil palm plantation area in Indonesia increased by approximately 50 per cent, from 4,500 hectares to 6,800 hectares. However, approximately 67 per cent of the districts in Indonesia do not have any smallholder oil palm plantations.
Restricting the sample to only those districts that do have smallholder plantations, the average district area of smallholder plantations is 18,100 hectares and the average production is around 41,400 tonnes, respectively. However, if we focus on the main production regions, in Sumatra only 28 per cent of the districts do not have any smallholder plantations, while the corresponding figure in Kalimantan is 31 per cent. The average district smallholder plantation area in those districts that do have smallholder plantations is 22,600 hectares in Sumatra and 11,000 hectares in Kalimantan. Table 4.3 below presents summary statistics on smallholder area and smallholder production for all districts in Sumatran and Kalimantan irrespective of whether the districts have smallholder plantations.
Table 4.3. Descriptive statistics for district-level smallholder data in Sumatra and Kalimantan, years 2003-2006.
Average (2003-2006) 2003 2006
Smallholder area (ha) 13,300 10,200 15,700
Smallholder production (tonnes)
26,500 19,600 35,500
Nationwide, smallholder oil palm plantations represent, on average, one per cent of the district land area, but there are large differences across the regions. In Sumatra, there are districts where up to 15 per cent of the area is covered by smallholder plantations, while in Kalimantan, the proportion varies from 0 to 5 per cent.
In addition to the smallholder data I employ the complete production data (both smallholders and large plantations) for selected provinces in Kalimantan.22 The area data cover the provinces of West Kalimantan, South Kalimantan and East Kalimantan (but not Central Kalimantan), and the production data cover the provinces of West Kalimantan and South Kalimantan (but not Central Kalimantan and East Kalimantan). Both the production data and the area data cover the time period 2004-2007, although data for the year 2004 is only available for West Kalimantan.23,24 In these provinces the average district area devoted for oil palm plantations is 34,000 hectares when focusing only on those districts that do have plantations. Table 4.4 below presents summary statistics on total area of plantations and total production for selected provinces in Kalimantan irrespective of whether the districts have oil palm plantations.
22 Area data cover immature, mature and damaged plantations.
23 Due to the inability to get comparable household expenditure data for the year 2008, the year 2007 is only used in health specifications.
24 I also have data on complete production (both smallholders and large plantations) for the whole of Indonesia for the years 2005 and 2008. However, due to an inability to get household data for 2009 the use of these data will be left for future research.
The annually implemented SUSENAS survey is a nationally representative household survey. Each year, in late February or early March, a new set of roughly 200,000 households are interviewed as part of the core of the national socio-economic census (SUSENAS). The dataset includes results from a small consumption module, consisting of 15 food items and 8 non-food items, that combines purchased and own-produced items.
The household per capita expenditure is deflated to 2007 prices using the consumer price index of the province capital. In addition there is data on household characteristics, such as education and health status of the household members, and housing conditions. The average log monthly per capita expenditure is 12.57 over the surveys, equivalent of 360,000 Indonesian rupiahs, or US$40. Very few household members report symptoms of asthma;
only 1.7 per cent of the individuals aged 10 and above reported suffering from asthma over the preceding month, but the incidence of asthma increased from 1.4 per cent to 2.8 per cent over the survey period.
In addition to the core datasets, two other data sets are employed in order to construct the instruments for the IV estimation. First, PODES data is used to construct a historical measure of palm oil production in a district. The PODES data come from a survey of over 65,000 villages throughout Indonesia, which rotates themes such as agriculture, economic and population. The 2003 PODES Agriculture Survey includes a section on village-level plantation crop production, and both the area and production of the five most important plantation crops are listed here.25 It is notable that the 2003 PODES survey was implemented in 2002 and therefore the production data refer to the year 2002. This is important as my study period starts in 2003. I aggregate palm oil production levels and oil palm plantation areas in all villages in a district, in order to construct a historical district-level measure of palm oil production and plantation.26 Importantly, the 2003 PODES survey covers the whole geographic area of Indonesia.
Satellite data on district forest cover is used to construct an alternative instrument for oil palm plantations and palm oil production. The proportion of district area covered by forest
25 As there are not too many plantation crops my judgement is that if oil palm is not listed among the five most important crops then the cultivation of oil palm is likely to be only a marginal activity.
26 It is not stated specifically whether the area and production refer to smallholders or to total production.
However, as the village head would probably not have access to data on area and production for private companies, it is therefore likely that this measure best refers to smallholders.
was provided by the Geographic Information Science Center of Excellence, South Dakota State University (see Broich et al., 2010 for further information). The district area in square kilometres was then employed in order to calculate the forest area of a district in 2000. As discussed earlier, large areas of forest have been converted into oil palm plantations and, therefore, district forest area prior to the study period is likely to be a good source of exogenous variation in palm oil production.27 Because the forest area is calculated prior to the study period, the possible direct correlation between household welfare and forest cover is eliminated. However, despite of these advantages, there are some disadvantages related to this instrument. First, forest data used in this study is not pure, in that it also covers mature plantations. The satellite data employ a definition of forest as areas where there is a tree canopy exceeding 25 per cent coverage, and greater than five meters in height.
Unfortunately, when using this definition, it is not possible to distinguish between natural forest and mature oil palm plantations. Second, data on forest cover are only available for Sumatra and Kalimantan. However, as Sumatra and Kalimantan are the main production regions, and together account for approximately 95 per cent of the national production I argue that this limitation is unlikely to bias the results to any significant extent.
Finally, I have monthly rainfall data covering the period of 1951-2007. This data is taken from two sources. Data for the period 1951-1998 was provided by Kirono et al.
(1999), while that for the period of 1999-2007 come from the Indonesian Meteorological Agency (BMG). These data are used in robustness checks in sections 4.5.1 and 4.5.2.