Dynamics of technical and environmental efficiency in rubber and oil palm land use systems

With the data collected in the first and the second round of the survey, an (still short) panel will allow us to look into efficiency changes over time. This will be done for both the technical efficiency analysis based on the full household survey (described in more detail in C07), and the environmental efficiency analysis based on the plot level survey carried out in this project. The third round of this survey will be carried out in the second half of 2016 for about a third of the households (n=215) covered in the full household survey. We will continue to collect data at the plot level on several environmental indicators, e.g., plant abundance data and species richness, and biomass measurements (as described in B06). Additionally soil samples of the plots will be collected, and analysed in cooperation with A04, allowing an insight on the dynamics of soil exploitations. Since the data collection focuses on a subset of the full household survey, we can easily combine the environmental indicators with the socio-economic information in order to assess environmental efficiency of smallholders in the presence of undesired by-products, such as loss of biodiversity or soil erosion. The repeated measurements of this data will allow us to assess the change of soil fertility and biodiversity due to management choices, and participation in specific marketing channels.

The methodological foundations will build on the successfully applied approaches from the first phase; stochastic frontier analysis using flexible functional forms has proven a viable tool for adequately describing the technology. For the analysis of technical efficiency, a production function based translog frontier is suitable.

For the analysis of environmental efficiency, a multi-output representation of technology is required. Both output and directional output distance functions are useful tools in this context.

The findings from the first phase indicate several important pathways through which technical efficiency is affected. Further analysis is required to check the robustness of these findings, and to examine the potential role of changes in efficiency over time. Such changes might be triggered on the one hand by learning processes and on the other hand by changes in some of the variables which we identified as drivers of technical efficiency differences. Such variation over time will also allow us to disentangle the role of learning from other drivers of technical efficiency.

With the completion of the third round of data collection in Jambi, we will examine productivity growth and dynamic technical efficiency. Dynamic technical efficiency has been analysed in agricultural production before (e.g., Serra et al. 2011) but there are no prior applications to perennial crops. In particular, the consideration of replanting, where actual experience is still scarce in Jambi province because the development of smallholder based oil palm cultivation only started in the nineties, is an important element that has not been considered before. Although the frequency of actual replanting is expected to be still low in the second phase of CRC 990, we will develop an adequate conceptual model of the existing approaches to dynamic efficiency analysis, building on experience from other Sumatran provinces. The necessary information will be collected with the support of Prof. Dr. Zulkifli Alamsyah by interviewing key stakeholders. We plan to talk to representatives of both oil palm plantation managers and producer groups in North Sumatra and Riau. In both provinces, replanting takes already place at a substantial scale so that information on the costs for clearing and replanting and on the benefits (mainly through higher fresh fruit bunch yields) can be collected.

Similarly, the necessity to incorporate non-marketed goods in an analysis of productivity growth has received increasing attention in the literature. In many macroeconomic oriented studies of productivity growth, pollutants and other undesirable outputs are often included. At the household level, only few studies do similarly because data on non-marketed outputs are often difficult to obtain at this scale. With our unique dataset, we are able to overcome this limitation and can obtain reliable estimates of the trade-off relations between environmental ecosystem services like biodiversity or soil quality and economic provisioning services (rubber and palm oil production, respectively). In the second phase, we will also extend our analysis of environmental efficiency by incorporating further environmental aspects for which a materials balance perspective is more appropriate. This approach is specifically tailored to situations where the negative environmental impact stems from emissions or leakages caused by inputs used in the production process. The environmental ´bad´ hence originates from the quantity of input use which is not taken up in the output. For a single environmentally detrimental input x (e.g., nitrogen fertilizer) used to produce a single output q, the negative emission or leakage b can be described (Hoang and Nguyen 2013) by b = ax - ßq with a and ß as technological coefficients (e.g., in the nitrogen case, the nitrogen contents of the fertilizer and the output, respectively).


Bernhard Dalheimer is Doctoral Speaker of Group C