Abstract

One of the issues encountered during survey implementation at Statistics Indonesia (BPS) is the high frequency of surveys conducted within a single year. Consequently, there has been a growing issue regarding the integration of several surveys into a unified survey. Survey integration involves combining various survey activities, which are typically conducted independently, to be implemented simultaneously, both in terms of instruments and survey methodology. For example, in the Agricultural Economic Survey (SEP), the economic activities of each agricultural unit are queried, covering food crops, horticulture, estate crops, livestock, forestry, fisheries, agricultural services, seedling production, and so on.
Traditionally, most agricultural surveys at BPS employ Probability Proportional to Size (PPS) to select enumeration areas (primary sampling units), followed by systematic sampling to select a number of enterprises (ultimate sampling units). With this method, some desired target units may not be obtained, particularly in rare cases.
To address the challenges posed by surveys with multiple targets, the Multivariate Probability Proportional to Size (MPPS) method was developed. MPPS presents a potential solution to the aforementioned issue. Unlike PPS, which uses a single measure of size, MPPS employs several measures of size in the sampling process. Consequently, with MPPS, each target sample has a certain probability of being selected.
However, MPPS has not yet been applied to surveys at BPS, thus necessitating a study to evaluate the feasibility of implementing MPPS in the relevant surveys. Therefore, a comparison will be made between the method commonly used at BPS (PPS) and the MPPS method. The comparison will be conducted using sample selection simulations, followed by an evaluation of several indicators such as relative bias, mean squared error, and relative efficiency.
Based on the simulation results, it is evident that MPPS produces a better sample representation compared to PPS, especially in agricultural activities where the unit population is small within a certain domain. Moreover, the indicators generated by the simulation show no significant difference between PPS and MPPS.