Small area estimation for rice and corn statistics

The role of quality, reliable and effective micro-level statistics to formulating developmental programs and policies can not be underestimated. The study looked into the possibility of using results of existing survey to generate municipal agriculture related characteristics based on results of existing surveys. Several estimators namely: Synthetic, Direct, Ratio-type, Composite and Imputation-like estimates were empirically compared using the results of the 1991 Census of Agricultural and Fisheries (CAF) in the provinces of Bukidnon, Isabela and Laguna [Philippines]. The sampling design of the Rice and Corn Production Survey (RCPS), conducted quarterly by the Bureau of Agricultural Statistics (BAS) was employed in generating samples and eventually estimates of agriculture related characteristics at the municipal level. Results show that the Ration and Imputation-like estimates provided more accurate and precise estimates. Mean Percent RAD [relative absolute difference] as low as 2.1 percent for these estimates was observed. However, the ratio-type (Ratio separate and Ratio combined) estimators showed more consistent reliable estimates. Also, it is recommended that the nature and type of municipalities must first be established as these estimates failed to provide reliable estimates for a few municipalities. The study recommends: a) future studies to determine the potentials of the ratio-type and imputation estimators to other provinces; b) further research to improve the properties of the ratio-type estimates and establish the limits of acceptable differences in municipal and provincial ratios, since the ratio-type depends largely on the assumption that municipal ratios of the variables are equal to the provincial ratio; c) future research to explore and exploit the positive property of the imputation techniques; and d) the behavior of the estimates seen to yield good results be studied for other survey design with varying sample size and stratification variables