The study focused on estimating food poverty incidence of households in agriculture and fisheries sectors in the Philippines. Three estimation procedures were used: direct, regression-synthetic and empirical best linear unbiased prediction (EBLUP). Direct, estimation uses only data from survey while regression-synthetic and EBLUP borrow strength from consensus and administrative data sets. Using the 2003 Family Income Expenditure Survey, direct estimates of provincial food poverty incidence were computed. For both sectors, incidences of food poverty have wide range and vary largely. However, majority of provinces have high incidences of food poverty among households agriculture sector and low incidences of food poverty among households in fisheries sector. Likewise, majority of the direct estimates are not reliable. Using data from 2000 Census of Population and Housing, 2000 Census of Agriculture and Fisheries, 2000 and 2003 Field Health Service Information System Annual Report, 2000 Enrolment Report and 2000, 2001 and 2003 Vital Statistics Resport, the best predicting model to generate regression-synthetic estimates for agriculture and for agriculture and for fisheries sectors were formulated. Meanwhile, direct and regression-synthetic estimates were combined to generate EBLUP estimates using weight relative to total variation. For both sectors, regression-synthetic estimates slightly vary with the direct estimates while EBLUP estimates are close with direct estimates.