Page 381 - Invited Paper Session (IPS) - Volume 2
P. 381

IPS280 Erniel B. Barrios et al.
                           Model-assisted estimation using panel data
                           and data from independent samples for the
                            Philippine Family Income and Expenditure
                                              survey
                            1
             Erniel B. Barrios , Divina Gracia L. del Prado , Mae Abigail C. Oberos , Desiree
                                                                              2
                                                      2
                                            2
                                     Robles , Jo Louise Buhay
                                                             2
                               1 School of Statistics, University of the Philippines
                          2 Statistical Methodology Unit, Philippine Statistics Authority

            Abstract
            The domains for the Philippine Family Income and Expenditure Survey (FIES)
            are the regions. Design-unbiased estimation for the provinces (smaller areas)
            would necessitate an increase the sample size to about four times. Considering
            further  that  FIES  is  conducted  semi-annually  (at  the  end  of  reference
            semester),  using  the  same  sample  households  for  the  survey  year  and
            administered  with  a  lengthy  questionnaire,  response  burden  and  refusals
            increases along with problems with willingness of interviewers to complete the
            task for the entire survey duration. This problem is addressed by keeping the
            regions as survey domains but partially rotating the samples in the two rounds
            of  data  collection.  This  study  then  proposes  a  model-assisted  estimation
            procedure in combining panel data and data from independent samples to
            generate annual estimates at the provincial level. Simulation studies indicate
            the advantages of model-assisted estimation procedure over design-unbiased
            method in generating estimates at the provincial level. Furthermore, sample
            size  requirement  using  model-assisted  approach  is  only  about  half  of  the
            required sample size for the design-unbiased method for reliable provincial
            level estimates.

            Keywords
            Rotated sample; skewed population; cut-off sampling

            1.  Introduction
               Nationwide surveys intended to generate official statistics often incur huge
            cost and complex logistical issues in survey operations, hence, often constraint
            statistical organizations to consider bigger areas (in some cases, national) as
            the domains. However, more granular data is also needed to monitor progress
            among countries towards achievement of sustainable development goals. The
            Philippine Family Income and Expenditure Survey (FIES) is the source of many
            indicators for various goals, the most important of which are poverty-related
            indicators. FIES has always been a panel survey where same households are
            visited every semester. While this has been observed to generate “reliable”
            data  for  income  and  expenditures,  it  easily  causes  respondents’  fatigue


                                                               368 | I S I   W S C   2 0 1 9
   376   377   378   379   380   381   382   383   384   385   386