Integrated Household Survey 1996-1997

The sample size of the 1996-97 PIHS survey was fixed by the Federal Bureau of Statistics (FBS) at a level high enough to allow most estimates to be obtained for each province and region (urban/rural). In addition, for key variables, the aim of the sampling strategy was to provide estimates with confidence intervals of plus or minus 5 per cent. In the case of the 1991 PIHS, however, the sample size for the survey was considerably smaller than that for the 1995-96 and 1996-97 PIHS surveys.

A two-stage random sampling strategy was adopted for the 1995-96 and 1996-97 PIHS surveys. At the first sampling stage, a number of clusters or Primary Sampling Units (PSUs) were selected from different parts of the country. Enumerators then compiled lists of all households residing in the selected PSUs. At the second sampling stage, these lists were subsequently used to select a fixed number of households from each PSU for interviews using a systematic sampling procedure with a random start. This two-stage sampling strategy was used in order to reduce survey costs, and to improve the efficiency of the sample. The number of PSUs to be drawn from each strata in the first stage was fixed so as to ensure that there were enough observations to allow representative statistics to be derived for each main strata of interest.

Use of this particular sampling procedure means that households residing in different parts of the country have been selected for the PIHS surveys with differing probabilities of selection. In order to derive representative statistics for each of the provinces, as well as for the country as a whole, raising factors (i.e. sampling weights) have been applied to the 1991, 1995-96 and 1996-97 PIHS data sets. These raising factors take into account the sampling strategy adopted in both of these surveys, and result in data for different households being weighted by a factor that is inversely proportional to their probability of selection in the sample.

Under this second round of the PIHS, data was collected from some 12,622 households living in 905 different primary sampling units (PSUs) selected throughout the Punjab, Sindh, NWFP, and Baluchistan (see Table 1.2). The data was collected between July 1996 and June 1997, apart from NWFP where the enumeration was completed in October 1997. All households were therefore not enumerated at the same time of the year. The survey is also being implemented in Azad Jammu and Kashmir, FATA, and FANA, the results of which will be presented in a later report.

In each of the selected PSUs, a fixed number of households were selected at random (12 in each urban PSU, 16 in each rural PSU), and a detailed household questionnaire was administered to each of them. In addition, in each PSU, a community questionnaire was also completed which gathered information on the quality of infrastructure, the provision of services, and consumer prices prevailing in the community.

Before moving on to discuss questionnaire content, two issues are worth elaborating with respect to PIHS sampling in Round II: sample size and household selection. With respect to sample size, it was recognised after Round I that the rural Baluchistan sample was too small to make accurate estimates of key variables. Accordingly, the number of Baluchistan rural households was increased from 950 in Round I to 1,111 in Round II. Regarding household selection, in Round II of the PIHS the enumeration team sought to re-interview two-thirds of the households from Round I, and replace one-third with new households. If the original household from Round I could not be found then they were not replaced. The enumeration teams were successful in locating almost all the original households from Round I.