The methodology of the sample design used by Gallup Pakistan was consistent with the Pakistan Federal Bureau of Statistics (FBS). The survey was designed to cover all four provinces in Pakistan (Baluchistan, Sindh, Punjab, and Khyber Pakhtunkhwa), except FATA, GB and AJK. These three regions represent less than 4% of Pakistan's population. The main reason for their exclusion is inability to carry out standard random sampling/probability procedures as well as due to logistical and law and order constraints.
The survey used the population frame of the FBS, which divides the entire country into rural and urban enumeration blocks. The latest census on which these enumeration blocks were based was the Population Census 1998. The enumeration blocks in rural areas were fixed in 1998, those in urban areas in 2003.
The following provides an overview of the sampling process:
1st stage: Province/ Sub province distribution is made such that adequate and statistically significant analysis can be done on 13 specially developed regions as well as at the 4 provinces. Unlike in 2008, the sampling strategy for the 2015 A2F survey gave an over-weight to smaller provinces, so as to ensure that province-level results are statistically significant.
2nd stage: Once the sample has been distributed in the 13 regions, the spreads between the urban and the rural population within these regions is specified according to the split provided in the1998 Census.
3rd stage - Distribution by settlement: in urban areas, census blocks were used and in rural areas villages. The selection of villages and census blocks was performed using a probability proportional to size (PPS) method.
4th stage Starting points for Random Walk: in urban areas we randomly selected the house number and in rural areas a central point (such as a mosque or a public building) as a starting point; once the first household is contacted, every third household on the right side of the route is contacted thereafter until 15 interviews are completed.
5th stage Respondent selection: a Kish grid was used to select the respondent in each household.
The resulting sample distribution is as follows:
The size of the sample selected for this survey is consistent with the standard formula used to derive the size of nationally representative samples. With the obtained sample size, the error margin for this survey would be +/-1 % at 95% confidence Interval.