Pakistan Survey Methodology


Survey Methodology

2010 Data

The 2010 data on which the analysis is based was sourced from a national level survey conducted by Gilani Research Foundation in 2010.

METHODOLOGY
The information and analysis presented in this Report is based on a nationally representative survey of more than 18,000 respondents. The details of the sample are given below:

The survey procedure was adapted from the internationally practiced methods documented in “Kent, Audience Measurement (1994)” and the practices of the National Readership Survey conducted in the UK.

Population Estimates:
Population estimates are based on the projections made by the Government of Pakistan/ Economic Survey of Pakistan provided in June 2010. The projections take the 1998 Census as the starting point. The GOP/ Economic Survey have given a detailed explanation of their current estimate and estimates for previous years, which they have revised retroactively. Obviously, it creates some degree of inconsistencies for reports which relied on their previous estimates. \the GOP/ Economic Survey has suggested the procedure to rectify and address such issues. For that please refer to Economic Survey of Pakistan 2009/2010, page 235, chapter/section titled: Population, Labour Force and Employment.

Sampling Information:
The sample for this survey included over 18,000 respondents. They include a cross-section of men and women from the rural and urban areas of all the four provinces of the country. The sample details are the following:

The tabulated results are weighted and correspond (with some approximation) with the census distribution of the rural and urban populations in the four provinces of the country.

The procedure of stratified random sampling was used to select the respondents from villages and urban locations situated in all the four provinces of Pakistan.

Field work dates: July-September, 2010
Mode of Interview: Face to face
Methodology limitations:
1- Caution on Segment Analysis: Since sample sizes drop, the data as to be read with caution. Example: Kalat-Makran Southern KP
2- Political unrest led to field work problems in several regions notably parts of KP and Balochistan. This could have led to certain biases in the data for those regions or sub-regions.

2008 Data

The survey utilized here was implemented in 2008 on behalf of the British Broadcasting Corportation by the market research firm IMRB International.

Research Design- An Overview
The Universe
The universe for the survey is defined as all adults in urban as well as rural Pakistan. An adult, for the purpose of this study, is defined as a person aged 15 years or more.
The research covered the total population comprising both the urban and rural markets.

Sample design
A stratified multi stage sampling procedure was adopted for the survey.

Provincial classification
Pakistan’s 4 provinces were taken into consideration for the purpose of the study

  • Baluchistan
  • NWFP
  • Punjab
  • Sind

The following areas were excluded from the fieldwork:
FATA – Federal Administrative Tribal areas, Provincially Administered Tribal Areas (PATA), Northern Areas, Jammu & Kashmir, Military restricted areas and areas that are out of reach due to law and order, flood etc. The excluded areas would account for approximately 10% of the population of Pakistan.

Sampling Methodology
Pakistan is divided into two domains – urban and rural which is approximately 33:67 split in the population. However we propose 50 percent of sample to be from the urban areas that is the sample will be divided equally in urban/rural areas to cover for the heterogeneity of the urban population.  The results will be reweighed to actual population proportions.

The details of the urban/rural population of Pakistan and the proposed sample break up are given below:
The total population of Pakistan is around 16,000,0000 that is divided into urban /rural areas in the ratio of 33: 67.

Thus the total population of the two areas is-
Urban: 52,800,000
Rural: 107,200,000
The population is divided almost equally amongst males and females. Pakistan has four provinces with distinctive characteristics, cultures, politics etc. Similarly there are 490 cities and towns in the country of different population sizes. The Research Design therefore needs to represent Pakistan at the aggregate levels of:

  • Key Demographics – Urban / Rural
  • Main strata of town populations and
  • Rural

Urban domain
The Urban population is divided into these provinces and further sub-divided into four tier of cities based on their population as described below:


Rural domain
The rural Pakistan is divided in the four provinces as under-


Sampling Methodology
Sampling procedure-Urban:
The sampling design was a multistage procedure described below
Stage 1: Selection of cities within each province cum population size (town class) strata using PPS
Stage 2: Selection of sampling units within each selected city – using frame of circles published by Pakistan Census Organization
Stage 3: Selection of households – using random walk method within the selected circles
Stage 4: Selection of individuals – using Kish grid

Stratification plan and allocation of sample to strata

  • The  Sampling Zones were the four population Strata
  • Within the four zones the adjustments were made such that there was a slightly higher sample allocation for the larger towns.
  • The three large cities account for 38% of the urban population and the seven largest cities account for 51% of the total urban population. These therefore were included in sample
  • For the remaining sample random selection of the cities will be done to represent their strata/cells. 

Stage 1

  • All large cities were included in the sample (the top 7 cities)
  •  For small cities/towns systematic random sampling will be done to ensure maximum variations across geographies and cultures in all provinces.

Stage 2

  • Circles will be treated as the sampling unit.
  • The sampling interval (k) was determined by dividing the total available number of sampling units in each city/town by the required number of sampling units to be covered in that city/town.
  • A random number smaller than the sampling interval was selected. First sampling unit corresponding to this number was selected.
  • The next sampling unit was obtained by adding the sampling interval to the random number selected and choosing the sampling unit corresponding to this number and so on.

Stage 3

  • Each Circle was divided in 3 to 4 continuous Blocks of 200 to 250 Households. One Block will be randomly selected from the chosen Circle.
  • The first random starting point (household) was selected from the list of households of the selected Block using simple random sampling procedure.
  • For the male interviews, we contacted every third household falling on the right hand from the first household till the male interviewer has completed the specified number of complete male interviews.
  • Simultaneously the female interviewer contacted every third household falling on the left hand from the first household till the female interviewer has completed the specified number of complete female interviews.

Stage 4

  • Selection of target respondent of the given gender, was done using the Kish grid method.
  • Sometimes on first contact it was possible that the randomly selected respondent is not available at home. In those cases three call backs were to be made to find the required respondent. If it became impossible to contact the randomly selected respondents in all three attempts, substitution would be made. While substituting the respondent, his/her age would be kept in mind and the replaced respondent must be in the age bracket of +2 years of the randomly selected respondent with other parameters of profile matching as far as possible.

Sampling procedure-Rural:
There are about 52,000 villages distributed in 106 districts and four provinces of Pakistan. Like cities/towns the villages are of different sizes but villages in each cultural belt are more homogeneous as compared to cities/towns.
The sample would be stratified by provinces and the sample design would be a multistage procedure described below-
Stage 1: Selection of rural circles within provinces.
Stage 2: Selection of households within selected circles.
Stage 3: Selection of Target respondent within selected household.

Stage 1

  • Circles were treated as the sampling unit.
  • The sampling interval (k) was determined by dividing the total available number of sampling units in each province by the required number of sampling units to be covered in that province.
  • A random number smaller than the sampling interval was selected. First sampling unit corresponding to this number was selected.
  • The next sampling unit was obtained by adding the sampling interval to the random number selected and choosing the sampling unit corresponding to this number and so on.

Stage 2

  • Each Circle was divided in 3 to 4 continuous Blocks of 200 to 250 Households (in large villages). One Block was randomly selected from the chosen Circle.
  • We selected the random starting point (household) from the list of households of the selected Block using simple random sampling procedure.
  • For the male interviews we would contacted every third household falling on the right hand from the first household till the male interviewer has completed the specified number of complete male interviews.
  • Simultaneously the female interviewer contacted every third household falling on the left hand from the first household till the female interviewer has completed the specified number of complete female interviews.

Stage 3

  • Selection of target respondent of the given gender, was done using the Kish grid method
  • Some times at the first contact it was likely that the randomly selected respondent was not available at home. In those cases three call backs would have to be made to find the required respondent. If it became impossible to contact the randomly selected respondents in all three attempts, substitution would be made. While substituting the respondent, his/her age would be kept in mind and the replaced respondent would have to be in the age bracket of +2 years of the randomly selected respondent matching the other features of the profile as much as possible.

Sample Distribution
Urban Domain

Rural Domain

(Note: The sample was divided equally between males and females).

We proposed 5 male and 5 female interviews to be conducted at each PSU. This would facilitate comparison of results across genders. Only one interview was conducted in each household.

Fieldwork quality control procedures
Trained interviewers and supervisors drawn from the fieldwork agency’s field force conducted the fieldwork.

All interviews finished in the first two days of fieldwork were scrutinized by the Field Manager / Executive / Project Leader and feedback was provided to the interviewers / supervisors.

Atleast, 10% of the interviews were accompanied and 15% were back-checked either by the Field Executive or Supervisor so as to ensure the quality of data and whether all instructions were adhered to.