Ghana - Ghana Time Use Survey 2009

Sample design

The sample for the 2009 GTUS was designed to provide estimates of key indicators at the national and regional levels as well as for urban and rural areas in Ghana. A representative sample of 4,800 households was drawn randomly from the list of Enumeration Areas (EAs) of the 2008 Ghana Demographic and Health Survey (GDHS), which served as a frame for the GTUS sample. In the selected households all individuals aged 10 years and older were interviewed.

The sample frame was first stratified into the 10 administrative regions in the country, then into urban and rural EAs. GTUS used a two-stage stratified sample design. At the first stage of sampling, 300 EAs were selected. These are a sub-sample of the 412 EAs selected from the 2008 GDHS. The second stage involved selection of 16 households from the 2008 GDHS listing in each selected EA.

The Primary Sampling Unit (PSU) was the EA, while the Secondary Sampling Unit (SSU) was the household. In the selected households all individuals aged 10 years and older were interviewed for the 24-hour activity diary. The following factors were considered in the selection of EAs and households:

a) The regional population and average household size in the 2000 Population and Housing Census. The larger the average household size, the smaller the proportion of sampled households in the EA.
b) A confidence interval of 95% with an error margin of 0.025.
c) The number of EAs for each region in the 2008 GDHS.
d) Allowance for a non-response rate of 20 percent for households. The rationale here was to eliminate the need for substitution of unfound or non-responding households during the fieldwork. Giving the option of substituting households to supervisors would have led to a biased sample and therefore field officers were not allowed to substitute. Furthermore, the selection of households considered the average household size of the regions and hence aimed at achieving an adequate sample of individual respondents who were the observation units.
e) Increasing the number of selected households to compensate for the exclusion of the population under 10 years old in the households.
f) As variations in the variables to be studied in the GTUS are expected to be higher in rural areas, it was decided to draw a larger sample (77% of EAs in GDHS 2008) for these areas than for urban areas (67% of EAs in GDHS).


Sample Selection
The regional samples of EAs selected from the 2008 GDHS EAs were done using SPSS syntax that applies a systematic simple random sampling procedure. However, the sampling weights were calculated on the basis of the population size of the EAs and their totals in the region.

The households were also selected using a systematic simple random sampling procedure in Microsoft Excel© using the 2008 DHS listing information. A sampling interval and a random starting number were determined. The random starting number served as the first household to be selected. The remaining 15 households were selected by adding multiples of the sampling interval to the random starting number until the desired number was achieved.


Response rates at the household and individual levels. The response rate for the 2009 GTUS was 99.5 percent at the household level and 86.5 percent at the individual level. As can be seen, the response rate at the individual level was higher in rural areas (87.2%) compared with urban areas (85.5%). It was also higher overall for females compared with males (88.1% against 84.8%). This can be explained by the fact that individuals are more likely to be absent from home in urban areas than in rural areas and females are more likely than males to be present in the household premises at the time of the interviewer's visit. It should also be noted that diary questionnaires that could not be linked to a fully completed household questionnaire have not been maintained in the sample for analyses.


Sample Weights
The 2009 GTUS sample was weighted to the 2009 national population of Ghana. The overall sampling weight was adjusted to suit the variation between the selected and responding sampling units as well as the 2000 census population projections for 2009.

The calculation of sampling weights takes into account the level of representation of sampling units. The survey observations are then adjusted accordingly.

The sampling weight Wij for jth EA in ith region is calculated as follows:

Wij = WGDHS * (1/si) * (Mi / mij)

where:
WGDHS is the weight of the EA in the 2008 GDHS.
si is the number of EAs selected from the ith region.
Mi is the total population of 2008 GDHS EAs in ith region.
mij is the population of jth EA in ith region during the 2008 GDHS.
The sampling weight Wjk for kth household in jth EA is calculated as follows:

Wjk = (Nij / 16)

where:
Nij is the number of listed households in jth EA of ith region during the 2008 GDHS.
The overall sampling weight Wijk for household k in EA j in Region i is the product of EA and household weight and is calculated as follows:

Wijk = (Wij * Wjk)

The overall sampling weight was adjusted due to the variation between the selected and responded sampling units as well as between the 2000 census population, the 2008 GDHS population and the 2009 GTUS population. The household adjustment factor was calculated by dividing the actual listed EA population by the estimated EA population from the survey.
The EA adjustment factor was calculated by dividing the estimated 2009 GTUS regional population by the 2008 GDHS regional population.