Ghana - Multiple Indicator Cluster Survey (MICS3) 2006, Third Round

The sample for the MICS 2006 was designed to provide estimates on a large number of indicators of the health status of women, men, and children at the national level, for urban and rural areas, as well as for the 10 administrative regions in the country.
A representative probability sample of 6,302 households was selected nationwide. The list of enumeration areas (EAs) from the Ghana Living Standards Survey 5 (GLSS5) served as a frame for the MICS sample. The frame was first stratified into the 10 administrative regions in the country, then into urban and rural EAs. 660 EAs {-281 urban and 379 rural}

1) Twenty households per EA were selected
2) 25 per EA for rural EAs in Northern, Upper East and Upper West.
3) All women aged 15-49 and children less than 60 months in these selected households were eligible for interview.
4) Males in every selected third household aged 15-49 were also eligible for interview.
5) This is different from DHS whereby males aged 15-59 are eligible for interview.

The MICS 2006 used a two-stage stratified sample design. At the first stage of sampling, 300 census enumeration areas (124 urban and 176 rural EAs) were selected. These are a sub-sample of the 660 EAs (281 urban and 379 rural) selected for the GLSS 5 and fisheries. The clusters in each region were selected using systematic sampling with probability proportional to their size. The distribution of EAs between regions is not proportional to the 2000 Population and Housing Census, mainly due to over-sampling in the number of EAs for Northern, Upper East and Upper West Regions.

A complete household listing exercise covering all the GLSS 5 EAs was carried out in May through July 2005 with a few selected EAs listed early 2006. At the second stage, a systematic sampling of households was selected based on this list. The MICS households were selected systematically from the household listing provided by GLSS 5 after eliminating from the list households previously selected by the GLSS 5 (15 regular with 5 replacement). The reason for selecting different households is that the GLSS 5 interviews are long and demanding for respondents. It therefore seemed preferable to keep the two household samples separate in order to avoid respondent fatigue and possible high rates of refusal in the households falling in both samples as they were being conducted concurrently. For the MICS, 20 households per EA were selected in all the regions except in Northern, Upper East and Upper West regions, where 20 households per EA were selected in urban EAs and 25 households selected from rural EAs. The objective of this exercise was to ensure an adequate number of complete interviews to provide estimates for important population characteristics with acceptable statistical precision per region. Due to the disproportional number of EAs and different sample sizes selected per EA among regions, the MICS 2006 household sample is not self-weighting at the national level. For reporting national level results, sample weights are used. A more detailed description of the sample design can be found in Appendix A.

Sample Size and Sample Allocation

The sample size for MICS 2006 was calculated as 6,300 households using basically the same sample of clusters selected for DHS 2003, as well as for similar sample size. The resulting number of households from this exercise was a minimum of about 500 (except for Upper West Region) households which is the sample size needed in each region - thus yielding about 6,500 in total. The average cluster size in MICS 2006 was determined as 20 households (except in rural clusters in Northern, Upper East and Upper West Regions with 25 households) based on a number of considerations, including the budget available, and the time that would be needed per team to complete one cluster. Dividing the total number of households by the number of households per cluster, it was calculated that the selection of a minimum of about 25 clusters would be needed in each region.

The allocation of the total sample size to each of the ten regions follows almost same allocation than the DHS 2003. Therefore, a minimum of 25 clusters were allocated to each region, with the final sample size calculated at 6,300 households and 300 clusters in total. In each region, the clusters (primary sampling units) were distributed to urban and rural domains, proportional to the size of urban and rural populations in that region. The table below shows the allocation of clusters to the sampling domains.

Table SD.1: Allocation of Sample Clusters (Primary Sampling Units) to Sampling Domains
Region N clusters Urban clusters Rural clusters
Western 29 11 18
Central 26 10 16
Greater Accra 43 38 5
Volta 24 6 18
Eastern 32 11 21
Ashanti 47 24 23
Brong Ahafo 24 9 15
Northern 30 8 22
Upper East 24 4 20
Upper West 21 3 18

Total 300 124 176

Sampling Frame and Selection of Clusters

The frame for MICS 2006 is the DHS 2003 sample frame (also being a sub sample of the 660 clusters for the Ghana Living Standard Survey GLSS-5), selected systematically and PPS (probability proportional to size) The first stage of sampling was thus completed by selecting the required number of enumeration areas from each of the urban and rural areas separately, as well as for each of the ten regions separately

Listing Activities

Since the sample frame (the 2000 Population and Housing Census) was not up to date, household lists in all selected enumeration areas were updated prior to the selection of households during DHS 2003 and the 2005/2006 GLSS 5 samples. A complete household listing exercise covering all the GLSS 5 EAs was carried out May through July 2005 with a few selected EAs listed early 2006. At the second stage of selection, a systematic sampling of households was done from such list

Selection of Households

Lists of households were prepared by the listing teams in the field for each enumeration area. The households were then sequentially numbered from 1 to n (the total number of households in each enumeration area) at the Ghana Statistical Service, where selection of 20 households in each enumeration area was carried out using systematic selection procedures


No deviation of the original sample design was made


Response rates are important as high non-response may affect the reliability of the survey results. Table HH.1 presents information on the results of the household and individual interviews. A total of 6,302 households were selected for the MICS. Of these, 6,264 were found to be occupied and interviews were completed for 5,939 households which represent s a 95 percent response rate. A total of 6,240 women age (15-49) were identified from every selected household and then interviewed, while 1,909 eligible men (age 15-49) from every third selected household were identified for the individual interviews. Interviews were successfully completed for 5,891 women and 1,743 men, yielding response rates of 94 percent and 91 percent respectively. In addition, 3,545 children under five years were listed in the households. Questionnaires were completed for 3,466 children corresponding to a response rate of 98 percent. Taking into account the non-response at the household level, the overall response rates for women, men and children under five were 90 percent, 87 percent and 93 percent respectively.

Regional differentials in response rates regarding household interviews, eligible women, and children were similar (over 90 percent, respectively). However, overall response rate for women, men and children varied slightly by place of residence. The response rates are higher for the rural than the urban sample and among women than men. The main reason for non-response among households and eligible individuals was the failure to find these individuals at home despite several visits to the households.

Table HH.1: Results of household and individual interviews
Numbers of households, women, men, and children under five by results of the household, women's, men's and under-five's interviews, and household, women's, men's and under-five's response rates, Ghana, 2006
Area Region Total
Urban Rural Western Central Gt Accra Volta Eastern Ashanti B.A Northern U East U West
Sampled households 2,480 3,822 580 520 861 480 641 940 480 710 580 510 6,302
Occupied households 2,470 3,794 577 520 856 478 637 936 476 706 574 504 6,264
Interviewed households 2,327 3,612 561 510 802 447 589 881 442 673 561 473 5,939
Household response rate 94.2 95.2 97.2 98.1 93.7 93.5 92.5 94.1 92.9 95.3 97.7 93.8 94.8

Eligible women 2,546 3,694 560 434 939 414 606 850 471 824 632 510 6,240
Interviewed women 2,385 3,506 537 426 859 375 565 808 452 790 598 481 5,891
Women response rate 93.7 94.9 95.9 98.2 91.5 90.6 93.2 95.1 96.0 95.9 94.6 94.3 94.4
Women's overall r r 88.3 90.4 93.2 96.3 85.7 84.7 86.2 89.5 89.1 91.4 92.5 88.5 89.5

Eligible men 739 1,170 165 121 277 133 176 303 133 260 193 148 1,909
Interviewed men 660 1,083 154 118 237 117 163 272 120 248 179 135 1,743
Men response rate 89.3 92.6 93.3 97.5 85.6 88.0 92.6 89.8 90.2 95.4 92.7 91.2 91.3
Men's overall resp rate 84.1 88.1 90.8 95.7 80.2 82.3 85.6 84.5 83.8 90.9 90.7 85.6 86.6

Eli children under-five 1,030 2,515 319 263 330 245 346 426 245 595 399 377 3,545
Mother/ctaker Interviewed 1,012 2,454 316 262 326 236 337 415 242 576 389 367 3,466
Child response rate 98.3 97.6 99.1 99.6 98.8 96.3 97.4 97.4 98.8 96.8 97.5 97.3 97.8
Childoverall res rate 92.6 92.9 96.3 97.7 92.6 90.1 90.1 91.7 91.7 92.3 95.3 91.4 92.7


Sample weights were appended to all data sets and analyses were performed by weighting each household, woman, man or under-5 with these sample weights-( hhweight, Chweight)

Calculation of Sample Weights

The MICS 2006 sample is not self-weighted. Essentially, by allocating not a proportionally numbers of households to each of the regions, different sampling fractions were used in each region since the size of the regions varied. For this reason, sample weights were calculated and these were used in the subsequent analyses of the survey data.

The major component of the weight is the reciprocal of the sampling fraction employed in selecting the number of sample households in that particular sampling domain:

Wh = 1 / fh

The term fh, the sampling fraction at the h-th stratum, is the product of probabilities of selection at every stage in each sampling domain:

fh = P1h * P2h * P3h

where Pih is the probability of selection of the sampling unit in the i-th stage for the h-th sampling domain, i.e.,

P1h is the selection probability in the GLLS 5 survey;
P2h is the sub selection rate for clusters used in the 2005 Ghana survey from GLLS 5 survey; and
P3h is the sub selection rate for households in the cluster.

Since the estimated numbers of households per enumeration area prior to the first stage selection (selection of primary sampling units) and the updated number of households per EA were different, individual sampling fractions for households in each EA (cluster) were calculated. The sampling fractions for households in each EA therefore included the probability of selection of the EA in that particular sampling domain and the probability of selection of a household in the sample EA.

A second component which has to be taken into account in the calculation of sample weights is the level of non-response for the household and individual interviews. The adjustment for household non-response is equal to the inverse value of:

RR = Number of interviewed households / Number of occupied households listed

After the completion of fieldwork, response rates were calculated for each sampling domain. These were used to adjust the sample weights calculated for each cluster. Response rates in MICS 2006 are shown in Table HH.1 in this report.

Similarly, the adjustment for non-response at the individual level (women, men, and under-5 children) is equal to the inverse value of:

RR = Completed women's (or under-5's) questionnaires / Eligible women (or under-5s)

Numbers of eligible women, men, and under-5 children were obtained from the household listing in the Household Questionnaire in households where interviews were completed.

The unadjusted weights for the households were calculated by multiplying the above factors for each enumeration area. These weights were then standardized (or normalized), one purpose of which is to make the sum of the interviewed sample units equal the total sample size at the national level. Normalization is performed by multiplying the aforementioned unadjusted weights by the ratio of the number of completed households to the total unadjusted weighted number of households. A similar standardization procedure was followed in obtaining standardized weights for the women's, men's, and under-5's questionnaires. Adjusted (normalized) weights varied between [lowest weight] and [highest weight] in the 300 enumeration areas (clusters).