Survey Methodology- Liberia 2008

Overview

The universe sampled was the adult population aged 15 years and above living in Liberia.

A random sampling method was used to identify respondents participating in the study. A total of 1,542 interviews were achieved in Liberia, between 11th November and 18th November 2008.

The research study was conducted in the 15 regions: Bomi, Bong, Gbarpolu, Grand Bassa, Grand Cape Mount, Grand Gedeh, Grand Kru, Lofa, Margibi, Maryland, Montserrado, Nimba, Rivercess, River Gee, Sinoe.

Universe

The population size for Liberia was taken to be 1,869,925 adults aged 15 years and above as per the 2008 adult population estimate derived from the CIA World Fact Book.

Sampling Procedure

Sample structure

The urban interviews were allocated in proportion to the urban populations in the fifteen regions in Liberia and counties within the regions. In rural areas, all interviews were allocated to regions and counties in the same way. There was no boost applied in urban areas.

Small adjustments were made to the allocation by grouping counties with small populations so that each county selected was represented by a minimum of five interviews.

The table below indicates the number of interviews achieved by region:

          


Sampling point selection
Easily identifiable points such as schools, churches and police posts were marked out randomly as the starting point in each sampling area. The supervisor, in close consultation with the field manager, carried out the mapping. A maximum of 10 interviews were conducted at each sampling point.

Household selection
At each sampling point the starting point was identified using the “date summation score” method. If the date of the interview was 15th then a quick addition of the two digits yields 6 and consequently the interviewer targets the 6th household on the left hand side of the starting point. Subsequent households were then selected using the left hand rule, which required interviewers to skip four households after every successful household, targeting the fifth household on the left. At the end of the street the interviewer took the first right turn, at the end of the next street, the first left turn, and so on.

Contact Record Sheet
The interviewer kept Contact Record Sheets for each sampling point that they worked at. These were then coded as indicated in Table B below:



Using the above code list, the interviewer recorded all contacts made on the contact record sheet, whether successful or unsuccessful. For example, if the interviewer, after following the household selection criteria, randomly selected a non-residential building such as a hospital or hotel, they would then code ‘1’ as the outcome of contact code on the contact record sheet. If they selected a residential household, but after several knocks on the door, there was no response, they would then code it as ‘2’.
Interviewers did NOT record a code 9 (“skip household”) on the CRS for households that were skipped as part of the random route rules.  Code 9 was only recorded for any households that were due to be contacted (according to the random route) but was skipped for some other reason.

Respondent Selection
Random selection of respondents at household level was done by the use of a Kish Grid. Using the Kish Grid, the interviewer wrote down the names of all eligible members in the household (in this case, all adults aged 15+) in a systematic order (in order of age starting with the oldest). The educational status for each eligible household member was also recorded.  The total number of eligible household members and a pre-assigned random number (the last digit of the questionnaire number) were used in the Kish Grid to identify the household member to be interviewed.

If there were no adults aged 15+ living in the household, an outcome of Code 4 (“No adults 15+ in household”) was recorded on the CRS.

If the selected respondent was available and an interview was conducted, an outcome of Code 7 (“Achieved interview – original respondent”) was recorded.

If the selected respondent was at home but refused to co-operate, an outcome code of ‘Selected adult refused’ (Code 5) was recorded on the CRS. A substitution was then made in such an occurrence. A substitution was also made if the selected person was not available for interviewing after the interviewer had made up to three call-backs at different times of the day (including evenings).  

The substitute respondent matched the demographic characteristics of the original selected respondent in terms of gender, age and level of education.  Age bands used in substitutions were 15-24, 25-34, 35-44, 45-54, and 55+ (i.e. a substitute respondent had to fall into the same age band as the original respondent that was being replaced).

If a substitute respondent was required, the interviewer followed the same rules as before for selection of the next household.  However, when the interviewer listed the individuals in the household, they were permitted to take any respondent who matched the criteria they were searching for.  If that person was available for the interview, a Code 8 (“Achieved interview – replacement respondent”) was recorded.  If that person was not available, then the interviewer selected another individual using the Kish procedure and attempted to make a substitution at the next household selected. This was continued until the interviewers exhausted their sampling point or conducted a successful replacement interview.

Interviewer briefings
Prior to fieldwork, a general briefing session served to familiarise interviewers and supervisors with the questionnaire.

An Executive conducted a comprehensive face-to-face briefing in Monrovia with all interviewers and field involved in the study, prior to the start of the actual fieldwork. During the briefings the interviewers were taken through all the project documentation in detail, i.e. briefing instructions, recruitment questionnaire, contact record sheet, main questionnaire and interviewer summary sheet. 

The questionnaire was work-shopped and dummy interviews were conducted. This served to identify areas that were not clear to all interviewers. These questions / concepts were then thoroughly explained until everyone fully understood all aspects of the questionnaire and other related documentation.

It was ensured that the interviewers were clear on how to select respondents.  They were also required to familiarise themselves with the questionnaire routing and as such conducted dummy interviews with potential respondents, before the actual fieldwork commenced.

Interviewers were under strict instructions not to tell respondents whom the research was for, even if they figured out who the client was.

The Interview / Questionnaire

The final questionnaire provided in English was translated into Vernacular Liberian English (“Broken English”) by RMS and confirmed by Ipsos MORI. Any comments or changes were discussed with RMS and agreed upon before the changes were effected.  The BBC approved the final version of the questionnaire.

In Liberia the interviews were conducted in English and Broken English.  However, Broken English was used for most of the interviewing. The interviews took an average of 47 minutes.

Fieldwork

Fieldwork was carried out in Liberia over a period of 7 days between 11/10/2008 and 18/11/2008.

The interviews were spread across the seven days of the week. Actual interviewing was conducted starting at 9 am and running until late in the evening.  Call backs took place at three separate times during the same day of the initial contact, to maximise the chances of interviewing the selected respondent.

Response Rate Analysis

The information from the CRS was punched and a response rate analysis was produced. 

The final response rate achieved was 73% based on all residential contacts made. The response rate at the total level is shown in the following table C.



Quality Control
RMS conducted the following quality checks to ensure that the data collection process was accurate:

• Training and piloting was conducted to induct the interviewers with the study process.
• Supervisors accompanied each interviewer to 10% of their interviews to ensure that they were conducting the study as per the briefing they received.
•   In addition, the supervisor conducted a 15% back check to validate the answers recorded in the interview.

 Results

Weighting
Population statistics were provided for Liberia.  The source of these were the Liberia Demographic and Health Survey 2007 and the 2008 National  population and housing census (Preliminary results).
The data was weighted by the following parameters:

•    Age within sex
•    Education
•    Regions within urban/rural

The data were also grossed up to a 2008 adult (15+) population estimate produced by the 2008 adult population figure derived from the CIA World Fact Book.   

The final weighting matrix used was as follows:

For more information, email us at info@audiencescapes.org