The FinScope 2005 was a nationally representative survey measuring access, usage, perceptions and demand patterns on financial services and issues to create insights into how consumers source their income and manage their financial lives. The survey aimed to establish credible benchmarks and indicators of access, provide insights into regulatory and market obstacles to growth and innovation, and highlight opportunities for policy reform and innovation in product development and delivery.
Kind of Data
Sample survey data
Unit of Analysis
v1: Edited, anonymised dataset for public distribution
This version was first received by DataFirst from Financial Services Deepening Zambia (FSDZ) in November 2016. This is the first version that has been made available for public release.
Individuals: age, financial services used, financial behaviour, knowledge of financial products and institutions, bank status, access to transport, constraints on financial product access, unexpected events, self-reported quality of life, employment, asset ownership, insurance, credit access, business operations, savings behaviour, cellphone access, household facilities, psychological health, marital status, education, language, literacy.
The survey covered all usual household residents age 16 and above in Zambia.
Producers and sponsors
Bank of Zambia
Department for International Development
Swedish International Development Agency
The sampling frame was based on data generated by the 2000 Zambian population census. The Finscope 2005 used a stratified, multi-stage cluster sampling approach. Stratification was done at the province and urban/rural level before cluster sampling was performed using enumerator areas as PSUs. 160 EAs were selected with probability proportional to size. Within each EA 25 households were randomly selected. In the third stage, one eligible (aged 16 and above) member of the household was randomly selected using a Kish grid. We should therefore expect the sample population to be equal to 4000 (160 EAs in which 25 households were selected).
The report states that, "a CSO senior statistician has assisted FinMark Trust in the cleaning and weighting of the FinScope™ dataset, and the comparison of the FinScope™ sample to the Zambia Census 2000 data to assess the representation of the FinScope™ sample to Zambia's overall adult population." We interpret this as meaning that the sample was weighted to match estimated Zambian population proportions based on demographic models informed by the Zambian Census 2000. It is unclear from the report the exact methodology used for the calibration of the post-stratified weights.
Deviations from the Sample Design
The sample is two observations short of what is expected. It is unclear from the available documentation why those two observations were omitted.
Final_weight: this weight variable appears to be the post-stratified weight. When applied, it gives population totals that agree with what we assumed is a demographic model informed by Zambian Census 2000 figures.
Final_weights_2: the application of this weight variable yields the correct proportions, but does not inflate in line with estimates of target population.
Dates of Data Collection
Data Collection Mode
Data Collection Notes
FinMark Trust and its local implementation partners worked with the Zambian CSO for this survey. The CSO staff assisted the market research firm contracted by FinMark Trust to undertake the fieldwork (DCDM) in designing the FinScope sampling frame, advising on the respondent selection criteria to be used, and preparing all the fieldwork maps required. Thirty surveyors were contracted by DCDM to undertake the face-to-face interviews, and eight CSO supervisors were contracted to oversee the surveyors in the field and report to the DCDM Project Manager. A pilot was conducted in Lusaka district to test the questionnaire. The fieldwork was conducted between 7th November and 22nd December 2005. Each team was provided with detailed maps prepared by CSO of the SEAs selected with their respective province, together with all fieldwork materials necessary to complete the survey. The households were selected randomly using the Random Route Methodology, and the interview respondent within
the selected household was identified using the Kish grid. One individual was interviewed per sampled household. Two call-backs were allowed for each selected respondent in addition to the initial contact. In cases where selected respondents were not available or refused to be interviewed, substitutions were made. These substitutions were carried out by the supervisor in accordance with CSO standard procedure and involved selecting the next household in line using the Random Route Methodology.
Where possible, the face-to-face interviews were carried out in English. Where use of a vernacular language was required, a native speaking surveyor carried out the interview. Show-cards in both English and all vernacular languages accompanied the questionnaire and were used to aid the respondent in answering several of the questions. In cases where a respondent was unable to read, the surveyor was instructed to read through and repeat each option to ensure the respondent
comprehended the statement and possible options for answering the questions.
De Chazal du Mee
A single questionnaire was used for the Zambia Finscope 2005. The FinScope core questionnaire was adapted to the Zambian context in close consultation with FSDP Working Groups and other financial sector stakeholders. The questionnaire is provided as an external resource.
The original version of the Zambian Finscope 2005 had a large number of geographic variables, many of which ended up being excluded. Some were excluded because they weren't actually usable without a private codebook, others because they could potentially lead to identification of individuals in the dataset. The final set of geographic variables included were province, district, and an urban/rural binary variable.
The original datafile also contained a number of variables that detailed surveyor, supervisor, and visitation information. These were also excluded from the distributed datafile. Also excluded were variables used only for informing placement in the Kish grid.
The data file received by DataFirst is two observations short of what was expected. That is, there are 3998 observations in the data file when there should be 4000.