The Central Statistics Office conducted the fifth Household Income and Expenditure Survey (HIES) from November 1993 to January 1995. As the title suggests, the main focus of the exercise was on household income and expenditure. Notwithstanding that, as a by-product, data on other socio-economic characteristics have been made available through the exercise.
Answers to economic questions such as: " Are households now spending more money on transport than on housing?", "Are the poor households getting poorer?" etc. - hinges on the results of the HIES. In the context of planning for economic development, studies of household income and expenditure are invaluable. These studies are helpful in evaluating the changes which occur, as a result of economic development, in household consumption patterns, levels of income, income distribution and the extent of the inequality, and trends in the preference of the different segments of the society. Viewed from another angle, the levels of these variables may be useful in determining the speed of socio-economic development in the foreseeable future.
Objectives and uses
It is probably appropriate to start by defining a household income and expenditure survey. A household income and expenditure survey is a survey designed to collect information on various sources of income (money or in kind) received by the households and details as to how they dispose of this income (on expenditure, remittances etc.). In essence, all the details of receipts by the households and those relating to the acquisition of goods and services for own consumption are recorded within the given reference period. The reference period, like in the previous survey, was one month. The key words in this definition, i.e. Household, Income, and Expenditure are defined in details under the section dealing with concepts and definitions.
The main objectives of the 1993/94 Household Income and Expenditure Survey were: -
i.) to determine household consumption expenditure patterns for urban towns, urban villages and rural areas so as to revise the weights used in the cost-of-living index. Information collected on itemized expenditure is useful for checking the existing basket-of-goods to ensure that the basket remains representative of national expenditure patterns.
ii) To determine the sources of household income, estimate income levels and distribution hence ascertain the extent of the inequality. Combined with details of the household structure and other socio-economic variables, such data are invaluable to planners and policy makers.
iii) to provide an independent source of information to estimate and improve the figures on "private final consumption" for National Accounts.
iv) to provide consumption data that enable the construction of a "Poverty Datum Line".
v) to provide business investors with information on consumption of specific products so as to determine potential consumer demand.
vi) to provide a range of baseline data for researchers.
In brief, the objective of the survey was the provision of comprehensive data on household income and consumption patterns for socio-economic analysis and planning.
Kind of Data
Sample survey data [ssd]
Unit of Analysis
The survey covered all households living in private dwellings, apart from households of foreign diplomats and their families. Also excluded from the survey are hotels, army camps, nurses hostels and other institutional accommodation. It should be mentioned, however, that Botswana Defence Force families living in ordinary private dwellings were included. The Ngamiland delta was not covered because of the difficult terrain.
Producers and sponsors
Central Statistics Office (CSO)
Ministry of Finance and Development Planning
Of the many factors that influence the sampling design of a particular survey, the nature of the subject is the most paramount. It is well known that the distribution of income among the households is uneven, with a few households accounting for a relatively large proportion of income. Botswana is no exception in this regard and indeed the 1985/86 survey revealed a high rate of inequality. There is a significant difference between income levels of urban towns, urban villages and the rural areas. Consumption expenditure depends, to a large extent, on income hence the arguments about income levels equally apply to consumption. The reason for going at length to elicit these problems is to give a background against which the 1993/94 Household Income and Expenditure Survey was conducted.
As in the 1985/86 survey, a two-stage stratified sample design was adopted. All surveys conducted under "The Household Survey Programme" have been two-stage stratified design. The multiple stages have been for ease of sample selection as well as for the fact that up-to-date sampling frame of the elementary units (households) are not available. On the other hand stratification is employed to provide separate estimates for the stratification factors as well as for gain in precision. Precision is gained when there is a reduction in the variance of the estimates.
In the 1985/86 Household Income and Expenditure Survey, the sampling frame which comprised the 1981 census was sorted into five strata.
- Freehold farms.
Following the 1991 census, nineteen of Botswana's villages are now classified as "urban", i.e. fewer than 25 percent of their workforce are working in traditional agriculture. Nonetheless, other characteristics of these villages may still be markedly different from the more established urban areas such as Gaborone, Lobatse, Francistown etc. Consequently, it was proposed that, for the 1993/94 HIES, an "urban-village" stratum, comprising these villages be created. The remaining villages were combined with lands areas, cattleposts and freehold farms into a "rural" stratum. For most practical purposes the difference in income levels between the areas constituting the rural stratum does not justify splitting the group into separate strata. Therefore, for the 1993/94 HIES the strata are:
- Urban villages
Unlike before, these stratification groups allow for presentation of separate results for each stratum.
Note: See detailed sampling procedure which is presented in the final report.
Dates of Data Collection
Data Collection Mode
Data Collection Notes
Data collection was mainly through daily visits by the field staff recording all household transactions. There were twelve field teams of which ten comprised three enumerators and a supervisor - the other two (mainly urban-based) comprised four enumerators and a supervisor. Thus, in total there were 50 field staff involved in the data collection. Supervisors covered a few dwellings, in addition to their supervisory role.
Field work started with the listing of all dwellings in selected blocks, thereby constructing a frame from which to select a sample of dwellings. The field staff were to visit all households in selected dwellings for six days each week, throughout the survey month. The working week was from Tuesday to Sunday with day-offs on Mondays. There was a four-day break in between rounds, i.e. one being a day-off, while the other three were for administrative arrangements and the listing of the dwellings in the block, in preparation for the next survey round. In lieu of the remaining dayoffs and public holidays, interviewers and supervisors were paid instead. For teams whose work areas were far from Gaborone, payment arrangements were made with the nearest District Commissioner's Office.
In order to establish some rapport with the households, on the first visit interviewers asked questions on demographic particulars and other less sensitive topics. Questions on income and expenditure were to be asked on the second or third visit. A supplementary record sheet was provided to households whose members were able to read and write. The idea was that each member could move around with the sheet recording any transaction they make. Interviewers would then visit daily to transcribe the information into the daily note book, handing back the sheets for subsequent transactions. Problems did arise when a member lost the supplementary sheets and could not recall all the transactions made. There were few such cases though. Besides that, this arrangement proved quite useful to members of the household who had to travel away from home for a few days. The task of the interviewer depended much on the household's level of income and the type of region. Households with high income often had many transactions and thus there was a lot to record. On the other hand for low income households, especially those in rural areas, there were often no transactions made for a whole week.
At the end of every round all the questionnaires for the completed round were dispatched to the office through registered mail. Before dispatch the team supervisor checked the questionnaires for completeness and other minor mistakes. When satisfied with the quality of work, the supervisor completed an accompanying summary form (HIES 11) which detailed out the number of books and supplementary record sheets for each household as well as all other materials dispatched to the office. Upon receipt of the batch at the office a receipt clerk counted all the forms received and cross-checked with details on the summary forms.
In order to check on the quality of fieldwork, occasionally a technical officer visited teams in the field and attended a few interviews. These visits were quite useful in that it could be ascertained whether the interviewers were asking the questions correctly or not. Apart from that, the field trips gave the technical officers insights into the applicability of some of the concepts in practice.
Although the main theme of the survey was income and expenditure a whole range of topics were covered, as a by-product. Information collected in the HIES falls naturally into two categories:
1. that which could be collected from single interviews (Book 1), and
2. that collected on a day-to-day basis over a period of one month (Book 2).
The questionnaires and other forms used in collecting auxiliary data are presented in the appendix.
The single interview questionnaires (Book 1) comprised:
1. i) Demographic data, and ii) Economic activities and employment (Section A)
2. Sources of household income (Section B)
3. Housing data (Section C)
4. Household enterprises (Section D)
5. Crops and livestock (Section E)
6. Employment earnings and deduction (Section F)
7. Major expenditure during past 12 months (Section G)
8. Regular monthly and annually payment (Section H)
9. Miscellaneous (Section I)
The day-to-day questionnaires were combined into the daily notebook (Book 2) which is subdivided into the following schedules:
1. Daily expenditure and other disbursements (Schedule D-1)
2. Cash receipts (Schedule D-2)
3. Goods and services received (Schedule D-3a)/given (Schedule D-3b)
4. Business receipts (Schedule D-2)
5. Business expenditure (Schedule D-5)
6. Own produce consumed (Schedule D-6)
Schedules D-1 to D-5 covered the full survey round of 30 days. On the other hand schedule D-6 covered a period of seven consecutive days within the survey month.
The in-house Data Processing Unit was responsible for data entry, maintenance of data entry and
validations systems, and the production of tables. As for data entry, questionnaires were entered by
one data entry operator and verified by another.
The fact that the HIES is a very difficult exercise to undertake cannot be overemphasized. A number of the questionnaires contained inconsistent data and this was not wholly attributable to the field staff. The complexity of the survey also had a bearing on that. Besides inconsistencies and cases of item non-response which of course were expected, records for items dealing with income were often not true. Editing teams had the enormous task of sorting out these problems. In addition, they had to contend with hundreds of records - up to a maximum of 500 records for one household in some cases.
Editing was done in two stages. The first stage involved checks for consistencies and completeness. Questionnaires whose data were not comprehensible were referred to the field supervisors for correction. The second stage needed more care as it involved some calculations and transcribing of records from one section to another since not all sections were entered directly. Care had to be exercised in transcribing data moreover that some items (e.g. rent) may have been recorded in three sections. Editors undertook limited imputations particularly in the case of missing price quotes for small items. Transcribing records from the daily notebook was rather cumbersome in that records were first grouped and then summed by item type. At the validation stage checks were made to ensure that information transferred was not duplicated.
Manual editing and coding began four months after the start of field work. The delay was allowed in order to accumulate more batches since editing and coding tend to be fast.
Most of the HIES questions were precoded hence there were few questions to be coded. This is not really to understate the volume of work the coders had to do. In Book 1 the questions whose answers needed coding were: subject of training, occupation, and industry. The bulk of the records to be coded were in the daily notebook. Coding was fairly straight-forward, although coders had to contend with the huge volume of records.
Checking of Editing
Upon completion of editing and coding questionnaires were checked by more senior staff for editing and coding errors, and any anomaly that may have been overlooked at the previous stages. This task was repeated twice, the idea being to filter out as many errors as possible. Finally, the questionnaires were scanned to establish whether they were complete enough for the data to be usable.
The bulk of the imputations involved questionnaires for households who were covered for fewer days than the survey round, but nonetheless for at least 15 days. The procedure adopted was to gross up to 30 days the expenditure for items the household was likely to have purchased on a daily basis. This was done for selected items in the list purchased during the days covered. Imputation for item non-response for the not-so-obvious answers, particularly for Book 1 records, was delayed until the end of the survey. Responses for some of these were imputed using the hot-deck techniques.
Imputation was just limited to fairly complete questionnaires. There were instances where the household provided reliable data on Book 1 and few or no records at all on the daily notebook. It would have been very unrealistic and indeed there was no basis to impute daily records for such households during the survey month. It was therefore decided to have two data sets - the first set being for households with both Book 1 and Book 2 complete while the second comprised the first plus households with only complete Book 1. The first set was used in the income and expenditure analysis while the second was just used in deriving population estimates.