Enquête Permanente Auprès des Ménages 1986-1987 (Wave 2 Panel)
Côte d'Ivoire Living Standards Survey (CILSS) 1986-1987
The Côte d'Ivoire Living Standards Survey (CILSS) was the first LSMS Survey to have field tested the methodology and questionnaire developed by LSMS. It consists of three complementary surveys: the household survey, the community survey and the price survey. The household survey collected detailed information on expenditures, income, employment, assets, basic needs and other socio-economic characteristics of the households. The Community Survey collected information on economic and demographic characteristics of the rural communities to which each cluster of households belonged. This was designed to enable the linkage of community level with household level data. The price survey component of the CILSS collected data on prices at the nearest market to each cluster of households, so that regional price indices could be constructed for the household survey.
The Côte d'Ivoire Living Standards Survey (CILSS) was undertaken over a period of four years, 1985-88, by the Direction de la Statistique in Côte d'Ivoire, with financial and technical support from the World Bank during the first two years of the survey. It was the first year-round household survey to have been undertaken by the Ivorian Direction de la Statistique.
The sample size each year was 1600 households and the sample design was a rotating panel. That is, half of the households were revisited the following year, while the other half were replaced with new households. The survey thus produced four cross-sectional data sets as well as three overlapping panels of 800 households each (1985-86, 1986-87, 1987-88).
Kind of data
Sample survey data [ssd]
Domains: Urban/rural; Regions (East Forest, West Forest, East Savannah, West Savannah)
Unit of analysis
Producers and sponsors
Direction de la Statistique
Ministère de l'Economie et des Finances
The World Bank
The principal objective of the sample selection process for the CILSS Household Survey was to obtain a nationally representative cross-section of African households, some of which could be interviewed in successive years as panel households.
A two-stage sampling procedure was used. In the first stage, 100 Primary Sampling Units (PSUs) were selected across the country from a list of all PSUs available in the sampling frame. At the second stage, a cluster of 16 households was selected within each PSU. This led to a sample size of 1600 households a year, in 100 cluster s of 16 households each. Half of the households were replaced each year while the other half (the panel households in 1986, 1987 and 1988) were interviewed a second time.
It is important to note that there was a change in the sampling procedures (the sampling frame, PSU selection process and listing procedures), used to select half of the clusters/households interviewed in 1987 (the other half were panel households retained from 1986), and all of the clusters/households interviewed in 1988. Households selected on the basis of the first set of sampling procedures will henceforth be referred to as Block 1 data while households based on the second set of sampling procedures will be referred to as Block 2 data.
>> Sampling Procedures for Block 1 Data
The Sampling Frame. The sampling frame for the 1985, 1986, and half of the 1987 samples (except for Abidjan and Bouaké) was a list of localities constructed on the basis of the 1975 Census, updated to 1983 by the demographers of the Direction de la Statistique and based on a total population estimated at 9.4 million in 1983.
The Block 1 frame for Abidjan and Bouaké was based on data from a 1979-80 electoral census of these two cities. The electoral census had produced detailed maps of the two cities that divided each sector of the city into smaller sub-sectors (îlots). Sub-sectors with similar types of housing were grouped together by statisticians in the Direction de la Statistique to form PSUs. From a list of all PSUs in each city, along with each PSU's population size, the required number of PSUs were selected using a systematic sampling procedure. The step size was equal to the city's population divided by the number of PSUs required in each city. One problem identified in the selection process for Abidjan arose from the fact that one sector of the city (Yopougon) which had been relatively small in 1980 at the time of the electoral census, had since become the largest agglomeration in Côte d'Ivoire. This problem was presumably unavoidable since accurate population data for Yopougon was not available at the time of the PSU selection process.
Selection of PSUs. Geographic stratification was not explicitly needed because the systematic sampling procedure that was used to select the PSUs ensured that the sample was balanced with respect to region and by site type, within each region. The main geographical regions defined were: East Forest, West Forest, and Savannah. Site types varied as follows: large cities, towns, large and small villages, surrounding towns, village centers, and villages attached to them. The 100 PSUs were selected, with probabilities proportional to the size of their population, from a list of PSUs sorted by region and within each region, by site type.
Selection of households within each PSU. A pre-survey was conducted in June-July of 1984, to establish the second-stage sampling frame, i.e. a list of households for each PSU from which 16 households could be selected. The same listing exercise was to be used for both the 1985 and 1986 surveys, in order to avoid having to conduct another costly pre-survey in the second year. Thus, the 1984 pre-survey had to provide enough households so as to be able to select two clusters of households in each PSU and to allow for replacement households in the event that some in the sample could not be contacted or refused to participate. A listing of 64 households in each PSU met this requirement. In PSUs with 64 households or fewer, every household was listed.
In selecting the households, the "step" used was equal to the estimated number of households in the PSU divided by 64. For example, if the PSU had an estimated 640 households, then every tenth household was included in the listing, counted from a random starting point in the PSU. For operational reasons, the maximum step allowable was a step of 30. In practice, it appears that enumerators used doors, instead of housing structures, in counting the step. Al though enumerators were supposed to start the listing process from a random point in the PSU, in rural areas and small towns, reportedly, the lister started from the center of the PSU.
>> Sampling Procedures for Block 2 Data
The Sampling Frame. The sampling frame for Block 2 data was established from a list of places from the results of the Census of inhabited sites (RSH) performed in preparation for the 1988 Population Census.
Selection of PSUs. The PSUs were selected with probability proportional to size. However, in order to save what might have been exorbitant costs of listing every household in each selected PSU in a pre-survey, the Direction de la Statistique made a decision to enumerate a smaller unit within each PSU. The area within each PSU was divided into smaller blocks called `îlots'. Households were then selected from a randomly chosen îlot within each PSU. The sample îlot was selected with equal probability within each PSU, not on the basis of probability proportional to size. (These îlots are reportedly relatively small compared with the size of PSUs selected for the Block 1 frame, but no further information is available about their geographical position within the PSUs.)
Selection of households within each PSU. All households in each îlot selected for the Block 2 sample were listed. Sixteen households were then randomly chosen from the list of households for each îlot.
>> Bias in the Selection of Households within PSUs, Block 1 Data
Analysis of the four years of the CILSS data revealed that household size (unweighted), dropped by 24 percent between 1985 and 1988. Three possible explanations were considered: (1) area l demographic change; (2) non-sampling measurement errors were involved; or (3) some sort of sampling bias. Investigation ruled out the first two possibilities. The third possibility clearly was an issue because the sampling frame and listing procedures had indeed changed in midstream and this was likely to have had an effect. In fact, the investigation found that the substantial part of the drop in household size over the years occurred between the first and second panel data sets in 1987, i.e. the tail end of Block 1 data and the start of Block 2 data. From this, it is reasonable to assume that differences in the sampling frame and sampling procedures between the two blocks were indeed responsible.
The listing procedures for Block 1 data indicate d that the selection of households within PSUs was likely to have been biased toward the selection of larger dwellings. Based on a discussion with Christopher Scott, statistical consultant, Demery and Grootaert explain as follows: "In the selected primary sampling units, where the listing of households was to occur, enumerators were instructed to start the listing process at a random location in the primary sampling unit and from this point to select every nth household, that is, with a given fixed "step" until sixty-four households were listed. There are two sources of potential bias in this listing procedure. First, the selection of the starting point might not have been random, but subject to motivated bias on the part of the enumerator (such as the selection of a point where there are numerous dwellings or that is easily accessible). Second, in practice, enumerators counted doors to achieve the "step", rather than counting actual households. This method leads to sample selection bias if the number of doors varies across households. Households with two doo rs will have twice the probability of selection as those with one door. Given that larger dwellings are more likely to have more doors than small dwellings, counting households on the basis of doors may have caused a bias in the sample leading to over-enumeration of large dwellings, and, thereby, large households. In fact, mean dwelling size recorded in the CILSS was significantly higher in 1985 and 1986 than in 1987 and 1988, supporting this interpretation." As explained by Christopher Scott, the sample selection bias towards larger households is reinforced because "where more than one house hold shares a door, the survey structure implicitly required that only one be selected. In the absence of specific instructions there would be a natural tendency to choose the main one, thus reinforcing the bias towards large households". One set of weights is provided to reconcile this bias (Household Size Weights).
>> Problems Due to Inaccurate Estimates of PSU Population
Scott and Amenuvegbe address the problem arising from the fact that the estimates of population used to select PSUs with probability proportional to size, are often outdated and inaccurate. The degree of inaccuracy increases with the number of years that have elapsed since the previous census. The extent of inaccuracy becomes clear when the listing process is completed for the PSU and the 'correct' population size becomes available. This deviation between the PSU's estimated population size and the 'correct' size needs to be addressed either by varying the sample 'take' in each PSU or by assigning corrective weights for each PSU, m' i/ni or m'i/mi, where m'i is the number of households found in the ith PSU at the listing stage and ni or mi is the measure of size (n=population, m=households) used in the first stage PSU selection. [Note: Omission of this weight has two effects: (1) a bias in favor of PSUs whose population has grown relatively slowly (or diminished) since the census, and against those that have grown exceptionally fast; and (2) a bias in favor of PSUs whose current mean household size is relatively large (Chris Scott, personal communication).]
As has been mentioned earlier, the CILSS implemented a fixed 'take' of 16 households per cluster, and one cluster per PSU. Given these rules, any attempt to address this issue would have to rely on corrective m'i/ni weights. However, it is not possible to calculate these weights either for Block 1 data (m' I is not available since only 64 households per cluster were listed); or for Block 2 data (m' i is not available since the enumeration area was the îlot and not the PSU).
>> Classification of Clusters by Geographic Location
A list of all CILSS clusters, their region, sub-prefecture (for rural clusters), and year of interview, are presented in Table IV. Regional classification was constructed so that clusters in Abidjan would be classified as Abidjan; clusters in other major cities in East Forest, West Forest and Savana would be classified as "Other Cities "; and only rural clusters would be included in the East Forest, West Forest and Savanna. Thus, the urban-rural classification is as follows: Urban - Abidjan and Other Cities; Rural - East Forest, West Forest and Savanna. The precise location of the clusters is noted on the map on the inside cover. The list of Sample Clusters is available in document "Basic information for users of the data".
>> Non-Response and Replacement of Households
In order to maintain the size of the sample, the 1985 CILSS instituted a procedure by which households in the original sample that refused or were unavailable were replaced by one of the 48 remaining households in the same cluster, enumerated during the pre-survey. The pre-survey questionnaire collected summary information about household size and a few socioeconomic characteristics of the household. Thus, for the first year of the CILSS, households were replaced (when necessary) by one of the remaining 48 households that most resembled it in terms of size and characteristics. During the first year of the survey, a total of 124 of the 1600 original sample households (7.8 percent) were not interviewed and were thus replaced by other households. The most common reason for non-response in the first year was the inability to locate the address or housing unit. Only 14 households (0.9 percent of the sample) were found but refused to participate, and these were all in Abidjan.
In the second and subsequent years, the effort to replace non-responding households with other like households was abandoned. Households were replaced by the first household on the list of those remaining from the pre-survey or listing operation. (In 1986, there was a maximum of 32 households left in each cluster that were unused from the pre-survey.)
>> Selection of the Respondent for the Section on Fertility
The survey collected fertility information for one woman per household. The woman, aged 15 or older, was selected at random from among household members, so that each eligible woman in a household had the same probability of being selected. The method used was to generate random permutations of household members' personal identification numbers on adhesive labels. Each questionnaire was affixed with a label. In order to select a woman at random, the interviewer scanned the list of personal id codes on each label until arriving at the id number of the first eligible female household member. Besides providing a random means of selection, this methodology ensured that the interviewer's selection of the respondent was verifiable and replicable by the supervisor.
Dates of collection
Mode of data collection
The description below is extracted from
This document provides more detailed information on changes in the CILSS questionnaires across years.
>> HOUSEHOLD QUESTIONNAIRE
The household questionnaire was administered to all households in two visits or "rounds", with an interval of two weeks between the two interviews. The two-round scheme had certain useful advantages. For sections in Round 2, that collect information on household expenditure and income, it was possible to use a bounded recall period with the bound being the first visit of the interviewer. Correction and verification of doubtful data from the first visit was also possible during the second visit to the same households. The two-round scheme had the added advantage of breaking up what might have otherwise been a prohibitively long interview into two smaller sessions.
The following description of the CILSS household questionnaire contents outlines the main topics covered in each section and points out the changes that took place in the questionnaires over the survey years. The questionnaire was almost entirely pre-coded. However, the responses to a few questions were not pre-coded.
SECTION 0: ADMINISTRATIVE INFORMATION
Part A collects information on the ethnic group of the household head, language of the interview, whether the household was a replacement household, date of verification by the supervisor, date of data entry, and date of supervision of printouts. In Part B, the interviewer records the date that individual sections were completed.
SECTION 1: HOUSEHOLD ROSTER
Part A: Household Roster
This section lists all people w ho normally live and eat their meals together in the dwelling as well as those who spent the previous night in the household, but cannot otherwise be considered household members. To qualify as a household member, a person must have resided in the household for 3 months or more during the previous 12 months. Since the number of months the person was resident in the household is recorded, users of the data can choose a more restrictive definition if they wish. For example, one study used six months of residence, rather than three months, as the criterion for household membership. The household head, newborn babies and new spouses are considered household members even if they do not satisfy the months of residence criterion. CILSS always excludes servants and boarders from household membership because they are considered to be separate households.
Part B: Information on Parents
Information on schooling and occupation of the parents of all household members is collected in this section, even if the parents are deceased. If the parents live in the same household, then the parents' ID codes are noted so as to enable linkage of information about the parents with information about the child.
SECTION 2: HOUSING
This section includes a description of the dwelling, housing expenses (including rent and utilities), source of water, cooking fuel and light, garbage disposal and type of toilet.
SECTION 3: EDUCATION
Part A: Education of Household Members
This section includes self-reported literacy and numeracy, years of schooling, number of hours of attendance at school in the past seven days, whether the person is currently enrolled in school, distance to school, time taken to get to school, apprenticeships, technical training, highest diploma obtained, and expenditure on education in the past twelve months for those who have been enrolled. The CILSS questionnaire collected this information for each household member five years or older so that welfare differences across individual household members could be studied.
Part B: Education of Children Living Elsewhere
The purpose of this section is to evaluate the total investment in education that was made by the household, including children who are not currently household members. Age, sex, and educational achievement of children below 30 who have left the household are recorded.
SECTION 4: HEALTH
This section focuses on the economic costs of health problems of all household members. The key questions are whether the household member was ill or injured over the past 4 weeks (self-reported), how many days of activity were lost due to illness, who w as consulted for the illness and the time and monetary cost of consultations.
SECTION 5: EMPLOYMENT ACTIVITIES
Part A: Time Use
Six questions are asked to determine whether the person did any work (salaried work, own account including farming, or working for a household enterprise) in the last seven days or over the past twelve months. An individual who did not work during the past seven days is asked a set of standard questions about job search. If an individual did not work over the past year either, then the interviewer is asked to skip to Part 5H.
Part B: Main Job in the Past Seven Days
For those who did work over the past seven days, questions are asked about the occupation and industry, time devoted to the job, income earned from it, and also whether their parents did the same work. Detailed information is gathered for wage earners: cash and in-kind income, distance to work, and time spent commuting, size of the firm where the individual works, whether the firm has a union and whether workers receive various types of social benefits.
Part C: Secondary Job in the Past Seven Days
Pertains to the secondary job held over the past seven days.
Part D: Search for Additional Work
This part is asked of everyone who worked in the last seven days.
Part E: Main Job in the Past Twelve Months
This section pertains to the main job held during the past twelve months.
Part F: Job History
For the latest job prior to the current one, the individual is asked to describe the occupation, the type of industry, whether the individual was self-employed or working for someone else in their own household, and how long the individual was employed in that job.
Part G: Secondary Job in the Past Twelve Months
This part pertains to the secondary job held during the past twelve months. Questions are similar to those in Part 5C about the main job over the past seven days.
Part H: Other Activities
This part is asked of all individuals to whom Section 5 was administered. It records the amount of time spent by each person in housework activities, such as cleaning, food preparation, laundry, shopping, and fetching water or wood. In addition, a set of questions is asked about periods of unemployment during the past twelve months, including a question about seasonal unemployment in the agricultural sector.
SECTION 6: MIGRATION
This section is asked of all household members 15 and older. Information is obtained on the place of birth, the most recent migration (to the current place of residence) and the total number of times a migratory move was made.
SECTION 7: IDENTIFYING RESPONDENTS FOR THE SECOND INTERVIEW
This section serves the purpose of identifying the household members who will be the respondents for certain sections in the Round 2 interview (two weeks later) and making appointments with them. Appointments are made with respondents who will answer questions from the Round 2 sections on farming, non-agricultural businesses operated by the household, and household expenditures. Also, one woman per household is selected to answer questions on fertility.
SECTION 8: CHARACTERISTICS OF HOUSING
Information is recorded on the construction material of the walls, floor, roof, and windows of the household's dwelling. The floor area of the dwelling was measured and is recorded in square meters.
SECTION 9: AGRO-PASTORAL ACTIVITIES
Part A: Land
This part asks about the amount of land owned, rented and cultivated by the household, land sales, gifts and trades, and land sharecropped in and out.
Part B: Crops
For each of 32 crops grown, Section 9B collects information on: (i) the number of hectares devoted to each crop, including those already planted but not yet in production; (ii) the amount and value of the crops sold in the last twelve months; (iii) the value of the output retained as seedlings for next year's crops; and (iv) the value of the amount given to workers or as gifts; and (v) whether there was any inter-cropping of that crop.
Part C: Age of Tree Crops
For coffee, cocoa, and other tree crops, questions are asked about the proportion of plants too young to produce, in full production and at the end of their productive lives.
Part D: Agricultural Inputs and Expenses
This part records the use of agricultural inputs by the household, their cost and for what crops. It asks whether the input was obtained with cash or credit and the source of the input (market or government agency?). It also asks about the extent to which exchange of labor took place with other farmers, the incidence and terms of sharecropping, the extent of food storage, and the availability of agricultural extension services.
Part E: Processing of Food Crops
For each processed crop, how much was received as revenue from the sale of the product, and how much was spent for the cost of processing inputs?
Part F: Livestock
For each type of livestock, the questionnaire records the number of and value of livestock currently owned, the number of and value of livestock sold over the past year, the number of and value of livestock purchased over the last year, and the number of livestock lost over the year due to other reasons.
Part G: Animal Products
This part records the animal products produced and the amount that was received over the past year from their sale.
Part H: Mutual Aid for Livestock
This part requests the number of person-days devoted to a traditional system of mutual help among farmers and the number of contacts with government extension agents regarding livestock are recorded.
Part I: Livestock Expenses
Costs associated with raising livestock are recorded in this section. Where livestock inputs were obtained is also recorded.
Part J: Farm Tools
For a list of the main small tools used by Ivorian farmers, the number of tools owned by the household is recorded, by type of tool.
Part K: Farm Equipment
For each type of farm equipment (not tools) such as tractors, carts, vehicles and draft animals, questions are as ked about the value of the current stock, the value of transactions that took place within the past twelve months, and whether the household made money by renting equipment out to other farmers.
SECTION 10: NON-FARM SELF EMPLOYMENT ACTIVITIES
Information on up to three businesses per household is gathered in this section. If a household operates more than three businesses, the household head is asked to identify the three most important ones.
Part A: Revenue Information
Information on revenues from household businesses is obtained. In addition, questions about the number of workers, presence of a union, and provision of benefits are asked.
Part B: Expenditures
This part records costs incurred by each business, either in the form of c ash expenses or inputs from within the household, from a list of all major expenditures that may occur in a household enterprise, e.g. hired labor, raw materials, transportation and electricity.
Part C: Capital and Inventory
This part records the value of productive assets and stocks, such as unsold goods, buildings, vehicles or equipment.
SECTION 11: HOUSEHOLD EXPENDITURES AND INVENTORY OF DURABLE GOODS
Part A: Daily Expenditures
The amount spent since the enumerator's last visit is recorded for a list of non-food items that the household can be expected to purchase daily or very frequently.
Part B: Annual Expenditures
Expenditures on all remaining non-food items, including taxes, are recorded, with respect to two reference periods: since the last visit of the enumerator during Round 1; and over the last twelve months.
Part C: Inventory of Durable Goods
For a list of commonly owned durable goods, the following information is recorded: age of the item, payment made at the time of purchase, and an estimate of the value of the good at the time of the interview.
Part D: Expenditures on Remittances
Expenditures on remittances to non-household members are recorded. Other questions include: relationship of the recipients to the household head; the geographic destination of the remittances; and whether the money is to be considered a gift or a loan.
SECTION 12: FOOD EXPENDITURES AND CONSUMPTION OF HOME PRODUCTION
Part A: Food Expenditures
This part records expenditure on purchased food for the period bounded by the initial visit of the interviewer and over the last twelve months.
Part B: Consumption of Home Production
This part records the monetary value of the household's consumption of home produced food.
SECTION 13: FERTILITY
This section records the total number of pregnancies and live births for the randomly selected adult female in the household. Other questions in the section pertain to the use of maternity services and extent of breastfeeding for the latest birth; the number of still births and miscarriages; and the respondent's marital status.
SECTION 14: OTHER INCOME
Part A: Non-Labor Income
Records non-labor income from various sources such as social security, pensions, dividends and interest, tontine, gifts, inheritance and scholarship. Revenues from sales of household assets over the past twelve months are also recorded.
Part B: Remittance Income
Remittances received by the household from outside sources are recorded. Details include: amount of the remittance, relationship of the sender to the household head, geographic location of the sender's place of residence, and whether part of the remittance is to be repaid to the sender.
SECTION 15: CREDIT AND SAVINGS
Part A: Loans Made and Received
This part records the total amount of loans provided by the household to others, total amount borrowed from institutions or from other people.
Part B: Credit
If the household borrowed money, information is gathered about the lender and on the terms of the loan.
Part C: Savings
Total value of all savings is recorded.
SECTION 16: ANTHROPOMETRICS
This section is not printed in any of the questionnaires, which was introduced during the seventh month (in September 1985) of the first survey year. Section 16 was designed to obtain weight and height measurements for all household members interviewed, including both children and adults. The interviewer recorded the date of measurement, weight in grams and height in centimeters for each household member measured. If the person was not measured, the interviewer had to record the reason (person is away, ill, etc).
This section was introduced during the second year of the survey to enable the identification of panel households and panel household members from one survey year to the next and to track the movements of those who are no longer members of the household. This section was generated by computer as a unique form for every panel household, based on that household's data from the previous year. The first column contains the ID cod e of each household member interviewed during the previous year. The second column contains the name associated with that ID code. [Because only household members from the previous year are listed, and not visitors and other non-household members, there are sometimes gaps in the "old" ID codes on the list.] Also, contained in a small box next to the name, are the age and sex of that household member. This was designed to assist the interviewer in identifying the household member in question, but was not reentered in the data entry program. All of these were generated by computer and provided to the interviewer. The answer to Question 1, "Is this person found in the household roster this year?", determines whether or not that person is still part of the same household. For those household members who are no longer in the household, this section asks questions about the reason for their absence (death or moving away). If the household member moved away, questions are asked about why he moved, and if he moved out of the cluster, where he moved to. If the person is still part of the household (i.e. answer to Q.1 is yes), then the new ID code for that person is copied from the household roster as answer to Q.2. Having the old and new ID codes enables the linkage of panel household members from one year to the next.
>> COMMUNITY QUESTIONNAIRE
The Community Questionnaire measures the access of the community to economic infrastructure and basic social services. It was completed in each rural cluster. The objective of the Community Questionnaire was to measure characteristics common to all households in a cluster. The questions were asked of a group of respondents who had detailed knowledge of the village. Typically, the group of respondents included the village chief and other prominent members of the community.
>> PRCE QUESTIONNAIRE
A price survey was conducted in parallel with the CILSS household survey as part of the same project. The main purpose of the price survey was to provide price data with which regional or cluster level price indices could be constructed.
The items in the price questionnaire were chosen to include:
1. Frequently purchased items, according to the results of the 1979 Côte d'Ivoire Budget-Consumption Survey
2. Items for which data were already being collected by the Price Index division of the Direction de la Statistique
3. Food items included in the section of t he household questionnaire on consumption of home-grown produce (this would allow for the conversion of the value of home production consumed by the household into quantities).
The price survey collected information on prices of both food and non-food commodities in the main market of each of the clusters in which the household survey was conducted. The price data were collected at the same time that the households in the c luster were surveyed. When there were at least three vendors selling an item, up to three prices were collected for each item, usually from different parts of the market. Weighing scales were used to determine the exact weight of food items. Weights were to always be recorded in grams. None of the items weighed were purchased by the survey. If a vendor refused to allow his products to be weighed, then a client had to be awaited so that information based on the sale could be recorded.
For the first two years, 1985 and 1986, information was collected for 22 commodities, including 18 food and 4 non-food items. Each item in the price survey was numbered; these numbers were used in naming variables in the price data sets. In 1987, the number of commodities was increased.
>> HEALTH FACILITY SURVEY QUESTIONNAIRE (CILSS 1987)
During the third year of the CILSS Survey, a new Health Facility Survey was implemented to collect more detailed information on the quality and type of health services available to CILSS households. The Health Facility questionnaire was administered to the dispensary and maternity clinic closest to or located within each cluster of households, in both urban and rural areas. Based on the community questionnaires of the second year, a list was made of all the nearest dispensary and maternity clinics to each cluster. Except for urban clusters, these facilities were generally not in the same place as the households being interviewed. This survey was conducted in 1987 only.
The purpose of collecting the health facility data was to analyze household demand for medical care as a function of the characteristics of the services. Accordingly, the facilities surveyed are not a random sample of all health facilities in Côte d'Ivoire. Therefore, the data should not be used to generate national or regional statistics on health services.
Other forms of data appraisal
The general consensus is that the quality of the CILSS household data is very good. An informal review of data quality conducted by Ainsworth and Mehra (1988) assessed the 1985 and 1986 CILSS data in terms of their accuracy, completeness, and internal consistency. The CILSS household data were found to score high marks on each of these three counts.
One measure of data quality is the extent to which individuals in question respond for themselves during the interview, rather than having proxy responses provided for them by other household members. The investigation of CILSS household survey data for 1985 and 1986 showed that 93 percent of women responded for themselves to the fertility section and that 79 to 80 percent of all adult household members responded for themselves to the employment module. The percent of children responding for themselves to the employment module was far less, 43 to 45 percent. Nevertheless, these rates were found to be higher than for the Peru Living Standards Survey (29 percent).
Investigation of several variables and modules in the CILSS (sex, age, parental characteristics, schooling, health, employment, migration, fertility, farming and family business), found that missing data in the household survey are rare. Rates for missing data were found to be close to 0 (0.01 to 0.05 percent) in many cases, but in any case, no higher than 0.76 percent.
>> Internal Consistency of Data
The Household Questionnaire was almost entirely pre-coded, thus reducing errors involved in the coding process. Also, the decentralized data entry system allowed for immediate follow-up on inconsistencies that were detected by the data entry program. Household and personal identification codes were recorded in each section, facilitating merging data across sections.
A comparison of the distribution of household size in the 1988 Côte d'Ivoire Census with the 1988 CILSS showed a strong similarity. Grootaert (1992) provides a brief assessment of the CILSS data quality. "...Quality control during CILSS data collection and data entry were extraordinary in comparison with usual survey practice (see Ainsworth and Munoz, 1986; Grootaert, 1986; Daho, 1992). The many analyses undertaken with the data have shown a high degree of internal consistency in the data. ... Selected demographic variables from the survey have been compared with the Côte d'Ivoire Fertility Survey and were found to be consistent (Ainsworth, 1989). Farming information in the CILSS was evaluated against other sources and found to be quite good (Deaton and Benjamin, 1988). .... The pattern of household expenditure observed for Côte d'Ivoire a s a whole is entirely consistent with the pattern of macroevolution recorded in the national accounts and in other macro data. Both the upturn in 1986 and the decline in 1987-88 are picked up by the CILSS data. In summary, we feel confident that the CILSS data are a valid data source to explore the evolution of welfare and poverty..."
Another assessment of CILSS data quality, "The CILSS: A Preliminary Assessment of Data Quality" by James Daniel and Lionel Demery, examines some demographic variables from the household roster and finds that errors as a percentage of total observations are very small and that data consistency improved over time. The paper also examined consistency among the panel households from one year to the next and found that, while the gender variable had a very low error rate (1%), the error rate for age was a higher 10.7%. (If the difference in reported ages was more than one year, allowing for the fact that panel households were interviewed one year apart, that was defined as an error). However, since many people in rural areas of Africa do not know their age with accuracy, and age reporting involves some degree of guesswork, this finding is probably not reflective of any underlying data problems.
Detailed assessments of the quality of CILSS Community and Price Surveys are not available.
World Bank LSMS
In receiving these data it is recognized that the data are supplied for use within your organization, and you agree to the following stipulations as conditions for the use of the data:
1. The data are supplied solely for the use described in this form and will not be made available to other organizations or individuals. Other organizations or individuals may request the data directly.
2. Three copies of all publications, conference papers, or other research reports based entirely or in part upon the requested data will be supplied to:
Institut National de la Statistique
The World Bank
Development Economics Research Group
LSMS Database Administrator
1818 H Street, NW
Washington, DC 20433, USA
tel: (202) 473-9041
fax: (202) 522-1153
3. The researcher will refer to the 1986 Côte d'Ivoire Living Standards Survey as the source of the information in all publications, conference papers, and manuscripts. At the same time, the Institut National de la Statistique is not responsable for the estimations reported by the analyst(s).
4. Users who download the data may not pass the data to third parties.
5. The database cannot be used for commercial ends, nor can it be sold.
Disclaimer and copyrights
The user of the data acknowledges that the original collector of the data, the authorized distributor of the data, and the relevant funding agency bear no responsibility for use of the data or for interpretations or inferences based upon such uses.