3.1 Information to be Collected
3.2 Sources of Dairy Consumption Data
3.3 Primary Data Collection
3.4 Field Instruments
The concept of consumption may be understood to have either of two related meanings, both involving the notion of an object's "disappearance" or transformation. In economic terms, consumption of an object may be considered to have occurred when it has been removed from the market by some exchange mechanism, such as when a transfer takes place between a seller and a buyer of a good. The more commonly understood meaning of consumption is its physical meaning of having been eaten. It is this latter meaning which applies in this characterisation methodology. Some other important definitions are:
Dairy product: A dairy product shall be defined as milk or any product derived from milk. Within the general class of dairy foods, different products will be differentiated by their physical composition or form, or where the market or consumer differentiate them.Consumption pattern: The combination of products consumed, the frequency of their consumption and the quantities consumed.
Mode of consumption: The manner in which the dairy product is consumed or used, e.g. milk may be taken directly or in tea and either cold or warm; butter may be used as a cooking fat, an edible spread, or a cosmetic.
Regulated market: A market in which government exercises control, for instance through price regulations, mandatory standards of hygiene, location-specific marketing licenses, etc. Conversely, an unregulated market is here defined as one which effectively operates beyond the control of government and is for all practical purposes "self-monitored", such as in dairy markets dominated by "over-the-fence" sales to neighbours.
The information necessary for the characterisation of the consumption system derives from three sources: i) the functional parameters necessary for the description of the system, ii) the performance indicators required for assessing the effectiveness of the system, and iii) the hypotheses to be tested in the characterisation phase of the Conceptual Framework.
Table 2 sets out the hypotheses contained in the Conceptual Framework which directly relate to the consumption system. In summary, they postulate that dairy product consumption is dependent on household location, consumer purchasing power, and consumer tastes and preferences. Also listed in the table are the data required to test each of the hypotheses. Not all of the hypotheses advanced will be tested during the characterisation phase but some will more appropriately be addressed at a later stage in the research process. The hypotheses in Table 2 have been identified as "core" hypotheses, and are considered essential for elucidating the forces underlying current dairy-consumption trends. Given, however, that consumption systems in different areas will exhibit their own unique characteristics, researchers are encouraged to formulate additional hypotheses, the testing of which will add to the collective understanding of dairy consumption systems.
Note that it is important at this stage to identify potential sources of the needed data. Not only will the identification process assist in evaluating the testability of the hypotheses, but it is also necessary for selecting the techniques to be used for gathering this information. Identifying the information sources and data-collection methodologies will furthermore facilitate organising the research.
Table 2. Phase 2 core hypotheses on consumer demand and dairy consumption.
|
Hypothesis |
Phase 2 |
|
|
Data required |
Sources of information |
|
|
(20) Consumption levels of dairy products are positively correlated to population concentration |
Quantities of domestic dairy products consumed/caput |
Consumers |
|
Population density of the smallest administrative division where the house is located |
Secondary sources |
|
|
(21) Tastes and preferences are more important determinants of dairy consumption than HH income |
Ranking of tastes and preferences for available dairy products |
Consumers |
|
(47) Dairy product share of household food budget increases with household income |
HH food expenditures |
Consumers |
|
(C1) Household location is a more important determinant of dairy product consumption than household income |
HH location |
Consumers |
|
(C2) Cooling facilities are a prerequisite for expansion of dairy consumption |
Availability of cooling facilities |
Consumers |
|
(C3) Levels of HH dairy consumption are significantly correlated with the personal attributes, e.g. education, income, etc. of the HH food budget manager |
Food budget manager personal attributes |
Consumers |
|
(C4) Price is the most important factor for consumers in deciding among dairy products |
Dairy product prices |
Key informants; consumers |
|
(C5) Milk is regarded as a food only for children |
Intra-household dairy product consumption/distribution |
Consumers; key informants; secondary sources |
|
(C6) Negative cultural beliefs and misunderstandings about dairy products inhibit dairy consumption |
Intra-household dairy products consumption/distribution |
Secondary sources; key informants; consumers |
|
(C7) Low per capita dairy consumption in SSA is by choice rather than the result of external constraints |
Dairy consumption levels |
Consumers; key informants; secondary sources |
|
(C8) Among HH's producing milk, HH consumption needs take priority over milk sales |
Ratio of milk sold to milk produced |
Dairy producers |
N.B. Numbers in parentheses correspond to the hypotheses on consumption originally presented in the Conceptual Framework base document, "Improvement of dairy production to satisfy the growing consumer demand in sub-Saharan Africa: a conceptual framework for research", Rey et al, 1993. Those preceded by "C" have subsequently been added as core hypotheses.
Table 3 indicates the key functional parameters and performance indicators for the description and evaluation of the consumption system.
Table 3. Information required for the characterisation of the consumption system.
|
Data |
Functional parameters |
Performance indicators |
Data sources |
|
Structure of consumption (consumption pattern, consumption profile) |
Dairy products consumed |
Kg LME/cons. unit/year |
Consumer |
|
Consumption environment |
Urban population and pop. growth rate (persons/km² at administrative level for both national and site levels) |
|
Gov't census |
|
Government taxes/subsidies on dairy products and their substitutes (Y/N) |
Price ratios: dairy products to their substitutes |
Key informants, secondary sources (Boards, Gov't.) |
|
|
HH location (urban/pert-urban/rural) |
Consumer perception of constraints to consumption |
Consumer |
LME: liters milk equivalent; HH: household.
These data, together with those required for testing the hypotheses associated with Phase 11, generate the minimum data set necessary for the system's characterisation. Table 4 states explicitly the information to be collected. The table also identifies the origin of the data need, and in the far right column, notes the specific question(s) in the household questionnaire which will generate the data.
Table 4 is a compilation of the information needs identified in Tables 2 and 3 as interpreted at the level of the consumption unit. As can be seen from the table, the consumption unit in this case is the household, and it is at this "level of observation" that consumption data are collected, i.e. the focus is on the consumption behaviour of the household members as a group rather than on the behaviour of any particular individual.
Table 4. Minimum data set for characterisation of the consumption system.
|
Parameter to be measured |
Origin |
Question(s) |
|
Number of household members (taking at least one meal a day at home) |
PI |
22 |
|
No. of HH consumer units |
PI |
23 |
|
Electricity supplied to house: If yes > > > fridge? |
H-C2 |
8, 9 |
|
Household food budget manager: sex, age, ethnic group, religion, education level; is FBM the HH head? |
H-C3 |
10, 11, 13-17 |
|
Age, sex, religion and ethnic group of HH head |
H-C3 |
18-21 |
|
Types of dairy products eaten |
FP, PI, H21 |
24, 25 |
|
Quantities of dairy products eaten |
FP, PI, H20, H21 |
25 |
|
Frequency of consumption of each dairy product |
FP, PI, H21 |
24 |
|
Source of each dairy product: regulated vs unregulated market, home production |
FP |
25 |
|
Purchase price of each dairy product |
H-C4 |
25 |
|
Consumption form/mode of each dairy product |
FP |
25 |
|
Complementary foods eaten with each dairy product |
FP |
25 |
|
Substitute foods for each dairy product |
FP |
25 |
|
Shortage periods (number of months and specific months) |
PI |
25 |
|
Place and frequency of consumption outside home |
FP |
25 |
|
Intra-HH distribution of milk consumption |
H-C5 |
30 |
|
Beliefs/taboos about dairy consumption |
H-C6 |
27-30 |
|
Preferences within dairy product categories (liquid milk, sour/fermented milk, cooking fats, edible spreads, cheeses, cosmetic fats) |
H21 |
31 |
|
Constraints to increased dairy product consumption |
FP, PI, H-C7 |
25, 26, 32 |
|
Is consuming household also producing milk? |
FP |
33 |
|
Is producing household also trading/selling dairy products? |
H-C8 |
33 |
|
Average weekly food expenditure |
FP, PI, H47 |
34 |
|
Household income category or level |
H21 |
35 |
PI = Performance indicator; HH: household; H: Hypothesis number; FBM: food budget manager; FP: Functional parameter.
For current purposes, the household shall be defined as being comprised of those individuals who take at least one meal a day at home. To maximise data accuracy, attempts should be made to collect consumption information from the individual who is responsible for acquiring food for the household, referred to here as the "food budget manager". While it is assumed that the majority of households will be purchasing the dairy products they consume, many households may also be dairy producers. This, however, should not prejudice the data collection. Similarly, it is expected that consumption of dairy products outside the household will vary across sites. Key informant surveys should provide insights on consumption of dairy products outside the household and thereby indicate whether or not a survey of food institutions is in order.
Although the household is identified as the primary level of observation, it is still important to consider using secondary data sources. Key informants are often reliable and accessible sources of information on population descriptors, income range, dairy products generally available and consumed, and cultural practices. Published statistics can also be valuable sources of information on distribution of income, retail prices, urban population density and population growth rates, and imports of dairy products.
3.3.1 Survey method
3.3.2 Definition of the population and sampling technique
For collection of information on household consumption, the method proposed is a formal survey of a representative sample of households within the dairy shed2 under study. The characterisation survey is a single visit survey with a reference period of the last one week. Based on previous household level survey experience, information collected in this manner is sufficient for characterisation of the consumption system. Notice must be taken, however, of variation in consumption during the year, i.e. intra-annual consumption variation, due to such factors as production fluctuations between rainy and dry seasons, seasonal off-farm labour demand, and religious or traditional festivals, e.g. Ramadhan, Christmas etc. Unless there is special interest in consumption during high or low peaks, surveys should be planned so as to avoid these periods. At sites where there is evidence of strong seasonal consumption, repeat visits to the same household will permit better characterisation of the consumption system. Where feasible, repeat visits will also benefit survey efforts by permitting corroboration of previously collected data and improving estimates of consumption.
2 The "dairy-shed" concept applied in this working document series has both a geographic dimension and a specific market orientation. The dairy systems with which this methodological series is concerned are focused on supplying dairy products to urban consumption centres. Geographically, the dairy shed extends to that local area supplying dairy products to the urban area under consideration.
Necessarily, the representative household sample must include non-consuming as well as consuming households, and the sample is therefore selected from the household population in the dairy shed. The households may be situated in urban, pert-urban or rural environments. Among the consuming households, however, the sample is intentionally restricted to those households which consume dairy products on a regular basis, which for purposes of the characterisation survey shall be any household having consumed any dairy product in the last week, excluding holidays or other special occasions. Because the size of this population of consumers is unknown in advance, the sample of households to be interviewed cannot be randomly chosen with a given sampling intensity. Based on previous sampling experience and resource considerations, however, a sample size in the range 120 - 150 households is proposed.
The sampling methodology recommended for identifying the sample households is the adaptive cluster procedure (Thompson, 1992), a modified application of traditional cluster sampling (Cochran, 1963). Adaptive cluster sampling is appropriate in situations where the population size is unknown and is particularly useful for adoption or impact studies. The final sampling unit is a cluster of elements, in this case, a group of households. Depending upon the sample size to be drawn and the data required to test the postulated hypotheses, it may be desirable to stratify the population using pre-determined or naturally defined groups.
To begin the sampling process, individual households are randomly selected from the total list of household units. These households serve as the starting points for the formation of household clusters (see Figure 2) of sizes 1 to n (defined later). Sampling continues from these points according to specific rules. Two types of household units are defined consuming units and non-consuming units. Eligibility rules of the adaptive cluster procedure for sampling of the consumption system are as follows:
A household is eligible for observation if:
1. it is part of a selected cluster, and;2. it has consumed any dairy product during the past seven days, apart from festivities or holidays; or,
3. it constitutes an edge unit3, i.e. a non-consuming household bordering a cluster of consumers;
4. it is a consuming household outside a selected cluster but does not lead to a sub-sample size greater than 5% of the total sample size; and,
5. the household is less than 1 km from the preceding household.
3 Whether an edge unit is merely an infrequent consumer of dairy products or a complete non consumer, there is still valuable information to be gathered from these households that will contribute to our understanding of determinants of dairy consumption. Information on household composition, education and income level, consumption constraints, food budgets and food consumption patterns may provide important insight into household consumption behaviour. The questions to be posed to edge units are designated in the proposed household consumption questionnaire found in Annex 2.
Bearing these rules in mind, the sampling procedure may then be described as follows:
1. Determine the number of primary consuming households, n, required (a sample of size n> = 25 is suggested) [this defines the number of clusters];2. If stratification4 is to be applied, allocate the clusters to each stratum, e.g. urban, peri-urban or rural household clusters, in proportion to the size of that stratum in the population if known, i.e. if a particular stratum represents 50% of the population then it should be allocated 50% of the total sample clusters;
3. Sampling will be without replacement, i.e. a cluster will be surveyed only once, and only households having the desired attributes specified under eligibility rule 1 (specified above) will be interviewed. Households selected randomly that do not possess these attributes will not be interviewed; and
4. For each of the initially selected primary households, all surrounding household units will be interviewed that also meet the observation criteria. Terminate the cluster formation when an edge unit is encountered or a maximum 5% total sample size is attained.
4 For further comments on stratifying samples see Annex 4
The adaptive cluster method will lead to the formation of a maximum of n clusters varying in size from 1 (where the first household sampled fails to meet the interview criteria) to 0.05 × n. Figure 2 below illustrates the adaptive cluster sampling procedure. Details of computing the mean and variance of data values collected using this methodology are given in Annex 5.
As a preliminary step to collecting information on the consumption system, secondary sources should be reviewed and discussions held with knowledgeable key informants. Information supplied by the key informants will, in some cases, fill data requirements of the characterisation process; in other instances, information from key informants will be needed to properly design the household survey instrument and to guide cluster sampling. Annex 1 describes the information which should be collected through key informants at the dairy-shed level. The example form presented in the annex suggests how the data collection can be organized to facilitate efficiency.
It is anticipated that much of the information required for characterisation of the consumption system will be gathered through primary data collection at the household level. In the preceding sections, the methodological background, hypotheses! minimum data requirements and sampling technique underlying this data collection have been described. The survey instrument, here a household questionnaire, brings these components together into a single, logically constructed document, the administration of which will generate the required data, it is expected, in an efficient manner. While there always exists the possibility of generating poor-quality data due to improperly worded questions, unskilled enumerators and other sources of non-sampling error, this possibility can be greatly reduced by closely following the preceding steps. Doing so will also result in the minimum number of questions that need to be asked. The probability of poor or unuseful data can be further reduced by ensuring that questions are to the point and unambiguous in their meaning. A proposed household level questionnaire developed according to the minimum data set (Table 4) is presented in Annex 2. In addition, ILCA will, upon request, furnish to users of the characterisation methodology, computer diskettes containing screen designs for data entry and a limited number of programmes for generating the functional parameters, performance indicators and hypothesis tests.
