FAO STATISTICAL DEVELOPMENT SERIES
PROGRAMME FOR THE WORLD CENSUS OF AGRICULTURE 2000
Food and Agriculture Organization of the United Nations
A programme as large as a census of agriculture has many methodological problems to consider. This chapter, while not exhaustive, reviews the theory on a number of problems that will be encountered.
The first subjects covered are objectives and scope, which are primary considerations in the planning process. Other subjects dealt with include coverage, complete or sample enumeration, and the census frame.
4.1 Three basic objectives of a census are:
4.2 Since the agricultural census is not a frequent data collection activity, it is natural to associate it with those aspects of agricultural structure which change relatively slowly. Most slow changing aspects relate to organizational structure. Analysis of these aspects also helps to explain, in a general way, causal structure relating to technological changes, economies of scale and supply response. It is necessary to bear in mind that benchmark data from a census may refer to a year affected by drought, disease or other phenomena that makes it atypical.
4.3 The second census objective is to provide a frame for other types of sample surveys on agricultural holdings. Such a frame may be a list of holdings or some other characteristic that helps to identify agricultural holdings.
4.4 The census is the preferred method of providing consistent information for local areas, legal districts or agro-ecological zones. Such local-level estimates are necessary for making policy decisions at local level.
4.5 Census data are also useful in preparing sample designs (such as probability proportionate to size sampling, stratified sampling, etc.) and in obtaining estimates through ratio and regression estimation techniques. Useful stratification criteria are included in the structural aspects of agriculture: sex and age of holder, total area, area by crops, number of livestock, number of trees, integration of holding with enterprises, etc.
4.6 The basic objectives, mentioned above under 4.1, may be too restrictive for some countries. However, countries where statistical systems are not well developed are cautioned, based on prior experience, from expanding the objectives. This caution results from the negative experiences of many developing countries attempting agricultural censuses with wide objectives. Countries with developed statistical systems and sufficient resources may, however, wish to add one or more of the following objectives:
4.7 When adding more census objectives, countries should ensure that the basic objectives mentioned in paragraph 4.1 are not adversely affected. To establish their census objectives, countries should carefully evaluate available resources against requirements.
4.8 The scope of the agricultural census concerns the data to be collected. Consistent with the basic census objectives, the scope is broadly defined as:
These items are described further in Chapter 5.
4.9 The above list omits many aspects relevant to agriculture (including production and quantities of inputs used such as fertilizers, seeds, etc. and yields), which although very important, are not suitable for collection in a multipurpose single enumeration period census. These aspects should, preferably, be covered by sample surveys or other data sources more appropriate to current statistics.
4.10 If a subject included in the census scope, such as shifting cultivation, is not relevant to a particular country, it should be disregarded.
4.11 For general guidance in determining the census scope, countries are recommended to critically review their level of statistical development based on the following criteria:
4.12 Countries ranking low in all or most of these criteria are recommended to restrict their census scope to those items presented in Chapter 5 which are marked as "essential" as based on the list in paragraph 4.8. Countries at a medium level of statistical development may wish to include also those structural items for which tabulations of separate administrative units and agroecological zones are considered essential. Countries with fully developed statistical systems may wish to further enlarge their census objectives (as mentioned in paragraph 4.6). However, all countries are cautioned against overloading the agricultural census with such extensive coverage that collecting, processing, storing, retrieving and disseminating data becomes too burdensome. Small, costefficient sample surveys, based on an adequate frame provided by the census, are the recommended sources for obtaining such additional data. In all cases, if accurate structural data are readily available from other sources such data should be excluded from the census scope.
4.13 The statistical unit for the agricultural census, for which required data items are gathered, is the agricultural holding as defined and explained in paragraph 5.7. It should be noted that the agricultural census will not include communal grazing land, fallow land under shifting cultivation and, generally speaking, all land not operated by agricultural holdings, which may be important categories in many countries.
4.14 The agricultural census should, in principle, cover agricultural holdings in the entire national territory, including both rural and urban areas. Complete coverage is particularly important for providing a frame for other subsequent agricultural sample surveys. In many countries, a minimum size limit is adopted for holdings included in the census. The rationale for this minimum size limit is that generally there is a large number of small holdings which makes a very small contribution to total agricultural production but whose inclusion in the census greatly increases the workload. Although this argument is acceptable in some countries, it cannot be defended in most countries where very small holdings may contribute substantially to total agricultural production. Small holdings are often a significant part of the agricultural structure; without information on such holdings a complete picture cannot be provided. Countries that exclude small holdings from their agricultural censuses are strongly urged to set the minimum size limit as low as possible and to take steps to collect data through special sample surveys from excluded holdings. No uniform lower limits for production factors such as area, number of trees, livestock, volume or value of output, for example, are suggested.
4.15 Some countries exclude urban areas and/or semidesert regions from the agricultural census. Omission of urban areas containing gardens used mainly for vegetable production or intensive animal husbandry activities, such as dairying or poultry farming, may result in loss of valuable information. Similarly, to exclude semidesert regions may omit important livestock resources - nomadic tribes may keep large numbers of livestock in such regions. As mentioned under census frame", (see paragraph 4.29) a population census, if taken prior to the agricultural census, can identify holders in urban areas and semidesert regions.
Role of women in agriculture
4.16 It is widely recognized that women's participation in agriculture is of great importance, yet their contribution to agricultural development is in most cases inaccurately reported and often underestimated. The lack of relevant data on women in agriculture has been identified as a major constraint to the integration of women in development planning and to effective project formulation. Therefore, it is recommended that emphasis on the need to collect census statistics, disaggregated by sex, be maintained throughout the process of census planning, questionnaire design, data collection, processing and dissemination. In this connection, countries should pay particular attention to:
4.17 The Programme for WCA 2000 refers to a national agricultural census taken during the tenyear period, 19962005. Countries are recommended to take at least one agricultural census during this period. However, where rapid changes are occurring in their agricultural structure, countries may prefer to conduct censuses at fiveyear intervals. If the census scope is kept very limited, a fiveyear interval agricultural census may be feasible and can provide a frame for intervening annual sample surveys that is kept more up-to-date.
4.18 The census reference year is a period of 12 consecutive months, generally encompassing the various time reference dates or periods of data collection on individual census items; it may be a calendar year or an agricultural year. For the census reference year, countries are asked to use a 12month period covering as much as possible of the calendar year 2000, or, if this is not possible, a year close to 2000, to make international comparisons more meaningful. In general, the time reference for specific items is the agricultural year or the day of enumeration. The reference period for specific items is indicated in Chapter 5.
4.19 The actual period in which the census enumeration is taken will be during or immediately after the census reference year; however, the enumeration period should be short. This is essential to avoid omissions or duplications because of variations in information collected, such as changes in livestock numbers created by movements. This also applies when census enumeration is carried out using more than one round, in which case even greater care should be exercised to ensure meaningful tabulation of data obtained from individual holdings. These are additional reasons for limiting the census scope.
Complete or sample enumeration
4.20 An agricultural census may be conducted through complete enumeration, or sampling or a combination of both. A complete enumeration involves obtaining information from all holdings in the country (see paragraphs 4.144.15) whereas sampling involves collecting information from only a predetermined number of holdings. In the past, an agricultural census conducted through sampling was referred to as a sample census of agriculture. In this Programme, the term sample enumeration is preferred, as the word census implies a complete enumeration. Sample enumeration is also consistent with the terminology used by the UN Statistical Office. Complete and sample enumerations are sometimes combined in various ways to conduct an agricultural census: completeenumeration for a few items and sampling for the remainder; or complete enumeration of certain areas or certain holdings (such as institutional holdings or holdings above a predetermined size) and sampling of other areas or the remaining holdings.
4.21 In a national statistical programme for food and agriculture it is desirable to have the agricultural census carried out by complete enumeration, if its full objectives are to be met. Only in this way is it possible to obtain statistics for every geographical, agroecological or administrative region, irrespective of size. This information is important for a comprehensive understanding of agricultural structure. A sample enumeration cannot provide estimates sufficiently precise at low levels of geographical areas. Complete enumeration also provides a frame on which to base the design of efficient sample surveys, such as an annual production sample survey or a farm management sample survey. The complete enumeration is also useful to identify certain holdings with special features, such as holdings operated by women, or holdings with only rented land, etc. This information is extremely important for rural development programmes oriented towards specific target groups. A sample enumeration will not provide such a complete list of holdings with these special features. Prior information on holdings is not required for a successful complete enumeration, whereas prior information is essential for designing an efficient sample enumeration.
4.22 On the other hand, complete enumeration is extremely demanding on statistical and other resources including transport and communications systems; management of organizations both centrally and in the field, which need to be staffed with numbers of qualified personnel; certain basic facilities, such as detailed maps, field supplies and data processing equipment; and funds for salaries, per diem, travel and other costs.
4.23 Sample enumeration is undertaken when these resources are inadequate to allow complete enumeration. A well designed sample enumeration requires more limited resources, may be more manageable and may produce equal or better quality data than a complete census programme at the national or large regional area levels. It should be noted that a successful sample enumeration requires expertise in sampling theory. Frame preparation, stratification, sample selection, developing formulae for estimates and their sampling errors are best accomplished by professionals trained in and with considerable experience of sampling techniques.
4.24 The census objectives and practical resource limitations must be balanced when choosing between complete or sample enumeration. Countries are urged to undertake research and preparatory studies before deciding on the most appropriate method to use.
4.25 A frame provides the means of identification of the population of interest (i.e. agricultural holdings) and can be either a set of lists or maps identifying the holdings. The frame for an agricultural census should cover all holdings in the country and is needed whether a complete or sample enumeration is to be undertaken.
4.26 An ideal census frame is a complete list of all holdings. In such a frame, each holding is shown without omissions or duplications and does not include any units other than holdings.
4.27 A register of holdings (farm registers) which is well established in some countries may be close to an ideal frame. These registers contain regularly updated information which takes account of continuity of holdings over time, as well as their appearance and disappearance. A farm register has a fixed reference number for each holding and basic information on its characteristics is entered periodically. The register is particularly useful when combined with an adequate computer storage information system, thus facilitating data retrieval, tabulation and analysis. If a complete farm register is developed independently of an agricultural census, then the second census objective (providing a frame for specialized sample surveys, as described in paragraph 4.3) can be omitted. Under such conditions, the census scope can be expanded, and can more rationally become a sample enumeration. However, the work volume and difficulties in establishing a farm register with basic information for each holding and its subsequent updating, in particular, is so great that even in statistically developed countries only a few basic items can be maintained in the register. Accordingly, while the possibilities of building up computerized farm registers and keeping them updated should be considered by all countries with the necessary facilities and regulations, the need for a census will continue.
4.28 One form of partial farm register may exist in the form of records of state, cooperative and institutional holdings, or list of parties contracting to the state, in countries with centrally planned economies. Almost inevitably there will be holdings which do not appear on such a register and other means of completing the frame will be necessary.
4.29 Most countries do not have a farm register of any sort and, therefore, need a frame constructed for the census purpose. One such frame can be obtained from a population census, if it is carried out prior to the agricultural census, and if relevant questions for identifying holders are included in the questionnaire. Such a source is useful, even in countries where a farm register is maintained, in order to update or complement the register, particularly in cases where the register does not include holdings of holders having other occupation(s). It is important to note, however, that a frame obtained from a population census may be outdated if the time lag between the population and agricultural censuses is too long. Countries are strongly recommended to coordinate the programme for these censuses. Such coordination is, in fact, necessary within the general framework of a national statistical programme for food and agriculture, where all data collection activities are interrelated not only with respect to scope but also with respect to operations, including timing, personnel, equipment, etc.
4.30 Other possible frames for an agricultural census include a list of enumeration areas prepared for the population census, a list of localities (villages), aerial photographs and maps. A population census enumeration area has clearly identified limits so that, while all national territory is covered, there is no overlapping. In rural areas, one enumeration area may be a village but larger villages are sometimes divided into several enumeration areas. In some rural areas the population live well dispersed on their holdings and great care is needed to ensure that enumeration areas are well defined.
4.31 If a complete enumeration is carried out, all holdings in each enumeration area or locality have first to be identified by means of an interview with the local authorities and a visit to each household. Identifying all holdings is tedious. If the census is to be conducted using sample enumeration a sample of enumeration areas or localities is selected as the firststage sampling units within which the process of identifying the holdings is carried out. For an efficient sampling scheme, prior information relevant to holdings is needed on each enumeration area or locality. In cases where information such as number of holdings, main types of crops grown and number of livestock is not available, one may have to be satisfied with the total population of the enumeration area or locality. The more information that is available the better the sample designs can be.
4.32 Maps, satellite imagery and aerial photographs can also be used for construction of a frame, particularly for sample surveys. In this case, a sample of welldefined areas is chosen for collecting data. In area sampling, rules are developed to associate data collected on sample areas with holders operating land in those sample areas. If aerial photographs and maps are not readily available, their cost should be compared to the cost of other frames when deciding upon the frame to be used, taking into account that an area sample frame is often believed to be superior to other frames because it requires less frequent up-dating. Satellite imagery has been used increasingly for construction of sample census/survey frames and its use is recommended, particularly for improving sampling design, through better stratification and for improvement of cartography.