No 6
ROME, 1995



The purpose of this chapter is to describe the procedures and problems encountered in the construction of the main types of frames needed to conduct agricultural censuses and surveys.

The construction of the frame is the most important preparatory activity for an agricultural census. For censuses conducted on a complete enumeration basis, it refers, in most cases, to the preparation of a list of enumeration areas with the approximate number of holdings/holders' addresses in each area. The agricultural census frame is therefore essential for organizational purposes and for ensuring the proper coverage of all agricultural holdings during data collection.

For censuses conducted by sample enumeration, a sampling frame is needed and has to be constructed for each sampling selection stage and a non-zero probability of selection has to be assigned to each sampling unit of the frame. There are two basic types of sampling frames: area frames and list frames corresponding to area sample designs and list sample designs. There are also multiple frame designs that combine an area sample with a list sample design, and for which the area and the list frames should not overlap. The construction of these three types of frames is treated separately.

Geographic Information Systems utilizing topographic charts, satellite images and aerial photos as well as Geographic Positioning Systems may also be utilized to improve the traditional methods of area frame construction and sample selection.

Definition of frame

7.1 The sample design of an agricultural census may have one or more sampling selection stages. As defined in survey sampling texts, and already mentioned in Chapter 6, the sampling frame for a given sampling selection stage is the set of sampling units from which the sample is selected, together with their probabilities of selection. Since the census characteristics or items that one wishes to study are defined in the holdings, if the sampling units are not the holdings, rules should be established to give a value of the characteristics in each sampling unit as a function of its values in a group of associated holdings.

7.2 A list frame is a frame used for the last selection stage of a list sample design. List frames are formed by lists of holdings or holders' addresses. In particular, the frame of a census conducted by complete enumeration (100% sampling) is the list of all holdings or holders addresses in the country.

7.3 An area frame is a list of all segments (land areas) of the country used for the last stage of selection of an area sample design.

7.4 A multiple frame design is a combination of the above two.

Practical approaches of frames

7.5 A sampling frame, as defined above, most often does not exist in practice. For example, a complete list of all holdings in the country is never available before conducting a complete enumeration census. Therefore, in practical applications, to be able to enumerate all holdings a preliminary frame is indispensable. In applied survey sampling, for convenience such preliminary frames are simply called frames. Therefore, a frame for each sampling stage of an agricultural survey design is a set of physical materials (maps, census statistics, lists, directories, records) that cover all holdings and allow for providing suitable sampling units. From now on, this will be the definition of the frame adopted.

7.6 A prerequisite for the organization of an agricultural census, whether conducted by complete or sample enumeration, is the preparation of suitable frames.

7.7 In developing countries it is advantageous to use the opportunity of conducting an agricultural census to prepare an efficient sampling frame which can be utilized not only for the census but also for subsequent annual or seasonal agricultural surveys, thus improving the agricultural statistical system.

7.8 The preparation of a frame for an agricultural census conducted by complete enumeration of all holdings, and similarly the preparation of the frames for an agricultural census conducted by sample enumeration, requires a large proportion of the total effort, time and resources invested in a census programme.

The frame of a census conducted by complete enumeration

7.9 In order to conduct an agricultural census by complete enumeration of all holdings, it is necessary to estimate in advance the approximate location of holders' housing units and to assign to census enumerators well defined areas of work, usually determined with the help of different types of maps and eventually sketches if small-scale maps are not available. The total area of the country is unambiguously divided into identifiable areas in such a way that the enumerators' workloads are approximately equal. These non-overlapping subdivisions of the country are called enumeration areas (EAs). The accurate mapping of the EAs would ensure proper coverage and avoid omissions and duplications. In general, the EAs are defined and delineated in such a way so that the enumeration work in each area can be handled by a single enumerator during the census data collection period.

7.10 More precisely, the agricultural census EAs are geographic areas such that:

  1. They constitute a complete subdivision of the land, with no overlapping, covering all holders' housing units;
  2. The boundaries of an EA should not cross urban, rural or political subdivisions of the country; and preferably the EA should have recognizable permanent physical boundaries;
  3. An EA should be a compact piece of land, so that an enumerator can walk or travel between any two points of the EA without crossing its boundaries. In particular, a large river should never cross an EA; and
  4. Their area should correspond to approximately equivalent workloads weighting:
    • the approximate number of holders' households;
    • the distances and difficulties of access to the holders' dwellings;
    • the average time needed for each interview and the established length of the census enumeration period.

7.11 An agricultural census frame is a list of enumeration areas, with the estimated number of holdings in each enumeration area.

7.12 The EAs often are the smallest subdivisions of the country for which agricultural census data will be available.

7.13 In some countries, the census EAs are defined as subdivisions of relatively small administrative or political divisions (districts, for instance) if good maps are available. In other countries, a village is a well identifiable unit and village maps are available showing boundaries. Then, a village can be, with certain modifications, adopted as an enumeration area. And, in countries where no such maps exist, sketches are used. In other countries, the latest population and housing census enumeration districts can be grouped to form agricultural census EAs. This latter possibility is often used since, in most countries, for the population census organization and data collection, enumerator areas (that are different to those of the agricultural census) are defined. The population census EAs are defined as a function of the workload for enumerating the total number of households or housing units with a certain average time for completing a questionnaire, and a certain census enumeration period, factors which are generally different for those of an agricultural census.

7.14 The possible sources of available information for the construction of an agricultural census frame are the following:

  1. The statistical data and maps of the EAs of the latest agricultural census including a list of EAs with their number of holdings.
  2. The statistical data and maps of the EAs of the most recent population and housing census and a list of EAs along with their number of households/housing units;
  3. Listings with the addresses of holdings not directly associated with holders dwellings;
  4. Registers, surveys, cadastral and other cartographic materials, which include listings of holdings or holders' addresses or their approximate location;

7.15 In principle, the latest agricultural census (item (i) above) would be the most adequate pool of data and maps to construct an updated agricultural census frame. However, the latest agricultural census data and maps may not be available or, if available, they may be either entirely or partially useless because they are too old or incomplete to define the new agricultural census frame. In such cases, the most recent population census data and maps (item (ii) above) may provide the most adequate data and maps to construct a frame for an agricultural census.

7.16 The construction of an agricultural census frame must also cover the agricultural holdings that are not directly associated with holders' households, such as large plantations or cooperatives. If a list of holders is obtained by screening a list of households or housing units from the population census, it is quite possible that those holdings will be omitted from the list. A special effort is required, therefore, to compile a list of such large holdings from other sources in order to ensure their coverage (item (iii) above). Although the population census may provide very useful data, it should not constitute the sole source of information for the preparation of an agricultural census frame.

7.17 The listings and cartographic materials indicated in item (iv) above are generally useful in preparing or verifying only a small proportion of an agricultural census frame. In fact, such listings and maps are usually less complete than those available from the censuses of population or agriculture. To a large extent, therefore, the construction of anagricultural census frame is determined by the quality, timeliness and coverage of the data and maps of the latest agricultural and population censuses.

7.18 In a number of developing countries it might be necessary, due to lack of reliable information, to prepare the census frame by conducting a preparatory survey. And, it might even be necessary, due to lack of good maps, to prepare sketches for the EAs.

7.19 It is preferable that the EA be small in size, say less than 50 holdings, as is the case in some countries. In such cases, the holdings will be visible from one point or will be located along a road. Even if they are spread around, their small number should reduce errors in enumeration except in cases of extreme carelessness.

7.20 If large EAs are used, including in some countries as many as 500 holdings or more, spread over several square kilometres, upon arrival in the EA, the enumerator cannot start visiting the holdings without a pre-established order. Without a definite plan of enumeration there is a definite risk that some holdings may be enumerated twice while others may be omitted. For this reason, a plan of listing is necessary before the enumerator can start calling on the holdings. If a reasonably good cartographic map is available, the enumerator may start the listing from one corner of the area and proceed systematically, say clockwise until he/she completes the visits to all the households. In the case of urban areas, all the households are usually divided into blocks and these blocks are numbered on the maps with street names or numbers. The listing can be done by blocks, starting from a fixed point of the block. However, in rural areas the houses are not usually arranged in blocks, nor do well-defined streets with names and house numbers exist. In many countries, people of the same ethnic group or families closely related stay in the same compound. These compounds can be numbered and listing of the holdings can be completed by compound. In such listings, the help of the chief of the families in the compound may be obtained. In the list of the households, the names of the heads of households and other particulars to identify the holders living in the household, may be written. If more than one holder exists in a household, the names of all the holders should be written one below the other. It is very important that all households are visited and all holders listed serially. This will ensure enumeration of all holdings and complete coverage.

Construction of the frame using the population and housing census

7.21 National population and housing census data and maps can be used in various ways to construct a frame for the agricultural census. It is important to improve the coordination between the population and agricultural censuses which in most countries are the largest and most expensive statistical programmes.

7.22 The population census cartography, that generally determines the boundaries of the urban, suburban and rural zones, is used in particular to define the coverage of the agricultural census frame.

7.23 The use of the population census to prepare the frame of an agricultural census is particularly useful in countries where most of the agriculture is covered by the rural households.

7.24 It is often convenient to define the agricultural census EAs by grouping contiguous population census EAs within each political subdivision and urban or rural zones.

7.25 In rural areas, if the number of holders' housing units can be estimated from the total number of occupied housing units by EA, then suitable groups of contiguous population census EAs can be used to define the agricultural census EAs.

7.26 In urban areas where the proportion of holders' households is generally low, it may be useful to define provisional EAs by grouping contiguous population census EAs for the purpose of organizing the field screening that would permit the identification of holders' households and the definition of the final agricultural census EAs. It should be noted that the agricultural census covers only a small proportion of urban households.

7.27 The EAs prepared for carrying out the latest population census may not have always been corrected and adjusted on the basis of the information collected during the census. This is also often the case for the agricultural census EAs. In these cases the EAs should be updated if it is considered worthwhile.

7.28 Population census questionnaires, in most cases, do not allow a direct link to be established between agricultural holdings and households. However, the population census questionnaires generally identify the population whose main activity is agriculture within the context of a fairly short time-reference period.

7.29 If a population census is to be conducted shortly before an agricultural census, consideration should be given to the following aspects of the population census questionnaires:

  1. The inclusion of questions aimed at identifying agricultural holders' households, such as obtaining the main and secondary occupation of the population during the agricultural year, questions which allow distinguishing between work performed on the respondent's holding and work performed on other holdings.
  2. The inclusion of questions designed to provide information on basic characteristics of agricultural holdings.

7.30 Modifying the population census questionnaires for the purpose of undertaking an agricultural census shortly thereafter requires careful consideration. The disadvantages are the additional complexities and costs involved in several areas of the census programme, such as the training and selection of enumerators and the census field data collection. It should also be kept in mind that such changes would produce useful additional data for a small proportion of households - namely agricultural holders' households - while producing lower quality data for the majority of the population.

7.31 An alternative to conducting the agricultural census shortly after the population census would be to conduct both censuses simultaneously, as is done in some countries, particularly in Latin America and the Middle East. There are advantages to such an approach. The listing of households and agricultural holdings addresses can be done in one operation. The same mapping material and the same field staff can be used providing there is timely and adequate planning. Furthermore, coordination of actual collection of data of interest to both censuses, such as demographic data and data on employment in agriculture, is possible. This coordination may be achieved either in the data collection phase, or by matching related computer records in the data processing phase of both censuses. In practice, it appears that this joint approach may be applicable in countries where the same office, such as the Central Statistical Office, is responsible for both censuses. Otherwise, the coordination of the two operations becomes very difficult.

7.32 It is feasible to design the population census EAs, and in some cases to adapt the census questionnaires, to construct adequate frames for censuses and list frame sample surveys of agriculture. However, the specific concepts involved in an agricultural census are much different from those of a population census, and require enumerators with specific knowledge of agriculture. Also, as these two types of censuses are not always theresponsibility of the same national organization practical difficulties have arisen in attempting to apply the population and housing census to the construction of an agricultural census frame.

7.33 It is extremely important to coordinate the construction of the EAs of the population and agricultural censuses, or to elaborate them jointly, because this will save considerable resources. These savings are particularly significant for countries that can dedicate only very limited resources to national statistical programmes, and to countries where a high proportion of the population is in the agricultural/rural sector. In these cases the only feasible way of conducting an agricultural census by complete enumeration or based on a list sample design may be to closely link it to the population census.

7.34 It may be convenient in some cases to improve the boundaries of the latest agricultural census EAs, or improve the coverage of certain areas of the country by using the population census data and maps. In such cases, the population census is used indirectly to construct the agricultural census frame.

7.35 Population and housing census data, enumeration districts and maps often provide a highly important component in constructing the agricultural census frame or the list frames of a sample survey. In addition, in most cases the population census information is the best tool for detecting inconsistencies in and verifying the validity of agricultural census frames.

The frame of a census conducted by sample enumeration

7.36 In order to conduct a census by sample enumeration, as mentioned in Chapter 6, there are two basic types of sample designs concerning the final stage sampling units and their probabilities of selection, namely, list sample designs and area sample designs. The preparation of the frames needed for these two types of sample designs, as well as the preparation of the frames for multiple frame designs is discussed in the following paragraphs.

Preparation of the frames for list sample designs

7.37 List sampling frames of holdings or holders' addresses come from previous agricultural, housing or population censuses, from an accumulation of lists prepared by political or administrative subdivisions, farmer's associations, tax records or other sources. There is usually information on population and housing units, availability of markets, schools, drinking water, electricity, health assistance etc., and possibly information on holdings size, crops grown, and livestock for each holding that allows them to be classified into groups with similar characteristics (strata) which greatly improves sampling efficiency. Some of this information may be used for checking the internal consistency of the census results.

7.38 The frame for the last stage of selection is generally constructed from new listings obtained by direct field enumeration of holdings or holders' households.

7.39 For the implementation of list sample designs good maps are even more important than for complete enumeration censuses because there is a high risk of double counting or omitting from the frame some agricultural holdings. In a census conducted by complete enumeration, a holder interviewed twice will inform the second enumerator about this fact, so that corrective action can be taken. It is important to note that for the implementation of list sample designs detailed maps are needed only for areas included in the sample.

7.40 There are many different ways of preparing a list of agricultural holdings. Local knowledge should be used to list not only farm households but also holdings under different legal status, such as cooperatives, government farms, enterprises, etc.

7.41 In the case of countries or areas which maintain up-to-date land records (a cadastre), it may be easy to prepare a frame through reference to the land records. In the land records the name of the holder is normally entered against each of the fields operated by him. By looking into the land records, it may be possible to sort out the list of all the holders of a given area or village. Checking the internal consistency of the cadastre is essential. By covering each and every field of the given area, it should be possible to completely list the holders. The procedure has several shortcomings. Firstly, the holder staying in that area but operating elsewhere may not be listed. Secondly, those who do not operate land but are engaged in poultry or dairy farming may be omitted. Thirdly, the cadastre often records names of owners rather than names of holders.

7.42 In some countries where rural reconstruction and development programmes have been initiated, a list of households may be readily available and this can be used to screen the farm households. In case the land tax is paid by the holder, it will be necessary to maintain a list of holders which can serve as a useful frame to start with. In India, in some states a list of actual holders of land is maintained by the village officials for many development activities. Such a list may be a good starting point for the preparation of the list of holdings. Such a list, however, does not include those who do not cultivate any land but are engaged in livestock or poultry production.

7.43 In case no list of households or holdings is found, it will be necessary to prepare a new listing of households operating some land or keeping some livestock or doing both. The enumerator may be instructed to start from a fixed point of the village/enumeration block and systematically number serially each of the houses in the enumeration block. This will ensure that no house has been left out. In each house there may be more than one holding. Only some of the members of the household may be agricultural holders. The enumerator should proceed from house to house, listing households along with the information whether or not its members are holders. Although in the agricultural census the enumerator is concerned only with the list of the agricultural holders, it will be desirable to prepare the list of all members of the households for the sake of completeness and internal checking. Later, when the actual data relating to the agricultural census are to be collected the inquiry may be confined only to those household members eligible as agricultural holders. Sometimes, a holder may operate more than one holding. For example, the holder may operate some land with the help of the members of his own household and some other land jointly with the members of other households. According to the agricultural census concept, there will be two holdings in this case, one individual holding and the other a joint holding. In the preparation of the frame, such situations should be covered and there should be provisions for listing such holdings.

7.44 For an annual or seasonal agricultural survey, a list frame of holdings or holders addresses prepared for the census is often incomplete or inaccurate, contains an unknown amount of duplication, and is rapidly outdated. If the list is a few years old, many of the names will no longer represent a holding due to sales, deaths, and abandonment, and also new holdings are not represented. It is also common in list building to unintentionally include more than one name that could be associated with a specific holding. In short, it can be seen that it is difficult for a list of holding addresses to fulfil the requirements for an adequate periodic agricultural survey frame, particularly if it is a large list. To be truly effective, a large list must be continuously and systematically updated which is expensive, time-consuming and requires a large staff. When the list frame is inaccurate the samplecannot be assumed to be a probability sample and the survey should be considered a subjective survey.

7.45 As mentioned in Chapter 6, list sample censuses often include some strata of special holdings that are completely enumerated, or with a high sampling selection fraction. The strata of special holdings may consist of large holdings, holdings with the largest area for a given crop, with the largest livestock herds, highly specialized holdings or those corresponding to a very local area of production. Such lists of special holdings are fairly easy to update since the holdings involved are usually visible and well known. Information can be obtained from extension agents, producers' associations, banks, tax records, agricultural censuses and from government agencies that either control and/or purchase production of certain crops. The preparation of such lists should include the accumulation of data on each operation such as holding size, crops grown, type of livestock and inventory, etc. for stratification purposes if the list is to be sampled.

7.46 The use of the census EAs from a population census or the use of the census EAs from the latest agricultural census as Primary Sampling Units (PSUs) is a usual form of cluster sampling for list sample agricultural surveys. It should be noted that although the PSUs are geographic areas, the associated probabilities for the sample selection are not proportional to their areas: they are usually proportional to the number of holdings. All holdings are listed within selected EAs and a sample of holdings (as represented by holders) is chosen in the second and final stage. Data collection usually consists of accompanying the holder to his/her holding, measuring the fields and gathering whatever other data is needed to complete the survey questionnaire.

7.47 The average size of PSUs varies, in general, from 50 to over 200 holdings, of which 4 to 10 (at times even 30) are selected for the final sample.

7.48 Some of the problems involved in the use of list samples are the following:

  1. In comparison with other sampling designs, a larger sample may be necessary due to the between cluster and within cluster variances.
  2. It is often not easy (sometimes impossible) to establish the boundaries of the PSUs whether they are villages, enumeration areas or administrative subdivisions.

Preparation of the frames for area sample designs

7.49 As defined in Chapter 6, an area sample design is a sample design in which the final stage sampling units are land areas called segments, and the selection probabilities are proportional to their area measures (size). Most area sample designs of agricultural censuses consist of a stratified probability sample of segments. The segments in each stratum are considered to be of equal size and selected with equal probability, and the estimation formulas used correspond to a single-stage design. The strata are defined by intensity of cultivated land, predominance of certain crops, special agricultural practices, average size of cultivated fields, agro-urban areas, or other land-use characteristics.

7.50 Two types of segments have been used for the area sample design of agricultural censuses in developing countries:

  1. Segments that have recognizable physical boundaries.
  2. Segments that coincide with the land of agricultural holdings.

Preparation of the frames for area sample designs with segments that have recognizable physical boundaries

7.51 An area sampling frame is an ordered list of land areas, called frame units, with their assigned number of segments, such that:

  1. The frame units form a complete subdivision of the total land area of each land-use stratum, with no overlap.
  2. The frame units provide a clear-cut means of identifying each segment.
  3. The number of segments assigned to each frame unit facilitates the probability sampling of segments.

7.52 A frame unit is an area of land with readily visible physical terrain features that contains a number of segments. All the land in an area frame is divided into frame units, without any overlap or omission. They are units that provide a means of identifying and counting segments without actually drawing off each segment. For identifying a selected segment it is only necessary to partition the corresponding frame unit into a number of segments equal to its assigned number of segments.

7.53 The frame units are subdivisions of the strata and preferably also administrative sub-divisions of a province or even sub-divisions or aggregation of contiguous census EAs if they were defined with recognizable physical boundaries. This would allow for additional agricultural or demographic information to be used.

7.54 For the area sample designs described in this section, each stratum has a different target segment size from the other strata. Small segments usually correspond to strata of intensely cultivated land or to segments in the urban or agro-urban stratum.

7.55 The target segment size in a given stratum is generally set so that it is equal, on the average, to a day's work of an enumerator. An acceptable segment size in a stratum of intensely cultivated land is one quarter of a square kilometre (25 hectares). Segments may have 10 to 15 tracts.

7.56 Assuming that an area frame has been constructed for a province, an area sample size may be 400 segments. The target size of frame units (say 4 to 10 segments) is set based on the target segment size for the stratum (say 25 hectares) and the availability of good boundaries.

7.57 The preparation of an area sampling frame, in order to select a sample of segments that have recognizable physical boundaries is a very demanding task. To prepare an area frame, the first requirement is to have up-to-date cartographic material (maps, satellite images, aerial photos) on which the land to be included can be visualized. The resolution or detail of the material must be sufficient to allow stratification according to intensity of land use (proportion of land cultivated, predominance of certain crops or other uses of land, etc.) and the subsequent subdivision of these land-use strata into frame units also with recognizable physical boundaries. Land-use strata and frame units are identified on satellite images or a mosaic of aerial photography and then transferred to topographic charts and measured. There might be from 6 to 10 land-use strata in a province. Frame units are constructed generally with maps on which the boundaries of the land-use strata have been transferred. In each land-use stratum, each frame unit will have to be measured and assigned a target number of segments of equal size. The number of segments assigned to each frame unit are summed to provide the total number of segments in the stratum. Then a sample of segments is selected from each land-use stratum using a systematic or random replicated selection procedure, all segments having the same probability ofselection. Each sample segment is constructed on small mosaics of aerial photography on which the boundary of the corresponding frame unit have been transferred. The selected sample segments are located on appropriate aerial photo enlargements used to control field data collection.

Preparation of the frames for area sample designs with segments that coincide with the land of agricultural holdings

7.58 In this case, a grid is overlaid on the strata and a sample of points is selected. Then, the points are identified on the ground and the corresponding holdings form the area sample. Thus, the design can be considered a stratified sample of holdings selected with probability proportional to their areas.

Preparation of the frames for multiple frame sample designs

7.59 For an agricultural census based on a multiple frame design, any duplication (overlap) of list frame elements in the area frame must be removed, an operation that requires special attention and resources.

7.60 The list frame component of a multiple frame survey can be a large, nation-wide list of holdings requiring a heavy investment in computer hardware and software a very controlled field operation for its periodic updating (if the frame is used periodically), and involving great difficulties for its combined use with the area sample. This type of multiple frame surveys, although the most accurate, is generally not adequate for developing countries.

7.61 The case of a relatively short complementary list frame, enumerated completely, used with an area sample is the multiple frame survey design described in the FAO manual on multiple frame agricultural surveys since these are generally the most adequate multiple-frame methods applicable in developing countries.

Geographic Information Systems (GIS)

7.62 Geographic Information Systems (GIS) can be defined as a set of application software, working on specific hardware, for collecting, storing, retrieving and displaying spatial data. GIS are used in many countries for a variety of purposes related to agricultural statistics. A GIS called CASS (Computer-Aided Stratification and Sampling) is currently used in the United States for area frame construction and sample selection. These procedures are highly automated and the material needs, mainly maps, are difficult to meet in most countries. The methods are important because they are adopted at present in the largest and oldest area frame project in the world and because they may represent the trend of the evolution of area frame construction and sample selection from manual to computerized procedures. In other words, these computerized methods may be partially or totally adapted to different conditions and requirements in other countries.

Suggested reading
Cotter, J. and Nealon, J. (1987). Area Frame Design for Agricultural Surveys. National Agricultural Statistics Service, USDA, Washington.
FAO (1995). Multiple Frame Agricultural Surveys-Agricultural Surveys based on Area and List Sampling Methods. FAO Statistical Development Series No. 7, Rome.
Houseman, E.E. (1975). Area Frame Sampling in Agriculture. Statistical Reporting Service, SRS No. 20, USDA, Washington.
UN (1982). Non-sampling errors in household surveys: Sources, assessment and control. NHSCP technical study.
UN (1992). Handbook of population and housing censuses: Part I, Planning, organization and administration of population and housing censuses. Studies in methods, Ser.F, No.54.