The purpose of this chapter is to present an overview of the main survey designs used for conducting agricultural censuses and surveys. 

After presenting the main types of survey design, list sample design and area sample design, other important uses of sampling techniques in the statistical work, are also mentioned, such as the application of sampling in quality control of the field and data processing operations, pre-testing of questionnaires, and various statistical operations. The construction of frames is discussed in Chapter 7. 

The statistical theory and practice of survey design for agricultural surveys is beyond the scope of this publication, and the reader may consult the list of suggested references on survey methods for agricultural censuses and surveys included at the end of the chapter. 

Objectives of agricultural censuses and surveys

6.1 In this publication, an agricultural census is defined as a large-scale, periodic, statistical operation for the collection of quantitative information on the structure of the agriculture of the country. As mentioned in the FAO Programme for the World Census of Agriculture 2000, the basic objectives of an agricultural census are:

1. To provide aggregate totals for fundamental agricultural data for use as the benchmark for inter-censal estimates.
2. To provide a frame for other agricultural sample surveys.
3. To provide data for small administrative units and detailed cross-classifications of holding structural attributes.

6.2 FAO recommends that each country should conduct at least one agricultural census in each census programme decade. The FAO Programme for the World Census of Agriculture 2000, for instance, corresponds to agricultural censuses to be undertaken during the decade 1996-2005.

6.3 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. Census data are also useful in the design of annual or seasonal agricultural sample surveys. For example, stratification criteria for current surveys are included in the census questionnaire: sex and age of holder, total area, area by crops, number of livestock, number of trees, integration of holding with enterprises, etc. Clustering procedures and assignment of probabilities of selection for current agricultural surveys are also often based on agricultural census data.

6.4 Besides the census, a well-designed national agricultural statistical system includes surveys aimed at collecting more specific and timely information, for example, data on crop yields, prices, and production that is not feasible to collect through an agricultural census.

Sample design and survey design

6.5 The sample design of a survey or an agricultural census conducted by sample enumeration refers to the techniques for selecting a probability sample and the methods to obtain the estimates of the desired characteristics from the selected sample.

6.6 The survey design refers to the definitions and established methods and procedures concerning all phases needed for conducting the operation: the sample design, the selection and training of personnel, the organization of the logistics involved in the distribution and receipt of questionnaires, the data collection and data processing procedures, and the analysis of data needed for the release of the final results.

6.7 An agricultural census may have a number of different survey designs. For instance, it can be carried out using a number of different enumeration procedures (e.g., by personal field interview, by expert observation, self-administered questionnaires, by mail or, perhaps, by telephone), by complete or sample enumeration, different types of sampling units, sampling selection methods and estimation procedures, by including or not the identification and objective measurement of agricultural areas or other agricultural commodities in the selected enumeration units, etc.

6.8 In agricultural censuses, most data are obtained by enumerators through personal interviews with the holders by using a questionnaire for each holding. Indeed, in developing countries it is not generally feasible to use, on a large-scale, other data collection procedures, such as self-administered questionnaires, mail, or telephone.

Complete enumeration censuses versus censuses carried out on a sample basis

6.9 Agricultural censuses can be classified into two categories: censuses conducted by complete enumeration of all holdings or by a sample enumeration. The main characteristics of these two categories are the following:

6.9.1 Complete enumeration.
For agricultural censuses conducted by complete enumeration (which are called simply agricultural censuses in many countries), the enumerators complete a questionnaire for each holding. The result for each characteristic is obtained from the values of the characteristics in all holdings. Therefore, the census results include only non-sampling errors.

6.9.2 Sample enumeration.
Agricultural censuses based on sample enumeration are probability sample surveys, that is, surveys for which a probability sample (of sampling units) is selected, and for which the methods of estimation for each census characteristic allows establishing their statistical precision. This requires, in particular, defining the sampling units and their probability of selection, from a known universe (frame).

Advantages of carrying out a census on the basis of complete enumeration

6.10 The census results can be obtained for small administrative and other area units. Such information is sometimes required by law, for local planning, or for practical purposes such as irrigation projects or projects on regionalization or on an agro-climatic and ecological basis. (Sampling methods cannot provide accurate information for small administrative sub-divisions.)

6.11 Some crops, although cultivated only to a limited extent, may be of great economic importance. Information on such crops can only be reliably obtained from a complete census. The situation is similar with regard to rare species of animals, or any other rare variable. (Sampling methods cannot provide accurate information on variables that occur infrequently.)

6.12 Current agricultural statistics in most of the countries have to be collected through annual sample surveys. These sample surveys can be planned much more efficiently if census results are available for small area units. First, the census listings can be used as a frame for the selection of the sample. This may prove an important saving because the preparation of the frame itself generally consumes a significant part of the budget allocation for the sample survey. Furthermore, census data can be used for improved sampling design and better estimation procedures, which may lead to more reliable estimates of the characteristics under study.

6.13 The planning and implementation of an agricultural census conducted on a complete enumeration basis requires fewer highly qualified statistical personnel than a census conducted on a sample basis. This constitutes an important advantage in countries with limited technical expertise.

6.14 Processing data from a complete enumeration is straightforward and does not involve the calculation of sampling errors or expansion factors and therefore requires less skills. (Processing the data from a sample enumeration is technically more complicated.)

Disadvantages of carrying out a census on the basis of complete enumeration

6.15 In a country with a large number of holdings, to conduct a census by complete enumeration is, in practice, more expensive and time consuming than a census conducted on a sample basis. This is a particularly important consideration in areas with poor communication and transport facilities. (Sample enumerations are generally considerably less costly that a complete enumeration.)

6.16 For a complete enumeration census, a very large number of enumerators and supervisors are required. Quite often candidates with the desired qualifications are not available in the required number so the standard has to be lowered with a consequent effect on the quality of data. (A census conducted on a sample enumeration basis requires a smaller number of enumerators and supervisors than a census conducted by complete enumeration. Consequently, the quality of data collected can be expected to be much better because of the employment of better trained enumerators and supervisors and better quality control.)

6.17 The quantity of data to be processed is very large for a complete census. The results may be considerably delayed if insufficient data processing equipment is available. The cost of processing the data will be higher due to the large volume to be processed. (Sample enumerations require less processing capacity (fewer forms to process) and the results are usually available considerably faster than for a complete enumeration census.)

Factors for consideration in choosing between a complete or a sample census

6.18 The sample enumeration is a very attractive proposition where there is a severe limitation of funds and personnel and the aim is confined to securing data with reasonable accuracy for major administrative units and the country as a whole.

6.19 A decision whether to carry out a complete enumeration or to plan a sample survey of holdings will depend on the level at which the results are required, that is whether the results will be tabulated for the entire country, for individual provinces, for individual districts or even for smaller administrative sub-divisions. In practice, this decision is either a complete enumeration or a sample enumeration but could be a combination of both.

6.20 Even those countries which lack resources in terms of funds and trained personnel should seriously consider the possibility of undertaking at least a minimum part of the census on a complete enumeration basis. This is to ensure a good base for preparing an efficient sampling design for the collection of detailed data on important items of the census, for planning future agricultural surveys to collect current agricultural statistics and to be able to produce at least some data for small administrative units. Many countries adopt a phased approach to their agricultural census by conducting an initial listing exercise as the first stage to obtain a complete frame of all agricultural holdings in the country and collect some limited information on key variables. The second stage of the census is a sample enumeration to collect the required detailed information.

6.21 In practice, control of non-sampling errors, including those of non-response, may be possible only on the basis of a sample. Even though a census may have been planned on the basis of a complete count, sampling techniques will have to be used for controlling the census operations, the response errors and those of data processing. However, in order to carry out these tasks effectively there is a need for highly-qualified persons to design and execute the sampling plans.

Main types of sample designs for agricultural censuses

6.22 For agricultural censuses based on a sample enumeration, there are two basic types of sampling methods commonly used concerning the final stage sampling unit and their probability of selection, namely list sample methods and area sample methods. In addition, multiple frame sample methods are also used, these being survey designs that combine an area sample with a list sample to obtain the census estimates.

List sample designs

6.23 List sample designs are the most commonly used sampling procedures. In this case, the last-stage sampling units are generally the holdings or the holders' addresses.

6.24 List sample censuses often include some strata of special holdings that are completely enumerated, or have a high sampling-selection fraction. Such strata consist of holdings that either correspond to a significant proportion of the total estimated value of important census characteristics, or, if selected in the sample, whose characteristics may distort the results. For example, 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 localized production.

6.25 Cluster sampling is commonly used to account for geographic contiguity in the first stages of the sampling method. A sample of holders can be selected indirectly by first selecting a sample of villages (Primary Sampling Units-PSU's) with probability proportional to their total population (or housing units) since such information is usually available in most countries and approximates to the number of holders. Hopefully some additional information about the villages, such as farm population and primary agricultural activity, will also be available for at least rudimentary stratification. Other small administrative subdivisions, such as districts, counties or other administrative divisions could be used as well as villages as PSU's. The total population of each selected village would then be screened for agricultural holders and a sample of holders selected in a second stage of sampling.

Data collection procedures for list sample designs

6.26 During data collection, the enumerator usually completes a questionnaire for each selected holding by conducting an interview with the holder. In addition, in some cases the enumerator measures the fields and gathers whatever other data is needed to complete the census questionnaire.

Area sample designs

6.27 An area sample design is a probability sample method in which the final stage sampling units are land areas called segments, and the selection probabilities are proportional to their area measurement (size). Therefore, the entire survey area is considered to be completely subdivided into non-overlapping segments that are the sampling units.

6.28 Most area sample survey designs for agricultural censuses consist of a stratified probability sample of segments. The strata are defined by intensity of cultivated land, predominance of certain crops or other land-use characteristics.

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

6.29.1 Segments that have recognizable physical boundaries.
The boundaries of a segment are physical terrain features, such as roads, rivers, canals, railroads, etc., that are readily found and provide an unambiguous identification of the segment. In this case, for each stratum, the segments are defined with approximately equal size (area) and a constant sampling expansion factor is used to obtain the estimates derived from data in the segments. The sample usually consists of a number of selected independent sample replicates in each stratum, that facilitates the rotation of the sample to reduce respondent burden if the sample is to be used periodically over time. The sample design can be considered a stratified cluster sample of tracts, which are the parts of holdings (or non-agricultural areas) included in the segments.

6.29.2 Segments that coincide with the land of the agricultural holdings (point sampling).

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 (size).

Data collection procedures for area sample designs

6.30 Area sample designs with segments that have recognizable physical boundaries generally involve an annual (or seasonal) field data collection carried out by enumerators who complete a questionnaire for each tract included in each selected sample segment but can also be used in taking censuses. The data collection may involve objective measurement of agricultural areas utilizing aerial photo enlargements (or maps or scale drawings), called segment photos. For each tract of a given sample segment the enumerators delineate on the segment photo the boundaries of the tract included inside the segment and the boundaries of all fields included in the tract, and verify the crops planted and other uses of land for each field, and also the information provided by the holder. Such identified agricultural areas in each sample segment are later measured in the office using a planimeter or a computer graphic system providing the basis, through sample expansion procedures, for an objective estimation of agricultural areas.

6.31 For the objective measurement of agricultural areas, the measurement on aerial photos is a very important advantage particularly if the interviewed holder does not know or does not want to report the area of land operated.

6.32 Area sample designs with segments that are the land of the holdings are generally enumerated by enumerators who complete a questionnaire for each selected holding. The data collection may not involve objective measurement of agricultural areas. If objective measurement of agricultural areas is required, this may be accomplished with list samples, namely by measuring the holdings and fields on the ground, without using aerial photos for measurement.

Estimation procedures in area sample designs with segments that have recognizable physical boundaries

6.33 Since sampling units are land areas (segments) that may not coincide with the land of a holding, it is necessary to establish a criterion to associate each segment with a holding so that a census characteristic value can be assigned to each segment as a function of its value in the associated holding. Then, a uniform expansion factor is applied in each stratum to obtain the census estimate for the characteristic.

6.34 For agricultural censuses based on an area sample with segments that have recognizable physical boundaries, two types of direct expansion methods have been used to associate segments with holdings and to expand the results obtained in the holdings associated with the selected sample of segments.

6.35 The direct expansion estimations refer to the way the holding data are distributed among the segments before they are expanded by the inverse of the sampling fraction. The direct expansion methods used are called the weighted segment method and the open segment method.

6.36 Ratio-estimation methods can also be used to improve the direct-expansion census estimates. However, except for direct-expansion methods, no other type of estimation procedures are covered in this outline.

The weighted segment estimation method

6.37 Let us consider a segment in a given stratum. The direct weighted segment estimation method involves the following steps:

1. The weighted segment method uses all holdings with a tract in the segment, by associating to the segment all holdings with any land inside the segment.
2. The value of a given census characteristic in each tract (e.g., the number of cows) is defined as the value of the characteristic in the holding multiplied by a factor equal to the ratio between the area of the tract divided by the area of the holding. For instance, if 10% of a holding's area is in the tract, 10% of the total number of cows of the holding are assigned to the tract.
3. The value of the characteristic in the segment is defined as the sum of the characteristic in each of its tracts, as defined above.
4. The census estimate of the characteristic in the stratum is the sum of the characteristic in all its segments multiplied by the expansion factor of the stratum.
5. Finally, the census estimate of the characteristic is the sum of the characteristic in all strata.

6.38 The weighted-segment estimator does not require establishing precisely who is the holder, the location of his residence or the holding headquarters, but it is obviously necessary to identify the holding, calculate its area inside the selected segments, and find an informant who can provide data for the total holding. The use of the weighted segment estimator is the recommended estimator for censuses based on area sample designs.

The open segment estimation method

6.39 The direct open segment estimation method involves the following steps:

1. The open-segment method associates to the segment all holdings with headquarters inside the segment. For this purpose, clear rules have to be established to define a unique reference point for each holding, called the headquarters. There are several ways to do this but the most common procedure is to define the headquarters as the dwelling (residence) of the holder. In this case, a one-to-one correspondence between holdings and headquarters can be established, with an additive rule in case of joint holders living in different dwellings.
2. The value of the characteristic in the segment is defined as the sum of the characteristic in each of its associated holdings, i.e., the holdings with headquarters inside the segment.
3. The census estimate of the characteristic in the stratum is the sum of the characteristic in all its segments multiplied by the expansion factor of the stratum.
4. The census estimate of the characteristic is the sum of the characteristic in all strata.

6.40 Open segment estimators are not generally recommended due to the difficulty that arises in identifying (creating a one-to-one correspondence between holdings and headquarters) and locating the residence of the holders in urban areas.

Multiple frame sample designs

6.41 A survey design that combines an area sample design with a list sample design, is called a multiple frame sample design. Multiple frame estimates combine area sample with list sample estimates for each census characteristic. The results obtained from the list sample are expanded and added to the area sample estimate.

6.42 The list sample component of a multiple frame design is called the list of special holdings, and is usually formed by those holdings with the largest total area and the largest area for a given crop, those with the largest number of livestock and poultry, those with the largest revenues, with the largest number of agricultural workers, highly specialized holdings and those corresponding to a localized production. The list of special holdings should all be enumerated if possible. In this case, the results obtained from the list are added to the area sample estimate with no contribution to the overall variance.

6.43 A list sample is a necessary addition to an area sample in order to provide adequate estimates for important agricultural characteristics. In fact, as already mentioned for list samples, many important agricultural characteristics have a skew distribution, concentrating a significant proportion of the total estimate in a small proportion of the holdings. For some of those characteristics, the list sample component (the list of special holdings) should account for the skewness of their distribution so that the multiple frame estimates will be significantly more precise than the area sample estimates.

6.44 For multiple frame designs, any duplication between the list of special holdings and holdings partially or totally included in the selected area sample segments must be eliminated from the selected segments. This operation of removing duplications of holdings requires special attention and resources. For this reason, it is particularly important in developing countries to utilize a relatively short list of special holdings, that would be feasible to inspect.

Choice of sample design

6.45 The statistical survey method on which to base an agricultural census or survey in a given country should carefully consider the local conditions, resources and requirements. The choice of survey design should be made taking into account the trained personnel, the resources available and the desired accuracy of estimates of the principal characteristics. The sample design should also be simple enough to operate in the field with the help of available personnel. Experience indicates that it is difficult to make adjustments for any significant deviations occurring from the sample design (missing questionnaires, errors in selection of the sample, errors in the frame, etc.) if the sample design is not simple. Also, the size of the sample has to be fixed at an adequate level so that it can be enumerated within the time limit prescribed for the statistical operations. The total cost and requirements of personnel (the number and the period), the construction of the frames and the other required facilities should be clearly assessed and definite government approval obtained for incurring the expenditure. In case funds and other resources required are beyond the capacity of the country, the sample designs have to be adjusted keeping in view the resources actually available and the authorities have to be informed of the type of results that will be achieved with this change in the sample design. It may also be said that resources promised originally are not always made available in full and at the time required. This could lead to failure of the sampling design unless this factor is kept in mind while planning the sample design. It is perhaps advisable to plan on the basis of resources which are somewhat less than promised. Alternatively, a sampling plan which can easily be adjusted according to the actual resources available at the time of carrying out this phase of the statistical operation should be developed. This really requires a great deal of ingenuity on the part of the statistician, a periodic review of the resources and an intimate knowledge of how the census operations are proceeding. Of course, once the original design has been carefully chosen, it should not be abandoned or modified except for serious reasons.

6.46 The sample size in a given situation is difficult to determine. The appropriate sample size depends upon many factors, such as the efficiency of the sampling design adopted and the number of areas and classifications for which estimates are required. If the data are needed for planning at administrative sub-division level, a much larger sample will be needed; a sampling fraction of 2 to 10 percent of the holdings may be adequate for providing reliable estimates for all the principal characteristics. However, if such information is needed at only the broad provincial level or national, a sampling fraction of 1 to 2 percent of holdings or agricultural areas may do. As a rule of thumb, no attempt should be made to make sampling estimates for areas or groups of holdings for which less than 200 to 300 sample questionnaires have been completed. A critical examination of the sample size should be made at the time of the pilot census. One of the objectives of the pilot census should be to study the variability of different characteristics and the time and cost involved in obtaining information on them.

6.47 For an agricultural census to be conducted by sample enumeration, the sample design should be chosen by considering the characteristics of multiple-frame methods and list sample methods along with their comparative advantages, disadvantages and requirements.

6.48 Comparisons of the different types of sample survey designs require special statistical knowledge that is beyond the scope of this publication; the reader could usefully refer to the publication Multiple frame Agricultural Surveys- Agricultural Surveys based on Area and List Sampling Methods, FAO (1995). Nevertheless, a few simple indications will be given in the sequel to illustrate the factors to be considered when choosing an appropriate sample design for an agricultural census. For such purpose, a few definitions and specifications are given below.

6.49 A sample frame will denote here, as is usual in survey sampling texts, the total set of sampling units, that is, the set of units from which the sample is selected. Accordingly, a list frame is a list of all holdings and an area frame a list of all segments (land areas) of the country (see Chapter 7).

Multiple frame designs versus area sample designs

6.50 Multiple frame designs that combine an area sample with (at least) a short list of special holdings to be completely enumerated during field data collection are preferable to area sample designs since they can provide more accurate estimates of important census items (characteristics) and because the extra work involved for its design and implementation will generally not be significant.

Multiple frame designs versus list sample designs

6.51 When further referring to an area sample it will be assumed to be the area sample component of a multiple frame design that also includes a list frame of special holdings.

6.52 For an agricultural census to be conducted by sample enumeration, the following preliminary considerations can be used to compare the different types of sample designs:

6.52.1 Unbiased estimates.
An area sample can generate unbiased estimates since it is based on a sample from a frame that provides complete coverage of the area of interest. This cannot be said for a census based on a list sample since, in practice, a complete and perfectly updated list of holdings, valid during the data collection period, cannot be established: list frames of holdings are often incomplete and outdated. Coverage errors are a major problem in list sampling, but not in area sampling provided the rules of association linking holdings with selected segments are performed correctly.

6.52.2 Precision of the estimates.
The multiple frame method, which combines an area sample of segments with recognizable physical boundaries, obtains more precise estimates of agricultural areas, a key variable studied in all agricultural censuses, than a list sample. In fact, by definition, in area sampling the probabilities of selection and the sampling expansion factors utilized are precisely proportional to agricultural areas. This is true for an area sample design of the type considered in which the main crop areas in the selected segments are identified and delineated on aerial photographs or detailed maps during the field data collection and then measured in the office. In other words, a multiple frame design involves objective measurements of agricultural areas, which is an important advantage for providing precise area estimates.

6.52.3Non-sampling errors and objective measurement of areas.
In area sample designs in which segments have recognizable physical boundaries, non-sampling errors associated with area measurements are reduced by using aerial photographs of the selected segments that clearly indicate the holdings and the fields. The photographs are used to check reported area of fields and total area of the holding itself. The holder is more inclined to be truthful when confronted with questions about specific portions of his holding that are also being observed by the enumerator at that moment. Such area sample designs provide more precise estimates than list sample designs. Many list sample methods do not require measurement of areas but rely on respondent information. For list sample designs that require measurement of areas, the measurement is usually done only in a subsample of holdings during the field data collection. These procedures are generally very slow and somewhat cumbersome to apply, and are impractical for measuring holdings formed by parcels located a long distance from one another. It should be mentioned that in many countries, or in large areas of developing countries, the areas reported by holders are not considered reliable and therefore objective measurements of areas are required to obtain reliable estimates. This need arises partly on account of all kinds of arbitrary local units of measurements in use in different parts of the same country, and partly because of the general tendency among holders to under report their areas and production.

6.52.4 Unbiased annual estimates.
If an annual agricultural survey is to be implemented using the census sample design, it is worth noting that the area frame is generally far more durable than a list frame of holdings. An area frame can be used over a period of years (say 5-10) without updating the sampling units in areas where agriculture is stable. An area frame does not become outdated unless the population extends into areas not covered by the frame. Changes in land use, or in the number and location of holdings, may reduce the precision of the estimates but they do not introduce bias.

6.52.5 Basis for a crop cutting yield survey.
An area sample with segments that have recognizable physical boundaries provides the means for selecting a statistical sample of fields needed to conduct crop cutting yield surveys and therefore for estimating crop production when crops mature and forecasting crop yield estimates by measuring plant characteristics at certain stages of growth during the crop year. In other words, an area sample provides the means to better estimate crop yields.

In developing countries, holders are often not able to report reliable estimates of crop yield and production and there is the problem of local measures which often vary from village to village or even from holder to holder. Crop-cutting techniques are recommended to collect objective estimates of yield. Crop-cutting methods for estimation of crop yield have been adopted by many developing countries as a standard technique. From the sampled fields a standardized plot, usually in the form of a square or a circle, is taken for measuring the yield of a given crop. In the case of rice and wheat, a plot of 1 to 5 sq.m. could be sufficient. For maize and tubers 10 to 25 sq.m. plots are preferred while for widely spaced crops 100 sq.m. may be more adequate. For obtaining reliable estimates of yield at district or administrative sub-division level, at least 100 to 200 crop cuttings should be made. The crop-cutting method is very time consuming and expensive. Furthermore, it is subject to various biases, such as border bias as yield is different at the borders than inside the field. Other biases arise because of inadequate training of enumerators or because of inadequate supervision. Countries should, therefore, find which method will best suit their conditions and also consider organizing special studies for evaluating local measuring units in order to improve production estimates provided by holders.

6.52.6 Complexity of implementation.
The implementation of an area sample design with segments that have recognizable physical boundaries requires more technical expertise than the implementation of a list sample design.

6.52.7 Cartographic requirements.
The selection of an area sample requires accurate cartography on which to identify and measure areas. It requires the availability of suitable topographic charts, and preferably satellite images, as well as scale-transfer and area measurement instruments. Aerial photos of the selected segments are a great advantage if objective measurement of areas is required.

6.52.8 Proximity of the holder or respondent to the holding.
It may not be feasible or even possible to use an area sample in some countries due to difficult terrain or due to certain social mores of the rural population. Area sample methods should not be used if, for instance, the information obtained from holders who do not live close to their holdings or who are difficult to locate corresponds to a large percentage of the total value of important survey variables.

6.52.9 Costs.
There are high costs involved in the selection of an area sample of segments with recognizable physical boundaries, and these costs may be higher than those needed for the selection of a list sample. However, such a high investment may easily be justified if the samples are to be used on a regular survey basis.

6.52.10 Lack of permanent boundaries.
For an area sample design with segments that have recognizable physical boundaries, the lack of permanent boundaries in the maps, satellite images and aerial photos constitutes a serious problem. In tropical areas, such as West Africa, because of the climatic conditions and shifting cultivation systems, boundaries change more frequently or get covered by bush and are not visible on the cartographic materials.

6.52.11 Distinguishing characteristics.
A distinguishing characteristic of multiple frame sample designs is that they have incorporated important technological advances in computer data processing to a larger extent than list sample methods. In fact, area sample methods can utilize satellite imagery or even digital satellite data as part of Geographic Information Systems, hand-held Geographic Positioning Systems and generally a variety of automated procedures and techniques for sample selection and data analysis.

Other uses of sampling techniques

6.53 Besides conducting field data collection on a sample basis, there are many other ways and reasons to use sampling techniques for completing the work of an agricultural census. Some of the most important and indispensable applications of sampling techniques for an agricultural census are listed below:

1. Arranging collection of data in two parts: a simple questionnaire by a less qualified set of enumerators on a complete enumeration basis, and a more complex questionnaire by enumerators of higher qualifications on a sampling basis.
2. Combining a complete enumeration of large holdings with a sample enumeration of small holdings.
3. Checking the completeness of a frame of holdings or households especially when the frame available is rather old (see Chapter 7).
4. Pre-test surveys and pilot censuses for checking the questionnaires and various census procedures (see Chapter 13).
5. Arranging supervision of field work on a rational basis with a view to providing a measure of the quality of data collected, and to provide for a correction factor wherever possible.
6. Arranging the field work in a set of replicate samples such that each is capable of providing valid estimates of characteristics under study so that these estimates observed together provide a measure of the reliability of the census data.
7. Designing supplementary sample surveys covering special subjects, such as crop-yield objective measurements, livestock and poultry in urban areas not covered by the census, smallholdings below size covered by the census.
8. Post-enumeration surveys on the completeness of data collected and evaluation of response errors (see Chapter 16).
9. Rapid preparation of some preliminary results of the census.
10. Quality control of errors in the data processing: coding, verification of data entered, etc.
11. Final tabulation of the data on the basis of a sample as an emergency solution, in situations where resources are not sufficient for carrying out the analysis of all the data collected.

6.54 The approach where sampling is used together with complete enumeration for increasing the scope of the census by augmenting, with the help of a sample, the information collected through complete enumeration, may be particularly suitable (see ii) above). This means that for basic items in the agricultural census, complete enumeration of all holdings is undertaken, but for various additional items only a sample of holdings is enumerated. These additional items are usually of a more complicated character and involve careful questioning of the respondents. A more qualified type of enumerator is required and can be entrusted with this work if it is limited to a sample of holdings. Normally, a single sample would be selected for whatever additional information is sought, but it may also be distributed over different samples, a limited amount of information being collected from each sample, so that all samples put together provide the total additional information proposed to be secured by sampling without placing an unduly heavy burden on the respondents. This operation of securing additional information from a sample can either be done simultaneously with complete enumeration or be staggered so that supervisory staff can be employed for enumerating the sample, after they become free from the supervision of the main census.

Suggested reading
Ardilly, P. (1994). Les techniques de sondages. Technip.
Cochran, W.G. (1977). Sampling Techniques. Third edition, John Wiley and Sons.
Cotter, J. and Nealon, J. (1987). Area Frame Design for Agricultural Surveys. National Agricultural Statistics Service, USDA, Washington.
Desabie, J. (1966). Theorie et pratiques des sondages, Dunod.
Dubois, J.L. et Blaizeau, D. (1989). Connaître les conditions de vie des ménages dans les pays en développement. Ministere de la coopération et du développement, France (3 volumes).
FAO (1989)-Kish, L. Sampling Methods for Agricultural Surveys. FAO Statistical Development Series No. 3, Rome.
FAO (1995). Multiple Frame Agricultural Surveys-Agricultural Surveys based on Area and List Sampling Methods. FAO Statistical Development Series No. 7, Rome.
Hansen, M. H., Hurwitz, W.N. and Madow, W.G (1953). Sample Survey Methods and Theory. Vols. I-II, New York: John Wiley and Sons.
Houseman, E.E. (1975). Area Frame Sampling in Agriculture . Statistical Reporting Service, SRS No. 20, USDA, Washington.
Kish, L. (1965). Survey Sampling. John Wiley and Sons.
Vogel, F.A. (1986). Sample Design and Estimation for Agricultural Sample Surveys. Statistical Reporting Service, NASS/USDA, Washington.