|Multiple frame agricultural surveys. v. 1: Current surveys based on area and list sampling methods. (FAO Statistical Development Series - 7) (1996)|
With the global shift towards market economies, the need for timely and reliable agricultural information has become more important than ever before in the decision making process at international and national levels. Much of the required information for the agricultural sector, such as crop production estimates, livestock inventories and basic social and economic data pertaining to the sector, is obtained through Current Agricultural Surveys, which are periodic (annual or seasonal), national (or large-scale), multiple-purpose, agricultural data collection programmes. The establishment and development of such survey programmes is, therefore, a fundamental component of the agricultural information system.
From one source or another, annual estimates are available for most types of agricultural commodities in almost every country. Moreover, several estimates may exist for the same commodity. Often these estimates are based on nothing more than an educated, or worse an uneducated, guess. There is usually no way to judge their accuracy because they are not based on any procedure that allows for statistical analysis or evaluation.
The reasons why current agricultural statistics are poor for many countries are many. Lack of political support for data collection, the high cost of agricultural surveys, the shortage of requisite skills and the failure to identify the most appropriate methods are among the most commonly identified. Timely and reliable national statistics of a country's agricultural sector can only be provided by the establishment of an adequate, periodic, national agricultural survey based on probability sampling methods.
The Current Agricultural Surveys considered are based on probability sampling and estimation methods, which are the only ones that could provide timely and reliable basic data of a country's agricultural sector.
This manual describes the sample design, organization and implementation procedures of current agricultural surveys based on Multiple Frame Probability Sampling Methods.
Procedures are presented in the context of the considerations and steps needed to develop and maintain a current agricultural data collection programme in countries where experience with area frame and multiple frame sampling methods is lacking.
The intention is to introduce the subject in a straightforward and practical manner, and describe list and area frame construction and sample selection methods that require a minimum of resources and specialized staff taking into account the constraints faced by developing countries. In order to illustrate with specific examples the use of multiple frame methods in a range of countries and conditions, volume II will include summaries of multiple frame and area frame agricultural surveys conducted in the Albania, Brazil, Canada, the Czech Republic, France, Honduras, Italy, Morocco, Nicaragua, Pakistan, Spain and the United States. Briefer references will also be included to the area sample surveys undertaken in a number of other countries.
The survey designs proposed consider the world-wide increasing availability in recent years of satellite images, computerized area measurement and scale-transfer instruments, and even geographic information systems and hand-held global positioning systems, as tools for the application of the proposed multiple frame agricultural survey methods in developing countries. They may be the most practical way for a country to produce the required annual basic data for the agricultural sector.
The agricultural survey methods for developing countries presented are derived from those developed to conduct the multiple frame agricultural survey of the United States, which constitute a record and experience with multiple frame agricultural surveys of primary importance; and take into account the experience of applying area and multiple frame sampling methods in many countries in North Africa, Asia, North, Central and South America and Western Europe.
The Multiple Frame Agricultural Survey Design Considered
This manual describes the overall survey design of a Current Agricultural Survey Programme: a periodic (annual or seasonal), national (or large-scale), multiple-purpose, agricultural probability sample survey designed in order to obtain timely and reliable basic data for the agricultural sector.
Survey estimates are required, it is assumed, for most of the following variables: crop areas (prepared, planted and harvested), crop yields (forecast and achieved), crop production, livestock inventories, grain stocks, farming systems, cost of production, farm expenditures and social and economic characteristics of the agricultural holdings. The determination of the variables to be studied and their required level of accuracy should be clearly established at the outset of the survey planning since this has a direct bearing on the overall sampling design and, in particular, to the type of multiple frame estimators used, the questionnaire design and data collection procedures.
It is assumed also that, in terms of the timeliness of the survey design discussed, survey data should be collected, tabulated and published within a two month period. However, in the case of cost of production surveys or other special surveys that include variables that require a complex questionnaire and data collection procedure, the period to obtain the final results can be allowed to be longer.
The multiple frame sampling methods described combine a probability sample of land areas called segments, selected from an area frame, with a complementary short list of special agricultural holdings to be completely enumerated during the survey field data collection. The multiple frame estimates combine estimates from the area sample with estimates obtained from the list of special agricultural holdings.
Area sample component of the multiple frame survey
The area sample design consists of a stratified probability sample of segments, with a replicated selection procedure. The area sampling frame considers the territory divided into a number of land-use strata defined by proportion of cultivated land, predominance of certain crops or other land-use characteristics. The strata and sample segments must have identifiable physical boundaries (roads, paths, rivers, etc.) that can be located both in the field and on the cartographic materials used for their identification (satellite images, mosaics of aerial photos and maps). For the replicated sample selection, each land-use stratum is completely subdivided, by similarity of agricultural characteristics or following a geographic distribution criteria, into areas with equal number of segments called substrata or zones, which provides a further level of stratification. In each stratum, the area sample consists of a number of independent replicates. Each replicate is formed by one segment in each substrata, with equal probability of selection within the stratum. The area sample design can also be considered as a stratified, cluster sample of tracts, a tract consisting of the part of a holding (or non agricultural areas) included in the segment. Area sample surveys with segments that coincide with the land of agricultural holdings or with square segments are illustrated in several case studies included in volume II.
List sample component of the multiple frame survey
The complementary list of special holdings ensures the inclusion of those holdings which make a significant contribution to the total estimate of some important survey variables. Such list of special holdings may consist, for instance, of those holdings with the largest total area, those with the largest area for a given crop, those with the largest number of livestock and poultry, those with the largest revenues, those with the largest number of agricultural workers, those corresponding to a localized production and those concentrating on highly specialized types of production: these relatively small categories may not receive appropriate representation in an area sample with the consequence that the precision of the estimates for certain variables may be significantly affected. The technical difficulties of adding a short list of special agricultural holdings (the area frame component) to an area sample design are relatively minor.
Data collection procedures for the multiple frame survey
The area sample component involves an annual (or seasonal) field data collection carried out by enumerators that complete a questionnaire for each tract included in each selected sample segment. The enumerators collect the data for each tract by personal interviews with the holder or other respondent who can provide information on the tract. The data collection, in addition to completion of a questionnaire, often involves identification and measurement of agricultural areas. For each sample segment, the enumerator use an aerial photo enlargement (or a map or scale drawing), that includes the boundaries of the segment. This is called the segment photo. For each tract within a given sample segment the enumerator delineates on the segment photo the boundaries of the tract and the boundaries of all fields included in the tract. The enumerator verifies the crops planted and other uses of land for each field, information provided also by the holder. During the interview, the enumerator may also use a transparent grid on the segment photo to verify, approximately, the reported area of fields. Such identified agricultural areas in each sample segment can later be measured in the office using a computerized measurement instrument or a planimeter. The checking of area estimates made by holders and/or enumerators, provide a very important feature concerning data reliability. The enumerators also collect information on the complementary list frame of special holdings applying a questionnaire to each such holding.
Since the technical difficulties of adding a short list of special agricultural holdings (the list frame component) to an area sample design are relatively minor, most of the manual discuss the design and implementation of an area sample survey with segments that have identifiable physical boundaries, as described. Area sample surveys with segments that coincide with the land of agricultural holdings or with square segments are illustrated in several case studies included in volume II.
The decision to base a Current Agricultural Survey on the proposed multiple frame sampling methods should take into account local conditions and resources and also alternative survey methods. For this reason, Part I of this volume refers to the concepts and definitions required to describe and compare the main types of survey designs used to conduct such current agricultural surveys. Part II describes the sample design, organization and implementation procedures of agricultural area sample surveys with segments that have identifiable physical boundaries, that is, the area sample component of the multiple frame survey. Part III describes the preparation of list frame component of the survey which is the list of special agricultural holdings, and the multiple frame estimators.
PART I: Current Agricultural Surveys
Chapter 1 refers to the initial considerations required for the establishment of a current agricultural survey programme based on probability sample methods that could provide timely and reliable basic data for the agricultural sector; and Chapter 2 describes the main types of current agricultural survey designs. The chapter also discusses the problem of choosing an appropriate current agricultural survey design and provides a number of comparisons between multiple frame and other alternative survey methods.
PART II: Area Sample Designs with Segments that have Identifiable Physical Boundaries
Chapter 3 introduces the general characteristics of the survey design; the basic concepts, definitions and procedures used in area sampling and the criteria for determination of the main parameters used for area frame construction and sample selection: for example the total sample size and frame size, the number and type of land-use strata, the minimum size of compact blocks within strata, the allocation of the sample to strata, the target size of segments and primary sampling units as well as the permitted range of actual size in each land-use stratum.
Chapter 4 presents the survey estimation methods: the formulae for the direct expansion and variance calculations for the closed, weighted and open segment estimators along with their associated strengths, weaknesses and applications. Radio estimators and their variance formulae, both for within- survey and between- surveys, are also presented in addition to their strengths, weaknesses and applications.
Chapter 5 covers in more detail the resources required for the implementation and maintenance of the area sample survey, including personnel, equipment and materials. It refers to the cartographic materials (maps, satellite images and aerial photographs) and instruments needed in order to identify, transfer and measure the area units involved in area frame construction and sample selection.
Chapter 6 deals with the area frame construction procedures, that is, land-use stratification and subdivision of the strata into Primary Sampling Units (defined as areas with physical boundaries used to determine the exact size of the area frame and to facilitate the subdivision of strata into substrata and the sample selection). It covers, also, the use of different combinations of cartographic materials (satellite images, maps and aerial photographs), area measurement and scale-transfer instruments and equipment. In particular, it includes a section on the current use of satellite images for area frame construction.
Chapter 7 deals with the area sample selection of segments, that is, with the procedures for selecting the random (or systematic) sample replicates. It includes also a point sampling procedure for selecting segments in strata that contain few easily recognizable physical boundaries. And deals with the preparation of segment photo enlargements or scale drawings to assist data collection.
Chapter 8 describes the survey preparations needed for data collection, including preparation of the survey questionnaire and other forms, timing, hiring and training of supervisors and enumerators, materials and equipment for field staff. And covers the survey field data collection procedures used by enumerators to conduct direct interviews with holders with the assistance of photo enlargements of the segment.
Chapter 9 summarizes issues concerned with the survey data processing and analysis, including questionnaire editing and review, data entry, and the preparation of preliminary tabulations of the final results.
PART III: The Multiple Frame Survey Design
Chapter 10 discusses the preparation of the complementary list frame of special agricultural holdings to be enumerated during the field data collection.
Chapter 11 presents the multiple frame estimation procedures and provides some actual examples of survey results.
Chapter 12 features a selection of topics that require special attention and knowledge in the design and implementation of a multiple frame agricultural survey. It also presents some concluding views and basic recommendations for the planning and implementation of periodic, multiple frame agricultural surveys in developing countries.
The multiple frame agricultural survey methods described represent, in several ways, an improvement on those methods based exclusively on a list sample of agricultural holdings or holders' addresses currently used in most countries as the basis for their current agricultural surveys. Multiple frame methods result in grater precision of estimates of agricultural areas, the main crop areas and other key variables of all multiple-purpose agricultural surveys, since the area sample component involves a practical procedure for the objective measurement of agricultural areas on aerial photos. In addition, the area sample component provides the means for selecting probability samples of fields needed to conduct crop cutting yield surveys, that provide objective crop production and crop forecasting estimates.