FAO STATISTICAL DEVELOPMENT SERIES
CONDUCTING AGRICULTURAL CENSUSES AND SURVEYS
FOOD AND AGRICULTURE ORGANIZATION OF THE UNITED NATIONS
The census questionnaire is considered to be the basic census document, and is of key importance to the success of the census. This chapter discusses factors to be taken into account when preparing the questionnaire such as: size of the questionnaire, ability of agricultural holders to provide information required, kinds of questions, electronic processing of the questionnaire, etc.
Other primary topics closely related to questionnaire formulation are pre-test surveys and pilot censuses, which can test the questionnaire in the field (Chapter 13) and the Tabulation Plan (Chapter 9). The questionnaire is the means of collecting data required for the proposed tabulation plan. It is generally considered, that the tabulation plan should be developed before the questionnaire.
No examples of a census questionnaire are included in this chapter. Readers involved in questionnaire formulation are advised to find examples of questionnaires from previous censuses, other agricultural surveys, or from neighbouring countries. No suggestions have been made, regarding specific data items to be included in the questionnaire, as this subject is covered in the FAO document "Programme for the World Census of Agriculture 2000".
Development of the census questionnaire
8.1 Once the decision on the scope and coverage of the census has been taken, a questionnaire can be developed in order to secure the relevant information in an orderly and coordinated manner. The census questionnaire is the most basic document in the census programme since it becomes the vehicle for collecting the desired information. Any deficiencies in the questionnaire design will lead to incomplete and inaccurate data being collected. Considerable thought should be given to formulating the questionnaire and input sought from available experts on this subject matter.
8.2 Generally three methods, (i) self-enumeration, (ii) interview method and (iii) objective measurements, are considered for obtaining census data. In a number of developed countries where the holder is literate and maintains records, or in the case of modern holdings/plantations in developing countries, self-enumeration is the preferred approach. For example, in the U.S.A. census data are obtained mainly through the mail-out/mail-back method. This consists of mailing the questionnaire to holders with several mail or telephone follow-ups when necessary, and, as a last resort, a personal visit of enumerators. Alternatively, self-enumeration may be organized by enumerators who visit holdings twice, first to leave the questionnaire and provide instructions, and second to review and collect the completed questionnaires with the possibility of transforming the second visit into an interview when necessary. In developing countries, where many holders are illiterate the interview method is necessary. Objective measurements of areas and yield are used in countries where holders are not familiar with standard units of measurement and where cadastre information does not exist.
8.3 Questionnaire, form or schedule are generally synonymous in statistical literature. Whether any distinction is made or not, the format of the questionnaire will depend on themethod of inquiry, whether by interview, by self-enumeration, intended for objective measurement or a combination of these methods.
8.4 The size and form of the questionnaire are the first items to be considered. The questionnaire should not be too large and should be easy to handle in the field. Its size and shape should be such that the enumerator can easily handle it in the field while recording the respondents' answers.
8.5 The temptation to use the census to ask a great number of questions of interest to official and private data users should be resisted even if it is often argued that once the holder has been contacted, maximum advantage should be taken in collecting the necessary census information as it is more costly and time consuming to meet him/her than to obtain data through the mail. However, this argument is not valid for many reasons. Frequently the data requested is not readily known to the holder who may need to consult the records and other members of the household, which takes time. Furthermore, if the questionnaire is lengthy, the holder, who at the outset is prepared to reply to the questions, may become less cooperative after being questioned for a long time. It is, therefore, very important that the questionnaire is not too lengthy. However, it is difficult to establish the ideal length of a questionnaire because this depends not only on the number of questions it contains, but also on their degree of complexity which reflects directly on the length of time the holder needs to answer. Not only is the holder affected adversely by the length of the questionnaire and duration of the interview, but the enumerator also becomes tired and makes careless mistakes in recording the data. As a general rule, an interview should not exceed 45 minutes.
8.6 The definitions and concepts to be used in the questionnaire should be carefully studied and care should be taken to make sure that they are easily understood by the holder and the census field staff. An endeavour should be made to follow the recommendations of international programmes, for purposes of data comparison at regional and global levels.
8.7 Census data collection should normally be limited to structural items which do not change quickly over time, while more detailed information should be collected through specialized surveys subsequent to the census.
8.8 If it is observed that the questionnaire is too lengthy after carefully studying the subjects to be included and the corresponding questions, various possibilities may be considered.
8.9 One possibility is to distribute the questions in two or more questionnaires. It is almost certain that on smallholdings agricultural machinery is not used, nor are labourers hired for farm work, nor is there an irrigation system, so these items could be included in another questionnaire. In this case the first questionnaire would be of an acceptable length and the second would be rather small. This procedure is used also to widen the scope of the census, the first questionnaire applying to all the holdings and the others relating to only specialized items like vineyards, greenhouses, nurseries, etc., in a complete enumeration or to a sample of them. In addition, when considering the subjects which need more in-depth study, the reliability of the sample results has to be taken into account when deciding what questions are going to be included in each questionnaire.
8.10 Another possibility is to use different questionnaires for different provinces when these differ considerably in crop and cropping practices. In this case various items could be removed completely from the questionnaire of one province and its length reduced considerably. For example, if one province is known to be almost exclusively a livestock production area and owing to its physical characteristics has no permanent crops, thequestions regarding crops may be reduced and those relating to livestock expanded. This may, however, involve printing a different set of questionnaires to meet the requirements of different provinces.
8.11 Once the decision on what subjects are to be included in the questionnaire has been taken, attention must be paid to question sequence, that is to say, they should be set out in a logical order so that it is easy for the holder to supply the requested information. All the questions on one subject should be grouped together and enough space left between them so that they are distinct and can be easily located.
8.12 The questions should be formulated in a clear simple language, using, wherever possible, the vocabulary familiar to the holder. This is not always possible because in the majority of countries there are local differences, and expressions which are very common in one part of the country may be unknown in another. However, when there are terms commonly used by the holder, although these may not be correct idiomatically, they should be employed in preference to others. Similarly, measurement units sometimes vary from province to province. It is desirable to record the data in local units and to convert into standard units later in the census office.
8.13 At times it is advisable for a smooth transition between subject matter or to lead into a subject to use introductory questions or statements, which are not tabulated, but serve as a control or to introduce another question so as not to lose the informant. For example, the area rented from others could be asked directly, but it is preferable to ask first whether any land was rented from others and, if so, how many hectares were rented. Furthermore, such introductory questions may be very useful in the data entry stage and are discussed later in this chapter.
8.14 Attention should be given to the quality of paper used. Thin paper should not be chosen because during the field work the questionnaire is often subjected to very unfavourable climatic conditions and to constant handling during the distribution of the document and the subsequent tabulation of the data.
8.15 Another aspect to be considered is the colour of the paper. When an agricultural census is taken at the same time as a population census, it is very useful if different colours are used for the two sets of questionnaires, so that they can be easily distinguished and errors avoided in handling the census papers. The same reasoning applies when two types of questionnaire are used for collecting information, one of which will be answered only by a sample of holders. In general, since a large quantity of questionnaires are handled in a census, it is advisable to use different colours for each different questionnaire. Light colours which do not strain the eyes and on which it is easy to read should be chosen. Colours should change for each census so that questionnaires from different censuses can be easily distinguished if used during editing and processing.
8.16 The size of the print should be easily read even when light is not adequate. This occurs frequently when the holders are interviewed in the evening, since in many rural areas there is minimum lighting. Often, census offices, in order to keep questionnaires to a reasonable size, use a small print which is unacceptable.
8.17 Efforts should be made to use a different type of print for the questions and for notes or instructions to the enumerators. The questions must be easily distinguished since most of them will have to be read aloud to the holder. However, very heavy print should not be used as the questionnaire will look overloaded.
8.18 The space for replies should be large enough so that there is room for responses to be entered and the lines should not be printed too close together. If lines are close the enumerator, when correcting some of the answers given by the holder, may rub out the preceding answer with the result that the question has to be repeated or the data is omitted or made illegible.
8.19 The questionnaire must be uniform in style, that is to say the readings should all be printed with the same type of lettering, the explanatory notes with another type or in brackets or shaded, and in a certain position with respect to the question, after or below it. Similarly the coding system, if any, should be printed in the same position and be of the same size in each section of the questionnaire.
8.20 Each question should be numbered so as to be able to refer to it easily in the instructions and elsewhere. The same applies when the answers are recorded in different columns; each column must bear a number or letter.
Census questionnaire working group
8.21 Bearing in mind that the information obtained in the census will be used for the country's agricultural planning, it is very important that a working group be formed whose task will be to work with data users to determine the specific questions which the questionnaire should contain. The group should include staff who are involved in agricultural planning, in collection of statistics or as a data user within the agricultural sector, so that they know the informational needs and can visualize how the data obtained will be used. They should hold responsible posts in their respective departments. It is not advisable for the group to be large. A basic group of three to five senior and experienced officials could be formed. The group can be assisted by specialists from the different ministries according to the subjects to be discussed. For example, when aspects of irrigation and drainage are considered, experts from the ministry in charge of the country's water and irrigation resources should be present. A data processing specialist should also be associated with the formulation of the questionnaire.
8.22 For assistance on the subjects to be included, reference can be made to the international recommendations contained in the FAO document "Programme for the World Census of Agriculture 2000", which has attempted to identify the basic information needs of the agricultural sector for the decade 1996-2005.
8.23 Reference should be made to the last agricultural census taken in the country, if available. Starting, with the questionnaire used in the previous census, each of the items included therein could be carefully examined, the content of the questionnaire compared with international recommendations, and each of the questions examined, studying the difficulties encountered and the use made of the information collected.
8.24 As the economic planning in a country becomes more refined and as agriculture changes, the information needs change and items which were not included in the census taken earlier may now be important. For example, the application of fertilizers and the use of improved varieties of seeds were practices which were perhaps not very widespread previously, but which are now important. The reverse may also occur, that is to say that subjects considered useful in the previous census are no longer of any value. The questionnaire must meet the needs of current data users and consider future expectations.
Use of questionnaires developed in other countries
8.25 It is advisable to study the questionnaires used by other countries, especially those of the same region, because it is quite possible that their information needs will be similar and they may have similar data collection problems. It may be possible to profit from their experiences, utilizing ideas and approaches not only as regards to the items included, but also the presentation, taking care, of course, to examine whether these innovations are applicable to the country.
8.26 The validity of various questions in the questionnaire can be evaluated by conducting a pseudo-tabulation. Through such tabulation it is possible to determine whether all information targeted from the census can be obtained. It is therefore very important to design the tabulation plan at the same time as the questionnaire. Each question appearing in the questionnaire could then be studied to determine whether it could supply the data needed in the tabulation plan. As a general guide, data not intended for tabulation should not be collected. There are some exceptions, such as: identification (name of the holder, address, etc.), introductory questions mentioned above and questions intended for data validation.
Processing the questionnaire
8.27 It is very important to analyze whether the information recorded on the questionnaire can be processed easily. For this purpose full collaboration is necessary between the group in charge of designing the questionnaire and the data processors. The questionnaire design must assure that the presentation is simple. Many printed codes can cause difficulty for enumerators in the field. In case of conflict between data collection and data processing requirements, priority is given, in principle, to data collection requirements for the simple reason that enumerators are often working under adverse conditions. A number of important aspects which may affect the ease of processing the questionnaire are given below.
Indicate area under different tenure forms:
|TENURE FORM||AREA IN Ha|
|1. Land owned|
|2. Land rented from others|
|3. Land rented to others|
|TOTAL LAND OPERATED|
( 1 + 2 - 3 )
Frame 8.1 Example of numerical value question
8.28 An identification code should uniquely define each questionnaire and should always be numerical (not alphabetical). It should be as short as possible, although some redundancy or control code may be desirable in order to minimize possible errors and to help locate the correct identification code in case an error occurs. It should distinguish different questionnaires and in case of complicated hierarchy of questionnaires (e.g., several parcels per holding, several fields per parcel, several crops per field, etc.) each part will have to have its own code to permit sorting and linking of data. In case of sample enumeration, identification should provide sufficient information for assignment of expansion factors (strata, primary sampling units, area segments, etc.). The identification code should also allow distinction between administrative (or other) areas for which tabulation is required.
8.29 From the point of view of data processing one can distinguish five different types of questions:
Indicate legal status of the holder
(check one box only):
Frame 8.2 Example of multiple choice question
Indicate type of aquaculture installation used for fisheries
(check one or more boxes):
Frame 8.3 Example of multiple answer question
Any livestock kept on this holding?
Frame 8.4 Example of introductory question
8.30 For data processing, questions of types (i) and (ii) create no problem. Type (iv) is very useful, particularly at data entry stage, while types (iii) and (v) create problems and it is advisable to avoid them when possible.
8.31 Type (iv) introductory questions are useful as enumerators can skip whole sections without needing to enter all zeros as might normally be required. They are obliged, however, to give an answer to the introductory question even if the answer is no, otherwise the editor cannot determine if the section was just overlooked or properly skipped. Similarly,at data entry stage, introductory questions allow skipping to the next section with one keystroke.
8.32 Type (iii) multiple-answer questions, although easily processed, create some confusion for the tabulation plan and record design. It is better to replace such questions with multiple choice questions (type (ii)), sacrificing part of the information (in the example given above, the question may simply be "main type of aquaculture used for fisheries").
8.33 Type (v) open-ended questions are frequent in agricultural censuses and create problems. Category "Other - specify" is difficult to process properly. At best it provides useful information for planning the next agricultural census. It is advisable, therefore, not to plan tabulation of these additional categories, but to keep them together under the title "other". If details are important a list of these additional items should be obtained from other sources or from the pilot census. The only other way to use these data are to develop codes for minor items and code the answers before data entry.
8.34 For crops, it is preferable to provide a list of those crops which are of interest as a part of the questionnaire and provide pre-codes for the enumerators' use. When a short list is printed and the enumerator records many names of crops in the space allotted to "others" the coding takes time and can become a difficult task prone to errors. Generally the major crops of the country are known and can be listed.
8.35 Most of the questions in the agricultural census are such that precoding on the questionnaire is possible and it is strongly recommended in order to simplify post enumeration coding and facilitate processing.
8.36 Fully precoded questionnaires are often used in agricultural censuses. The need for this arises because agricultural census questionnaires contain many zero entries; many holdings have no livestock, machinery or hired workers. Only a few holdings in many countries use fertilizers or pesticides, and they will rarely have more than a few crops. Under the circumstances and in order to save time for data entry and reduce the memory space needed on the computer, a different code is assigned to each possible entry in the questionnaire (usually pre-printed in a corner or in front of the box made available for the data). Depending on data processing methods, great economy is possible by entering into the computer only non-zero entries and their code. With modern equipment and software where data entry uses screens for visualizing parts of the questionnaire and permits skipping large parts of questionnaires with "zero' entries, this approach is gradually being abandoned. For this reason, introductory questions inside the questionnaire, permitting skipping parts of the questionnaire by one key stroke, have become very important.
8.37 Another characteristic of the agricultural census questionnaire which creates difficulties in data processing is the complicated hierarchy of parts of the questionnaire such as parcels, fields, and plots. This can be simplified by choosing only one parcel/field/plot as the unit of enumeration. This is justified in countries where most of the parcels have just one field or plot. Further simplification is found in Europe and many countries in the Americas where crop data are collected at holding level. In this case aggregation is normally done by the holder, who probably knows the totals better than plot-by-plot. Similarly, data on population can be collected in the form of aggregates (such as number of males under 15 years of age, etc.; the approach was used in FAO Programmes for 1950, 1960 and 1970). Simplification of the questionnaire in respect of hierarchy of various data sets may be the decisive factor in simplifying the organization of data processing and is, therefore, strongly recommended.
Testing of census questionnaires
8.38 Although the members of the group who designed the questionnaire may be very competent, it is essential that its functionality be evaluated by means of a series of pre-test surveys and pilot censuses. This issue is dealt with in Chapter 13. After taking into account the field testing experience and evaluating data inconsistencies and illogical replies, which might indicate that the questions were not understood by the holder and/or the enumerator, the questionnaires should be very carefully revised. All questionnaire tests should be completed well in advance of the actual census to allow time to make the necessary changes to the questionnaires, and consequently to the instruction manuals and, if necessary, to again pretest. In view of the large number of questionnaires and instruction manuals required, sufficient time must be allowed for printing.
FAO (1965). Some problems of agricultural census taking with special reference to developing countries (by V.G. Panse).
UN (1982). Survey data processing: A review of issues and procedures. NHSCP technical study.