Previous Page Table of Contents Next Page


CHAPTER IV. METHODOLOGICAL GUIDELINES


4.1 SOURCES OF INFORMATION
4.2 AGRICULTURAL CENSUSES

The production of statistics must be carefully monitored at every stage of the process to ensure that they take gender considerations into account.

Although they are not the only source of information on the sector, agricultural censuses are central to, and act as a link among, all other related surveys. Some of these other surveys are made as post-census updates of certain specific indicators, while others seek to investigate topics too complex to cover in a census. Some surveys cover topics outside the agricultural sector as such, but still supply data that are relevant.

Among other useful tools that supply data on the sector are population censuses, administrative records and various types of surveys. Each of these sources of statistical data has its advantages and drawbacks.

A unique virtue of agricultural censuses is their geographical coverage. They are the only source that report on all agricultural establishments, large and small, throughout a country. Such operational scope does, however, imply that only a very few simplified questions can be investigated on the topics covered, and this reflects on both the depth and the quality of the data that can be captured.

Some of these shortcomings can be offset by ad hoc sample surveys designed to report on specific topics in more detail. For example, time-use studies or food consumption surveys can include more questions than more general surveys. Such surveys also have access to highly trained staff and greater control over field enumerators and the data processing stages. A major limitation of any survey is, however, that it can either supply significant data for large geographical aggregations or else for a restricted number of areas, depending on the sample design, which depends, in turn, on the specific survey objectives and available budgetary resources. Sample design and application problems can also arise at the field level.

Despite its imperfections, any source that has been treated with statistical rigour can be of great help. Valid inferences can be made once the possible biases, gaps and scope of a source are known.

The database should be produced according to a statistical system that brings together interlinked sources sharing a common conceptual and methodological basis. Barring that, there must be mechanisms to ensure that the data collected complement one another.

4.1 SOURCES OF INFORMATION


4.4.1 Population and housing censuses
4.1.2 Food consumption surveys
4.1.3 Household income and expenditure surveys
4.1.4 Employment surveys
4.1.5 Time-use surveys
4.1.6 Natural resource management and environmental degradation

The following is a brief overview of the various sources of data on the agricultural sector that are open to the incorporation of a gender perspective. A detailed description of the characteristics of agricultural censuses - the backbone of the agricultural sector data system - follows.

4.4.1 Population and housing censuses

Population and housing censuses supply a wide range of useful data on the general population. They describe the socio-demographic characteristics of the members of the household, their main occupations and data that are useful for estimating internal migration. Housing censuses, which are normally carried out in tandem with population censuses, furnish data on the buildings in which people live and the services they enjoy.

These two censuses tap important data on the rural population and should be fully exploited, but their very nature means that they do not supply a global overview of all those involved in agriculture. The snapshots that are obtained from data collected at a given moment or over a very short period (e.g. the week before the census) are insufficient for measuring rural employment because farm work is so highly seasonal. Moreover, a population census normally records only the task which the respondent considers to be his or her main activity. If that is not farm work,50 the agricultural activity will go unrecorded, leading to the undercounting of the number of persons making a contribution in agriculture. Another crucial problem is the under-representation of women's farm and non-farm activities, as discussed in Chapters 1 and 2.

50 Population censuses usually leave it to the respondent to decide whether his or her main economic activity is that which is: most lucrative, most time-consuming, most stable, or the one in which he or she has worked longest. When the respondent engages in several activities, it is always recommended that precedence be given to an economic over a non-economic one, so a retired person or a homemaker can still be identified as a working person. Prevailing prejudices, however, sometimes defeat this precaution.
The population census offers many advantages for an understanding of the agricultural sector. One of the most important of these is the fact that its universal coverage of both urban and rural areas makes it an excellent framework in which to define households and identify small production units.

Ideally, the population census should include a specific question on agricultural work, because a general question on additional work activities will not necessarily produce data on agriculture. The respondent may not even consider farm work as a secondary activity. Many women farm parcels of land, for example, and sell the produce directly or as processed, ready-to-eat foods such as tortillas, tamales or cooked soft maize. But when questioned about the work they do, they usually describe themselves as merchants who process food or provide a service, failing to mention that they are also farmers.

Although some countries, such as Costa Rica, have conducted agricultural and population censuses in tandem, this is not usually advisable because of the operational complexity. The inclusion of a specific question about the respondent's farm activities in the population census questionnaire should make it possible to identify which households to follow up on with the agricultural census or sample survey questionnaire (those who answered in the affirmative). Obviously, the two surveys should follow one another fairly closely, and full collaboration between the two census teams is essential, although this may not be viable when the two surveys are done by different institutions.

4.1.2 Food consumption surveys

These surveys are intended primarily to collect data on food consumption and sometimes also seek to identify nutritional conditions. The data on food consumption refer to global household consumption, which makes it very difficult to find out about the diet of individual household members, especially when they all eat together, often out of the same dish. The average per caput consumption in a household does not reflect the real consumption of each member. In general, given their poor level of nutrition, it is assumed that women in very poor households eat less than the rest of the family.

The measurement of nutritional status requires a sophisticated approach, for example, anthropometric measurements such as upper arm circumference and height to weight ratio should be taken. But such surveys are expensive and require specialized teams and equipment.

4.1.3 Household income and expenditure surveys

This type of survey usually collects data on income and expenditure (taking the household as the unit of analysis and observation) and may include questions on the main occupation of each household member of working age. In other cases, questions may be directed to the head of household only. When everyone is interviewed, the limitations are the same as those for the population survey (i.e. only the main occupation is captured). But when only the head of household is interviewed, the data are highly limited as only one person is recorded, and this is unlikely to be a woman.

Analysis of the distribution of household and income expenditure refers solely to the head of the household - a built-in gender bias that results in both the contribution of each member of the household and the way in which income is distributed remaining invisible. Another serious limitation of such surveys is the (sometimes deliberate) omission of rural areas or households involved in agricultural production. The most useful feature of household income and expenditure surveys is that they allow extrapolation of the expenditure for food.

4.1.4 Employment surveys

The coverage of employment or labour force surveys varies, and they are usually based on probability sample surveys. Some countries include only the major cities, others only urban areas. Such surveys may even be designed to cover the total population with no subdivision such as rural/urban, or any sort of area differentiation.

Employment surveys can supply data on people living and working in rural areas (including unpaid family workers), people working in small production units or dayworkers on large establishments, provided that they have been specifically designed for this purpose. Seasonal changes are captured through periodic surveys.

It is essential that the same problem that is inherent in population surveys - i.e. the omission of agriculture when it is not considered the main occupation - be avoided. Data can be collected on population segments that are usually missed out - children, older people and, above all, women. Fortunately, the major problems can now be overcome by using the backlog of statistical experience that is available.

4.1.5 Time-use surveys

The best tool for evaluating gender contributions to economic and non-economic activities is the time-use survey, which is ideal for examining the gender division of both paid and unpaid labour within the household.

Time-use surveys are especially useful for measuring the gender contributions of rural people in farm work. Unfortunately, the blurring of economic activities and domestic tasks makes it more difficult for them to be accurate. People also have different ways of perceiving time, and it may not be measured in the same way in rural as in urban areas. In such cases it is advisable to bolster statistical with anthropological techniques. Concerning this, the following point can be made:

"The concept of "time worked" in the agricultural industry is much more difficult to apprehend than in other branches of economic activity. There is no fixed place of work, as farm work includes working in the fields, preparing agricultural products for marketing, taking farm products to market, bringing farm requisites from town, keeping farm records, etc. Part of the work is done on the holding, another part in the holder's dwelling, still another in the markets or in the town, etc. Travel between these different sites of work may be long and time-consuming. It is therefore advisable to include all relevant periods of work and travel time when recording the time worked by the holder, family workers and paid workers."51

51 FAO. Collecting statistics on agriculture, population and employment. Economic and Social Development Paper No. 7, p. 119. Rome.
There is a growing body of experience in this field, and national statistics bureaux are now making increasing use of time-use studies to quantify unpaid work and other concepts for inclusion in the System of National Accounts.

4.1.6 Natural resource management and environmental degradation

Faced with the universal environmental decline, succinctly referred to as "global change" (which includes the thinning of the ozone layer, global warming and the loss of biodiversity), research efforts and action plans with very diverse objectives, coverage and methodologies have been initiated all over the world. Of these, the most relevant to the topics covered in this paper include the efforts to generate data on the exploitation of wood and non-wood resources, especially where these resources are used as a supplement in subsistence agriculture.

4.2 AGRICULTURAL CENSUSES


4.2.1 The scope and objectives of the census
4.2.2 Questionnaire design
4.2.3 Tabulation plan
4.2.4 Census publicity and respondents' cooperation
4.2.5 Statistical unit and coverage
4.2.6 Designing the sample
4.2.7 Staff recruitment
4.2.8 Instructions and training for enumerators
4.2.9 Instructions and training for supervisors
4.2.10 Choosing the respondent
4.2.11 The pilot census
4.2.12 Data processing
4.2.13 Data presentation
4.2.14 Dissemination

Agricultural censuses collect data on the agricultural holding. The common focus is on production - harvests, livestock and agricultural inputs. Human resources are completely overlooked in many countries and, in others, they receive only the most marginal attention. Few nations design questionnaires that include detailed questions on household members and hired labour.

In 1995, FAO52 its voiced concern about the problem of universal coverage and expressed interest in the activities of members of households in which there is a one-to-one correspondence between household and holding.53 This new outlook is bound to mark a turning-point in the history of agricultural censuses. Putting the emphasis on production and factors of production in isolation from the labour force has left little room for capturing data on people involved in the various agricultural productive processes.54 The data obtained and recorded on the size and characteristics of the labour force have, therefore, been quite scanty.

52 Op. cit., footnote 25.

53 This concerns the proposal for a household type linked to one agricultural production unit (op cit., footnote 25, p. 29), and the consequent need to capture the socio-demographic characteristics of all members of the household and their participation in farm and non-farm activities (Ibid, p. 30-31). These aspects are also covered in Chapter 3.

54 It was argued that a figure for the number of farm workers could be derived from the population censuses, until the limitations of this source of data in capturing all of the economically active population actually working in agriculture became clear.

One common drawback of agricultural censuses is the exclusion of smallholdings, a subsector where women and other family members play a particularly important role. This oversight is partly the result of technical reasons and partly because, by excluding smallholdings, agricultural censuses can save money and simplify enumeration. In the past, the main objective was to describe the large production units where output was concentrated. However, now that poverty, food insecurity and environmental degradation are under the spotlight, there is an urgent need for appropriate methods to discern the characteristics, inner logic and functions of that part of the population, women as well as men, whose subsistence depends on the output of their small farms.

In some countries, holdings located in urban zones and/or those below a certain (subjectively defined) size are not normally included in the census. The two reasons usually given for this omission - limited financial resources and the practical difficulties of identifying such properties - hide a third reason, which is the predominance of a way of thinking that arose in the developed countries, where such holdings do not, in fact, contribute much to overall agricultural output and are therefore considered "secondary" or "minor" activities. However, the following points need to be considered:

· Excluding urban areas may mean underestimating women's contribution to agricultural production, especially in the developing countries.

· Many households depend for their survival on small-scale production, so information on the activities and characteristics of such units is decisive for policy formulation.

· The ambiguity about what is rural and what is urban has become such a serious and wide-ranging problem that the United Nations has never ventured to spell out specific guidelines. In many countries (including Mozambique), "urban" is a legal category determined by policy and administrative requisites, regardless of population density, available services or economic activity.

Each country will seek to solve this problem in terms of its own statistical infrastructure and financial resources. Some possible solutions are:

· Differentiated questionnaires for the same agricultural census - one for smallholdings and the other for large and medium-sized holdings;

· A built-in probability sample survey to collect data on small units within the agricultural census;

· Probability sample surveys on agriculture and rural employment (using the agricultural census as a sample frame) with the specific objective of collecting more detailed data on small units.


Whatever method is used, it is essential that smallholdings, and the women, children, young and elderly people who work on them, be represented in statistics. A major prerequisite for gender-sensitive data production is the close monitoring of all stages of the data collection process. A census is a major undertaking, and should be embarked on only when success is certain, a legal framework is in place to identify the institution responsible (and its duties and responsibilities), the budget has been prepared, and sources of financing are secured. Ideally, users should participate at various stages of this preparation process; in addition to the technical advantages, this would also serve to enlist users' support in securing the funding that such a complex effort requires. Once the decision has been made, work teams are formed and an efficient coordination system is devised. Each group has specific and clearly outlined responsibilities in the following areas:

· Conceptual design;
· Administration;
· Fieldwork;
· Data processing;
· Social communication.
The various teams have a great many tasks to perform, the most salient of which are mentioned here.55
55 See op. cit., footnote 25, and other FAO documents for a more in-depth treatment of each topic.
The design team first reviews the objectives and scope of the census, determining whether a sample survey is needed. The team designs the data collection tools, instruction manuals, training materials, tabulations and data presentation formats, defines the criteria for data verification and reviews criteria for the classification and grouping of variables.

The administrative team is responsible for the logistics, for ensuring that supplies arrive on time and for acquiring and monitoring the budget.

The field team's participation ranges from mapping (or reviewing existing maps) to enumeration. It coordinates the training of supervisors and enumerators in collaboration with the data processing team. It defines the critical paths for data collection, takes the census and provides supervision. It reviews the documentation of all material used in the census and, if necessary, field-tests the sampling frame, following guidelines worked out by the design team.

The data processing team designs the data collection and verification programmes and compiles data in accordance with the design team's indications. It participates in the design of the questionnaire and checklist format (with the field team) to ensure that it will both facilitate the teams' work and avoid errors in data collection, compilation and processing.

The social communications team is responsible for the basic work of sensitizing respondents in order to enlist their cooperation, reduce bias and disseminate the most relevant data to the public.

The following are the main stages in the planning and carrying out of a census:

· Definition of census objectives and scope;
· Design of questionnaire;
· Tabulation plan;
· Census publicity campaign and enlisting of respondents' support;
· Enumeration and coverage;
· Sampling;
· Staff recruitment;
· Production of instruction manuals and training of enumerators and supervisors;
· Checking of respondents' credentials;
· Pilot census;
· Data processing;
· Dissemination programme.

4.2.1 The scope and objectives of the census

The adoption of a gender perspective implies the introduction of new themes or, at the very least, the adoption of existing international guidelines and recommendations. Discussion and agreement are an essential first step in ensuring that the collected data accurately measure men's and women's work, capabilities and access to available resources. Indeed, such discussion is inherent in the planning of census operations and the setting of objectives.

The data needed must be identified and the objectives and scope of the survey accurately determined, while respecting specific country interests and policy considerations. Another essential step at this stage is to review existing studies and surveys from other centres so as to avoid duplicating efforts, ensure conceptual coherence and lay the foundations for an integrated system of agricultural statistics.

Briefly, the census objectives are to:

· Obtain, process and disseminate basic data on the agricultural and forestry sector, i.e. evaluate the product generated (supplying food and raw materials), the contribution of the men and women who worked to produce it, and the material resources used. Precautions must be taken to avoid the omission of small production units;

· Enumerate all holdings in the country (the broadest possible coverage) to ensure a relevant, high-quality product;

· Develop sampling frames to serve as the basis for a system of periodic agricultural and forestry surveys.

4.2.2 Questionnaire design

From the gender point of view, statistical quality depends on the questionnaire. If it is not carefully and appropriately designed, the concepts and definitions, as well as the order, formulation and wording of questions, can all adversely affect the answers, and may even give rise to errors. The credibility and relevance of the answers given is heavily dependent on questionnaire design.

A gender perspective can be incorporated by revising the concepts and definitions, reformulating the questions and reorienting current working methods and procedures. Attempts to collect gender-related data are often squashed with the argument that the process will make the questionnaire too long or too repetitive, but new questions are not usually necessary for the incorporation of a gender perspective.

Various user groups should be asked to cooperate in the design of the census questionnaire. An advisory committee of persons sensitive to gender considerations should be brought in at the questionnaire design stage. The committee should be knowledgeable about the kind of data that a survey or census can capture, and about the analytical use to which the statistical data may be put. Such a committee can help to clarify and defend gender concepts and overcome conventional resistance to the issue of gender.

In the past, statistical definitions and concepts tended to be elaborated without taking gender specifics into account. Essential indicators of women's situations were simply ignored. In recent years, many of the international efforts to improve the quality and coverage of statistics from the gender point of view have focused on revising standard definitions, particularly those concerning employment.56 Concepts and definitions have been elaborated by experts in a range of special fields - including economic statistics, employment and demography - based on the direct experience of the national statistical institutes, the findings of specific research studies and case studies, and have done much to improve the measure of gender-specific situations. Countries should make every effort to adopt the international recommendations, because they offer the possibility for data comparison and the for the incorporation of updated country data into the international organizations' databases. The questionnaires should make every possible attempt to avoid the most well-known biases. Typical causes of error include:57

56 Important steps in this process were the 1982 definitions of the economically active population and of the unemployed adopted made by the International Conference of Labour Statisticians; the 1993 revision of the United Nations System of National Accounts; and the 1993 adoption of the first definition of the informal sector by the International Conference of Labour Statisticians. See Chapters 2 and 3.

57 Adapted from Perucci, F. & Hedman, B. Engendering statistics. (mimeograph)

Inadequate definitions and concepts. These fail to reflect accurately the gender differentiations common in rural areas, such as the conventional definitions of head of family, head of holding, economic activity, and the like.

Erroneous wording of questions. A typical case concerns the question about work on the holding, which is often so badly worded that work is construed solely as the regular exercise of a remunerated activity. As a result, many women are not recorded in the census.

Selecting the wrong respondent. In this case, the respondent selected is not in a position to report correctly either on people belonging to the holding or people working on it. Male respondents may, for example, report women who are actually working on the holding as not economically active.

Using the wrong enumerator. Enumerators can introduce biases and personal values in the way they formulate questions as a result of their own prejudices, insufficient training or simple carelessness.

Communication problems. These arise when respondents fail to understand the content or language of the questionnaire. This occurs most frequently in interviews with women where the wording of the question is too technical or the terminology too complex.

Obscuring the truth. In this case respondents deliberately give a wrong answer, either to meet some socially acceptable norm or because they are fearful or suspicious about why the question is being asked. A man may deliberately deny that his wife works on the holding, for example.


Conceptual framework

The first stage in the development of the questionnaire is the identification and specification of concepts and definitions. This is pivotal to measuring the work that men and women do on the holding, their socio-economic characteristics and gender access to resources. These aspects have all been discussed in detail in the preceding chapters, where it has been seen that women's work and the work done on female-headed holdings can easily be underestimated when the wrong definitions are adopted.

Questionnaire concepts must be framed as concrete questions, according to the following steps:

· Identify the concepts that are to be explored with reference to the scope and objectives of the census;

· Determine the practical application of definitions and specify the details of the criteria to be adopted;

· Specify the criteria by which possible options are classified - the main categories are covered by international recommendations, but adaptation to specific national circumstances is always advisable.


All elements, concepts, definitions and criteria for application must appear in the questionnaire, which must be as self-explanatory as possible, with brief and simple instructions. Manuals are not designed to solve the problems that arise out of a badly designed questionnaire; they are aids to solving atypical situations, avoiding the emergence of biases during the census and providing training support to enumerators and supervisors. Situations should be illustrated by examples to make correct interpretation easier to achieve.

Questionnaire design is complex and can be difficult to understand. Census programming tends not to allow sufficient time and resources for questionnaire design, overlooking the fact that a poorly designed questionnaire defeats the whole purpose of this lengthy, expensive, multistage endeavour.

Given the diversity of country situations, there is no magic formula for questionnaire design. Those responsible for developing the questionnaire should review earlier experiences within both the national and the international contexts, especially in comparable countries. Field testing is the most important step in the process.

If the statistics are to reflect a gender perspective, questionnaires must:

· Record all data that identify smallholdings that correspond, on a one-to-one basis. to households;
· Identify the men and women farmers who actually head the holdings.
Framing the question

Prevailing prejudices can lead to erroneous answers and respondents can easily misunderstand inadequately formulated instructions and questions, so special attention must be paid to the wording. Interviewers may also have preconceptions about the various issues involved, including gender-related ones, and this can have a significant bias on the results.

Every possible precaution must be taken to ensure that the answer is not inherent to the question. Mistakes can be avoided by knowing where and when biases tend to arise and amending the question accordingly. Secondary "filter" questions may have to be added to recover data that might have been lost during the first question.

The preceding chapters looked at some of the major potential biases. Women tend to think of themselves as homemakers and to underestimate their economic activities, especially in the agricultural sector, where women farmers traditionally do unpaid work on the family holding. Such economic activities as planting, harvesting, husking and shelling are viewed as domestic tasks solely because the concept of "work" is identified, in women's minds, with paid work. Furthermore, in certain social settings, men may be reluctant to admit that their wives and daughters work because of the social stigma attached. There is an assumption that the man is the bread-winner, and even interviewers (especially male interviewers) often classify women as homemakers.

Keywords need to be carefully defined and used. Words that are used to express concepts can easily be misinterpreted. In questions such as: "Are you working as...?" or: "What is your main occupation?" the word "work" is often taken to mean "paid employment". Studies have demonstrated that the measure of women's work varies according to the keyword used to ask the question. The question: "Have you worked on a holding", for example, could be replaced by the question: "Have you been active in any of the following: rearing poultry, weeding, sowing, harvesting, etc.?".58

58 Dixon-Mueller, R. & Anker, R. 1988. Assessing women's economic contribution to development. Geneva, Switzerland, ILO. In 1988, a United Nations Statistical Office (UNSO) project in Pakistan revised the questionnaire for the national labour force survey. In addition to altering the format and sequence of questions on working conditions, a section was added that contained direct questions designed to capture the work done by women. Such specifications as the purpose of production (for market or home consumption) were included. This brought the procedure into line with the recommended 1988 ILO definition of economic activity. By simply modifying the question, without changing the definition, the rate of female participation rose from 13 percent to 47 percent.
Choosing the phraseology

The answers given on questionnaires reflect the phrasing of the questions. Certain terms lend themselves to misinterpretation or gender bias. There is a mental association between the expression "man hours", for example, and men, while the phrase "homemaker" almost inevitably conjures up the image of a woman. Even some of the most gender-neutral expressions, such as "head of household" are often thought of as representing a man, owing to the common perception of gender-assigned roles. With a bit of effort,59 gender-neutral terms such as "person hours" and "domestic tasks" can be found.

59 The Spanish language, in particular, is male-biased, and requires extra efforts in wording.
In countries where gender access to education and training is skewed, complicated phraseologies usually put women at a disadvantage, particularly in the countryside, where women have fewer opportunities for public contact, higher rates of illiteracy, less social life, little opportunity to travel, limited experience with media and restricted access to training. In this context, women will obviously find it difficult to understand unfamiliar terms. The use of technical terms is also inappropriate in dialogue with people who are unfamiliar with censuses and surveys.

In some countries, indigenous languages are more widespread than the official national language, especially in rural areas. Women are those most likely to be monolingual, making dialogue difficult when the interview is conducted in the official language. Census planners should take this into account and, if they lack the resources to have the census material translated into the vernacular tongue, they could recruit bilingual interviewers (who are trained to interpret the questions without introducing gender bias).

Identification coding in the questionnaire

As mentioned in Chapter 3, identification coding should identify the exact geographical location of each holding. Where two or more holdings are associated, it should be possible to correlate data from the household(s) with those from the corresponding holding(s), and vice versa. The identification list should include four features:

· Data that identify the exact geographical location of the holding with corresponding codings for the political and administrative division;

· A unique reference number for the holding, and additional digits for its components, if required;

· A one-digit space which will contain a number code defining whether or not the holding is associated with a household. (If only a specific number of units are to be analysed, these data can be filled in immediately by computer.) For instance, the number 1 could be used to designate holdings associated with at least one household; 2 for entrepreneurial holdings not operated in connection with the holders' households; and 3 for mixed operations (holdings operated as enterprises but organized in connection with the holders' households);

· A unique reference number that identifies the household. Different identification codings must be used for households and holdings, because they do not necessarily correspond.

The following is the coding that should be used when recording the various cases:
· The letter "a" stands for the holding;
· The letter "b" stands for the household;
· Consecutive numbers should be assigned to "a" and "b";
· The series should be restarted for each new geographical location;
· When recording a new holding, increase the number for "a" by 1;
· When recording a new household, increase the number for "b" by 1.
The ratio of households to holdings, and their correspondence, can thus be clearly established.
The following are some examples of identification coding.

a) Household operating more than one holding


Holding code

Type of holding

Household code

Geog. location

a

1 or 3

b

Geog. location

a + 1

1 or 3

b


b) Holding operated by one or more persons from different households


Holding code

Type of holding

Household code

Geog. location

a

1 or 3

b

Geog. location

a + 1

1 or 3

b + 1


c) One household with one holding operated by persons from a single household


Holding code

Type of holding

Household code

Geog. location

a

1 or 3

b


d) Holding organized as an enterprise, operated by a hired contractor, and with no corresponding household


Holding code

Type of holding

Household code

Geog. location

a

2

0


Identification of holder

Chapter 3 discusses the concept of holder and its limitations from the gender standpoint. The holder, whether male or female, is the person who makes the major decisions in the production unit. This responsibility may be shared by more than one person on the holding.

The major statistical constraints arise when surveying small units (which may be hard to identify as productive units) and identifying the persons responsible for the unit without under-representing women. The first step is to determine whether a household has at least one agricultural production unit - this is sometimes self-evident but in other cases it has to be clarified.

The reference period is that in which the agricultural production unit was active (i.e. not idle). This varies by country. In Latin America, for example, a six-month reference period is more than adequate. Even when seasonal migration occurs, some labour-intensive activity, such as harvesting or sowing is bound to have occurred during the previous six months. Mexico's National Survey of Rural Employment, for example, asks respondents which of the following they have done:

over the last six months, that is from -----------------------------------until-------------------------------------

Read every option and tick whichever of these options applied, knowing that you can tick more than one question

1. Did you work the land and/or participate in agricultural services?

2. Did you rear or care for livestock60?

3. Did you hunt or fish?

4. Did you gather plant products and/or forest products?

5. None of these.

60 The Mexico survey had added "for sale" to this question (while a preceding version was worded as: "Did you look after livestock and/or a kitchen garden for home consumption?"). The concepts of production for home consumption or sale can be combined under the new international recommendations, so the specification "for sale" has been dropped.
This type of question helps to verify whether or not a specific household corresponds to an agricultural holding (this can be adapted to fit the context). Where all household members answer all the questions in the negative, the interview ends there, but if at least one answers in the affirmative, there is assumed to be a holding and, therefore, the rest of the questionnaire should be filled out.

The following procedure is recommended to ensure that all those actually responsible for production are counted:

1. Having established that a household corresponds to a productive unit, all members of that household should be enumerated. The procedure followed can be the same as that used in household surveys, i.e. the head of the household is identified (as the person acknowledged as such by other household members)61 and all other household members are listed with their relationship to the head, sex and age.
61 Who may be the main earner, the main producer or the eldest person - subsequent data will clarify this. It is important to record all household members.
2. Following this a set of questions will be addressed to all persons who are above a specified age (normally those who are old enough to make decisions concerning the productive activity). The questions usually refer to specific decisions, e.g. Do you decide which crops to plant? Do you decide which animals to rear? and Do you decide what and when to sell?
Appropriate questions need to be formulated for each separate situation in order to clarify who is responsible, who makes important decisions, i.e. who is (are) the (male or female) holder(s). Having identified the holder(s), the interviewer will know who can best answer questions about production (age and sex data have already been entered).

It is very likely that more than one person responsible for the holding will have been identified by this point, because any individual can specialize in a particular type of work or participate in decision-making on various activities. For example, a woman may be responsible for the poultry and a man for the crops. In such cases all data must be recorded, and no detail omitted. Ideally, each person should report on his or her specialization.

4.2.3 Tabulation plan

A basic rule in taking censuses is that the tabulation plan needs to be prepared concurrently with the questionnaire, so that it can be ready before fieldwork begins. This is especially important when new ideas are being explored and alternative procedures adopted, as in the production of gender-sensitive statistics. Carefully prepared tabulation plans ensure that no important question will be omitted from the questionnaire, and that none of the questions asked will be irrelevant. A question that is badly phrased cannot be linked to a specific heading or item and will produce irrelevant answers, which is wasteful of resources. A pilot survey should include the basic tabulations to allow partial evaluation of the questionnaire.

It is not enough merely to know the percentage of female heads of holding and female-headed households; a gender-oriented analysis of all the major factors in access to resources and aspects of production needs more than this. In addition to the "sex" variable, the following need to be investigated: characteristics of holders (for holdings); characteristics of heads and members of household (for households); and characteristics of workers - relatives and hired hands - on the holding (for individuals). Sex is not a separate or additional variable, it is one of the essential characteristics for defining holdings in various contexts and for multiple-entry cross-tabulations.

4.2.4 Census publicity and respondents' cooperation

The national statistics bureau or other agency designated to conduct agricultural censuses in each country usually has its own policies for announcing and publicizing a census. The main issues are to alert the public and sensitize potential respondents about the importance of measuring the work of women accurately. The objective is to combat stereotyping and discrimination. Posters, television and radio announcements, press releases and other media can be harnessed to explain the objectives of the census and illustrate the breadth of women's contribution to agricultural production in both rural and urban areas. This is a job for professional communicators, who need to bear in mind and respect the concerns, needs and priorities of census designers.

Publicity and support-raising campaigns should be aware that women have less access to media, higher rates of illiteracy and less contact with the world outside their homes. The use of graphics has vast potential, for example, posters depicting images that women can identify with should be put up in places where women tend to congregate, such as bus terminals, markets, health clinics and schools. Radio programmes can be targeted at women. Grassroots organizations and women's groups can also help to get the message across.

4.2.5 Statistical unit and coverage

The recommended statistical unit for agricultural censuses is the agricultural holding. Ideally, the census should cover all holdings, both rural and urban. In the developing countries, it is essential that smallholdings are included because they represent the majority of production units. In addition, the majority of women farmers whose main activity is production for home consumption are on smallholdings. Another reason for including them is the existence of multiholdings in this subsector. Each member of a household who is responsible for operating a separate holding should be considered as a separate statistical unit, despite the difficulties that can arise in making this distinction. In this way, women's contribution is less likely to be omitted (see Chapter 3). All of the holdings that contributre to overall food output must be recorded.

A complete, up-to-date list of holdings provides an excellent frame for the census, giving all basic data on each agricultural holding. Unfortunately, such a list is very costly to compile and assumes that the kind of statistical and administrative systems normally found only in rich countries are in place and operating. In many Latin American countries (but not Cuba, which lists all holdings), only the large agricultural establishments may appear on the list.

Coverage errors are often the result of an inadequate census frame owing to an incomplete list of agricultural holdings, an out-of-date population survey or some other similar shortcoming. It is very difficult to measure the impact of these errors on data quality.

4.2.6 Designing the sample

When a list of holdings is not available, the population census is used to develop the sampling frame. The enumeration of smallholdings and subsistence units depends on having a well-prepared sampling frame. A good probability sample survey that can produce reliable measurements of the critical gender issues in the agricultural sector is wholly dependent on the use of a good sampling frame.

If neither holding lists nor population censuses are available, enumeration will have to be based on geographical areas or lists of villages and settlements. In this case, the enumerator will need to identify all the holdings in all localities, either by visiting each household in the area or by interviewing local authorities, who are assumed to know everyone in their area.

In countries where statistical systems are constrained by meagre resources or other problems, the census can be approached on the basis of a probability sample survey, which is used where there is no possibility of surveying an entire area in order to develop a sampling frame. In this case, designing the sample is a two-stage process:

Stage one: divide the national territory into geographical enumeration areas (segments) and effect an initial selection of these areas;

Stage two: make a complete survey of the selected areas in order to enumerate the holdings and select the units to be interviewed. Alternatively, subareas can be defined for a second selection, and only these used to draw up the list of production units covered by the questionnaire.


The sampling design and testing of the sample should be done by a sampling expert in accordance with the objectives established by the project design team.

The sampling design needs to give careful consideration to the factors that affect men's and women's activities, such as:

· Seasonal work. Agricultural work is intermittent, so samples can easily exclude holdings that operate on a seasonal basis. In order to cover seasonal work, most of which is done by women, it is recommended that a rotational sample be made on a monthly, quarterly or seasonal basis. When this is not feasible, useful data (although of poorer quality), can be obtained by asking retroactive questions.

· Geographical differences. Gender disparities can be looked at from a geographical point of view. Women and men usually work in different types of agricultural production, so some activities may be rare in certain areas and common in others. The sampling design may need to guarantee coverage of certain designated areas to guarantee the inclusion of all activities. This is a prerequisite for adequate sampling.

· Holding size. Because most smallholders are women, who are often dealt with superficially or even omitted from agricultural statistics, it may be necessary to design the sample specifically to guarantee that smallholdings are covered. Contemporary sampling techniques can ensure that the sample is statistically representative.

4.2.7 Staff recruitment

Agricultural statistics teams are usually made up of men only, either because few women have the necessary statistical expertise or because gender stereotyping tends to classify agriculture as a male domain. Female participation at the administrative, technical and field levels is essential, however. Even though women are more likely to be aware of gender issues, it is also vital that training and sensitization activities be addressed to them. Women may be more open to the gender approach because of their everyday experiences and observations.

Good enumerator-respondent interaction is crucial to the success of the census operation. Enumerators, both male and female, need to communicate easily with the men and women they are interviewing. They need to be aware of cultural traits that may influence the respondents' answers, and able to detect gender stereotyping that can give rise to bias. A male respondent may deliberately give the wrong answers to a woman enumerator, for example, as an act of refusal. Ideally, women interviewers will have the basic agricultural expertise that allows them to make light of a misleading initial response by turning it into a joke, thus earning the respect of the respondent and securing the right answer. Male interviewers must be extremely respectful, especially when interviewing women.

Population census-taking operations in many countries have demonstrated that the sex of the interviewer is not necessarily a determinant of correct answers from those interviewed. This variable has not been sufficiently evaluated in agricultural censuses but, in societies where there are cultural barriers to women talking with men, both male and female enumerators will obviously be needed.

4.2.8 Instructions and training for enumerators

Intensively trained enumerators and carefully prepared instruction manuals can do much to reduce gender bias in data collection. A good training programme should emphasize the objectives and goals of the census, and the fact that the quality of the census data depends primarily on the enumerators. This is essential for topics in which gender disparities traditionally arise. Enumerators should apply all the recommended criteria to avoid gender bias. They must also be specifically instructed on how to cope with gender stereotyping. Balanced examples, representing both men and women in different situations, must be included in the concepts and definitions section. The desire to highlight data on women can produce the negative effect of failing to report accurately on men's situations. As was pointed out in Chapter 2, the gender perspective aims at capturing and comprehending gender differentiation and gender interaction, and not just the situation of one of the two sexes.

The following is a recommended training frame for enumerators:62

62 Adapted from UN. 1988. Improving statistics and indicators on women using household surveys, UN DIESO, Series F, No. 48. New York.

1. Explain why it is important that the data collected mirror the true situations of men and women in society;

2. Train enumerators to use gender-neutral speech, avoiding such expressions as "man months" and gender-biased phrases and articles when referring to the head of household or holding;

3. Point out to enumerators the concepts and definitions that cause the most problems in gender-specific data (such as "status in employment", "economic occupation" and "head of family");

4. Ensure that enumerators have experience in interviewing persons of both sexes and are aware of the possible implications;

5. Provide manuals that show balanced examples of both men and women and that depict men and women performing non-traditional roles;

6. Discuss the problems and stereotypes that emerge in measuring women's occupations;

7. At training sessions, balance interviews with both sexes equally and discuss the differences between the two sets of results.

4.2.9 Instructions and training for supervisors

The supervisor's work, which is particularly intensive during the enumeration phase, is vital when detecting and correcting errors during the census-taking process. Gender stereotyping can emerge in the course of field activities, even when the main problems have been dealt with in questionnaire design, manuals and support materials. The supervisor must be able to correct biases and confused reporting from the initial interviews so as to avoid data contamination at later stages.

Supervisors also play a vital role in training enumerators, who must be encouraged to pay careful attention to gender considerations. Supervisors need to undergo intensive training, including sessions on gender and statistics, before they can properly instruct enumerators.

4.2.10 Choosing the respondent

Selecting the right respondent affects data quality in the following ways:

· If the respondent is not the person who is best informed about the operations of the holding, erroneous answers may be given.

· Men may be reluctant to admit that the women in their families work.

· Men may be uninformed about certain aspects of their wives' productive activities.

· Women may be unable to answer questions concerning the work their husbands do.

It is also important to consult manuals that illustrate interview techniques in order to get a clear picture of what must and must not be done. For example, the presence of others during the interview should be avoided, particularly when they are people who are not connected to the farm, because this can lead to inexact and misleading answers being given. Every possible effort must be made to ensure that each agricultural producer reports directly on his or her own agricultural activity, without interference from others.

4.2.11 The pilot census

Pilot testing can be decisive for assessing the validity of statistical tools and procedures. In addition to testing and checking the questionnaire, pilot testing can be helpful in training staff to gather field data, correcting the instruction manual and assessing training needs. The logistics of the operation are also sampled, providing a glimpse of all phases of operations and indicating any changes that are needed.

Enough time must be left between pilot testing and the census to allow analysis of the basic tabulations, evaluation of the operational aspects and detection of problems and errors in time to solve them.

For statistics that incorporate a gender perspective, sampling is particularly important in areas where the quality of the answers is subject to the influence of social and cultural factors. Common discrepancies between theory and fact can be identified by the interviewer. It is also crucial to pilot test the questions, phraseology and content of the questionnaire. Answers such as "none of the above" are more commonly given by women than by men. Major gender differentiation in the answers is an indication that something is wrong with the questionnaire.

4.2.12 Data processing

There are several crucial stages between field enumeration and publication of the finished product. These range from coding to preparation of the statistical tables. At all stages of the process, great care must be taken to ensure the incorporation of a gender perspective.

Coding

All the data collected should be coded numerically. Pre-coded answers are an option when such systems as multiple-choice questionnaires are being used. For open or semi-open questions, however, the full range of answers must be codified and classified. A gender perspective must be maintained in the adoption, revision and, if necessary, the reformulation of classifications. Any gender differentiation in the answers should be clearly indicated.

Great accuracy in the coding process is necessary to avoid altering the field data by, for example, excluding men or women from specific categories on the basis of personal convictions.

Verification and imputation

The percentage of total and partial non-responses is measured to assess the quality of the field data. Total non-response may be the result of a respondent's refusal to answer, geographical isolation (steep terrain, flooding or some such), the absence of the person to be interviewed or failure to identify the holding. The reasons must be listed meticulously, because each case has a bearing on the overall picture. Partial non-response may be owing to the absence of a respondent, so the interviewee is only able to give a partial response, or it may be the result of the respondent's deliberate concerning the data requested.

Verification of data processing and checking for consistency are important steps in the process of taking an agricultural census. The procedures for correcting inconsistent data or providing missing entries should only be applied when strictly necessary, based on quantitative and qualitative criteria determined in accordance with experience in the specific country. Great care must be taken to avoid erroneous imputations or changes arising from gender stereotyping.

There may be some systematic differentiation between non-respondents and those responding. In this case the existence of a possible gender differentiation should be verified.

Classification criteria

It is crucial to ensure that all the effort expended to produce carefully codified, gender-specific classifications is not subsequently nullified by grouping the data in such a way that major gender differences are lost at the tabulation stage.

Data are enhanced by being presented in detailed groups. General and specific systems are not incompatible. Major groupings can be constructed to maintain data compatibility and ensure that the data are comparable at both the national (historical series) and international levels; a far more detailed or specific analysis, such as gender analysis, can also be done at the same time.

4.2.13 Data presentation

The data presentation formats must be acceptable to and meet the needs of a wide range of users. Graphs and tables must be plain, simple and attractive. Only then can gender-specific statistics help to promote gender equality and encourage positive changes in women's situations.

Correct data interpretation is highly dependent on the selection of indicators and method of dissemination. Simply presented, well explained data can reach a wider audience.

Two fundamental questions need to be addressed when preparing tables and graphs:
· What is the most important information the reader can extract from this presentation?
· Have the tables and graphs been designed to meet users' needs?
It is essential that each table and graph provide specific data for users. The message will vary according to the way in which the data are presented. Care must therefore be taken to avoid the loss of vital data owing to unclear presentation.

Tables and graphs should not be overloaded. Every column and line in a table should be clearly headed and explanatory notes included where necessary. Clear data presentation has a direct bearing on content relevance.

A basic rule is that data on men and women should be presented in the same table. It is difficult to compare data that are listed separately, and this defeats the purpose of gender-specific statistics. If the data cannot all be fitted on to the same sheet, they can be presented by geographical area or time period on separate sheets.

It is recommended that all values be represented in absolute figures to give the user greater freedom of analysis. When values are calculated as percentages, the universe to which the percentage refers (100 percent of what?) must be given. Either the gender disaggregation of a given category can be shown, or else the percentage of women exhibiting a given characteristic with reference to all women (or men) in the survey. Both types of data are helpful in presenting the full panorama of gender statistics. It is important to decide which is the most appropriate presentation, including what technical information is needed for a correct reading and analysis of the table.

Lourdes Ferrán suggests the following:

· Observation of ILO recommendations on gender in all statistics concerning employment;
· Gender-specification of statistics on production and sale of agricultural products;
· Gender-disaggregation of data on remuneration, number of workers, type of work, technology, etc.;
· Review of classifications to ensure coverage of small-scale production.
Other indicators can be published, in addition to the multiple-entry tables in percentages or absolute numbers. The proportion of persons sharing a specific attribute with respect to the total population, or the proportion of active women compared with all women, or the ratio of consumers to producers within the household, by sex of the head of the family, are examples of some relevant statistics and indicators.

Modern data processing methods make it possible to explore information in greater depth than before, constructing subsets for categories such as women-headed households, which are then classified by variables such as marital status and age, and cross-tabulated with other major variables such as holding size and resource use.

Subsets can be constructed on the basis of social, cultural, ethnic and religious criteria when there are significant variations within a country. The gender variable can be one criterion for the definition of a subset or a constant variable for all tabulations - the choice depends on the specific analytical objective. One way of examining the informal subsistence agriculture sector, for instance, would be to classify groups in accordance with the type categories suggested in Chapter 3: viable; potentially viable; and non-viable. Another classification might be: self-sustaining; and dependent on external transfers.

Raw, unorganized data can be sent directly over the net to specialized users. To safeguard confidentiality, the individualized identification data can be masked to avoid recognition of enumerated units (especially disaggregated data based on unusual variables).

4.2.14 Dissemination

A great many field data are never analysed and never reach users. Many of the data on human resources in agricultural censuses only occasionally appear in print and are, therefore, rarely used.

The producers of statistical data should review data dissemination media, consulting representatives of users' groups to evaluate the demand for data and adapting their products and services to users' needs, thus fostering and contributing to the utilization of statistical data by academic and governmental institutions, NGOs and other users.

The dissemination of statistical data to official bodies, such as ministries, planners and other organizations responsible for the design and execution of policy, is extremely important because the data are used in the formulation of development plans and policies. Census data are also crucial for bringing about decisive changes in sustainable rural development and agriculture and for monitoring and evaluating these new developments.

The statistics and indicators that emerge from agricultural censuses are crucial to an understanding and analysis of aspects such as food production, poverty and access to resources, training and extension, which are of vital importance at the national level. These issues affect most of the population in many predominantly rural countries, and they are essential for subsistence. Despite this, official policy-makers still have limited access to the basic statistics they need to formulate policy, monitor changes and promote innovative measures.

The time-lag between the enumeration stage and the actual publication of the census results is one reason for official decision-making bodies' lack of interest in using statistical data. This interval can be substantially shortened by producing a pre-publication showing the preliminary, gender-disaggregated results, with the full and detailed version following at a later date.

A dissemination plan should take account of the following factors:

· Characteristics of the target groups;
· Users' needs and demand for statistics;
· Available human and financial resources;
· Most widespread means of communication in the target country.
If users participate in all stages of the census, they are kept informed as to the kind of data available for use in their specialist field. The following recommendations suggest ways of promoting the use of statistics, which can in turn highlight findings that challenge the prevailing prejudices about men's and women's roles:
· Distribute fliers, brochures and posters publicizing selected basic census data (and the related topics of highlighting and differentiating gender contributions) at meetings and discussions on agriculture.

· Present census data to policy-makers, planning bodies, NGOs, research institutions, gender advocates and the media.

· Publicize data in the form of simple posters and other attractively presented media such as radio announcements pointing out the contribution rural women make in agriculture, and ensure that the message reaches rural women.

· Organize workshops and seminars to train users in how to handle census data.

· Publish the most interesting census findings in the press, government and NGO bulletins and other publications.

· Organize television and radio talks on the census findings, ideally with the participation of well-known local personalities.

· Design graphic displays of census data for use at conferences, workshops and other official gatherings.

Once the preliminary report has been published, further reports of the census will follow in the form of volumes devoted to specific topics and/or geographical regions. Census data can also be published electronically, on floppy disc and/or CD-ROM, to meet the needs of more specialized users such as statisticians and university professors. A special publication should be planned to cover gender-related statistics, so as to ensure that gender considerations emerging from the data and data analysis reach the widest possible audience. The publication should be attractively presented, easy to handle, highly readable, illustrated with simple graphs and tables and offer brief explanations and definitions of the indicators presented.

People need to be made aware of women's contribution to agricultural production, rural development and general welfare. Informative posters and radio and television programmes designed to enliven the cold statistical facts are a good way of promoting gender equity.


Previous Page Top of Page Next Page