1. Mapping & preparation of frames and Sample design
Technical review meeting on World Programme for the Census of Agriculture 2020
Volume 2 – Operational guidelines on implementing census of agriculture Rome, Italy
30-31 January 2017
Chapters 12 and13:
Mapping & preparation of frames and Sample design
Item 4
Oleg Cara
Consultant, Agricultural Censuses and Surveys Team
ESS FAO
Naman Keita

2. CONTENTS Mapping Purpose of maps used for Agricultural Census (AC)
Cartographic work and types of maps
Preparation of frames for AC
Types of frames and sources of information
Some guidelines on building frames
Frame requirements for various census modalities
Use of sampling for census enumeration
Main types and choice of sample design
Advantages & disadvantages of complete enumeration and sample enumeration censuses
Discussion points

3. Purpose of maps Census planning and frame preparation
Provide the cartographic base for planning (including budget estimation of) the census.
For a sample enumeration can be used to improve sample design by providing information for stratification and other auxiliary information.
Organization, conduct and supervision of field data collection
Maps (or sketches) are used to organize the field work.
GIS provides a computer-based design of EAs and significant automation of map production tasks.
The use of maps should be an essential part of the enumerators' training.
Presentation and analysis of census results
Maps can be used to relate statistical data with the corresponding geographic area, facilitate the understanding of statistics and assure a more extended and appropriate use of data.
Maps (or sketches) and/or photos for identifying EAs are used in the field by enumerators. These are large-scale field maps (1/10,000, 1/5,000 or larger), which should help the enumerator locate the agricultural holders or holdings' addresses to assure complete coverage of areas without omission and duplication.

4. Cartographic work and types of maps
The cartographic work should be conducted during the early stages of preparatory census activities.
It should start with the inventory of available geospatial information and maps and the evaluation of their suitability for AC purposes. B
A large proportion of the cartographic preparations for an AC consists of delineating and identifying the enumerators' areas (EAs) of work.
In many countries EA maps are prepared as part of the cartographic work of the Population and Housing Census (PHC). In some countries the cartographic work is conducted for both PHC and AC (e.g. Brazil).
Adapting, revising and updating available maps to the census requirements.
Types of maps: topographic sheets, other government maps, satellite images, aerial photography, communication maps, land-use maps, economic maps, city and tourist maps, EA maps from PHC.
One of the first activities in the planning of an agricultural census or survey should to explore in detail the feasibility of using existing geospatial information and particularly hard copy maps

5. Types of frames
The List Frame is a list of agricultural holdings and/or households. However, when the list of agricultural holdings and/or households is missing, the list of EAs could serve as a starting point for the development of a census frame.
The Area Frame is a set of land elements, which may be either points or segments of land.
The Multiple Frame, in which part of the population is covered by a list frame (e.g. special holdings) and the remainder (e.g. other holdings) by an area frame.
There are three main types of frames that can be used for an agricultural census:

6. Sources of information for the construction of an AC frame
The statistical data and maps of the EAs of the most recent PHC and a list of EAs along with their number of households/housing units.
The statistical data and maps of the EAs of the latest AC including a list of EAs with their number of holdings.
Statistical farm registers, cadastral and other cartographic materials, other statistical and administrative data sources, which include listings of holdings and/or holders' addresses or their approximate location.
It might be necessary, due to lack of reliable information, to prepare the census frame by conducting a listing operation (prior to starting the enumeration of the holdings).

7. Some guidelines on building frames for AC
To be able to enumerate all holdings a preliminary frame is indispensable. It is made of a set of physical materials (cartographic maps, EA maps, etc.).
When PHC does not include specific questions on agriculture, the economic activity status must be considered together with occupation and industry in order to provide an approximation for households involved in own-account agricultural production.
The minimum data items recommended to be included in the PHC to better identify such households are: (i) Whether the household is engaged in any form of own-account agriculture production (ii) measure of farm size, e.g. area of land & the number of livestock.
Use of farm registers & other statistical/administrative sources to build a frame.
Preparation of sampling frames corresponding to each sample selection stage, including detailed mapping to support the field data collection (when sampling is used). The frame requirement is directly related to the type of sampling.
Using new technology for building frames, e.g.: GPS, GIS, Remote Sensing.
When complete enumeration is implemented, this set of material will be used to plan and organize the field data collection.
In practice very few countries have a complete list of all holdings available before conducting a complete enumeration census.
Therefore, to be able to enumerate all holdings a preliminary frame is indispensable which is made of a set of physical materials (cartographic maps, EA maps with number of households or agricultural holdings, list of farms with addresses from registries, directories, etc.).
When complete enumeration is implemented, this set of material providing a preliminary frame will be enough to plan and organize the field data collection.
 When sampling is involved the frame requirement is directly related to the type of sampling as discussed in Chapter 13.

8. Frame requirements for various census modalities (1)
Frame requirement for a classical census
An initial frame of holdings in the household sector and in the non- household sector is needed.
The total area of the country is unambiguously divided into non- overlapping subdivisions - EAs in such a way so that the enumeration work in each area can be handled by a single enumerator during the census data collection period.
Coordination of the construction of the EAs of the population and agricultural censuses.
Use of available cartographic materials. It might be necessary (due to lack of good maps) to prepare sketches for the EAs.
Conduct a listing operation due to lack of reliable information, if needed.
Establish a sampling frame (when sampling is used).
In general, the EAs are defined and delineated in such a way so that the enumeration work in each area can be handled by a single enumerator during the census data collection period.

9. Frame requirements for various census modalities (2)
Frame requirement for a modular census
As for the classical census, an initial frame of holdings in the household sector (or EAs) and in the non-household sector is needed to conduct the complete enumeration for the core module (CM).
The CM should provide the necessary information for building appropriate sampling frame(s) for the subsequent supplementary module(s), SM(s).
A built-in process should be developed for generating sampling frame(s) for SM(s), according to the design(s) decided.
Frame requirement for an integrated census and survey modality
As for the classical census, an initial frame is needed for organizing the complete enumeration for the CM.
The frame requirements for rotating modules are similar to those for SM of the modular census.
Frame requirement: modality based on use of registers as a source of AC data
The frame requirements for field operation component are identical to those discussed in the modalities above.
The points discussed under frame requirements for core module are similar to those for complete enumeration for classical census.

10. Use of sampling for census enumeration
In a classical census sampling may be applied when:
using the short-long questionnaire concept (the short questionnaire is administered to all target population of holdings, while the long questionnaire is administered only to a sample of such holdings); or
conducting a sample-based census as a single one-off operation.
In the modular approach sampling is needed for selection of holdings to apply the supplementary module(s).
When the census is part of an integrated census/survey modality rotating modules are conducted on a sample basis.
Use of registers as a source of census data could also be combined with filed enumeration on a sample basis.
using the short-long questionnaire concept (the short questionnaire is administered to all target population of agricultural holdings, while the long questionnaire is administered only to a sample of such holdings)
while the long questionnaire (with more detailed information)
conducting a sample-based census as a single one-off operation (a large sample of holdings is selected for census enumeration conducted in one operation).
All modalities of census taking may apply the combination of sample enumeration with complete enumeration
Uses of sampling at other census stages: In pilot censuses: to test census instruments and procedures
For quality checks during field operations
In PES
For the preparation of census preliminary results

11. Sampling techniques
Sampling techniques: simple random sampling (SRS), systematic sampling (SYS), stratified sampling (STR), sampling with probability proportional to size (PPS), multivariate probability proportional to size (MPPS), cluster sampling, etc.
One-stage (element) sampling:
a fairly complete and up-to-date frame is required.
is not always feasible when conducting ACs, especially in countries with a not well established system of agricultural surveys.
Cluster sampling:
a sampling frame at the element level is not needed for the whole population.
cost efficiency, that is, the relatively low cost per sample element.
in practice, clusters tend to be internally homogeneous, and this intra-cluster homogeneity increases standard errors and thus decreases statistical efficiency.
The multi-stage sampling procedure, which involves multiple sampling at different stages, aimed at achieving maximal accuracy of statistics for allowed costs, is largely employed for ACs and surveys.
Use of sampling for census enumeration
In a classical census sampling may be applied when:
using the short-long questionnaire concept (the short questionnaire is administered to all target population of holdings, while the long questionnaire is administered only to a sample of such holdings); or
conducting a sample-based census as a single one-off operation.
In the modular approach sampling is needed for selection of holdings to apply the supplementary module(s).
When the census is part of an integrated census/survey modality rotating modules are conducted on a sample basis.
Use of registers as a source of census data could also be combined with filed enumeration on a sample basis.
Advances in sampling theory, such as calibration, ratio and regression estimation may also be used.
such techniques as PPS and STR have strong requirements for prior auxiliary information for each element in the population.
Cluster sampling:
a sampling frame at the element level is not needed for the whole population, but cluster-level frames are often accessible, for example when they are defined administratively or geographically, e.g. for districts, villages, enumeration areas (EAs), etc.
cost efficiency, that is, the relatively low cost per sample element, because a lower costs of both listing and data collection (locating)
in practice, clusters tend to be internally homogeneous, and this intra-cluster homogeneity increases standard errors and thus decreases statistical efficiency
therefore, when constructing a sample design for agricultural censuses and surveys, more clusters will need to be selected and then subsampled using measures of size.
A variety of sampling techniques can be considered for the construction of probability sample designs for agricultural censuses and surveys, including: simple random sampling (SRS), systematic sampling (SYS), stratified sampling (STR), sampling with probability proportional to size (PPS), multivariate probability proportional to size (MPPS), cluster sampling, multi-stage sampling, etc. Advances in sampling theory, such as calibration, ratio and regression estimation may also be used to improve the reliability of census data collected by sample enumeration.
 For one-stage (element) sampling (using such techniques as SRS, SYS, PPS or STR) to be cost- efficient two key things are needed: first, a fairly complete and up-to-date frame (listing of the elements) of the target population. Second, locating the elements and collecting the data must be feasible and economical. Besides, such techniques as PPS and STR have strong requirements for prior auxiliary information for each element in the population. Therefore, element sampling is not always feasible when conducting agricultural censuses, especially in countries with a not well established system of agricultural surveys. For a given sample size, it can be shown that sampling errors will be smaller in a single-stage sample design than if clustering is applied. However, the former involves a larger frame development and data collection costs because the sample in that case is more widely distributed than, for example, if two- stage sample design is used.
 Cluster sampling:
Typically, the sampling design for an agricultural census consists of a combination of various sample selection techniques. A manageable sampling design often involves clustering and several stages of sampling.
Global Strategy (2015): Handbook on Master Sampling Frames for Agricultural Statistics, p.39
 A manageable sampling design often involves clustering and several stages of sampling.

12. Main types of sample designs for census enumeration
Sample designs can be classified according to the type of sampling frame used, i.e. based on:
List frames
Area frames and
Multiple frames.
The main difference between these types of sample designs is whether the final stage of sample selection is a listing of holdings or households, or is land based, or if a combination of a sample of area elements with a list sample is applied to obtain the census estimates.
Handbook on Master Sampling Frames for Agricultural Statistics (Global Strategy, 2015).
The reader can find detailed explanations of each type of frames and related sample designs in the publication Handbook on Master Sampling Frames for Agricultural Statistics (Global Strategy, 2015)

13. List sample designs (1)
The most commonly used sampling procedures in ACs.
List sample designs often include some strata of “special holdings” that are completely enumerated, or have a high sampling fraction.
List sample designs for classical censuses
Element sampling (using SRS, SYS, STR, PPS, etc.).
Single-stage cluster.
Multi-stage sampling.
Country examples:
Croatia : One-stage sample design based on STR
Suriname: Combination of complete enumeration with cluster sample design
The most commonly used sampling procedures in agricultural censuses.
The sampling frame is generally a list of agricultural holdings or households (e.g. when a wider census is conducted), which represent the ultimate sampling units.
Both one-stage and multi-stage list sample designs can be applied.
In the former case the units of interest for the census are selected directly from a list frame. In the latter case, the frame for the first stage of sampling is a complete listing of administrative units or geographical areas as primary sampling units (PSUs).
The sampling frame used at the final stage is the list of holdings or households, developed within the selected clusters in one or more stages of sampling. Despite the fact, that PSUs (in some cases SSUs as well) applied in such a sampling procedure are area units, this type of sample design, for the purpose of this publication, also refers to a list sample design, because the ultimate sampling units (i.e. holdings or households), are selected from a list sampling frame.
A sampling frame should include relevant auxiliary information such as measures of farm size, e.g.: total area of the holding, area by main land use types, number of land parcels, number of household members, number of livestock of the main types. This auxiliary information is very useful in the construction of an efficient sampling design (facilitating STR, PPS techniques, etc.) and further, at the estimation stage.

14. List sample designs (2) List sample designs for modular census
A multistage sampling for supplementary modules:
Suits very well in-depth surveys when the application of detailed and time consuming questionnaires is needed.
It is commonly applied, where EAs are the PSUs and holdings are the SSUs.
Depending on the availability of information about PSUs, SSUs, etc., the use of STR or selection with PPS would be appropriate to improve the efficiency of the sample design.
Country examples:
Togo: two-stage sample design for supplementary modules
Burkina Faso: two-stage probability sample design
India: combination of one-stage and two-stage sampling designs.

15. Sample designs based on area frames and multiple frames
Well suited for estimating parameters related to land areas such as cropland and can be used to evaluate the quality of the data collection.
Has major limitations for estimating other parameters required by a census (such as on livestock, demographic and social characteristics, etc.):
Many important agricultural characteristics have a skewed distribution;
The sampling of area units alone increases the sampling variance. When land of a large holding belongs to a sampled segment, the expansion factor can lead to inaccurate estimators for some variables. For instance, farms that have livestock and do not have their own land are difficult to pick with an area frame.
Sample designs based on multiple frames
The joint use of area and list frames: a list of relatively few “special holdings” complement the larger population of smaller farms covered by the area frame.
Country examples of agricultural censuses conducted in Fiji and Puerto Rico.
The main types of land elements in an area frame: i) Segments with natural (or physical) boundaries ; ii) Segments with a regular geometric shape, such as squares; iii) Points.
The main types of land elements in an area frame: i) Segments with natural (or physical) boundaries; ii) Segments with a regular geometric shape, such as squares; iii) Points.
There are basically three methods of defining a reporting unit when the area sample design is used: i) Closed Segment; ii) Open Segment; iii) Weighted Segment Estimator.
Well suited for estimating parameters related to land areas such as cropland (especially when direct observation is applied) and can be used to evaluate the quality of the data collection.
A major problem of using solely an area sample design is that many important agricultural characteristics have a skewed distribution, concentrating a significant proportion of the total estimate in a small proportion of the holdings. As a consequence, the sampling of area units alone increases the sampling variance. When land of a large holding belongs to a sampled segment, the expansion factor can lead to inaccurate estimators for some variables. For instance, farms that have livestock and do not have their own land are difficult to pick with an area frame.

16. Choice of sample design
Achieving maximal accuracy for available total budget (resources), or minimal costs for required accuracy/ level of aggregation for census data: two ways of stating the aims of efficiency.
Resources include the availability of trained and experienced personnel in sampling techniques, financial and information resources.
The sample design should be simple enough to operate in the field.
List sample designs versus area sample designs.
Multiple frame designs versus area sample designs.
Multiple frame sample designs versus list sample designs (pros and cons determined by strengths and weaknesses of the area component of the former):
Precision of the estimates.
Basis for a crop yield estimation.
Complexity of implementation.
Mapping requirements.
Proximity of the holder or respondent to the holding.
Distinguishing characteristics.
Local conditions, resources and requirements should be carefully considered .
Achieving maximal accuracy for available total budget, or minimal costs for required accuracy are two ways of stating the aims of efficiency, depending on which of the two (budget or accuracy) is fixed or less flexible.
The sample design should also be simple enough to operate in the field with the help of available personnel.

17. Complete enumeration versus sample enumeration census
Advantages
Reliable census results for the smallest administrative and geographic units and on rare events (such as crops/livestock types)
Provides a reliable frame for the organization of subsequent regular infra-annual and annual sample surveys. In terms of frames, it is much less demanding in respect of the holdings’ characteristics
Requires fewer highly qualified statistical personnel with expert knowledge of sampling methods than a census conducted on a sample basis. This is particularly important in countries with limited technical expertise
Aggregating data from a complete enumeration is straightforward and does not involve statistical estimations.
Is generally less costly that a complete enumeration
Contributes to decrease the overall response burden
Requires a smaller number of enumerators and supervisors than a census conducted by complete enumeration. Consequently, the quality of data can be expected to be much better because of the employment of better trained enumerators and supervisors and better quality control
Require less processing capacity and the results are usually available sooner.

18. Complete enumeration versus sample enumeration census
Disadvantages
High cost and administrative complexity
High overall response burden
Requires a very large number of field staff. As a result:
candidates with the desired qualifications might not be available in the required number
the standard might be lowered
adequate training of a large number of field census staff in a short period of time is also challenging, with a consequent effect on the quality of data
The amount of data to be processed is very large. The results may be considerably delayed if not sufficient data processing capacities are in place.
The amount of subnational data and cross-tabulations that can be produced is limited
Cannot provide accurate information on events that occur infrequently
May not ensure an adequate or complete frame for subsequent agricultural surveys
Requires a reliable sampling frame
Requires personnel who are well trained in sampling methods and analysis
Analyzing the data from a sample enumeration requires the use of more complicated techniques.

19. Discussion points Relevance of:
Sections on: mapping, frames, sample designs, advantages and disadvantages of complete enumeration and sample enumeration censuses.
Boxes and country examples.

20. THANK YOU