1. Regional Roundtable on
World Programme for the Census of Agriculture 2020
Nairobi, Kenya
18th - 22nd September 2017
Jairo Castano
Senior Statistician
Leader, Agricultural Census and Survey Team
FAO Statistics Division
Methodological Considerations
Technical Session 4
1
2. 2
Contents
• Background
• Overview of census modalities:
description, implementation steps,
advantages and limitations:
1. Classical approach
2. Modular approach
3. Integrated census/survey modality
4. Use of registers as a source of census
data
• Country examples
3. Background
The WCA 2020 broadened the approaches
introduced in WCA 2010, acknowledging that the
census of agriculture can be conducted in different
ways, using four main modalities:
Classical approach
Modular approach
Integrated census and survey modality
Use of registers as a source of census data
Main aim is to help countries to implement a census
in the most efficient way, taking into account
countries’ particular conditions. 3
4. Overview of census modalities
i) The classical approach: a census conducted in a single one-off
operation (usually by complete enumeration) comprising the
universe of agricultural holdings.
ii) Modular approach: comprises: a) a core module undertaken by
complete enumeration including all relevant frame items and
eventually other items; and b) supplementary modules conducted
using sample enumeration.
iii) The integrated census and survey modality: integrates a
multi-year programme of censuses and surveys. One option is
AGRIS, a modular survey programme which has to be
articulated with the agricultural census programme and
conducted on an annual basis between two censuses.
iv) Use of registers as a source of census data: registers and
other administrative sources are used as a source of census data in
combination with field data collection. 4
5. Comparison of census modalities
Classical Modular Integrated
census/survey
Use of registers
Enumeration
phases
One-off field
operation
Multiple phases:
a) core module
b) supplementary
module(s)
Multiple phases:
a) census core module
b) rotating thematic
modules
One or more field
operations and use
of admin sources
Coverage of
items
All census
items collected
in one-off
operation
Core module
includes items
required at the
lowest geo/admin
level, and those to
establish frames for
supplementary
module(s)
Census core module
includes items
required at the lowest
geographic/admin
level, and those to
establish frames for
rotating thematic
module(s)
Census items are
collected through
field operation and
use of administrative
sources.
Enumeration
method
Complete
enumeration
exclusively or
combined with
sample
enumeration
Complete
enumeration for
core module and
sampling for
supplementary
modules(s)
Complete enumeration
for census core module
and sampling
for rotating thematic
modules
Complete
enumeration
exclusively or
combined with
sample enumeration
5
6. Alternative census modalities
• The classical approach is the most extensively used census
modality. However, more and more countries would use
alternative census modalities in the near future.
• There are important reasons for using other census
modalities:
i. budget limitation for census;
ii. need to produce more frequent and timely agricultural
statistics;
iii. fast growing digital and mobile technology;
iv. increasing availability and access to data from
administrative sources and technical capacities to
handle such data;
v. reluctance of some population groups to participate in
the census and need to reduce respondent burden. 6
7. Some common issues for any
census modality
Identify data content of the census:
The minimum requirement for a census, is to include
all essential items, in order to enable
national/international comparison and frame items for
census modules or/and follow up surveys.
The final list of census items should be established in
consultation with main stakeholders depending on
country’s requirements, availability of reliable
administrative and other data sources, financial and
other resources.
In countries with well-developed registers, the use of
administrative data sources (ADS) to cover census
data items should be considered. 7
8. Some common issues for any census
modality (cont’d.)
The frame for the census must be carefully
established to ensure that all holdings are
covered with no omissions or duplications.
Use of sample enumeration
efficiency considerations (precision versus costs);
desired level of aggregation for census data;
use of the census as a frame for ongoing sample
surveys;
data content of the census;
capacity to deal with sampling methods and
subsequent statistical analysis based on samples.
8
9. Some common issues for any
census modality (cont’d.)
Use of thresholds
◦ In many countries, a minimum size limit
is adopted for holdings included in the
census.
◦ The rationale for this could be:
in the country there is a large number of
very small holdings which make a
marginal contribution to total agricultural
production; and
their inclusion in the census greatly
increases the workload and census budget.
9
10. 1. The classical approach
• It is conducted as a single one-off operation in
which all the census information is recorded. It
also includes the short-long questionnaire
concept.
• All items collected at the lowest geo/admin level.
• Can be conducted by complete enumeration,
sample enumeration, or by a combination of both.
• It is appropriate, for instance, when countries have
an integrated census/survey programme or wishing
to collect census items at the lowest geo/admin
level. 10
11. 1. The classical approach: types of
questionnaires
• Single questionnaire: administered to all agricultural holdings covered by
the census without regard of their type. It is easy to apply in the field.
• Short-long questionnaire: The short questionnaire is administered to all
holdings on a complete enumeration basis (to collect basic info), while
the long questionnaire is administered only to:
holdings identified according to certain criteria (to collect more detailed
info), such as being above an established threshold or belonging to a
particular segment of the population;
a sample of holdings.
• Other types of questionnaires - specific census questionnaires:
to fit different segments of the target population (such as household and
non-household based holdings); or
for different provinces when these differ considerably in cropping and
livestock systems, and in agricultural practices. 11
12. 1. Classical approach: advantages
• Snapshot of the entire target population at a specified period.
Comprehensive data sets at the lowest admin/geo level.
• Data can be produced at lowest admin/geo levels with no
sampling error. Tabulations can be done in line with high user’s
requirements, including data for small administrative units and
information on rare events, such as emerging crops, rare crops
and types of livestock.*
• Complete enumeration is much less demanding in respect of the
characteristics contained in the frame than the sample-based
census.
• A good basis for building up a statistical farm register and an
exhaustive sampling frame for subsequent regular agricultural
surveys.*
* When census is conducted by complete enumeration
12
13. 1. Classical approach: limitations*
• Cost and administrative complexity.
• It implies a high burden on respondents.
• Risk to overburden the census questionnaire because
of the high pressure from some policy makers or other
stakeholders to include detailed items to collect data
at the lowest administrative level.
• Logistics: very large number of enumerators and
supervisors required.
• The amount of data to be processed is larger.
* Mainly when census is conducted by complete enumeration 13
14. 2. Modular census
• Has a clearly distinguishable core module (on a
complete enumeration basis) and one or more
supplementary modules (on a sample basis).
• An essential condition: data from the core module
used as frame for the supplementary module(s). The
census using the short-long questionnaire in one
operation is not considered as modular census (no
frame use).
• A ‘module’ is defined as a group of data items to be
collected on a specific target population (e.g. holdings
with livestock).
• All essential items should be covered by the core and
supplementary module(s). 14
15. 2. Modular census (cont’d.)
• Core module (CM) must include the items required at the
lowest geographic or administrative levels, and/or needed
to establish sampling frames for the supplementary
module(s).
• CM should include all frame items. It should also cover
rare events (unusual crops or livestock), which would not
be possible to estimate from SM(s) because of high
sampling errors.
• SM(s) use the frame generated by the core module to target
specific populations and should include the rest of the
agreed census items that are not included in the core
module.
15
16. 2. Modular census modality: ways and
steps for implementation
Two ways of implementing the modular approach:
i. implementing core and supplementary modules
separately (common way)
ii. implementing the core module and the supplementary
module(s) as part of a single data collection operation
(difficult in practice).
Sources of frame data for supplementary modules:
a) The census core module;
b) An agricultural module/section of the population
census;
c) Administrative registers.
16
17. 2. Modular census: advantages
• The combination of core’s and SM’s items allows the
modular census to produce a wider and country-specific
range of data, for which small area estimates are not so
important.
• More effective use of available budget to collect country
relevant information.
• Allows a focused and more detailed training of field
personnel.
• Countries with a not well established system of
agricultural surveys and limited budget may find the
modular approach as a logical first step towards the
creation of a system of integrated agricultural censuses
and surveys.
17
18. 2. Modular census: limitations
• Risk of having a CM with too many items (due to pressure from
stakeholders).
• The availability of well trained professional staff in sampling could be a
challenge.
• There are limitations in terms of cross-tabulation between variables in
the core and in the supplementary modules (SMs) or between variables
in different SMs not conducted jointly.
• If the time lag between the implementation of the CM and SMs is too
long, the benefit of having a good frame from the CM disappears. A
frame update would be needed adding to the cost.
• If SMs are undertaken at the same time as the CM there exists the risk of
loss of data quality if not properly organized.
• It could be difficult to mobilize additional funds for SMs.
• In the absence of good planning, the conducting of the CM and
publication of results can jeopardize a better preparation of SMs. 18
19. 3. Integrated Census/survey modality
• This new modality features a census core module (to be
conducted on a complete enumeration basis) and a number
of several rotating thematic modules (to be conducted
annually or periodically on sample basis over a 10-year
period).
• An example of a survey programme with rotating thematic
modules is the new Agricultural Integrated Survey
programme (AGRIS).
• All essential items should be covered by the census core
and rotating thematic modules.
• The census core module should mainly provide frame data
needed to implement rotating thematic modules.
19
20. 3. What is AGRIS?
• The AGRIS is a modular survey program which is to be
articulated with the agricultural census programme and
conducted on an annual basis between two censuses.
• It consists of one annual production module (crop and
livestock production) and four rotating modules:
‘economy’, ‘labour force’, ‘machinery and equipment’,
and ‘production methods and environment’. Additional
modules can be added as needed.
• In the integrated census and survey modality, AGRIS
annual-production and rotating modules are
synchronized with the agricultural census core module
and operate over a 10-year cycle. 20
21. 3. Integrated census/survey:
recommended modules flow
Years 0 1 2 3 4 5 6 7 8 9 10
Agricultural Census core module (•) and inter-
census survey (o)
•
o •
AGRIS
Production
Module
Holding roster (ID & socio-demo) • • • • • • • • • •
Crop/livestock production • • • • • • • • • •
Key thematic issues • • • • • • • • • •
Rot. Module 1 Economy • • • •
Rot. Module 2 Labour • •
Rot. Module 3
Machinary, Equiment, Assets and
Decisions
•
Rot. Module 4 Production Methods and Environment • • 21
22. 3. Integrated census/survey modality:
advantages
• Effective use of available budget to collect
country relevant information on an annual basis.
• More detailed information available on topics of
interest.
• Focused training of field staff.
• Wider set of census items.
• Allows the establishment of a system of integrated
agricultural censuses and surveys.
22
23. 3. Integrated census/survey modality:
limitations
• The risk of expanding too much the census core
module resulting in high cost which will reduce the
relative benefits of this modality.
• Conducting the census core module and AGRIS annual
production at the same time (when relevant) requires
good survey planning and sampling capacity.
• The limitations in terms of cross-tabulation between
variables in the production and in the rotating thematic
modules or between variables in different rotating
thematic modules not conducted jointly.
• Difference in the reference period for different census
items collected in the census core and the items in the
rotating thematic modules. 23
24. 4. Use of registers as a source of
census data
• A meaningful part of the census items for the entire
population of the CA (or for part of it) comes from
existing admin sources created for non-statistical purposes.
• The data could come from one or several administrative
sources. Usually these are used in combination with field
data collection.
• Thus the definition of the population of interest and the
data collection protocols commonly is out of the control of
the census agency.
• As the objects already exist in the data source, a selection
need to be made of objects (holders or holdings) and
variables that are relevant to the census. 24
25. 4. Use of registers: Quality considerations
Some information in administrative sources could be of low importance for
administrative purposes which may result in lower statistical quality.
The purpose and method used to collect data should also be taken into account
to detect possible systematic errors or systematic bias in the register.
The following quality aspects need to be carefully considered:
Relevance (content) of the administrative data sources.
Accuracy: an admin source is of good quality if a large proportion of the variables
required for the census exists in the register and data are reliable.
Data coverage: one important accuracy category: under-coverage or over-coverage of
units, and misclassification of items.
Unreported events: data related to the reference period are not available in the source
referring to an event from which parameters are to be derived: e.g. births, deaths or
loss, sales, etc. of livestock.
The information from administrative sources has to be coherent and
comparable with data from other sources and over time.
The timeliness dimension (difference between the reference period and the
availability of the results) when using administrative data source.
Accessibility in terms of physical access to the information in the admin
source. 25
26. 4. Use of registers: Ways of using
1. The starting point (like in all census modalities) is to define the data content of
the census in terms of data items to be covered.
2. Next step is to decide on the way to use the admin source. Possible ways or
approaches are:
Split data approach: admin sources are used to provide some of the
variables for all of the holdings.
Split population approach: Data from admin sources are used for some
holdings where these data are of sufficient quality, and other statistical
sources are used for the remainder of the holdings.
Combining the two approaches: administrative sources replace the field
data collection for some of the variables for part of the holdings.
Pre-filling answers in the questionnaires which are to be checked by
respondents during the census.
Totally replacing the census data collection, on all census characteristics
and for all units in the census with data coming from one or more admin
sources (no common).
26
27. 4. Use of registers: Implementation
1. Assess the usability of the source/register and quality (see
before);
2. Assess legal basis (to access and use); consultations with
owners (concepts, definitions, classifications used, reference
period, coverage); public support (to statistical use).
3. Design a government-wide project to outline responsibilities
and the necessary work:
Make an inventory of sources, usability/quality.
Develop the content for the census.
Decide items to be produced from what kind of data
sources.
Determine steps to link datasets.
Develop communication strategy for the use of registers.
Plan the budget for the particular preparatory work related
to the use of registers for the AC.
27
28. 4. Use of registers: Advantages
• Reduction of cost of census data production.
• Reduction of the burden on respondents.
• By combining the data collected in the field with the data from
administrative sources, new derived variables can be created.
• It allows compilation and publication more frequently and faster.
• Non-response rate can be either significantly reduced or
eliminated (when data on the entire target census population are
included in the administrative sources).
• It could significantly improve the quality of the source and leads
to a substantial harmonization of certain information between
different institutions.
• As a result of more efficient and faster operation the public
perception of statistics may become more favourable.
28
29. 4. Use of registers: limitations
• Linkage of datasets are difficult/impossible if legal background is inadequate.
• It could be very difficult to establish a good cooperation with register owners.
• The cost for the access to the admin data could be too high.
• When there is different population coverage, admin sources can be used for
pre-filling the questionnaires only on common population while other units
need to be enumerated.
• Incoherence of concepts, definitions, classification and reference periods could
hamper the use of admin sources.
• Problems related to linking data from various data sources.
• Problems on quality or stability may arise due to political changes.
• Timeliness and punctuality.
• If an admin source is abolished, it is difficult to provide comparable statistical
data series.
• Substantive or technical changes in the admin sources may not be detected
immediately.
• Comparability over time is strongly influenced by the change in the level of
coverage in the different years and can give misleading results. 29
30. Country examples
•The 2011/2012 Census of Agriculture was a sample survey of
approximately 15% of rural households in the 2003 Census of
Population.
•The sampling design was a stratified two-stage design with
districts forming the strata and the first-stage units being the EAs.
First stage units were selected systematically in proportion to the
2003 total number of households. The second-stage sampling units
were holders within EAs from which a systematic random sample
of five (5) holders was selected for enumeration.
•The enumeration period spanned for 7 months.
•Seven structured questionnaire forms were used (listing,
demographic, area measurement-GPS, harvest, community,
groundnut, mango).
Gambia 2011-2012 Census of agriculture: sampled-
based census.
31. Country examples
• The CAP 2009-2010 followed the modular approach.
• Section G - Agriculture and Livestock - from the 3rd
Population Census in
2007 served as Core Module and Sample Frame for Complementary
Modules.
• The 3rd
Population Census included 6 of the 16 questions recommended
by FAO as part of the core items of the Agricultural Census.
• Six complementary modules were defined: Crops, Livestock, Labour
Force, Aquaculture, Food Security and Agricultural Services and
Practices
• A two stage sample method was used for the complementary modules:
• First stage: enumeration areas (EA) were selected systematically
• Second stage: ten agriculture households were selected randomly in each
EA.
31
Mozambique Agriculture and Livestock Census
2009/2010 (CAP) - Modular approach
32. Country examples
The agricultural census frame for the AC 2010 was the administrative
farm register (AFR) of the Ministry of Economic Affairs, Agriculture and
Innovation. Farmers have to register by law. The AFR contains names,
addresses and other characteristics of holders or holdings and a unique
registration number.
The agricultural census applies a higher threshold than the AFR, thus
only agricultural holdings above the threshold are taken into account.
This threshold is applied to separate professional from hobby farmers,
and to minimize processing burden.
Information on the census items existing in the AFR is taken directly
from the register for the whole census population.
The Netherland - Agricultural Census 2010
(Combined agricultural census with use of administrative registers)
The use of Computer Assisted Personal Interview (CAPI) (see Section 4.2.1) can facilitate the field work. The devices can be programmed to use information collected in the short questionnaires to identify the sub-population to apply the long questionnaires. For example, the device can be programmed in such a way that the long questionnaire pops-up when a holding meets the threshold requirements or criteria related to the subpopulation of interest.
Countries should make this decision depending on their national requirements, taking into account cost and resource availability, including financial and human resources.
In addition to those advantages, for countries that do not have a functioning agricultural statistics system, this modality being a 10-year integrated census/survey programme, lays the foundations for the creation of such a system to produce a regular flow of basic data. It will allow the generation of 65% of the Minimum set of core data recommended by the Global Strategy and will also provide basic data for monitoring the relevant Sustainable Development Goals (SDGs). Finally, AGRIS is expected to generate the flow of quality data required to monitor regional policy frameworks, such as the African Union Comprehensive Africa Agriculture Development Programme (CAADP).
Some countries may have 5-year census programme instead of 10-year programme. In this case, the implementation of AGRIS should be adapted to this particular situation with possibly more frequent implementation of the rotating thematic modules.
See .......
The cases where only the identification items are taken from administrative sources (such as name, address and location of the holding) are not considered as part of combined census modality, in these cases data from administrative sources can be used for construction of the census frame.
Possible problems are: missing data when some or all characteristics for a given unit are missing; errors in the variables when erroneous values for certain variables are registered (e.g., when the respondents give erroneous identification numbers for administrative data sources in the questionnaire).
Pre-filled questionnaires have three main benefits:
They help reduce the burden on data suppliers by saving them time (assuming that checking and correcting is quicker than finding and entering data)
They allow to check the quality of the administrative data
They are very flexible, particularly where the coverage of administrative data is not as complete as would be required for the other ways of use discussed above. They allow an easy data revisions/changes and adding additional information. The main disadvantage is the risk of bias introduced because some respondents may simply accept the pre-filled data without checking them, or may choose not to spend time correcting errors.