The FAO-UN Women partnership to support countries in producing sex-disaggregated data on agricultural landownership: SDG indicator 5.a.1

The DHS Program has collected, analyzed, and disseminated accurate and representative data on various topics including population, health, HIV, nutrition, women’s economic empowerment and violence against women through more than 400 surveys in over 90 countries since its inception.

This study focuses on the DHS-VII surveys which were fielded starting in 2018. For the calculation of SDG indicator 5.a.1, the sources for the minimum data requirements are specific data items from the three questionnaires of the survey, briefly described below:

Figure 2.

First six definitions of agricultural households.

For the denominator of 5.a.1, we need to identify the agricultural population, which is defined as “adult individuals living in households where at least one member is involved in agriculture as an own account/self-employed worker”. Thus, the first step is to identify the agricultural households, where someone in the last year has cultivated any land or raised any livestock or poultry. Although the DHS-VII questionnaire does not explicitly collect this information, there are three pieces of information that could be used for this purpose:

Household level ownership of land and/or livestock (1 and 2) is not sufficient to identify those who have been working in agriculture, thus these two are complemented by taking into account the working sector and status of interviewed individuals (3).

Initially, FAO identified 6 possible definitions derived from the three dimensions above, in which the household level ownership of agricultural land was always taken into account and combined with the other two to find a suitable proxy. These six options are summarized in Fig. 2.

The rationale behind always including the household level ownership of agricultural land as a criterion in the denominator was to mitigate an issue with the numerator, which will be further discussed later. In short, individual questions on land ownership (used for the numerator) are not specific to agricultural land, but refer to any land, thus restricting the denominator to only households with ownership of agricultural land in particular would have helped in improving the proxy.

After these six initial definitions, internal discussions led to the idea that the household ownership of agricultural land could also be moved from defining the agricultural households of the denominator to help discerning individual general landowners from individual agricultural landowners at the numerator level. In other words, instead of restricting the denominator only to households who have some agricultural land (as well as one or two of the two other criteria), we could restrict the numerator only to individuals in agricultural households who possess land themselves and that live in a household with ownership of agricultural land. This opened up to the possibility of 4 further definitions for the denominator, which did not necessarily account for household agricultural land ownership at this stage, these are summarized in Fig. 3.

Figure 3.

Four additional definitions of agricultural households.

Figure 4.

Different types of answers to main work activity found in the data and corresponding limitations.

On review, the closest definition to the metadata would be definition 10, which looks at the individual employment sector and status of the men and women, which is found in the questionnaire usually administered to individuals between 15 and 49 years of age. Thus, any household where at least one of the interviewed members is classified as self-employed in agriculture has been identified as an agricultural household. In turn, any adult (18+) member of those households, regardless of direct involvement, is considered part of the agricultural population, i.e., the denominator of SDG indicator 5.a.1.

The indicator’s metadata recommends including households where the involvement in agriculture is through self-employment activities at least for one of the members. Thus, it is recommended to filter out the households where all the members that are involved in agriculture are wage workers. Usually, this is achieved by adding another question that specifically asks for the type of employment. However, we do not have precise information on this distinction in the men’s questionnaire of the DHS dataset. In fact, although the women’s questionnaire includes the question: “Do you do this work for a member of your family, for someone else, or are you self-employed?”, the question is not present in the men’s questionnaire in all the countries analyzed except for Indonesia.

Despite the inconsistency between the two questionnaires, it is possible to retrieve the employment status of men by analyzing the answers to the question on the main work activity in the men’s questionnaire. Unfortunately, the answers to this last question were not collected uniformly across all countries, leading to inconsistency in the responses as summarized in Fig. 4.

The choice of denominator definition has important implications in terms of the sample that can be used to estimate the proxy for SDGi 5.a.1. In the DHS, household level information is collected through the household questionnaire that is administered to all sampled households. Individual level information is collected through the women’s questionnaire, which is administered to all eligible women in all households; and through the men’s questionnaire, administered to all eligible men in a subsample of households, usually one half or one third.

Whenever we use individual employment information to define agricultural households, and thus we use individual data to define household level characteristics, i.e., definitions 3, 4, 5, 6, 8, 9 and 10, if we estimate the indicator on the whole sample, there will be a bias between households where both men and women have been interviewed and those households where only women have been interviewed. The proxy measure methodology thus uses a subsample of the survey households consisting of households where both men and women are interviewed.

The numerator of indicator 5.a.1 is defined as the number of people in agricultural population with ownership or secure rights over agricultural land. To determine whether an individual is said to have ownership or secure rights to agricultural land three conditions (proxies) are defined in the indicator’s metadata:

Including alienation rights is necessary to include de facto ownership in the indicator, which is especially relevant in countries where the penetration of formal documentation is low but individuals still hold customary and non-formal ownership rights.

The collection of land ownership information in the DHS survey differs from the recommended methodology for 5.a.1. The DHS questionnaire has three questions that relate to individual land ownership, that are administered through the male and female individual questionnaire:

Thus, these three questions posed a few challenges in terms of estimating 5.a.1. First of all, individual land ownership questions do not refer specifically to agricultural land, but they refer to any type of land; this challenge was mitigated by taking into consideration the household level question that was specific on agricultural land ownership:

Only individuals who declared having any type of land and at the same time lived in a household that owned agricultural land were considered for the indicator’s numerator.

Secondly, no alienation rights data are collected, which could imply a significant loss of information on individuals who might be the de facto owners and exercise ownership rights even if they do not have a document.

Given the above discussion, three different approaches for the calculation of the numerator were evaluated based on the availability of proxies in the DHS as described in Fig. 5.

Figure 5.

Three definitions of agricultural landowners.

Whenever the three proxies are not available, the recommendation from 5.a.1’s methodology is to prefer the most inclusive option, to include non-formal and customary forms of ownership. Using option B or option C would mean to exclude many situations where non-formal but de facto ownership is relevant to women’s and men’s empowerment. Thus, option A has been selected as the one providing the best approximation of SDGi 5.a.1.

DHS surveys have the potential to be a valuable source of data to estimate SDGi 5.a.1. The discussion in this paper has touched upon how the available data can be used to approximate the estimate of the indicator, albeit with a few important limitations. Some can be easily solved and some are more difficult to overcome – for instance the fact that relevant questionnaires are administered to an age-defined subset of the population (usually 15–49 years, for example, for women). We would like to offer some recommendations on how to revise the existing DHS questionnaire to include all the data requirements to estimate SDGi 5.a.1.

The main limitation to the identification of the denominator, i.e. the agricultural population, lies in the different ways employment questions are collected for men and for women. Collecting a question regarding the type of employment (whether the individual works for a family member, for someone else or is self-employed) not only for female respondents but also for male respondents would eliminate this issue. The proposal for improvement would simply entail adding one question in the male questionnaire to reflect the same structure as the female questionnaire, as layed out in Fig. 6, where the proposed addition is highlighted in green.

Figure 6.

Proposal for improving questions capturing the agricultural population according to 5.a.1 definitions.

Figure 7.

Proposal for improving questions capturing land ownership according to 5.a.1 definitions.

In terms of the numerator, there are two main limitations. First, land ownership questions are not specifically referring to agricultural land, but rather refer to both non-agricultural and agricultural land together. Asking the same 3 existing questions separately for non-agricultural and agricultural land would allow to identify agricultural landowners and discern them from other types of landowners. The second limitation is found in the lack of questions on alienation rights, which means that the survey does not capture a very important section of de-facto ownership, which proves particularly relevant for female landownership in many countries; within the flow of the current questionnaire, this dimension could be easily captured by adding two questions for each type of land that is collected. The proposal is layed out in Fig. 6 where changes and additions are highlighted in green.

The two proposals in Figs 6 and 7 provide a simple and effective way to align the current questionnaires to the SDGi 5.a.1 recommended methodology which allows for a good estimation of the indicator. Individual countries may wish to expand on this proposal and add questions on the type of land ownership, to be able to disaggregate the indicator by type of tenure, for instance freehold tenure, customary tenure and so on. Such an exercise would need to be preceded by an analysis of the existing land tenure system in the country and corresponding adaptation of the questionnaire. Then, a question on the type of tenure should follow question 1 of proposal 7, and questions 2–5 should be repeated for each type of tenure.

With reference to the key requirements for estimating indicator 5.a.1 and its sub-indicators, the EHVCM collects three types of information that could be used to identify agricultural households:

There are three main challenges in identifying the denominator arising from this structure of the EHVCM survey questionnaire:

Figure 8.

Operational definition of agricultural households for 5.a.1 in EHVCM surveys.

First of all, although it fits the definition, Section 17A cannot be used for purposes of identifying the denominator of SDG indicator 5.a.1. The reason is that the individual ownership questions are located in the “fields and parcels” (Section 16A) section of the survey, and thus they are administered only to households that have been identified as cultivating land. Livestock-only households cannot be included in the denominator because their members’ numerator status is undefined since they are not asked about their ownership of land. Thus, the proxy for 5.a.1 estimated using this survey will have the limitation of having a denominator and thus a reference population restricted to crop farming households only.

Secondly, the question in Section 16A that identifies households where at least one member has cultivated land has a shorter reference period, the winter season, than recommended by the indicators’ metadata, the past 12 months, and this represents a limitation of the estimate computed using these surveys.

And thirdly, although Section 16A does not explicitly consider only individuals who are self-employed in agriculture, it still implicitly does so by collecting information based on a roster of parcels cultivated by the household. The tenure options of these parcels are: owner, free loan, tenant farming, sharecropping, pledge or other, clearly indicating that agricultural wage workers are not captured by this section. For simplicity, it was decided not to use employment information from Section 4 to further confirm whether they were self-employed. Households that are self-employed in agriculture in Section 4 but at the same time they are not captured in Section 16A will still be excluded given that they were not asked ownership questions. Households who are involved in agriculture exclusively through wage work would not be captured by Section 16A in the first place.

Given the challenges and considering the limitations above, only adult individuals who live in households who declare cultivating land in Section 16A were included in the denominator of 5.a.1, the main driver of this decision was the placement of ownership questions only in that section. For a visualization of this operational definition, refer to Fig. 8.

Figure 9.

Two definitions of agricultural landowners in EHCVM surveys.

As defined in Section 2, to capture the individual ownership of agricultural land, the methodology for SDG indicator 5.a.1 recommends identifying individual land ownership through the collection of the three proxies of ownership rights. Collecting data on alienation rights allows to capture the de facto owners of agricultural land even in contexts where legal documentation is not widespread.

Figure 10.

Proposal for improving questions capturing land ownership in EHCVM surveys according to 5.a.1 definitions.

After having identified the data source of the denominator, i.e. the agricultural population, we need to identify the data source for the numerator of 5.a.1, namely the adult men and women in the agricultural population who are owners of agricultural land. Individual ownership of agricultural land in EHCVM surveys is collected for each of the parcels listed in the “fields and parcels” section1616. For each parcel declared by the household, respondents are asked four relevant questions regarding the ownership, that allow both for identification of individual reported owners, as well as individual documented owners of the parcels.

These questions allow for two options for the identification of landowners, these two options are visualized in Fig. 9.

Broadly speaking, Option A, reported ownership or possession, is relatively less reliable than Option B, documented ownership. However, in a situation of scarce data on formal documentation along with missing information on alienation rights, reported ownership is a useful alternative for comparing ownership between men and women without excluding de-facto ownership. Thus, in this case individuals identified as being the reported owners (as shown in option A) are considered as owners for the numerator of SDG indicator 5.a.1. Using option B would mean to exclude many situations where non-formal but de facto ownership is relevant to women’s and men’s empowerment.

It is important to note that an important limitation of EHCVM surveys concerns the identification of agricultural land, as defined in Section 2.

The EHCVM questionnaire asks to list all the field and parcels that have been exploited by someone in the household, thus covering LU1 and LU4; however it does not include pasture land (LU2 and LU5) and fallow land (LU3). Thus, the numerator will be underestimated since the survey does not collect information on some subcategories of agricultural land.

EHCMV surveys have the potential to be a valuable source of data to estimate SDGi 5.a.1. The discussion in this section has touched upon how the available data can be used to approximate the indicator, albeit with a few important limitations. We would like to offer some recommendations on how to integrate the existing EHCMV questionnaire to include all the data requirements to estimate SDGi 5.a.1:

These four recommendations provide an effective way to align the current EHCVM questionnaires to SDG 5.a.1’s methodology.