Individual land rights: Filling data gaps with the 50x2030 Initiative

4.1Adoption of the integrated approach in the Initiative’s survey tools

The 50x2030 Initiative has developed a set of reference questionnaire instruments, which continue to be expanded and refined.1616 These survey tools have been designed to include the module on measuring land rights at the individual level discussed in Section 3, encouraging the national level collection of these data in 50 countries by 2030. Before addressing the manner in which these questions on land rights were integrated into the Initiative’s tools, a brief overview of the 50x2030 survey programs is provided.

As discussed in Villarino, Buenaseda Tejada, and Patterson [15], the survey programs supported by the 50x2030 Initiative are structured such that a series of survey tools, each with specific topical focuses, are implemented on a rotating basis while core questions on agricultural production are asked on an annual basis. The two programs, the Agricultural Survey Program and the Integrated Agriculture and Rural Survey Program (henceforth referred to as the Integrated Program), differ in that the Integrated Program expands the sample from the inclusion of household and non-household sector farms to also include rural, non-agricultural households.1717 The survey tools developed by the Initiative are the same for both survey programs, with the exception of the tool on Non-Farm Income and Living Standards (ILS-HH) which is only administered in the Integrated Program. A summary of the survey tools is provided in Table 1, which is adapted from “A Guide to 50x2030 Data Collection: Questionnaire Design” [17].

As seen in Table 1, the placement of the questions for monitoring land rights at the individual level is dependent on the survey program being implemented. In the Agricultural Program, the questions are integrated into the Farm Income, Labor, and Productivity questionnaire (ILP-AG) tool, while they are included in the Non-Farm Income and Living Standards questionnaire (ILS-HH) in the Integrated Program. The rationale behind the distinction is that, in the case of the Integrated Program, the questions should be asked for both agricultural and non-agricultural land in order to also inform SDG 1.4.2 (albeit only for the rural population unless the sample is nationally representative). Since non-agricultural households are not administered the ILP-AG tool in the Integrated Program, it is necessary to administer the land related questions in the ILS-HH tool. For a heavily detailed discussion of the placement of the questions on land rights in both survey programs as well as the calculation of Indicator 5.a.1 using 50x2030 survey tools, see “A Guide to 50x2030 Survey Tools and SDG Indicator 5.a.1: Measuring Gender Parity in Ownership and Tenure Rights Over Agricultural Land” [18].

The survey tools of the Initiative incorporate the version of the questionnaire tool (discussed in Section 3) that allows for proxy respondents and collects data at the parcel level. The selection of the parcel level approach is a natural one, given the analytical value of having parcel-level data and the structure of the 50x2030 that already collect data at the parcel level (for agricultural production, for example). The allowance for proxy respondents in the reference tools was a decision taken based on considerations for the trade-offs between ease of implementation, budgetary constraints and data quality. Though the 50x2030 reference tools allow for proxy respondents, countries are encouraged to adapt this default protocol to collect self-respondent data for land rights, as feasible.

4.2Supporting the measurement in the Initiative’s partner countries

In addition to integrating the harmonized land tenure module into the Initiative’s survey tools, the Initiative works to ensure that partner countries are assisted in the adaptation of the survey tools to their local context and survey needs. In the initial stages of collaboration with a partner country, advocacy and technical sessions are normally held to explain (i) the value collecting data on individual land rights and SDG 5.a.1; (ii) how the questions can be best integrated in the main survey tools; and (iii) how the data can be analyzed.

An initial step in this support phase is confirmation of whether SDG 5.a.1 is already monitored (or supposed to be monitored) by other survey programs in the country. For instance, Georgia initially excluded SDG 5.a.1 from the national agricultural survey program because the indicator was already planned for inclusion in the national household survey. However, the indicator was subsequently excluded from the household survey and GEOSTAT amended its plans and scheduled the inclusion of the data needed in SDG 5.a.1 in the 2023 agricultural survey.

A second important discussion point is related to the value of individual-level data vis-à-vis household-level data. Initial reactions to this issue vary considerably across country partners. In some countries, the awareness on the importance of sex-disaggregated data is very high. This is the case of Uganda, where the national statistical office collaborated with UNSD on the MEXA study1818 and with FAO on the Global Strategy’s pilots [8], and has a long tradition of collecting sex-disaggregated data. These countries are obviously more eager to include the required questions and more proactive during the customization process. Other countries are less sensitive to the topic and it takes time and effort to appreciate the value of shifting from household-level to individual-level data. In some cases, the enthusiasm on the topic depends on the institution in charge of the agricultural survey program. National Statistical Offices tend to be more receptive than Ministries of Agriculture due to their experience with other types of survey programs, such as national household surveys and multi-topic surveys.

Once countries decide to collect the data for SDG 5.a.1, the third and final ‘crossroad’ is about the method – i.e., whether to collect the data at the parcel or individual level and if a self-respondent approach is feasible. As expected, countries choose the level that fits better the main survey questionnaire. For instance, Uganda and Senegal collected data at the parcel-level because they already had a parcel module in their questionnaire; Cambodia inserted individual-level questions in the household roster, given the absence of parcel level data in the main questionnaire. With respect to the respondent approach, countries are generally reluctant to adopt the self-respondent approach. This reluctance is generally rooted in a limited awareness of the potential discrepancies between proxy-respondent and self-respondent data, and by the fear that the self-respondent approach could substantially increase the fieldwork duration and complexity, especially when considering the larger demands of the survey as a whole.

Figure 6.

SDG 5.a.1 in Uganda. Estimates derived from the Uganda Annual Agricultural Survey 2018 and 2019.

As mentioned above, mainstreaming 5.a.1 questions requires reaching the right balance between customization and standardization. The questions on land rights for SDG 5.a.1 (and 1.4.2) are standard and should remain as such in order to ensure comparability across time and space; nonetheless, the list of legally recognized documents and the tenure types are country specific and should be adapted without changing the original concept. For instance, the list of documents for a given country should be exhaustive and clear and should always allow the analyst to distinguish the legally recognized documents from the informal ones during the processing stage. Consulting tenure experts can become crucial, especially in countries where different tenure systems and types of documents exist.

The experience of the 50x2030 Initiative to date has illustrated that using a standard and constant set of questions is very important for the stability of the SDG estimates. For instance, the 5.a.1 sub-indicators remained very stable in Uganda in 2018 and 2019 using the same questions (see Fig. 6), whereas different statistics were obtained in previous years using questions that diverted from the standard ones.

4.3Methodological validation

Methodological validation, to test and understand the implications of different approaches to collecting data on specific topics, is central to the 50x2030 Initiative’s mission to improve the quality and availability of agricultural data. The Initiative is purposefully designed to foster innovation and validate new tools and technologies for data collection, which are then documented in the form of public guidance tools and scaled up into national-level 50x2030 survey operations. The measurement of individual land rights is no exception.

Because multiple versions of the tool to collect data for monitoring SDG 1.4.2 and 5.a.1 were developed in an effort to allow integration into different types of surveys, it is necessary to conduct methodological validation around the different approaches. Though the different versions, as described in Section 3, are designed to ensure the questions are asked in a consistent manner despite the differences in the respondent approach and level of data collection, it is important to understand how the computation of the SDG indicators, as well as the resulting understanding of the specific land tenure rights included therein, vary with the implementation of the various versions.

Though the 50x2030 Initiative has already adopted a version of the land tenure module into its reference questionnaires, using the parcel level and proxy-respondent approach, which was selected with consideration for the demands of the larger survey as described in Section 3, these reference questionnaires can be adjusted based on findings from methodological research. Such findings can also encourage countries to adopt more refined approaches during the country customization stage, including the use of a self-respondent approach. In countries where a self-respondent approach is not feasible, it remains important to understand the data quality trade-offs and potential biases in SDG estimates stemming from the use of the proxy-respondent approach.

The 50x2030 validation efforts around the collection of data on individual land rights, both for the computation of SDG indicators and for the analysis and understanding of various aspects of land tenure, build and expand on previous literature on the gender-differentiated impact of proxy respondents on the measurement of assets in general [7, 11] and various aspects of land rights in particular such as Kilic, Moylan and Koolwal [12], Deininger et al. [13], and the studies included in the UN guidelines [7]. While these previous works focus on the impact of the respondent approach, research under the 50x2030 Initiative extends this work into another geographic context and includes the analysis of the impact of the level of aggregation of land tenure data. The design of and findings from the Initiative’s initial methodological study, conducted in Armenia, are discussed below.

Armenia Land Tenure and Area (ALTA) Study: Design & Key Findings

The Initiative’s first, but not last, methodological validation study on the measurement of individual land rights was conducted in Armenia, a 50x2030 partner country. Armenia was selected for the initial study based on country demand and interest in conducting such research. The World Bank, who spearheads the Methods & Tools Development Component of the Initiative, partnered with the Statistical Committee of the Republic of Armenia (ArmStat) and the International Centre for Agribusiness Research and Education (ICARE) to implement the Armenia Land Tenure and Area (ALTA) study.

The ALTA study, which was fielded from September to December 2019, was conducted in three marzes of Armenia – Ararat, Kotayk, and Vayots Dzor. These marzes were purposefully selected to ensure the sample covered areas with variation in intensity and type of agriculture, topography, and population density. Within these three marzes, a total of 100 enumeration areas (EAs) were randomly selected, using the 2021 population census as the sample frame. A household listing operation was conducted in each selected EA, and from that listing, 12 households were randomly selected, resulting in a total sample of 1200 households. In order to achieve the objective of analyzing the implications of the various land tenure modules with respect to SDG computation and the underlying land tenure rights, households were randomly assigned to one of four questionnaire versions, or treatment arms. The assignment of treatment arm was done at the headquarters level directly through the computer-assisted personal interviewing application, the World Bank’s Survey Solutions program,1919 to ensure there was an equal and random distribution of questionnaire versions and so that the assignment could not be altered by enumerators. The treatment arm assignment was randomized within EA, such that within each EA three households were administered each of the four versions of the questionnaire tested, to account for any differences at the EA level that could have affected responses to the land tenure questions, such as village-specific gender norms, for example.

The four questionnaires tested, and therefore the treatment arms, included all combinations of the respondent approach options (proxy or self) and the levels of land data collection (parcel level or aggregate), as defined in Table 2. In the arms employing a self-respondent approach, Arms 1 and 2, up to three adult household members were randomly selected for interviews, again directly through the Survey Solutions application to prevent biases introduced by selective assignment. Based on the amassing literature on the importance of self-respondents, and the a priori assumptions that individuals are best informed about their own rights, rather than the rights of other household members, and that reporting at the parcel level reduces cognitive burden through the elimination of the need for aggregation, we consider Arm 1 to be the best practice and therefore the benchmark against which the other arms are compared.

Figure 7.

SDG 5.a.1(a), by gender and Arm. Figure extracted from Gourlay et al. [19].

Given the experimental design of the ALTA study, we are able to draw conclusions on the implications of implementing each of the four questionnaire versions, in the context of Armenia, through descriptive analysis. The results discussed here are limited to those related to SDG 5.a.1 and the agricultural population, given the relevance to the 50x2030 Initiative and appropriateness of the Initiative’s sample strategy for complete monitoring of the Indicator, though ALTA results related to the urban and rural populations and SDG 1.4.2 can be found in Gourlay et al. [19].

Table 3 presents, by gender, summary statistics on the land tenure rights included in SDG 5.a.1, including legal documentation, the right to sell, and the right to bequeath, for the agricultural population, as measured by the different treatment arms. Immediately evident is the gender gap in and rights, with only 48% of women holding documented rights to land (in Arm 1) while 72% of men hold documented rights (in Arm 1) for example, reinforcing the need for the collection of individual land data. From a methodological perspective, the notable findings are the differences found in the measurement of these rights across treatment arm, and how those differences vary across gender. The estimates of the shares of men and women with legally documentation does not vary to any statistically significant degree across treatment arms. Alienation rights, however, do vary, but more so for men. While the share of men with the right to sell land significantly differs from the benchmark when estimated with Arms 3 and 4, there is no significant difference across Arms for women. The estimates of both men’s and women’s right to bequeath is sensitive to treatment arm, with the questionnaire version using a proxy respondent approach and parcel level data resulting in an underestimation of rights holders relative to the benchmark of Arm 1.

Though differences are observed across arms in the measurement of the alienation rights, the estimation of SDG Indicators 5.a.1 (a) and 5.a.1 (b) shows more consistency across treatment arm. Figure 7 depicts SDG 5.a.1 (a), the proportion of total agricultural population with ownership or secure rights over agricultural land, for men and women separately, by Arm. While the point estimates differ across treatment arms, the means are not significantly different in Arm 1 than in the other treatment arms (within gender). The same is true for the estimates of 5.a.1 (b), share of women among owners or rights bearers of agricultural land, which are estimated at 45%, 43%, 42%, and 45% in Arms 1, 2, 3, and 4, respectively, but with no statistically significant difference between the benchmark (Arm 1) and the other treatment arms. Point estimates for both SDG 5.a.1 (a) and 5.a.1 (b) are reported in Table 4.

The stability of the SDG 5.a.1 estimates across treatment arms, and therefore across proxy respondent approaches and levels of land data collection, lends support to the use of the proxy respondent version that is currently integrated into the 50x2030 reference questionnaires as well as the other versions made available to the public. This finding, however, is potentially driven by the relatively high rates of documentation in Armenia vis-à-vis other developing contexts, coupled with the manner in which 5.a.1 is estimated (whereby the presence of a legal document or alienation rights is sufficient for an individual to be considered as having secure tenure), and therefore extrapolation of this finding to other contexts should be done carefully. Previous research conducted in contexts with lower documentation rates, such as Kilic, Moylan, and Koolwal [12], have shown more significant biases stemming from the use of proxy respondents. Results of ALTA, however, do suggest consistency in the estimation of SDG 5.a.1 in the Armenian contexts, and potentially other contexts with a similar land tenure landscape. Despite the consistent SDG estimations, the understanding of specific land tenure rights, in particular the rights to bequeath and sell, is biased by the use of proxy respondents and data aggregation, with use of use of the proxy respondent modules resulting in underestimations of these alienation rights relative to the suggested self-respondent approach. These biases should be taken into consideration when selecting the appropriate method of collecting land tenure data for analysis of different land rights.

Future Methodological Research & Scale-up Avenues

The experience of the ALTA study contributes greatly to our understanding of the implications of various survey design decisions on the resulting SDG estimations and understanding of land rights, but it in itself is not sufficient to make conclusions at a global scale. Under the 50x2030 Initiative, additional studies such as ALTA will be conducted in different contexts, including in Africa and other Asian countries, to inform individual land data collection practices in those 50x2030 operations.

The ALTA study has informed the development of a guide for cognitive interviewing around the land tenure module, irrespective of the questionnaire version selected, that will facilitate cognitive interviewing around land rights prior to survey fieldwork, ultimately resulting in improved data on individual land rights Hovhannisyan, Gourlay, and Grigoryan [20].

Finally, parallel work of the Initiative on women’s empowerment has recognized the importance of land rights and pushed forward with further adoption of land rights measurement. The International Food Policy Research Institute (IFPRI) and Emory University, together with the 50x2030 Initiative, are working on the Women’s Empowerment Metric for National Systems (WEMNS) project to develop and validate a measure of women’s empowerment, which includes critical questions individual land rights. Methodological validation of WEMNS tool is being conducted in Bangladesh, Malawi, and Nepal prior to large scale implementation.