How good are livestock statistics in Africa? Evidence from Ethiopia

Livestock supports the livelihoods of around 1 billion people in low- and middle-income countries (LMICs) (Thorne and Conroy 2017, Baltenweck et al. 2020). However, growth and productivity of the livestock sector in many LMICs are not keeping pace with the increasing demand for animal-source foods. Boosting the sector’s productivity is crucial for poverty reduction in LMICs, which continue to face multifaceted challenges and shocks that threaten the sustainability of food systems. This, in turn, requires reliable livestock data for informing livestock policies and investments. However, there is a significant gap in the quality and reliability of livestock data, which inhibits evaluations of the livestock sector’s role in livelihoods and national economies and therefore impedes evidence-based livestock policies and investments.

Livestock statistics in most LMICs rely on self-reported, survey-based measures. However, respondents may face various challenges in accurately reporting livestock ownership. Despite the large potential of the livestock sector to drive poverty reduction efforts and economic growth in many LMICs, including Ethiopia (the focus of our research), these data gaps are likely to undermine our understanding of the contribution of the livestock sector.

What drives mismeasurement in livestock assets?

Measurement of livestock assets in most LMICs relies on self-reporting by selected respondents. This approach is mostly preferred because of cost-efficacy and logistical feasibility. In most household and agricultural surveys, livestock data is collected as part of extensive multi-topic modules. These modules are typically administered in a single-visit household survey, making the marginal cost of collecting self-reported livestock data very low. However, self-reported data can be prone to various sources of bias and underreporting.

1. Recent research show that self-reported assets and income data suffer from substantial measurement errors (Meyer et al. 2015), although many of these studies are from high-income countries. Measurement errors associated with livestock assets may arise from number of factors. First, most livestock data is collected as part of multipurpose agriculture modules, which are usually long and hence suffer from fatigue and recall bias (Ambler et al. 2021, Abay et al. 2023). Given that livestock modules usually come late in the survey when respondents are ‘fatigued’, the quality of livestock data may be disproportionally affected.
2. Respondents may have some motives to intentionally underreport their assets and income, particularly if they believe that the data may be used for purposes that may affect their tax payments or access to various public programs. This is the case if respondents perceive that the data they are reporting to interviewers may not be kept confidential. For example, despite variations across regions, some regions in Ethiopia have introduced livestock taxes in the last few years and livestock ownership is one important criterion for inclusion in the national Productive Safety Net Program (PSNP) (Berhane et al. 2014).
3. Respondents in developing countries lack the level of literacy and numeracy that is needed to accurately count and aggregate all types of livestock owned by all members of the household (Dillon and Mensah 2024). This can be described as aggregation bias (Dillon and Mensah 2024).
4. Livestock assets in many LMICs are owned by several members of the household, and hence one respondent, including the household head, may have some level of asymmetric information about all types of livestock assets owned and managed by each member of the household. This is plausible as different types of livestock are usually managed by different household members.
5. Rural farmers in developing countries such as Ethiopia are not used to counting their livestock and related assets, mainly because of social and cultural norms related to counting assets or other belongings. In the absence of methodologically appropriate methods that can address these challenges, self-reported livestock data are likely to suffer from under(over)-reporting.

When and why mismeasurement in livestock assets matter

Under(over)-reporting of livestock asset ownership can lead to under(over)-estimation of the contribution of the livestock sector to national economies and livelihoods. Given the multiple purposes livestock assets serve in LMICs, including as source of livelihood, income, food and nutrition, draft power and transportation services, such under(over)-reporting of livestock asset ownership can mislead public and development policies. In most cases, estimating the contribution of the livestock sector to national Gross Domestic Product (GDP) entails estimation of the livestock population as well as associated livestock products and services (Behnke et al. 2010). Underreporting and hence underestimation of livestock population can undermine the contribution of the livestock sector to national GDP.

Similarly, under(over)-estimating livestock population in a specific geographic unit can mislead public and private investment priorities as well as climate adaptation and mitigation strategies, especially given the significant contribution of livestock production systems to global warming (Havlík et al. 2014). Finally, as livestock assets are widely used as indicators of rural economic wealth and hence used for targeting of poverty reduction programs and interventions, misreporting associated with livestock asset ownership can lead to mistargeting of development programs.

In terms of implications on predictive inferences, the consequences of measurement error associated with self-reporting of livestock assets depend on whether these potential inaccuracies are systematic or random. Some of the mismeasurements and inaccuracies associated with self-reported livestock data can be random and hence less consequential while some of the above sources of mismeasurement are more likely to be systematic and hence distort predictive statistical results using these self-reported data.

Direct counting and nudges to improve livestock asset measurement

In recent research (Abay et al. 2025), we introduce a novel set of survey and measurement experiments to improve livestock statistics in Africa. First, with the aim of addressing some of the potential sources of underreporting in livestock assets described above, we introduced an explicit nudge to a random subset of survey respondents. The nudge reminds respondents that the livestock data collected will be used solely for research purposes, and not to identify beneficiaries of social protection programs or for tax purposes. This was delivered midway through the survey and just before the start of the livestock module.Second, we arrange for direct counting of livestock assets by enumerators and local livestock experts. As shown in Figure 1, we show that that self-reported data on livestock ownership suffer from significant underreporting. While the average effect of the nudge appears to be negligible (although it affects the reporting behavior of households with larger stocks of livestock), direct counting increases total livestock ownership by 39% and the reported number of cattle by 43%.

Figure 1

Implications for improving measurement and informing livestock policies and investments

Our findings have important policy implications, including on the debate related to the role of the livestock sector to national economies and livelihoods. The pervasive underreporting of livestock asset ownership can lead to an underestimation of the livestock sector’s contribution to national economies and livelihoods. For example, the contribution of the livestock sector to national GDP is a function of the estimated livestock population as well as associated livestock products and services. Underestimating the livestock population can undermine this contribution.

Underestimation of livestock population and associated returns can also lead to misallocation of public and private investments while also distorting our understanding of rural economic conditions. Beyond distorting descriptive patterns and statistics, systematic measurement error in livestock asset ownership can mislead predictive inferences relying on these data, including those related to the contribution of livestock to household livelihood, nutrition and income. Indeed, the empirical and inferential implications of such mismeasurements in livestock asset ownership merits further investigation.

Kibrom Abay is a Senior Research Fellow with IFPRI’s Development Strategies and Governance (DSG) Unit; Hailemariam Ayalew is a DSG Research Fellow; Zelalem Terfa is a Scientist, Gender Quantitative with the International Livestock Research Institute (ILRI); Joseph Karugia is an ILRI Principal Scientist; Clemens Breisinger is a DSG Senior Research Fellow and IFPRI Kenya Country Program Leader. This post first appeared on VoxDev. Opinions are the authors’.

This work was supported by the CGIAR Research Program on Policy Innovations.

Referenced paper:
Abay, Kibrom A.; Ayalew, Hailemariam; Terfa, Zelalem; Karguia, Joseph; and Breisinger, Clemens. 2025. How good are livestock statistics in Africa? Can nudging and direct counting improve the quality of livestock asset data? Journal of Development Economics 176(September 2025): 103532. https://doi.org/10.1016/j.jdeveco.2025.103532