86440 Testing New Approaches Household sample surveys are rich in information but collect information from a to Mapping Livestock relatively small sample of households, which is rarely if ever large enough to produce Variables in Uganda: statistically reliable estimates at lower levels of disaggregation, such as for districts or wards. Integrating Census and For instance, multi-topic household surveys could provide information to estimate livestock Survey Data income for some representative household, but cannot provide details on the districts where There is evidence that livestock provide a households benefit the most (or less) from their multitude of livelihoods services to farm farm animals. households. They are a source of food, cash Censuses collect data on a limited set of income, draught power and hauling services, variables but on a complete enumeration basis, saving and insurance, dung and social capital. i.e. from all farm holdings/rural households in There is also evidence that a large share of rural case of an agricultural census, or from a sample households keep livestock, including the poor. large enough to produce estimated at lower Targeted investments in the sector could thus administrative units. However, the collected definitely contribute to improved livelihoods data are, on their own, insufficient for and a decline in rural poverty. formulating effective investments. For example, Data integration, the combined use of data from a livestock census provides detailed information different sources, is an effective way to derive on the distribution of the cattle population at detailed policy and investment indications, village level, but not enough information to which could not be obtained by using individual estimate livestock income at household level. datasets on their own. Indeed, the second pillar Small Area Estimation (SAE) technique can be of the Global Strategy to Improve Agricultural used to integrate household sample surveys and and Rural Statistics targets integration of Census Data. The method behind SAE is different data systems for more effective relatively straightforward. First, comparable policies and investments. variables are selected from the survey and the This brief summarizes the results of a livestock census. These variables must be similar in mapping exercise whereby household survey and mean, standard deviation, frequency livestock census data are integrated to generate distribution at the national level, collection highly spatially-disaggregated maps for Uganda. method, and definition. Second, an estimation It is the result of a joint work undertaken by model is fitted on the survey, where the IFPRI/HarvestChoice, the LSMS-ISA project of dependent variable is one missing in the census the World Bank and FAO-WB-ILRI Livestock Data data (e.g. income). Third, parameters estimates Innovation in Africa Project. More details are from survey data are applied to census data, available in C. Azzarri and E. Cross (2012) allowing mapping at the lower administrative Livestock Data Innovation in Africa BRIEF Testing New Approaches to Mapping Livestock unit indicators generated in the survey. The and Livestock Income in Uganda, available at method is explained in detail by Elbers, www.africalivestockdata.org/afrlivestock/conte Lanjouw, Lanjouw (2003; Micro-Level Estimation nt/papersreports of Poverty and Inequality, Econometrica, 71:1, pp. 355-364). Methodology and Data Numbers for Livelihood Enhancement www.africalivestockdata.org SAE has been predominantly used to impute though statistical diagnostics are not fully measures of consumption or income into census satisfactory, the two maps provide evidence data using estimates based on survey data. that by virtue of survey-to-census prediction, it Here, the 2009/2010 Uganda National Panel is possible to draw higher spatially- Large ruminants: number/skm Panel Survey Census Predicted Predicted income-related livestock variables disaggregated maps and for more US$ PPP per households % of total household income than using census or indicators Predicted survey data on their own. These and similar maps could be valuable tools for decision makers. Conclusions Data integration is essential for effective investment decisions as it allows generating information than different surveys, on their own, could not produce. SAE represents an effective method to integrate data from different sources, Survey and the 2008 Uganda National Livestock including from the 2009/10 Uganda Census are integrated using SAE technique to National Panel Survey and the 2008 Livestock draw highly spatially-disaggregated maps Census. targeting livestock related variables than would For data integration to be feasible it is have been possible using survey or census data important that the various methods of data alone. collection be consistent. For instance, the statistical unit should be the same, as well as Results definitions and classifications; even questions Three sets of results are presented. First, the should be formulated consistently. Finally, the density of large ruminants (heads/sqkm) at the timing of the various surveys is also relevant as sub-county level are predicted and then it may be of little use to compare data from compared to actual values in the census. This surveys undertaken in different years, allows testing the reliability of the prediction particularly when addressing issues pertaining method used. The model has a relatively high to fast, or relatively fast growing sectors. explanatory power (see top figure above), which suggests that SAE can be a viable and reliable For further information please visit: method to estimate spatial distribution of www.africalivestock.data.org missing information through prediction. Or contact: The other two estimated models allow Carlo Azzarri, IFPRI estock Data Innovation in Africa BRIEF predicting and mapping livestock per capita c.azzarri@cgiar.org income (US$ 2005 PPP) and household’s share of income from livestock at sub-county level. Even Elisabeth Cross, IFPRI elizabethcross85@yahoo.com Numbers for Livelihood Enhancement www.africalivestockdata.org