FUTURE DIRECTIONS IN PRICE DATA MAPPING FOR FOOD SECURITY ANALYSIS

However, current price data analysis is hindered by patchy data, which is both too sparse in terms of commodity/time/space coverage and too large, with too many observations rarely synthesized in a way useful for decision-making. Methods exist to create usable price datasets, including interpolation, data summary methods, and analysis and drawing policy implications. Interpolation describes a toolkit of methods for "filling in" data where raw data do not exist. There are numerous interpolation techniques and models, many of which have been developed for geo-spatial analysis and have not been applied to price analysis. Interpolation can be spatial (filling geographic gaps) or temporal (filling gaps in a time series) or some combination of the two. Price data interpolation is generally temporal. In this paper, we will look at spatial methods too. Interpolation is necessary in all FEWSNET countries due to missing and incomplete data. Comparison of prices over time and space requires consistently defined data. Several reasons exist why, in the conventional "market information system" model, consistent time series are lacking. These include the problem of low transactional volumes (too few valid observations when small markets are quiet), low compliance with data collection protocols by agricultural extensive officers, inconsistent investment in market information, and data loss through poor coding. Temporal interpolation methods include using autoregressive/lag models, general consumer price deflation, application of more specific CPIs (e.g., rural/urban, higher frequency), explicit estimation of seasonal effects using NDVI (i.e., association of intra-annual variability of de-trended data with explicit determinants of seasonality), and NDVI-based measures to predict price variability. Regression modeling is also used to test interpolation methods. For price data to be useful for policy-making, it now needs to be summarized. Single commodity data is hard to map and interpret. The key method for summarizing price data is in indices (plural of index). The index aggregates price data from multiple commodities. The indexing is done depending on what story you want to tell. For FEWSNET, a lowest common denominator is used, i.e., data that are available in all FEWSNET countries. This ensures that the methods and reporting are comparable across countries. Interpolation allows visualization of price data in map form. It is hard to get a picture of market prices from a few scattered market points. Interpolation will allow the possibility of building maps that show market "hotspots" where prices are anomalous for some reason. The paper proposes a pathway of steps required to produce meaningful map products for price analysis. It covers the methods for handling patchy data (interpolation), data summary methods, and analysis and drawing policy implications. The paper sets out a roadmap for generating meaningful "heat maps" of market prices that is similar to the WRSI and NDVI products. It proposes to build tools for analyzing price data, which will include on and offline mapping capabilities. Advances in GIS make it possible to generate continuous (raster) map surfaces from point data. Market food prices are a key determinant of food insecurity, but price data are not systematically considered in analyses. Stronger analysis would benefit policy-making both in terms of identification and quantification of the scale of the food security problem (i.e., spotting those areas where needs assessments need to be focused) and identifying the best responses. A key reason for inadequate price data analysis is patchy data. The paper proposes to develop tools that are applicable in data-scarce FEWSNET countries, using a lowest common denominator, i.e., data that are available in all FEWSNET countries.