The existence of rich micro-level household data sets that combine welfare indicators with socio-economic and policy variables has increased in the past decade and a half. Such data sets provide a valuable tool for analyzing poverty and its causes.
The bulk of the analysis carried out using these surveys has been country-specific: the surveys were designed to provide policy-relevant data to governments. But increasingly the data from these surveys, in terms of poverty and inequality measures, are used for comparisons across countries. (For example, the estimates of the “global” poverty aggregates in Chen and Ravallion, "How Did the World's Poorest Fare in the 1990s?," draw on 297 surveys for 88 countries spanning 1985-2000.) This has created a need for a better understanding of the content of the welfare measures, better documentation of them and an effort to generate greater consistency over time and across countries.
Two sets of problems have been identified, which this subcomponent of the research program will aim to address. Internal consistency
Accessibility Internal consistency
Firstly, there is a pressing need for enhancing the internal consistency of existing data and to better document existing inconsistencies, many of which will no doubt persist. The Bank now has direct access to the micro data for many hundreds of household surveys. The compilations done by the Bank in the World Development Indicators (drawing on the GPID) are all done from primary data (either unit record or tabulations). These compilations are up-dated over time at varying frequencies; for example, an up-date of the Deininger-Squire data base is currently underway in DECRG (supported by the Research Committee), and the GPID is up-dated annually. However, only minimal filters are applied to the source data to iron out inconsistencies, such as differences in definitions of income or consumption, which can vary between countries and even over time for a given country. Serious comparability problems have been documented in the few cases in which existing data compilations have been closely inspected.
A subset of the surveys are within the LSMS. Here there is already some degree of internal consistency. However, as each survey was analyzed separately, there are inconsistencies in the definitions, calculations and coverage of the social indicators. Going beyond LSMS, the comparability problems magnify.
Generating more comparable indicators would improve the quality of cross-country comparative work in the Bank and the development community more generally. It would also open up new areas of research. Of course, there are comparability problems that cannot be eliminated, short of mounting new surveys. However, given that so little work has gone into re-constructing the derived data sets from the micro data (with users relying heavily on existing aggregates from the survey data), there is also scope for a more standardized data base for distributional analysis. This effort would also expose comparability problems that cannot be resolved, but should still be properly documented for users.
Accessibility
Secondly, users of micro data sets in the research and policy communities currently face unnecessarily high start-up costs. At present anyone wanting to do (say) a cross-tabulation of school enrollment rates by level against household income per equivalent single adult for a given country will first have to get access to the micro data, often requiring permission from the statistics office in the country concerned. Naturally, access to micro data at country level is constrained by the letters of agreement between the Bank and the Governments concerned. These agreements vary from country-to-county. Users currently face transaction costs in obtaining access that can be sizable, particularly for cross-country comparative work. Users face a quite significant skill and technology hurdle; mistakes are to be expected and are common. The entry cost rises more than proportionately if aiming to do cross-country comparisons, such as for a Bank regional report or a central cross-country study, given the heterogeneity of micro data (as noted above). Yet it is feasible to greatly reduce these access costs to users, thus greatly expanding the potential set of users, while avoiding common sources of error. This requires a reduced, cross-country data set containing comparable indicators at micro level that is web based.
To ensure that these types of analyses are done correctly and to provide a much-requested service to our clients, we propose to create the Comparative Living Standards Project (CLSP). This will aim to provide the most internally consistent and well-documented library of integrated micro-data and basic summary tabulations from those data available for developing and transitional countries. The core of CLSP will be an interactive on-demand query system of cross-country household-level data sets that responds to client requests with tables and analyses created by the clients from the Bank’s survey data archive. CLSP will be a unique effort to create a well-documented, standardized and accessible micro data base for measuring poverty and inequality at country level. In doing so, the project will fill the increasing demand from World Bank staff, governments, and researchers for comparable poverty and inequality measures as well as simple analyses — frequencies, cross-tabulations, and means — on various topics in specific countries and across countries. These clients want the ability to do quick comparisons by poverty group, by gender, by country, by region, and all of the above. Prior clearances will be sought for all data sets to be included in CLSP. This means that a user will not need further clearances unless she wants to go to the original micro data. For many purposes, however, the software available within CLSP will suffice. For users within the Bank, it will also be possible to download the data for use with other software. (For users outside the Bank, clearances will still be required for downloading the micro data.)
In essence, the outcome of the CLSP will be a “World Development Indicators” at household level (rather than country level), complementing what the Bank’s SIMA database does at the macro level. It will serve a similar function for the developing world as the Luxembourg Income Study (LIS) has done so well over the last decade for the developed countries. CLSP will offer a potentially powerful new data tool for development research and policy analysis.
|
RSS