The Global Consumption Database is composed of individual country surveys to form a database on household consumption patterns in developing countries. The data has been standardized following a six-step process:

Step 1: Annualizing consumption or expenditure data:

In simple cases, this amounts to using a multiplying factor determined by the recall period (the period in which households are asked to recall their expenditure during that period). For example, food data collected for the last 7 days would be divided by 7, then multiplied by 365; monthly values by 12 etc.

Step 2: Detecting and fixing outliers:

Expenditure values were flagged to be outliers if they exceeded the average amount consumed in the third quartile plus 5 times the interquartile range (the difference between the first and third quartiles of the data).

Any flagged values need to be confirmed before imputations are made. If three or more non-food values are flagged as outliers for a household, it was assumed this indicates a rich household; hence the flags were removed. Households in the top two consumption quintiles were also assumed to spend unusually large shares of their income on education and jewellery. Outlier values that did not fit either of these criteria were replaced with the weighted mean of the non-extreme values for the consumption variable in question.

Step 3: Mapping commodities to the ICP/COICOP classification:

Commodities found in each survey dataset were mapped to a standard classification of products and services, and then aggregate standard products and services into sectors and categories. This used the International Comparison Program (ICP) classification which is equivalent to the International Classification of Individual Consumption According to Purpose (COICOP).

Step 4: Extrapolation to 2010:

Extrapolations were undertaken to convert all consumption and population data to a common reference year, 2010. For example, for the 2007 survey conducted in Guinea: final consumption expenditure per capita in LCU was 3,177,774 in 2010 and 1,547,012 in 2007 (the survey year). All survey values were therefore multiplied by 3,177,774/1,547,012=2.054137.

Consumption data were converted from local currencies to international dollars adjusted for purchasing power parity (PPP$).

Step 5: Review and validation:

Data was compared with other sources, notably the respective survey reports, and the World Bank’s poverty dataset, Povcalnet.

Step 6: Production of summary tables and metadata:

The World Bank generated of a standard set of tables for each country showing consumption and demographic patterns across consumption segments.

For more information on the Global Consumption Database methodology see: http://datatopics.worldbank.org/consumption/detail under the ‘Standardization of Data’ tab.

As the World Bank’s Global Consumption Database draws on a variety of country surveys which differ in design, methodology, and timing, there are limits to the extent to which surveys can be standardized. Therefore, cross-country comparisons should be made with caution. For more information see http://datatopics.worldbank.org/consumption/detail under the ‘Note on comparability’ tab.

All figures reported are based on national totals. The World Bank notes “each survey is composed of ordinary households only; “institutional households” (prisons, military barracks, hospitals, convents, and others) are not covered by household surveys. Homeless and nomadic populations and visitors present in a country during a survey are also excluded from the sample.”

The surveys used in the database were conducted between 2000 and 2010. For more information see http://datatopics.worldbank.org/consumption/detail under the ‘Sources of Data’ tab.

Sources

Data Published By: World Bank Global Consumption Database

Data publisher source: National household consumption or expenditure survey datasets. For a comprehensive list see: http://datatopics.worldbank.org/consumption/detail under 'Sources of Data' tab.

Link: http://datatopics.worldbank.org/consumption/