Expert Meeting on ICP Expenditure Data
World Bank, 3-4 February, 2006
Introduction
The agenda and list of participants are attached as annexes.
The meeting was concerned only with the expenditure weights for household final consumption expenditure (HFCE). Two topics were discussed:
· National accounts metadata that should be maintained by each of the Regional Offices and shared with the Global Office;
· Methods of editing the expenditure weights and resolving problems with implausible weights.
Metadata
Metadata for the expenditure weights should answer the following questions:
· Is household final consumption expenditure estimated independently or is it obtained as a residual by deducting other components of final expenditure from GDP estimated as the sum of value added?
· Are the weights for Basic Headings within HFCE based on a household budget survey? If so:
o When was the survey carried out?
o What households were covered?
· Are the weights for Basic Headings within HFCE based on a supply-use table?
· Are the Basic Headings on a national basis (i.e. expenditures by resident households both in the domestic territory and abroad), or on a domestic basis (i.e. expenditures on the domestic territory by both residents and non-residents)?
· Are there any exceptions:
o Goods and services that should be excluded but that are included in the expenditure weights?
o Goods and services that should be included but that are excluded from the expenditure weights?
Metadata maintained by the IMF for the GDDS/SDDS programmes cover all the above points.
The IMF and the Global Office will arrange to convert the existing GDDS/SDDS metadata records into a standard format and send the converted metadata records to the Regional Offices.
The Regional Offices will be required to:
· Verify with their countries that the metadata records are complete and up-to-date;
· Collect metadata in the same format for participating countries that are not covered by the IMF GDDS/SDDS programmes.
The Regional Offices will return the amended set of metadata records to the Global Office.
Procedures to be followed by the Regional (or Sub-Regional) Offices in editing the expenditure weights
Objective
The objective is to detect those weights for particular Basic Headings (BHs) in particular countries that appear unusual or unreasonable compared with the weights for those BHs reported by other countries in the group – Region or Sub-Region – to which that country belongs.
Sub-regional groupings
Editing should be carried out for groups of relatively homogeneous countries. Africa and Asia Pacific countries should therefore be grouped into sub-regions before editing; the Latin America, West Asia and CIS Regions are already quite homogeneous and do not need to be split into sub-regions.
Techniques for detecting anomalies
There are three basic tools for detecting outliers – common-sense, variability and correlations between expenditure shares and per capita income:
· Common sense. Common sense may suggest, for example, that a poor country in a warm part of the region should not have expenditure shares for footwear that are substantially higher than the shares reported by richer countries in a cooler part of the region;
· Variability. Variability tests usually involve identifying weights that are one, two or more standard deviations outside the average.
· Correlations. For some BHs, it has been well established either that per capita income and the expenditure share rise together (“luxuries”) or that expenditure shares fall as per capita incomes rise (“necessities”). Ranking countries by income levels can help to detect anomalies for these BHs.
Variability tests
A standard procedure is to calculate the coefficients of variation (CVs) of the country weights for each basic heading and highlight those which are particularly high. Very high CVs require that the weights reported by all countries are investigated. This CV test is combined with an examination of standard deviations in which countries are high-lighted if their weights fall outside one, two or more standard deviations from the mean.
Two additional procedures have been developed for ICP purposes:
o The Quaranta tables which are used to detect outliers among price observations can also be applied to the expenditure weights
o The Dikhanov tables have been developed specifically to detect outliers among expenditure weights. The Dikhanov tests are carried out at the BH level and can be run by the Global Office on BH shares submitted by the Regional Offices.
Correlations
The two graphs below show how a simple correlation approach can be used to detect anomalies in expenditure weights.
It is well established that expenditure shares on most food items decline as per capita incomes rise. The graphs below show expenditure shares on rice for two sub-regions within the Asia Pacific Region. Countries in each group have been ranked by per capita income from the richest to poorest so that expenditure shares should be higher for countries towards the right-hand side of the graphs. In the South Asia Group, Pakistan’s weight appears to be too low and in the South East Asia group, the expenditure share for Vietnam should also be higher in view of its position on the income scale.
Ranking the countries can be done using World Bank estimates of per capita GDP or using shares of expenditure on food and non-alcoholic beverages as a proxy for per capita income (High shares indicate low per capita income)
The graph below shows expenditure shares on actual and imputed rents of dwellings. The 1999 OECD-Eurostat comparisons covering 45 countries showed a clear positive correlation between per capita income and expenditure shares – higher shares as incomes rise. The graph below shows no discernable pattern, however, and this suggests that all the expenditure shares should be reviewed.
Note that correlation tests are not applicable to all items of household final expenditure. They can be used for most food items and for a few non-food items where studies have clearly established correlations between expenditure shares and per capita incomes. These include items such as culture and recreation, dwelling services and expenditures on private transport. For most non-food items however correlations between expenditure shares and per capita incomes are weak or non-existent.
Editing by the Regional Offices and Global Office
The Regional offices should carry out editing procedures of the kind described above for the Basic Headings. The Global Office will make checks at the level of aggregation agreed for publication.
Note that the Regional Offices are expected to use their own judgement as regards the choice of editing procedures and may decide to use methods other than those described above.
Procedure when anomalies are detected
The Regional Office will notify the country when an anomalous weight is detected. There are then three possibilities:
· The country provides a plausible explanation as to why the weight is higher or lower than expected and the Regional Office accepts the weight as correct;
· The country agrees to modify the weight and the revised weight is considered acceptable by the Regional Office;
· The country refuses to adjust the weight but cannot provide a plausible explanation.
In the second case, the revision to the weight may result in a change to the level of the Gross Domestic Product. In this case, the country must agree to incorporate the revision in its official series of national accounts either immediately or at the time of the next revision of the official series.
In the third case, the Regional Office should not use the weight (or the accompanying price data) in calculating PPPs for that Basic Heading. The weight may either be discarded altogether or be assigned to a miscellaneous Basic Heading for which reference PPPs are used. The country will be informed.
Accountability
The Regional Offices are responsible for defending the PPPs, per capita expenditures and price level data which they publish for their Region. This means in particular that they have the right to reject expenditure weights that they find anomalous.
The Global Office has the same responsibility with regard to data published for the global comparisons. This means in particular that the Global Office has the right to exclude from the global comparisons countries whose data are considered to be of an unacceptable standard.