Research Agenda: Compilation of PPP Time Series and the Rolling Survey Approach

Compilation of PPP Time Series and the Rolling Survey Approach

Extrapolations of PPPs at the aggregate level over long periods intervening the benchmark comparisons, coupled with major methodological innovations introduced in the benchmark comparisons, severely limit the ability to reconcile extrapolations from different benchmarks. This was certainly true in the case of the 2005 ICP results published after a gap of 12 years from the 1993 comparisons; and a relatively smaller gap of six years in the case of the 2011 and 2005 ICP comparisons. The significant differences between the extrapolations and the benchmarks led to significant revisions to estimates of the size of the world economy in real terms and its distribution across regions. The World Bank had to make major revisions to its estimates of global poor and global inequality in 2005 and 2011.

Inconsistencies between benchmark and extrapolated results arise due to a number of factors including changes in expenditure weights, commodity baskets, aggregation methods, linking regional comparisons and other considerations (Deaton and Heston, 2010; McCarthy, 2013; Annex D by Inklaar and Timmer to McCarthy, 2013; Deaton and Aten, 2017; and Inklaar and Rao, 2017 for details). A way to resolve this issue is to conduct more frequent comparisons as is the case with the Eurostat-OECD PPP program.

This issue of inconsistences between benchmarks was given prominence in the recommendations of the UN Statistical Commission:

“One of the key conclusions of the initial report on the 2011 round was that a six-year interval between ICP rounds was too long. This view was expressed by almost all stakeholders during the evaluation process and was also reflected in the reactions at the session of the Commission. It was expressed that the acceptance and relevance of ICP results depend on their frequent and timely availability (at least every two to three years with extrapolations to annual results). The Commission at its forty-sixth session expressed its preference to shorten considerably the interval of ICP worldwide comparisons with the intention of having the next benchmark occur as soon as possible, preferably in 2017”. (Paragraph 45, E/CN.3/2016/9, UN, 2016)

Establishing a framework for the production of PPPs and international comparisons of GDP and its components on an annual basis is a complex and challenging task. The Global ICP Unit has taken this task seriously and had undertaken some preliminary work in this direction. In the process, the Global ICP Unit has identified four major components to its work program as it moves towards the implementation of rolling price surveys and the compilation of annual time series of PPPs. 

Component 1. Moving Towards Rolling Price Surveys  

[see Task Force on Country Operational Guidelines and Procedures

The notion of rolling price surveys is not entirely new to ICP. During the implementation of the 2005 and 2011 ICP benchmarks a pseudo-rolling approach, whereby prices collected in different years were moved backwards or forwards using appropriate deflators, was used. While this was commonly used in different regions, the approach was somewhat ad hoc and procedures were crafted on a case by case basis. The current thrust towards a more formal approach to rolling price surveys will ensure a consistent approach across all regions and thereby enhances quality of the comparisons.

The rolling survey approach, to be implemented starting in 2017, involves spreading the price data collection over a three-year time period around the reference year. PPPs for the basic headings not based on the reference year are to be extrapolated to the reference year. The Global ICP Unit identified the following data related issues:

  • Determine the frequency of data collection (annual, biannual, quarterly) for the rolling survey for each aggregate of the GDP;
  • Assess availability and quality of national account deflators, CPIs and other price indices by GDP component to be used in the extrapolation to the reference year;
  • Determine how to link PPPs from regions with different timetables for data collection; and
  • If a three-year rolling survey approach is implemented then it is necessary to assess methods of extrapolating PPPs for the two years between individual benchmarks, before the full rolling survey method is implemented (i.e., when PPPs are computed every year). This is also related to Component 3 below.

Component 2. Linking Interim Regional Updates into a Global Comparison

[see Task Force on PPP Compilation and Computation

Since the completion of the 2011 ICP, various regions conducted interim updates. However, these regional efforts were not mandatory and largely depended on the initiatives of the Regional Implementing Agencies. The Global ICP Unit faces the challenge of linking these regional updates and make them useable in compiling meaningful global comparisons. The following issues need to be investigated and addressed:

  • Effect of different timings of regional surveys ranging from 2012 to 2016;
  • Impact of different degrees of coverage of the GDP;
  • Effect of changing regional composition and of countries not participating in the regional interim updates at all; e.g., several Arab countries from Africa are also participating in the Western Asia comparisons, whereas China is not part of the Asia update;
  • Effect of discrepancies in global core list coverage by region on inter-regional linking; and
  • Effect of methodological differences between regions on the linking process.

Component 3. Building PPP Time Series for the Interim Period 

[see Task Force on PPP Compilation and Computation]

As work on 2017 ICP is underway, the Global ICP Unit recognizes the need to fill the gap for the intervening period. The current timetable is to release comparisons for the years 2012 to 2016 around the same time as the release of results from 2017 ICP. Simultaneous release of the 2017 benchmark and comparisons for the intervening years will avoid scrambling by users to fill this gap by themselves. Thus, the PPP time series will be built using the 2011 benchmark results along with data generated by regional interim updates (see Component 2 above). The main tasks involved are:

  • Assess availability and quality of national account deflators, CPIs and other price indices by GDP component (as in Component 1 above);
  • Examine consistency of CPI weights and national account weights, and assess the weights’ effects on PPPs;
  • Suggest steps for statistical capacity-building to ensure that CPIs and national account deflators are consistent over time and comparable across countries; and
  • Ensure consistency between the 2011 and 2017 results and the comparisons for the intervening years.

Given the tight timeframe for achieving these goals, the Global ICP Unit has initiated work in this direction. Progress has been made in setting up an extrapolation database and establishing procedures to standardize the data provided by the countries through their respective regional implementing agencies. Detailed time-series data for expenditures, deflators, CPIs, population and exchange rate data are being collected and consistency checks are being conducted. A bi-annual submission schedule was agreed at the 2nd Inter-Agency Coordination Group (IACG) meeting. All the six regions (Africa, Asia and Pacific, CIS, Eurostat-OECD, Latin America and Caribbean and Western Asia) have reported data, but with differing degrees of coverage and quality. The unbalanced reporting from countries and sparse coverage of detailed time-series data for many countries call for careful integration of data in the compilation of time series of PPPs and in ensuring the quality of linked comparisons.