A New Approach to the Estimation of Quarterly GDP:Toward production of "Flash" Estimates
(Provisional translation)
May 1999
[Executive Summary of the Report by the committee on earlier dissemination of GDP data]
Chapter 1 Purpose of the Committee
Quarterly national accounts aim to provide a measure of quarterly changes in macroeconomic aggregates. They allow economic agents to study business cycles and to measure the effects of economic policy. But in Japan the first preliminary quarterly estimates of GDP is released around two months and ten days after the end of the quarter. Many people point out that the preliminary estimate is mere "a picture of the economy during a past period." And due to the long delay, it is said to be of limited use for monitoring and analyzing the current developments of the economy.
Among the member countries of the OECD, the United States releases "the advance estimates of GDP" about one month after the end of the quarter, while the United Kingdom does so about three weeks after around. The European Union plans to publish GDP estimates for the EU as whole within 40-45 days of the end of the quarter in question.
Against these backgrounds, a committee on earlier dissemination of GDP data was established in July 1998, as an advisory body to the Director-General of the Economic Research Institute of Economic Planning Agency. Its purpose is to investigate the feasibility of releasing quarterly GDP more early. Since then, the committee has met several times to discuss mostly technical issues, centering on the development of alternative methodologies and timing of publication.
As a result, the committee has concluded that the timing of publication of the first preliminary quarterly estimates of GDP should remain unchanged because sufficient principal source data for the first preliminary estimates are not available. Instead, the committee has formulated what the members discussed into six recommendations, including production of "flash estimates" of quarterly GDP.
Chapter 2 Six Recommendations
Recommendation I: Production of Flash Estimates of Quarterly GDP Approximately One Month and Ten Days After the End of the Quarter in Question
There is currently a long delay before the first preliminary estimates are published. The Committee recommends that flash estimates, which represent the earliest picture of the economy according to national accounts concepts, should be produced approximately one month and ten days after the end of the quarter. And we don’t think that estimates should be named "preliminary estimates" or "QE", because they are based on more incomplete information than the first preliminary estimates. In addition, estimates for gross fixed capital formation should be substituted for separate estimates for residential investment, non-residential investment and public investment because primary souse data, such as Quarterly Financial Statements of Corporations, are not available at the time when the flash estimates are to be compiled.
Recommendation II: The Need to Publish Relevant Basic Statistics More Timely in Order to Disseminate the First Preliminary Estimates More Quickly
The Committee concludes that the timing of the dissemination of the first Preliminary Quarterly Estimate of GDP should remain unchanged because sufficient principal source data for earlier estimation of the first preliminary estimates are not available. The Committee strongly requests relevant ministries and agencies that basic statistics needed to compile the first preliminary estimates be provided more quickly and that the publication schedule be released at the beginning of the year.
Recommendation III: Publication of the Second Preliminary Estimates approximately Four Months After the End of the Quarter in Question
Under the current practice, the second preliminary estimates for the quarter in question are to be published about five months and ten days after the end of the quarter, at the time when the first preliminary estimates for the following quarter is to be released. But the Committee concludes that the second preliminary estimates for the quarter in question should be published about four months and ten days later, at the time when the flash estimates for the following quarter are disseminated.
Recommendation IV: Disclosure of Methodology of Quarterly National Accounts and Accountability for Revision to the Already Released Estimates
Because economic statistics, such as national accounts, are public goods, everybody has a right to access the relevant information. The Committee recommends that the SNA authorities should disclose its methodology concerning quarterly national accounts. In addition, the Committee recommends that the SNA authorities should not only publish figures used in the estimates but also explain reasons for revisions. The SNA authorities need to make every effort to maintain credibility concerning their national accounts estimates.
Recommendation V: Improvement of the Estimation Method and Enrichment of the SNA Authorities
The Committee mustn’t forget that there is no perfect estimation method. The sources and methods used for flash estimates are very different from those for the first preliminary estimates with respect to gross fixed capital formation in particular. Therefore the SNA authorities should make every effort to improve methods for the flash estimates and heighten the accuracy of those estimates.
In addition the staffing in the SNA authorities in Japan is inferior to that of other OECD member countries. The Committee recommends that the institution that compiles quarterly GDP estimates in the country should be enriched.
Recommendation VI: Remaining Problems to be Addressed in the Long Term
The Committee recognizes that there are remaining issues, including monthly GDP and quarterly GDP by the production approach, that need to be addressed in the long term.
Revision Policy for quarterly GDP estimates
Flash estimates of quarterly GDP are prepared in approximately one month and ten days after the end of the quarter in question. And they are revised a month later (the first preliminary estimates), and are then again revised two months later (the second preliminary estimates). No further revisions are until the annual revisions, which usually are made in December and cover the two most recent years that are completed. Following the second annual revision, no further revisions are made to the estimates until the comprehensive revisions (often referred to as "benchmark revisions",) which are usually made every five year.
Chapter 3 Some Remarks on Flash Estimates of quarterly GDP
The Committee recommends that SNA authorities should stress the following points when they release the flash estimates.
(I) Implication of Limited Availability of Source Data on the Characteristics of Flash Estimates of quarterly GDP
National accounts are secondary statistics, which are based on source data. Therefore the more the source data are available, the more accurate become the estimates and, consequently, the closer the estimates come to true values.
The less available source data are, the less accurate the estimates become and, consequently, the character of the estimate nears flash figures of the first preliminary estimates.
The Committee concludes that estimates made approximately one month after the end of the quarter have two characteristics. They are "actual results" and "flash figures." Because some components of GDP is estimated by using alternative methods and data source. The estimates for gross fixed capital formation and private inventory investment in particular can be characterized as flash figures of the first preliminary estimates. Therefore revisions to these estimates could be large, at the time of releasing the first preliminary estimates.
With regard to private final consumption expenditure and export and import of goods and services, however, estimates made approximately one month after the end of the quarter can be characterized as "actual results" because the estimation methods are the same and almost all source data are available at that time.
(II) Relation between Accuracy of Flash Estimates and the Level of Aggregation for Presenting Flash Estimates to the Public
There is always a trade-off between timeliness and accuracy. While there is substantial demand of users for fresh information, there is the risk of releasing wrong data which would misguide economic agents or market participants.
How to take the right balance between the two varies among countries. The Committee observes that the United States and the United Kingdom put more weight on timeliness while Canada, France and Germany give more weight to accuracy.
The Committee concludes that in spite of their less accuracy, flash estimates are of great use as an instrument in detecting the phase of economic cycles, highlighting peaks and troughs and suggesting policies or measures that should be taken to achieve short-term quantitative targets for main macroeconomic variables.
But since errors in the flash estimates would relatively be large, the Committee recommends that for the flash estimates higher level of aggregation for presentation of concepts should be taken than that for the first preliminary estimates.
The Committee recommends that the following presentation forms should be considered with regard to the flash estimates.
Form 1: Percentage changes from the previous quarter in seasonally adjusted estimates at constant prices by component;
Form 2: Percentage changes from the same quarter of the previous year in original estimates at constant prices by component.
And as a supplement, percentage changes from the previous quarter in GDP as a whole at constant prices.