SiteMap •  Links • Search

Revision of Official Statistics – Is It Cheating?

Concerns were raised, quite understandably , about the revision of foreign trade statistics released recently by the National Statistics Office (NSO). Some of these concerns were articulated objectively, but some, unfortunately, degenerated into convoluted accusations about cheating by government statisticians!

On August 2, 2005, the NSO released revised import figures for 2002-2004 : from 35.4 million to 39.2 million in 2002, 37.5 million to 40.5 million in 2003 and 40.3 million to 44.0 million in 2004, all in US Dollars, valued free-on-board (f.o.b.) or at exporter's customs frontier. Seemingly unknown to the misinformed accusers, on April 22, 2003, the NSO also released revised import figures for 2000-2002: from 31.4 million to 34.5 million in 2000, from 29.6 million to 33.1 million in 2001 and from 33.6 million to 35.4 million in 2002², valued f.o.b. in U.S. dollars! The first set of revisions covering the period 2000-2002 was based on an initial survey-based methodology approved by the NSCB on January 15, 2003. The second set, corresponding to the 2002-2004 revisions, which ranged from more than 5% to almost 12 % was based on a second methodology approved by the NSCB on May 18, 2005. This was the result of the work done by a Task Force created by the Monetary Board in June 2001 and officially tasked by the NSCB on January 7, 2002 to study the issue on the under/over valuation of exports/imports. The initial recommendations of the Task Force³ which was composed of representatives from the NSO, the BSP, the Bureau of Customs, the PEZA, NEDA and the NSCB, were based on a survey of 15 electronic companies⁴ jointly conducted by the NSO and the BSP. These companies accounted for 90% and 75% of the total output of the electronic industry for 2000 and 2001, respectively. The Task Force benefited from the cooperation and collaboration of the Semiconductor and Electronics Industries in the Philippines, Inc. or SEIPI.

In fact, one of our most intelligent, hard-working and truly dedicated congressmen, Cong. Joey S. Salceda from Bicol, also called the attention of the NSCB on the trade statistics issue in early 2003. If I may add, Cong. Salceda did this in a very professional manner and since then, the NSCB has constantly benefited from his wisdom, expertise and challenges.

Revision of official statistics is standard fare recognized and appreciated by those who truly know! Not revising official statistics certainly cannot count among the best practices of a national statistical system. And to the credit of the Philippine Statistical System, even before the IMF got to pursue its advocacy for statistical offices to come up with revision policies, the NSCB thru Resolutions No. 8 Series of 1997 and No. 7 Series of 2005, approved policies for the revision of the national accounts and agricultural statistics, respectively. Thus, the NSCB undertakes a series of revisions on the quarterly national accounts: during the immediately following quarter, during the May round of estimation for each of the three immediately succeeding years and during overall revisions which are conducted generally after at least ten years. NSCB records show that from the first quarter of 1997 to the first quarter of 2005, the GDP growth rates had a mean revision of 0.1 percentage point and a mean absolute revision of 0.2 percentage point between the initial estimate and the first revision. For the GNP, the corresponding averages were 0.1 and 0.4 percentage point, respectively. For the revisions between the initial and the final/last estimate, the averages for the GDP were 0.1 and 0.7 percentage point: for the GNP, they were 0.0 and 1.0 percentage point. In the case of the U.S. , for the period 1983-2002, the mean revision on the real GDP estimate was 0.51 percentage point between the initial and the first revision and 1.29 percentage points between the initial and the latest estimate⁵.

Revision of official statistics is NECESSARY AND DESIRABLE for a number of reasons.

First, towards improving the timeliness of official statistics, these statistics are generally released based on incomplete data/reports. Statistical offices used to be and still are being ridiculed on the long time lag in the release of statistics – aanhin pa raw ang damo kung patay na ang kabayo! We of course agree that statistics with the wrinkles of your favorite mother-in-law are useless. So, even if not all the supporting data have come in, we release official statistics based on best efforts and on methodologies that follow international practices and standards. For example, as many readers would know, the NSCB releases the quarterly national accounts two months after the reference quarter, except for those for the fourth quarter which are released one month after yearend. Thus. the quarterly GDP estimates are based on data available to the NSCB as of the estimation period deadline, which are almost always incomplete. Later on, after the NSCB has released the estimates, additional data would be submitted which could lead to revisions of the preliminary estimates. A necessary consequence of this is the use of standard imputation techniques and judgmental estimates based on available indicators as part of the methodology! And may we be reminded that one reason for the incompleteness of the data is the delayed submission by companies of the accomplished survey questionnaire to the NSO! Talking about good corporate citizenship and corporate social responsibility, I wish the top management of each company would appreciate, understand and live the importance of their cooperation in the statistical surveys and the administrative records they are being asked to submit to government bodies. Only then, could we look forward to less revisions in official statistics!

Second, official statistics is an evolving science that must respond to changes in the economy and society, its structure and its priority areas as well as to developments in the global community. As a result, concepts, definitions and classifications change and international guidelines are revised and updated. For example, in the compilation of national accounts, the 1968 UN SNA ( United Nations System of National Accounts) has been superseded by the 1993 SNA. Necessarily, when a country shifts from the 1968 UN SNA to the 1993 SNA, national accounts estimates will be revised comprehensively. Two weeks ago, the author attended an Experts Group Meeting on Distributive Trade Statistics in New York upon the invitation of the United Nations Statistics Division. The task of the Experts Group is to recommend changes in the so-called M57 Series of the UN, the Principles and Recommendations on the Compilation of Distributive Trade Statistics. One change being envisioned is the exclusion of the repair of household goods from the coverage of distributive trade, which would result in the revision of back data series on wholesale and retail trade.

Third, improvements in statistical methodologies and the need of researchers and econometric modelers for comparable, consistent and long data series require the statistical offices to revise past estimates computed using old methodologies, even if this is not always possible due to resource constraints. In the statistical profession, we recognize that many, if not all methodologies for generating statistics can stand improvement. When we change these methodologies, we are aware that the unenlightened few may make villains out of us, and accuse us of every conceivable sin to cheat for the good image of the current administration! Just like what happened when we revised the methodology for poverty assessment! But, we are professional statisticians, we will not be deterred and we will not be cowed. Surely, our accusers could have never been my students at the UP Statistical Center ( now the UP School of Statistics ) , but if they were, I might have had to invent a grade not as high as 5!

Fourth, revisions are inherent in some statistical tools. For example, in the seasonal adjustment of time series, when forward seasonal factors are used, they change the estimates when new data points are added to the series, even if there are no revisions in the other parts of the series. Or in constant price estimation, a change in the base year will automatically cause revised estimates of levels.

And fifth, of course human errors do occur and unless we are dishonest, we will revise our estimates to correct the errors. Anyone who thinks that such human errors should never happen, is out of his/her mind and our enjoyably imperfect world would be so much better off without them!

Revisions of official statistics happen everywhere. In the U.S. , one needs only to browse the website of the Bureau of Economic Analysis (www.bea.gov ) to learn about the history of revisions the BEA has made on the national income and product accounts of the U.S, many of which are not statistically different from zero. On the U.S. Consumer Price Index, the Boskin Commission Report⁶ is probably still fresh in the minds of many, which led to the sixth comprehensive revision of the U.S. CPI! In New Zealand , they have a Revision Advisory Committee that formulates recommendations on the revision of different data series produced by Statistics New Zealand. In Australia , since its first compilation in 1960, the CPI has been reviewed and reweighted thirteen times!

Yes, indeed, revisions of official statistics happen even in the richest countries of the world, with the most developed statistical systems, and with a political will much stronger than ours to invest the necessary resources on statistics!

In recognition of the need to revise statistics, manuals and guidelines on revisions have been published by the United Nations Statistics Division, the Eurostat ( the statistical office of the European Union ) and the IMF. In fact, the IMF is currently preparing its recommendations on revision policies for the national accounts. It ought to be mentioned that the Philippines is one of the original subscribers to the Statistical Data Dissemination Standards or SDDS of the IMF. Subscription to the SDDS is limited to countries whose statistical systems are willing and able to abide by the IMF-recommended guidelines on the production and generation of a number of important statistics, such as the national accounts, prices, balance of payments and labor force statistics, among others. The recommendations include the generation of quarterly national accounts estimates one quarter after the reference quarter ( some countries like India did not produce quarterly national accounts until recently), the release of the monthly CPI one month after the reference month and the release of quarterly BOP data one quarter after the reference quarter. The Philippines is more than in full compliance with respect to these three data series. And no matter what we think of the IMF, it must be admitted that these recommendations are essentially sound and are meant to improve statistical practices all over the world. If a country's statistical system cannot cope with the IMF recommendations ( I never liked to refer to them as IMF requirements!) subscription is not accepted or recognized even if the country wishes to subscribe. True to the IMF character, unfortunately, the IMF has used subscription to the SDDS as a precondition to access to the international financial market! With such behaviour, how can we honestly talk about halving poverty by 2015 in next week's Millennium Summit?

Having despaired, we would not lose however, and we would like to reassure our public that there are many of us, professional statisticians in government, who will continue to live up to the standards of our profession. Under the difficult circumstances we face amidst the resource constraints of the statistical system, we will continue to do what we think is best, to maintain our position among the more respected statistical systems in the world. And that, as sure as the tilaok ng tandang ni Tatang in the morning, or the exquisite Manila sunset along Roxas Boulevard at dusk, includes revisions, revisions and more revisions!

 

Reactions and views are welcome thru email to the author at ra.virola@nscb.gov.ph.

 

 

___________________________

¹ Secretary General of the National Statistical Coordination Board (NSCB) and Chairman of the Statistical Research and Training Center (SRTC). He holds a Ph. D. in Statistics from the University of Michigan in Ann Arbor , U.S.A. and has taught mathematics and statistics at the University of the Philippines . He is also a past president of the Philippine Statistical Association.

² Note that the import figure for 2002 was revised for the second time!

³ On March 5, 2004, the Task Force was elevated by the NSCB into an InterAgency Committee on Trade Statistics one of whose mandates is to seek improvements in the generation and dissemination of trade statistics including distributive trade. This IAC is presently chaired by Margarita Songco, the first woman Deputy Director General of the NEDA.

⁴ The Task Force decided later on that the survey was not sustainable and the methodology it initially recommended was revised.

⁵ Fixler, Dennis. Revisions to GDP Estimates in the U.S. http://www.bea.doc.gov/bea/papers/fixler_gdp_revise.pdf . Note that the percentage point revisions cited for the U.S. are in terms of changes in levels; for the Philippines , the percentage point revisions refer to changes in growth rates.

⁶ The 1996 report of the Advisory Commission to Study the Consumer Price Index. The Commission was appointed in 1995 by the U.S. Senate Finance Committee and chaired by Michael J. Boskin.

Posted 12 September 2005

 

Statistically Speaking

Main Page

Related Links

Press Releases

Announcements

For the Record

E-Newsletter

Media Services

Events

 

	E-mail the webmaster	Terms of Use					Home • Top of Page
		1997-2010, National Statistical Coordination Board Makati City, Philippines