No. 37 — September 29, 2000

 

Weekly Review

DATA SOURCES AND SEASONAL ADJUSTMENT
--- by Arthur J. Alexander

Japanese statistics are becoming more accessible. Most of the commonly cited data series now can be downloaded from the Web sites of ministries and agencies or from the sites maintained by industry associations and other groups. As a general rule, the Japanese-language sections of Internet sites include more information than the English versions. However, there are variations among organizations. The Bank of Japan, for example, provides virtually the same information in both languages, while the Ministry of Posts and Telecommunications offers very little English-language material. Moreover, while in the past, analysts often had to wait days or even weeks for the latest information on trade flows or the gross domestic product, these statistics can be accessed within minutes of their release.

CD-ROMs - Several important economic data series are being published on CD-ROM. A disk containing the national income and product accounts, for example, comes with the Annual Report on National Accounts, which is published by the Economic Planning Agency. The CD-ROM contains more information than the printed report and includes time series going back to 1955 or to the beginning of a particular series. And, unlike the printed version, which is annotated for the most part in Japanese, the disk includes tables in both Japanese and English.

The data in the printed and the digital media are not published until the books are closed on the fiscal year. The National Accounts volume dated 2000, therefore, includes annual information through calendar or fiscal 1998. However, selected tables now are being added to EPA's Web site (http://www.epa.go.jp) as they become available, making some of the main figures accessible sooner.

Since 1997, BOJ has offered its Economic Statistics Annual only in a CD-ROM format. The disk includes most of the data series that had been in the annual printed edition and that are in BOJ's still-published Financial and Economic Statistics Monthly, with time series going back to the beginning of a particular entry. Most of the data are monthly. The information is presented in several sections, including financial data, very detailed information from the quarterly tankan (survey of short-term business prospects) and price statistics. The 2000 version, which has data through the end of 1999, came out in April 2000. BOJ's Web site provides information on ordering the disk. It costs about $40.

Of interest to those in search of data on earlier times, the Historical Statistics of Japan: 1868-1985 also is available on CD-ROM. The five-volume, 2,700-page printed edition was priced around $900. The disk costs about $40 and appears to contain all the information found in the printed version. One of the real bargains in the world of statistics, it can be ordered from the Japan Statistical Association (http://www.jstat.or.jp). This group also publishes many of the reports compiled by the Statistics Bureau of the Management and Coordination Agency, including the Japan Statistical Yearbook in both printed and CD-ROM versions. The Statistical Yearbook also is available at the Web site of the Statistics Bureau (http://www.stat.go.jp). All the tables can be downloaded in spreadsheet form.

I have copied the English-language versions of these three disks to my computer's hard drive, putting a comprehensive collection of Japanese economic information at my fingertips. On my computer, the national accounts data occupy 25.8 megabytes, the BOJ disk takes up 38.4 MB and the historical statistics require 66.7 MB for a total of 130.9 MB.

Seasonal Adjustment - The wealth of available information on Japan's economy can be blinding, obscuring the need for discrimination and effective analyses. Until recently, a weak point of much Japanese economic data was the method used to account for seasonal variability.

With the most modern seasonal adjustment routines built into many statistical packages, Japanese government ministries and agencies were negligent in their treatment of their own information. Because of inadequate seasonal adjustment, much analysis was based on comparing current information with year-earlier data. This procedure automatically removes the complexities arising from seasonal patterns while introducing a different problem.

A well-known statistical property of year-to-year changes is that turning points are displaced from their true positions. Consider, for example, a symmetrical, V-shaped 12-month decline of an economic variable followed by a 12-month recovery. For six months after the curve bottoms out, year-to-year changes will fail to disclose an upturn. Economic policy based on such an analytical method could be biased. A preferred method is to consider month-to-month changes, but doing so requires good seasonal adjustment.

BOJ was one of the first Japanese agencies to adopt the so-called X-12-ARIMA method for seasonal adjustment developed by the Department of Commerce's Bureau of the Census. The X-12 routine counts among its improvements the forecasting of data beyond the current period, thereby reducing the need for later revisions. Another key feature is its ability to adjust for the specific days in a week in a month or a quarter. For example, many production facilities begin the workweek slowly as employees return from a weekend closure. Data for a month with more than the average number of Mondays would seem to indicate a production slowdown when output just was following the normal pattern. Adjustments of the X-12 type are important for processes that vary within the week.

The X-12 routine also adjusts for differences in the number of working days in a period and such anomalies as leap years. In the past, the Ministry of International Trade and Industry did not take these factors into account in its mining and manufacturing production index. As a consequence, this important index, which covers 25 percent of the economy, was a poor gauge of current economic activity. Beginning in 2000, MITI improved its seasonal adjustment of this vital measure.

Even improved methods of seasonal adjustment will not correct for inadequate sampling procedures. Many Japanese statistical series still exhibit very high monthly variation following seasonal adjustment. In part, this may reflect actual behavior. For example, the liabilities of failed companies vary greatly due to the occasional major bankruptcy. However, such other data as the production index bounce around for unexplained reasons but probably as a result of poorly conceived or implemented survey procedures.

This problem can be resolved by using moving averages. However, to avoid displacing turning points, the moving average must be centered. That is, a five-month moving average, which often is adequate to remove random noise, should include the two months before a given month and the two months that follow. Of course, that means that several months must be forecast in order to come up with a figure for the current period.

This is where the ARIMA part of the Census Bureau's method comes in. Most statistical packages include this autoregressive, integrated moving-average process to generate predictions based on internal movements within a data series. Centered moving averages of seasonally adjusted data using these forecasts often reveal changes in trends, whereas the raw data show only noise and the year-to-year comparisons are bound to be out of date.

Recommended Reading - JEI Report filled a niche not satisfied by any other single publication. However, the gap can be partially filled. The Economist, for example, produces weekly reports on various aspects of Japanese business and the economy but cannot treat these issues with the regularity or the thoroughness of a specialized publication. Similarly, while the Financial Times delivers excellent daily reporting on Japanese economic and financial issues, Japan is only one of many countries covered. The Oriental Economist is a 16-page monthly newsletter focused on Japan, with news, analyses, interviews and what might be called insider gossip. It includes invited contributions as well as articles from its regular writers. The publication's disadvantage, however, is its monthly schedule, which can render it less than timely.

Many banks and other financial services companies operating in Japan produce regular reports for their clients. These often are of high quality, and they benefit from proximity to decisionmakers in government and business. However, since the goal of these firms' analyses is to be responsive to the immediate investment concerns of their customers, they are less likely to have the longer horizons of the pieces that often appeared in JEI Report &emdash; a unique and perhaps indispensable publication.

The views expressed in this report are those of the author
and do not necessarily represent those of the Japan Economic Institute.

This is the final issue of JEI Report.
All members of the JEI staff wish to thank subscribers to the publication
for their loyalty and support over theyears.

Issue Index aaaa 2000 Archive Index aaaa Subscriber Area aaaa Home