Economic Research Institute
Economic Planning Agency
We welcome recent active discussions and critical remarks on Japan's GDP figures, because they show a wider attention to the state and the measurement on Japan's economy. Some of the criticisms, however, seem to be based on the insufficient information on the process of making GDPs and on the nature of GDP figures.
The main points of these opinions are as follows: (1) GDP figures are too volatile, and there may be problems on the method of seasonal adjustment. (2) GDP figures have often be subject to large revisions. (3) Large discrepancies exist between GDP figures and monthly supply-side figures, such as IIP (Indices of Industrial Production). (4) Current structural changes are not taken account of in the process of calculating GDP figures, for example, investment on information technology (IT) is not included in fixed capital formation, and the expenditures of single-person households are excluded from consumption. (5) The improvement of quality is not considered for the deflator, and the real value may be underestimated.
We would like to clarify some of the points about our current estimation methods.
(1)The volatility of the GDP figures, and the method of seasonal adjustment
The following may be some causes of the volatility of GDP figures: 1) The current economic condition itself is volatile; 2) The basic statistics do not reflect the economic condition accurately because they own the statistical bias; and 3) The factor of seasonal adjustment.
As for the method of seasonal adjustment, the U.S. Census Bureau X-11 system is applied. The X-12-ARIMA system, which is developed as the new seasonal adjustment system in the U.S. Census Bureau, has not been introduced by various reasons. For example, on private consumption series -- the main component of GDP -- statistically significant results have not been obtained that a leap-year or trading-day effect had existed. The results of these studies have been press-released in November 1997. (*)
In short, it is not necessarily true that the more stable the data are, the better the data. We are trying to achieve the maximum accuracy of data with the minimum lags of delivery.
(*) In addition, on the stability of the seasonal adjusted series, both well-developed and worse-affected series are found using the X-12 system.
(2)The revision of GDP figures
The preliminary GDP figures are revised annually every December ('Annual Estimation'). The annual GDP figures are estimated by the Commodity-Flow method. The method is, in contrast with the preliminary estimation method from demand-side statistics, based on the annually available supply-side statistics that cover economic activity widely and contain the thorough aggregation of the each commodity's items. The method requires more time, too long time lags for the Preliminary Estimates, but provide more accurate and detailed information. (**)
As will be explained in (3), the monthly supply-side statistics in Japan possesses some limitation. So it is the best way to use the monthly or quarterly demand-side statistics that contain rich information for estimating quarterly GDP figures under the current conditions of economic statistics. However, in the case that monthly supply side data are available, such as those of new car sales and medical services, supply side statistics are used in estimating the preliminary private consumption.
Moreover, on the annual estimation, the seasonally adjusted series are re-estimated using the new seasonal index calculated from the new original series of added four-quarter data. This has caused the revision of seasonally adjusted GDP figures since 1955.
Incidentally, as for the difference of real GDP between preliminary estimates and annual estimates, while the average difference on the quarterly growth rate of Japan is larger than that of the U.S., the average difference on the level of real GDP on Japan is almost the same as that of the U.S. according to our calculation.
(**) The preliminary estimates are calculated mainly from the demand-side statistics, but supply-side statistics are also taken into account. In detail,
- The preliminary estimates are calculated by extrapolating the benchmark of the previous year, which are estimated on the supply-side statistics.
- On the other hand, the year-over-year change rates that are calculated from the demand-side statistics in general are used as extrapolating rates.
(3)The discrepancy between the GDP figures and monthly product-side statistics
The reasons for the discrepancy between the preliminary GDP figures (demand estimates) and monthly product-side statistics may be as follows:
- Aside from the fact that preliminary GDP figures are estimated using mainly demand-side statistics, while the product-side statistics are supply-side data, preliminary GDP figures are based on value added statistics, while the product-side data are on the output volume.
- Monthly product-side statistics, such as the Index of Induced Production, are estimated as the weighed average based on 1995's fixed weight. So there exists a discrepancy between this weight and the current commodity weight on the demand-side statistics, while preliminary GDP is composed on the current commodity weight.
- Product-side statistics such as that of consumer services are hard to cover comprehensively because of the lack of data.
(4)The changes in economic trends and GDP figures
With respect to the Information Technology (IT) spending, the purchase of computer software by enterprises is counted as intermediate consumption in the current system of national accounts (1968SNA). Still, the software installed in computer hardware, however, is totally counted as investment in the current national account in Japan.
Moreover, with the introduction of the new system of national accounts (1993SNA) in Japan this October, order-made computer software purchased by enterprises will be newly counted as investment, as is software installed in computers at present. On the contrary, general-purpose or in-house-developed software will be counted as intermediate consumption because of the limit of basic statistics.
Concerning preliminary estimates of private consumption, as for consumption per household, 'Household Survey' for non-farming family and statistics for farming families are used. On the other hand, as for the number of households, total number is estimated including single-person households.
(5)On the deflator
Some have argued that for the deflators of GDP figures in Japan, especially deflators on fixed capital formation, the improvement of quality is not considered, and that this causes the underestimates of the real values. However, quality adjustment is operated for the WPI, the basic statistics in the estimating deflator of fixed capital formation.
Incidentally, in the estimation of private consumption deflator, WPI is partly used as basic statistics such as the deflator of personal computers.
The pace of changes is so fast that we have to be alert every moment to make our statistics follow the vicissitude of the economy. Specifically, we established a "Working Group on Preliminary GDP Figures'' to study the following matters: (1) Estimation methods on private consumption, considering the availability of quarterly data on expenditure of single-person households which are planned to be newly released. (2) Revaluation of the seasonal adjustment methods. (3) Estimation methods of preliminary GDP corresponding to the introduction of the 93SNA system, which recommends counting the order-made computer software as investment.
We appreciate comments on Japan's GDP figures because they help us reexamine some of our concepts and procedure.