Behind the Numbers U.S. Trade in the World Economy (1992) / Chapter Skim
Currently Skimming:
Appendix E: An Alternative Seasonal Adjustment Procedure for Merchandise Trade Data
Appendix E: An Alternative Seasonal Adjustment Procedure for Merchandise Trade Data
Pages 237-244
The Chapter Skim interface presents what we've algorithmically identified as the most significant single chunk of text within every page in the chapter.
Select key terms on the right to highlight them within pages of the chapter.
From page 237... ...
For example, the mean monthly merchandise trade balance for 19851990 is about -$10.5 billion, the mean monthly export figure for the period is $23.5 billion, and the mean monthly import figure is $34.0 billion: the difference is a substantial proportion of the means. We have considered alternative methods of seasonally adjusting the data because an initial analysis of the data leads us to believe that the X-ll seasonal adjustment procedure currently used by Census Bureau not only fails to remove much of the seasonal noise from the data, but also adds more noise to the data.)
|
From page 238... ...
Once we have the optimal decomposition at the optimal variance parameter values, the model-based seasonally adjusted estimate is simply the nonseasonal component.3 The first model-based seasonally adjusted trade balance is given in Figure E-3. The graph is considerably smoother than the X-11 output.
|
From page 239... ...
For yearly change, the X-ll procedure yields a series with a standard deviation of $2.7 billion; the model-based procedure yields a series with a standard deviation of 2.1 billion. The standard deviation of the first difference and yearly difference fall by about $0.1 billion each when historical data are used.
|
From page 240... ...
240 / APPENDIX E 4 , o 4 -8 -12 -16 I 1 1 ~ I 1 1 1 1 1 1 1 1 1 1 1 1 975 1 977 1 979 1 981 1983 1985 1987 1989 1991 Year FIGURE E-1 Merchandise trade balance, not seasonally adjusted. 4~ o 4j 6a so m -8 -12 -16 1975 1977 1979 1981 1983 1985 1987 1989 1991 Year FIGURE E-2 Merchandise trade balance with X-ll seasonal adjustment.
|
From page 241... ...
_ ._ m -4 -8 -16 ~ ~ 1 ~ ~ ~ 1 ~ 1 ~ 1. ~ ~ ~ 1 ~ :~\ 1 ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ I ~ ~ / \J 1975 1977 1979 1981 1983 1985 1987 1989 1991 Year FIGURE E-4 Merchandise trade balance, seasonally adjusted with historical data.
|
From page 242... ...
265.6 X-11 801.0 1,120.7 1,229.2 Ratio: optimal/X-11 0.443 0.493 0.216 Concurrent Filter Optimal 460.7 703.5 287.0 X-11 901.4 1,191.1 1,258.6 Ratio: optimal/X-11 0.511 0.591 0.228 continued that the optimal model-based filter does significantly better than the X-ll procedure: for the homoskedastic model-based procedure using monthly figures, the root mean square error of the model is just 0.43 as large as that for the X-ll procedure; for the monthly change figures, the model-based procedure is 0.48 as large; finally, using yearly change data, the model-based procedure is 0.19 as large. Evidently, much of the variability in the trade balance data is created by an adjustment procedure that is extremely nonoptimal.
|
From page 243... ...
Thus, using only data available at the time of the initial announcement of the trade balance produces a seasonally adjusted series that has considerably less variation than the Census Bureau's X-ll procedure, which uses historical data. Considerable room for improvement thus exists in the X-ll procedure used for trace balance data.
|
From page 244... ...
We expect that the resulting weighted average would eliminate much of the variability before the revisions are made.
|
Key Terms
This material may be derived from roughly machine-read images, and so is provided only to facilitate research.
More
information on Chapter Skim is available.