Accurate and early predictions of the current quarter's real GDP growth require perceptive use of the best and earliest source data. We at the Chief Economist's office of the U.S. Department of Commerce attempt to anticipate quarterly GDP growth prior to its advance release. Therefore, we have studied the source data for one of GDP's most volatile and influential components, producers' durable equipment (PDE). Because we do not get any information from the Bureau of Economic Analysis (BEA) before the time of public release, we have to use only publicly available information.(1)
What we have learned may prove of interest to other business economists. Recently, we estimated the change in 1997 fourth quarter real PDE to be 0.3 percent at an annual rate, when many analysts were expecting a growth rate of about 8 percent. (BEA's advance release reported a decline of 3.9 percent.) For 1998Q1, we predicted an increase in real PDE of 23.7 percent, not too far away from BEA's advance estimate of 28.8 percent.
We find that one should look at the available data on the Census Bureau's nondefense capital goods shipments excluding aircraft to estimate the quarterly percent change in nominal PDE excluding motor vehicles and aircraft ("core PDE"). Adding four extra shipments categories available in the M3 data release(2) and adjusting for exports and imports improve the prediction. To predict PDE aircraft expenditures, one should look at aircraft shipments as reported in Census's Current Industrial Reports, and for motor vehicles, BEA's monthly data on unit sales and average purchase value should be used.
WHAT BEA DOES(3)
In their annual summer revision, BEA adjusts nominal PDE components to reflect new data on about 350 product categories from the Census Bureau's Annual Survey of Manufacturers.(4) In between revisions, BEA extrapolates from these data by using monthly manufacturers' shipments from Census's M3 database. For each industry M3 category, a product ratio is established as to how much of that industry's production is included in PDE. The ratio is usually less than 1.0, because both intermediate goods production and consumer and government purchases of final goods are excluded from the total. Finally, the remaining shipments are adjusted to exclude exports and include imports. The final importance of a category is thus less if exports of that industry exceed imports (such as for communications equipment), but becomes greater where imports exceed exports (computers, furniture). The individual data must then be deflated by the appropriate price series to obtain real PDE. As might be expected, the huge decline every year in computer prices has a large impact on the total, so that real PDE after the reference or base year will always be greater than nominal PDE.
Table 1 shows almost all of the M3 industry shipments categories used in PDE. The numbers in parentheses attempt to mimic the product ratio and represent the rough percentage of M3 shipments in each category included in nominal PDE.(5) These percentages were calculated by dividing nominal PDE shipments by Census shipments in each category. They are higher than the actual category shares that BEA uses by about a fourth for two reasons: (1) PDE shipments are taken at purchaser value, which includes retail and wholesale margins plus transportation charges, whereas Census shipments are valued at lower factory value;(6) (2) small amounts of PDE shipments in most categories represent capitalized services.(7) However, these differences are roughly equal across industries.
The motor vehicles and aircraft components of PDE are treated differently in that M3 shipments data are not used. For motor vehicles, unit sales data from dealers are used, multiplied by an average vehicle price to get the total value. For aircraft, data on the value of shipments of complete civil aircraft are taken from the Census Bureau's Current Industrial Reports (available on their web site, www. …