Will the Real Economy Please Stand Up?

Business Economics, Jan, 2000 by Robert Parker

How do these new measures of GDP change the picture of economic growth and inflation in the current expansion compared to previous estimates? For the current expansion--through the second quarter of 1999, the new numbers show more growth. Previously growth had averaged about 3.1 percent per quarter, at an annual rate, and that's now revised up to 3.5 percent.

And inflation? Inflation is less. A lot of the lowered measure of inflation in our new numbers comes from the introduction of the geometric-mean measurement of consumer prices for 1978 to 1994. Correspondingly, of the 0.4% per quarter upward revision to growth in the current expansion, the early part is due primarily to downward price revisions, and the latter part is due to the combination of price revisions and other changes.

Which GDP components were stronger, and which were weaker over the course of the expansion? The revisions were widespread among the major components. Consumer expenditures were revised from 3.4 to 3.7 percent growth per year. Fixed investment from 7.8 to 8.1 percent. Government consumption and investment, a little bit more from 0.5 to 1.0 percent.

Can you speculate about what implications these upward revisions to real growth might have for labor productivity growth? Other things equal, these revisions should revise productivity growth upward. If the only revision was due to our measures, which is the numerator, one would expect 0.4 to 0.5 percent per year.

You added software and revised the treatment of banking services in these revisions. How much has software added to the level and to the growth of nominal and real GDP? Is the boost more important recently than earlier, and by how much? Software added about $125 billion to nominal GDP in 1996. In 1998, it was closer to $160 billion in current dollars. In 1987, it was about $45 billion. So it tends to be relatively smooth in nominal terms. I can't tell you right now exactly how much total software as investment added to real growth. For the major component of software, which is in private equipment and software, a lot of the growth came in the late 1980s and mid 1990s. It continues to grow, but at perhaps somewhat of a lesser rate in the late 1990s.

Will the recent pattern of growth in software--say up through the third quarter or maybe a little bit before that--show corporations' efforts to get Y2K compliant? I don't think we can identify Y2K efforts in these estimates. Y2K creates problems for the software estimate. With any equipment, you have maintenance and repair, and you have new investment. We cannot separate what Y2K has caused between new investment and maintenance. No doubt, government agencies and private companies completely redid some of the software that they've been using for many, many years. Clearly, that's investment. But where investment starts and changing a few codes in a program ends is just not something we know enough about right now.

Software does not include what is preloaded on PCs? Pre-loaded software has always been included in our estimates as part of the value of the PC.

So this is software in addition to that? That's correct.

What is meant by own-account software? Own-account software is that which a company develops in house with its own programmers and systems analysts. The alternatives are to buy pre-packaged software, or to hire a consultant or a software firm to come in and design it. The latter is custom software. But a lot of the mainframe software was developed in house, and is considered part of own-account software.

Regarding software in imports and exports, am I correct in thinking that you do not separately break it out, or perhaps even take account of it in those components? The trade and goods data for imports and exports do include software, but it is not separately identifiable on an ongoing basis. We know that it's being included, but in order to break it out requires us to get the Census Bureau to do special tabulations.

On pricing of software, I know you're using hedonic adjustment for software prices. But the prices, at least over the most recent years, are essentially flat. So they're clearly not behaving the way they are for equipment. Is there any difference in method between software and hardware hedonic prices? We use the hedonic index for the price of prepackaged software; it shows a decline. Not as much as the decline in computer prices, but a substantial decline. For the own-account software, we are using an input-cost approach to measuring the price. Lacking a comparable market price, we have to use, as we do elsewhere, an input-cost approach. We are not adjusting that for any changes in the productivity of programmers. And so that series tends to rise, as these costs rise, which of course, mostly are wage costs. For the price of custom software, we take a weighted average of the pre-packaged hedonic index, and the input cost, own-account index. And that tends to be flat. When you average the three together, you t end not to get a sharp decline, or much of a decline at all, in the price.