Scrutinizing industrial R&D data. (Perspectives).

The September 2002 issue of the Institute of Electrical and Electronics Engineers' (IEEE) Spectrum Magazine contained a special report on R&D. It was generally upbeat about industrial R&D spending, noting that, "the world's Top 100 Spenders increased R&D expenditures by an average

A former industrial R&D executive and I had a discussion about the current state of industrial R&D. He was skeptical about the Spectrum message because his peers provided anecdotal evidence of substantial layoffs and cuts in research divisions in 2002. He suspected that the true story was being underreported, in part because of the lack of timely data. Some of his colleagues insisted that he was getting a skewed perspective because his peers primarily come from the telecommunications industry, which has been hit hardest.

So, which perspective is more accurate? A variety of data sources track the magnitude and character of the R&D work to help answer the question.

R&D Data Sources

The National Science Foundation conducts an annual Survey of Industry Research and Development (www.nsf.gov/sbe/srs/indus/start.htm). A key advantage of the NSF data is the number of variables, in addition to R&D expenditures, it provides. The data are very useful for retrospective studies, but currently have two limitations: First, there is a significant lag time between the collection of the data and their presentation; e.g., the latest available data at the time of this writing were for calendar year 2000 (albeit there are indications that future NSF reports will lag only one year in reporting data). Second, the NSF data are aggregated by industry; disaggregated information at the company level would be more helpful to analysts.

Is there a good data source for more up-to-date R&D trends? This issue of Research * Technology Management includes the Industrial Research Institute's "R&D Leaderboard," which is based on Standard & Poor's Compustat database (see p. 21). The data show the changes from fiscal years 2000 to 2001.

The IRI also publishes the results of its annual member survey, "R&D Trends Forecast," in this issue (p. 17). And R&D Magazine publishes its own forecasts every January, based on work by Battelle and Schonfeld & Associates. There are also a variety of information/media companies that sell proprietary data.

Standard & Poor's bases its data on company financial statements that are reported to the Securities and Exchange Commission (SEC). Publicly traded firms are required to report their R&D expenditures in each reporting period. Financial Accounting Standards Board's FAS 2, which governs Accounting for Research and Development Costs, states, "Disclosure shall be made in the financial statements of the total research and development costs charged to expense in each period for which an income statement is presented" (2).

Public companies report R&D expenditures quarterly on their form "10-Q" and annually on form "10-K." Quarterly data provide an up-to-date and near real-time view of the industrial R&D landscape. Unfortunately, some of the largest R&D spenders only report data annually; e.g., Ford and GM, the two largest R&D spenders in 2001, do not report quarterly R&D data. It is also interesting to note that while there is a consensus among companies, investors and academics that R&D fuels future growth, the data are often difficult to locate in financial statements, or absent in the case of Ford and GM.

Current State of Industrial R&D

The Table, next page, shows R&D expenditures for the latest two quarters reported (e.g., Q3'02 + Q2'02), and those equivalent two quarters from the prior year (e.g., Q3'01 + Q2'01) (3). It then shows the change year over year for those two quarters. The fifth column shows the percent change in R&D from the IRI R&D Leaderboard data. The sixth column shows R&D Magazine's forecast for 2002. The last column shows IEEE Spectrum's percent change from 2000 to 2001. The companies chosen are the top nine R&D spenders according to the IEEE Spectrum article plus a manufacturer, Boeing, to add some industrial variety.

In every case except Pfizer and Boeing, R&D Magazine forecasts were too optimistic. In every case except Pfizer and Intel, extrapolating numbers from the IEEE Spectrum article overstates the actual 2002 R&D results.

The actual results show that 6 of the 10 firms have shrinking R&D expenditures in 2002; whereas, the R&D Magazine forecast predicted only 2 of the 10 would shrink, and the 2000-'01 Spectrum change shows only 3 of the 10 had shrinking expenditures. It is obvious that the other data sources paint a very different picture than the more recent quarterly data.

There are major discrepancies between the R&D numbers reported by the articles. For example, R&D Magazine reports Motorola's 2001 R&D expense as $5.4 billion but IEEE Spectrum reports it as $4.4 billion, and the financial statements report it as $4.3 billion. This is because the 2001 data from R&D Magazine is an estimate made by Schonfeld &Associates, but that is not clearly stated.

There is also a significant difference between the IRI data and the data reported in IEEE Spectrum, even though they are both based on the same data source. For example, Spectrum reports Cisco's 2001 R&D spending as $4.8 billion but the financial statements show this to be $3.9 billion. The explanation could be that Spectrum is counting in-process R&D ($855 million in 2001) as part of Cisco's R&D spending whereas, the IRI correctly subtracts out that figure from the S&P data.

In-process R&D is buried in the footnotes of the financial statements and the SEC requires that it be specified distinctly from R&D expenses. My accounting friends question whether in-process R&D should be lumped together with regular R&D expenses since they are accounted for in very different ways; i.e., in-process R&D includes allocations of goodwill from business combinations whereas R&D expense is actual costs incurred by the company (4). In other words, combining in-process R&D with R&D expenditures is like lumping apples and oranges together, and comparing them across time or companies is not a good idea.

My R&D executive friend was particularly concerned about the telecommunications industry. Lucent's trends may be instructive since it is the largest telecommunications R&D spender. The yearly sources show that its expenditures are shrinking, but they mask the extent of the reductions (5). From a peak of $1,012 million in the first quarter of fiscal 2001, R&D expenditures have declined six straight quarters, to less than half the peak, or $480 million in the third quarter of 2002.

This simple exercise demonstrates that both my executive friend and his colleagues are probably correct. There are companies where industrial R&D spending is shrinking significantly, such as Lucent, and others where expansion is still healthy (Pfizer for one). However, the overall state of industrial R&D is not nearly as rosy in 2002 as it was in 2001, and that is much more apparent in the quarterly data.

IRI's president, F.M. Ross Ambrecht, Jr., predicted this last January at a Dept. of Commerce Corporate R&D Roundtable, saying. "As to the overall level of funding for industrial research in the coming year, it's my belief that it's probably going to be difficult to reach a level that would stay even with inflation.... So I believe we're going to see maybe for the first time in a number of years a drop in total technology supported by industry in real dollars."

The Need for Better R&D Data

The exercise also demonstrates that there is a significant need for better and more timely R&D data. Any control systems engineer will tell you that lag times in signals cause instability. If technological innovation and change have speeded up, which has become a consensus belief, then R&D measurement should be more timely. Reducing the lag times should improve management and policy decision-making.

As Harvard professor David Hart has pointed out, the story of the past decade in technology policy circles is that industrial R&D spending overtook federal R&D spending in a major way. This has led to complacency among policymakers that the U.S. R&D enterprise is more than healthy and increases in the federal investment might actually crowd out private R&D spending.

At the IEEE-USA, for example, we have been complacent with respect to the overall health of the U.S. R&D enterprise, in large part because of the rapid rise of industrial R&D investments. We will be re-thinking this position at our upcoming Engineering R&D Symposium in Washington, D.C., in March. The Symposium is being co-sponsored with three other engineering societies (ASME, ASCE and AlChE).

The other observation is that the quality of the data has major limitations. In a day when many proclaim that information is power, it is surprising that the quality of industrial R&D data is so poor. In-process R&D is often lumped together with R&D expenses. The most up-to-date restatements of R&D data do not always get presented. The nature of the R&D expenses and personnel is rarely described in any detail.

Of course, one of the main reasons these data are not available is that they are viewed as proprietary information by companies, which would prefer not to reveal it to competitors. However, the presentation of accurate, consistent and timely R&D data should lead policy-makers to more informed decisions.

R&D spending is only one way to measure the health of the R&D enterprise, as the IRI R&D Trends Forecast rightly points out. In my next article, I will examine data issues with respect to the R&D workforce including supply, demand and the pipeline. Many in technology policy circles are concerned that the supply of science and engineering U.S. Ph.D.s is not keeping up with demand. What data is this belief based on?

Quarterly R&D Spending (in $millions) vs. Published Trebds
(sorted by percent change past 6 months) *

                                       Percent
               Trailing 2  Equivalent   Change
                Quarters   2 Quarters   Past 6
   Company       FY '02      FY '01     Months

Pfizer          $2,458      $2,144        15
Johnson &
  Johnson       $1,763      $1,588        11
Intel           $2,006      $1,914         5
Bristol-Myers
  Squibb        $1,048      $1,003         4
Microsoft       $2,322      $2,361        -2
Boeing            $844        $883        -4
IBM             $2,333      $2,493        -6
Cisco           $1,655      $1,941       -15
Motorola        $1,836      $2,258       -19
Lucent          $1,004      $1,763       -43

                  IRI
                  R&D         R&D        IEEE
                Leader-     Magazine   Spectrum
                 board      Forecast   2000-'01
                2000-'01    for '02     Change
   Company      % Change      (%)        (%)

Pfizer              9           7           9
Johnson &
  Johnson          23           8          24
Intel              -3          13           0
Bristol-Myers
  Squibb           17           9         158
Microsoft          16          13          16
Boeing             34          -5          -3
IBM                22           3          22
Cisco              45           0          17
Motorola          -10          12          -9
Lucent            -12         -31         -30

* A complete analysis would include data for the top 100
spenders; nevertheless, the sample is illustrative of the
different outlook that more up-to-data provide.

References and Notes

(1.) Goldstein, Harry. "They Might Be Giants." IEEE Spectrum, Sept. 2002, pp.43-48. The data source is Standard & Poor's, and is likely to come from financial statement disclosures, although this is not specified.

(2.) Financial Accounting Standards Board. FAS2-Accounting for Research and Development Costs, June 1, 2002.

(3.) The trailing quarters for FY '02 and FY '01 were collected from the EDGAR database at www.sec.gov. The data come from the latest two filings in 2002 and the equivalent 2001 filing.

(4.) See Financial Accounting Standard 2.

(5.) Actually, there is a discrepancy in the IRI Leaderboard's presentation of Lucent's R&D spending. Lucent's 10-K filed for FY '01 restates the FY '00 R&D spending from $4,018 to $3,179. The Leaderboard does not take this into account in calculating the percent change from '00 to '01. The restatement is probably due to the spin-off of Avaya from Lucent at the beginning of 2001.

TechTalk

"Somehow people in the technology industry think that every one of their customers wakes up every morning saying, `I wish I had more technology. I can't wait to learn more about what computers can do.'"--Retired IBM CEO Louis Gerstner, Jr., in his book, Who Says Elephants Can't Dance? (Harper Business, 2002)

Ron Hira, Ph.D., P.E. Center for Science, Policy, and Outcomes, Columbia University, Washington, D.C. rh2107@columbia.edu; www.cspo.org, and Chair, R&D Policy Committee, IEEE-USA