Assessing the Value of Research in the Chemical Sciences (1998)

David A. Hounshell

Carnegie Mellon University

In the century-long history of industrial research, three problems confronting large, diversified manufacturers have remained essentially intractable, or, more precisely, perennial. I say, "essentially intractable," because they have not been solved; I say, "or, more precisely, perennial," because managers of industrial research have over the years thought they had solved these problems once and for all, only to see the same problems reemerge a few years later with equal or even greater prominence. All three problems revolve around resources. The first problem is how best to deploy research and development resources: in a central organization removed from the business unit, in a decentralized organization closely affiliated with the business, or through some combination of the two approaches. Over the years firms have moved from predominantly centralized research organizations to predominantly decentralized organizations and then back again, only to swing back yet. Each organizational form has its costs and its benefits, and the combination of the two creates new organizational complexities with their own costs and benefits. Movements from one form to another have been largely cyclical, and firms have tended to move in herds as they are wont to do in so many domains of business.

The second problem is a close cousin to the first: how to allocate resources between the short term and the long term. This is not a simple optimization problem because, unlike very short term R&D investments, the benefits of long-term research are highly uncertain. The longer the time horizon, the greater the uncertainties. Yet if firms invest strictly on a short-term basis, they risk being ruined by what the economist Joseph Schumpeter called the "perennial gale of creative destruction"¹—technological obsolescence, market displacement, and the like.

Finally (and flowing directly out of the second problem), measuring the returns on investment in R&D has proven intractable. If people tell you they have an accurate and infallible way to measure ROI in R&D at the firm level, the industry level, or the national level, take it with a grain of salt. It is likely

that the claimant has an agenda that conflicts with the goal of truly assessing the costs and benefits of R&D. I say this because research—especially long-term research—is highly uncertain, and also because the benefits of research are highly complex, going well beyond the abilities of most models to comprehend them except in a crude probabilistic sense.

As a historian, I cannot—and I will not—say that history "proves" my argument. But I can definitely say that in the past, some very smart people have wrestled with the issue of measuring the return on investment in R&D and have frankly admitted that it is an intractable problem. Moreover, they concluded that any scheme they might develop was so flawed as to be dangerous if used alone for decision making; consequently, they relied on other criteria.

This paper centers principally on the case of the DuPont Company, which until relatively recently has consistently been one of the nation's leaders in industrial research. In 2002 the company will celebrate the bicentennial of its founding and the centennial of the founding of its first major industrial research laboratory. DuPont was among a handful of what I have elsewhere termed "the R&D pioneers" in the United States.² DuPont's scientific and technological achievements are many, and some have even become legendary.³ DuPont also holds a special place in the annals of business history, for as Alfred D. Chandler, the dean of American business history, has shown, DuPont pioneered in 1921 the organizational innovation of the multidivisional governance structure for a diversified multiproduct firm—the classic M-form organization.⁴ Even though both structural developments and fashion have served in recent years to undermine the extent and stability of the diversified, multidivisional firm—especially the vertically integrated firm—the M-form remains a fundamental organizational framework in business today.

Even before DuPont arrived at the M-form of organization, its executives had been pioneers in another area of business management: the formulation of a method to calculate return on investment and the development of decision rules or norms based on ROI calculations. The DuPont innovation of ROI calculations represents one of the most significant turning points in the history of modern accounting and management. As Chandler emphasizes, it allowed for the first time the integration of financial accounting, capital accounting, and cost accounting. ⁵ I suppose one might be tempted to leap to the conclusion that this development marks the triumph of the bean counters, but that would be a dangerous conclusion to reach in the case of DuPont, at least for much of its history. The development of the ROI calculation was the work of F. Donaldson Brown, an executive in DuPont's Treasurer's Office. Brown became treasurer of DuPont during World War I and in 1921 vice president for finance at General Motors, a company then headed toward bankruptcy but rescued by DuPont and an alliance of DuPont family members and executives, who together gained controlling interest of the automaker.

Donaldson Brown's ROI formula eventually became a landmark in textbooks in control, accounting, and finance.⁶ In 1993 I interviewed a man who had been an assistant professor of accounting at the Harvard Business School in the early 1940s and later president of a Fortune 10 firm in the early 1960s. Without prompting he recited the formula for me and proceeded to give me a short history of Donaldson Brown's development of the formula and its application in corporate accounting and decision making. His command of this information was as fresh and as complete as if he himself had just developed the ROI formulation rather than having taught it more than five decades earlier. (The interviewee, by the way, was Robert S. McNamara.⁷)

I call attention to Brown's ROI formula not only because of its importance but because it was merely one analytical method developed at DuPont to guide its executives in making decisions about the allocation of assets. Brown's work was done in the context of DuPont's bold program of diversification, which over little more than a decade moved the company from being predominantly an explosives manufacturer to becoming a diversified chemical giant. The company's executives needed objective methods to guide their resource allocation decisions. Would DuPont realize a greater return by investing in this business rather than that one? Should executives fund the expansion of this plant rather than that one? Brown's methods helped guide these executives, and it also allowed them to measure the performance of existing DuPont businesses.⁸

In acquiring companies, DuPont's principal method of diversification, executives also needed methods for evaluating the worth of potential acquisitions. How much should DuPont pay for a company that made a commodity product in plants that were on average 10 years old versus a company marketing a branded product in a plant that was 20 years old? What value should the company assign to trade secrets compared with products and processes protected by patents? These are but some of the questions faced by Walter S. Carpenter, a brilliant young executive who carried out most of the evaluations of potential acquisitions. Like Donaldson Brown—and indeed like almost all DuPont executives—Carpenter had attended engineering school.⁹ He approached problems methodically and analytically. Despite the high degree of uncertainty in such evaluations—cooked books, inaccurate plant appraisals, and other drawbacks—Carpenter derived a simple way to assign worth, which guided executives in carrying out one of the most successful chapters in diversification in American business history.¹⁰ Despite his youth, Carpenter replaced Donaldson Brown as treasurer of DuPont in 1919 and continued to develop analytical methods for cost analysis and asset appraisal and to refine Brown's ROI formula. (As a member of

General Motors's Board of Directors, Carpenter later assisted GM's Brown and Alfred Sloan in solving some seemingly intractable problems in executive compensation and pensions.¹¹)

Both Brown and Carpenter were nurtured by Pierre S. du Pont, the man who can be credited most for the transformation of DuPont into a highly profitable modern corporation. Trained in engineering at MIT, Pierre was simply a brilliant businessman. His career was distinguished by one principal mode of operation: gather the best information possible—both current information and historical data; assemble it into comprehensible charts, graphs, and tables; and make decisions based on reasoned analysis of the data. Only those who have spent time reviewing Pierre's business records can fully appreciate the extent to which Pierre's brilliance was the product of rigorous information gathering and reasoned analysis of data, rather than flashes of insight or daring decision making.¹²

I have focused on Pierre du Pont, Walter Carpenter, and Donaldson Brown to make a point. These men believed in quantitative data, cost analysis, the benefits of ROI calculation in decision making, and management of the business through tight accounting and financial controls. Yet they never assumed that the firm's research could be managed by the numbers. That is, they never thought for a moment that the firm's investments in research could be evaluated by the same means it used in evaluating whether a new plant should be built, an existing plant expanded, or another company bought. Let us examine more closely what they thought about the management of R&D and how they actually proposed to evaluate its returns on investment.

For a brief period in the first decade of this century, Pierre du Pont found himself with the responsibility for the direct oversight of the Experimental Station, which from its creation in 1903 was responsible for research related to all the company's products and for monitoring and evaluating technologies developed outside the company. Pierre was well aware of a conflict within DuPont' s executive ranks about how best to organize industrial R&D. One of his fellow executives, Hamilton Barksdale, had been responsible for creating DuPont's first industrial research laboratory in 1902, the Eastern Laboratory. Unlike the Experimental Station, the Eastern Laboratory was focused on one line of research: high explosives. It worked on both product and process research, and in both of these areas it brought substantial, quick returns on modest investment.

As early as 1904, Barksdale had begun to campaign against the broad, general mission of the Experimental Station. He argued that DuPont would get the greatest return on its investment in R&D if it organized research for its other principal products—smokeless powder and black powder—in the focused manner of the Eastern Laboratory. Pierre du Pont shared the views of his business mentor, Arthur J. Moxham, the head of DuPont' s Development Department, who believed that the Experimental Station had a much broader mandate and that ultimately it would bring major returns to the company. DuPont's Executive Committee argued strenuously over which approach was best, and it ultimately stalemated, leaving DuPont with two approaches to industrial research, one centrally managed and focused on corporate-wide research and the other a narrower, business-unit-focused laboratory with a

clear mandate and relatively short term objectives.¹³ During the brief period of Pierre du Pont's oversight of the Experimental Station—1907 and 1908—Pierre instructed the director of the Experimental Station to think broadly and for the long term.

Pierre du Pont thus made a remarkably clear statement that the returns on research investment cannot be measured solely by their direct effects—a new product or a new process, an improved product or an improved process—but also must be evaluated on their secondary and tertiary effects, which in many instances transcend the primary effects.

In 1911, in a major reorganization of the top management of the company, which came about entirely independently of debates about research management, Hamilton Barksdale became the general manager of the company, or what in today's parlance would be the chief operating officer. Barksdale's philosophy about the management of research has already been noted; he believed that it should be closely aligned with the operating or manufacturing units of the company and must be measured by short-term criteria. Even before taking the reins of the company, Barksdale moved to make his point about the different performances of the Experimental Station and his near-and-dear Eastern Laboratory. He convinced the Executive Committee to request each research organization—the Eastern Laboratory and the Experimental Station—to prepare a retrospective three-year evaluation of what it had contributed to the company's profits.¹⁵ Charles L. Reese, the founding director of the Eastern Laboratory, had an easy time of it. He simply took four projects at Eastern that had resulted in new products and processes; reported on the sales, earnings, and savings stemming from these innovations; and compared these figures with the total costs of the Eastern Laboratory's operations over the same period. He was able to show that for each dollar the High Explosives Operating Department spent on R&D, the lab had returned roughly three dollars to the company. As Reese crowed, "In consideration of the fact that only four of the many subjects worked upon at the Eastern Laboratory are included in the estimate of saving, it is safe to say that the Eastern Laboratory has justified its existence." ¹⁶

Reese's counterpart at the Experimental Station—by this time, Pierre's younger brother Irénée du Pont—had a far more difficult time. He could not demonstrate in any rigorous manner any direct returns to the company's investment in general research. The time horizon was too short, he stressed in his report. However, he believed that the Experimental Station had definitely generated many of the secondary and tertiary benefits that Pierre had identified four years earlier. Much of the work of the Station had been devoted to meeting new regulations being imposed on the company by the government.¹⁷ He might have added that the research the Station had done in the area of smokeless powder had kept the company ahead of the U.S. government, its sole customer for the product and a potential competitor. This fact would soon be borne out.

Under pressure from the anti-big business forces that had begun to build in Congress in 1908 (shortly after the Justice Department launched its antitrust case against DuPont), the Army and the Navy had been pressured into finding means within each service to make the nation less dependent on DuPont for both the development and the manufacture of smokeless powder. In 1912 the Justice Department won its case against DuPont for violations of the Sherman Act. The company was forced to accept a consent decree that called for DuPont to divest two-thirds of its manufacturing capacity in explosives, establishing two new competitors, Atlas (which eventually became ICI's U.S. operations) and Hercules. But the military services interceded, defending DuPont as a progressive company whose research in military propellants was critical to the nation's security. In the consent decree of 1913, DuPont was allowed to keep all of its military propellants capacity. World War I broke out the following year, and DuPont could not have been better positioned to make vast profits from the sale of military propellants.¹⁸ In this case, the tertiary effect of the research program at the Experimental Station was simply enormous.

Hamilton Barksdale attempted to bring greater "relevance" to the Experimental Station and to narrow its research areas and shorten its development horizon in 1911 by putting his favorite research director, Charles Reese, in charge of the Station in addition to the Eastern Laboratory. Ironically, over the next 8 years or so, Reese built a large, powerful central research organization that Barksdale would have found highly objectionable. But by 1915 Barksdale was pushed out of the company's general manager position and replaced by Pierre du Pont, who had emerged as the leader of the faction of the du Pont family that gained controlling interest in the DuPont Company following a major schism in the family. Any immediate attempt to evaluate the return on investment in R&D by narrow financial criteria was soon abandoned. The R&D organization played a critical role in the company's diversification efforts in ways that have been discussed elsewhere.¹⁹

During the next 20 years, as DuPont's research organizations grew, some executives made occasional efforts to bring DuPont's R&D programs under the same ROI regime that the rest of the company operated by. But those efforts failed—or, I should say, cooler heads prevailed. In the late 1920s, DuPont's general managers—the heads of its diversified businesses who today would be called group vice presidents—began to compare notes about how each operating department determined how much to spend on R&D. To the amazement of Charles Stine, the head of DuPont's central research organization, there were essentially as many rules of thumb for allocating research as there were operating

departments. One general manager allocated research expenditures on the basis of how many pounds of product his department made each year; another simply spent a certain percentage of his department's sales; yet another tied R&D spending to earnings; still another let instinct be his guide. How these research monies were actually spent within each department also varied widely. Some departments devoted most of their research expenditures to product research, others spent more on process improvement, while still others invested in more basic research.

Under Stine's leadership, both while he served as the head of the company's central research organization and after 1930 when he became the member of DuPont' s Executive Committee who was in charge of the company' s research portfolio, DuPont moved toward the standardization of R&D accounting across the company. It established five classifications of R&D expenditures (chemical control, improvements to existing processes and products, development of additions to established lines of product, development of new products or processes in entirely new fields, and fundamental research). Once implemented, this classification method allowed executives to see more clearly how the company was spending its R&D dollars, and it also improved the coordination of research across the company.²⁰

The implementation in the early 1930s of standard accounting procedures for R&D expenditures soon led to a new round of thinking and debate within the Executive Committee about how much the company should be spending on R&D, now that it knew how and where it was spending R&D money. What was a good benchmark? Was it 3 percent of sales, or should it be 6 percent of earnings? Or should it somehow be tied to new investment in plant and equipment? These were tough questions, especially in the context of the Great Depression, when one-fourth of U.S. workers were unemployed. For a while, the company settled on tying R&D to sales, which it began to do in 1930. But soon members of the Executive Committee sensed that the general managers who reported to them, not on the operations of their respective departments but on their departmental profit-and-loss statements, were cutting back too much on R&D to keep their balance sheets attractive. Led by the brilliantly analytical Walter Carpenter, the Executive Committee changed its policy regarding how much it would spend on research, beginning in 1934—a revision that held until the mid-1960s. Rather than tying R&D expenditures to sales, DuPont would thereafter make decisions on research allocations based on the merits of the research itself—that is, on the research opportunities in the various domains covered by the company. Carpenter stated categorically that DuPont would fund any "well conceived" R&D project that "we are prepared and willing to undertake . . . with perseverance, enthusiasm, and ability."²¹

This policy echoed the earlier policy articulated by Pierre S. du Pont. Research was of such a nature that it could not be managed by the bean counters, who could not possibly capture in their numbers all the returns on investment possible in research. Opportunities in research must be judged by research managers; research budgets would be determined not by formula but by the opportunities for improvement in products and processes, for new products and processes, and for the production of new and useful knowledge. DuPont would fund any and all projects judged to have significant scientific, technical, and commercial merit. And so it was at DuPont for more than three decades. Whereas capital asset allocation had to clear DuPont's hurdle of 15 percent ROI, allocation of R&D monies had simply to meet the new criteria. Executives assumed that as long as a project remained interesting—as long as it

possessed technical merit—the company would support it, even though no direct benefits could be demonstrated or measured. The promise of primary benefits was sufficient, and certainly the secondary and tertiary benefits might even outweigh any of the primary ones. But these could not be measured accurately and would therefore not be measured at all. This policy is all the more remarkable when we recall that it was made during the Great Depression, when uncertainties about the company's future were perhaps at their highest in two decades.

This, then, was the DuPont formula for the allocation of R&D monies. This formula was derived by the same men who were absolutely hard-nosed about setting ROI hurdles when investing in capital equipment and plant and when making major acquisitions. These were the same men who built an enduring organizational structure that has had enormous global impact. These were the same men who not only helped to make DuPont one of the most profitable and successful corporations in U.S. history, but also helped to rebuild General Motors from its near-bankrupt state in 1920 into a phenomenally profitable company over a long period, using the same organizational structure and hard-nosed decision making as at DuPont.

The obvious question is this: Why did these men treat industrial R&D differently from other types of investment decisions and asset allocation steps? Why did they not apply the same rules? Why did they resist measuring the returns to investment in R&D in a formulaic manner? One answer might be that these men were incapable of devising the proper measures of R&D productivity. But I think not. They were brilliant executives who proved their analytical abilities many times over in all aspects of overseeing the largest firm in the U.S. chemical industry as well as the largest firm in automobile manufacturing.

Another answer might be that DuPont was so profitable and had so many assets that its executives did not need to worry about the firm's R&D spending. The historical record will not support this argument, because these executives worried about every penny the company spent. They watched the firm's assets carefully.

Yet another answer might be that the company spent so little on R&D relative to its sales and earnings that its executives did not waste time on measuring returns on R&D. Here, too, the record does not support such an argument, for the same reason as above.

To me at least, the most plausible explanation is simply that these executives understood research. They especially understood the high degree of uncertainty that accompanies long-range research. They understood that the further out on the horizon R&D moved, the greater the uncertainty in terms of specific, predictable results. But they also understood that even though the primary effects of research (that is, new and improved products and processes) might be difficult to predict and to measure, the returns on longer-term R&D were not limited to primary effects. Secondary effects (such as continuity of programs, improved capability for recruiting scientists and research engineers, and increased organizational capabilities) were perhaps more certain but by no means easily or accurately measured. Tertiary effects were also considerable but no less difficult to capture in a model or formula.

Most certainly, DuPont's executives knew that the returns on short-term, more precisely focused R&D investment were both more certain and more easily measured. But when the company moved its research objectives out on the horizon just a little bit, uncertainty grew rapidly. Further out on the horizon, the only certainty was that if the company did not do research, it could not remain competitive.

Some DuPont executives invested in long-term research and development as a matter of faith. Their experience with research supported their faith in research; that is, their belief in research inevitably paid

off. Still other DuPont executives saw research as a form of gambling. Predicting in advance the exact payoff from any particular investment in R&D could not be done, but over a large number of investments the odds of winning any particular gamble (or investment cycle) could be established, at least roughly. And as long as the payoffs were sufficient over a large number of cycles to pay more than the wagers, and as long as the differential exceeded the returns from other forms of investment, DuPont's executives were willing to continue gambling on R&D.

Both the faith-based decision making and the probabilistic decision-making heuristics worked reasonably well as long as executives could judge the scientific and technical merits of the research they were funding. Only later, when executives lost the ability to evaluate the merits of proposed research or to judge research opportunities—or failed to develop suitable mechanisms to accurately appraise the merits of research—did the company run into serious trouble when evaluating its research programs. It was at this point in the company's history that executives sought supposedly ''more sophisticated means" to anticipate the expected returns on investment in R&D. This was largely an act of desperation, for there was no real substitute for informed judgment.

The forces that today are driving both public and private investment in scientific and technical research to be justified by supposedly rigorous or "hard" analytical methods signal, I believe, a similar state of desperation. This desperation is borne out by a loss of faith in the inherent benefits of enlightened research and by the inability of policy makers to judge the merits of both proposed research and research opportunities more broadly, not to mention the capabilities of the researchers themselves. The nation is in for some rough times. Lying with numbers, cooking data or inventing numbers outright, and distorting programs to ensure that some arbitrary investment hurdle or public benefit criterion will be met will surely follow. DuPont went through such a period in its research history.²²

Audience Member: In the current environment, with the discussion of the nation' s largest companies cutting back on research and development, where is DuPont in the 1990s, and what is going to happen in the future with regard to R&D at DuPont?

David Hounshell: You will have to ask a DuPont spokesman about what is happening now. I can answer any question about DuPont's research until 1980. I'm a historian.

Audience Member: What about this perception that people have that big companies are cutting back because of foreign competition?

David Hounshell: With the end of the Cold War, a general malaise seems to be setting in with regard to R&D. There is intense debate about public investment in research. There is this question about the forces of globalization: To what degree is global competition driving incentives for investing in R&D down, lowering those incentives? These are very complex issues.

There are a number of non-U.S. based corporations as well as some U.S. corporations that have actually increased their R&D spending since the end of the Cold War. They have tended to move toward more fundamental research in some cases. So I would say that there is not a uniform pattern. We do have

a few notable examples where large, previously very successful central R&D laboratories, in terms of scientific and technical output, have been shut down as companies have been acquired. One example is the acquisition of Gulf in the 1980s by Chevron, where they closed one of the major petroleum research organizations. Another resulted from GE's purchase of RCA and the subsequent closing of RCA laboratories, as well as the impact of the deregulation of the telephone industry and Bell Laboratories. I am happy to report that I hear from people at Lucent that Bell Laboratories has never seen more commitment on the part of its owners toward long-term research.

All in all, I would say there is no uniform pattern. There is certainly a sense that it is not the way it used to be, and we know that is the case.

Robert Lichter, Camille and Henry Dreyfus Foundation: As a historian, could you say a bit more about your views on why the transition in the 1960s occurred? You described its occurrence, but it wasn't clear why the change took place then. What were the factors that contributed at that time?

David Hounshell: The factors go back to a misreading of a major period in DuPont's history that began in 1926 when DuPont moved to establish a new central research organization, a small fundamental research program, which in one month, March of 1930, led to the discovery of the first wholly synthetic rubber and of the first wholly synthetic fiber. The outcome of the first led to the commercialization of neoprene. The outcome of the second led to the discovery in 1930, and then the development and commercialization in 1940, of nylon. Nylon was an enormously successful product. It was the product of a fundamental research program that contributed in a major way to polymer chemistry.

Then, there were the very powerful technologies that came out of World War II. DuPont was involved in the development of one of them, the Manhattan Project and the atomic bomb. It was very clear in a publication in January 1939 that fission, the splitting of the nucleus, could lead to the development of a powerful weapon—the atomic bomb. This was achieved in 1945. This and the other technological leaps forward in World War II (radar, etc.) were very powerful confirmations of the power of basic research. DuPont saw these developments and in the postwar period invested very heavily in the expansion of its fundamental research program.

What the company lost sight of was that it would not have been possible to commercialize neoprene or nylon or to develop an atomic bomb without tremendous technical and organizational capabilities, and that merely investing in research would not be sufficient. For a period of 15 years, the company invested heavily in research, not thinking about the commercial aspects and the organizational aspects of what it was doing, and it began to see diminishing returns on its investment in research. In an act of desperation, in the 1960s it established what came to be known as the New Venture Program, in which it tried essentially to remove some commercial constraints and it wound up basically bankrupting the company. The company had never borrowed money previously, and the efforts that came out of the New Venture Program eventually led to the company having to borrow money for the first time. So, it was a failure to appreciate that research is important, but other factors—technical capabilities, market knowledge, marketing expertise, manufacturing expertise—were critical in the success of its own stellar products, neoprene and nylon.

Thomas Manuel, Air Products & Chemicals Inc.: To contribute to the point of cyclicality you brought up in your presentation on DuPont, I would like to note that they have a tremendous record and history, and that there have been some very interesting articles published recently by Joseph Miller and Perry Norling of DuPont, who describe cycles of productivity of generation and development and exploitation and so on, which have repeated throughout the history of the corporation. I forget the period.

David Hounshell: Fourteen to 16 years.

Thomas Manuel: Yes, 14 to 16 years. In a sense that is reassuring because, like most sine waves, this suggests that investment in R&D will tend to turn up again. DuPont's history strongly suggests that.

David Hounshell: At least Joe Miller is hoping that is the case. He, of course, has reasons for finding that periodicity.

Vern W. Weekman, Mobil R&D Corporation: Did DuPont use any type of a rolling average over a 5- or 10- or 15-year period of the return on the basic research programs in the Experimental Station? If they did, did this influence their faith in long-term research?

David Hounshell: In the 1920s, some of the executives, who were keen on the return on investment calculations, wanted to do this. There was extensive debate in DuPont's management, which I've only glossed over here, on how to measure these returns. There were DuPont managers who derived models that made use of rolling averages, smoothing functions, and so on. These individuals were very sophisticated with their analytical methods, and they tried to use these techniques to place the ROI from basic research on a firm footing. This was finally brought to a head in 1934, when both Stein and Carpenter concluded that none of these models works. The models simply could not capture all of the possible returns on basic research, particularly the secondary and tertiary returns. Stein and Carpenter decreed that DuPont was not going to use any of the proposed ROI methods for basic research.

There was another factor behind this decision. As the DuPont executives gained increased experience with their ROI measures, they realized that, although these measures led to better informed decision making, their decisions were not based solely on the ROI calculations. There were many times when they made investment decisions where the ex ante ROI calculations were 8 and 10 percent rather than 15 percent. So the ROI calculations were merely advisory. They were no substitute whatsoever for judgment. That is what was critical. The company's executives, Carpenter in particular, believed that as long as the top executives were knowledgeable and talented enough to know both the scientific and THE commercial merits of an enterprise, they would make the right decisions.

James Fry, University of Toledo: In the state of Ohio, the Science and Technology Council, which advises the governor, is making a serious attempt to demonstrate that science and technology development have an impact on society and economic development. Is there any message from the studies you've done that would suggest there is a right way to approach the public on these matters, as opposed to an incorrect way?

David Hounshell: Yes, I think there is a better way. One can look at the West, the history of the West from the 18th century forward, to see that those nations that have been committed to enlightenment—that is, to the advancement of human understanding about the world and about humans themselves through systematic research—have done better on average than those societies and nations that have not done so. I take it as a matter of enlightened faith. I realize that one cannot base all policy on faith, but I would submit to you again that any set of calculations that try to predict the return on investment for research, especially fundamental research, is subject to daunting problems. As soon as the measures go into effect, people will start "cooking" numbers. The result will not be better ROIs, simply better numbers.

We need to look more intensely at our own history in Western society, recognize the enormous

benefits that we have gained through scientific and technical change, and identify the sources of that change. Much of the change has come about through research and development. One cannot understand the history of the 20th century without understanding the development of industrial research and development laboratories, and the heavy commitment by corporations to scientific and technical change. If we look at the 1993 science indicators, we find that corporations spent 68 percent of the nation's R&D dollars. They employ the bulk of scientists and engineers in the United States. These numbers are very consistent with other trends in the 20th century. We reaped tremendous benefits from these investments. Unfortunately, I cannot demonstrate it rigorously, quantitatively.

Andrew Kaldor, Exxon Research and Development Corp.: My understanding is that somewhere in the 1900s, the accounting principles for R&D changed. Early in the century, R&D was considered to be part of the capital investment made by the company, but later it was considered to be expense.

David Hounshell: That is correct.

Andrew Kaldor: I wonder whether this is a fundamental problem facing the R&D enterprise in industry. In industry, we trade in options. It's a perfectly acceptable legitimate business principle, yet I don't see the options concept, and how you manage it, applied to research. Did DuPont, or some other company, investigate this approach?

David Hounshell: Pierre du Pont explored the options concept at the outset. As early as 1902, when the Eastern Laboratory was established and the Executive Committee was created essentially to oversee DuPont, Pierre strove to obtain a better accounting for research. One of the things that led to the establishment of Donaldson Brown's ROI formulation was the work that Pierre had done on accounting, which was very different than the accounting procedures used by the railroads. In particular, he worked on capital accounts as opposed to expensing or cost accounting.

Pierre initially thought that you could place R&D funding entirely under a capital account, treating it in the short term as an expense account and then readily converting it to a capital account. He investigated this concept intensely between 1902 and 1904. By 1904, he had abandoned this concept entirely. He just didn't think that it worked. He was not satisfied with an expense method either. That is why he wanted to treat it very differently than the other expenses incurred by DuPont. Certainly, in the overall ROI calculations of the company, it was treated as an expense, except where it had generated identifiable intellectual property that could be assigned value—patents, for example. Patents were handled on the capital account, on the capital side of the ledger rather than on the expense side.

Andrew Kaldor: I wonder if you can confirm another piece of DuPont folklore or history, I don't know which. At a recent meeting in San Francisco, someone was talking about DuPont's retrospective study of their technology. They apparently have a list every year of their top 10 technologies and the bottom 10. They have followed these technologies over a 15-year period to see how they fared. The result was remarkable! As I remember, 8 of the top 10 technologies failed, and roughly the same number of the bottom 10 succeeded. This certainly raises some questions.

David Hounshell: Yes, this is essentially correct. I call your attention to the dissertation that Joan Adams is working on. She is doing a number of case studies on polymer innovations. She's studying several cases from DuPont, but many other cases in the chemical industry as a whole. You might want to talk to her. She is much more aware of what's taking place at DuPont today than I am.