Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
MEASUP.ING ECONOMIC EFFECTS OF FEDERAL RESEARCH AND DEVELOPMENT EXPENDITURES: RECENT HISTORY WITH SPECIAL EMPHASIS - ON FEDERAL R&D PERFORMED IN INDUSTRY Nester E. Terleckyj National Planning Association This paper reviews ache recent research on the economic effects of federal research ant development (R&D ~ expenditures . An in~cerpreta~cion and synthes is of this research are attempted. It may be appropriate to begin by noting that existing economic indicators and measurement conventions have certain serious limi~cations in measuring the effects of R&D. Some of the limitations are statistical: The data systems simply do not cope adequately witch technical change. As a result, the data, especially the price and quantity data, are deficient particularly in the most technologically d~n~nic sectors of the economy. Such deficiencies can be remedied by improvements in measurement. But, deeper than that, there are fundamental conceptual problems deriving from the nature of index numbers and problem of measuremen~c of real output of public goods to which scheme are no solutions. These problems can be approached only with reasonable approximations or explanations. This paper focuses on research re rating to government - funded R&D performed on contract by U.S. industry. Such R&D is only a part of the total federal R&D funding, but, since other papers in the workshop deal win federal spending for basic scientific research programs and with federal Rho spending for agricul~cure, space, energy, epic., this paper will deal with chose subjects only Inca round out ache p icture . Analysts of industrial history in the United States have noted that many very important applications to ache private economy have resulted from products developed originally by industry under contract performed in support of defense and space programs. Indeed, it would be hard to argue that ache aviation and computer industries have not been built in a large measure on founda~cions developed in federal contracts, or that much of the progress of celecoonnunica~cions has not been based on federally developed satellite technology. It would be idle to speculate how those industries would have developed in ache absence of U.S. government- suppor~ced R&D, but it is possible to attempt an assessment of some of the effects of such R&D on the priorate economy. . - 149 -
Initially, the economic research an federal R&D has explored the direct effects of government R&D on priorate sector productivity. When the direct effects are not readily apparent, the indirect effects on private app lications and spec tfically on priorate R&3) expenditures have been examined. Often, foreign observers comment that American industry enj oys competitive advantage by performing contract R&D for the A. S . government. Cat R&D gives U. S . industry a head secant in developing the technologies likely to have great eventual value in private markets. Also, the advantages of government purchase of the resulting products help deco nurse civilian markets in ways that foreign producers do not enj oy. On the other hand, some American observers have expressed concern that performance of contract R&D for the government disp laces or crowds out the opportunities for privately funded R&D by absorbing the qualified R&D manpower, by driving up the cost of R&D, 0~9 perhaps, by preempting entrepreneurial or in~renti~re energies of US S . firms . Analysts are not yet at a stage where they can sort out all the possible effects of federal R&D and their incidence in particular sectors and at particular times . S tit1 unknown are the dynamics of the short. and long- tens effects on salaries, use of R&D professionals, arid supply of scientis~cs and engineers for Red, as well as the economic consequences of shifts of scientists and engineers between R&r) and other acti~ri~cies. However9 some progress has been made. We can begin with the results obtained thus far by economic research. The effects of government- supported it&l) activities on ache priorate economy, its productivity growth, and its organizational structure vary among government progress. Expenditures for R&D by business firms are more homogeneous than go~rerrwent expenditures for Rho. Private R&~) expenditures represent a particular type of business inves mend, and, sub ~ ect to variations in risks and market condi~cions, all such expendi~cures generally can be analyzed within the sue framework. Because federal R&D programs are quite different from each ocher, across ~ the-board generalizations and analyses based on aggregation of data for all R&D expenditures of ache U. S. governmen~c are much less likely to be Haloid than generalizations of research results for private R&D. DIRE:C1: Ef SECTS OF GOtlER~ME~ R&D ON PRODUCTIVITY For some government research programs, most s~crilcingly in agriculture, go~rernment-financed R&D was undertaken clearly and directly to increase pri~ra~ce productivity. It has succeeded very well in this objective (see Griliches ). A positive relation 150 -
between government expenditures for agricultural research and productivity growth of American2agricul~cure has been found over long historical periods ~ see Evenson ); Use of agricultural research expenditure by the government is now standard in estimation of production and productivity models of agri culture. A study of returns from agricultural R&D in different periods ~ see Evenson, Waggoner, and Ruffian ~ found continued high rates of return from agricultural R&D and no evidence of a secular decline in productivity o f that R&D . In adcti~cion, studies have been made of several smaller gove~nen~c R&D programs and their effects on productivity. An example is a study by the Charles River Associates of the effects on industry growth of semiconductor technology R&D supported by the National Bureau of S tandards (NBS ~ . There is no reason to think that government- financed R&D programs aimed directly at raising productivity of particular industries or of the economy in general do not succeed in their goals . However, the principal obj ectives of most of the other government expenditures for R&D are different. Economic retunes have not been identified for such programs, and there is a question whether one should expect to find such benefits or whether economic returns as measured currently by economic data systems provide meaningful indicators of ache output of the programs, especially in the basic sciences. A positive relati.onsh.ip of ache intensity of industrial research and development activi~cies ~ ratio of R&D outlays to outpu~c, or of R&D manpower to Coccal employment) deco productivity growth has been established rather well for the manufacturing industries (see Terleckyj5 and Minasian6. After further distinc~cions were made between private and government-financed R&D conducted in industry, the estimated effect of private R&D on productivity became larger and more strongly evident, while no effect of government- financed R&D on the productivity of industries performing i~c could be identified (see Leonard and Teriecky; ). These results, showing weak indications, at most, of relatively smaller effects of government- funded Red), were confirmed in a number of subsequent studies for different industry units and periods. These include, most recently, a rather detailed analysis of total factor producti~ri~cy growth in 27 manufacturing industries over the period 1957-1976 by Griliches and Lichtenber~ and a macroeconomic analys is by Levy and Terlecky]. O Comparable differences in ache estimated productivity effects of private and government- funded indus~crial R&D were found for R&D embodied in purchased inputs. the embodied R&~) was estimated from deco a~ctribut~ng the cost of R&D conducted in the incus tries performing it through the flows in tit transactions and capital input-output matrices (see Teriecky; ~ ). Occasionally, posi~cive estimates were obtained for the effects of private R&D embodied in purchased inputs, but such resul~cs for government R&D were not stable and sometimes even nega~ci~re. This is puzzling because some obvious
maj or adaptations of the results of defense and space R&D were made in the private economy, notably in civil aviation, communications, computers, and heal th care . Never~che less, no s tab le s Ice tis Scat relationship could be estimated. Sectoral leered analyses of "downstream productivity effects in industries using the new products produced similar results. An analysis of productivity change in commercial aviation by Lesaz9 Machnic, and Elkins found that expend) Cures for aircraft R&D have made large contributions lea airline produc~ci~rity. However, in ache analys is, government and private R&D expend) tures were comb ined ~ and ache effects of the government R&D expenditures were not examined separate ty . Subsequent sec~coral study by Terlecky; t4 found positive effects of priorate R&D performed in five aircraft manufacturing industry on the producti~ri~cy of the airl ine industry, but no effects were found for government- funded aircraft and pelts R&D . In another Study, of Ache communications sector, Teriec3~j examined the re lationships of R&D in communications and e lectronsc components Deco product:~vi ty o f Ache communications industry and found strong positive association between private Red) and industry productivity, but no effect of governmen~c Rho. For the entire private bus iness economy, government R&D t~ eated as a capital stock had a positive effect on aggregate praductivity growth, but the estimate of Chat effect was statistically weaker and four times smaller "hen ~cha~c for Ache corresponding effect of the stock of privately funded industrial R&D. In discussing these resui~cs in an earlier paper, Terlecky 16 concluded that while government-financed R&D apparently created oppar~cuntties for additional prince R&D by making such priorate inures t:ments more productive, it apparently could not have direct productivity effects without additional private adap~cation R&D, because products developed for the government usually were not suited for private markets. One would expect that in allocating private R&~:) expenditures stimulated by Ache results of go~rernmen~c R&D, business enterprises would inkiest their funds in the same manner as they would invest in _ field of technology developed with priva~ce funds only, Mae is, equalizing on the margin the risl<-adjusted relearn from all types of R&J) invastment~. ~hen, in the case of all government R&D programs in which Ache purpose of government spending is to support missions of government agencies rather than productivity of dacha private sector, only the induced private R&D would have an observable effect on productivity growth. In a recent paper, Levyt7 argues that the reason no producti~ri~cy effects can be observed directly for goversment-funded industrial R&D is that such R&O is a public good, used by all users · 1~2 - c
to the extent where its marginal product is zero. Therefore i its contribution to productivity cannot be observed statistically by tradi tional techniques and approaches . This appears to be a reasonable explanation of the general failure of models predicated on estimating marginal products to produce s ignificant direct estimates of any effects of contract R&D on produc~ci~rity of industries. the indications that at the aggregate level there may be an observed residual effect of contract R&D on productivity may mean Chat there are some unappropriated benefits accruing to the users of newly developed private products derived from the antecedent results of federal contract R&D. An important clarification is in order. The fact that federally funded industrial R&~) does not produce any observable productivity effects and probably should not be expected to produce such effects does not mean that it has no economic value. As was pointed out some time ago by Forte and Buchanan, there is an inherent conceptual limitation in the national income accounting (and gross national product) data in that they attempt to measure the real product of the public sector by the real cost of producing it. While the resource Cost used in the public sector can be identified, the real output of public goods cannot be measured because its marginal value product and implicit price are always zero. Measuring them at cost to force the identity of real factor cost and real output is inappropriate and only confuses the issue, as Forte and Buchanan pa int Out, Thus, the contribution of government- funded R&0 to the productivity of the economy has to be approached empirically through its contribution to the productivity of private R&D. This is, however a formidable research task, one that barely has begun. There are great difficulties in identifying marginal and average contributions of governmen~c R&I) to the productivity of priorate R&D. This applies not only Deco contract R&D but perhaps with even greater force ea basic research in the natural and life sciences, where the connections between the results of basic research and Ached applications in industry may be more difficult to trace because they are much more diffused and may involve longer time lags. Because of the difficulties in direc~cly identifying the contribution of contract R&D to productivity of the economy, and because of the increased awareness of the in~cerdependence of government and priorate R&D, much recent research. has been concerned with the effect of federal R&D on the level of privately funded industrial Rho. S=vIES RElATING GOVERNMENT R&D 'to PRIVATE R&D I~JES1~7T A number of research studies completed in the past six years indicate that, in most cases, government R&D expenditures have been pos itively 1 ~ 3
re laced to private it&l) expenditures . Table 1 presents the q~anticative results of . these studies . The macroeconomic analyses are s',mm~rized in Part I of the ~cable, the industry and sectoral analyses in Part II, and the macroeconomic estimates in Part ITI. The studies differ greatly in methodology, data sets, levels of aggregation' time periods, model specification, and ocher variables. Some of ache studies were cross - sec~cional, others used a time - series approach In some cases, they included a~ctemp~cs to estimate Ate lag structure of the rela~cionshlp between government and private R&3) expenditures . Carmichaei19 related federally and privately funded R&D performance for /;6 transportation companies in 1976 and 1977. His results indicated a negative correlation between federal and priorate funding of - . 08 . In a study of the determinants of t~8 composi~cion of private R&D spending of 27S companies is 1977, kink included federal R&~) expenditure among the explanatory variables. He found significant relationships between federally financed R&D performed in the company and the scowl company R&D spending, with ~ coeff~cien~c of .094, and also between federally financed R&D and company expenditure for development, with a coefficient of .086 . The estimated coefficients for batta basic and applied research were negative; the coefficient for applied research was very smal1. Neither was significant at the 5-percent level. Mansfield and Switzer2\ conducted a microeconom~c study of 25 major firms, in it&D-intensive indus~cries, involved in 41. federally funded energy R&D projects. They questioned how an increase or decrease fry the mount of government funding for fiche energy projects would affect the companies ' expenditures of their own funds for R&D . The companies' responses were asymmetrical between increases and decreases in federal funding. For increases, the respondents indicated that priva~ee R&D expenditures would be increased by 6 cents per dol Alar in each of the fires two years after a one- time increase in federal funding. For decreases, they indicated that private R&D would be reduced by 2S cents for each federal dollar in the first two years following the cut and by an additional 19 cents in ache third year. Cumulated for echoes years, the responses imply a total reduction in prince R&D expenditure of 69 cents resulting from one dollar cut in go~rernmen~c R&D spending. This may reflect diminishing returns, as the authors point out. However, the responses also may reflect an expectation of the decline in government market for the results of pri~ra~ce R&D implici~c in the cut of government R&D support: in the energy field. In that cases such an estimate may be too high as an indicator of Ache magnitude of the impact of government R&D on private R&D in general. In a recent study of the possible effects of company e,~endi~cures for adver~cistng and for R&D on seller concen~cra~cion, Scott 1~4 _
TABLE 1 estimates of the Change in Private Industry R&D Expends Cures Associated with $1. 00 Change in Government R&D Expenditure Studier Change in Pxi~ra~ce R&D Method, Data, And Period of Estimate Part I: Microeconomic Analyses Carmichael, 198 Link, 1982 To Cal R&D Basic research Applied research Deve topmen~c .~.ansfield and Swi Hazer, 19 84 - . C8* . 09* - 08 - . 01 . 09* Increase in gov. .12 ernment R&D Decrease in goner- .69 ernment R&D Scott, 1984 Lichtenberg, 1984 Cross - section regres - s ions, 46 transporta- ~cion companies, 1976 and 1977. Cross - section regress ions, 27 S UO S . manufacturing companies, 1977. Survey responses, 2S companies for energy R&D, 1977. Significance no sac tes ted . .08* to .~* - O 23* to .10* - . 26* to - . 17* Cross - section regressions, 431 companies reporting for 259 TIC lines of bus iness categories, 1974. Cross - section regres - sions, company R&D/ sales return on fed- eral it&D/sales for 991 companies, 1967, 1972, and 1977. Changes in the above ratios, 1967-1972 and 1972 - 1977. _ , .~: ° Not estimated No~c estimated Not "s~cimated Not estimated Not estimated 2 years: .06 ea. 3 years: 2 years; 3 rd year .25 for .19 in Not estimated Not estimated Not estimated
TABLE 1 ~ continued) Method, Data, - Change in And Period Time Lag S Judy Priorate R&D of Estimate Es animate Part II: Industry and Sector Analyses Goldberg, ~ 979 .02 Time series model, 14 Sum of distributed manufacturing incus - lags ~ O ~ and ~ - 1 ) tries, L9 5 8 - 19 7 5 . S fig - nificance not tested. Levin and Re is s 1984 (Table 8~5, columns 2-5) Lichtenberg, 1984 . 08* to . 12* Pooled data for 20 manufacturing industry es, 1963, 1967, and 1972. Pooled regressions, 12 manufacturing industries, 1963°1979. S ignif icance not tested. Not es Mated Expenditures - .08 deco .12 Sum of d~s-~ioured legs (O) to (I) Frnploymen~c - .31 to .10 Sum of distributed lags (O) to (-2) Terlecky;, 1984 .34 Sectoral model Sum of distributed equation for elec- lags (0) to (-4) trical equipment and comsouslication Rho. Significance not Scented. . 31* 1948-1980 ~ -2) . 1~0* 1~948~ 1963 . 56* 1964~ i980 Terlecky, 1985 Sectoral model (-2) equation for aircraft and parts R&D, 1948-1976. . 21* 1948-1976 ~ 17 194801960 - O 12 1961-1976 ~ 156 -
TABLE 1 ~ continued) Me shod, Data, Cnange in And Period Time Lag Study Private R&I) of Estimate ~Estima~e Part iII~ Macroeconomic Analyses Levy and Terleck~i, 1983 (Table 1) Macroeconomic model equations, UO S O bus iness sector, 1949 - 1980 . Government 27* Current year R&D performed only in industry All other goner- .19 Year (-3) only ernment R&D Leery and Terlecicyj,, 198S O 47* to . S2* Estimates of aggregate ir~restment function for 1~. S . R&D including private R&D variables for Europe and Japan, 1963 - 1983. * S tatistica3~1y significant at least at S -percent level . Sources: See References. Not estimated _ - . .: / -
examined the relationship between government- financed R&D and company- financed R&D at the level of individual lines of bust ness within the she companies. He used 3, 388 detailed observations for 437 companies that reported their activities in 259 lines of business to the Federal Trade Commission for 1974. For the entire group of observations, Scott found significant coefficients ranging from . 076 to .10 for government ~ funded R&D in the e qua tion for priorate R&D spending. Signifi,~3nt negative and positive correlations were reported by Lichtenberg between the ratios of company and f ederal R&I) to sales in 991 companies at three different dates, but only significant negative coefficients were found among the changes in these ratios. In a study by Goldberg, 24 government- financed R&D outlays were introduced as a factor determining the productivity of privately financed R&D and, indirectly, Ache demand for private R&D investments Working with ~cime-series data for 14 industries for 19S8-197S, Goldberg estimated an R&D investment model. lathe modes contained estimates of the effect of federal R&D expenditures on private R&D expenditures lagged one year, and of untagged R&D. A longer lag structure was no t cons idered . Me sum ~ f the two - year es timates yields a positive but ~ ow estimate of the total effect on private R&D . Baled on data for 20 manufacturing industries, Levers and Reiss estimated a cross - sectional model of R&D and remarket structure involving concentration ratios, advertising, private R&D, and government R&D O The ir es timates o f the end tion for priorate research and development expenditure included in that mode 1 contain significant coefficients for goverrman~c R&D. he relevant version of this equation indicate that one dollar of federal R&D expenditure induced addi "tonal private R&D expenditure of between 8 and 12 cents . In analyzing pooled regressions of262 manufacturing industries over the period 1963-1979, Lichtenberg examined a number of equations, including distributed lag equations. He found both negative and positive coefficients, with a predominance of negative effects . Terieckyj27 examined the relationship between private and go~rernmen~ ) expenditures in the electrical equipment and communications industry. For the period 1948-1980, he found a series of distributed lag coefficien~cs for up to four years with a son of .34 and a similar and sigslifican~c coefficient of .31 for gover~=ent- funded it&l) lagged two years. For the latter rela~cionship, he also found sta~c~s~ically significant but different coefficients for two subperiods, .10 for 1948-1963, and .56 for 1964-19800 In a Similar sectoral analys is of aircraft and parts, Terlecicyj found a significant coefficient of .21 for ache period 1948-1976, but insignifican~c coefficients for shorter periods. - 158 -
In a macroeconomic study by Levy and Terleckyj,29 estimates of ache s simulation of private R&D expend) cure by go~rerTlment R&D spending were developed within the framework of ache economic growth model of the National Planning P~ssocia~cion (see Terlecicyj, Leery, . and Kanelco ). Me macroeconomic analysis explored a number of issues regarding governmen~c R&D. Specifically the study inquired whether, in ache economy as a whole, primate spending for industrial R&D is influenced by ache total government expenditure for R&D' or only by government- financeti R&D performed in industry ~ Aggregate estimates indicated an effect of 27 cents per dollar for government contract R&D performed in industry, and 19 cents per dollar for governmen~c R&D performed elsewhere ~ in government, universities, and nonprofi~c institutions-~. Stronger statistical results were obtained For the go~rernmen~c- financed R&D performed in industry. lye Levy and Terleckyj analysis31 did nor succeed in iden~cifying clearly the time lags involved in the effects of government R&D. They did no~c find any lag structure for government R&D performed in industry and found a possible indication of a three-~~ar lag for other government R&~ he Mansfield and Switzer survey indica~ces similar three-year lags O These seem deco be relatively short lag periods. Some of the longer time lags, such as those inherent in basic research, simply may be too weak or too variate te deco be identi f fed from ache data . On the o cher hand ~ the Mansfield and Switzer study has indicated that companies often initiate proposals for federal R&~) Prod eces Deco complement R&D programs already under way witch their own funds. Pse~raienc. of that pattern on a relatively large scale would tend deco shorten lag periods appreciably and might even result in negative anneal lag situations reflecting private R&D spending induced by expectation of future government- financed proj ects . SOME INFERENCES AM HYPOTHESES The results discussed above suggest that go~rernmen~c R&D stimulates private R&I). Such evidence was obtained at the level of companies and of industries that perform government- financed R&D, and at the aggregate leered of the priorate bus iness sector ~ S ignifican~c s~ca~cistical estimates of the stimulating affect on private industry R&D were obtained by Levy and Terlec~y],3 both for government. financed R&D performed in industry and for that performed in all etcher organizations. These posi~ci~re estimates probably are interpreted best as the aggregate net effects. They are consistent with smaller indirect effects and with partial, local, and short-term negative effects or crowding out. Me nega~ci~re correla~cions may reflect different motivation and behavior for different types of companies. One would expect firms to forgo funding at least some basic research if The government will fund the ir performance o f that research . Link reports such - L:9 -
possible offsets for basic and perhaps for applied research, but not for development. Another reason for the observed offset may be that companies selling deco the government may support their R&D staff working on government products with their own funds tO bridge the time gap or in preparation of R&D proposals for the government. When ache government awards such companies R&D contracts, then that R&D effort becomes supported by federal funds. What complicates the picture is the Independent Research and De~relopmen~c (IR&~) allowance, which allows federal con~cractors to fund some research and development as part of the overhead on federal contracts, both on contracts for procurement of goods and services and on R&D contracts. Many companies report private R&D outlays to ehe Securi~cies and Exchange Commission in amounts similar in magnit ude tO the average percent of JR&D expenditures for all large contracts as reported by the Department of Defense and the National Aeronautics and Space Administration. It is ~ ikely that percentage overhead allowances for different contractors tend co be equalized, but no JR&D deco are access Able for individual companies, and this is only a surmises Some of the substi~cution between priorate and public funds may involve JR&D funds, which are viewed legally as a cost of doing bus iness and are reported as pri~ra~ce funds by the companies . However, those refunds of it&l) spending received frotn ache federal government represent a distinct ins~citutiona1 mechanism for funding of R&D; since that mechanism is not available to companies selling only in priorate marke~cs, those funds should be considered separately. Me research rests summarized in Table 1 suggest that direct stimulation of further private R&D investments by the performers of government contract R&D is not the only way in which government R&D may s timula~ce addi Lionel priorate R&D expend) cure . Ins teed, the influence of government R&D operates across the economy from government to industry and, most likely, among companies and industries as well. Results of go~rernm.ent R&D perform.ed in a company may be transferred to other companies in the sue or in other industries in a variety of ways. In many cases, such transfers occur through movement of R&D personnel. Additional channels include technical and scientific j ournals and all types of written and oral communications . Some government agencies, notably the National Aeronau~cics and Space Administration, maintain elaborate programs two facilitate transfer of technical irtformeation and commercialization of their it&l) results. They also adopt policies, such as those regarding patent licensing, to facilitate transfer of government- financed technology to priorate sector ,'-~es . Also, successful adaptation of governments funded research to general market products is likely deco stimulate additional R&D spending by imi~cators. Therefore, it is not unreasonable to expect stimulation of private R&D to occur outs ide the companies performing the original go~rernmen~c R&D. Indeed, one may postulate the existence of a - 160
complex network of effects of government R&D on private R&D, ache elements of which could be identified by available research techniques. If all the estimating procedures were specified properly and the negative effects were outweighed by the positive effects in the aggregate of all levels of organization and in the long Nan, one would expec~c the estimate of the magnitude of ache direct impact of government- funded R&D deco be largest at the national level because it would reflect the secondary and ouches induced interactions . I ~ should be of an intermediate size at the industry level, because at that level it would include only the effec~cs of go-~err~ent R&D performed in the given industry, and smallest at the company level, because it would exclude the intercompany effects . By how much these estimates should differ and what negative offsetting effects may be involved, and where, are empirical questions. lathe differences may vary also among programs. For example, the magnitude of intercompany and interindustry transfers may depend on the policies that a government agency has witch respect deco facilitating transfers of R&D results from private performers and from the government. Other things being equal, one would expect the external effects of government R&D to be greater for the R&1) expenditures of those agencies that have active policies for facilitating such transfers . 35~e precariously mentioned study of the National Bureau of Standards R&D program in semiconductor technology illustrates how technology developed in a government laboratory may interact with priorate Rho. In one analysis, ache contributions to company growth made by company patenting activity were found to be significantly higher in the semiconductor companies that maintained frequent contact with the NBS laboratory performing the semiconductor R&D -hen in other companies in the industry. The national level coefficients obtained by Levy and Terieclcy] generally are larger than the coefficients at the industry leered, and the macroeconomic coefficients tend to be smaller and more frequently negative than the industry-le~rel es~cima~es. One may infer from comparing these magnitudes that, on ache average, positive relationships predominate and that there exist interproduct and in~cerfirm effects o f federal R&D on priorate R&D expenditures. Such effects may represen~c imitations of successful applications of federally funded R&D deco private markets, but there also may be complemen~cary effects between different product fields in the same company. If such patterns predominate, then one would expect ache magnitude of the statistical coefficient, specified properly for time lags and other relevant factors, to increase with the leered of aggrega~clon of industrial organization from firms, to industries, to the economy. It should also increase witch broadening of ehe scope of product fields within fates. One might then expec~c to progress from the coefficient on the order of . 29 identified within lines of business within companies by Scott to the macroeconomic coefficient on the order of . IS reported in the macroeconomic studies in Part III of Table 1. The author currently - 161 -
is conducting research to test for ache presence of incerproduc~c and interfirm spillovers, using ache Federal Trade Commission Line of Business data for the years 1974-1977, and hopes to report the resul ts in the near future . U. S . CONTRACT R&D AN INTERNATIONAL R&D TRENDS Me results of federally funded industrial R&D do not appear to remain for long ire the exclusive possession of the firms performing the R&D or the contracting government agencies. lye resulting innovations tend to be diffused to and imitated by other U.S. producers, as well as foreign firms, without very long time lags. Clearly, the results of federally funded teas lo research in the natural and life sciences;are disseminated widely and rapidly. It may be the" the difference between the availability of results of federally funded contract R&D and of teas to research is one of degree and time arising from the proprietary interests of ache performing firms and national security requirements. But, at least as far as the business applications are concerned, the barriers may not be Freer high. An analysis of bents in primate and government expenditures for incus trial R&D in the United S tates, Europe ~ seamen leading R&D performing collieries)-, and Japan presented in a recent paper by I-e my and Terlecky] found that' at the aggregate level, only private R&D expenditure in "he United Sees had ~ positive statistical rela~eionsh~p to privet. it&l) spending in Europe and Japan. In all the estima~ces, the apparent effect of U.S. contract R&D was insignificant. This suggests that ache diffusion of the results of U.S. contract R&D to other countries occurs through U.S. private R&D and resulting private sector applications, rather than directly. However, the study also found that government R&D expenditure in each economic area (Untted States, Europe, and Japan) had a significant relationship deco private R&D spending Kitchen that some area. To avoid possible estimation error arising from the mutual dependence of R&D spending in ache echoes economic areas, the echoes R&D investment functions were estimated by the ins tenemental variables method. The result (see Table 2) is consistent with the hypotheses that private Red) invest:ment in each area depends, with ~ lag, on ache level of economic activity, on the government-fur~ded RM) performed in industry, asked on the foreign private R&D in that area. For private R&D spending in Europe and Japan, only U.S. R&D was significant. For U. S . private R&D expenditure, no significant difference was obtained between European and Japanese private Rho, though both were significant, witch different time lags. lathe two were combined in the final estimate. There are very large differences between the coefficients reducing go~rerr~ent- funded and priorate R&D performed in industry in - 162
T-~' E 2 Estimates of an Interdependent System of the Investment Equation For Private Industrial R&D Expenditure in the United States, Europe, and Japan .A=de by the Instrumental '~?ar'ables Method (based on data for 1967-1983 expressed in bit lions of 1972 dollars; t-ratios in parentheses) Interdependent Ins trumen~cal Variables Estimates 1967- 1983 Variable and Lag U. S . Cons tonic GI)E (-1.~* GCRD fist. BRD E P(-1) U.S. °9.04 -8.67 (6 ~ 08) (7 . 35) 0 78 <'78 (10e 83) (1lo 34) .64 (2. DO) _st. BRD J P(-1) .40 (~.59 ) Est. BRD E+J P(ol) Est. BRD US P(-1) .51 (7O 99) Japan -3 ~ 2S (6.~4) O 0091 .0007 (S. 17) ~ . i6) 18.05 (4. 12) O 37 (I. 56) Europ e -3 .93 (2.77 ) .00,4 (5. 52) La 26 (] 67) ~ ~ JC (1. $8 ) .22 .32 (2. 82) (3 O 67) R2 corrected Durbin-Watson Statistic Rank serial correlation o f res iduals S tandard Error .990 .990 .986 .985 loS5 1~65 1~61 1~62 .038 .035 . 04 ° .015 *For the United States, gross output of the business enterprise sector. Source: lard M. Levy and Negator E. TerlecXyj. "Trends in Industrial RED Activi~cies in the United States, Europe and Japan, 1963-83. n Paper given at ache National Bureau of Economic Research Conference on Productivity Growth in Japan and the United States, Cambridge, MA, August 26-28, 1985.
the three economic areas . The coefficient for the United S tates is much smaller than that for Europe and vastly smaller than ache coefficient for Japan. Howe~rer' both absolutely and relatively, ache amounts of government-funded industrial R&D are very small in Japan and cons iderably smaller in Europe than in ache United S vacates a Also, the character of the industrial R&D funded by the government is quite different outside the united Stances. In Europe and Japan, government funding is oriented much more toward supporting private incus try innovations and productivity, while, tn the United S Sauces, it is oriented primarily to government obj ectives O ME I~IESTM~1T E1JNCTION FOR PRIVATE it&i) ., lathe best analytical cool for approaching the relationship between federally and privately funded industrial R&D is the investment function for private RScD expenditure. A number of the studies discussed in this paper have used tha~c approach. It has the advantage of attempting to include all the factors that bear or. R&~) investment behavior. Expenditures by private firms for research and development constitute every bit as much of an in~res~cment as expenditures for buildings or equipment. Indeed, R&D inves Amend is guided by many of the some variables as investment in plant and equipment. Among those variables are corporate profits, the growth of the economy, bus iness cycle, and corporate taxes, even if the coefficients describing the effects of the variables va::y among the Aches of invest:ment. Two recent developments provided new perspectives on the U33. R&D investment function. As mentioned already, Levy and Terleckyj identified significant positive effects of foreign R&D, specifically of private R&D expenditures in Europe and Japan, on U.S. R&D investment. Lichtenberg has shown ~cha~c separation of the R&D. intensive federal purchases of goods and services from ache rest of the gross national product (GNP) reduces the estimated effect of federal contract R&D on private R&33 and results in a better overall explanation of private R&D expenditures than does a model restricted to a single demand variable. in attempting co combine the results of recent studies, the author has estimated an aggregate R&D in~restmen~c function for the United States in which the priorate Rid) expenditure (PRD) depends, contemporaneously, on Me output of the U. S. private sector, excluding federal defense purchases of durable goods. (QExGDD); on these purchases (GI)I)) and private expenditure for industrial R&D in Europe and Japan (PRO E+3), all witch one year lag; and on the government contract R&D (GCRD). The estimate of this equation, made by the generalized least squares method, with the POD data for 1964-1984, expressed in millions of 1972 dollars (as deflated by the GNP deflator and with foreign data converted deco U.S. dollars by the purchasing power pari~cies ), is: ~ 6 4 _
PRO - - ~ 8 ~ 5 ~ . 0 0 ~ QExGDD ( - 1 ) (3 . 04) (6 . 82) .080 (lDD( -1) + .244 GCRD (S.03) -(9.30) + .249 (5 . 16) BRD E+J ( -1 ) Durbin-Watson Statistic o 1. 947; Adjusted R Sq. . 997 This result supports Lichtenberg' s finding that the government marice~c, and here, specifically, ache market for defense disables, has a much larger effect per dollar on R&D expenditures than the rest of ache market does. It also confirms the existence of significant aggregate- level effects of government- funded R&D . lathe magnitude of that effect a: . 24 is similar to that found- in earlier work by Levy and Terlec~rj ~ for the longer time period 1949-1980,, as reported in Table 1. The effect of foreign R&D estimated in ache above equation is also highly significant but appears to be smaller than the estimate by Levy and Terieckyj 2 for ache 1963~1983 period, also reported in Table I. The results obtained with the separation of the defense durables market appear more plans ible, but further attemp As at a new synthesis cannot be pursued here. There is also the question of the stability of U. S. I&) investment behavior in recent decades. There are indications that foreign competition may have induced increased investment in R&D in the United States Charting at some point in the 1970 ' s . The recent Levy and Terleckyi paper indicates that the U. S . R&D in~restment function may have shifted and that priorate R&D expenditures have become influenced more by government R&D and foreign R&D and less by ache growth of the domestic mare and the business cycle. Table 3 provides estimates of the investmen~c function for the period 1963-1983 and for Nero shorter periods within those years. Because ache time periods are so short, some variables witch smaller statistical effects were omitted. No meaningful. results were obtained with an attempt to separate the purchases of defense durables in these estimates. lithe present estimates, while suggestive, provide only a starting point for subsequent explorations, especially at the more detailed levels of sectors, industries, and firms, and involving further differentiation of the product markets. \ - 16; -
parentheses Variab le. TABLE 3 Estimates of the Investment Equation for U. S . Private Expenditure for Indus~crial R&D (all variables in billions of 1972 dollars- ; t- ranchos in (Generalized Least Squares Estimates 196 3 - 8 3 1963 - 72 1973 - 83 Constant -4 O 29 - 6 O 21 - 3 . 00 (-3. 73) (-4. 81) (-3. 30) Q ( ° 1) . 0042 . 016 . 0014 ( 2.44) ( 2.73) ( 1.21) G<:RD .52 .29 .7S ( 9.60) ( 1.30) ( 7.73) BED E+J Prier. o 67 ~ 11 ~ 6 I; (11- . 02) ( . 38) (12 ~ 68 ) R2 corrected .995 .990 ~. 000 Durbin-Uatson S tatistic Rank serial correlation of residuals S tandard Error 1.88 .038 .21 1 O 68 - . 01~7 .13 1.97 - .091 .22 Source: I:a~rid H. Levy and Nestor E. Terlec~yj. "Trends in Industrial R&D Activities in the United Stances, Europe and Japan, 1963-83." Paper given at the National Bureau of Economic Research Conference on Productivity Growth in Japan and the United States, Cambridge, MA, August 26-28, 198S. < - 166 a
REFERENCES Z~i Griliches. "Research Expenditures, Education and the Aggregate Agricultural Production Function, ~ American Economic Review, Sol . LIV, No . 6 (December L964), pp . 961- 974 R. £. Evenson. "A Century of Agricultural Research and Productivity Change Research, Invention, Extension and Productivity Change in U;S. Agriculture: An Historical Decomposition Analysis. In Research and Extension Productiviry in Agriculture.. Edited by Ahmed Araji. Moscow, Idaho: 1980, pp . 146- 228 . 3. R. E. Evenson, Paul E. Waggoner, and Vernon A. Ruttan. "Economic Benefits from Research: An Example front Agricul~cure, n Science, Vat. 205 ~1979 ), pp . 1101-1107 . 4. Charles River Associates, Inc. Produ~tivi~cy Impacts of Government R&D Laboratories: The National Bureau of Standards Semi cond uc For Tec*no 1 ogg Program - - A n Prod uc ~ i on Func t~ on Approach O n Report prepared for the National Bureau of Standards, U. S . Department of Co~erce, April 1981, CPA Report 49 3, Volumes I and II . S. 7. Nestor E. Terlecicyi. Sources of P-oductiv,cY Advance. A Pilor Study of Manuf acturing Indus ~cries, 189g ~1953 . Ph . D . Dissertation, Columbia University, 1960 . Jora Hinasian. ache Economics of Research and Development. n In The Rate and Direction of Inventive Activity: Economic and Social Factors. Edited by R. R. Nelson. Princeton: . Princeton University Press for the National Bureau of Economic Research, 1962, pp . 93 -142 (originally a doctoral disser~cation at ache Univers ity of Chicago, 1960 ~ . By N. Leonard. "Research and Development in Industrial Growth, " Journal of Pol itica] Econo - , Vol . 79, No . 2 (March/April 1971), pp . 232- 256 . 8. Nestor E. Terleclq3. Effects of R&D on the Productivity Growth of Industries, An Exploratory Study. Washington, DC: National Planning Association Report No. -140, 1974, 60 pp. 9. Z~ri Griliches and Frank Lichtenberg. PROD and Productivity Growth at the Industry I-evel: Is There Still a Relationship?" In R&D, Packets, And Productivity. Edited by Zvi Griliches. Chicago: University of Chicago Press, 1984. 10. Madrid M. Ivory and Nestor E. Terlec~;. "Effects of Go~rernmen~c R&D on Private In~resment in R&D and Productivi~cy," Bell Journal of Economics, Sol. 14~1983), pp. 551-561. 1 6 /
11. Terlecky:, 1974, op. cit. 12. Nestor i:. Terlecky] . "R&D and the U. S. Industrial Productivity in the 1970s. n In Transfer and Utilizat~or2 of Technical Knowledge. Edited by D. Sahal. Lexington, lIA: D.C. Heath, 1982. 12. Terlecky], 1982, op. cit. 13. Ralph C. Lenz, John A. Machnic, and Anthony W. Elkins. The Influence of Aeronautical R&D Expenditures Upon the froductivi to of Air Transportation. Dayton, Ohio: University of Dayton Research Institute, July 1981. 14. Nestor E. Terlecicy]. The Time Pattern of ache Effects of Industrial R&D on Productivity Growth. ~ In Incerindusery Differences in Productivicy Growth. Edited by John W. Kendrick. Washington, DC: American Enterprise Institute, forthcoming. IS. Nestor E. Terleclcy]. "A Growth Model of the U.S. Communication Industry, 1948-80.~ In Information and Com:3unicarions Economies: New Perspectives. Educed by Meheroo Jusawalla and Helene Ebenfield. Amsterdam: Norwich Holland Publishing Company, 1984. 160 Terlec~;, 1982, op. cit. 17. David M. Levy. Estimating the Produc~civity of Public Goods. ~ Public choice Working Paper 8436. George Mason University, 198S. 18. F^ancasco Forte and Jams Al. Buchanan. The Evaluation of Public Services, n Journal of Political Econo - , Val. INFIX, No. 2 (April 1961), pp. 1~07~1~21. 19. Jeffrey Carmichael. ·The Effects of Mission-Oriented Public R&D Spending on Priorate Industry, ~ Journal of Finance, Vol. 36, No. 3~1981), pp. 617-627. 20. A. N. Link. Can Analysis of the Composition of R&D Spending,. Southern Economic Journal, Sol. 48~1982), pp. 342-3490 21. E. Mansfield and L. Switzer. "Effects of Federal Support on Company-Financed ~ and D: The Case of Energy,. Management Science, Sol. 30, No. 5 (May 1984), pp. 562-571. 22. John T. Scott. "Fin Versus Industry Variability in R&D Intensity. n In Red, Patents, and Productivity. Edited by Z~i Griliches. Chicago: University of Chicago Press, 1984. · 168 -
23 . Frank Lichtenberg. "The Relationship between Federal Contract R&D and Company Rid), n American Economic Review, Vol. 74 ~ 1984), pp . 73 - 78 . 24. Lawrence Goldberg. The Ink 7uence of Federal R&D Funding on The Demand for and Recurns to Industrial Reck. CRC Report No. 3880 Alexandria, VA: Public Research Institute, Center for Naval Analyses, October 1979. 25. Richard C. Levin and Peter C. Reiss. "Tests of a Sch~peterian Model of Red) and Market Structure. ~ In R&D, Parents, and Productivity. Edited by Zvi Griliches. Chicago: Uni~rersit: of Chicago Press, 1984. 26. Lichtenberg, 1984, op. cit. 27. Terleckyj, 1984, op. cit. 28. 'rerlecky;, 1985, op. chic. 29. Levy and Terlec~r;, 1983, op. cit. 30. Nestor E. Teriecky;, David M. Levy, and Yuichi Kaneko. NPA Model of UO S ~ Economic Growth 1948-1983: Description and ~ escing. National Economic Projections Series, Vol. 83-N~3. Washington, OC: llational Planning Association, 1983. 31. Leery and Terleckyj, 1983, op. cit. 32e Mansfield and Cwitzer, op. cit. 33. Levy and Terlecky;, 1~983, op. cit. Link, op. cit. 35. Charles River Associates, Inc., op. cit. 36. Leery and Terleckyj, 1983, op. cit. 37. Scott, op. cit. 38. David M. Levy and Nestor E. Terieckyj. "Trends in Industrial R&D Activities in the United States, Europe, and Japan, 1963-83. " Paper given at the National Bureau of Economic Research Conference on Productivity Growth in Japan and the United States, Cambridge ~ ~A, August 26 - 28 ~ 198S . 39. Ibid. 40. Frank R. Lichtenberg. "Assessing the Impact of Federal Industrial Research and Development Expenditure on Priorate Research and Development Activity, " this conference O 169
41. Levy and Te~lecky;, 1983, op. cit. 42. Levy and Terlecicy;, 1985, op 43. Ibid. ; # . cit. 1 70
