Papers Commissioned for a Workshop on the Federal Role in Research and Development (1985)

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.

FEDERAL RESEARCH .~N'D DE\ELCP.~EN Ably U. S . AG2IC7-L=u~E: At?; AS S ES SCENT OF ROLE A.~3 PRODUC-IVI~ EF.- EC 1 S Robert D. Weaver Pennsylvania S tate Univers i ty The agricultural sector of the U. S . economy presents ~ sper ial case for the study of the role and impact of federal ly funded research and development (R&D). On the surface, the long history of both direct and indirect federal involvement seems to provide a record of an experiment from which important general conclus ions might be derived. Indeed, this possibility has not gone unnoticed by ?°3 icy designers and economists who have sought empirical evidence that might rationalize a federal role in A&) in various sectors of the economy. In fact, ache literature is voluminous and detailed. Hi s paper wit ~ describe the characters sties of the aged cultural sector and review briefs the nature of the federal role in Rap. Ha~r'ng Fresco ibed the context, the paper then will develop ~ taxonomy of the ef~ec-s of federal Rid). Avant fable e~r'der.ce concerning the extent of each of these effects will be presented and tics credibility assessed. THE STRUCTURE OF THE AGRI CIJL=-RAL S EC ~ OR The structure of the aged cultural sector of the U. S . economy has changed dramatically over ache last 200 years, yet several characteristics have persisted. In general, production takes place on family owned and operated farms, which produce ~ divers if ted product line, often involving both crops and livestock. During the las ~ 100 years, the average area of land operated has increased steadily, while farm population has declined dramatically as terminologies that allow intensification in use of these two limiting resources have been adopted. Detailed histories of the evolution of the agricultural sector can be found in Arndt et al. or Peterson and Hayami. Despi he these trends, farmers have remained atomistic actors in the agricultural sector's economy. Individually, their impacts on that economy are negligible. In the short run, they respond to, rather than dateline, existing price structures and available technology. Only the aggregate of their anticipated or actual responses affects prices and available technologies. On the 34 ~ -

demand s ice, co~.s~;ne~s of food products can be charac derided in the same cents- - their decisions are affected by, but do not dete ~ mine, price and available product variety. Although these characteristics conjure up an image of a sector that economists could justifiably label as perfectly competitive, several complicating characteristics destroy the appropriateness of that label. On tne farm side, producers operate a process over which ~ hey do not have complete control. Because climatic events play such a critical role in determining ache productivity of marketed, con~ro.iiable inputs, output petrels are uncertain. Input use and output patterns and levels are the result of initial planning and of adaptation to climatic events . The time in-ansi tenets of the production process not only al lows exposure to uncertain climatic events, but also means that the output prices a~-ailacle at haves. are uncertain at planting time. These two sources of uncertainty force farmers to malce production plans based on anticipation of 1~ kely future outcomes . More important, these uncertainties have led to the development or both private and pubic c institutional mechanisms to reduce the extent of uncertainty to which producers are exposed. One of the most important of those mechanisms during the last 50 years has been direc ~ federal intervene on into farm commodi ty markets to fix prices at levels tha- ensure adequate food suppiv and stabilize the profi~c;~ility of farming. Anile the nature of such programs has varied, a generai characteristic has been their provis ion for farmers of what has been called a "price support" or, in economist' s jargon, a guaranteed minim price. During the postwar period, excess domestic supplies of many commodities ted to the insti, union of requirements that farmers reduce acreage planted in certain crops as a condition for price support . The impor Lance of those policies for ache performance of the sector cannot be understated and is ~ subj ect that has received considerable attention. In general, the conches ion can be drawn that these policies have enhanced the level of and reduced the uncertainty of net farm income (profit) and, thereby, critically affected incentives and the budget available for investment in new technologies. An important structural element of the production side of agriculture is the supply industries that provide the farmer with sped ialized inputs, such as seeds, fertilizer, herbicides, pesticides, farm machinery and equipment, and farm building products . Me nature and cost of these products ~ accounting for about 85 percent of farm revenues ~ play a critical role in determining the level of fame income and the productivity of limiting resources. Their role has placed the supply industries in key positions as providers of new technologies and impro~red3inputs to the production side of the agricultural sector. Leibeniuf~c provides a useful overview of ache structure of these industries. - 348 —

A . he ccr~su~z~ on side, a simi1 ar revolution occurred in the era ? for ~o acrid par II. two principal trends during that period were a conch nucus shift of ache processing function from the household to an intermediary process ing industry and the accompanying alienation of the place or production from Ached of consumption. At present, the cost of marketing and processing food in the Unit ted States accounts for more than 75 percent of its retail price. The magnitude of that percentage emphasizes the importance of understanding the role of federal R&D ire food marketing and processing if the ultimate obj active is to understand the Apace of R&D on levels and costs of nutri ~ion. The above discussion provides ~ s imple, conceptual schematic or describing the ways in which federal R&D affects the agricultural sector. Clearly, the federal role ~ s pervasive if one focuses on all levels of the sector and a complete characterization of all ~ nputs, including power supply, transportation, and communica~cion sys Gems . In this sense, the agricultural. sector is not unique. However, the spatial decentralization of both demand and supply of food maracas, the Prague of information concerning price and climate flus tuations, and the importance of time in both supply and marketing suggest ~hat! for the agricultural sector, such infrastructural incus tries as transportation, communication, and information processing are crucial in determining income levels and variability. Despi be this complex) By, if focus is placed only on federal R&D that is related directly to one of the structural elements of the sector ~ the problem of tracing the federal role is s implified. Further, if the enquiry is limited to the production side of agriculture, the problem becomes feasible and within the scope and limits of this paper. TlIE FEDERAL ROLE IN R&D Ri:IATED 3)IREC FLY TO AGRI CULTURAL PRODUCTION HISTORICAL ORIGINS Overviews of ache federal role in R&D have been provided by the General Accounting Office, 4 to: Cooperati7e State Research Service, Johnson and Widower, Hadwi~er, F:izharris, ~ and Peterson did Fitzharris. Schultz, True, and. Knoblauch provide early histories of agricultural R&D in the United States which are useful for establishing the social motivation for public in~rolvernent in R&D. Even in the absence of recent criticisms of the appropriateness Of a public role (see, for example, Marshall 4 Lepkowski, . Science, 1 and National Academy of Sciencesi7), identification of such a rationale and assessment of the impacts of federal involvement seen necessary, especially given a long history of private sector involvement. _ 349 —

The private sector role is highlighted by Fitz~arris, l3'l9 who cites the i~mpor~anr~ of private initiatives in developing arid marketing improved mechanical equipment, in publ ishing Farm- related magazines and other information sources, and in forming agricultural sock sties in the 1800's. However, rapid expansion into new lands with variations in climatic, soil, and human resources conditions led to increasing demands for new and localized products and technologies. In addict ion to the rapidly enrolling set of small markets for new products, other factors limited the poten~cial of priorate profitability for new agricultural products and methods. Although mechanical innovations were developed frequently, ache identification and development of local plant varieties and anima' breeds were slower and more costly processes from which private benefits were not always appropriac'e. In most cases, the profits from new methods could be reaped quickly by those who had not ~ nvested in the development of the technology. This phenomenon, knows as free - riding, des boys incentives for research and development, and rationalizes a public ro ie . Fitzharr~s reports that, by the mid-~SOO's,. agricultural societies were able to obtain state funding to support some of their efforts. In this way, all Chose benefiting from R&I) would pay, at least indirect' y, through taxation. the political base in the agricultural societies was strong, and by 1862 the U. S. Congress had transformed the agricultura1 division of the U. S . Patent - Office (founded in 1839) into the U. S. Department of Agriculture (I;SDA) and passed the Harrill Land-Grant College Act. he latter provided federal ~ ends to states to establish colleges of agriculture, which were encouraged to create experimental farms to serve as teaching laboratories as well as research facilities for development of locally adapted seed varieties, animal breads, cultural practices, and mechanical innovations. By 1887, the Hatch Agricultural Experiment Station Act put full congressional support behind the concept of state experiment stations and a link between research and teaching. That those activities would be complementary and provide j oint benefits was presumed. A federal role in direct field education of farmers was established by Congress in 1914 with the Smith- Lever Act, which provided federal suppor ~ o f agricultural extension we're. Lois meant that the federal go~rerr£ment was involved directly in three activities; research , teaching , and extans ion , which j ointly provided problem recognition, research, developments and. diff,,-~ion of results. The state-based agricultural research system relied heavily at first on federal funding (see Table t), although state funding was continued and expanded over the years. In parallel with the state systesIl, Congress established authority for the USI)A to develop federal research facilities under the Agricultural Research Service. From this history, it can be concluded that the pal itical power of the agricultural societies facilitated public support of research and development services that the priorate sector appeared incapable - 350 —

T.~3L~ i 3^p?ro?ria~ior~s for Public Research (current dollars , ~ ~ millions) State Acri, cu ~ tural Year Experiment seat~onsa USDA 1915 1916 lal7 1918 1919 1920 1921 1922 1923 1924 1925 1926 lg27 1928 1929 1930 1931 1932 1933 1934 935 936 1937 1938 1939 940 1941 1942 1943 1944 1945 1946 947 948 949 950 951 952 1953 1954 406 3.8 3.8 4.2 4.2 5.0 5.2 6.3 7.0 7.6 7.3 8.9 9.3 11.4 12.0 13.1 12.S 12.1 ~4 11.1 11.1 12.1 12.9 4.8 15.6 6.8 6.7 7.7 17.5 18.8 19.8 23.6 28.1 35.3 39.9 48.2 50.0 56.4 60.0 68~0 - :51 - 6.0 5.2 5.8 6.3 6.9 7.7 7.8 8.2 8.5 8.4 9.3 10.2 10.5 11.7 13.8 15.5 16.7 16~1 13 11.1 11.4 14.4 16.4 18.0 23.3 22.1 21.4 22.0 21.8 22.0 22.9 27.6 33.2 38.2 46-0 46.8 45.1 45.0 45.3 46-0 (Conti nued )

~.~ i _ i ~ ccr~tiru~d; . . St ate Agri Cal ~ tural Year Experiment Stationsa USDA 1955 73el3 53~4 1956 ~ 85e 4 59 ~ 6 1957 92e ;? 86 ~ 6 1958 10509 83~7 1959 110.3 gg oO 1960 120 ~ ~ 105 ~ 2 1961 127.3 128 e9 lg62 142. 1 126~4 1963 151.3 136 ~ 1 1964 169e 3 14~3.8 196~; 18 1 ~ 8 ~ 192 ~ 5 1966 223~4 212~7 1967 239 e 7 218~5 1968 261.5 219 .S 1969 274~0 213~2 l9tO 296e 1 238 ~ 7 aFederal plus nonfederal funds available Excludes fees and sales. Source: S tate -Agricultural Experiment S Cations: Data for 1915 -1960 Taken from "Report on the Agricultural Experiment S cations, " published by the Office of Experiment Stations through 1953 and by ache Agricultural Research Seduce from ~ 9S4 through 19 60, Washington, OC . Data for 1961 -1973 taken from n Funds for Research at State Agricultural Stations, " Cooperative S tate Experiment S Cation Service, Washington, DC . USI)A: Data for 1915- i953 taken from "Repor~c to the Director of Finance, n USDA, Washington, DC. Data for 1954- t973 taken from "Appropriations for Research and Education, " prepared by the Office of Budge e and Finance, USDA, Washington, DC . - 3~9

of pro lid -.g at air -he level demanded privately or at ~ the leered that right He rest for socie By in the long run. Peterson and Fitzharr's2 argue that through the early 1900's, the principal focus or experiment s cation research was local adaptation and maintenance research ,. which attemp ted to achieve and maintain existing potential productivity levels in local conditions encountered during; exacts ion into new land areas . This argument suggests that during those years, eaten witch public involvement, ache demand for new technologies and products was not satisfied. CURRENT =VELS OF I~JOLV=.ENT Stepping from his ~orical origins teethe current situation, the Cooperative S Bade Res earch Service indicates that the 19 8 2 agricultural R&D budget for the United States was approximately $3 billion. Roughly 54 percent was funded by the private sector, 17 percent by s cares, and 29 percent by the USDA. Several changes in federal funding hare been made since the 1960's. In 1965, a researc-. budget was provided for the so-called ~ 890 schools (black land-grant schools ), although the first funding was in 1967 and for only $283, 000. Later, in 1977, a funding formula was adopted for those schools. Also in 1977, special funding was provided for animal health and disease research conducted in Colleges of veterinary medicines In 1965, a special grant program was established for direct funding of research in either ongoing or new promising areas of agricultural science as selected by the USDA. .\ similar though more broadly compete time grant program was introduced in 1977 to allow the USDA to encourage high-priority research. Tabl e 2 summarizes average annual agricultural research funding over the period i97 8 to 19 8 2, as reported by the General Accounting ()ffice . EFFECTS OF FEI)E:RAL R&D FOCUSED DIRECTLY ON AGRICULTURAL PRODUCTION The conceptual schematic developed earlier presents a useful framework for enumerating the effects of federal R&D on the agricultural sector. Again, for the purposes of this paper, the discussion is limited to R&D focused directly on the producer on s: de of agriculture. The schematic identifies both points at which R&D could affect the production process and the linkages between R6cD and the final effect. On the input side, R&D has aimed to expand the produccivi ty of limiting resources (~and and climate) through the use of purchased inputs . Innovations are al her embodied in new product charac~cerist-~ cs or manifested in new husbandry or management practices. In the latter case, the innovation is said deco be disembodied. The effect of either type of innovation is to change

i9~30 school furring I, 128 987 T-~3L~ 2 .~_ric~',''rcL1 Research t' Is Received by Region by Funding Category average fiscal year 1978 - i982 ~ , . . . North Central Northeastern Southern Western funding Type Region Region Region Region Total Hatch Act formu] a funds S 34,833 S 20,736 S 41,034 S 19,6g! S Il6,294 Competitive grants 5,483 3,677 2,490 3~417 15,067 Speci a] research grants 4,621 2,451 4,338 4~154 15,564 Animal heal th and disease l'638 514 I'309 1,005 4,466 TO, 893 Sub tote ~ 4 7 . 70 3 28 ~ 36 5 63, 949 28 ~ 267 158 ~ 28 4 ARS funding 70,187 94,2'97 102.122 64,025 3~0,531 Tota ] Federal ~117~890 122.662 166,071 2292 498,gI5 State Appropriationsa i25,`375 52~274 194 423 128.581 ~l, 153 , ~ Total S243, 765 S174~936 S360,494 S22D,873 S1,000~068 State/Hatch 3.39 2.38 4.47 6.16 4.06(mean) aComplete data were available only for 1981. Source: General Accounting Off' ce. Federal Agricultural Research r ending: Issues and Cor.cer``s. October 20, 1983 . - 354 -

the l.evel of potential agricultural production achievable from any particular bundle of inputs . Such changes in potential ~ imply that for a given set of 'reliefs alum probable output prices, for a given set of input prices, and for given levels of flows from limiting resources, the producer '~ill exploit new potential by changing both the mixes of inputs used and the outputs produced. Changes in producer demands for inputs and in supplies of outputs affect the producers ' markets and result in changes in input and output prices, Ache incomes of input suppliers, and the expenses of output consumers. Following this immediate, or short- run effect, price changes further induce changes in inputs supplied and outputs demanded, inducing more price changes. The final effect of this longer term adj ustment process is a new equilibria in which there are new mixes of inputs and outputs, new prices, and new laurels of incomes, profits' and expenses for those involved in the production economy. The relative importance of these possible effects depends on the nature of the innovation, the speed of its adoption, and the speed with which the market adapts to induced changes. In the short run, the effects can be summarized as: ~ 1) changes in the mix of outputs supplied, (2) changes in the mix of inputs demanded; (3) changes in productivity of limiting resources, and (4) changes in returns ~ ~ limiting resources. In the long run, changes in prices are induced, and those changes force all ret aced elements of the economy to adjus c resource allocations.- Given this general taxonomy of. the effects of R&D, the analys is of the impact of federal R&D on the a~gricul Mural sector requires empirical chararterizaticn of the R&D process, identification of the federal role in that process, and l~ nkage of that process deco the effects. Although it would be of interest to review the nature of pas. innovations affecting agriculture production and to identify the relative roles of federal, state, and private funding, only very limited information is available. Since 1965, the federal-state systems of experiment stations have collected a variety of information characterizing ongoing research and results of completed proj eats . The initial oh; ective of the resulting USDA data base was to facilitate centralized planning and allocation of research effort. In ache priorate sector, an even more limited record of R&1:) ac~civity is available. Although ache National Science Foundation reports expenditure and other indicator series for bo th pride and public sectors, the series are classified typically by discipline, no~c by area of final incus trial application. The weakness in available data implies that while the federal R&D role in the federal - state research establishment can be identified, the federal role cannot be identified in basic R&D conducted by nonagricultural departments in universe ties and colleges. These conclus ions deserve further clarification. 17: e problem of monitoring the R&D process depends upon an understanding of the general nature of the process as well as an enumeration and - 355 -

classification of possible processes by specific characteristics. lathe first condition, understanding the process, is one in which our knowledge base is meager, while, for the second condition, the appropriateness of She classification system depends on the user. Johnson and Wittier 4 classify research into three categories: basic or disciplinary research, multidisciplinary research, and applied or problem- solving research. This class~ficat~ on can be used to place federal agr~cul~cural R&D in the context of the broader A&) effort observed in the United S tates . Further 9 it allows ~ useful illustration of the problem of measuring the federal role in agricul, ural R&D . Johnson and Wittwer define disciplinary (DI~C) research to include the basic sciences--physics, chemistry, biology, mathematics, and economics--wh~ch focus on system studies and methods to solve generalized problems. `Multidisciplinary (MD) research is focused. on synthes in ing dis c ip 1 inary res en rch resul ts into a sys tem ~ f knowledge that allows analysis of an observed physical system; for example, agronomy, horticulture, and food science research fall into this category. The National Science Foundation definition of basic research would include both DISC and ED research. Problem- soldering (PS ~ research is similar to what the Na,tonai Science Foundation labels as applied research. It attempts to solve specific problem encountered with the functioning.of an observed sys tem. Johnson and Wi ttwer argue that the priorate sector and the agricultural colleges play parallel roles in MD and PS research =.e Johnson and Utetwer class if icat~on shows the difficulty of es Publishing ache federal role . Implicit in their class) fixation is a linear conditional linkage of research ac~c~ri~cy from disciplinary through multidisciplinary to problem solving. This establishes the dependence of direct MD and PS agricul~cural research on disciplinary research. Because of this linkage, even if we wanted solely to measure expenditures on R&D, the federal role would be difficult to establish because of the conditionality between various types of research. That is, although federal. expenditures on problem- solving research might be measured, ache dependence of the effectiveness of such expenditures on prior federal expenditures on disciplinary research makes measurement of such DISC expenditures a prerequisite of any understanding of the federal role. In the absence of further details on R&D effort and effects, or on outputs from the R&D processes that might affect agricultural production, researchers attempting to measure the effects of R&D have limited their enquiries to R&D activities at state experiment stations. Exceptions have been a variety of crop-specific or livestock- specific case studies that trace particular R&D activities and the effects of the adoption of innovations that they produced. Lee remainder of this paper will focus on historical studies of the effects o f experiment station R&D activities . 3S6 -

EVIDENCE Or E: ^- ACTS OF EXPERIME:~T ST.AT~ ON R&D ACTIVIT YES ON AGRI CULTURAL PROI:)UCTI ON Empirical measuremen~c of the effects of R&D on an economic system requires: ~ 1) characterization and measurement of outputs of the R&D process, (2) an empirical characterization of the economic system, and ~ 3 ~ a model of the lir~lcage between the R&D process and the economic system. In the case of the role of R&I) in agriculture, Ale maj ority of evidence takes ~ sharp focus on the relationship between R&D expenditures by state experiment stations and characteristics of the agricultural production process: output level, productivity of particular inputs or focal factor productivity, and, in a few studies, utilization of inputs. These studies ignore second-round effects of output or input changes on price charges. A second the of study is the case study of ache effects of particular innovations. PROBLEMS IN MEASIJRF~ OF RED EFr-CTS ON PRODUCTION Before reviewing evidence presented ~ n the expenditure studies, a brief discussion of the problems involved will place the studies in context. The general problem of relating the R&D process to a production process in an empire Cal model is complex. The vast li Denature that studies R&D processes in the private and public sectors makes clear that these processes are extremes y diffi cult to characterize due to the variations in final obj ectives, the available human and physical resources, the time horizons over which they are conducted, and the performing organizations. Were is general agreement in the literature that the R&D process involves a continuous application of a variety of inputs. However, in contrast to the traditional view of industrial or agricultural production processes, in an R&D process the application of inputs can be said at most to condition the probability of discovery or attainment of ob; ectives . This appears to be a characteristic of basic, applied, and development - oriented research . There also appears to be a consensus that there exist certain critical levels of potential discovery at which the response of ache probability of discovery to further input use changes. In general, the time and cost of a discovery are highly uncertain at ache initiation of a research activity, and the degree of uncertainty often is not reduced as time passes and expendi-=ure~s are made. This uncertainty implies that at any time during an R&D process, ache necessary arid feasible input requirements for completion of the process are identifiable only in general terms. In economics terminology, the frontier of technological possibilities of any R&D process is uncertain arid can be described only through approximation or identification of minimal critical levels of sets of particular inputs. Awhile these characteristics make it difficult to plan research activi. ies, they also complicate our ability to evaluate

empirically ache efficiency of Rid:) processes or, for tonal matter, e':en to characterize their levels empirically. On the output s ide of the R&I) process, the literature leaves As in a similar situation: The output is a mix of a variety of characteristics Gnat may not have any phys ical, unifying manifestation. Those characteristics have led many economists to conclude that use of a production- function concept for describing an R&D process is inappropriate . However, despite those conclus ions, there remain consensus and evidence that R&I) output levels are conditional, though possibly not ~ n a continuous relation, on it&l) input leve is . Empirical evidence supporting the existence of a relationship has found s tatis tically s ignif icant and ~ nCuition- re inforc~ng ret rations between various indicators of output ~ for example, patent numbers, value of the firm, sales, or profits ~ and a variety of indicators of input flows (for example, R&D expenses, personnel counts, or capital goods investmer~t) . However, the overwhelming conclus ion that can be draw from this literature is that, at best, we can hope to observe statistical indicators of an R&D process, and the indicators are likely to be, "~c best, s stochastic measures of the unobservable input characteristics and output flows involved in the R&D process. On ache surface, empirical characterization of a production process might appear more straightforward than the task of characterizing an R&l:) process. Although this is a safe conclusion, the problem on the production side is not without its difficulties. As noticed ~ n a rec en" review paper ~ see Weaver ), the specification of an econometric model of a production process mus t rely on a combination of prior knowledge of both the production process and the economic behavior of the managers of the process, and on the artful use of economic and econometric theory. Knowledge of the economic behavior of managers is required because economists do not have laboratories in which the production process can be observed in isolation. Instead, input and output combinations have been filtered through a manager' s response to economic incentives; that is, they are a result of economic behavior. In the case of agricultural production decis ions, the earlier description of the sector highlighted several of their characteristics. Weaver 6 ~ concludes that models of aged cultural production decisions should recognize ache following elements of these decisions: (~) the multiple-output, multiple- input nature of production, (2) the importance of climatic events, (3) the role of output price uncertainty, and (4) the presence of government programs that affect incentives. The link between R&D and the production process in agriculture is one that introduces further difficulties into the problem of measuring the effects of R&D. lathe spatial decentralization of agricultural production means that diffusion and adoption of - 358 -

innovations depend cr' tical7 y upon the speed of the information f' off and the nature of incentives and constraints on adoption. The inertia imposed on diffusion and adoption by existing institutions and infrastructure complicates further the continuity of ~ he relationship between R&D activity and i As effects on production. To summarize, empirical measurement of the effect of federal R&D on agricultural production is complicated by (1) the fundamental nature of R&D processes, which renders them difficult to describe empirically; (2) Me complexity of the agricultural decision process, yet the necess icy of modeling aches process if ache underlying; agricultural production process is to be quantified; and ~ 3 dye development and adoption process, which separates the products of the R&D process from application in "he production proces s . EVIDENCE OF R~ Erg ACTS ON AGRI=~ PRODUCTION Within this context of specification of the empirical relationship between R&D ac~civi By and its effects on production, available evidence for the agricultural sector can be summarized and interpreted. Me literature has f ocused on ex posse evaluation of the rate of return to expenditures by state experiment stations2 A review of methods and- results is presented by Norton and Davis 8 Two approaches have been taken. In the firs t, the elrect of research expendi.~r~s on ache supply of aggregate agricultural output is estimated. By j oint cons ideation of the supply function shift and consumer demand, an. es timate of the change in net economic benefi As is derived. The alternative approach has focused on ache direct effects of research expenditures on production by introducing~hem as an input in an econometric model of the production function. However, an ~ mportant difference ~ n this second approach is its conceptualization of a generalized it&l) process as producing a continuous flow of i nnovattons that lead to continuous effects on production and supply. As noted in thp previous section, this characterization may be inappropriate. Schultz29 estimated the supply function effect by calcllla~cing the value of agricultural inpu~cs that had been saved through the use of new production technologies. The value was interpreted as an · economic benefit and compared with an esthete of the research cost. In his rough approximation, Schultz evaluated the value of inputs saved by comparing the cost of producing the aggregate output observed ire 1950, using 1950 technology, with the cost that would have occurred if 1910 technology were used. Aggregate output per unit of aggregate input was used to characterize the two technologies. Although Schultz was only suggestive in interpreting his results, those results indicated that for an expenditure of $ 7 billion over 40 years, the lowest estimate of inputs saved was about $9 . 6 billion per year. Although Schultz never argued that his - 359 —

results were anything more than suggestive, they stimulated mare systematic cons iteration in later hi _erature . Griliches30 presented the first such detailed analysis, but carefully inserted the caveat, 'The calculated rate or return is an es timbre, subj ect to a wide margin of error, '' and claimed that i. could be used bes. as an indicator of only the order of magnitude of the "true n social rate of return. Griliches calculated the value of the return to hybrid corn R&D. A mail surrey was used deco estimate RED expenditures in both the priorate and the public sectors. Social returns to the innova_~on were estimated by a valuation of observed increased corn production adjusted for price changes. The higher cost of producing hybrid corn was subtracted to derive an annual series of net social benefice . Using wo di fferent interest rates, present (3 955) values of the stream of benefits and research costs were calculated. Griliches concluded that the rate of return to the R&D, as of l9S5 ~ that is, excl uding future benefits ), exceeded 700 percent per year. Pe~c~rson31 applied a generalization of this method to evaluate the returns to poultry- research. Although the approach was similar in pr~r.ciple to that used by Gr~liches, Peterson estimated ache rate of return to state experiment station and private sector research on poultry starting in i9~5 and found that net annual recurs remained negative until the mid-1930's; thereafter, however, p-osi~ive annual net benefits were realized. Using these escalates, he calculated a 17 percent internal rate of return. Perhaps the most sophisticated application of this methodology is ~ at used by Schmitt and Seclcler, 2 who estimated the net rate of return to mechanical tomato harvester research. Because adoption of this new technology displaced labor, Schmitz and Secular based estimates on different rates of compensation to displaced ~ abort Estimates of rates of return of between 8 arid 1, 288 percent were repo reed . Application of this methodology in other countries has resulted in an extensive Literature that has refined the approach, while, at the some time, established its limitations. A general conclusion can be drawn that the methodology is app lied most appropriately to specific products in which demand and supply functions can be estimated with some confidence. A further lesson of repeated application of this supply function shift is the sensitivity of calculated sites of r2 turn deco measurement of research costs. For example, Nelson argues that the critical knowledge facilitating the development of hybrid corn was the result of teas ic genetic research, the cos Sac of which is difficult deco measure in the absence of full knowledge of the path of discovery and development that led to hybrid corn. Certainly, a survey ennui ry such as Chat used by Griliches could be designed to uncover the roles of such basic research, but, in general, such expenses are left - 360 —

unaccounted. Given this omission, estimates of races of return may be b iased upward severely . Conceptually, the approach of valuation of a discrete shift supply function be preen two time periods is ~ v~ria~c~on of a more general approach in which the aggregate flow of innovations is perceived as continuous over time. In this second general approach, if an ir~dicator of that flow could be identified and related to ~ functional description of production technology, ache effects of R&1) on technology or on the production function could be estimated. Furthers if o~oserved economic behavior is. assumed to fol low some rational process of exploit ing technological oppor.~ni~=es deco optimize an obj active, the effect of R&D on economic behavior could be modeled. This type34f approach has been taken in an extensive literature. Griliches pioneered the approach in an application that age' n s stirred the imaginations of both researchers and policy designers . The basic approach focuses on empirical characterization of a production -~c~c~ on, relating inputs to outputs. As noted earlier, this element is essential in understanding the relationship between R&D and production. However, the production- function approach proceeds by simple Eying severely or abstracting from the ocher elements of the frmn~woric outlined above. First, research expenditures are used as an exact measure of the I've of R&D process, abstracting from the essential characteristics of ~ he process. Second, the approach simplifies the lilac between R&D and produc Lion by specify ng R&D expend) Cures as an input to production with either instantaneous effects or, as in later a~oplications, with a time phi chin of effects on production. For example, Evenson, ~ Fishelson, 7 and Cline make various assumptions about initial time 1~ gs in effects of Rho ~ha- produce "V" or ntJ't shaped time pat terns of effete t . The obJ active of the approach is to measure the ~ ncrementa1 output that could be attributable to an incremental change in the R&D input as measured by expenditures. This marginal product of R&D is used along with output price and R&D expenditure per unit output to calculate ~ Dance of return. Griliches used agricultural census data for 1949, 1954, and 1959 to estimate an aggregate U. S. production function for agricultural output. Again, he was careful in interpreting results, but his calculations suggested that the rate of rerun to state experiment station research expenditure was in excess of 300 percent, and possibly as high as 1, 300 percent. If a six-year period is allowed for the R&D effect to take place, Peterson and Hayami4 point out that Griliches' estimates would imply an internal rate of retune' of about 53 percent. Results of other studies adopting this general approach are summarized in Table 3. The approach also has been applied extensively Deco research4~ffect~ r,g international agriculture; see, for example, Kahlon et al. and AD en al. 2 — 361 —

T^~3LE 3 Econometric Studies of Productivi~y Retu~n to Agricu'~ura1 Researc-` in the United S tates Author (Date) Comodity Time Period Rate of Return 6riliches (1964) Aggregate output 1949-5g 35.40 Latimer ( 1964) Aggregate outpu ~ 1949-59 not s ~ gni f i cant Peterson (1967) 45 Poultry .1915-1960 21 Evenson (19~) 46 Aggregate 1949-1959 47 C1ine (1975)47 Aggregate 1939-4B 41-50 Knu tson ~ Tweeten ( 19 79 ) 48 Aggregate 1949-58 39-47 195g-68 32-39 1969-72 28-35 8redahl ~ Peterson (1976) 9 Cash grain 1969 36 Pou 1 try 1969 37 De i ry 196g 43 Li vestock 1969 47 Dav i s ~ lg75l ) 50 Aggregate 1949- IgS9 66- 100 1964- 1974 37 Evenson (1979);1 Aggregate 1868-1926 65 1927-19~ 95 (appl fed RSD) 1927-19~ 110 (basic RLD) 1948-1971 45 (basic RlD) Davis ~ Peterson (1981)52 Aggregate 1949 100 1954 79 1959 66 1964 37 1969 37 1974 37 Norton (1981) 53 Cash grain 1969 31* Pou 1 try 1969 27 Oairy 1969 56 Ll vestock 1969 30 Cash grain 1974 44 Pou ~ try 1974 33 Dairy 1974 66 *~ased on caximum lag 1engtt' esti~ated (9 years). - 36z -

CRITICISMS OF FiIi)~N'C~ OF R&D EN FECTS fine conclusion char ~ s typically dram from evidence such as that presented in Table 3 is -=hat the rate of return to state experiment station research expenditures is quite high. However, before acc~pt~i ng this conclusion, it is useful ( 1) to consider its validity by criticizing the data and methodology upon which the estimates are based and (2) to assess carefully The implications of the results for R&D investment policy. Before proceeding, however, it is worth noting that ache studies reviewed were completed over a long period or time ~ over 20 years ~ during which econometric methods have improved substantially, allowing a series of refinements in rnet;nods and specifications . I ~ is also worth noticing the remarkable robustness of ache order o f magni tude o f the es timates des? i te tines a re f inemen~cs . To proceed in criticizing the data and methodology upon which these estimates are based, the fundamental issue is whether alternatives exist currently that would likely lead to significant changes in tne orders of magnitude of past estimates. Concerning data, two types must be considered. As noted above, estimates of R&D payoff rely upon empirical measures of (1) the R&D process and (2) the production process. Perhaps the most important criticism of past data sets is their reliance upon research expenditures at stare experiment stations. Conceptually' ~ ~ is not clear, and no evidence seems available to support the hypothesis, that expendi.-ures are a useful measure for indicating R&D performance. W~hile no one can deny the evidence of correlation between expenditures and production measures, no evidence has been presented to establish causality. In fact, the i;L=~ition of Gu~tman5 is supported by Huffman and >£iranowslci :: in their evidence that expenditures are endogenous to economic, including agricultural, performance. While economists studying other sectors have explored the potential and appropriate usefulness of other indicators of R&D, little effort has been devoted to this issue in the agricultural sector. Evenson presents an exception in his effort to construct a weighted average stock of knowledge measure based on percents. Weaver57 suggests us ing an approach ~chat would emp lay mul~cip le indicators of the R&D process, but recognizes that each indicator can be interpreted at best as an erroneous measure that is related only stochastically to unobservable measures. In other words, following our meager knowledge of the R&D process, it is proposed that a full set of available indicators of unobservable inputs and of outputs be used to describe the R&D process s structurally. Similar approaches have been useful in modeling relationships among indicators in psychology and sociology. New methods that show promise for identifying the dynamics invo lured among R&O input indicators, output indicators, and product' on decisions are available also. However, in both cases, it is difficult to judge ache impact of these new methods on the estimated orders of magnitude for the rate of return to R&D inves tments . - 363 —

In addit~cn to measures of the R&D process, the second component of the analytical model of R&I) effect outlined above is a model of agrtcul rural production. In this area, the agricul tubal sector has data resources that leave Little room for improvement in extent of product detail and spatial da saggregation. hilt several efforts have been made at updating methods used for aggregating products and prices ~ Nero indexes ~ only small5ghane,es :; index series have resulted; see, for examp le. Lu 9 Brown, and Ball . A critical remaining issue, however, is the effect of output aggrega~cion. lathe use of an aggregate output index is appropriate only if input choices cay be made independent' y of output choice . In earlier work (see 'feaster i), the validity of this restrictive condition was not supported by data from the spring wheat region of North and S ouch Dakota. Alternative specifications exist and deserve exploration. A final issue of some concern regarding data is che choice of geographical area within which far=-level date will ce aggregated, or over which estimated coefficients will be assumed invariant. Several issues concerning specification of the modal of agricu~ tura1 products or production decisions are left unresolved by co oh a priori observation and economic theory. These issues rang. from the selection of commodity groups to be represented by aggregate indexes to the choice of functional form. While an en~era~i of or the 1 ssues can be found e, sewhere (see, for example, Weaver°: ~ 3), theme r net effect is Icy introduce considerable uncertainty into the process of model specification. While, in many cases, resulting estimates may be robust under alternative specify cations, their sensitivity is an empirical issue deserving investigation if resulting estimates are to be granted credibility. Unfortunately, ache estimates in Table 3 were typically reparked in the absence of such specification analysis. A final issue concerning specification of ~ e empire Cal model upon which these estimates are based is their failure to model the diffusion process explicitly. Although descriptive empirical relations are established through the use of ad hoc structures for research expendi~cures, this approach ignores the respons iveness of adop tion to economic incentives and, equally important, to economic constraints. Empirical characterization of the diffusion process seems particularly important in agr: culture, where federal policies have tinkered so s readily with incentives and cons train~cs . - 364 _

REr ERE->iCES 1. -. M. At lid., D. G. Dalrymple, and ~ J. Ruttan. Resource Allocation w-2d Produc~-vity in National and International Agricultural Researc.i. Minneapolis: University of Minnesota Press, 1977. U. L. Peterson and Y. Haymni. "Technical Change in Agriculture. n In A Survey of Agricultural Economics Pi rerarure , Volume I Edited by L. bear ~in. Minneapolis: University of Minnesota Press, 1977. 3. Robert F. Leibenluf~. "Competition in Farm Inputs: An Examination of Four Industries. ~ Policy Planning Issues Paper, Federal Trade Gore' ss, on, Office of Policy PI anning, Februa~- 1981. 4. General Accounting Office. Federal Agr~cul rural Research Funding: Issues and Concerns. October 20, 1983. . Cooperative State Research Service . "Research 1984: The State Agricultural Experiment Stations. n Experiment Station Commit~cee on Organizat' on and Policy, 1984. &. Glenn L. 30~.nson arid Sylvan H. Wittier. "Agricultural Techr.olog- IJntil 2030: Prospects, Priorities and Poll cies . " Special Report No ~ 12, Michigan S tare University Agricultural Experiment S Cation, July 19 84 . 7. 13. F. Hadwiger. The Fo} ~ Tics of Agricul rural Research. Lincoln: Univers ity of Nebraska, 1983 . See also review by J . B . Hendrick, Jr . American Scientist, Vol . 71, No . 5 ( 1982), pp . 53 8 - 53 9 . 8 . J . C . Fitzharrts . n Science for the Farmer: The Development of the Minnesota Agricultural Experiment Station, 1868-1910, Agricultural History, Vol. 48 (January 1971), pp . 202-214. 9. J. (:. Fitzharris. The Development of d~nnesora Agriculture, 1880-1970: A Study of Producriv' ty Change. Department of Agricultural and Applied Economics Staff Paper ~ 74-20. St. Paul: Uni~rersi~cy of Minnesota, September 19 74. 10. W. L. Peterson and J. C. Fitzharris. "Organization and Productivity of the Federal-State Research System in the United States." In Resource Allocation and Productivity in National and Incerna~cional Agricul turn Research. Edited by T. M. Arndt , I) . G . Dalrymple , and V . W. Ruc~an. M' nneapolis: University of Minnesota Press, 1977. 11. T. W. Schultz . The Economic Operation of Agricul tore. New York: McGraw-Hil', l9S3. — 36: —

A. C. True. A History of Agric~,l aura J Fxperimenc~ticr: and Research in the United Scares? :607- 7995, precluding a .;Iis~ory of - The {Jn~-ed Seances De?arement of Agricul sure. USDA Miscellaneous Pubic cation No. 2S1. Washington, DC: L'. S . Government Printing Off' ce, i937. 3 . H. C. Knoblauch et al . State Agrica ~ rural Experiment Stations: History or Research Policy and Procedure. 13S:)A Miscellaneous Publication No. 904. Washington, DC: U. S . Government Printing Office, 1962. 14 E1~o~c Marshall. "USDA Research Under Fire, " Science, Vo] . 217, No. 45~441982), p. 33. 15 . Hill Lepkowsici. "Shakeup Ahead for Agricu1 Mural Research., " Chemical and Engineering News, November 22, 1982 ), pp . ~ -15 . "White }Iouse Plows into Ag Research, ~ Science, Sol. 217 (September 24, 1982), pp. 1227-1228. 17. National Research Council, Division of Biology and Agr~cul~cure. Report or the Commi tree or: Research Advisory to the U . S . Department of Agr~cu ~ Cure. ~'ashington, DC: National Academy of Scat ences, ~ 972. . Fit~harris, 1971,. op. cite. lD . Fi ~zharris, 19 74, op . city . 20 . Peterson and Fitzharris, op . cf. . 2t . Coope fire Stance Research Service, op . cit . 22. General Accounting Offices op. cit. 23 U. S. Department of Agriculture and Association of State Universities and Land Grant Cold eyes. tJaciona1 Program cuff Research for Agriculture. Washington, DC: U.S. Government Printing Off ice, 19 6 6 . 24. Johnson and W~:twer, op. cit. 2S. R. D. Heaver. "multiple Input -Multiple Output Production Choices and Technology in the U.S. Wheat Region," American Journal of Agricul rural Economics, Vol . 65 (February 1983 ), pp . 45 - 56 . 26. R. I). Weaver. "The Theory, and Measurement of Provisional Agricultural Production Decisions. ~ Ph.D. Thesis, IJniversity of Wisconsin, 1977. 27. Weaver, 1983, op. cit. — 366 -

28. G. W. Norton and J. Davis. "Review of Methods Used to Evaluate e Returns Deco Agricultural Research. n Staff Paper P79-15, University of Minnesota, Department of Agricultural Economics, May 1979. 29. Schultz, op. cap. 30. Z. Griliches. Research Costs and Social Us: Hybrid Corn and Related I~ons," Journal of PoliCical Economy, Vol. 66 (October 1958 ), pp . 419 -431. 31. W. L. Peterson. "Return ~o Foul~c~ Research in the United States, n Journal of barn ~corlomics, ~Jot. 49(Aug,us~ 1967), pp. 656 - 669 . 32. A. Schmitz and G. Secicler. "Mechanical Aged ~uiture and Social Welfare: Me Case of the Tomato Harvester, " American Journal of Agricul rural Economics, Vo! . S2, No . 4 (November 1970 ), pp . S 69 - 578 . 33. R. Nelson. "The Simple Economics of Basic Scientific Research, " Journal of Political Scions, Vol. 67~1959), pp. 297-306. 34. Z. Griliches. "Research Expenditures, Education, and the Aggregate Agricultural Production Function, ~ American Economic Review, Sol. 54(I:ee.mber 1964), pp. 967-974. 35. R. E. Evenson. nThS Contribution of Agricultural Research to Production, n Journal of Fat Economics, Sol. 49~Derember 1967), pp. 141S-142S. 3 6 . R. E . Evenson . "The Contribution of Agric~ul~cural Research arid Extension to Agricultural Production. n Ur~olished Ph.D. 13'ssertation, University of Chicago, 1968. 37. G. Fishelsoh. Ale turns to H',man and Research Capital, United Stances Agriculture, 1949-1964. ~ Unpolished Ph.I). Dissertation, North Carolina State University, 1968. 38. P. L. Cline. "Sources of Productivity Changes in United Stances Agriculture. n Unpublished PheI)e Dissertation, Oklahoma State University, 1975. ~9 . Griliches, 1964, op . cit. 40. Peterson and Haying, op . cit. 41. A. S. Kahlon., M. K. Bal, P. N. Saxena, and D. Sha. ~Ren~rns Deco In~rest:ment in Research in India ~ In Resource Al location and Productiv:~cy in National and International Agricul rural Research. Evinced by T. M. Ar`~dt, D. G. Dalr~ple, and V. W. Ruttan. Minneapolis: University of Minnesota Press, 1977. — 367 —

42. Arndt et al., op. Cat. 43. Griliches, 1964, op. cit. 44. R. Latimer. ~ Some Economic Aspects of Agricultural Research and Education in the U. S . n Unpublished Ph. O. Dissertation, Purdue Univers i by, 19 64 . . Peterson, 1967, op. cit. 46. Evens on, 1968, op. cit. 47. Cline, op. cit. 48 . M. Knutson and L. Twee ten. "Toward an Optimal Rate of Growth in Agricultural Production Research and Extension, ~ America Journal of Ag~icul Cural E:cor2omics, Job . 61, No . 1~19 79 ), pp . 70- 76 . 49 . M. Bredahl and U. Peterson. "The Producti~ri~ y and Allocation of Research: TJ.~. Agricultural Experiment S~cations, n American Journal of Agric,~1 c,ural E~or~omics, Sol 58, No . 4, Part 1 (November 1976), pp. 684-692. 50. J. Davis. "A Comparison of Alt~rr.ative Procedures for Calculating the Rate o f Return to Agricultu. al Research Us ing ache Produc Lion Func t i on App roach . ~ Mimeo, Unive rs i ty 0 f Minnes o ta, Department of Agrict~lCur~l and Applied Economics, March 1979. 51. R. Evenson, P. E. Waggoner, and V. Ruttan. "Economic Benefits from Research: An Example from Agriculture, " Science, Sol. 20541 979), pp. 1101-1107. :2. J. Debris and W. L. Peterson. "Stability of ache Research Coefficient for U. S. Agriculture. ~ Presented at Symposium on Me thodolo gy for Agricultural Research Evaluation, Minneapolis, En, May 1980. 53. G. W. Norton. "The Productivity and Allocation of Research: U. S. Agricultural Experiment Stations, Revisited, n North Central Jou~.a] of Agricultural Economics, Vol. 3, No. 1~1981), pp . 1- 12 . 54 . Joel M. Gunman n Interes t Groups and ache Demand for Agricultural Research, n Journal of Politick Economy, Vol. 86 (1978 ), pp . 467 -484. . E . Huffman and J . A. Miranowslci . "An Economic Analys is of Expenditures on Agricultural Experiment Station Research, American Journal of AgricoiCural E:conom~cs, Vol. 63 (1981), pp . 104- 118 . — 368 —

56. R. I. Emerson. PA Century of Agricultural Research and Productivity Changes, Research, Invention, Extens ion, arid Productivity Mange in U.S. Agriculture: An Historical Decomposition Analysis. n Paper prepared for the Symposium on Agricultural Research and Ext ens ion Evaluation, Moscow, Idaho, May 21- 23, 1978 . R. D . Weaver . n Private Sector R&D and Agricultural Productivity: A New Approach for Measurement. n Staff Paper No. 49, Pennsylvania State Uni~rersi~cy, Department of Agricultural Economics and Rural Sociology, 1982. Presented at the Symposia on Priorate Roles in Fos Bering Agricultural Productivity, Logan, IJ:ah. 58 . Y~o-Chi Lu. "Measuring Products viny Change Southern Journal of Agricu1 Cural Economics, PP. 69-75. in U . S . Agr.cu7 ~cure, " (December 1975 ), 59. Randall S. Brown. "Produc~civity, Retuners, and the Structure of Productivity in U.S. Agriculture: 1947-1974.~ Ph.I:). Dissertation, University of Wisconsin--Madison, Repayment of Economics, 1978. 60 . Eldon V. Ball. "Measure. ng Agricultural Productivity: A New Look. n USI)AiERS/NES Staff R2part, 1984. 61. Weaver, 1977, op. cit. 62. Wearer, 1982, op. cite. 63. Weaver, 1983, op. cit. 369 -

ADDITIONAL BIBLIOGRAPHY H . ~ . Ayer and G . E . Schuh . " Social Rates of Return arid Other Aspects of Agricultural Research: One Case of Cotton Research in Sac Paula, Brazil, ~ Amer, can Journal of Agricul rural Economics, Vol. 54, No. 4, Par. I(November 1972), pp. 55?-569. M. E. Bredahl.. "lathe Productl~rity and Allocation of Research at U. S Agricultural Experiment Stations . n Unpublished Ph. D . Dissertation, University of Minnesota, 1975. I. L. Cline and Y. C. Lu. "Efficiency Aspects of the Spatial Allocation of Public Sector Aged cultural Research and Extension in the United States, n Regional Science Perspectives, Vol. 5 (1976), pp . 1-16 . K. W. Easter and G. Norton. "Potential Returns from Increased Research Budget for the Land Grant Universities, ~ Agricul tu-al Economics Research, Sol . 29, No. October 1977), pp. 127-3 33 . - L. Fishel. "l~e Minnesota Agri~ul~cural Research Resource Allocation Information System and Experiment. n In Resource Allocation in Apical rural 2esearc.5 . Edited by W. L. r ishel . Minneapolis: University of Minnesota Press, 1971. W. L. Fishel. R-sor~-ce Allocation in Agricul rural Resear-n. Minneapolis: IJniversi~cy of Minnesota Press, 197.1. Z. &riliches . Sources of Heasured Producti~ri~cy Growth: I]. S . Agricu' tare 1940-1960, ~ Journal of Political Economy, (1963) . F. (I. Jarring and R. K. Lir~dner. "Research Benefits Revise ted, n Rev' ew of Marke~ci~g and Agricul rural Economics, iIo1. 45, No . 4 (December 1977 ~ . R. K. Lindner and F . G . Jarrett ~ ~ Supply Shifts and the S ize of Research gene fits ~ ~ American Journal of Agric~l rural Economics, Vol . 60, No . l (February 1978 ), pp . 48 - 64 . ':. Lu, Is. Q~e, and C . L. Liu. n fro; ecting Agricultural Productivity and Its Economic Impact, ~ American Journal of Agricu] rural Economics, Sol . 60, No . S (December 1978 ), pp . 976 -980. W . Norton and J . [Davis . Devaluating Re turns to Agricultural Research: A Review, n American Journal of Agricultural Economics, Vol. 63, No. 4~1981~. W. L. Peterson. "Ran turns to Poultry Research in th" United States Unpublished Ph. I). Dissertation, University of Minnesota, 1966. _ 310 - ~1