Research Costs and Social Returns: Hybrid Corn and Related Innovations Author(s): Zvi Griliches Source: The

Journal of Political Economy, Vol. 66, No. 5 (Oct., 1958), pp. 419-431 Published by: The University of Chicago Press Stable URL: http://www.jstor.org/stable/1826669 Accessed: 23/11/2010 12:48
Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use. Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at http://www.jstor.org/action/showPublisher?publisherCode=ucpress. Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission. JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

The University of Chicago Press is collaborating with JSTOR to digitize, preserve and extend access to The Journal of Political Economy.

http://www.jstor.org

RESEARCH COSTS AND SOCIAL RETURNS: HYBRID CORN AND ]RELATED INNOVATIONS1
ZVI GRILICHES of University Chicago
INTRODUCTION AND SUMMARY

tures on "research and development" have grown very rapidly in the last decade. Quantitatively,however, we know very little about the resultsoftheseinvestments. have some We idea of how much we have spent but very little of what we got in return. We know almost nothingabout the realized rate of returnon these investments, that it must though we feel intuitively have been quite high. This article presents a first step towardansweringsome of these questions. However, all that is attemptedhere is to estimatethe realized social rate of return,as of 1955, on public and private funds invested research,one of the outin hybrid-corn standing technologicalsuccesses of the century.The calculated rate of return is an estimate, subject to a wide margin of error,but it should provide us with an order of magnitude for the "true" social rate of returnon expenditureson hybrid-corn research.Actually,I believe that my estimate is biased downward1 for, whenever I had to choose among alternativeassumptions,I chose the assumptionthat led to the lowestestimate.
I This article is an outgrowth a largerstudy of of the economics of hybrid corn. See my article, "Hybrid Corn: An Exploration in the Economics October, of Technological Change," Econometrica, 1957. I am indebted to A. C. Harberger,Martin J. Bailey, Lester G. Telser, and T. W. Schultz and to the National Science forvaluable comments Foundationand the Social ScienceResearchCouncil support. forfinancial

and publicexpendiBOTH private

This estimate witlnot tell us the global rate of returnon researchexpenditures, but even a modest step in that direction may be of some use. Tihe followingprocedure is used to arriveat the estimate: First,private and public research expenditureson hybrid corn,1910-55, are estimatedon the basis of a mail survey and other data. Then the annual gross social returnsare estimated on the assumption that they are approximately equal to the value of the resulting increasein cornproduction plus a price-change adjustment. The additional cost of producinghybrid seed is subtracted from these gross returns to arrive at an annual flow of net social returns. Using first a 5 and then a 10 per cent rate of interest, I bring all costs and returnsforward to 1955, when the books are closed on this developmentand a rate ofreturn computed. is Research costs are expressedas a capital sum, and returnsare converted into a perpetual flow.The estimatedperpetual flowof returnsis divided by the cumulated researchexpendituresto arrive at a rate of return that will equalize the presentvalue of the flowof returns with the cumulatedvalue ofresearchexpenditures. This procedure leads to the estimate that at least 700 per cent per year was being earned, as of 1955, on the average dollar invested in hybrid-corn research. Since thisis not the onlyway in which a rate of returncould be computedfrom these data, sonie alternative ways of

419

420

ZVI GRILICHES

definingand estimating the social rate of returnare exploredbriefly. Comparisons are also made with estimates of in returns some otherareas oftechnological change. Finally, I discuss the limitations of the procedureused and the implications of the results. In particular, I shall emphasize that almost no normative conclusionscan be drawnfromthese fewestimates.
RESEARCH EXPENDITURES

Inbred lines and hybrids have been developed by state agricultural experimentstations,the United States Department of Agriculture(USDA), and private seed companies.The distinction between the firsttwo developing agencies is mainly in the source of funds.Except for funds spent on research and coactivitiesat Beltsville,Maryordinating land, most of the USDA funds were spent on co-operativecorn-breeding research at various experiment stations. A mail inquiry to ascertain expenditures on hybrid-corn research was sent to all the agricultural experimentstations, and usable data were obtained fromtwentyof them.The twentystates represented theserepliesincludemost by of the importantcornstates in the country.Expendituresof non-responding stations were estimated by setting the expendituresof each of then equal to the of expenditures a "similar" station.2 Data on USDA expenditures "cornon productionresearch:agronomicphases" with 1931 were obtained from beginning the AgriculturalResearch Service and extrapolated back to 1910. They over-

estimate the USDA contributionsubstantially,because they include various other aspects of corn research besides hybrid corn. Moreover, some of the USDA fundshave already been counted in the expenditures of agriculturalexperimentstations. The research expenditures of one of the major private seed companies for the years 0925-55 were extrapolated back to 1911 and divided by that firm's estimated share of the total market for hybridseed cornto arriveat an estimate of the researchexpendituresof the "private" segmentof the industry.' The figures 1955 may be used as an for example of the numbersinvolved.I estimate that in 1955 the USDA spentabout $300,000 on hybrid-cornresearch, the experiment stationsabout $650,000, and the private companiesabout $1,900,000.4 The historical research expenditure data, deflated by the Consumers Price Index (1955 = 100), are reproduced in column 1 of Table 1. In view of all the assumptions made, these figuresshould be taken with several grains of salt, the dosage increasingas one goes back into the past. In particular, for the years 1910-25, the figures littlemore than are guesses. For the purpose of estimating rate the

3 This will again overestimate expenditures, because "public" hybrids make up 25-30 per cent of the total market,and researchexpenditures on these have already been countedonce. 4In 1951 M.T. Jenkins, the USDA, estimated of the total annual expenditureson hybrid-corn. researchas follows:USDA, $220,000;states,$600,000; and private industry, $1,100,000. M\,y own independentestimate for 1951 is: USDA, $190,000; states, $550,000; private industry, $1,300,000. The two totalsare $1,920,000and $2,040,000,respective2 The pairing was made on the basis ofgeographic ly. The agreementis very close, consideringhoew about the in- arbitrary some of my assumptions are (see M. proximityand general information For example,Indiana expenditures wereas- T. Jenkins, "Corn Breeding Research-Whither dustry. sumed to equal the reportedIllinois expenditures; Bound," Proceedingsof the Sixth Annual Hybrid Oklahoma'sto equal Kentucky's, and so forth. These CornIndustry-Research Conference [Chicago: Amerpairings probably overestimatethe total expendi- ican Seed Trade Association,November,1951],Pp. tureson hybrid-corn research. 42-45).

RESEARCH COSTS AND SOCIAL RETURNS

421

ofreturn theseexpenditures, assume on I that the public sector will continue to invest in hybrid-cornresearch at an annual rate of$1 million, and the private sector at an annual rate of $2 million. No incrementalreturns,however, will be ascribed to these expenditures.I assume them to be "maintenance" exin penditures face ofa malevolentnature.
COST OF ADDITIONAL TO PRODUCTION RESOURCES OF HYBRID DEVOTED SEED

from the subsequent estimate of gross returnsto arrive at a net returnfigure.7

THE VALUE OF HYBRID CORN TO SOCIETY

7 Throughoutthe period these computations use the average corn acreage planted in 1939-48; they disregardannual fluctuations total corn acreage in and seeding rates. For any year before 1956, the extra cost of seed equals the percentage planted with hybridseed times $98 million (90 millionX [7.5/64] X $9.50 - $2 million [research expenditures]= $98 million). 5 This is somewhathigherthan the marketprice 8 For example: "Plant breeders conservatively of corn because of the betterquality of seed corn estimate increase in yields of 15 to 20 per cent and the labor that would go into its selection. fromusinghybridseed underfieldconditions.They Since open-pollinated seed is now quoted at about expect about the same relative increases in both $3.00 to $4.00 a bushel, this assumptionalso con- low-and high-yielding areas" (USDA, Technology tributesto an over-alloverestimate cost. of on theFarm [Washington, 1940],p. 7). 6 This result is reached as follows: 80 million 9 This figure based on a USDA demand analis acres X 90 per cent in hybridsX (8.6/64) average ysis (see R. J. Foote, J. W. Klein, and M. Clough, seedingrate X $9.50 = $92 million.Subtracting $2 The Demand and Price Structure Corn and Total for millionresearchexpenditures, get $90 million Feed Concentrates we [Technical Bull. 1061 (Washing(source: Agricultural Statistics, 1955). ton: USDA, October,1952)]).

I assume that the price of hybrid seed, approximately$11 per bushel in 1955, measures adequately the value of resources devoted to its production. If there were no hybrid corn, farmers would use mainly home-producedopenpollinated seed, which I value at $1.50 per bushel.5The quantity of hybridseed used annually was estimated by multiplyingthe reportedcornacreage planted withhybridseed by the average seeding rate of corn. Multiplyingthe result by between the price $9.50, the difference of hybridand non-hybrid seed, and subresearchexpenditures, tracting million $2 I get $90 millionas my estimate of the additional resources currentlydevoted each year to hybrid-seed production.6 Using the average 1939-48 corn acreage (90 million), the 1951 seeding rate (7.5 pounds per acre), and the percentage planted with hybrid seed, I computed the additional cost of hybrid seed for the years 1933-55 and subtracted this

As everyone knows, hybrid corn increasedcornyields.The figure mostoften quoted for this increase is 20 per cent. For my purpose, I assume that the superiorityof hybridover open-pollinated varieties is 15 per cent, the lower figure in most estimated ranges.8 The value ofthisincreaseto "society" will be measured by the loss in total corn production that would have resulted if there were no hybrid corn. This hypotheticalloss will be valued at the estimatedequilibrium price of corn plus a price-change adjustment, a procedure equivalent to computing the loss in "consumer surplus" that would occur if hybrid corn were to "disappear." The amount of this loss will depend on our assumptions about the relevant demand-and-supply elasticities. As will be seen fromthe formulaspresentedbelow, theseelasticities have onlya secondorder effect, and hence different reasonable assumptionsabout them will affect the results very little. I assumed that the price elasticity of the demand for corn is approximately -0.5.9 Since we

422

ZVI GRILICHES In this case, the loss is measured by the area Q2P'P'Q1 in Figure 2. Instead of assuming that the supply curve shifts upward, we now assume that it shifts k per cent to the left. The rectangle Q2PlP1 Ql measures the loss in corn production at the old price P1. The triangle P1P2P1 can be viewed as the additional loss in consumersurplusor as an adjustment for the increase in price fromP1 to P2. The total loss is now given by the formula Loss 2 = kP1Q1(1+
/

know much less about the supply elasticity of corn, I shall firstexplore the extreme consequences of two different assumptionsabout it. that in the long Let us assume, first, elastic; runthesupplyofcornis infinitely that is, we face long-runconstant costs. The "disappearance" of hybrid corn would shift the supply curve upward by the percentagereductionin the yield
U

V~~~~~~~~
P27

kn).

It is easily seen that the second assumption leads to a higher estimate of S pi the loss. It can be also shown that the two estimates bracketestimates implied D by assuming other intermediatesupply elasticities.The ratio of the loss under QUANTITY assumption2 to the loss under assumpQ2 Q1 tion 1 is (1 + Ikn)/(1 - 'kn). In our FIG. 1 case, this ratio is approximately1.07.10 between these two exof corn. The "loss" to society, in this The difference treme assumptionsimplies only a 7 per case, is the total area under the demand curve betweenthe new and the old supply curves. In Figure 1 this area is P1P2P2P1. This area can be interpreted Cr as the increase in the total cost of producingthe quantityQ2 in the new situa- p22 tion, the rectangle PiP2P2P1, plus the loss in consumersurplus caused by the rise in price, the triangle P1P2P1 A of linearapproximation thisarea is given D by the formula
/

m1<~~~~~~~~~~~~1
~~~~P

LU

Loss 1 = kP1Q1( 1-2 kn), wherek is the percentagechangein yield (marginalcost and average cost), P1 and Q, are, respectively,the previous equilibrium price of corn and quantity of corn produced, and n is the absolute value of the price elasticity of the demand forcorn. it Alternatively, could be assumed that the elasticityof the supply curve is zero.

Q2 FIG. 2

Q1

QUANTITY

in cent difference the final estimate of the total loss. Because this difference is so small and because I am striving fora lower-limit estimate,I have chosen
10Assumingk = 13, i.e., 15/115, and n = 0.5, the ratiois (1 + 0.5 X 0.13 X 0.5)/(1 - 0.5 X 0.13 X 0.5) = 1.07.

RESEARCH COSTS AND SOCIAL RETURNS

423

the first assumption-that of an in- I assume that the average 1943-52 level elastic long-runsupply of corn. ofproduction-approximately3,000 milfinitely To calculate the loss, we must assume lion bushels annually-will continueand an equilibriumprice of corn. I shall use that the percentageplanted with hybrid $1.00 per bushel in 1955 dollars as a seed will stabilize at 90. Both these asminimal estimate of the value of corn sumptionsare conservativeand will reof to society. The currentprice of corn sult in an underestimate returns. Assuming that k, the relative shift by about $1.25, is affected the esixtence programsand probably in the supply curve, is 0.13 (15/ 15); of price-support the social value of corn." that PQ is $3,000 million; that n, the overestimates is Because not all corn acreage was or demand elasticity, 0.5; and that h, the is planted with hybrids,I multiplythe currentand futurefractionof all corn percentageshiftk by It, the percentage acres planted withhybridseed, is 0.9, we of all corn acres planted with hybrids can calculate the currentand expected (loss = hkPQ[1 ->-hkn]). This proce- annual flow of gross social returns dure disregardsthe fact that the acres from hybrid corn as follows: 0.9 X first planted to hybrids were higher- 0.13 X $3,000 million (1 - - X 0.9 X yieldingacres than those planted later, 0.13 X 0.5) = 0.117 X $3,000 million and hence the procedureunderestimates (1 - 0.029) = 0.117 X 0.971 X $3,000 million= $341 million. Subtractingthe total returns. from hybrid projected annual cost of hybrid-seed past returns In estimating corn, I ignore annual fluctuations in productionand research $93 millionprices and production,basing my com- we get $248 million as the currentand putations on the average 1937-48 level expected annual flow of net social reof productionof 2,900 million bushels'2 turns. and a real price of corn of 4t.00 per Similarly,returnsfor the past years, bushel in 1955 dollars. On the returns beginningwith 1933, are calculated by side, only the percentage planted with multiplying2,900 million 1955 dollars, hybrid seed varies over tine. To cal- the average total value of corn producculate the annual flowof futurereturns, tion, by the percentage of total corn 11 An approximateformulafor determining the acreage planted withhybridseed in each price of corn in the absence of price supports is year. Subtracting the estimated past givenby Marc Nerlovein "Estimates of the Elasticcosts of hybrid-seedproduction,we get Commodiities of Supply of Selected Agricultural forthe years 1933XXXVIII (May, the net social returns ties," JournalofFarm Economics, 1956), 497: dp/po= (dq/q)/(n+ e), where po is 55 shown in column 2 of Table t.13
price,n and e are the demand-andthe equilibrium supply elasticities,and dq is the quantity placed under loan. In recent years about 7 per cent of the annual corn crop, on the average, has been placed underloan withthe CommodityCreditCorn poration. The assumptions = 0.5, e = 0.2, imply price is about 10 per cent above that the current the equilibriumprice. The currentprice is about $1.25 per bushel, which implies an "equilibrium" price of corn of about $1.t13.But this estimate does not take into account the impact on corn prices of the eliminationof price supports on all other agriculturalcommodities.Taking this into account,the equilibriumprice would be closer to $1.00 per bushel. In any case, it is unlikely to be lowerthan $1.00.
12 This assumption was made to simplifythe part of the period,produccalculations.In the first and, sinceI use a relativetionwas below thisfigure, ly high rate of interest,this will resultin an overplantedwith But the percentage estimateof returns. was also low then,whileit was muchhigher hybrids during the period when production exceeded its average, and this will result in an underestimate of returns.On balance, the second effectshould by outweighthe first a fairmargin. 1' These are calculated from the following formula: h X [0.13 X $2,900 million (1 - 1/2 X 0.9 X 0.13 X 0.5) - 90 millionX (7.5/64) X $9.50 $2 million]= h X $268 million.This procedure is since h should also have entered an approximation,

424
TABLE 1

ZVI GRILICHES
CALCULATION OF A RATE OF RETURN

HYBRID CORN: ESTIMATED RESEARCH EXPENDITURES AND NET SOCIAL RETURNS,

1910-55

(Millionsof 1955 Dollars)

Total Research Expenditures (Private and Public) 0.008 Net Social Returns*

Year 1910.......

1911 1912 1913 . 1914 1915 . 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927.. 1928 1929 1930 1931 .0.395 1932. 1933.0.584 1934 1935.0.593 1936. 193 7 1938 1939 1940 1941 1942 1943 1944 1945. 1946. 1947.1.660 1948.1.660 1949.1.840 1950 1951 .2.110 1952. 1953 1954. 1955. Annuallyafter 1955

0.011 0.010 0.016 0.022 0.032 0.039 0.039 0.039 0.044 0.052 0.068 0.092 0.105 0.124 0.139 0.149 0.185 0.210 0.285 0.325 0.495 0.564

0.661 0.664 0.721 0.846 1.090 1.100 1.070 1.390 1.590 1.600 1.820

2.060 2.180 2.030 2.270 2.790 3.000

0.3 1.1 2.9 8.3 21.2 39.9 60.3 81.7 105.3 124.3 140.4 158.7 172.6 184.7 194.3 203.7 209.8 209.0 218.7 226.7 232.1 234.2 239.1 248.0

research expenditures. social returns zero Net are before 1933.

* Net of seed production cost but not net of

into the secondpart of the first termof the formula, the "triangle." I neglect this. Because It is less than 1 and because that part is always subtracted, total returns. this procedureagain underestimates

Table 1 presents estimates of costs and returns.There are several ways in could be summarized whichthesefigures and a rate of returncalculated. My procedure is as follows:Consider the development closed as of 1955. Future expenditureswill not increase returns,nor will there be an expansion of hybridcorn acreage. Standing in 1955, I cumulate and bring forwardto 1955 all past at costs and returns a reasonableexternal rate of interest.To explore the impact rates of interest, of two quite different I performthe calculations twice, using first a 5 per cent and then a 10 per Past researchcosts cent rate of interest. are cumulatedand expressedas a capital are sum. Past returns cumulatedto 1955, and a 5 or 10 per cent rate of return on these cumulated returnsis projected into the future. The estimated flow of futurenet returnsis added to the flow to from past returns arriveat a perpetual flow of net social returns from hybrid dividedby thecumulated corn.This flow, researchexpenditures, gives us our estimate of the realized perpetual rate of return. Table 2 presentsthe calculations that lead to my estimate of approximately $7.00 as the annual returnin perpetuity, as of 1955, for every dollar that has been invested in hybrid-corn research. Actually, even if we ignore all past returns completely,the figureis still very high-approximately$4.00 annually (using the 5 per cent interest rate) for everyresearchdollar. This way of calculating a "rate of return" is not really different from a ratio calculation. It may be benefit-cost useful to bring out explicitlythe relabetweenthesetwo concepts.The tionship preceding rate of return is defined as

RESEARCH COSTS AND SOCIAL RETURNS

425

100 (PR X k + AFR)/RC, follows:r where PR= cumulated past returns,k = the external rate of interestused to cumulate or discount returns,AFR = annual futurereturns, and RC = cumulated researchcosts. A benefit-cost ratio fromthese same data would be BIC = (PR +AFR k)/ RC. Hence r= 100 k (BIC), and we can translateour calculation into a benefitcost ratio, and vice versa. Using 5 and 10 per cent as the external rates of interest, benefit-cost the ratios forhybridcorn researchexpendituresare 150 and 70, respectively.14 When we recall that most Bureau of Reclamation watershed projects have ex post benefit-cost ratios of 1 or less, this does imply a certain misallocationof public resources.15 These calculationsuse an externalrate of interest to bring all sumns forward to 1955. It is reasonable to assume that the marginal productivityof capital in alternativeinvestments between5 and is 10 per cent and to use these rates as conversion factors for funds expended or earned at different dates. Alternatively, however,one could calculate an internal rate ofreturn-that rate ofinterest which will equate the flow of costs and the flow of returns over time.'6 Such a rate has to be calculated using an iterative procedure, changing the rate used until the cumulated costs are equal
14If beforecumulatingwe were to subtractthe researchcosts from net returnsannually (that is, have in our denominator onlyresearchexpenditures before 1934, the year when net returnsbegan to exceed researchcosts), the benefit-cost ratios would be substantiallyhigher (about 700 and 200, respectively),and so would also the rate of return as defined the text. in
15 For an evaluation of public investmentsinl watershedprojects see E. F. Renshaw, TowurdResponsibleGovernment (Chicago: Idyia Press, 1957).

to the discountedreturns.The two procedures will give different answerswhen the time shape of costs differs markedly from the time shape of returns,as in our case. The internal rate of return on hybrid-corn research expendituresis between 35 and 40 per cent."7My objection to this particular procedure is that it values a dollar spent in 1910 at $2,300 in 1933. This does not seem very sensible to me. I preferto value the 1910 dollar at a reasonable rate of returnon some alternativesocial investTABLE 2 RATE OF RETURN ON HYBRID-CORN RESEARCH EXPENDITURES AS OF 1955

(Millions of Dollars) (1) Net cumulated past returns .................. 4,405 (2) Past returnsexpressedas an annual flow 220 .......... (3) Annual future gross returns 341 .................. (4) Annual additional cost of and research.. 93 production (5) Total net annual returns, (2)+(3)-(4) ............ 468 (6) Cumulated past research expenditures ............ 63
(7) Rate of return 100 X (5)/(6) ................. 3
r=0.05 r=0.10

6,542 654 341 93 902 131

689

ment. Also, this procedure gives tremendous weight to the early expenditures, which are subject to the largest error of measurement. Actually, however, the two estimates are not very far apart. The estimateusingan external rate ofinterest says that a dollarinvested in hybrid-corn research earned 10 cents annually until 1955 and $7.00 annually thereafter.The internal rate estimate says that the dollar earned 40 cents annually throughoutthe whole period. If the average delay betweeninvestment and fruitionis about ten years, then

16I am indebtedto Martin J. Bailey forsuggest17 That is, 40 per cent is too high and 35 is ing this alternative way of calculating the rate too low. The iterativeprocedure was not carried of return. further.

426

ZVI GRILICHES periorityof hybridsare both conservative and probably result in an underestimate of the real returnsto society. I have also assumed that all past research has already borne all its fruitand that all future researchon hybridswill result in no benefitwhatsoever.Nor has credit been given for the impact of hybrid corn on otherfields:the researchon hybrid poultryand hybridsorghumwhich it stimulatedor the reductionof farmer resistanceto new technology a result as of the spectacularsuccess ofhybridcorn. tHence, faras costsand returns as from hybrid corn per se are concerned, the estimate is too low. One troublesome problem,however,remains to haunt us. Does it really make sense to calculate the rate of return on a successful "oil well"? What is the point of calculating on the rate ofreturn one of the outstanding technologicalsuccesses of the century? Obviously,it will be high. What we would like to have is an estimate that would also include the cost of all the "dry holes" that were drilledbefore hybridcorn was struck. The estimate does include the cost of all the "dry holes" in hybrid-corn research itself. Hybrid corn was not a unique invention--it wias an invention of a methodof inventing. Many different combinationswere triedbeforethe right ones were found. One major seed company annually tests approximatelyfifteen hundred different combinationsof inbred lines. Of these, at most three or fourprove to be successful.The cost of the unsuccessfulexperimentsis included in my estimate.What is excluded is investmentin various other areas of agricultural researchwhichhas not borne fruit. The problem of dry holes, however, can be reduced ad abszrdmr. What is the relevantsegmentforwhichan aggre-

the two figuresrepresentdifferent ways of saying the same thing.'8

LIMITATIONS

The estimateof 700 per cent is probably too low. At almost every point at which therewas a choice of assumptions to be made, I have purposelychosen those that would result in a lower estimate. This is an attemptto arrive at an estimatedlower limit of the social rate of returnfromhybridcorn. Both the public and the private research expendituresare probably overIn statedsubstantially. fact,theexpenditure estimatessupplied by some experiment stations are obviously too high. perhaps less Of the USDA expenditures, than half were devoted to hybrids. I before did leave out all expensesincurred 1910, but these in total could not have been more than a few thousand dollars. This should remind us, however, that the estimated rate of return is mainly a rate of returnon applied rather than basic research.The basic idea of hybrid corn was developed between 1905 and 1920, with the help of very littlemoney. The rate of returnon thisbasic research, if it could be calculated, would be much higher. However, the idea had to be translated to commercial reality, and separate adaptable hybrids had to be areas. These are developed fordifferent the activities reflectedin my estimate of researchcost. The returns,on the other hand, are understated.The assumed price of corn of $1.00 per bushel in 1955 dollars and the assumption of a 15 per cent su18 An annual flowof 40 cents discountedat 40 per cent has a presentvalue of $1.00. An annual flowof $7.00, discountedat 40 per cent, will also have a presentvalue of $1.00 if there is a lag of ten approximately yearsbetweenthe date of investment and the date at which the perpetual flowof returns be-ins.

RESEARCH COSTS AND SOCIAL RETURNS

427

gate rate of returnis to be computed? Is it really reasonable to ascribe the cost of unsuccessfulgold explorationto the oil industry?If one takes this kind of reasoningtoo seriously,there is only one rate that has any meaning-the rate of returnon research for the economy as a whole. Nevertheless, the rate of return on a successfulinnovationmay be of some interest.In particular, may be useful, it ex ante, to break down the probable rate of returninto two components:the rate of returnif the developmentturns out to be a success and the probability that it will be a success. The approach outlined here is a way of estimating the firstcomponent.An estimate of the probability of success, however. must be made on the basis of data otherthan those presentedin this article.
RETURNS IN SOME OTHER AREAS

receipts fromfarm marketingsin 1951, minusapproximately per cent to allow 10 forthe impact of the supportprograms), and assuming a price elasticity of demand for agriculturalproducts of -.25 and an infinite supply elasticity,we get k (1-Akn) X$30 billion= 0.037 X (1 - 0.5 XO.037 XO.25) X$30 billion= 0.037 X0.995 X$30 billion= $1,110 million. Using the upper limit estimate of 18.5 per cent saving in inputs,we get 0.185 (1- 0.5 X0.185 X0.25) = X $ 3 0 billion $ 5,4 3 0 million as an upper-limitestimate of the gross annual social returnfrom the technical change that occurredbetween 1940 and 1950. Schultz also provides an estimate of total public expenditures on agricultural research for the years 1937-51. I assume that all theseexpenditures were used to produce the increase in output in 1951. This leaves out the contributions developed fromfundsspent before1937, but, on the other hand, it disregards the possible returns after 1951 from the 1937-51 expenditures.On balance, we will probably overestimatethe funds spent on the 1940-50 improvementin technology. I will assume that total private agricultural research expenditureswere of about the same magnitudeas the public expenditures.This is approximately half the corresponding ratio for hybrid corn but is probablyan overestimate agrifor cultureas a whole.20
20 Mighell has estimatedthat recentannual expendituresby industry researchon agricultural for products and on machineryand materialsused in agriculturewere in excess of $140 million.At the same time the USDA and state agricultural experi-

T. WV.Schultz has provided us with estimatesofcosts and returns research of forUnited States agriculture a whole.19 as His data can be used to estimate the rate of return on agriculturalresearch as a whole. Schultz gives an upper-and a lower-limit estimateof how much more input it would have taken to produce the 1950 output with 1940 techniques and inputs. His upper estimate is that it would have taken 18.5 per cent more input; his lowerestimateis 3.7 per cent. Let us use Schultz's figuresto estimate the perpetual gross annual returns,beginningwith 1951, fromthe agricultural research that bore fruit between 1940 and 1950.Using thelowerestimate,there would be a loss of 3.7 per cent in output if the new technology were to disappear. Taking the total annual value of farm output as $30 billion ($32 billion cash
19 TheEconomic Organization Agricullure of (New York: McGraw-HillBook Co., 1953), pp. 114-22.

428

ZVI GRILICHES

Multiplying the 1937-51 public expendituresby 2, deflatingthem by the ConsumersPrice Index, and cumulating them at the rate of 5 per cent, I get the figureof 3,180 million 1951 dollars as my estimateof total cumulated research expendituresin agriculture.Comparing this with the two estimated limits of the annual social.returnsof $1,110 and $5,430 million, I get a lower limit of 35 and an upper limit of 171 per cent as estimatesof the annual rate of return research. per dollar spent on agricultural These are substantiallylower than the estimatedreturnsfromhybridcorn but are comparable to estimates made by Ewell forthe economyas a whole (100200 per cent per year per dollar spent on "researchand development") and to figures quoted by major industrialcompanies on theirreturnson research.21 Of course, these estimates are quite consistent with the estimated returns on hybrid corn, if the probability of success in research on innovations like hybridcorn is on the orderof one-tenth or one-twentieth. Nevertheless,all these figures indicate that,in spite of the large growthin research expendituresduring this century,the social returnsto this activityare still veryhigh.

example, hybridsorghum.The development of hybridsorghumbegan seriously only afterWorld War II but has gained momentum rapidly since. Hybrid sorghum is now being introducedcommercially. Very little was planted in 1956, but substantial amounts were planted in 1957. The experimentaldata to date suggest that the superiorityof hybrid over previousseed may be even sorghum greaterthan that of hybridcorn.22 The 1956 value of the grain sorghum crop was approximately $232 million. Assuming a 15 per cent superiorityof hybrids,a demand elasticity of -1.0, elastic supply curve, and an infinitely the estimated gross social returnsfrom sorghum hybrids would be about $37 million annually.23Assuming that the extra cost of hybridseed will be about $7 million annually, the net annual re.24 turns would be about $30 million I data on researchexpendinave no official tures on hybrid sorghum.The head of one of the major seed companies has estimatedthat to date all public and private expenditures on hybrid sorghum total approximately millionand that $1 currentexpendituresare at an annual rate of $300,000. Doubling his estimate and projectinginto of past expenditures the futurean annual research expendiHYBRID SORGHUM ture rate of $500,000, I get $10 or $13 dependingon the rate of interest The approach previouslyoutlinedcan million, be used to estimate the probable rate 22 SorghuZtm Hybrids (USDA, ARS, Special Report of return on a new development: for 22-26 [Washington, May, 1956]).
meant stations spent ai)ut $118 million annually
2

23 This figure is derived as follows: 0.15 (1 + X 0.15 X 1.0) X $232 million= 0.16 X $232 = on research. However,on.iv one-third theindustry $37 million.The figurewould have been approxiof research was in aid of farm production,mainly mately twice as large if I had used the value of the in machineryand cheminicals: rest was used the 1957 crop-$493 million-as mybase. of in product rc:^earch, while fPur-fifths public ex2 The additional cost of hybridseed, seven milpenditureswereforresearchin aid of fiarrn produc- liolndollars, is estimatedfromthe followingdata:

tion (see R. Mlighell, A mcwrioan A oricul-bure [New

York: JohnWiley & SoTIs,1955J, 130). p.

21 R. H. Ewell, 'Role of Research in Economic Growth,"Chlemical (andEngineering News,XXXIII (1955), 298-304.

10 millionacres sown to sorghum;ten dollars differencebetween the average prices per hundredgrainsorghum weightofhybridand open-pollinated seed (Agricultural Prices, April, 1958, p. 42); and an average seedingrate of seven pounds per acre.

RESEARCH COSTS AND SOCIAL RETURNS

429

used, as my "estimate" of cumulated hybrid-sorghum researchexpenditures in 1967. I choose 1967 as the reckoning date on the assumptionthat it will take ten years forhybridsorghumto capture most of the sorghumseed market.While the projected rate of development is faster than that of hybrid corn in the United States as a whole,it is equivalent to the rate of acceptance of hybridcorn in Iowa. It is reasonable to assume that hybridsorghumwill spread much faster than hybridcorn did. Sorghumproduction is more localized than corn production, almost all sorghum is grown for commercial purposes, and hybrid sorghum will probablyencountermuch less resistancethan hybridcornencountered. I also assume that the use of hybrid in sorghum the United States will follow the same time path that the use of hybrid corn followed in Iowa.25 This asallowsme to estimatethesocial sumption returnsduring the "transitionperiod," 1957-67. Table 3 outlines the calculation of the estimatedrate of returnon research expenditureson hybridsorghum,which 400 is approximately per centper annum. While this is somewhat lower than the estimatedrate of returnon expenditures on hybrid-corn research,it is still very high indeed.
SOME IMPLICATIONS

corn breeding and because sorghumgrowing is much more localized than corn-growing. Therefore,adaptable hybrids will have to be developed for a much smaller portion of the United States. Nevertheless, the estimated returns from hybrid sorghum are lower than those fromhybrid corn. Why? To some extent this lower rate may be a result of overestimatingresearch costs for hybrid sorghum,but the principal explanation is that the total value of the sorghum crop is substantially smaller than that of corn; sorghum a relatively is
TABLE 3
ESTIMATED COSTS AND RETURNS OF HYBRID-SORGiHUM RESEARCH AS OF 1967

(Millions Dollars) of Cumulated socialnetreturns, bValue in 1967of returns Total value of net returns in
1967.................... ............... yond 1967 1957-66.................
r=O0.05 r=O.10

155

171 295 466

590 745

research Cumulated ex-:enditures 9.4 .................... Benefit-cost ........... ratio 79

"Rate of return"in per cent

13 36 360

........ perannum

395

One mighthave expected, on a priori grounds, that the rate of returnon exresearch penditures on hybrid-sorghum would be higherthan the returnon hybrid-cornresearch. The cost of hybrid sorghum has been and will be lower than the cost ofhybridcornboth because of the cumulated experience in hybrid25 The percentage of all cornplanted withhybrid seed in Iowa followedthe followingtime path by years: 0.02, 0.06, 0.14, 0.31, 0.52, 0.73, 0.90, 0.99, and 1.00.

unimportantcrop. It has recentlybeen suggested that we should redirect our research effortsaway from the major commoditiesthat are in "surplus" and away from commoditieswith low elasticities of demand, where technical improvementsresult in reduced total returns to farmersin the long run. I-owever, if we assume, in the absence of that I any otherinformation, technological change operates somewhat like a percentage increase in yield, and that the cost of achieving a given percentage boost in yield is the same for different crops or at least independent of their price elasticities and relative "importance," then the highest social returns

430

ZVI GRILICHES

per research dollar are to be found in formof lower prices and higheroutput. the important,low-elasticitycommodi- Entry into the hybrid-seed industry was ties. For it can be easily shown that easy, and in the long run no "abnormal" the absolute social gain from a given profits were made there.By valuing the percentage increase in yield will vary extra cost of seed productionat the marproportionately with the total value of ket price,I have countedas a cost whatthe crop and that the impact of different ever profit was made in this area by demand-and-supplyelasticities is of a privateproducers, and the resulting estisecond order of magnitude. The latter mate of returnsconsists almost entirely affect only the "triangle,"or the magni- of social rather than private returns. tude of the price-change adjustment. These social returns were diffused widely Amongall the different of factors,the total among consumers cornand cornprodvalue of the crop is by far the most ucts. Given the difficulty patenting of importantdeterminantof the absolute most of the valuable ideas in this area, social gain froma given percentage in- the shortlifeof a patent,and the general crease in yield. precariousness of a monopoly position No matter how we calculate them, in the long run, the incentiveforprivate there is little doubt that the over-all investment was very much smallerthan social returns publiclysupportedtech- that impliedby the social rate of return. on nological researchhave been very high. While a divergence betweensocial and It is not clear, however,whetheror not private rates of return is a necessary thisfacthas any normativeimplications. reason for public intervention, is not, it I am afraidit has veryfew.More knowl- by itself,a sufficient reason. We must edge than is now at hand is required ask not only whethersocial returnsare forpresciption. higher than private-this is also true It is clear that we have not succeeded of many private investments-but also in equalizing the returns on different whether the private rate of return is kinds of public investments. The returns too low,relativeto returns alternative on from technologicalresearch have been private investments, induce the right to much higherthan the returnsfromrec- amount of investmentat the righttime. lamation and watershed projects. But The social returns from nylonwereprobshould we have had more technological ably many times higher than DuPont research? Surely,we have yet to reach but the latter were high enough profits, the optimal level of expenditureson reto inducethe development nylonwithof search, but this can only be a hunch. out a public subsidy,although,perhaps, Should the public support agricultural not soon enough. To establish a case research? My analysis illustrates and quantifiesone of the major arguments for public investment one must show are that,in an area where social returns forpublic investments this area-the in because of the nahigh,private returns, divergencebetween the social and private rates of return.Almost none of the ture of the inventionor of the relevant are calculated social returns from hybrid institutions, not highenoughrelative to other private alternatives. This was corn were appropriatedby the hybridseed industry by cornproducers.They undoubtedly true of hybrid corn, and or were passed on to consumers in the it is probably true of many other areas

RESEARCH COSTS AND SOCIAL RETURNS

431

of agricultural research and basic researchin general.But it is not universally true. Hence a high social rate of return is not an unequivocal signal for public investment. In this paper I have estimated that the rate of returnon public investments in one of the most successfulventures of the past has been very high. This may give supportto our intuitivefeeling that the returns to such ventures in general have been quite high and to

our feeling that "research is a good thing." But that does not mean that we should spend any amount of money called "research."The moral on anything some sort is that, though very difficult, calculation is possiof cost-and-returns ble and should be made. Conceptually, the decisions made by an administrator of research funds are among the most economicdecisionsto make and difficult to evaluate, but basically they are not from any other type of very different decision. entrepreneurial