Professional Documents
Culture Documents
Alston, 2018
Alston, 2018
Sixty years ago, T.W. Schultz introduced the idea of the productivity “residual” to agricultural
economics (1956). Schultz’s main message was that growth in conventional inputs accounted for little
of the observed growth in agricultural output, and that there was work to be done by agricultural
economists to understand and ultimately eliminate this unexplained residual called “productivity.”
Thus was launched the economics of agricultural productivity as a sub-field within agricultural
economics, along with the economics of agricultural R&D and innovation and related government
policy. Much progress has been made in the decades since. Still, critical issues remain unresolved.
This matters because agricultural innovation and productivity matter, and so do the related policies
that rest to some extent on our established understanding of the economic relationships. In this paper
I review some unsettled issues related to economic models and measures applied to agricultural
R&D and productivity, and some unfinished business in terms of economic and policy questions that
are not yet well answered. Before doing that, I present some evidence on agricultural productivity
and why it matters. Next, with a nod to “factology,” I present available productivity measures from
the U.S. Department of Agriculture (USDA) and International Science and Technology Practice
and Policy (InSTePP) Center, and compare them in the context of translog cost function models. In
subsequent sections I use these and other data to develop new evidence related to two contentious
questions: (a) Do farmers benefit from public agricultural R&D? (b) Has U.S. agricultural productiv-
ity growth slowed in recent decades? The answers are revealed within.
Sixty years ago, T.W. Schultz (1956) called “productivity.” Schultz and his student
introduced the idea of the productivity Zvi Griliches (1963) made initial inroads into
“residual” to agricultural economics. The this task, identifying improvements in quality
main message in that seminal paper was that of inputs, including education of farmers, and
growth in conventional inputs accounted for changes in technology as major sources of
little of the observed growth in agricultural productivity growth—attributable ultimately
output, and that there was work to be done by to investments in science, education, and for-
agricultural economists to understand and ul- mal and informal R&D by individuals, firms,
timately eliminate this unexplained residual and governments.
Thus was launched the economics of agri-
cultural productivity as a sub-field within agri-
Julian Alston is a distinguished professor in the Department of
cultural economics, along with the economics
Agricultural and Resource Economics, and Director of the of agricultural R&D and innovation and
Robert Mondavi Institute Center for Wine Economics, both at related government policy. Much progress
the University of California, Davis. Alston is also a member of
the Giannini Foundation of Agricultural Economics. The work made in the decades since, elaborating on the
for this project was partly supported by the University of ideas introduced by Schultz and Griliches,
California and the Giannini Foundation of Agricultural refining the concepts, and improving the meth-
Economics. Connie Chan-Kang, Jarrett Hart, Yijing (Olivia)
Wang, and Yanuo Zhou provided excellent research assistance; ods of measurement and measures. Still, criti-
Rob Fraser, John Freebairn, Lilyan Fulginiti, Jeff Hopkins, Phil cal issues remain unresolved; much is in
Pardey, Dick Perrin, Josef Schmidhuber, Daniel Sumner, and
Jim Vercammen provided helpful comments and advice. This
dispute, and in many aspects the evidence is
good help is all gratefully acknowledged. Correspondence may unconvincing or the state of the art is
be sent to: jmalston@ucdavis.edu.
otherwise unsatisfactory. This matters because and real value of agricultural output more
agricultural innovation and productivity mat- than trebled, even though land in agriculture
ter, and so do the related policies that rest to increased by only about one-tenth, making
some extent on our established understanding for an era of unprecedented agricultural
of the economic relationships. abundance (Alston and Pardey 2014). This
The economic policy stakes are large. remarkable accomplishment belied wide-
Agricultural innovation, increasingly driven spread prophecies of doom in the early 1960s,
by organized agricultural R&D, has served as coming from a perspective—based on tradi-
an engine of economic progress, priming the tional agriculture and conventional inputs—
pump of the agricultural transformation that that did not anticipate the transformation
has been an integral element of economic of agriculture that was to come soon after
progress around the world. But incentives are for large parts of the world. Investments
weak and attenuated in many areas of agricul- in agricultural science and technology and
tural R&D, leaving a crucial role for govern- other “nontraditional” inputs such as knowl-
ments to play both in establishing the edge and education, and improvements in the
institutions that encourage private innovation, quality of material inputs and people, played a
including investments in R&D, and in filling crucial role, helping to shift agriculture to a fir-
the gaps that remain by funding public invest- mer footing and capitalize on agriculture as an
ments in socially valuable research areas that engine of economic growth (Timmer 2009;
the private sector nevertheless neglects. Alston and Pardey 2014, 2017).
Evidence of remarkably high sustained rates The experience of the United States illus-
of social payoffs to both private and public trates the value of farm productivity growth
investments in agricultural R&D testify to a in the context of the agricultural transition.
significant failure of government to fully ad- In 1916, the U.S. farm population peaked at
dress the underinvestment problems caused by 32.5 million, 31.9% of the total U.S. popula-
the market failure. Moreover, if anything, in tion. Since then, while the U.S. population
high-income countries like the United States, continued to grow, the farm population de-
agricultural R&D policies seem to be trending clined to an estimated 4.6 million in 2013, just
in the wrong direction, making matters worse. 1.5% of the total. A dramatic transformation
Is this to some extent our fault—a failure of of agriculture, with farms becoming much
the profession to provide enough of the right larger (and symmetrically many fewer) and
types of evidence and ideas at the right time to more specialized was achieved through the
enable better policies to be made? Or is it sim- progressive introduction and adoption of a
ply a reflection of the limited role of economic host of technological innovations and other
evidence in policymaking—especially as it farming improvements that enabled much
applies to global public goods like agricultural more to be produced with less land and a lot
science and climate change, where the lags be- less labor. Productivity grew rapidly. For ex-
tween action and outcome are extremely long ample, Schultz (1956) wrote: “From 1923 to
and variable, the benefits are diffuse, and 1929 only about one-half—or a little more—
much uncertainty surrounds who precisely will of the increase in output appears to have
benefit, when, and by how much. been achieved by additional inputs. During
In what follows I review some unsettled the depression years, 1930 to 1940, none of
(and perhaps unsettling) issues related to eco- the increase in output seems to be explained
nomic models and measures applied to agri- by additional inputs. . . . The war years called
cultural R&D and productivity, and some (to forth substantially more output, yet from
some extent consequently) unfinished busi- 1940 to 1948 perhaps only a fifth to a fourth
ness in terms of economic and policy ques- of the increase in output can be explained by
tions that are not yet well answered. Before additional inputs.”
doing that, I present some evidence on agri- The same phenomenon continued over the
cultural productivity and why it matters. next 60 years, as U.S. farm output more than
doubled in spite of reductions in land and es-
pecially labor employed in agricultural pro-
Productivity Matters (and the Agricultural duction. Specifically, between 1949 and 2007
R&D Policy Paradox) the InSTePP (Pardey et al. 2006) index of
U.S. farm output (Q) increased from 100 to
Over the past half-century, while the world’s 269, but none of this increase can be
population more than doubled, the quantity explained by increases in inputs (adjusted for
300
280.3
Multi-Factor Productivity
268.7
250
Output Index
200
150
100 95.9
Input Index
50
0
1949 1959 1969 1979 1989 1999
compositional changes) because the index of ag- technology, developed and adopted as a re-
gregate inputs (X) actually fell from 100 to 96 sult of agricultural research and extension,
such that the index of multifactor productivity and other sources of innovations.
(MFP ¼ Q/X) grew from 100 to 280.1 Hence, Another way to look at this counterfactual
the unexplained differential between output is in terms of the quantities of inputs that
growth and input growth—called “productivity would be required to produce the 2007 quan-
growth”—accounts for more than 100% of the tity of output (2.7 times the 1949 output) us-
postwar increase in U.S. farm output.2 ing 1949 technology (i.e., productivity and
Clearly agricultural productivity growth is factor shares): that is, 2.7 times the 1949
enormously valuable. Of the actual farm out- quantities.3 An increase to 2.7 times the ac-
put in 2007, worth about $330 billion, only tual quantities of land and labor (along with
one-third (i.e., 100/280 ¼ 0.36) or about $118 capital and other inputs) used in 1949 would
billion could be accounted for by conven- require adding 2.0 billion acres (an area the
tional inputs using 1949 technology, holding size of the contiguous United States or
productivity constant. The remaining two- Australia, much more than the total agricul-
thirds (i.e., 180/280 ¼ 0.64) or about $212 bil- tural area in either country) to the 1949
lion in that year alone, is attributable to the quantity of land used in agriculture, and an
factors that gave rise to a 180% increase in additional 34 billion hours of operator, fam-
productivity since 1949—including improve- ily, and hired labor (or about 12 million
ments in infrastructure and inputs (if not cap- FTEs); the required increase would be
tured already in the indexes), as well as new closer to 2.5 billion acres over the 2007
quantity of land and a fivefold increase over
the 2007 farm labor force! That these
1
required increases seem absurd speaks
Pardey et al. (2006) describe the University of Minnesota’s
InSTePP (International Science and Technology Practice and
Policy center) Production Accounts, Version 4 (documentation
available at http://www.instepp.umn.edu/products/instepp-pri-
3
mary-data-documentation-us-agricultural-input-output-and-pro- An alternative measure is 2.8 times the actual quantities of
ductivity-series); Version 5, used here, is similar. land, labor, capital, and other inputs used in 2007 (i.e., using 2007
2
If, instead of falling, inputs had been held constant at the factor shares), which would be less dramatic but nonetheless
1949 quantities, output would have increased by a factor of 2.8:1 striking. Neither of these counterfactuals is entirely reasonable
instead of 2.7:1. or ideal for our purposes.
to the remarkable “productivity” perfor- farm commodity policies (trade policies and
mance and the transformation of U.S. other farm subsidies).
agriculture in the second half of the twenti- In an analysis of policies as they applied in
eth century. 2006, for example, I estimated that, for every
Available evidence from much study of dollar of U.S. government spending on farm
the subject indicates that a very large (and subsidies, farmers received about 50 cents,
probably increasing) share of U.S. agricul- landlords received about 25 cents, domestic
tural productivity growth is attributable to and foreign consumers received about 20
organized agricultural science—whether con- cents, and 5 cents was wasted (see Alston
ducted in the private sector or the public 2009). Additional amounts are wasted col-
sector. The annual value of farm productivity lecting the taxes to finance the spending and
growth is large, as we have seen, both in in administering the policies—costing tax-
absolute terms and relative to the annual payers perhaps another 20 cents per dollar.
investment in farm productivity-enhancing By these measures, eliminating expenditure
R&D. Partly for that reason, the returns to on farm programs of $20 billion (with a social
those investments have been very favorable, opportunity cost of, say, $24 billion) would
indicating a systematic sustained pattern of cost farmers $10 billion and would avert a net
underinvestment from the point of view of social cost of $1 billion (or $5 billion if we
society as a whole (see, e.g., Rao, Hurley, count the social opportunity cost of the
and Pardey 2016). funds).
These main claims are not in dispute (al- In contrast, agricultural research involves a
though some economists working in the area “deadweight gain.” Our estimates (Alston et
may disagree with one another about specific al. 2010a, 2011) indicate that U.S. federal and
aspects of the measures of research returns state government expenditure on agricul-
and their intepretation). Remarkably, never- tural research and extension generates
theless, during the twenty-first century in benefit-cost ratios of at least 10:1 (more
real terms, U.S. public funding for agricul- likely 20:1 or 30:1)—evidence of a serious
tural and food R&D has stalled and more re- underinvestment. Pardey and Beddow
cently shrunk, and within this now-shrinking (2017), echoing Pardey, Alston, and Chan-
agricultural and food R&D portfolio, the Kang (2013), suggested that a reasonable
balance has shifted away from farm first step would be to double U.S. public in-
productivity-enhancing R&D (see, e.g., vestment in agricultural R&D—an increase
Pardey, Alston, and Chan-Kang 2013; of, say, $4 billion over recent annual expen-
Clancy, Fuglie, and Heisey, 2016). The U.S. ditures.4 A conservatively low benefit-cost
government does not show any sign of re- ratio of 10:1 implies that having failed to
versing those trends; an exacerbation seems spend that additional $4 billion per year on
more likely. public agricultural R&D imposes a net social
It appears to be widely understood that cost of $36 billion per year—an order of
market failure is widespread in the context of magnitude greater than the annual $1–5 bil-
agricultural science and technology. It may lion social cost of $20 billion in farm
be less well understood that governments subsidies.
have not done nearly enough to address the These rough calculations are crude but
consequences of these market failures, that plausible in view of the value of productivity
government failure is also widespread in this growth and the role of R&D in driving it, and
context. Agricultural economists have the compelling evidence of sustained high
invested a great deal in modeling and mea- rates of return to R&D. We can quibble
suring the social costs of government failure about whether the benefit-cost ratio would be
to choose farm subsidy policies or agricul- so large for such a large increase, and how
tural trade policies that will maximize na- such an increase might have to be phased in
tional economic surplus; very little, by
comparison, has been done on analyzing and
measuring the social cost of government fail- 4
Compilations of estimates of public and private spending on
ure in agricultural R&D policy. But the size agricultural and food R&D in the United States are available
of the economic losses from poor agricultural from the USDA (see, e.g., Fuglie and Toole 2014) and InSTePP
R&D policy decisions may be orders of mag- (see, e.g., Pardey, Alston, and Chan-Kang 2013). Either source
would support an estimate of public spending of a bit over $4 bil-
nitude greater than the losses from the much lion in recent years (roughly one-third of the total public and pri-
more extensively quantified distortions from vate spending on agricultural and food R&D).
over time to allow for developing the capacity depending on the use to which the data are to
to spend the funds well and wisely; but any- be put, it is especially important for the user
one who has doubts about the scale of the po- to undertake an investigation in factology be-
tential benefits relative to the expenditures is fore telling the story.”
advised to consider the scale of the past bene- In his Presidential Address to the
fits attributable to farm productivity growth American Economic Association, Zvi
and the implications of an absence of that Griliches (1994) made some similar points,
growth for either massive output reductions with specific reference to productivity mea-
or huge additional resource requirements to surement: “[We] often misinterpret the avail-
sustain production. Perhaps this topic able data because of inadequate attention to
deserves more attention from economists in- how they are produced and that same inat-
terested in the social costs of distortions in tention by us to the sources of our data helps
agricultural policy—including those that re- explain why progress [in understanding the
flect errors of omission (as in agricultural sources of productivity growth] is so slow. . . .
R&D) as well as errors of commission (as in Great advances have been made in theory
subsidies). and in econometric techniques, but these will
Here endeth my attempt to provide a moti- be wasted unless they are applied to the right
vation for an interest in the economics of ag- data.”
ricultural R&D and productivity and related Available measures of U.S. agricultural
policy, and for some concern over whether inputs, outputs, and productivity and their
we are getting the relevant concepts and evolution illustrate the issues well. The
measures right. What follows is a selective USDA Economic Research Service (USDA-
sampling of some unsettled issues among ERS) has played a longstanding role, led well
economists and some unfinished business in by Eldon Ball for many years, in developing
economics research related to agricultural measures of U.S. agricultural productivity
R&D and productivity and related policy. A (see, e.g., Ball et al. 2016). As discussed by
discussion of factology is a good place to Gardner et al. (1980), and by Shumway et al.
start. (2015, 2016) in the most recent episodic exter-
nal review of that activity, these measures have
certainly come a long way in terms of address-
ing the conceptual and measurement issues
Factology and Productivity Measures raised by Schultz and Griliches. Since the early
1980s, a parallel project has been underway at
Twenty-five years ago, in his Fellow’s Lecture the University of Minnesota, led by Philip
to this Association, Bruce Gardner (1992) Pardey and, since 2003, conducted under the
discussed the importance of data creation auspices of the International Science and
and of having econometricians, policy ana- Technology Practice and Policy (InSTePP)
lysts, and other data users know how the data Center (see, e.g., Pardey and Craig 1989, Craig
they use were created: “Agricultural econo- and Pardey 1996). The two projects have
mists and other social scientists tend to take evolved in parallel, to some extent competi-
data as facts. . . The problem is the data are tively, with mutual learning and improvement
not facts. Facts are what is really there. Data in both and some convergence over the course
are quantitative representation of facts, of more than a quarter of a century. Today, a
which statistical workers and economists con- researcher can download annual series of state-
coct. . . I call the study of how primary statisti- specific and national indexes of prices and
cal information is made into economic data quantities of inputs and outputs, and productiv-
‘factology’. The neglect of factology risks sci- ity, from either InSTePP for the years 1949–
entific ruin.” 2007 or the USDA for the years 1948–2013
Factology is important in the context of the (only 1960–2004 for state-specific indexes).5
measurement of agricultural inputs, outputs,
and productivity, a context where analysts in
constructing the “data” make a great many
choices that can matter much for the meas-
ures and their interpretation. As Gardner 5
The USDA data (1948–2013) are available at https://www.
(1992, p. 1071) noted specifically in the con- ers.usda.gov/data-products/agricultural-productivity-in-the-us/ag-
ricultural-productivity-in-the-us/#National Tables, 1948–2013,
text of productivity measures: “For data that and the InSTePP data (1949–2007) are available at http://www.
themselves tell a story, and a different story instepp.umn.edu/united-states#Version 5.
Which should we use? Does it matter?6 How possibilities, table 1 reports results from esti-
should we decide? mating a Translog cost function model using
As discussed by Andersen, Alston, and the national annual data from the two data
Pardey (2011), and earlier by Acquaye, Alston, sets for the years for which both were avail-
and Pardey (2003), the two sets of measures able, 1949 to 2007.8 The models were esti-
have much in common in terms of their meth- mated subject to the conventional
odological underpinnings and data sources. homogeneity and symmetry restrictions and,
Both sets of estimates apply the same index less conventionally, imposing constant
number theory and use discrete approxima- returns to scale at the level of the industry,
tions to Divisia indexes, and both begin with U.S. agriculture, as a whole.
highly disaggregated inputs and outputs to re- The two datasets imply quite different esti-
duce the effects of pre-aggregation bias.7 But mates for the rates of factor-neutral technical
they also entail some notable differences in change (twice as fast using the InSTePP data
methods that give rise to some meaningful dif- compared with the USDA data), and some
ferences in indexes that ostensibly measure the substantive differences in the detail of the pat-
same conceptual quantity—and more so for tern of factor-biased technical change (though
some states than for the national indexes. both indicate that technical change has been
Andersen, Alston, and Pardey (2011) identi- land-, labor-, and capital-saving, and “other
fied differences in capital as being compara- inputs”-using, consistent with the gross trends
tively important. Certainly, their plots of in factor shares). Perhaps the most significant
InSTePP versus USDA data tell very different difference is with respect to the elasticities of
stories about capital use in U.S. agriculture factor demand (and corresponding elasticities
over the second half of the twentieth century. of substitution). The estimates based on the
Andersen, Alston, and Pardey (2011) showed USDA data imply a (statistically significantly)
that these differences could be attributed positive value for the Hicksian (i.e., output
largely to the use of the annually variable ver- constant) own-price elasticity of demand for
sus fixed interest rates in the computation of capital at the mean vector of predicted input
capital service flows, and they argued in favor shares, which is not consistent with the under-
of the fixed interest rate approach as used by lying theory. The other elasticities are largely
InSTePP, both on conceptual grounds and comparable between the two sets of esti-
because the resulting measures appeared mates: the other own-price elasticities are all
more sensible. The differences remain inter- plausible (and similar); the cross-price elastic-
esting regardless of which approach we ities are mostly positive, indicating Hicksian
might prefer in principle—perhaps more so substitutes, but the estimates using the USDA
if we are not sure. data suggest that capital is a complement for
Many studies use one of these datasets or both labor and “other” inputs, while the esti-
the other without apparent regard to fragility mates using the InSTePP data suggest that
from this source, and perhaps without any at- capital and land are complements.9
tention to factology. Andersen, Alston, and
Pardey (2011) identified numerous examples
of studies that utilized the USDA data in a 8
variety of applications, noting that the results Jarrett Hart did the main work of estimating these models,
under my direction. These estimates are not meant to be treated
from those studies that focused on input use too seriously as such—the models were not estimated with that
may be especially contingent on the measures intention and were not subjected to appropriate elaboration and
testing—but are provided merely to illustrate how results based
of capital service flows. To illustrate these on just one of these datasets might be fragile from this
perspective.
9
In evaluating the results from these estimations, I noticed
that the USDA-ERS price index for services from land is re-
6
In reviewing the USDA-ERS data, Shumway et al. (2015, markably volatile, dropping from 1.05 in 1996 to 0.16 in 2000 and
2016) paid remarkably little attention to the existence of the al- 0.12 in 2002 before jumping to 1.35 in 2004. These land rental
ternative, InSTePP data, the differences between the two data price gyrations have significant (and seemingly implausible)
sets, and what to make of the differences. implications for both the observed and predicted cost share of
7
Agricultural production uses some inputs (such as the fossil land (including some negative predicted values from the
water in the Ogallala aquifer) and produces some outputs (such Translog model) and could well have influenced the cost-
as greenhouse gases and environmental amenities) that are not function estimation results and other analysis using these data.
traded in markets. Such “environmental” impacts, which are a This feature of the land price index appears to be attributable to
potential concern for measures of production and productivity the practice of treating land as the residual claimant, for the pur-
more broadly (see, e.g., Feldstein 2017), are not explicitly pose of computing factor payments to land. In their review of the
addressed in either of the two sets of measures of U.S. agricul- USDA-ERS data, Shumway et al. (2014, 2016) discussed (and
tural productivity. Ball et al. (2004) illustrate the issues and a largely endorsed) this approach, but they do not appear to have
possible approach in one application. noticed its implications for the measures.
Table 1. Results from Translog Cost Function Models Applied to U.S. Agriculture, 1949–2007
Panel a. Elasticities of factor demand and elasticities of substitution, at sample means
Hicksian Elasticities of Factor Demand Elasticities of Substitution
USDA InSTePP USDA InSTePP
gkk 0.04 –0.19 rkk 0.30 –1.40
gkl 0.03 –0.08 rkl 0.20 –0.49
gkh –0.04 0.10 rkh –0.20 0.28
gko –0.03 0.18 rko –0.06 0.54
glk 0.03 –0.07 rlk 0.20 –0.49
gll –0.29 –0.31 rll –1.88 –1.80
glh 0.12 0.37 rlh 0.56 1.04
glo 0.15 0.01 rlo 0.30 0.02
ghk –0.03 0.04 rhk –0.20 0.28
ghl 0.09 0.18 rhl 0.56 1.04
ghh –0.40 –0.63 rhh –1.89 –1.78
gho 0.34 0.42 rho 0.69 1.25
gok –0.01 0.07 rok –0.06 0.54
gol 0.05 0.00 rol 0.30 0.02
goh 0.15 0.44 roh 0.69 1.25
goo –0.65 –0.73 roo –0.37 –1.56
Panel b. Annual rates of factor neutral and factor biased technical change
Input USDA InSTePP
Capital (k) 0.0001 0.0004
Land (l) 0.0011 0.0009
Labor (h) 0.0023 0.0016
Other (o) 0.0036 0.0029
Neutral 0.0076 0.0151
Source: Derived from estimates of cost functions, further details available from the author upon request.
Note: “USDA” and “InSTePP” denote sources of data used for estimating cost function models using data for the period 1949–2007. The indexes are “k” for
capital, “l” for land, “h” for “labor” and “o” for the aggregate of all other inputs. In Panel a, all of the coefficient estimates used in calculating the elasticities
were statistically significantly different from zero at the 95% level of confidence, as were all the estimates in Panel b, except for the rate of capital–saving
technical change estimated using the USDA data. In Panel b, a negative coefficient indicates factor-saving technical change.
It is not entirely surprising to see capital fea- Translog model is the right model or that we
tured in the differences in elasticities given the have shown that the InSTePP measure of capi-
observation that the most important difference tal is better constructed, though given the use
between the two datasets is in the capital se- of that model the results would seem to favor
ries (as noted by Acquaye, Alston, and Pardey using the InSTePP data.10
2003, and Andersen, Alston, and Pardey In what follows we will revisit this issue—
2011). These findings confirm the suggestion which dataset to use and how it affects
by Gardner (1992), echoed by Andersen, findings—in the context of other unsettled
Alston, and Pardey (2011), that differences in issues and other unfinished business. Among
decisions about data construction—in the con- the unsettled issues is whether farmers bene-
text of measures of agricultural inputs, out- fit from agricultural R&D in the long run.
puts, and productivity—can have Some of us may think we know the answer to
consequences for the inferences drawn from this question, but I am not sure we even
studies that use the data, whether largely know how to ask it. To some extent the
untransformed (as in computing productivity
measures or other descriptive use) or with
more complicated transformations (as in esti-
10
mating a Translog cost function model and the In preliminary analysis, applying the same approach but
with a Generalized Leontief model instead of a Translog model
implied elasticities of factor demand or rates yielded a different pattern of differences in results; generally less
of technical change). We cannot say the satisfactory estimates with both datasets.
answer may turn on the quality of the “data” associated with how the “producer surplus” is
available, but some of the issues are concep- distributed among factor suppliers—do land
tual. The same is true of some other unsettled owners benefit at the expense of suppliers of
questions related to measures of returns to farm labor, including farm operators, or vice
research and whether productivity growth versa? In particular, how is the total producer
has slowed. This is a large and diverse incidence (benefits or costs) distributed be-
agenda, and some of it will get only brief at- tween “farmers” (as the suppliers of land,
tention here. family labor, and managerial inputs used in
agricultural production) and “others” (as the
suppliers of inputs purchased by farmers, in-
Who Benefits from Agricultural R&D? cluding hired labor, and other agribusiness
inputs used in activities beyond the farm
A few years ago, colleagues and I wrote an gate).
article for the Annual Review of Resource We can contemplate and potentially mea-
Economics (Alston et al. 2009) in which we sure these outcomes using a variant of the
briefly summarized a decades-long discussion Muth (1965) two-factor, single-commodity
over the determinants of the total benefits market model in which research gives rise to
from agricultural R&D and the distribution factor-augmenting changes in technology,
of benefits between producers and consum- which imply shifts in factor demand and
ers. In the conclusion, we noted that, in spite product supply. Here, benefits to farmers cor-
of all that work, compelling evidence has not respond to producer surplus measured off the
been developed to show whether farmers supply function for the factor(s) supplied by
benefit from technological change in agricul- farmers and, under the maintained assump-
ture. This is serious unfinished business. tion of competition, total benefits are given
In the conventional model, research causes by the sum of changes in producer surplus
the supply curve for agricultural output to shift across factor suppliers plus consumer surplus
down and out against a stationary demand in the output market. In this context, farmers
curve, giving rise to an increase in quantity stand to benefit from technological change if
produced and consumed, and a lower price. it causes an increase in demand for the fac-
The total gross annual research benefits tors they supply, and thus an increase in ex-
(GARB) depend primarily on the size of the penditure on them.
research-induced supply shift and the scale of Using the solution to the Muth model in
the industry to which it applies. The distribu- Alston, Norton, and Pardey (1995), and inter-
tion of the benefits between producers and preting factor X2 as representing inputs sup-
consumers depends on the relative elasticities plied by farmers (and X1, all inputs
of supply and demand, the nature of the purchased by farmers), the proportional rate
research-induced supply shift and, less impor- of change (denoted by E) in expenditure on
tantly, on the functional forms of supply and inputs supplied by farmers (R2) (or income
demand. Specifically, producers are assured of accruing to farmers from farming) as a
benefits from vertically-parallel supply shifts function of the rates of factor-neutral and
but they may lose from pivotal supply shifts, X2-saving technological change, d and c,
depending on the elasticities. In particular, respectively, is given by:
producers do not benefit from a pivotal shift
unless demand is elastic. 1
ð1Þ E ðR 2 Þ ¼ ð1 þ 2 Þ
The possibility of losses to producers in ag- D
gregate is often discounted, on the grounds
either that demand is relatively elastic (which 1 s2
ðr þ 1 Þð1 gÞd þ rð1 þ gÞc
is valid in some contexts), or that a parallel s2
research-induced supply shift is relatively
likely (or that a pivotal shift seems compara- where i denotes the supply elasticity for fac-
tively unlikely), but concrete empirical evi- tor i, r denotes the elasticity of factor substi-
dence has been elusive to date. Thus, even tution, g denotes the absolute value of the
when we can be assured of benefits to the na- elasticity of demand for the final product, s2
tion, some uncertainty remains about the dis- represents expenditure on inputs supplied by
tribution of benefits between producers and farmers (s1 ¼ 1 – s2 for other inputs) as a
consumers. Further distributional issues are share of total expenditure (or total gross
farm income), and D ¼ rðg þ s1 1 þ s2 2 Þþ As noted, estimates using the two datasets
gðs2 1 þ s1 2 Þ þ 1 2 > 0:11 Since all of the (from USDA and InSTePP) both indicate
parameters are defined as positive magni- that technical change has been land-, labor-,
tudes, E(R2) is unambiguously negative if de- and capital-saving, and “other inputs”-us-
mand is inelastic (g < 1Þ, farmers are made ing.14 In a counterfactual prediction from the
worse off by both neutral technical change and estimated model using the InSTePP data for
X2-(farmer-supplied inputs)-saving technical 1949–2007, to produce the actual quantity of
change. If demand for farm output is elastic output produced in 2007 in the absence of
(g > 1Þ, however, the sign of equation (1) is any technological changes (i.e., eliminating
ambiguous—farmers are still made worse off the time-trend effects) since 1949, the cost-
by X2-saving technical change but they benefit minimizing quantities of land and labor
from neutral technical change, and the net out- would be 3.6 and 3.5 times the actual quanti-
come depends on the relative importance of ties; stated differently, technological changes
the two types of technological change.12 since 1949 reduced the demand for land and
Consider the case of U.S. agriculture. Is de- labor by 73% and 71% (for capital, by 67%
mand for aggregate U.S. farm output inelas- and for other inputs, by 37%) relative to the
tic? Has innovation been predominantly amounts that would have been demanded
either factor-neutral or biased in the direction otherwise to produce that actual 2007 output
of saving land, labor, and other inputs sup- quantity, thereby reducing average and total
plied by farmers? If the answer to both ques- costs by 62%. In contrast, using the estimates
tions is yes, then we would expect to have based on the USDA data instead, technologi-
seen a shrinking in payments to farmers after cal changes reduced the demand for land and
adjustment for the effects of population and labor by 61% and 60%, respectively.
income growth and other demand-side factors This is only a partial answer to the ques-
on the demand for U.S. farm output. Even if, tion, since we have not taken into account the
as I suspect, the elasticity of demand for U.S. effect of less-costly production on increasing
farm output is closer to –2.0 than –1.0, U.S. the total scale of production, and thus the de-
farmers as a group may have incurred losses mand for farmer-supplied inputs, when sup-
as a result of their adoption of largely land- ply shifts down—here is where the output
and labor-saving agricultural innovations.13 demand elasticity comes in. This output ex-
The estimated cost functions, presented pansion effect may be sufficient to more than
earlier, may be informative in this context. offset the effects of input savings from neu-
tral and biased changes in technology if de-
mand is sufficiently elastic. However, demand
would have to be significantly elastic
11
I opted to interpret the factor-saving technological change (g > 1:2) for the estimated 62% reduction in
as applying to inputs supplied by farmers since major innovations
have involved labor-saving and land-saving technologies, and
the unit cost of production to result in a large
land and labor are the principal inputs supplied by farmers. R2 enough expansion in consumption and thus
corresponds to net farm income, in that it represents the remain- total factor demand to offset a 75%
der after payments for purchased inputs, which goes to pay for
inputs supplied by farmers. The corresponding net welfare mea- innovation-induced reduction in demand for
sure would be producer surplus measured off the supply function land and labor, and leave farm incomes unaf-
for inputs supplied by farmers, X2, which will increase with fected. Demand would have to be much more
increases in R2 in this analysis.
12
To see why, suppose the demand elasticity is constant and ex- elastic for farmers to reap an appreciable
actly unitary. In this case, the total expenditure on all inputs will be share of the total national benefits.15
constant, regardless of their productivity and total production, and
consequently farmers will necessarily lose when technological
Evidence on changes in farm income—
changes cause a reduction in the factor share of the inputs they sup- from the balance sheet of the farming
ply. In contrast, if demand is inelastic, total expenditure on all inputs
will shrink when average and marginal costs fall, exacerbating the
effect of factor-biased technological changes on reducing the de-
14
mand for land and labor supplied by farmers. Only when demand is Some types of innovations, such as higher-yielding varieties
elastic can farmers expect to benefit from factor-neutral changes in and fertilizers, can be seen as distinctly “land-saving” while
farming technology, and demand has to be even more elastic if some, such as mechanical harvesters and bulk handling machin-
farmers are to benefit from factor-biased changes in technology that ery, are more clearly “labor-saving.” Some others, however, such
save the land and labor that they supply. as the tractor, saved both human labor and animal power, and in
13
Matt Andersen and I estimated the elasticity of demand for doing so also saved much land used to grow feed for horses and
U.S. aggregate farm output as –1.83 using the InSTePP indexes of mules (see, e.g., Gardner 2002; Olmstead and Rhode 2001).
15
prices and quantities for 1949–2002, in a price-dependent specifi- Another issue is distribution of producer benefits among
cation of demand (see Alston 2006, Appendix B). This estimate is producers. Even if we can be assured that producers as a whole
similar to estimates we derived using a synthetic approach, based would benefit, those who do not adopt the new technology will
on underlying commodity-specific elasticities of supply and de- not gain and, in fact, will be made worse off if the adoption by
mand, and national shares of global production and consumption. others leads to output price reductions.
120 50%
100
40%
80
30%
60
20%
40
NFI Share of VASP
Net Farm Income 10%
20
0 0%
1929 1939 1949 1959 1969 1979 1989 1999 2009
sector—reinforces the idea that farming inno- U.S. Net Farm Income (NFI) in real (2009)
vations have caused reductions in total in- dollars for the years 1929–2017. It also shows
come to farmers as a group. In its “Farm the Value of Agricultural Sector Production
Income and Wealth Statistics” the USDA- (VASP) and NFI as a percentage of VASP.
ERS publishes annual estimates of “Net In 1929, on the eve of the Great
Farm Income,” a measure that corresponds Depression, U.S. NFI was $62 billion. After
reasonably closely to the concept of income plummeting to $27 billion in 1932, it
accruing to farmers, from farming, contem-
plated here.16 Figure 2 shows the time path of
which the data were available I created an alternative measure:
“Returns to Operators” plus “Net Returns to Landlords,” which
includes land rent accruing to both farm operators and others.
16
This measure includes land rent accruing to farm operators This measure tracks Net Farm Income closely, but exceeds it by
who rent land to other farm operators, but does not include rent about 8% in recent years, with the difference representing
accruing to other landlords. For a subset of the time period for mainly land rent accruing to others (i.e., not farm operators).
fluctuated between $35 billion and $55 billion This finding refers to functional income dis-
until the United States entered World War tribution, at the level of the sector, rather
II, when it promptly doubled, reflecting the than personal income distribution (see, e.g.,
wartime boom in demand for farm output. Gardner 2002). Certainly, for those that re-
The 1940s saw NFI fluctuating in the range of main in farming, average income per farmer
$120 billion, with an historical peak in 1948 at is much greater today than it was 80 or even
almost $130 billion. During this 20-year pe- 40 years ago, but they are many fewer and
riod (1929–1948), land in farms continued to the effect of the reduction in numbers of
grow but average farm size grew even faster farmers outweighs the effect of the increase
(see Alston et al. 2010a, figure 2-5); mean- in income per farmer, such that total pay-
while three-quarters of the remaining 20 mil- ments for land, labor, and other inputs sup-
lion horses and mules were eliminated from plied by farmers has shrunk (see appendix
farm production, and much human labor was table A-1). This is relevant to the question,
also saved as farmers increasingly adopted looking forward: “Would today’s farmers as a
tractors, combines, and other machines (see group expect to benefit from investments in
Alston et al. 2010a, figure 3-1; Olmstead and agricultural R&D?”
Rhode 2001). The 1950s saw NFI declining This discussion of benefits accruing to U.S.
rapidly to return in 1959 to the 1929 value of farmers as a group has abstracted from im-
$62 billion. Since then, the trend in NFI has portant elements of the question of who ben-
been fairly flat, with notable fluctuations in- efits from agricultural R&D, which can be
cluding the OPEC-induced commodities boom considered in terms of the definition of the
of the early 1970s, followed by the bust of the relevant counterfactual. Crucially, when we
early 1980s, and another oil-induced boom- hold the demand facing U.S. farmers constant
bust sequence in the most recent decade. in our analysis, we are implicitly assuming
Remarkably, the estimate of NFI for 2016 that producers in other countries are not able
is $61 billion, slightly less than the 1929 value to adopt the technology in question.
of $62 billion, and it will be less again in 2017 Consequently, producers in the rest-of-the-
even though during those nine decades, total world (ROW) must lose when U.S. pro-
demand for U.S. farm output grew consider- ducers, facing a downward-sloping demand,
ably, reflecting an increase in total U.S. popu- adopt cost-reducing innovations and drive
lation from 122 million to over 320 million down the world price. Globally, the demand
combined with an increase in per capita in- for farm products is highly inelastic, even
come from $8,669 to $51,516 (in real 2009 when it is elastic for producers in any particu-
dollars; Johnson and Williamson 2017), aug- lar country.
mented by a significant new demand for bio- As illustrated in figure 3, when a subset of
fuels. Indeed, the value of sectoral output the world’s producers adopt a cost-saving in-
grew from $140 billion in 1929 to $363 billion novation, some (and potentially more than
in 2016 (in real 2009 dollars), but NFI as a all) of their benefits are necessarily at the ex-
share of VASP shrank from 50% in the 1940s pense of ROW producers (see, e.g., Edwards
to around 20% in recent years (see figure 2). and Freebairn 1982, 1984): agricultural inno-
It seems inescapable that the agricultural vation led by the United States could con-
innovations that made food much more abun- ceivably have an immiserizing effect on
dant and cheaper for consumers did so to farmers globally, even if it benefits American
some extent at the expense of farmers as a farmers. Likewise, an agricultural treadmill,
whole—more than offsetting the effects of as first described by Cochrane (1958, 1993)
growth in demand for output from the sector. and revisited by Levins and Cochrane (1996),
This finding is reinforced when we pay atten- might be evident as a global market phenom-
tion to the details of the timing. Specifically, enon even if agricultural innovation yields
the periods of the most rapid decline (or net benefits to farmers taking a narrow na-
slowest growth) in NFI seem to coincide with tional perspective. The same thinking applies
the periods of most rapid increase in farm if the producers not adopting the innovation
productivity—the 1940s to 1980s, especially in question are producers of other products
1950–1980, as identified by Andersen et al. that are substitutes in consumption: benefits
(2018)—consistent with the hypothesis that from innovation come to some extent at the
agricultural innovations have reduced net expense of competing farmers who are slow
incomes for U.S. farmers as a group. to adopt or never adopt the innovation.
Benefits to
Consumers A+B=(C+D)+B +
Home Producers E–C ?
$/Unit ROW Producers –B –
Total Producers E–(B+C) ?
S1
A B C D
E Excess
World
Demand Demand
growth in agricultural productivity. For ex- Andersen, and Pardey 2015; Andersen et al.
ample, using their preferred econometric 2017).18
model estimated using data for 1949–2002, The results were broadly consistent across
Alston et al. (2010a) projected MFP growth a variety of procedures (most transparently in
under two scenarios. In the “optimistic tests for constant growth rates using cubic
scenario,” R&D spending in each state and polynomial logarithmic trend models) and a
federally would grow after 2002 at the respec- variety of data sets (including national or
tive average rate for the period 1949–2002; in state-level data from either USDA or
the “pessimistic scenario,” R&D spending InSTePP). These findings support the view
would grow after 2002 at the average rate for that U.S. farm productivity growth has
the period 1990–2002. In the pessimistic slowed in recent decades. But more than that,
projection, MFP grows slowly and at a dimin- they also suggest that this slowdown came af-
ishing rate (averaging 0.52% per year ter a period of unusually rapid productivity
2003–2050), perpetuating the productivity growth in the middle of the full sample pe-
slowdown the authors identified as beginning riod, 1910–2007, with slower rates both in the
in the 1990s. earlier decades (i.e., 1910–1930) and more re-
In fact, public spending on agricultural cently (1990–2007). Alston, Andersen, and
and food R&D since 2002 has grown even Pardey (2015) conjectured that a wave of
more slowly than allowed for in that pessi- technological progress through the middle of
mistic projection. According to Pardey et al. the twentieth century—reflecting the progres-
(2016), between 2000 and 2011 inflation- sive adoption of various mechanical innova-
adjusted spending on public agricultural and tions, improved crop varieties and animal
food R&D increased only slightly (from 4.34 breeds, synthetic fertilizers and other
to 4.40 billions of 2009 PPP dollars, 1.4% in chemicals, each in a decades long process—
total over 11 years, or 0.12% per year). This contributed to an extended surge of faster-
trend reflects a decline in real public spend- than-normal productivity growth throughout
ing since 2009, which has continued in the the third quarter of the century, and a subse-
years since, and the productivity implica- quent slowdown that has extended into the
tions may be more serious given the drift in present era. These authors further speculated
emphasis away from farm productivity- that a one-time surge in productivity, like
oriented R&D within the total portfolio this, could be an inherent feature of the
(private spending on agricultural and food economics of the agricultural transition,
R&D has grown more rapidly, but a large the essential feature of which is to shift the
share is devoted to food R&D, and it is pro- majority of the farmers and their families
prietary, with different implications for cost out of agriculture—a one-time change.
savings on farms compared with public International evidence might shed some light
R&D). on this question.
The trends in public agricultural R&D
spending would give us cause to anticipate
and look for evidence of a slowdown in farm International Agricultural Productivity
productivity growth. But the long lags and Patterns
other complexities of the stock-flow relation-
ships between investments in R&D and Most countries do not have agricultural
effects on productivity, combined with the statistics nearly as good as those for the
complicated history of spending patterns, United States prepared by USDA-ERS and
make it difficult to pin down the likely timing InSTePP. Various studies of agricultural pro-
of a slowdown in MFP attributable to a slow- ductivity patterns around the world have re-
down in agricultural R&D spending. In an lied instead on FAOSTAT data, which are
initial foray on this issue, using data for 1949– readily available for many countries and
2002 we reported evidence of a productivity many years (FAOSTAT 2017). Using these
slowdown since 1990 (Alston et al. 2010a).
More recently, using data for 1910–2007, we
have developed more comprehensive evi- 18
These findings regarding the existence, timing, and extent
dence indicative of a slowdown in U.S. farm of a slowdown in U.S. farm productivity growth have been con-
productivity beginning much earlier, after a tested, mainly by economists working at the USDA-ERS. See,
e.g., Ball, Wang, and Nehring (2010); Wang (2010); Ball,
surge in MFP growth that peaked in the third Schimmelpfennig, and Wang (2013); Wang et al. (2015a, 2015b);
quarter of the twentieth century (Alston, Fuglie et al. (2017).
resources primarily, Keith Fuglie has com- In table 2, the regression results are similar
piled measures of inputs, outputs, and TFP when the models are applied to ten-year
for countries, regions, and the world as a averages of productivity growth rates by dec-
whole, in a useful form (see USDA-ERS ades, or to the (fairly volatile) annual meas-
2016). Studies using these or similar data have ures, and they are comparable (though
generally rejected the productivity slowdown different in a way that can be rationalized)
hypothesis, and more often have reported an across the alternative productivity measures.
acceleration of productivity—especially but In every case, the estimated models do not
not exclusively in developing countries.19 account for very much of the total variation
Whether this is right matters for under- in productivity growth rates, but the esti-
standing the prospects for the world food mated coefficients on the linear and quadratic
equation and other important determinants income terms are statistically significantly dif-
of well-being around the world. In what fol- ferent from zero. The maximum predicted
lows I first use these data to explore the evi- growth rate of TFP is reached when per cap-
dence from the cross-section related to the ita GDP is $1,879 (column 1 in table 2) or
conjecture that agricultural productivity $1,858 (column 2).21 Other evidence suggests
growth surges during the transition from an that the maximum growth rates are likely
agrarian to an industrial economy. Next, I re- to occur at higher per capita income. For ex-
visit the evidence from these versus other ample, in 2005 PPP dollars, U.S. per capita
data on the U.S. productivity slowdown and GDP was $18,830 in 1966, the year of maxi-
draw inferences for what we can say with con- mum MFP growth from Andersen et al.
fidence about international agricultural pro- (2018); in 2014, GDP per capita in China was
ductivity patterns. $10,890, but it was close to $1,800 in 1993.
The question of whether the agricultural These preliminary results using the interna-
transition entails a surge in agricultural pro- tional data are broadly consistent with the
ductivity is large and serious, and difficult, results from Andersen et al. (2018) using the
and at this stage I can only offer sketchy evi- U.S. state and national data, somewhat sup-
dence from preliminary analysis. Three alter- portive of the idea that the U.S. productivity
native measures of national agricultural slowdown may be reflective of a more general
productivity growth (G)—including TFP and phenomenon associated with the agricultural
partial factor productivities for land and transition and not simply an idiosyncratic re-
labor—were each regressed against a measure sult reflecting the path of U.S. agricultural
of the stage of economic development (I ¼ R&D spending—though both processes
GDP/capita in constant, 2005, international could be in play. This question is the subject
dollars), using annual data for a panel of 147 of continuing work from which we hope to
countries (or regional aggregates) over the derive more definitive results.
years 1961–2014.20 The models, which include One cause for reservations is concern over
fixed effects for year, t and country, i, are qua- the quality of the international TFP meas-
dratic forms in the logarithms of I, as follows: ures. As noted at the outset, “productivity” is
always an unexplained residual. Schultz and
ð3Þ Gi;t ¼ a0 þ a1 ln Ii;t þ a2 ln2 Ii;t þ li;t : Griliches identified elements of that residual
that could be reduced in various ways, and
significant progress in that direction has been
made by both the USDA-ERS and InSTePP
in their state and national U.S. agricultural
productivity data series. However, in spite of
19
See, for example, Fuglie (2008, 2010a, 2010b, 2012, 2015), Keith Fuglie’s heroic efforts to do as much as
various chapters in the volume edited by Fuglie, Wang, and Ball
(2012), the award-winning article by Trinidade and Fulginiti can be done with the fundamental data
(2015), and OECD (2016).
20
resources that are available, the USDA-ERS
The purpose of these estimations is not to attempt to estab- international TFP measures leave a great
lish “causality” between income per capita (as a measure of the
stage of economic development) and rates of agricultural produc-
tivity growth, both of which are jointly endogenous in the process
of economic transition from an agrarian to industrial and eventu-
ally to a post-industrial economy; nor is the purpose to “account
21
for” productivity growth, which would require a much more elab- The maximum predicted land productivity growth rate is
orate model. Rather, it is simply to ask whether the international reached later in the process when per capita GDP is $2,827 (col-
cross-sectional data exhibit patterns consistent with the conjec- umn 3) or $3,726 (column 4); the maximum predicted labor pro-
ture that the process of transition entails a surge in agricultural ductivity growth rate is reached when per capita GDP is $3,746
productivity growth. (in columns 5) or $3,292 (in column 6).
deal still in the residual—in particular for the per year over the longer period, 1961–2012
world’s poorest countries.22 (Figure 2.13).”
This aspect of these and other productivity In other words, by these measures, and as
measures based on FAOSTAT data is not al- Figure 2.13 illustrates, U.S. agricultural TFP
ways acknowledged, and even when it is ac- growth accelerated since 2003. In the next
knowledged, the implications are not always paragraph, however, the text acknowledges
appropriately highlighted by users of the that the “. . .international TFP comparisons
data. A recent publication by the OECD are hampered by data limitations. In particu-
(2016) illustrates this point; it presents a fig- lar, they only crudely account for certain im-
ure, based on the USDA international agri- portant inputs like pesticides and contract
cultural productivity data, showing annual services,” (OECD 2016), and points towards
agricultural TFP growth rates for selected the “more comprehensive” domestic produc-
OECD countries, 1961–2012 and 2003–2012. tivity accounts produced by USDA-ERS. But
Immediately prior to that figure, the text the reader is not cautioned to ignore (or even
(OECD 2016) reports: “U.S. agricultural pro- discount) the apparent evidence of
ductivity has experienced sustained high rates Figure 2.13 or the suggestion that farm pro-
of growth for decades. Using estimates from the ductivity growth has accelerated in the
ERS International Agricultural Productivity United States, even though the more reliable
accounts, Total Factor Productivity (TFP) evidence from national accounts suggests
growth in the United States averaged otherwise.23
2.1% per year over 2003–12 and 1.47% Table 3 compares measures of U.S. agricul-
tural productivity growth from three sources:
InSTePP, the USDA-ERS national accounts,
22
and the USDA-ERS international accounts.
Alston, Babcock, and Pardey (2010b) discuss some of the In each case the entries are computed either
data and measurement problems in reviewing a chapter by
Fuglie (2010b) in a book they commissioned and edited. These by regressing the productivity index in
authors conclude that “Fuglie’s estimates are the only available
estimates of agricultural TFP growth for many countries of the
world in the recent period. Even so, they should be used care-
23
fully, given the many constraints that data and measurement re- The same figure (OECD 2016) also suggests a significant
alities and choices place on generating accurate estimates, and productivity acceleration in Australia whereas various analyses
especially in relation to the question of a slowdown in productiv- of national data sources by economists at ABARES and else-
ity given that we have little basis for assessing their accuracy for where have reported a significant slowdown (see, e.g., Sheng,
that purpose.” Mullen, and Zhao 2011).
logarithms against a linear time trend the USDA “National” accounts data, it
(denoted “OLS”) or simply as the average of became evident later, after an acceleration in
annual growth rates (denoted “Mean”). In the 1980s.
principle, these measures are meant to repre- The upshot of this comparison is first to
sent the same concept, and the magnitudes confirm that the USDA “International” agri-
should be similar; in practice, the differences cultural productivity data could be highly
are striking. Consider, first, the last row of misleading as an indicator of U.S. agricultural
entries, representing average annual growth productivity patterns, compared with either
rates for the entire period (1961–2007) for of the available alternatives that are based on
which data are available from all three sour- a more complete accounting of the inputs, es-
ces. Both measures based on the USDA pecially capital. In addition, although we do
“International” data (columns 5 and 6), are not have available alternative measures
similar, at around 1.4% per year, implying based on national accounts that we can use to
total growth over 46 years of just 90%. The check directly for most other countries, it
two alternative sets of estimates based on seems reasonable to suppose that the mea-
national accounts both indicate much faster surement problems that bedevil the estimates
average annual growth rates: 1.6% or 1.8% for the United States would be at least as bad
from USDA-ERS “National” data (columns for other countries for which the FAOSTAT
3 and 4) and around 1.8% from InSTePP measures of inputs and outputs would be sub-
(columns 1 and 2), the latter implying total ject to even greater deficiencies.24
growth over 46 years of 127%, 1.40 times A part of the problem is that these meas-
the total growth implied by the USDA ures based on the FAOSTAT data have gone
“International” data. The patterns over time only a short way down the path required to
are also quite dissimilar. The USDA fully account for (and eliminate) the produc-
“International” data show a continuing accel- tivity residual—as proposed by Schultz and
eration in productivity growth from less than Griliches. Another part is that the underlying
1.2% per year before 1981 to 1.6% per year FAOSTAT data on inputs and outputs may
since 1981, and more than 1.7% per year
since 1991. Both the USDA “National”
accounts data and InSTePP show a slowdown
24
since 1991, compared with the previous pe- Josef Schmidhuber has advised me that he and colleagues
are engaged in developing an alternative set of measures of na-
riod, but in the InSTePP data that slowdown tional agricultural productivity growth rates, based on national
was already evident after 1981, whereas in accounts.
26
The socially productive work of creating and validating
25
Leontief (1971) justly complained: “. . .in all too many data resources is undervalued generally, if not derided. As Alice
instances, sophisticated statistical analysis is performed on a set Rivlin (1975) stated in her Ely lecture: “Disdain for data collec-
of data whose exact meaning and validity are unknown to the au- tion is built into the value and reward structure of our profession.
thor, or rather so well known to him that, at the very end he Ingenious efforts to tease bits of evidence from unsuitable data
warns the reader not to take the material conclusions of the en- are much applauded; designing instruments for collecting more
tire ‘exercise’ seriously.” appropriate information is generally considered hack work.”
USDA national accounts and InSTePP, the is close to the value in 1929 even though in
data from the USDA-ERS international the nine decades since, the U.S. population
accounts indicate much slower overall U.S. and per capita income and demand for farm
agricultural productivity growth since 1961. output have increased manifold.
Further, the USDA international accounts in- Drawing these elements together, it seems
dicate a huge acceleration, whereas both the inescapable that American farmers as a
USDA “National” accounts data and group, supplying land, labor, and managerial
InSTePP show a slowdown since 1991. inputs used in agriculture, have been made
Work remains to be done to resolve the de- worse off by the changes in technology that
bate over how best to measure the price and transformed America agriculture—which in
quantity of capital—the most important no way implies that American farmers can af-
source of difference between InSTePP the ford to cease to innovate. The comparison
USDA-ERS national accounts. I am inclined here is with a counterfactual scenario in
to favor the InSTePP measures for the rea- which (implicitly) agricultural technology
sons presented by Andersen, Alston, and could have been frozen not only in the
Pardey (2011, 2012), buttressed by the cost United States but also everywhere else. It
function analysis presented here. But clearly I would be even worse for American farmers if
cannot be entirely neutral on this question, they were to cease unilaterally to innovate
having some personal professional investment while farmers in the rest of the world did not.
in the InSTePP data. Other analysis here A related question concerns distribution
raises some serious concerns over the inter- among farmers. Even if U.S. farmers do stand
pretation and use of productivity measures in to benefit from technological changes of the
the USDA-ERS international accounts, which types experienced since WWII, given that
is especially bothersome because no compara- worldwide demand for farm output is inelas-
ble alternative exists for many countries, or tic, it follows that farmers somewhere will
for making international comparisons. Even have been made worse off. The question “do
when we have grave concerns about the data, farmers benefit from agricultural R&D?” has
the temptation to use them nevertheless to to be asked carefully (with attention to which
conduct analysis can be overwhelming. At a famers and under what corollary conditions)
minimum, I would prescribe a very large grain if we want meaningful answers. Farmers and
of salt and much caution as we wait and hope farm lobby groups appear to be much more
for improved estimates to become available. interested in seeking government support for
In much of my own work I have argued subsidized crop insurance and the like rather
that farmers should expect to benefit from ag- than agricultural R&D; and perhaps they are
ricultural R&D, but it has been troubling not well-advised in so doing. A better under-
to have direct, compelling empirical evidence standing of the ultimate incidence of research
to support this view (e.g., Alston et al. 2009). benefits might help us do better at finding sol-
Here, I have tried seriously to answer that utions to the problem of underfunding of ag-
question for the first time, and the answer ricultural R&D.
came as a surprise. First, a fairly simple but
transparent theoretical analysis suggests that
when demand for the farm product is inelas- References
tic, farmers as a group can expect to lose
from factor-neutral technical change or from Acquaye, A.K.A., J.M. Alston, and P.G.
factor-biased technical change that saves Pardey. 2003. Post-War Productivity
inputs supplied by farmers. Second, both the Patterns in U.S. Agriculture: Influences
cost function estimates and informal evidence of Aggregation Procedures in a State-
(technology factoids) indicate that technolog- Level Analysis. American Journal of
ical change in U.S. agriculture has been land- Agricultural Economics 85 (1) : 59–80.
and labor-saving; and, even though the Alston, J.M. 2006. Benefits and Beneficiaries
demand for U.S. farm output is probably from U.S. Farm Subsidies. Paper 4 pre-
elastic, it might not be elastic enough to offset pared for the American Enterprise
these farmer-supplied-input-saving effects. Institute project, Agricultural Policy for
Third, U.S. net farm income (NFI) has 2007 Farm Bill and Beyond, led by B.L.
shrunk as a share of gross farm income to an Gardner and D.A. Sumner. Available at
extent such that, in real terms, total NFI has http://aic.ucdavis.edu/research/farmbill07/
been roughly constant since 1960. NFI today aeibriefs/20070515_alstonSubsidiesfinal.pdf.
———. 2009. Efficiency of Income Transfers and the Treatment of Interest Rates: An
to Farmers through Public Agricultural Error and a Correction. American Journal
Research—Theory and Evidence from of Agricultural Economics 94(5) : 1247–8.
the United States. American Journal of Andersen, M.A., J.M. Alston, P.G. Pardey,
Agricultural Economics 91 (5) : 1281–8. and A. Smith. 2018. A Century of U.S.
Alston, J.M., M.A. Andersen, J.S. James, and Farm Productivity Growth: A Surge Then
P.G. Pardey. 2010a. Persistence Pays: a Slowdown. American Journal of
U.S. Agricultural Productivity Growth Agricultural Economics. In press.
and the Benefits from Public R&D Ball, V.E., C.A.K. Lovell, H. Luu, and R.
Spending. New York: Springer. Nehring. 2004. Incorporating
———. 2011. The Economic Returns to U.S. Environmental Impacts in the
Public Agricultural Research. American Measurement of Agricultural Productivity
Journal of Agricultural Economics 93(5) : Growth. Journal of Agricultural and
1257–1277. Resource Economics 29 (3) : 436–60.
Alston, J.M., M.A. Andersen, and P.G. Ball, V.E., D. Schimmelpfennig, and S.L.
Pardey. 2015. The Rise and Fall of Wang. 2013. Is U.S. Agricultural
American Farm Productivity Growth. Productivity Growth Slowing. Applied
University of Minnesota Staff Paper Staff Economic Perspectives and Policy 35 (1) :
Paper P15-02, International Science and 435–50.
Technology Practice and Policy (InSTePP) Ball, V.E., S.L. Wang, and R. Nehring. 2010.
Center. St. Paul: University of Minnesota. Agricultural Productivity in the United
Alston, J.M., B.A. Babcock, and P.G. Pardey. States: Data Documentation and Methods.
2010b. Shifting Patterns of Global Washington DC: U.S. Department of
Agricultural Productivity: Synthesis and Agriculture, Economic Research Service.
Conclusion. In The Shifting Patterns of Ball, V.E., S.L. Wang, R. Nehring, and R.
Agricultural Production and Productivity Mosheim. 2016. Productivity and
Worldwide, ed. J.M. Alston, B.A. Economic Growth in U.S. Agriculture: A
Babcock and P.G. Pardey, 449–482. New Look. Applied Economic
Ames, IA: Center for Agricultural and Perspectives and Policy 38 (1) : 30–49.
Rural Development. Clancy, M., K.O. Fuglie, and P. Heisey. 2016.
Alston, J.M., G.W. Norton and P.G. Pardey. U.S. Agricultural R&D in an Era of
1995. Science under Scarcity: Principles Falling Public Funding. Amber Waves.
and Practice for Agricultural Research Available at https://www.ers.usda.gov/
Evaluation and Priority Setting. Ithaca: amber-waves/2016/november/us-agricul-
Cornell University Press. tural-rd-in-an-era-of-falling-public-funding/
Alston, J.M., and P.G. Pardey. 2014. Cochrane, W.W. 1958. Farm Prices: Myth
Agriculture in the Global Economy. and Reality. Minneapolis: University of
Journal of Economic Perspectives 28 (1) : Minnesota Press.
121–146. ———. 1993. The Development of American
———. 2017. Transforming Traditional Agriculture: A Historical Analysis.
Agriculture Redux. In The Oxford Second Edition, Minneapolis: University
Handbook of Structural Transformation, of Minnesota Press.
ed. C. Monga and J. Y. Lin. Oxford: Craig, B.J. and P.G. Pardey. 1996. Productivity
Oxford University Press (in process). Measurement in the Presence of Quality
Alston, J.M., P.G. Pardey, J.S. James, and Change. American Journal of Agricultural
M.A. Andersen. 2009. A Review of Economics 78 (5) : 1349–54.
Research on the Economics of Edwards, G.W., and J.W. Freebairn. 1982.
Agricultural R&D. Annual Review of The Social Benefits from an Increase in
Resource Economics 1 : 537–565. Productivity in Part of an Industry.
Andersen, M.A., J.M. Alston, and P. G. Review of Marketing and Agricultural
Pardey. 2011. Capital Services in U.S. Economics 50 : 193–210.
Agriculture: Concepts, Comparisons, and ———. 1984. The Gains from Research into
the Treatment of Interest Rates. Tradable Commodities. American Journal
American Journal of Agricultural of Agricultural Economics 66 (1) : 41–9.
Economics 93 (3): 714–34. FAOSTAT. 2017. A Database of the Food
———. 2012. Erratum—Capital Services in and Agriculture Organization of the
U.S. Agriculture: Concepts, Comparisons, United Nations (FAO).