Is technological progress in US agriculture slowing?

M. Clancy
Institute for Progress
United States

This essay reviews evidence that technological progress in US agriculture – here understood as the increasing efficiency by which inputs are transformed into outputs – is slowing, at least relative to the common benchmark of constant exponential growth. The case for a slowdown seems to hold whether measured with yields or more sophisticated methods, such as total factor productivity (TFP). The slowdown may stem from agriculture-specific factors, such as stagnating levels of research and development (R&D) through much of the late 20th century. It may also be influenced by broader factors, such as slowing technological progress in other domains and a general tendency for innovation to get harder.

Why focus on US agriculture? Surprisingly, perhaps, agriculture is a good subject for the study of long-run changes in technological progress. Four reasons are elaborated below.

First, any study of progress needs data over time. US agriculture is unique in providing quite good data over a long period.

Second, the United States is traditionally seen as operating on the technological frontier. US public funding for agricultural R&D is by far the largest in the world, accounting for 25% of publicly funded agricultural R&D in high-income countries (Heisey and Fuglie, 2018).

Third, economists usually think of “technology” as the processes that convert inputs into outputs. In agriculture, the nature of an output has not changed much, at least compared to sectors such as communication, transportation and manufacturing. Corn is corn and the way it was measured 150 years ago is not that different from today. For something like automobiles, where accounting for the changing nature of goods produced is important, counting the number produced in a year might be controversial. However, comparing the annual number of corn bushels produced in 1920 and 2020 is not controversial in the same way.

Fourth, agriculture is full of technological progress. Two of the most important inputs to agricultural production, over the long run, have been labour and land. The increase of annual corn production in the United States by more than sevenfold between 1948 and 2022 was accomplished without any significant increase in these two inputs. Land use has stayed roughly the same, while labour has fallen dramatically.

One can already derive a crude but common measure of technological progress in agriculture using these data on output and land, namely “yield” (e.g. bushels of corn per acre). In economics, “constant technological progress” is typically defined to mean constant exponential growth in the efficiency with which inputs are converted into outputs – e.g. yields increasing by 2% per year. Since agricultural output fluctuates a lot due to weather, this essay uses a long-run indicator of progress – the percentage change in a measure of technological efficiency over 20 years.

Figure 1 below plots average US corn yields on the left and the 20-year growth rate of those yields on the right. It shows a dramatic slowdown in the rate of technological progress by this measure (though progress in yield growth today remains significantly higher than the stagnation that prevailed prior to 1940).

Measuring technological progress in agriculture with yields has the advantage of not depending much on theoretical constructs. This simple set of data shows a sharp slowdown in yield growth. This is mostly (but not entirely) because growth has been constant in absolute terms (growing by about 38.5 bushels every 20 years between 1970 and 2021), and therefore must be decreasing in exponential terms.

However, yield is also an unsatisfying measure of technological progress because it misses so many aspects of agricultural production. A way is needed to account for the varied agricultural products (not just corn); the diversity of inputs used; and other factors that affect agricultural production and that might have changed, such as the climate.

Ciliberto, Moschini and Perry (2019) nicely illustrate some of the ways yield is an unsatisfying measure of technological progress. A prominent technological innovation in US agriculture has been the genetic modification of crops. For example, in 2014, nearly 90% of US corn was genetically modified with a gene that confers resistance to the chemical glyphosate, a key pesticide. This modification makes it easier and less costly to control weeds. Another common genetic modification confers resistance to various species of corn rootworms, which reduces the need for insecticides. Both innovations are only indirectly connected to yield but are highly valued by farmers. By comparing demand for these seeds at various prices, relative to comparable seeds without genetic modification, Ciliberto, Moschini and Perry (2019) estimate farmers are willing to pay an extra USD 5-17 per acre for one of these traits.

To capture these kinds of improvements, a measure of technological progress needs to be created. It must account for progress that keeps yield the same but reduces farm labour (for example, related to weed control), or use of other inputs (such as insecticides). Moreover, to study technological progress in the agricultural sector as a whole literally requires a way to compare apples and oranges and everything else farmers grow. Fortunately, economists have many theories for aggregating baskets of goods over time by using data on spending and price changes. Using these techniques, from 1949 to 2017 (the last year for available data), growth in total agricultural output in the United States did not look that different from the trends seen in corn production: total US agricultural output, in inflation-adjusted terms, nearly tripled.

Calculating what happened on the input side is trickier. Technological progress in agriculture has involved waves of new technologies, which are gradually adopted by a larger and larger share of farmers (Pardey and Alston, 2021). It is a lot harder to measure these new kinds of inputs because they come in such a variety of forms (fertiliser, pesticides, tractors, silos, etc.).

Moreover, the quality of these inputs evolves over time due to technological progress. For example, one cannot just count the gallons of pesticides used over time since the nature of that pesticide changes. Instead, economists at the US Department of Agriculture (USDA) attempt to adjust for the changing quality of pesticides to measure the farm sector’s use of some kind of “constant-quality” pesticide. Similar adjustments are made for the other inputs used in agricultural production.

Those inputs are then aggregated in a way that weights their share of the value of intermediate inputs. This reveals at least two places where agriculture has increased, rather than decreased, its use of inputs. According to USDA measures, quality-adjusted fertiliser use more than tripled between 1948 and 2017, while quality-adjusted pesticide use increased more than fiftyfold over the same period (from a low base).

To some extent, the invention of cost-effective fertiliser and pesticides is itself a story of technological improvement since effective versions of these inputs were themselves inventions. However, taking the existence of fertiliser and pesticide for granted, if more intense use drives the rise in yields noted above, it calls into question the story of technological progress observed so far. Maybe increased output in agriculture arose simply from using more (non-land) inputs, not from any increased ability to get more from less.

It will take several steps to continue measuring technological progress as the ability to translate ever-fewer inputs into ever-more outputs. First, the basket of different inputs, use of which grew in some cases and shrank in others, should be aggregated into a single index of inputs. Second, the basket of different outputs should be aggregated into a single index of outputs.

Dividing an index of all agricultural outputs by an index of all agricultural inputs is analogous to how yield was one output (corn) divided by one input (land). It gives a more comprehensive measure of technological progress – TFP (sometimes called multi-factor productivity), which measures the capacity to produce more with less.

Economic theory provides a framework for constructing these estimates. Given data on all the different inputs used in production, under some standard assumptions, it is possible to weight indices by their share of total costs and add them up to generate an aggregate index measure. Many methodological choices go into constructing these data series, however. Ultimately, there is no simple and objective metric to check that these choices are right. These measures are ultimately theoretical constructs.

However, US agriculture is fortunate in having two different teams of economists – one composed of government economists affiliated with the USDA, the other led by academics affiliated with the International Science and Technology Practice and Policy group (InSTEPP). Together, they have tackled this measurement challenge using somewhat different methodologies (see Fuglie et al., 2017 for a discussion). The extent to which the two different approaches converge on the same findings gives some confidence in the results.

The different estimates can be seen in Figure 2, which plots estimates from InSTEPP and the USDA for the 20-year growth rate of agricultural TFP, and for comparison, the 20-year growth rate of TFP for the entire US economy. Among economists, it is not controversial to assert a slowdown in US TFP growth since the 1970s.

Figure 2 clearly indicates that TFP growth in agriculture has slowed. Notably, the magnitude of the decline in TFP is similar across all three series, though there is some considerable disagreement between InSTEPP and the USDA in the 1980s. However, setting this decade aside, each series hangs around the 40-50% range (total over each preceding 20-year period) in the first half of the series. Each series then ends in the 20-30% range (total over the preceding 20-year period).

There is some significant disagreement about when this slowdown began. InSTEPP showed declines beginning in the 1990s, while USDA placed them in the 2000s. In any event, the two different TFP estimates suggest a slowdown comparable to the slowdown in the entire US economy in the late 20th century and early 21st but with a later onset.

On the other hand, perhaps using TFP as a measure of technological progress is misleading. Like yields, TFP misses important contemporaneous factors that affected how much output was produced from agricultural inputs. For example, a worsening climate or pest burden could reduce agricultural output from a given set of inputs. This would also reduce measure of TFP. However, in this case, the decline in TFP growth would not be caused by a slowdown in technological progress. As discussed in Clancy (2021), these considerations do not appear to alter the core claim made above that the growth rate of agricultural TFP slowed in the late 20th century and early 21st century.

Another important dimension of agricultural production not typically included in TFP relates to the environmental sustainability of agricultural production. Unsustainable forms of production, which drew down natural resources (for example, soil quality) or produced harmful pollutants, may have shifted towards more sustainable practices. Since TFP does not typically measure use of natural resources and production of pollutants, a move to greater sustainability could result in lower TFP growth. Again, this would not imply any actual reduction in the rate of technological progress.

Expanding the scope of TFP to include these non-marketed inputs and outputs is an active area of research in agricultural economics (e.g. Bureau and Antón, 2022). However, long-run data are needed to determine if these measures also indicate a slowdown in technological progress.

If the available data are correct, and technological progress has indeed slowed, the next question is: why might this have occurred?

To begin with, note the following coincidence: over the entire 20th century the growth rate of agricultural TFP (as estimated by InSTEPP) has followed the same basic trajectory as the growth rate of non-farm TFP. However, it has a multi-decade lag. Pardey and Alston (2021) argue TFP growth in the non-farm sector increased through the 1940s, then declined through 1990. Conversely, agricultural TFP rose unevenly through the 1980s and then fell.

There are a few reasons to think this rise and fall, separated by decades, might not be a coincidence. Agricultural economists generally agree that new and better technologies have been the fundamental driver of US agricultural productivity gains. These technologies themselves emerged from earlier R&D (in some cases, many decades earlier). For reasons described below, in the short term (which can still be pretty long), specifically agricultural R&D might be most useful. Conversely, non-agricultural R&D is likely to be useful in the longer run.

Figure 3 plots US agricultural R&D over roughly the same time period as Figure 2.

At first glance, this does not look much like the TFP growth series. Rather than staying constant and then declining, R&D rises, stagnates and then rises (though its composition changes a lot). However, movements in contemporaneous R&D should not be expected to match those in agricultural TFP growth for two reasons. First, R&D only affects productivity with a lag. Second, falling TFP growth alongside stagnant R&D would be expected if the productivity of R&D is declining over time.

With respect to lags, there is a long literature in agricultural economics that attempts to pin down the correct time lag between R&D and productivity. Researchers look for correlations between R&D spending and productivity, both at the national and state level. Baldos et al. (2019), for example, adopts a Bayesian methodology to explicitly model uncertainty about this process. It finds a peak impact of R&D on productivity around 20 years. This is consistent with much other literature in this field, which tends to find a multi-decade gap between the onset of R&D and its impact on productivity.

With respect to declining research productivity, R&D effort is defined as current R&D expenditure levels divided by the prevailing wage for scientific labour. Constant levels of R&D effort yield diminishing proportional increases in any given R&D target across many fields. Thus, a time path of agricultural R&D that initially grows, then flattens, then grows again, might well generate a path of TFP growth that is constant, then declines, then remains constant.

Given at least a 20-year lag between R&D and its effects on productivity, one might then predict agricultural TFP growth to remain constant for at least 20 years after agricultural R&D began to stagnate in 1980. Beginning sometime around 2000, agricultural TFP growth might then begin to decline as stagnating agricultural R&D catches up to it with a delay. For the period illustrated in Figure 3, this actually fits the time path for TFP growth (as measured by the USDA). However, the decline in TFP growth as measured by InSTEPP appears to occur too early for changes in the productivity of R&D to be the whole story (though it could certainly exacerbate declines in later years).

Still, agricultural R&D itself builds significantly on R&D elsewhere. Clancy et al. (2021) measure the extent to which patented agricultural technologies rely on knowledge developed outside of agriculture. This includes, for example, farm machinery, fertilisers, pesticides, veterinary medicine, plant varieties and plant breeding techniques. They measure this by looking at the share of citations made by patents for agricultural technology to other patents and academic journals. They also look at the share of novel technological concepts in the text of agricultural patents that originate in other non-agricultural patents. They find, in most cases, that most ideas used in patented agricultural technologies do not originate in agricultural research.

Their findings suggest a model where technological developments in the non-farm economy seed promising avenues for agricultural research to adapt. Indeed, Gordon (2017) argues the surge in TFP growth for the non-farm economy is a crude measure of widespread technological innovations in the US economy. Such innovation begins with the automobile, roads and electrification, and is later followed by a revolution in chemical technology. To a large extent, the story of agricultural productivity in the 20th century is the story of these economy-wide innovations gradually diffusing out from cities into rural America. At that point, they are adapted for agricultural use via agricultural R&D and then diffuse across farms over the course of decades.

However, as Pardey and Alston (2021) show, changes in productivity on the farm only occur after the uptake of these innovations. They argue a process of economic reorganisation is necessary to reap the benefits of these new technologies. For example, if technology disproportionately improves the productivity of larger farms, the full benefits of technology will only be realised after a protracted period of farm consolidation (which is, in fact, observed). Furthermore, as Costinot and Donaldson (2016) show, better infrastructure may allow farms to sell on distant markets. If this happens, farms can specialise in growing crops for which they have a comparative advantage rather than a diverse set of crops to satisfy local market demand. This adjustment also takes time.

It does appear that technological progress in US agriculture has begun to slow, at least compared to a benchmark of constant exponential growth. This is visible with crude but robust measures like the growth in yields over time, but the result also holds true after incorporating changes in the mix and quality of inputs. Moreover, while this essay is focused on US agriculture, a slowdown in agricultural productivity growth is a global phenomenon. ERS/USDA has also produced an internationally comparable TFP database using a simplified methodology applicable across countries.1

As noted in Fuglie, Jelliffe, and Morgan (2021), these data show global productivity growth in agriculture fell from an average of 2% per year over the 2000s to 1.3% per year over the 2010s. Developing countries experienced much steeper declines. As in the United States, these declines may stem from slowing technological progress but also from non-technological factors such as climate change.

Moreover, TFP growth in countries further from the technological frontier is more likely to reflect the adoption of frontier technologies and practices, as well as the transition to efficient scales given new technologies, rather than the rate of frontier technological advance per se. Reviewing these factors, however, is beyond the scope of this essay.

In the United States, stagnating agricultural R&D in the late 20th century, in an environment where innovation gets harder, may well explain part of the country’s decline in agricultural productivity. However, at a deeper level, it may well be that slowing progress in agriculture is a long-delayed echo of a slowdown in innovation across the wider non-farm economy.


Baldos, U.L.C. et al. (2019), “R&D spending, knowledge capital, and agricultural productivity growth: A Bayesian approach”, American Journal of Agricultural Economics, Vol. 101/1, pp. 291-310,

Bureau, J. and J. Antón (2022), “Agricultural Total Factor Productivity and the environment: A guide to emerging best practices in measurement”, OECD Food, Agriculture and Fisheries Papers, No. 177, OECD Publishing, Paris,

Ciliberto, F., G. Moschini and E.D. Perry (2019), “Valuing product innovation: Genetically engineered varieties in US corn and soybeans”, RAND Journal of Economics, Vol. 50/3, pp. 615-644,

Clancy, M. (2021), “Is technological progress slowing? The case of American agriculture”, 24 November, New Things Under the Sun,

Clancy, M. et al. (2021), “The roots of agricultural innovation: Patent evidence of knowledge spillovers”, in Economics of Research and Innovation in Agriculture, P. Moser (ed.) University of Chicago Press.

Costinot, A. and D. Donaldson (2016), “How large are the gains from economic integration? Theory and evidence from U.S. agriculture, 1880-1997”, Working Paper, No. 22946, National Bureau of Economic Research, Washington, DC,

Federal Reserve Bank of San Francisco (2021), “Total Factor Productivity”, webpage, (accessed 28 November 2022).

Fuglie, K.O. et al. (2017), “Research, productivity, and output growth in U.S. Agriculture”, Journal of Agricultural and Applied Economics, Vol. 49/4, pp. 514-554,

Fuglie, K., J. Jelliffe, and S. Morgan (2021), “Slowing productivity reduces growth in global agricultural output”, 28 December, Amber Waves,

Gordon, R. (2017), The Rise and Fall of American Growth, Princeton University Press.

Heisey, P.W. and K.O. Fuglie (2018), “Agricultural research investment and policy reform in high-income countries”, Economic Research Report, No. 249, US Department of Agriculture, Economic Research Service, Washington, DC.

Pardey, P. and J. Alston (2021), “Unpacking the agricultural black box: The rise and fall of American farm productivity growth”, The Journal of Economic History, Vol. 81/1, pp. 114-155,

Metadata, Legal and Rights

This document, as well as any data and map included herein, are without prejudice to the status of or sovereignty over any territory, to the delimitation of international frontiers and boundaries and to the name of any territory, city or area. Extracts from publications may be subject to additional disclaimers, which are set out in the complete version of the publication, available at the link provided.

© OECD 2023

The use of this work, whether digital or print, is governed by the Terms and Conditions to be found at