The Productivity Plateau

Global agricultural productivity growth has fallen by roughly forty percent between the 2000s and the 2010s, and output growth is now the slowest in the six decades the data covers. This report separates the input contribution from the technology contribution and explains why the cost is not showing up in prices.

Measurement approach

For sixty years the direction of travel in agriculture was not in dispute. Output per unit of combined input rose, food got cheaper in real terms, and the land needed to feed an additional person fell. The most recent decade of data does not say that, and the change has happened quietly enough that most long-range planning assumptions in the food system have not been updated for it.

Total factor productivity is the right measure and it is not the one most people use. Yield per hectare can rise while productivity falls, if the extra output is bought with more fertiliser, more water and more machinery. TFP asks the harder question: holding the whole input bundle constant, how much more comes out.

USDA's Economic Research Service puts global agricultural TFP growth at 1.14 percent a year across 2011 to 2021, against 1.93 percent in the preceding decade. That is a fall of roughly forty percent in the rate at which the sector gets better at what it does.

Where the output is coming from instead

Output growth fell too, from 2.74 percent a year to 1.94 percent, the slowest decade in a series that starts in 1961 and averages 2.3 percent. But it fell by less than productivity did. The difference between the two is the input contribution, and it has widened.

When output growth exceeds productivity growth by a widening margin, the sector is meeting demand by using more rather than by using better. That has a price, and the price is not paid at the supermarket.

Those inputs are land, water, fertiliser and machinery. Each has its own cost curve and its own environmental constraint, and several of them are getting more expensive rather than less. A food system that leans harder on inputs each year is one whose cost base is increasingly exposed to energy and fertiliser markets rather than to agronomy.

The composition matters as well as the level. The slowdown is observed primarily in developing countries, where TFP growth roughly halved between the 2000s and the 2010s. That is where most of the additional demand to 2050 is expected to arise, which is an uncomfortable pairing.

Three contributing factors

Research intensity

Public agricultural research has the longest lag of any input and the clearest measured return. Work on the United States estimates that offsetting climate-induced productivity losses by 2050 would require sustained public research spending growth of 5.2 to 7.8 percent a year over 2021 to 2050, an additional 208 to 434 billion dollars across the period. The authors compare the scale to the research commitments that followed the two world wars. Nothing of that order is currently being committed.

Adoption lag

Growth in TFP implies that new technology or better management is raising average efficiency of input use. Where the gap between frontier practice and median practice widens, measured TFP stalls even though the frontier keeps moving. This is the component most amenable to commercial intervention, because it needs distribution and finance rather than new science.

Climate drag

The climate contribution is the least discussed and probably the largest. Econometric work combining weather effects with counterfactual climate scenarios estimates that anthropogenic climate change has already reduced global agricultural TFP by about 21 percent relative to where it would otherwise sit, with substantially larger effects in warmer regions. The same work finds global agriculture has become more sensitive to further warming, not less. That is the finding that should worry anyone extrapolating past adaptation.

Cost of the plateau

Not a price spike. Productivity sets the trend and weather sets the year, so a stalled trend does not announce itself through a bad harvest. It shows up three other ways.

• A thinner buffer. Each cycle, the stock available to absorb a genuinely bad year is smaller relative to consumption. The distribution of outcomes widens without the central case moving much.
• Input price beta. A system leaning on inputs transmits fertiliser and energy prices into food prices faster and more completely than one leaning on productivity.
• Land and water. Output growth met by extensification carries an environmental cost that is real, delayed and generally unpriced.

Conditions under which this would not hold

• A measurement artefact. If the input side of TFP is overstated, for instance because capital services are mismeasured as machinery becomes more capable, then true productivity growth is higher than reported. We think this explains part of a gap this size and not most of it.
• An adoption discontinuity. Precision application, gene editing and biologicals are all at a stage where measured contribution is small and potential is not. A step change in adoption across five years would show up as a TFP acceleration and would falsify the plateau reading.
• Reallocation rather than stagnation. Part of the global slowdown reflects composition: production shifting toward countries and crops with lower measured productivity. That is a real economic phenomenon but a different one from a technology stall.

We will revisit this when the next full productivity round is published. If the reading holds, the requirement side of the calculation will have to be revised upward, because the shortfall compounds.

Source register