Implications of Alternative Agricultural Productivity Growth Assumptions on Land Management, Greenhouse Gas Emissions, and Mitigation Potential
Justin S. Baker ([email protected]) is a research economist at RTI International. Brian C. Murray ([email protected]) is Director of Economic Analysis at the Nicholas Institute for Environmental Policy Solutions, Duke University. Bruce A. McCarl ([email protected]) is University Distinguished Professor at the Department of Agricultural Economics, Texas A&M University. Siyi Feng ([email protected]) is a research fellow at the Department of Agricultural and Food Economics, Queen's University of Belfast. Robert Johansson ([email protected]) is Deputy Chief Economist at the U.S. Department of Agriculture, Office of the Chief Economist.
This article was presented in an invited paper session at the 2012 AAEA annual meeting in Seattle, WA. The articles in these sessions are not subjected to the journal's standard refereeing process.
Future productivity growth in agriculture is necessary to satisfy rising food, fiber, and bioenergy demands, and to contribute to global environmental objectives, including greenhouse gas (GHG) mitigation. Conventional logic suggests that improved productivity per-unit area (e.g. from improved seed varieties of other scientific advances not controlled by the farmer) can spare additional agricultural land expansion, intensification on existing cropland, and land degradation (Burney, Davis, and Lobell 2010; Hertel 2011). Such changes can reduce net GHG emissions from the global agricultural and forestry system. However, this inference does not hold universally because of the complex relationship between exogenous productivity advancements, endogenous land management responses, and landscape heterogeneity. Consequently, the relationship between agricultural technology growth, land use, and GHG emissions is ill-defined in the literature, especially for coupled crop and livestock systems in regions with historically high levels of land use change.
Some research asserts that technology-driven yield improvements can increase deforestation and GHG emissions (Angelsen 2010). Supporting this hypothesis, Rudel et al. (2009) show that cultivated area has generally increased along with agricultural yields, often to the detriment of natural areas such as forests. However, these studies consider primarily localized effects, rather than investigating whether productivity improvements can be land-saving at higher scales of aggregation (national to global). Moreover, the studies do not explicitly consider agricultural expansion rates in the absence of such productivity improvements. Ceteris paribus, lower technological growth rates could stimulate more agricultural land expansion, but the effects of productivity improvements on land management decisions can be highly localized. For example, higher net returns under staus quo practices could lead to local intensification and land clearing if the localized yield improvements dominate any decline in market price that results from collective productivity improvements shifting out market supply. Yet such improvements can be land sparing globally if yield and productivity growth outpace demand growth. Recent studies have illustrated how productivity growth and management intensification can be “land sparing” at local and global levels (Choi et al. 2011; Burney, Davis, and Lobell 2010; Gockowski and Sonwa 2010; Shively 2001).
In high productivity systems in the developed world where crop and livestock operations are inextricably linked via input markets (such as in the United States), the story is less clear. On the one hand, technological change can lower net emissions by relaxing input use at the intensive margin. However, greater returns to crop production could cause land expansion, livestock herd expansion, or both. Technological advances in genetic breeding that allow for plant growth in suboptimal growing conditions could lead to additional agricultural land expansion in less productive areas. Furthermore, greater crop productivity could expand feeding opportunities in the livestock sector. Either case can raise emissions.
-
Does greater agricultural productivity increase or decrease agricultural expansion in the United States?
-
How might systematic productivity improvements impact crop-technology and management intensity decisions?
-
How might GHG emissions from livestock operations, which use crops as an input, respond to higher crop-productivity regimes?
-
Can agricultural productivity improvement be a source of GHG abatement by itself, and how might such improvement change GHG abatement potential for agricultural and forestry systems?
Statistical Examination of Technological Progress
Alternative productivity growth scenarios were developed for this study based on historical observations of yield growth. Specifically, we estimated productivity growth parameters for all major U.S. agricultural commodities across different time periods, then extrapolated future yields across several distinct scenarios. Note that these estimates include exogenous (technology driven) yield improvements, endogenous yield enhancement through input use (including nitrogen fertilizer expansion), and variability in yields driven by climatic factors. We do not differentiate between these separate yield effects in the statistical estimation phase, but as explained below, we do control for the endogenous yield effects in the model simulations. We use United States Department of Agriculture—National Agricultural Statistical Service (USDA-NASS) national data of crop and livestock yields from 1960–2009. Yields are estimated as a function of time, for both linear and log-linear functional forms, testing for structural break points from 1969–1985. A “best-fit” parameter is chosen for each commodity by minimizing the sum of squared errors over all regressions, with or without a structural break, and break year. The purpose of this relatively simple statistical process is to develop reasonable estimates of yield growth for use in dynamic simulation analysis.
Our baseline for this analysis, labeled “High” productivity growth, includes the set of best-fit parameters for each commodity, with or without structural break points. These are the highest productivity growth rates we examine, and could be considered overly-optimistic given recent literature on the potential impacts of climate change on grain yields in the United States. The “Medium” scenario extrapolates growth using only a linear functional form (with or without structural breaks). The least optimistic productivity growth scenario is named “Low,” and is formed by extrapolating yields using the lower boundary of a 90% confidence interval for the linear (Medium) growth estimation. Note, however, that we do not include a case with stagnant or declining growth, so it could be argued that all scenarios are optimistic in nature.
The “High” productivity growth scenario includes exponential growth rates for some important commodities, most notably corn (where the log-linear function form was chosen as the best statistical fit). The majority of the estimated parameters for the “High” growth scenarios were linear, thus there is little substantive difference between the “Medium” and “High” growth scenarios for most crops except for corn. However, given the importance of corn as a food and energy commodity, we separate this scenario from the others. Figure 1 illustrates the difference in productivity extrapolations by growth scenario for corn, soybeans (the “Medium” and “High” growth curves are identical for soybeans), and wheat.
Historic and extrapolated yields for key crop commodities (bushels per acre)
Dynamic Sectoral Optimization Methods
We examine the implications of future yield growth on land use, production, GHG emission levels, and GHG mitigation potential in the United States through dynamic economic modeling. The employed model was the Forest and Agricultural Sector Optimization Model with Greenhouse Gases (FASOMGHG; Beach et al. 2010), a partial equilibrium model of the U.S. forest and agricultural sectors, land allocation and management, and the GHG consequences thereof. The FASOMGHG model solves for multi-period, multi-market equilibrium between the sectors by maximizing inter-temporal economic welfare (the sum of producer and consumer surplus). Resources are allocated to the production of raw and processed commodities, and commodity output prices are endogenous.
Land use projections by productivity growth scenario (million acre)
The agricultural sector depicts production possibilities for 40 primary crop commodities, 25 livestock commodities, and more than 50 processed goods. Detailed budgets are associated with each production possibility, so the model chooses levels of a given production activity based on prices for outputs and inputs. Regional crop mixes are restricted to fall within observed regional area ranges, but can vary according to the relative net value of a particular crop. Primary crop commodities can be consumed domestically, processed into secondary commodities (such as biofuels), exported, or converted to feed. The crop and livestock systems are linked through the market for feed grains and extensive margin competition between cropland and pasture.
The structure of the forestry sector is based on the suite of models applied by the USDA Forest Service for Renewable Resources Planning Act (RPA) assessments (Adams and Haynes 2007; Alig et al. 2009), and depicts markets for most raw and processed forest products (including pulp, paper, and biomass energy). The FASOMGHG models land use competition between cropland, grazing land, and forestry, allowing for a portion of the agricultural and forestry land base to transfer across sectors (or between cropland and grazing land). Land use change is influenced by the relative profitability of each land use over time (Alig, Adams, and McCarl 1998; Alig et al. 2010; Baker et al. 2010), and the total land base shrinks over time to reflect development pressures (based on RPA estimates).
For this analysis, the statistically estimated yield growth parameters were incorporated into FASOMGHG exogenously. Only statistically significant parameters at the 10% level or greater are used in the simulations; otherwise, productivity growth is set to zero. The integrated modeling framework allows us to examine the interactions between exogenous (technology-driven) yield growth, and endogenous land management responses. Endogenous yield growth in FASOMGHG can occur through: land use or crop mix shifts; nitrogen fertilizer intensification to achieve higher yields (115% of base application levels); irrigation use; tillage method change; use of nitrification inhibitors; intensive grazing; alternative feed blends; livestock growth hormones; and crop/livestock mix change. In addition, possible intensive margin shifts in the forestry sector include choice of optimal rotation lengths, species mix, and forest management intensity class (see Beach et al. 2010 for more discussion).
Each productivity scenario is simulated under otherwise business-as-usual conditions for key market and policy variables, including the Energy Independence and Security Act's Renewable Fuels Standard biofuel mandates control case (USEPA 2010). Simulations are performed over a 70-year horizon from 2000-2070. To assess differences in GHG mitigation potential for the alternative growth scenarios, we introduce CO2 equivalent (e) price incentives ($15, $30, and $50/tCO2e, constant), following the same approach as in McCarl and Schneider (2001), Lee et al. (2006), and Baker et al. (2010). Further information on FASOMGHG can be found in Beach et al. 2010.
Simulation Results and Discussion
Model simulations illustrate how land management, production, and subsequent emissions trends might evolve under alternative productivity growth trajectories. Total annual production of most crop commodities decreases over time for the Medium and Low growth scenarios relative to the base. Between 2010 and 2030, annual corn and wheat production each decline relative to the baseline by approximately 1% per year for the Medium growth case, but 6% and 11%, respectively for the Low growth case. During the same time frame, simulated corn yields per unit area drop 5% and 17% for the Medium and Low Growth scenarios, respectively. However, projected wheat yields are actually higher under Medium growth than for the High growth scenario due to differences in land allocation and input use.
Deviations from the high growth base in total production and yields widen over time, especially for corn. From 2030–2050, corn production under the Medium scenario is 4.3%–7.3% lower than the High growth scenario, while yields per acre drop 17.9%–28.9%. As crop commodity production declines, so does the feed availability for livestock operations, causing higher feed prices, and lower livestock commodity supply. Commodity prices rise and exports decline for the Medium and Low growth cases.
Land Use and Management
Cropland use expands under diminished productivity (figure 1). Under the Medium case, total cropland use is roughly 1% higher than baseline from 2010–2030, and 3% or greater beyond 2030. For the Low case, cropland expands 1.3%–2% above baseline between 2010–2030, and over 4% beyond 2030. Results indicate a counteracting trend in pasture area. Use of grazing lands decreases for the Medium and Low cases, thus the herd size reduction effect of higher grain prices dominates the potential switch from intense-feeding to lower-cost pasture-grazed livestock systems. Most of the additional cultivated cropland used in the lower productivity scenarios is transferred from forests and pasture. Table 1 shows land conversion results. By 2030, the conversion of pastureland to cropland expands approximately 38.4% above the baseline (which is roughly 2 million acres). Cumulative deforestation increases 7.8% and 17.2% relative to the baseline, respectively, for the Medium and Low growth scenarios. Thus, results suggest strong extensive margin effects of reduced productivity growth.
| 2010 | 2020 | 2030 | 2040 | 2050 | |||
|---|---|---|---|---|---|---|---|
| Forest to Cropland | High | Total | 4.9 | 6.4 | 8.7 | 9.7 | 10.6 |
| Medium | Total | 5.1 | 6.9 | 9.2 | 10.1 | 10.6 | |
| Low | Total | 5.8 | 7.5 | 10.9 | 12.4 | 12.7 | |
| Pasture to Forest | High | Total | 2.4 | 2.5 | 2.5 | 7.1 | 14.2 |
| Medium | Total | 2.4 | 2.4 | 3.4 | 7.1 | 13.7 | |
| Low | Total | 4.2 | 6.1 | 6.1 | 8.6 | 14.3 | |
| Pasture to Cropland | High | Total | 2.4 | 5.2 | 5.5 | 5.5 | 5.5 |
| Medium | Total | 2.4 | 7.2 | 7.4 | 7.4 | 7.4 | |
| Low | Total | 2.4 | 7.4 | 7.4 | 7.4 | 7.4 |
In addition, we see some intensification of input use to at least partially counter the lower exogenous growth effects. Table 2 displays changes in input use intensity (per acre) from 2010–2030 and 2030–2050. Nitrogen fertilizer use and intensity per unit area expand, although only slightly. Phosphorous and potassium use intensity expands greatly over the long term (as crop production shifts to regions with relatively high synthetic fertilizer use). Energy use intensity expands for most fossil fuels, driven primarily by regional crop mix shifts. Water use effects are largely ambiguous at the national level.
| 2010–2030 | 2030–2050 | |||
|---|---|---|---|---|
| Medium | Low | Medium | Low | |
| Nitrogen | 0.1% | 2.3% | 0.0% | 0.8% |
| Phosphorous | 1.5% | 1.9% | 4.7% | 8.2% |
| Potassium | −0.9% | −1.6% | 3.0% | 8.6% |
| Water | 0.2% | 0.1% | −0.3% | −2.1% |
| Electricity | −0.5% | 5.5% | 0.0% | 0.9% |
| Gasoline | −0.5% | 5.5% | 0.0% | 0.9% |
| Diesel | −0.5% | −0.4% | −0.5% | −0.7% |
GHG Emissions
Even though crop area expands and input use intensifies under the lower productivity regimes, the net effect on GHG emissions from the agriculture and forestry system is relatively small. The primary reason for this small net change is that emissions from livestock increase with productivity growth, whereas crop production and forestry emissions decrease with productivity. GHG emissions from crop production increase slightly under lower productivity growth (less than 1%) due to the aforementioned extensive and intensive margin shifts. The net change in emissions from forestry is much higher on a percentage basis for the lower productivity cases, though the absolute effect is relatively small. This emissions increase is caused primarily by higher rates of deforestation for less-productive scenarios in early simulation periods. Net fossil fuel emissions displacement from bioenergy stays relatively constant across scenarios. Taking the different flux categories together, the net change in total annualized emissions (the sum over all aggregated accounts) is negligible. Net emissions, or the sum of all aggregated GHG accounts, increase less than 1% relative to the High growth baseline for both the Medium and Low growth scenarios (table 3).
| High | Medium | Low | |
|---|---|---|---|
| Crop Production | 324 | 326 | 326 |
| Livestock | 156 | 151 | 145 |
| Bioenergy | −118 | −117 | −117 |
| Forestry (including Afforestation) | 30 | 36 | 39 |
| TOTAL | 392 | 395 | 393 |
The primary reason for this small net change is that emissions from livestock increase with higher productivity growth, whereas crop production and forestry emissions decrease with higher productivity. Higher grain yields over time make feed less costly in the livestock sector, thereby raising herd size and total emissions from livestock operations. Total livestock head increase for the High growth scenario relative to the Medium and Low growth cases, as do pasture use and intensive grazing. Emissions from livestock are 4-7% lower annually for the Medium and Low growth cases, respectively, as lower cost feed is foregone. Thus, drawing conclusions about agricultural productivity effects on GHGs, while evaluating only land use change and crop management emissions, potentially ignores a large change in emissions from the livestock sector.
Mitigation Scenarios
While net GHG emissions change only slightly with varying productivity growth, GHG mitigation potential is drastically improved by higher productivity. Table 4 displays average annual mitigation potential for the sector, and several important observations can be drawn from these results. First, net mitigation potential from GHG price incentives is consistently lower under Medium and Low growth assumptions, and this difference grows over time as the gap in crop yields expands. The average difference in mitigation potential between the 2010–2030 and 2030–2050 periods widens greatly for both Medium and Low growth. The divergence in yields from higher-growth trajectories generates land rent differentials that raise the opportunity costs of allocating land for purposes of GHG mitigation.
| 2010–2030 Average | 2030–2050 Average | |||||
|---|---|---|---|---|---|---|
| High | Medium | Low | High | Medium | Low | |
| $15/tCO2e | 492 | 462 | 424 | 575 | 497 | 463 |
| $30/tCO2e | 811 | 809 | 747 | 1059 | 1013 | 959 |
| $50/tCO2e | 1119 | 1119 | 1075 | 1494 | 1423 | 1349 |
The second observation is that relative differences in mitigation potential across the productivity scenarios decline with the highest CO2 price incentive. This is especially true for GHG abatement options that come at the expense of conventional commodity production–the opportunity costs of sacrificing an acre (or a unit) of production increases at lower productivity levels due to the systematic effect that lower yields have on market prices and agricultural land rents. Thus, lower GHG mitigation price incentives may not be sufficient to lure land out of agriculture and into GHG mitigation. However, much like McCarl and Schneider (2001) find, the higher mitigation price incentives ($50/tCO2e) tip the balance in favor of substitute land uses that provide the greatest GHG benefits (such as afforestation and dedicated bioenergy feedstock production). This supports the notion that higher yields improve mitigation by freeing up more land for the provision of environmental services.
The third observation is that the mitigation portfolio changes under different yield growth assumptions. The High productivity growth offers the greatest mitigation potential from the livestock sector, bioenergy, and afforestation. For the Medium and Low growth baselines, emissions from crop production and forest management are greater than they are under the High growth baseline, and mitigation from these sources is improved. Greater baseline emissions from a particular source can expand mitigation opportunities for that source once incentives are introduced, thus offering additional lower-cost abatement options. Beginning with a lower-emissions baseline makes mitigation more difficult to achieve, especially at lower CO2 price incentives, as the system is operating at a high level of GHG efficiency to begin with.
Conclusions
Increased agricultural productivity holds the key to meeting the food, fiber and energy needs of a growing and more affluent population. It also will play a critical role in mitigating greenhouse gases. We perform statistical analysis of historic growth rates in U.S. agriculture to posit a range of plausible scenarios for future changes in agricultural productivity, then simulate sectoral performance under these alternative rates. We find that lower crop and livestock productivity growth leads to extensive and intensive margin shifts coupled with higher emissions from land use change and management, lower emissions from livestock operations, and a minimal net impact on aggregate emissions. The main reason for the negligible effect on emissions when the margins shift has to do with the complex relationship between crop and livestock systems, productivity, input use intensity, land use change and comparative advantage. Higher productivity growth may reduce the acreage and other inputs needed to grow crops, thereby lowering emissions from crop production, but it can also increase levels and lower prices of livestock feed, which raises production and emissions from the livestock sector.
The results also show diminished production and export potential under lower productivity regimes, which would increase international emissions, thus exacerbating the indirect emissions impact of reduced U.S. productivity growth. Havlík et al. 2012, shows strong global land use change effects of lower sustained crop productivity growth. Finally, while the baseline emissions effect is small in magnitude, we find substantial improvements in net GHG mitigation potential with increased productivity when mitigation price incentives are introduced. This result is consistent with the findings of the global analysis presented in Havlík et al. (2012). Thus, while productivity enhancement might not be a viable mitigation strategy in isolation for intensive agricultural systems, it can increase the effectiveness of direct GHG reduction incentives, both nationally and globally.
