U.S. Crop Yields Redux: Weather Effects versus Human Inputs

Over the 40-year period from 1960 to 2000 the dramatic increase in world crop production was the result of land expansion, but also the result from increasing yields through the use of chemicals, fertilizers, pesticides and water from irrigation systems (Tilman et al., 2002, PNAS). An additional 50 percent increase in world population and higher incomes are expected to double food demand by 2050. Studies show that agriculture production in certain regions like the cereal production areas of the United States, are operating near maximum yield potential (Grassini et. al, 2011, FCP). All this seems to indicate that the food production increases needed to satisfy future demand will put great pressure on existing cropland and natural resources. Climate change, a final source of concern, is likely to aggravate the situation (Schlenker and Roberts, 2009, PNAS). An important step towards understanding the evolution of agricultural production under different climate scenarios is to carefully estimate the effect that different temperatures have on agricultural productivity. This is the goal of this research. Most agronomic studies on the effects of weather on crop yields are based on field experiments. These studies account for the biological effect of different temperatures on specific crops (Ritchie and Nesmith, 1991, MP&SS.) Schlenker and Roberts (2009) consider the effect of weather on aggregate farm yields. They regress corn, wheat and cotton yields in counties east of the 100º meridian on weather variables during the years 1950-2005 and found that there is an increasing positive relation between temperatures and crop yield up to 29-32ºC (depending on the crop.) Temperatures above these thresholds are found to reduce yields significantly. Their regressions included precipitation, time trend, soils, and county effects for location-specific unobserved factors. There are two important omissions in this study. First, they only consider rain-fed counties, those east of the 100º meridian, while production increases have been directly related to irrigation developments mostly west of the 100º. Second, their study controls for natural characteristics like precipitation but does not allow for purchased inputs. These inputs have had a pivotal role on increased yields and are under the control of the farmer. It is important then to understand the degree of substitution and the contribution of these versus other inputs to the time trends they estimated. This research develops a county level biomass production function for an 800-mile climatic gradient from the Rocky Mountains to the Mississippi River (41o N latitude). A panel data set that includes 101 counties for the 1960-2008 period is developed. The quantity of biomass produced per hectare (from all crops) is hypothesized to result from the use of traditional inputs under farmers’ control such as land, fertilizer, chemicals, and irrigation and from environmental variables such as soil organic matter, precipitation and temperatures. Indexes are constructed for all variables at the county level. Given interest on climate effects, particular emphasis is placed in the development of county precipitation and degree-days indexes. A semi transcendental logarithmic production specification is jointly estimated with share equations for purchased inputs using a seemingly unrelated estimation approach with panel data corrections. The authors do not know of any other study for this region that provides this information. Results quantify the critical effects that high temperatures have on agricultural productivity in the region, after controlling for irrigation, other managed inputs, soil characteristics, precipitation, and technological change. We found a negative and substantially increasing (nonlinear) effect of temperatures over 30 ºC on crop yields. While a full day of temperatures between 30ºC and 35ºC decreases expected yield by 0.5%, a day of temperatures over 35ºC decreases yields by 36.5%. Our results are qualitatively similar to the findings in Schlenker and Roberts but provide additional information. First, the inclusion of irrigated land seems to diminish greatly the negative effect of higher temperatures. We estimate that in irrigated areas the harmful effect of temperatures above 35ºC is reduced by 62%. Second, it is also clear that semi-arid areas like western Nebraska and eastern Colorado and Wyoming, for example, compensate the lack of precipitation with high values of irrigation. Finally, the contribution of fertilizer and chemicals to yield changes is significant, substantially reducing the residual time trend due to unspecified factors. Technological change has been irrigation, fertilizer, and chemicals using.