Association for Public Policy Analysis & Management

Soil erosion degrades the environment and agricultural production. One of the most famous examples, the 1930s American Dust Bowl, is estimated to have persistently decreased annual agricultural revenue and farmland value by 20-30% from the 1940s to the 1990s in highly eroded Great Plains areas (Hornbeck, 2012). Several countries including the US have used afforestation as a strategy to counter the adverse effect of soil erosion. In the Great Plains Shelterbelt Project initiated by President Franklin D. Roosevelt, 220 million trees were planted in American Midwest at a cost of nearly $20 million of federal and local investment from 1935 to 1942 (Droze, 1977). Similar projects are being undertaken today in arid and semi-arid areas such as the Sahara and China. From the literature in agroforestry, the benefits of shelterbelts include containing moisture in the soil, reducing wind-related damage, and improving air quality and irrigation efficiency; on the other hand, potential harm include occupying arable land as well as sapping and shading effects that take away water and sunlight from the crops nearby if not properly maintained (Natural Resources Conservation Service, 2011). Thus, we do not know in practice whether shelterbelts are worth the investment. Existing scientific literature has only focused on determining the theoretical effects of shelterbelts, leaving a paucity of empirical evidence on the actual effects of shelterbelts on agricultural productivity. My paper sheds light on the efficacy of afforestation programs by assessing the short- and long-run economic consequences of one such program in the US, the Great Plains Shelterbelt Project.

To empirically estimate the effect of shelterbelts planted, it is usually problematic to draw a conclusion by simply comparing the regions with shelterbelts and those without, because trees may be planted where it is feasible or has lower opportunity cost. To tackle this endogeneity problem, I use the historical variation arising from a shelterbelt zone designated in the Great Plains Shelterbelt Project as the instrumental variable to predict the actual region where shelterbelts were planted. This zone formed a 100-mile-wide belt that stretched 1,150 miles from the Canadian border into northern Texas, and was pushed as far westward as possible considering the feasibility of climate and soil conditions. Hence, the proportion of a county included in the Belt was arguably exogenous when controlling relevant climate and soil conditions. My first-stage regression shows that counties within the 100-mile-wide Belt are nearly twice as likely to be covered by shelterbelts (or 15.3 percentage points higher in coverage) as the neighboring counties outside the Belt. The second-stage results show that a 10% increase in shelterbelt coverage in a county leads to a 9-15% increase in agricultural productivity with the effect growing over time. I find heterogeneous effects by levels of soil erosion caused by the 1930s Dust Bowl. The regions with medium level of soil erosion benefit from shelterbelts whereas low or high eroded regions do not. In addition, the increase in productivity occurs in pasture rather than cropland because livestock are not adversely affected by tree planting.