New models to estimate costs of US farm programs
Is Version Of
In this study, I extended the stochastic model built by Babcock and Paulson (2012) to conduct a one-year cost projection for crop insurance in order to investigate its feasibility of being solely provided by private firms. Based on the 52 years’ yield data from 1961 to 2012, the risk consequences from insuring crop yield and revenue against losses are estimated to be far beyond what private insurers could bear on their own. However, reinsurance from the government provides an attractive incentive to insurance firms. Among the six insurance policies researched in this study, the minimum expected net underwriting gains to private firms with government reinsurance in 2013 was $289.9 million, which is about 9.30% of retained premiums. The maximum loss that firms could have borne in 2013 was $4.9 billion. In addition, the impact of a proposal to eliminate premium subsidy for the harvest price option is also estimated. The total savings for taxpayers are estimated to be $1.3 billion, which is about $400 million more than CBO’s estimate in 2013, but only 67% of its estimate in 2015.
Based on the three-crop competitive storage model initiated by Lence and Hayes (2002), I also develop a better approach for a multiple-year cost projection by modeling the demand shock as a random walk. This approach is capable of preserving the correlations between national yields and prices, maintaining the relationships among national, county and farm yields, retaining the spatial correlations of yields across crops, and incorporating inter-temporal price correlations as well. More importantly, this approach is capable of simulating price draws with a desired volatility pattern: increasing over time but at a slower rate than square root of time t, as stated in Lence, Hart and Hayes (2009). Preserving these correlations and price-related features are crucial in conducting precise cost estimations and valid policy analysis. My analysis shows that the payments from Price Loss Coverage (PLC) with its time-invariant fixed guarantees would be significantly underestimated if both the price serial correlation and the increasing, concave price volatilities are ignored. For Agriculture Risk Coverage (ARC) and Supplemental Coverage Option (SCO), their guarantees are adjusted to reflect market conditions so the difference in estimated payments is modest. An easy fix for estimating the cost of PLC is to inflate the price volatilities used to generate random prices for budget scoring purposes.