Missing data recovery for farm financial data: an application of the matrix completion method

The Agricultural Resource Management Survey (ARMS) dataset is source of information on production and financial practice of farm businesses and the farm households in U. S. However, one variable missing from ARMS data is commodity-specific returns. While revenue is recorded on a commodity-specific basis, input usage data are generally not, meaning that commodity-specific returns can only be computed for single-commodity producing farms. Possessing commodity-specific return distributions for a sample of U.S. farms would be very useful for agricultural policy analysis. In this dissertation, I utilize a matrix completion approach to recover the missing commodity-specific net return values in ARMS dataset and estimate the suitable fitted distributions for those net returns for six major commodities. Overall, the matrix completion approach is efficient at recovering these missing values and allows policy makers access to highly useful information.