Many studies have used the U.S. Department of Agriculture’s (USDA) Agricultural Resource Management Survey (ARMS) to research various aspects involving the agricultural sector in the United States, including studies focused at the farm level. Since nonresponse and inaccurate reporting may cause significant bias in statistical analysis, research is conducted to estimate the magnitude of response error in the farm debt section of Phase III in ARMS. USDA Farm Service Agency (FSA) data on Farm Loan Program (FLP) borrowers with end of year debt are matched with ARMS respondents for years 2001, 2004, 2006, and 2007. A multinomial logit model is estimated to identify demographic, structural, and financial characteristics of FSA borrowers who refused to indicate if they had end of year farm debt, or who accurately or inaccurately classified their farm operations as having end of year farm debt on the ARMS. Additionally, estimates of the magnitude of response errors in ARMS for both FSA direct and guaranteed FLPs are estimated. The current study estimates that 12.9 percent of direct FLP respondents and 9.9% of guaranteed FLP respondents indicated “no” to the question on having end of year farm debt when they should have indicated “yes”. Also, those responding “no” are found to have their ARMS total debt outstanding less than their FSA total debt outstanding. Direct FLP operators are more likely to report “no” and, therefore, under-report end of year debt in the ARMS if they had a lower total FSA debt outstanding balance, had a greater value of crop production relative to total production, or had a lower gross cash farm income. Guaranteed FLP operators are more likely to under-report their debt in the ARMS if they had an operating line of credit loan, had a greater share of production from crops, or had a lower gross cash farm income. They are less likely to under-report their debt if they either had some college education, were eligible for socially disadvantaged loans, or were beginning farmer eligible. These results are of keen interest to those using ARMS data and those responsible for collecting ARMS data. The results allow researchers using ARMS data to appraise operator debt status to be better informed about potential data limitations. For example, estimates of the share of farm operations with little or no debt may be overestimated. Also, the results could assist the USDA National Agricultural Statistics Service in developing improved imputation techniques to estimate farm debt from the ARMS, especially for those respondents indicating “no” as to having end of year debt in the Farm Debt section of the ARMS.