Data on farm household income and characteristics come from USDA's annual Agricultural Resource Management Survey (ARMS). Changes in survey methods, sampling, or response rates can influence estimates derived from survey data. For 2019, the most recent survey year, the number of family farm records used to calculate farm household statistics was 13,070 farms, down from a recent high of 20,189 in 2017. Over the past 10 years, the highest response rates (67.4 and 67.9 percent) were in 2012 and 2014 when the Census of Agriculture and TOTAL Survey were conducted.
Changes Since 2012 and Effect of the Coronavirus (COVID-19) Pandemic on Data Collection
Beginning in 2012, two major changes affected how ARMS data are collected. First, ARMS was an all-mail survey in 2012-14, though most surveys were eventually completed through follow-up in person or over the phone. Second, the all-mail "core" version of ARMS Phase III that had been introduced in 2003, was discontinued. These survey design changes make farm household statistics from 2012 onward difficult to compare with earlier surveys and can also affect the precision of estimates.
Most surveys are mailed to respondents, but in most years the majority of records (50 to 82 percent) were enumerated in person. For the 2014 and 2019 data, just over 40 percent of the surveys were enumerated in person. Between 2009 and 2019, the share of usable records enumerated through the mail has ranged from 14.2 percent in 2011 to 46.3 percent in 2014; in 2019 33.5 percent of usable records were enumerated through the mail.
For the 2019 data, almost 9 percent were enumerated over the phone, which is much higher than the 2-3 percent seen in previous years. This reflects changes in collection methods that were made because of the coronavirus (COVID-19) pandemic of 2020.
The coefficient of variation (which measures the size of the standard error relative to the mean) for total household income has ranged from a low of 1.9 in 2010 to as high as 5.6 in 2019; the variation in mean off-farm income moved similarly, at 1.6 in 2010 and 6.8 in 2019. The coefficient of variation of the farm income component of household income was generally higher, with a low of 3.8 in 2014 and a high of 18.8 in 2009, consistent with greater riskiness and variation overall in farm sector income. See more statistics on the ARMS from recent years.