USDA will release the 2017 Census of Agriculture this week. As usual, there’s farmer buzz and speculation about the accuracy of the numbers because many farmers don’t fill out the survey.
Here’s how it works. All known farms, even small plots of land – whether rural or urban, are sent a survey if $1,000 or more of agricultural products were produced and sold, or normally would have been sold, during the census year. It is then up to the farmer to fill out the survey. While it is illegal to not complete the survey and NASS spends a lot of time and money to encourage participation, only 73% of known operators turned in a completed 2017 survey.
So how do the other 27% get accounted for? In the 2017 Census, NASS used capture-recapture methodology, which is an accepted statistical methodology to account for farms who either didn’t receive the survey, did not respond or were not properly classified.
Is it accurate? While that does leave some uncertainty, the uncertainty can be quantified. This is called a coefficient of variation and it’s published at all geographic levels for the Census. Which means we can access the coefficient of variation down to county-level data.
Additionally, as part of the agency’s commitment to accuracy, Census responses are compared to data NASS already has and other known information called administrative data. For example, NASS can use statistical models to compare the June acreage survey (which is more widely responded to) and information from the Census. Using these models, NASS can develop accurate and reliable estimates for U.S. agriculture.
Because the survey is dependent on operator response, errors can occur. Sometimes NASS contacts farmers to verify their answers. Because there have been previous concerns about the questions included in the survey, the 2017 survey was vetted by a committee to improve clarification, according to Veronica Nigh an economist at American Farm Bureau. “NASS then conducted two rounds of extensive cognitive testing on the proposed changes,” she adds.
