The increase of consolidation in agriculture is a major theme in your paper. How does the definition of “farm” used by the census hide this trend?

Think about what happens in the 1940s: There’s a lot more mechanization—like tractors—and you have artificial fertilizers. These mean you can farm much larger farms. So you start seeing consolidation. [Only 1.2 percent of the U.S. labor force are farming, according to USDA Economic Research Service data.]

As a society, we have a hard time thinking of agriculture as just a commercial enterprise. And there are a lot of forces that want to maintain this ambiguity. Think about it; if we say there are 200,000 farms that produce eggs, we’re not talking about the fact that, really, over 90 percent of America’s eggs are produced in fewer than 400 facilities.

For the agricultural lobby, extension researchers, and people whose jobs depends on agriculture, the legitimacy of their work is in part tied to the number of farms. And so there is a general resistance to admitting that the situation has changed dramatically and things are not the same as they were in the 1940s, ’50s and ’60s.

The way the number of farmers is tracked has also changed. The census collects data on the number of  operators per farm, but since 2002 it has allowed farms to report more than one operator. Can you talk more about how this change impacted the count of U.S. farmers?

These changes were driven by good reasons. From 1850 to 2002 you had a perfect correspondence between farms and farmers: There was one farmer per farm. What that tended to do [was] hide the presence of women in agriculture. To improve on inclusiveness, the census began adding other operators for each farm.

But still, there was one principal or main operator. What I think is much more problematic is that in the last census, we got rid of that primary operator. The main impact of this change has been that we’ve blown up the number of women that appear to be farmers, because now we’re not distinguishing between the roles of the women on the farm.

We haven’t made progress. We’ve just counted women in a different way than we were counting them before. It’s completely artificial, compared to how we were doing things in the past, and it makes it impossible for somebody 20 years from now to go back and see how things have changed in the first quarter of the century versus the second.

How has the census improved or gotten worse at accurately capturing the racial, gender, and ethnic breakdown of farmers?

The census is produced within what is happening in society. 1978 was the year where we separated race from ethnicity, and so we started counting Hispanic farmers. But we were underestimating how many Native Americans were working in agriculture, because there was just one person counted per farm. Not until much later did we start counting all the farmers who worked on commonly held reservation farms. Those were considered “abnormal” farms, which in and of itself is a loaded term.

Also, the census obviously has issues in terms of response rate, what people tell you, [and] how they understand the questions. The response rate for the census has been going down, particularly for large operations, because those people are busy and they have other things to do.

Under its current structure, who else is not represented in this census? Is the census accurately reflecting the reality of U.S. farms, especially when it comes to issues like underrepresented farmers and farm labor?

Farm labor is a really big one. The USDA collects some information on farm labor, which it gets not from the farm laborers themselves, but from the farm operators. [So the labor] population is very little understood. We know that a lot of them are undocumented. We know that a lot of them are tied to a specific visa and are not mobile from job to job, which has led to exploitation. But also, these are people who have trouble accessing services because they live in rural areas. I argue that the census’ focus on the farmer as the operator, while obscuring this role of agricultural workers, is problematic, particularly now.

The other category that the census basically completely ignores is landowners, and that’s really troubling to me, because most agricultural landowners are non-operating [not farming the land themselves], and about a third of U.S. farmland is owned by people who are not farmers. They tend to be pretty white. They tend to be older. Because of the way the tax laws are structured, they tend to keep the land within the family, so you have this generational wealth aspect. These people may not know anything about farming at all.

I think it’s really important that we get good information about these people, like, what kind of linkages they have with farming and what kind of role the farm money plays in their finances. This population is very understudied as well. The picture that people get from the census is not as illuminating as it could be if it took into account the consolidation in farming, the lifestyle farms, and these other understudied populations.

Does the Trump administration seem to be following the current census definitions of farmers when it comes to enacting policies?

I would say that administrations in general talk the talk of helping small farmers, but don’t walk the walk. If you think about the role that the USDA has historically played in delivering policy, it is very important, whether it’s loans, technical support, or research through the Agricultural Research Service. But we know that the support has been uneven because there’s been lawsuits, for example from African American farmers, that show the USDA discriminated against them in access to services and loans and information.

USDA is an equal opportunity discriminator in the sense that it doesn’t matter who’s in the White House. USDA has historically discriminated against minoritized farmers and against farmers who were doing things out of the box that didn’t fit the policy. If you want to be diversified, if you want to be organic, or if you want to focus on animal welfare, USDA is not producing a ton of research that matters to you. It’s producing a ton of research that matters for the big guys.

What are the potential consequences for the census if it isn’t updated to reflect reality?

I think that at some point, the chasm between reality and what the census shows is going to be too wide. And given what we are seeing now with the administration that we have, and what they’re doing to the statistical system—I think you have to be truthful and hold the line on these things now more than ever. Having good data is the foundation for making good science, and good science is the foundation for making good policy. You cannot do anything if this data isn’t right.

This interview was lightly edited for length and clarity.