We read this recent opinion piece in the New York Times after a skeptical friend sent it to us. A few days later, another friend, a science journalist, sent the same article to a group of farmers (including us), asking if anyone would draft a letter to the editor in response. After checking out the piece and its source–this unpublished working paper–we agreed that the author, Marc Bellemare, was overstating his results when he sounded the alarm that farmers markets are correlated with a higher incidence of food-borne illness per capita. To his credit, Bellemare reminds readers that the correlation that he found does not imply causation. But reports like this get picked up and repeated without the caveats that Bellemare included in his article. We agree with our journalist friend who said that the New York Times gave Bellemare a big megaphone for results that are less than conclusive.
To understand the difference between correlation and causation, imagine reading that states with more traffic accidents also had more food-borne illnesses. Another way of saying this is that there is a positive correlation between traffic accidents and food-borne illnesses. It would be difficult to imagine anyone arguing that traffic accidents caused the illnesses, or for that matter, that illnesses caused the accidents. Similarly, there is nothing in Bellemare’s results that would prohibit the conclusion that food-borne illness caused increases in farmers markets, rather than the other way around.
Still, a correlation between two variables is usually the first clue that they might be related. The stronger the correlation, the more likely it is to catch a researcher’s eye. If you look at the figure above, you see one of Bellemare’s main results. It shows a correlation between farmers’ markets per capita and a variable that is not well-explained in the paper, but which presumably represents all reported outbreaks of food-borne illness. The green dots are the actual data, the red line is a type of trend-line, and the blue shape is the uncertainty about the location of that trend-line. The steeper the trend-line, the stronger the correlation, so a flat line would indicate no correlation. Look at the actual data points (green dots). If you were looking at the data for the first time, without the blue shape and red line, would you think there was a relationship, or would this just look like a blob of points? If you do think there is a relationship, is it a strong relationship?
We don’t think so. To us, it looks like a shotgun blast of points that show no trend. But even if you think it does show a trend, it would still just be a correlation.
Mistaking correlation for causation is a common error that we have written about before. What bothers us most is the New York Times should have known better. By publishing the piece with the headline, “Farmer’s Markets and Food-Borne Illness,” the NYT is implying that shopping at farmer’s markets increases the risk of contracting a food-borne illness. This, despite the fact that in the conclusion of his unpublished working paper, Bellemare himself writes “from a policy perspective, it would be a mistake to take the results in this paper and discourage or encourage people to purchase food from farmers markets on the basis of our results.”
That said, we welcome a discussion of ways to reduce the risk of food-borne illness. It is the responsibility of small and large producers alike to ensure the food that they sell is safe. Regulators rightly focus on larger producers, whose products are available in supermarkets nationwide. Higher risk foods, like cheese and meat, are also more closely regulated, even at farmers’ markets. It is likely that small producers of lower risk foods, like vegetables, are not inspected because the risk of an outbreak is low relative the impact it would have, since each farm serves a small area, and not an entire region.
And therein lies the main safety mechanism for consumers. By shopping at a farmer’s market, you are buying food grown by a member of your community. No-one wants to make their neighbors sick, and we all know that word travels quickly if there ever is a problem. Farmer’s markets give customers the opportunity to make food safety judgements for themselves, by asking producers about their practices, or by visiting their facility. It is when food production is removed from the community that it becomes necessary for the government to step in and inspect on the consumer’s behalf. So until there is actual evidence that food purchased at a farmer’s market is unsafe, we can relax and continue to enjoy farmers’ markets for their fresh nutient-dense food, local economic benefit, and sense of community.
We did send a letter to the editor in response to Bellemare’s opinion piece, but the Times declined to publish it. Here it is:
To the Editor:
Marc Bellemare writes that certain outbreaks of food-borne illness are correlated with the number of farmers markets per capita in a state (“Farmers Markets and Food-borne Illness” Op-Ed, Jan. 17). He rightly concedes that correlation does not imply causation, so for the Times to publish his non-peer-reviewed analysis seems premature and risks sowing undue fear of local food. To imply that farmers markets are to blame ignores that, as Bellemare writes, “most…illness [is] caused by consumers who undercook or fail to wash their food,” no matter where it is purchased. Food at supermarkets is anonymous. But at the farmers markets where we sell in central Virginia, customers can ask about food safety issues including pesticide applications, use of growth hormones and antibiotics, and worker conditions. The Times’ ill-considered publishing of Bellemare’s results stands to misinform consumers.
Ben Stowe and Heather Coiner