I saw some recent discussion on Twittter of this post by Emily Cassidy in which she discusses her 2013 paper in Environmental Research Letters coauthored with Paul West, James Gerber, and Jonathan Foley. The subtitle of her post and paper is: "from tonnes to people nourished per hectare."
It's an interesting and thought provoking piece, and at the heart of it are figures like this one Cassidy posted on her blog:
In the published paper, the authors argue they, "illustrate where tremendous inefficiencies in the global food system exist today" and reach the normative judgement that, "shifting the use of crops as animal feed and biofuels would have tremendous benefits to global food security and the environment."
There are some methodological issues that I think are important in this discussion, some of which the authors themselves acknowledge and discuss, but I'll get to those in a minute.
First, I want to make the case that this state of affairs is not as "inefficient" or "irrational" as is often portrayed.
For one, take a look at the above figure. Is there some commonality between the locations with more green (more production for "food" - supposedly the "good" outcome)? These locations tend to be the spots that are relatively poorer, hungrier, and more malnourished. That ought to give us pause - that the locations with the supposedly "good" farming practices have some of the biggest challenges with under-nourishment.
Now, we shouldn't mistake correlation with causation (i.e., production for "food" probably isn't causing food security problems), rather I suspect this pattern is largely explained by income effects. What we're probably seeing in the above graphs relates not to production practices per se but to preferences of relatively rich people vs. relatively poor people. Our production practices are constrained by what people want to buy. In the same way one can argue it's "inefficient" for a relatively wealthy person to have a bigger car or bigger house or private jet, one can also point out that this sort of person has the means to pay for enjoyable things that are somewhat less efficient. If all we cared about was caloric/protein efficiency, we humans should eaten a spartan, undiversified diet of beans and rice. So, that's the first answer: people in relatively richer countries eat more meat because they like it and they can afford it. Maybe we shouldn't like or want to eat animal products, but as economists are fond of saying, de gustibus non est disputandum.
Beyond "preferences", why do we grow so much corn, soy, and wheat in the U.S.? A primary answer is that these plants are incredibly efficient at converting solar energy and soil nutrients into calories (they're the best, really the best). Moreover, these calories are packaged in a form (seeds) that are highly storeable and easily transportable - allowing the calories to be relatively easily transported to different times and to different geographic locations. Contrast these crops with directly-human-edible fruits/vegetables like kale, broccoli, or tomatoes. These plants are poor converters of solar energy to plant-stored energy (i.e., they're not very calorie dense), and they are not easily storeable or transportable without processing (mainly canning or freezing), which requires energy.
This gets to some of the methodological issues in these sorts of calculations. As I've discussed before using various analogies, there are two ways to view livestock. One is that they are inefficient - using up a lot of energy to make food. Another is that they are good at converting one form of energy that is highly storeable/transportable but untasty (field corn, soy, sorghum) to another form (eggs, meat, dairy) that we like to eat. Rarely do these sorts of research papers include the the calories (or energy) used in food processing. It is a mistake to compare the calories in steak to the calories in a wheat kernel. The wheat kernel requires energy/processing to convert to flour and then more energy to get pasta or bread. In the developing world (largely the green countries in the above graph), I suspect a lot of this processing isn't measured because it occurs in the household. The cowpeas, cassava, or beans require grinding and cooking to be human-edible, and the energy used to accomplish this isn't measured. The historian Rachel Laudan has written eloquently on this in a number of places (see her blog or book), and it is a feature of our modern food system that is vastly under-appreciated.
The other two issues the authors mention in their journal article as worthy of additional research are food waste and the ability of livestock like cattle to convert human-inedible calories from grasses into human-edible meat/dairy. On that last topic, there is a nice report by the Council for Science and Technology written Jude Capper and others. To those issues, I'd also add that we need to think about water use (the corn/soy/wheat in the U.S. is largely un-irrigated whereas fruits/veggies require comparatively large amounts of water often supplied by irrigation; of course, livestock consume water too) along with use of other inputs like pesticides and fertilizer (again, fruits/veggies can be relatively heavy users of pesticides).
Where does that leave us? I'm not going to say it's perfectly rational for the U.S. to devote the majority of it's cropland to corn/soy/wheat, but I think this discussion suggests it's not irrational either.
P.S. In terms of tonnes of production USDA data suggest in the 2016-2017 marketing year, 40% of the corn/sorghum/barley/oats produced and imported in the U.S. went to "food, alcohol, and industrial use", 32% went to "feed and residual use", 14% was in "ending stocks" (i.e., it was stored for future use), 14% was exported, and the small remaining amount was "seed use".