My dad explained to me that in the 1960s and 1970s, USDA assessors would just drive down the road and look at a field and scribble a random number into their notebook. The data might not be very accurate. The reason he said they never cared to write an accurate number is that after they gave the numbers to the USDA, the USDA would just write the number they wanted to hear down instead. It had something to do with keeping the price they needed to keep exports flowing. For farmers of course, it hits them right in the pocket book because it prevents the prices reaching where they should be. On the other hand, subsidies help alleviate that a bit.

I'm not sure if this is true still today. But, I doubt they're counting every successful crop correctly. And, there's a lot that can happen before they are actually harvested; floods, insects, blight, drought, etc. For example, there'd be a huge drought and the USDA would announce a bumper crop and he'd laugh. And, that was recent.

Either way, past data is probably pretty suspect. And, today's data might have the same problem. Which means if you're basing prediction models on this you might expect to be disappointed when things don't pan out. This is especially dangerous if we enter an era where food becomes more scarce. It is an accounting issue if you think about it.