Above: the Frontier Strip - the westernmost U.S. states before the Rockies - contain the 100th meridian (bold vertical line in maps) which is often used a marker of low precipitation and therefore low agricultural productivity. The argument has been that the population therefore drops off to the west of this; but if that's the case, then why does Canada not follow the same pattern?
It is commonly believed that the population drop-off seen in the United States at the 100th parallel has to do with changes in rainfall and agricultural productivity as one moves west in North America. This article shows that this is probably not the case. It more likely has to do with historical accident - about the time Americans had expanded west to the 100th meridian, we started building trains, and people could move to the more pleasant climates of the West Coast. Note that the "population distribution is by historical accident" theory applies more to the New World than the Old, because of the introduction of intensive agriculture and lack of a Malthusian limit. Maybe over time, there will be a smoother population gradient - it would be easier to argue that with thousands of years of agriculture, longer-term Malthusian limited populations in Asia and Africa populations do reflect agricultural productivity, and therefore also rainfall and longitude-degrees from ocean.
Presumably the same applies to latitude, although for different reasons. Very high latitudes are not great places to grow food. Latitude is still going to be affected by geography (if you're near coastline at 25 degrees latitude, life is much more pleasant than if you're inland) - but the solar radiation remains constant so the effect of latitude is a little less subject to being obscured by the details of geography than longitude would be. (Note that there may be constraints on population growth other than calories that I'm completely neglecting; e.g., maybe agriculture at the equator is better than 30 degrees, but disease slows population growth once you drop below the 30s.)
Geography junkies have likely already seen maps like the one above showing population by latitude, and the assumption is that agriculture just produces more calories at certain latitudes (similar to the provincial precipitation distribution that obtains in North America and which was assumed to influence our own population distribution). But if we're trying to pull out first principles, absolute population doesn't help much - what if there's just a lot of land at certain latitudes? That doesn't tell us whether that latitude is good for agriculture and therefore modern humans, it tells us where the land is on Earth's surface. If your denominator for population density includes ocean, then it doesn't really tell us anything about the carrying capacity of land at that latitude. So population alone won't work - it needs to be population density.
And even just population density per latitude won't work - it has to be population density per actual land area found at that latitude. I couldn't find population density per land area data. Fortunately this Brookings Institute Paper quotes World Bank 1997 figures for land area and population in bands of 10 degrees latitude (see pp 49-51), and I back-calculated population density per actual land area at this level of granularity.
Thinking it might be interesting to see how things stack up when the northern and southern latitude bands are combined, I did so:
(There wasn't full data below 40s south or above 50s north, which means we can't even do 50s when we combine them, which non-ideally eliminates most of Europe.) The question that emerges is, what's wrong with you people in the 30s and 40s south? Up here on the correct side of the world we certainly seem to be able to get along alright in those latitudes. The lion's share of the land in the 30s and 40s south is to be found in the southern half of Australia, most of Argentina and Chile, and South Africa. These aren't bad places to live - note that this contains 3 of the world's 5 Mediterranean climate zones! Again, even trying to smooth things out, historical details matter. These are places where intensive agriculture was only introduced in the last few centuries by Europeans. And geographical details matter, even after using population density instead of population - Australia is an ancient craton with terrible soil, and Argentina is drier than it might otherwise be due to the rainshadow of the Andes. That said, in the Malthusian future, we should expect to see the Pampas filling up much like another Mediterranean climate zone (California) is filling up as we speak.
Finally, it's worth glancing at Malthusian long-term intensive agriculture countries, to see whether this holds up. I've put population density, precipitation and elevation next to each other. China's marked parallels are 40-30-20. India's are just 30-20. (10 crosses near the southern tip of the subcontinent.)
 |  |  |
 |  |  |
The precipitation relationship holds up better for India than for China, but both of them follow elevation. This may be more true in rice-farming areas than other grains. The relationship of population to elevation has been profitably studied elsewhere; there's a notable outgroup in the central Mexican plateau, where grains that enjoy dry conditions are grown and near to where agriculture began in North America.
Fully stated then, the theory can then be summarized as "Population distribution is strongly influenced by agriculture which in turn is determined by climate and therefore latitude, but a) until Malthusian limits are reached millennia after the introduction of intensive agriculture, distribution will be mostly by historical accident as in the New World and b) precipitation and especially elevation dramatically affect the distribution as well."
