Claim: Populations should be segmented into relatively more homogenous groups before conducting elections. Avoid segmenting them into relatively diverse groups.
The obvious problem is that gerrymandering does exactly this. But Bath argues that there is an alternative way of doing it that avoids the standard gerrymandering problems, by parties. This seems like a good way to guarantee constant party free-for-all and demagoguery, but perhaps this would be regarded as an appropriate trade.
There are, in fact, reasons why we tend to do things by geography (one of the commenters notes that even tiny Lichtenstein divides itself into two regions for these purposes), and some of them are things that Bath seems to regard as weaknesses: for instance, we make use of geography in this kind of context because it is not as ideological as a party system and thus guarantees that we aren't merely chopped up into opposing ideological blocks, and because we often do not like tying elections to mere political identity rather than a broader shared identity, and because we often want something that guarantees at least the occasional relevance of geographical concerns even when they don't make it into national policies. With none of these is it so very obvious that we're dealing with a weakness in the system rather than exactly the sort of thing a good political system will involve.
There are other issues, less obvious but also important. Bath suggests that representation based on geography is "[s]omething you never would have thought of yourself" that we do just because it's the way we've always done it. This is certainly false. The primary reason we have representation based on geography is that we are already organized by geography for lots and lots of other things; it's not that geography is important because it is used to determine representation but that it is used to determine representation because it is already important. The colonies pre-existed their own legislatures; the states pre-existed Congress; European countries pre-existed the European Union.
A second issue, sometimes closely related, is that people are culturally, morally, and sometimes religiously invested in geography far more than Bath suggests. People in Austin differ quite a bit from people in El Paso, both of whom differ quite a bit from people in Amarillo or Llano, but it's not hard to find people in all of these places who are heavily invested in the culture of Texas, which incorporates, and is not independent of, its geography. It is quite natural in such a context to elect representatives qua residents of this geographical unit called 'Texas'. (And it is worth noting that it is already as much a political and cultural unit when so identified as it is a geographical one; this is quite clear with Texas, which has a geography so diverse that it includes five different climate regions, each of which is itself sufficiently large to exhibit a significant variation in geography. And while there are geographical borders, there are also purely artificial borders arising out of a mix of culture, politics, and historical accident.) This is not a purely Texan thing, although Texans take it to the level of both an art and a national sport; this appears to be true in every U.S. state, in every Mexican state, and in every Canadian province, and is probably the norm everywhere; it has certainly been true of every place I've stayed long enough or under such circumstances as to get some sense of the culture.
It's also the case the geography is a handy political mnemonic; we tend to think in terms of geography for lots of things, anyway, even if it's just a matter of deciding where to vacation. It's second nature. One of the often overlooked advantages of the Electoral College is that it makes the Presidential election extraordinarily easy to follow. I mean, think about it -- here is an election system that requires that people engage in a lot of basic math and arbitrary numbers, and people still can usually follow it very well if you just give them a map. And what is more, people love the maps. And we see analogous things in other election systems, with parliamentary ridings and the like. (I'm actually pretty certain that one reason, also often overlooked, that people tend to hate gerrymandering so much more than many other kinds of political manipulation is that it makes the maps ugly and unstable.) When we look at smaller scale issues, the kinds of political arguments people have close to home, we find that they don't just think in ideological terms but in terms of variations across geography, partly because there are important geography-tied political issues, almost everywhere, things like water rights, but also because it helps with things like organizing campaigns, which often have to take geography into account. Thinking about politics geographically is already virtually universal; this doesn't require that we have to think of representation, particularly, in geographical terms, but it does make clear that it's a fairly natural way to think.
