Thread by @JohnFPfaff

The London-Tokyo shift here is a great example of the sort of data change that can come abt for only two reasons:

1. Definitions change (Tokyo absorbs something, something secedes from London).

2. Something's wrong with the data.

Pop sizes can't jump like this "naturally."

https://twitter.com/mattyglesias/status/1489204666994937868

In this case, something is wrong. Tokyo (and we'll come back in a second to what "Tokyo" and "London" even are) has been bigger than London before 2000.

At least based off these graphs, which track the numbers in the chart fairly closely.

But any comparison between cities is hard, bc what do we count? As this Guardian piece shows, defining a "major city" is actually really hard: theguardian.com/cities/2015/fe…

What is included, what is excluded? It can really shape how we see things.

Even here: Tokyo Prefecture has ~14M ppl, London the city ~8M. Those are a lot closer than 6.6M and 26.4M, as in the graph.

It's even possible the chart is not comparing comparable MSAs. London's metro pop is 14M, Toyko's is ~35M. That's apples-apples.

And, of course, all these comparisons turns on national statistical agencies defining the borders even for statistical metro areas (which can overlook political borders) similarly. Which... maybe they do! But cross-national comparisons are always fraught.

The struggle to define a "city" can lead to confusion. Like, how do you compare per capita crime rates of Houston and St. Louis (MO) when one city is a giant Borg-like sprawl absorbing everything around, and the other is just the dense urban core?

Not at all equally "city-like."

But anyway: for London to be #1 in 1999 at ~6M, and then Tokyo takes over in 2000 at ~26M just isn't possible. Pop sizes can't shift by 20,000,000 in a single year (absent utter catastrophe) unless the change is definitional, which doesn't seem to be the case here.

And to be clear: my point here is that anything involving "cities" is really hard, and it's really easy to get tripped up, and I have no doubt I've stumbled dealing with how arbitrary-but-significant all the contrasting and competing definitions are.