Who counts for representation? – part I

The Trump administration has abandoned its plan, already blocked by the Supreme Court, to have a question about citizenship included on the next United States census, due to be taken next year.

Clearly one objective of the plan was simply to keep the issue of immigration, especially illegal immigration, in the public eye, so as to maintain enthusiasm among Trump’s fervently xenophobic base. But there was more to it than that.

If immigrants – including those who are in the country unlawfully, or who think they might be suspected of it – are asked to disclose their citizenship status, it’s thought that many would refuse to participate in the census, resulting in a systematic under-counting of areas where more immigrants live. And that, almost certainly, is what the administration intended.

Under-counting could matter for a lot of things, but a key one is representation: in Congress, in state legislatures, and in the electoral college, where population numbers are critically important.

And even if there was no actual under-counting (or not enough to matter), collecting data on citizenship would give Trump and his supporters ammunition for a campaign to allocate representation on a more politically favorable basis by ignoring non-citizens.

This all raises a set of very interesting questions. Who should count when it comes to representation? Should we exclude unlawful immigrants from the relevant numbers? What about lawful permanent residents who are not citizens? And why stop there – what about citizens who are below the legal voting age? Or those who are eligible to vote but not enrolled? Or who are enrolled but don’t bother to turn up? Or do but vote informal?

Most democracies don’t have this problem. If you have a direct election for president, for example, the people who count are the ones who turn up and cast a valid vote. The president, once elected, “represents” everybody, but those who don’t vote carry no weight in electing them.

Similarly if you have a legislature elected nationwide by proportional representation. Only actual votes get counted; there’s no issue about apportionment, and therefore no need to worry about which set of numbers to base it on.

But neither the US nor Australia works like this. One of the reasons that the US created the electoral college was that in a direct election, some sections of the country would be under-represented in relation to their population, because a large fraction of that population consisted of slaves who had no political rights.

The use of population numbers in the US comes in at two levels: firstly, in apportioning a number of seats to each state in the House of Representatives, which is also used for representation in the electoral college; and secondly in drawing district boundaries within a state, both for Congressional elections and for state legislatures.

For the first purpose, the law on how to proceed is apparently clear. The fourteenth amendment, adopted in 1868, states that “Representatives shall be apportioned among the several States according to their respective numbers, counting the whole number of persons in each State, excluding Indians not taxed.”

It goes on to provide for representation to be reduced in the case where a state restricts the voting rights of some of its – adult, male – citizens. This is no longer relevant (at the time it mattered a great deal for the rights of former slaves), but it shows that the authors of the amendment knew what “citizens” were. If they’d meant that rather than “whole number of persons” in the first sentence, they would have said so.

None of that has stopped the anti-immigrant lobby, including the president, from arguing that the inclusion of non-citizens is unfair. And in one sense they clearly have a point, since (as noted above) in a direct election, which every sane person agrees is more democratic, the non-citizens, like non-voters in general, would not be counted.

But since Trump only won the presidency courtesy of the electoral college, none of his supporters seem at all keen on its abolition, so one might reasonably question their devotion to democracy.

As I’ve argued before, the main problem with the electoral college is very much the same as the problem with single-member electorates in Australia: it’s not so much systematic bias as random error. In a close election, either system can get the wrong result – as 2016 did in the US, or 1998 in Australia – simply by one side being lucky (or tactically astute, as its supporters will maintain) in the distribution of its vote.

There are, however, systematic biases in the electoral college. One is the fact that seats include an allocation for senators (two per state) as well as representatives; that gives a bonus influence to small states, which currently works to the advantage of the Republicans, but the effect is very slight.

Another is the fact that non-voters count towards apportionment. How big a difference does that make?

To find out, I did the calculation myself, reapportioning seats in the electoral college on the basis of actual votes cast in 2016 rather than population. (If you want to repeat the exercise, here are the 2016 results, here are population estimates and here are details of the apportionment method, although regular Sainte-Laguë works just as well.)

In 2016, Trump won by 74 votes in the electoral college, with 306 to Hillary Clinton’s 232. That was based on an apportionment using 2010 census populations; if you revise it using 2016 population estimates, there’s almost no change – he wins by 308 to 230.

And if instead you apportion seats based on the actual votes cast in 2016, it’s almost exactly the same: 307 to 231.

But that conceals some quite large changes; it’s just that they mostly cancel one another out. Going by votes instead of population shifts seats away from the growth states of the south-west, where there are lots of children and immigrants, and towards places that are whiter or have more old people.

So, Texas loses ten seats and California loses eight, while Pennsylvania gains three seats and Colorado, Massachusetts, Michigan, Minnesota and Virginia pick up two each.

At the moment that’s pretty much a wash politically. But it could easily become important. Imagine, for example, that next year’s Democrat candidate pursues a “southern strategy” and picks up Arizona, Georgia and Texas, but at the price of losing support in the north and dropping Maine, Minnesota and New Hampshire.

In the actual electoral college, that would represent a net gain of 48 seats, for a clear if narrow victory. But in the hypothetical reapportionment according to votes cast, it would only be a net gain of 37, leaving Trump with the narrowest of wins, 270 to 268.*

So apportionment can make a difference; usually only a small one, but in a close election small differences matter a lot. What about the second use of population numbers, for drawing district boundaries? We’ll look at that in part 2.

 

* In each case I’m leaving in the extra two seats per state for the senators, but in a very close election they matter too.