COMPUTING SCIENCE
Machine Politics
Brian Hayes
Remaking the House
The U.S. House of Representatives reconstitutes itself every 10
years in a two-part process. The first part is reapportionment, in
which each state is allotted its share of representatives. This is a
purely arithmetical operation, and at first it appears quite simple.
Each state's share of the seats in the House is proportional to its
share of the total population, as enumerated in the decennial
census. But representatives come only in integer units, and so some
rounding of numbers is needed. That's where the trouble lies. There
are many schemes for rounding, some of which favor the larger states
and some the smaller.
From 1790 until 1940, reapportionment was a reliable source of
Congressional acrimony every 10 years. Since 1941, remarkably, it
has become a nonissue. What made the difference was writing an
algorithm into the law and thereby making the reapportionment
process automatic. Once the census returns are in, the allocation of
seats to states is immediately determined just by cranking the
numbers through the algorithm. There is no room for human judgment
or discretion. As it happens, the algorithm chosen is probably not
the best one for the purpose, but no major faction has been
sufficiently disgruntled to mount a serious challenge. This absence
of discord is perhaps the one bit of empirical evidence suggesting
that algorithmic methods might really have something to offer to
political science.
The second part of the decennial legislative makeover is
redistricting. Once a state learns that it will have k seats in
Congress, it must divide its territory into k districts, each of
which will elect one member. This is not an arithmetic procedure but
a geometric one, and it is severely underconstrained. For any
k > 1, there are many, many ways of drawing lines on the
map to create k districts. This superabundance of solutions
is an invitation to political (or meta-political) mischief.
The kind of mischief that comes to mind first is the partisan
gerrymander—a redistricting plan that favors one political
party over another. The term dates back to 1812, when Elbridge
Gerry, as governor of Massachusetts, presided over the creation of a
sinuous legislative district that opponents compared to a
salamander. There is also the "sweetheart" gerrymander,
where incumbents of opposing parties collude on a redistricting plan
to ensure their own re-election. Political gerrymandering is roundly
condemned by civic-interest groups, who argue that it gives an
unfair advantage to those who happen to be in control in a year
divisible by 10. They get to rig the electoral machinery to retain
their advantage throughout the following decade.
The legal status of political gerrymandering is uncertain. No
federal statute explicitly forbids it, and so far the Supreme Court
has never overruled a districting plan because of political bias.
Nevertheless, gerrymandering is widely considered unsporting, and
blatant instances are likely to offend the electorate.
Another form of gerrymandering,
which aims to dilute the political strength of racial and ethnic
minorities, was outlawed by the Voting Rights Act of 1965. Measures
taken to comply with this act were the issue in the recent North
Carolina and Texas litigation. Both states, under instruction from
the Department of Justice, created "minority-majority
districts" where African-American or Hispanic communities would
be able to elect representatives of their choice. The Supreme Court
ruled four of these districts unconstitutional on the grounds that
race or ethnicity had been the predominant consideration in drawing
them. (Note: I live and vote in the soon-to-be-former 12th district
of North Carolina, and the recent commotion over its status is what
inspired me to write on this theme. North Carolina districts serve
as examples in much of the discussion that follows.)
» Post Comment