A Fifty-Fifty Election Extracts Almost No Information From the Public

If you believe that elections matter, you believe something specific about them: that they extract information from the public and convert it into political power. The public wants something, the vote reveals what, and the winner is bound by it. On that view elections are important and participating in them is a duty.

I want to show that a fact everyone already knows about elections, that they land near fifty-fifty, is fatal to that view.

The Pattern

Look at any mature representative democracy over a few decades. The pattern is not that one side wins by twenty points and then the other does. The pattern is that the two large blocs sit within a few points of each other, election after election, and the winner is decided in the margin. Landslides exist, but they are the exceptions people remember precisely because they are exceptions.

I have argued elsewhere that this is not a coincidence or a sign of a divided country. It is what bundling issues into two packages does to a public that has clear majorities on every issue. The mechanism does not matter here. Take the pattern as given and ask what it implies.

Small Causes, Large Effects

If the structural expectation is an even split, the actual result is decided by whatever noise happens to be present. A clumsy remark by a candidate in the final week. A terrorist attack. A round of targeted advertising on a social network. The weather on polling day.

None of these has anything to do with what the public wants done about health, or housing, or war. Yet any of them can flip the outcome, and with it four years of policy.

When the margin is the noise, the result is the noise. A supporter of the electoral system has to accept that the government of the country is being selected by the least meaningful inputs available, because the meaningful ones cancelled out.

The Information-Theoretic Point

Here is a cleaner way to say what has gone wrong.

An election with two outcomes can carry at most one bit of information: A or B. That is not much to begin with. It is the answer to a single yes-or-no question about a country of millions with opinions on hundreds of things.

But a bit is only fully delivered when the outcome was uncertain in a way that reflects the underlying quantity. If the vote is structurally pinned near fifty-fifty regardless of what people think about the issues, then knowing the result tells you almost nothing about what they think. You learn which package edged ahead on the day. You do not learn which of its contents the public wanted, or by how much, or whether it wanted any of them.

I would guess, and it is only a guess, that a typical election delivers a small fraction of a bit about the public's actual preferences. The rest of the ballot's capacity is spent encoding the noise. For a mechanism that is supposed to be the channel between a population and its government, that is a terrible rate.

The Exception Proves It

There is one kind of election that does not behave this way, and it is worth noticing why.

When a single question dominates, the result moves away from even. Whether to withdraw from a failed war. Whether the sitting president is fit for office. Whether to accept a peace agreement on the table. In those elections the packages collapse into one issue, the public has a view on that issue, and the view shows up in the count.

Elections work exactly when they stop being about bundles. That is the tell. The mechanism is capable of carrying information; it is the packaging that destroys it. Reduce the vote to one question and the information comes through. Restore the package and it vanishes again.

What Follows

For anyone who values elections as the expression of the public will, this is bad news, and I do not think it can be argued away. Either the public has no clear views, which contradicts every issue poll ever taken, or it has clear views and the electoral mechanism fails to transmit them. The fifty-fifty pattern is the evidence for the second.

Something fundamental is broken in a system whose central act of consultation returns the same answer whatever the public thinks. The repair is not better candidates or better campaigns; those are inputs to a channel that is not working. The repair is to ask the questions one at a time, which is what the exceptions already do by accident and what a direct democracy would do by design.