Skip to main content
No. 1574:
Signal-to-Noise Ratio
Audio

Today, we try to measure beyond our ability. The University of Houston's College of Engineering presents this series about the machines that make our civilization run, and the people whose ingenuity created them.

The Bush/Gore vote-count has given us all a short-course on observational science. When you hear this program, the election may or may not yet be resolved. As I record on a sunny overcast Saturday, November 25th, the issue remains mired in cacophony.

However, this lesson in data-reading will linger for a long time. Vote-counting has always been imperfect, and when the outcome depends upon a difference smaller than that imperfection, we're left in a quandary. Suppose, for example, that we didn't use an electoral college. Then Gore would've won the popular vote by about one vote in five hundred. I seriously doubt that our counting is anywhere near that accurate. Hence, for all practical purposes, the popular vote was a tie.

However, since we live by an electoral system, it comes down to Florida. It might've come down to any of the other close states, but only Florida has enough electoral votes to dictate the outcome. And the percentage difference in Florida is far less than it is even in the national vote. I'm far from first to point this out, but the hotly-argued difference is presently about one vote out of every ten thousand. No one can reasonably call the Florida vote anything other than a tie.

And so we've watched unprecedented legal and procedural combat as both sides try to settle an issue that data (that is, vote-counts) are far too inaccurate to settle.

All this has its analogy in a science laboratory. How many nights have I stayed up trying to tease a conclusion from data that weren't precise enough to yield a conclusion? Technically speaking, the signal-to-noise ratio is far too low. We have statistical means for extracting results when that's true, but only by processing data in ways that obscure individual data points. Those techniques would never play in today's court of public opinion.

The combat in Florida is thus our only possible recourse. We face similar problems in horse races and gymnastics competitions. For fifty years I've toted up test scores and given students grades, knowing that my measurements aren't nearly accurate enough to absolutely tell every C+ from a B-. We likewise use terribly primitive measuring instruments in our presidential elections. I suspect we'll have to wait until we've become more scientifically literate before we accept fully electronic, or mathematically subtle, means for measuring the will of the people.

Meanwhile, each side tries to finish off the other using the only tools available -- the courtroom and the media. There are no villains here, only the fact that we don't have a sudden-death overtime to break the tie. Since we don't, I'm content to wait out the boxing match. Whichever candidate wins will be my president. No one can call democracy pretty as it struggles with the only tools at hand. It's messy, but do let me know when you find anything better.

I'm John Lienhard, at the University of Houston, where we're interested in the way inventive minds work.

(Theme music)


My argument about the vote being a tie for all practical purposes is not terribly original. Listeners Glen and Lark Finneman suggested a program on this subject as early as November 18th, 2000, and the November 22, 2000, Houston Chronicle included an op-ed piece by William Brody, president of Johns Hopkins University. Still, it is a point that needs to be reemphasized.

All of these images have been distilled, by modern statistical methods, from what would have appeared to be visual hash, even in the highest-power telescopes on the earth's surface.

Graph showing a tie in the presidential election

A tie in the presidential election