June 5, 1992
A straw vote only shows which way the hot air blows.
O. Henry
Time was when you had to wait until the ballots were counted before you knew who got elected. Now we are told who will win the election long before it takes place.
How accurate are polls? In the hands of people who don't know what they are doing, they are about as reliable as a ouija board. In the hands of experts, they are almost as reliable as the weather predictions are in the hands of expert weather people. With weather forecasts, 6 month predictions are worthless; a week before, they stand a fair chance of being right; the day before, predictions can be very good --but never certain. Only on the day of the weather, when the weather man looks out the window and sees rain, is his prediction right on the money. Since public opinion is at least as changeable as the weather, even the best of the election pollsters don't do any better than the weather people.
When you hear that a certain poll has a margin of error of 4%, what they are referring to is random error. That is the only error that can be calculated. We know that the chance of a coin coming up heads is 50:50. If you toss a coin 5 times and it comes up all heads(odds are 1 in 32), you can attribute that to random error. That does not mean that a poll that has a 4% margin of error will be wrong just one time in 25. It might be wrong one time in 25, just due to chance alone. It could still be wrong half of the time due to systematic error. If that is the case, you would do just as well tossing a coin as listening to the pollsters: heads it's Democrat, tails it's Republican.
If you take a gallon of a liquid and shake it up, then take out a hundredth of an ounce, it's a good bet that you will have a truly representative sample of what that gallon contains. Analyzing that hundredth of an ounce will tell you what that whole gallon contains. If you shook it well, the only error that comes into the picture is random error. If you consider 185,000,000 eligible voters, of whom half will never bother to vote, it's an entirely different matter. It's hardly a homogeneous group. If you polled all of the people in the New York Stock Exchange you would get a very different set of opinions from those in a homeless shelter in San Francisco. The critical element is being able to poll a representative sample. If the sample polled is not truly representative of the voting population, that is called systematic error and it can be very great. A slight error in the sampling can produce an immense error in the prediction. Polling several thousand people is polling one voter in 50,000. How will you know if you have a representative sample? Easy; if your prediction is right on the money, you have done it. If not, better luck next time. Just change the name of your organization and no one will be the wiser.
Bush beat Dukakis decisively in the last presidential election by getting 52% of the popular vote. What are the chances of polling 2,000 people, taken truly at random, and predicting the results of an election? Not very good. So the pollsters don't do it at random. They try to find a group of people who are representative of the voting population with regard to how they vote.
You really don't need a poll to know that Jerry Brown or Pat Buchanan are not going to be the nominees of their respective parties. As to who is going to win the election in November, it's anyone's guess. Throw Ross Perot into the election and it makes forecasting the weather look easy. If Perot gets enough Democratic and Republican votes, he could win the election. If no candidate gets a majority of electoral votes, the presidency will be determined by the House of Representatives, with each state getting one vote. The forthcoming election has to be the most unpredictable in history. Even when Henry Wallace and Strom Thurman formed two more political parties, it was still a two way race between Truman and Dewey. A split in the Democratic vote gave Lincoln the presidency and a split in the Republican vote gave Woodrow Wilson the presidency. This election will be the first real three-way race in American history. The pollsters must be climbing the walls.
In politics, no one trusts anyone else. It's not enough to rely on the Gallup or Roper polls; each party has its own pollsters. What are they supposed to do that Gallup can't? It certainly isn't just finding out what people think. Their goal is to influence public opinion --and they do. They, and the party's tricksters who will try to fix the election. So they will disclose a week or so before election day that President Bush has a Japanese mistress, Bill Clinton loves his mother and practices nepotism with his wife, and Ross Perot frequents the best little whorehouse in Texas.