ま だ 生 き て ま す 。

   In today’s information-rich and prediction-loving society, random selection has become so important that it is now done routinely by computers. These days, it is rare for anyone in the business of checking the effectiveness of drugs or testing public opinion not to be aware of the importance of using a genuinely random sample.

   It was not always so. In the U.S. presidential election of 1936, Republican Alf Landon was challenging the president then in power, Franklin Delano Roosevelt, a Democrat. In order to try to predict the outcome, the Literary Digest magazine conducted the biggest election poll ever, sending out over ten million question papers. When the results came in, they predicted that Landon would win by a huge majority.

   During the same election, a polling expect named George Gallup also conducted a poll.

His survey was much smaller than that run by the Literary Digest. He questioned a mere fifty thousand voters, on the basis of which he predicted a victory for Roosevelt.

   Roosevelt won. Gallup, with his poll of fifty thousand, was right; the Literary Digest, with its poll of ten million, was wrong. What made the difference?

   The secret of Gallup’s success lay in randomness. He selected his fifty thousand voters at random from the entire population and went out and asked them all how they would vote. The Literary Digest relied on its own subscriber list, club membership lists, and telephone books. As a result, its sample was anything but random. In particular, it excluded most of the poor, who were unlikely to subscribe to the Literary Digest, belong to clubs, or, in those days, have a telephone.

   Another problem for the Literary Digest poll was the response rate. It was a mail-in poll, and only 20 percent responded. For the most part, they were the people who were angry with Roosevelt and who wanted to see him replaced by Landon. By and large, the, the people who were content with Roosevelt did not bother to reply to the poll request.

   “What are the qualities of a good poll?” asks Bill Kaigh, a polling expert and a professor of mathematics. “The very first questions you need to ask are these. Will the survey truly reflect the opinions of the whole population? Is there a randomization involved? Did the survey select the sample, or did the sample select the survey? That is to say, did the questioner make sure that all the people he selected actually replied, or did the people who were questioned decide for themselves whether or not to reply?

   In the Literary Digest poll, the sample selected the survey: only the people with strong anti-Roosevelt feelings responded. For the same reason, the call-in polls you often see on television are not reliable. For a poll to be accurate, the sample must be chosen randomly from the whole population.

もううんざりだ。ルーズベルトとか日本史でやったよもう・・・




最後まで読んじゃった人。
僕の代わりに受験してください。