OCR Text |
Show GENERAL PJ-j . HUGH S. aa? Johnson M Jour: Washington, D. C. PUBLIC OPINION POLLS This is a piece about these dope-ster dope-ster polls that tell you all about elections elec-tions before they happen. I don't believe In them. Their accuracy depends too much on their timing, the way they are conducted and the wording of the questions. They can have great effect to sway voters. The conduct of them is a private enterprise en-terprise for profit There is no assurance as-surance of bipartisan or neutral control of their conduct. As campaigns cam-paigns reach climaxes there are few Americans whose sympathies are not aroused. It is human nature for the wish to be father to both conviction con-viction and conduct among poorly i paid enumerators also. These polls do not register secret ballot. They do not even register a sample "yes" or "no." They are frequently built up on a scries of more or less technical questions, sometimes so framed as to persuade a particular answer which the enumerator enu-merator Interprets. In view of their great influence, it seems to me that all polls require both investigation and regulation in the public Interest. They certainly do not deserve to be swallowed whole. I don't dispute their possible value (1) as some indication of shifts and trends in Dublic thinking, and (2) as an even better indication when actual actu-al choice closely approaches, and the simple question is "Do you favor X or Y?" But the folly of making conclusions conclu-sions on these polls Is apparent. Most people don't have time to study them. A flat statement such as Dr. Gallup recently made that his study showed 499 electoral votes for Roosevelt Roose-velt with 42 states and 32 votes for Willkie with 6 states, sounds almost as Impressive as the 1936 election returns when Mr. Roosevelt carried 46 states. But, however inaccurate, except as to trend, an examination of this and other polls shows a surprising reversal of popular opinion. In 1936, Mr. Roosevelt carried 27.4 million voters to Mr. Landon's 16,6 millions almost 66 per cent or a majority of 10.8 millions. The poll of American Amer-ican Forecasts Inc. a competitor of Dr. Gallup's predicts 23.7 millions for Mr. Roosevelt and 21.2 millions for Mr. Willkie a 53 per cent majority, ma-jority, or only a 2.5 million majority. Both poll conductors insist on a 3 per cent to 4 per cent margin of error as experienced in their own performances. Both also report very large "undecided" groups and surveys show that as many as 27 per cent of Mr. Roosevelt's current supporters are not definitely sure that they will vote for him." Consider Con-sider the states shown on the Gallup poll with less than a 4 per cent majority ma-jority (Dr. Gallup's claimed margin of error) for Mr. Roosevelt all of which are shown on the Dunn survey sur-vey as safely for Willkie. Consider also that the American Opinion Forecast Inc. poll is 2 per cent less favorable to Mr. Roosevelt than Dr. Gallup's. Consider both of these. I come to no such conclusions. I don't trust these polls either way. This is going to be a very close election. elec-tion. The Democrats are dumb if they are lulled by this sweet news. The Republicans are quitters if they become defeatists because of any such cheap stuff. The political practice about these modern polls is like Shylock with Portia, to praise them when their forecast is favorable and curse them otherwise. I can escape that charge. I have long regarded all these "sampling" polls as a public evil capable of vicious abuse. They won't tell their exact method. They resist investigation that would reveal re-veal more than their "general principles." prin-ciples." These are not enough to determine whether in the science of mathematics they stand even on a sound formula of probabilities. They refer to their record of accuracy. ac-curacy. Sometimes it has been remarkable, re-markable, but since they do not claim accuracy within 3 or 4 per cent and many an election has turned on less than that, it is not very convincing especially since they are very coy in reporting the actual number (not the percentage) or unaecmea answers oy location. That clouds their whole result There is a survey in this country based on a different method than "sampling." It is called the Dunn Survey. It has been conducted lor years by a scholarly, retiring sort ( fellow. It is not a poll taken &y part-time agents on a theoretically selected sampling. It is scientific analysis of several factors. There has been no attempts to commercialize commer-cialize or publicize it. It has been far more timely, or at least it has forecast results far sooner, than other polls. It is too ear.y for this column to repeat its 1936 stunt "Landon may carry four states, he is only sure of two," but I don't think we are going to hear any more of the Gallup poll after this election than of the Literary Liter-ary Digest poll after 1936. This, of course, is my own opinion. If it becomes appropriate to eat these words, I shall do it as gracefully grace-fully as possible, but I didn't have any literary indigestion last time and I don't expect to suffer galluping consumption in 1940. |