Show What People Think Roper Lists Possible Defects in Polling By Elmo Elm Roper Last week I 1 listed the ten questions questions ques ques- on public opinion polling I have as asked ed th the special committee of or the social science research council to comment on This week I 1 should like to spell out in a little more detail the nature of each of those problems 1 Did wo we Interview the wrong people I II I Obviously its it's essential to lo interview interview inter inter- interI I view the right people But Bul who are the they The problem in inan an in dice elec election tion lion pl prediction is to try by one means or or another to estimate which par partisans parsa sa s arc are actually going going- to vote The universe we vc ate are attempt attempting attempting at at- tempt tempting ng to measure is the uniVerse uni uni- erse Verse of actual voters and we c are arc attempting to measure it before belore e they vote vole If we failed to do that tha t this year why Would a sample which would have correctly forecast forecast forecast fore fore- cast titUs this election have ha correctly forecast I the last three If It not where arc we 2 I Did ld wo we ask the theron ron Wrung wrong questions ques It i is important to know just how to ask a question in order to get g-et a n frank answer In our previous three nR national election polls for example we never asked the direct question as to whom the respondent intended to vote for fOl because we c found a reluctance on the part pait of some to admit that the they intended to vote for Pres Roosevelt This time we used the simple direct question question and and Underestimated underestimated un- un the Truman vote Was Vas this perhaps our mistake Were ere there people this year who were cic either cither reluctant or afraid to say ay that they were w were re going to vote vole for 01 Truman 3 Was ns the timing of or the polls poll wrong Did people change chanc their minds the last a t week week week-or or the last da day Some of my mall mail has suggested that my most spectacular mistake uras as as to assume that campaigns did not change the voters' voters minds Others have said that while its it's true trite campaigns dont don't change chang the I voters' voters minds If the campaigns are arc roughly of oC equal potency these two particular campaigns s were not of anything like equal potency Dozens of letters have come from people who said they changed chang their minds late in the campaign n because because because be be- cause Mr Dewey refused to discuss discuss discuss dis dis- dis- dis cuss I issues sues and seemed so above I it all aU One letter came from a i Pennsylvanian who said he helped I contribute to our error because he told our own interviewer he was going to vote for Dewey Dewey and and later changed his mind Was Vas this widespread ad among the voters 4 d. Were Vere wr in error in assum assum- aS assuming lIm- lIm log ing the dont know voto oto would divide proportionately among the candidates Cc In other words was a dont know answer one me way of evading evading evad evad- ing nl an open commitment to our interviewers or vers or or an any one else else else- that people intended to vote vole for Pr Pres Truman 5 Did strong local loca candidates candidate I generate enough enthusiasm for forthe forthe forthe the Democratic side Ide to offset mild enthusiasm for Dewey Its It's being argued In some quarters quarters quarters that in a state like Minnesota for example a large larg-e number of people were apathetic about bo both h I candidates for the presidency but butery very ery enthusiastic for senator senator elect elect 1 Humphrey Did they decide at the thelast thelast thelast last minute not to risk mak mah ng nga a mistake about the candidate they were enthusiastic about and thus vote vole Democratic down the line The same argument has been advanced advanced advanced ad ad- for Connecticut Illinois and other states stales 6 Did more Dewey than Truman Truman Truman Tru Tru- man supporters sta stay awa away from the tho polls If so did the polls I I 1 or and/or the press contribute to that result This of cour course c. c IK is the contention contention contention conten conten- tion of some Republican politicians politicians clans many of whom are aie quite bitter about the polls I think the question asked inthe Inthe in inthe the parentheses is Important We Ve have always taken the tho position that the polls did not influence the outcome The bandwagon theory seems to have been definitely definitely deli deli- exploded but maybe polls have sop some c other effect For example example example ex ex- ex- ex ample maybe the effect is on the candidates and the campaigns thc they decided to wage wag If so we and the people should know it it 7 i. Is them hero evidence that there was in our possession information indicating a Truman victory or victory or ora a clo close e race race and and we wc ml misjudged judged or Ignored It For example should we have been warned by the fact tact that a Fortune survey done late in October showed the people as a whole wanted the new deal principles prin principles prin- prin continued even though thc they said saM they wanted them continued under a R new ships ship's captain 8 Is b there evidence that election prediction cannot c. be considered a proper sphere for tor public opinion research in view of today's development development development devel devel- of tho the science and the size of the possible error S So So far as I know all people who have studied the matter ate are agreed that to find out what people think at a given moment or what they did or bought yesterday is one kind of problem but to try to translate their yesterdays yesterday's mood into next Tuesdays Tuesday's action is another another an an- other kind of problem I have asked for the committees committee's opinion whether or not the error was inas in as assuming that an accurate tool for measurement of past actions and present moods could also be used to measure future action inan in inan inan an emotion-charged emotion situation We Ve cannot afford to be too generous about the margin of error we ve were willing to allow ourselves either If election polls had been taken talen I during the last 31 national elections elections elec elee- Hons and had they erred In the direction of overestimating the losers loser's strength by as little as 4 c each time they would have been wrong 16 times right 15 times Since the public insists on re regarding regarding regarding re- re garding a published figure as a I prediction despite any hedging or qualifications the pollsters them them- selves make we ought to face up to the need for greater accuracy cy than that if we ar are to use uso the tho tool for that purpose 9 Is there evidence c that either cither or polls polls' serve sene any so social lal or scientific pur purpose purpose purpose pur pose This question admittedly gets away awny from techniques and perhaps perhaps perhaps per per- haps the committee may decide it doesn't care to comment on it I think the question has to be answered answered answered an an- some time however because because because be be- cause if IC election polls serve neither a social nor a scientific purpose there is nothing to justify their existence 10 Is I. there an any et c c that either or polls poll do nn any disservice to the thin democratic H system j tem stem To me this f Is the most Important important important tant question of oC all For example does the possession of information that he faces a terrifically call up hill battle incline a good candidate to wait vh R aWhile le before running running- and thus insure the nomination of party varty hacks on both sides Do polls cause delegates delegates dele dele- g-a g gates tes to spend all their tune time worrying wor- wor i lying ying about who could be elected instead of or thinking n about who might mig mal make e the best president I am not going to offer my answers answers an- an to these questions at this time although it would be silly for me to pretend that after the investigation we have done I have no theories about them Public opinion polls pons have become such an integral part of both business and every day life that the public Is entitled to the reasoned convictions convictions convictions con con- of a group of impartial men such mch as are arc on the social science research council committee commit commit- tee It is expected that their answers answers an an- vers S will be ready in a few Cew weeks and when they are aie I intend intend in intend in- in I tend to publish them and along with them them them-or or following shortly thereafter thereafter-my my own agreement or disagreement on each of the above points points points-or or any others they decide to discuss Copyright N. N Y Herald Tribune I |