CFIF.org
Published: 10-27-08
The race for the Presidency is over finished in the bag for Barack Obama. The polls say so. The pundits amplify the polls. Why it’s going to be a landslide of epic proportions.
Oops! Not so fast kemosabe. At midweek three major national polls had the race at a statistical tie and some state polls were beginning to scurry about yet again. Next week who knows? The week after that well therein lies election day – the only poll that counts.
Since we are in the final stages of a presidential campaign at the same time that we are in God-only-knows what stage of a global economic meltdown it is perhaps wise to remember the words of Charles Mackey: “Men it has been well said think in herds; it will be seen that they go mad in herds while they only recover their senses slowly and one by one.”
In 1841 Mackey wrote a seminal investigation of crowd psychology and mass mania entitled “Extraordinary Popular Delusions and the Madness of Crowds” which studied among other follies several financial bubbles.
Most public pollsters are honest and work hard at their discipline. That’s far from easy. As Michael Barone wrote in The Wall Street Journal this week “political polling is inherently imperfect.” No political observer is better positioned to know. Barone who describes himself as a “recovering pollster” is co-author of “The Almanac of American Politics” and perhaps the only living American who has actually been in every voting district in the country. Barone believes that polling can be trusted but with “qualifications.”
Think about the most glaring imperfection. In the 2004 presidential election voters set a turnout record – 126 million. Given the “madness of crowds” that seems to have gripped the electorate in 2008 that record is almost surely to fall heavily.
Against that backdrop consider the unenviable job of the pollster who must find question and evaluate the responses of a valid sample of Americans. Those samples typically range from about 600 to 1500 citizens. Those tiny numbers of people polled produce the overall results with far tinier numbers representing the segments and sub-segments of a poll’s population.
The design of the overall sample is critical. If it’s just of voting-age Americans the poll may be interesting in terms of popular sentiment but is next to worthless in terms of evaluating elections. Polling registered voters is far better but far from conclusive. Polling “likely voters” using a screening process to arrive at that designation should all else being equal produce the most accurate results.
Nearing the end of an election “national” polls can be quite misleading because we actually vote in separate state elections many of which are virtually predetermined based on the political make-up of those states so polls in “battleground” states (which are themselves undergoing some remarkable shifts this year) take on a greater relevance.
At their most accurate polls have validity only as a brief snapshot in time of the inclinations of the pollees. One of the tied polls mentioned above saw a seven-point swing in one week. Some elections are fluid; this one is volatile driven more by emotion than careful consideration and emotions swing wildly. Emotions are also exacerbated by the polls themselves.
Pollsters are struggling desperately to achieve reliable weighting of their samples. Overweight or underweight any demographic group and that suspect sample leads to suspect if not misleading results. Even when polls are averaged which is becoming a major analytical tool given the sheer number of polls being conducted the average produces at best a lagging indicator of reality. Whatever the accuracy it is behind the reality which may or may not change rapidly.
This year there has been much discussion of the effect on polling accuracy by for example the almost exclusive use of cellphones by young voters or the so-called Bradley Effect in which voters may say they are going to vote for the black candidate but don’t. The truth is that no one knows.
The best advice to voters (which few will actually take we can reliably report without benefit of polling) is to study the candidates and their positions and absolutely ignore the raucous political sideshow that polling itself has become.
That’s just one way to avoid going mad in herds and “recover
