The veracity of the polls done during this last election period has been the subject of much discussion. There have been assertions that some polls were done to suppress the vote. When a well-known pollster was asked about that, the reply was “There was record voting.” The bigger question is whether the polls are impossible to get correct in the current environment. When that was addressed to career professionals, much of what we have been told appeared to be made-up inaccuracies.
If you listened to the talking heads in the closing weeks of the campaign and after election day, they were largely mouthing the same know-nothing comment about the shy Trump voter. When all the talking heads are saying the same thing, it is usually a good time to check out what the facts are with someone who has some knowledge about the matter. That is when I called my friend Ed Sugar who spent 30 years in the industry.
Ed spent his time providing services to business communities. Many political pollsters provide their services to private companies. Mr. Sugar reminded me the industry has radically changed since the time he entered the industry, and it is one of the reasons he stepped out of a day-to-day position while maintaining his relationships.
To understand the devolution of polls you need to understand the evolution of their source information.
Since the 1930’s polling data was based principally on one simple point of contact – a landline telephone. This operated in a standard manner until 2012. You may think that smart phones are the cause of the problem, but the process was deteriorating long before these devices became commonplace. The first chink in the armor of the process was the ubiquitous adoption of answering machines (remember those?) in the 1980’s.
Then came car phones in the 1990’s equipped with caller ID. Next was the cell phone that fit in your pocket which eventually led to smart phones. The death of home landlines began at a precipitous rate beginning in 2005. Sugar warned his bosses to start reading the obituaries as every reported death represented a deceased landline along with the individual. As everyone knows virtually no person under the age of 30 has a landline at home.
It was not just the switch to cell phones, but the portability of numbers that made it exceedingly difficult to geographically locate where responses were located. Americans are highly mobile as are their devices making it ever more challenging to collect proper info since many surveys are geographically oriented by either city or state.
The most recent challenge came with changes to TCPA (Telephone Consumer Protection Act of 1991) which happened in 2015. The update made it even more difficult and costly to dial cell phones with a significant surcharge incurred by pollsters. The cost of telephone interviewing in the exurbs and rural areas is now more expensive than in metropolitan areas. This happened when there was significant migration to less populated areas as the locales became developed technologically; thus, more attractive.
Pollsters turned to the development of various sampling techniques around 2008 as a method to compensate for the lack of individual contacts and lack of hard data. They began utilizing forecast modeling to help make up for the decline in the number of available respondents/ease of random sampling. This means they were projecting results based on smaller samples to provide answers where they used to have more broad-based hard data.
I then spoke to Lynn Stalone who has 35 years’ experience in polling. Out of the box she stated “There is no right way to poll today.” She went on to say “There is very little opportunity to capture accurate information from cell phones.”
Lynn defined additional difficulties with gathering data from cell phones separate from what we have already delineated. Pollsters get their telephone numbers from service aggregators who derive the data from providers (AT&T, Verizon etc.). The information provided is based on billing data. A single billing record can easily contain phone numbers for multiple individuals, not to mention multiple households perhaps living separately or even in different cities. Then the pollsters may be looking for age specific information and reach a 64-year-old male while wanting to speak to the 25-year-old daughter who may not live in the same location.
“No call” lists do not apply to true political polling, but that does not help if all the factors above are the reality for the polling firms. They often do not know who they are calling and/or where they are located.
Companies like Aristotle provide call lists from voter records, but Lynn expressed a concern as to the quality of the voter records. Between the people who have relocated, start new households or are deceased, there is lack of integrity. That was enhanced in 2020 because of people moving all over the place due to COVID-19.
Some of the wounds for pollsters are self-inflicted. Over the years Lynn has seen slanted questions employed many times. Or there is a series of questions leading up to a predisposed outcome. This is as prevalent as sampling errors. The questions often support the idea that one candidate is going to be the winner and people have a natural affinity to backing a winner. Thus, they may indicate to the pollster they are supporting A when they ended up voting for B.
“The elephant in the room is the non-responders.” Many people just do not want to answer surveys and thus pollsters do not know what their thoughts are Lynn told me. In the case of national polling, Lynn believes there were the “shy” Trump voters, but in other cases there are similar shy voters.
Adding all these factors together, it is extremely challenging to get accurate polls. That is why the polls were off in many states in the presidential campaign, but we likewise saw situations like Senator Susan Collins losing in every poll before the election and winning by eight percentage points. As Democrat campaign advisor David Shor stated, “This was a bad year for polling.” The question is when, if ever, will there be a good year?
Lynn stated the pollsters are adjusting quickly, and working hard to get it right, but they face a daunting task. If you add up all the changes that test their established methods, it may be something they cannot overcome.
The bigger question is why we as citizens need to be told that candidate Jones is leading in the polls by six points. National news organizations thirst for this data while it does nothing for us. The polls really add nothing and can only skew the eventual outcome if people are inclined to vote for the “winner.” It is clear the campaigns may want their own polling, but why do we need the Washington Post/ABC national poll and many others like them?
With the changing environment and the uphill climb to get accurate data from reluctant participants, it might be time to put these polls to rest and rely on the only one that counts – election day.