Veterans’ Day is one of those half-baked holidays – the stock market is open but the banks are closed; here in Washington, federal employees have the day off but federal contractors don’t, unless they work in a federal building that’s closed on Veterans’ Day, and so on and so forth. So normally, I would have expected a somewhat lighter schedule, but given the events of last Tuesday, that was far from the case. All my phones have been ringing like the bells of a nine alarm fire since shortly after 12:15 AM Wednesday, and I was booked solid from before sunrise until well after sunset, although it is, admittedly, November and the daylight here does grow increasingly limited this time of year.
Gretchen, like most Washingtonians, has been in shock since Hillary Clinton lost Florida. DC, a liberal enclave where over ninety-five percent of the population voted for the Democratic presidential candidate, has been a city of blank faced, staring zombies all week. With record numbers of people inside the Beltway calling in sick (i.e., drunk, depressed, on massive doses of Xanax, Valium, etc.) the last four days, I give her credit for nevertheless continuing to show up for work and stoically doing her job. Like every good Amish farm girl, she knows the cows have to be milked no matter what the weather, and that includes hurricanes and tornadoes.
No matter what the original subject of each consultation scheduled prior to Tuesday may have been, since Wednesday morning, the only thing I have discussed during any appointment, whether regular or ad hoc, has been the various implications, ramifications and potential consequences of a Trump presidency. With the exception of the Russians, who have been smugly celebrating all week, the international diplomatic community here in Washington is in a complete tizzy of uncertainty and confusion. Federal contractors are flummoxed beyond reason; they have no idea what the new administration is going to do with their existing projects. Despite being out on congressional recess, Democratic representatives and senators have frantically dispatched their staffers hither and yon in search of viable strategy options. A rare exception to the rampant state of stunned catatonia, these confused minions have been running around town like chickens with their heads cut off, desperately picking the brains of anyone who will talk to them for clues about how to react to what, for them and their bosses, amounts to nothing less than a political cataclysm. Federal employees, are, as would be expected, universally anxious about their cushy civil service and SES careers and little if anything else, even the ones at DoD. And truth be told, not everybody at the Pentagon is all that delighted with the prospect of having a sociopathic bully assume the office of Commander in Chief.
The first burning question on everyone’s lips is how this could have happened, of course – nobody can quite believe that the American electorate contains enough bigoted, xenophobic, ignorant, misogynist, racist morons to place such an obviously deranged monster in the White House. On that point, I would recommend they consult the works of Alexis de Tocqueville. The second question is, how come the public opinion polls all said that Clinton was going to win?
Puzzling out the answer to the second question is proving the more perplexing challenge, by far – it’s not like the pollsters know. Of that, I am certain, because my two o’clock appointment on Friday was with Dr. Percy Minkukel, Board Chairman of the American Political Opinion Institute, an association of polling organizations headquartered here in Washington.
“Tom,” he exhaled with an exhausted air as he plunked down wearily on the couch in front of the picture window overlooking the White House, “I’ve never been so humiliated in my entire life, and that includes when my fiancée left me at the altar and eloped to Tuscany with her amateur women’s soccer team goalie.”
“Understandable,” I observed. “As late as Monday night, you and your colleagues had everyone in the Establishment pretty much convinced Trump was toast. The Outlook section of the Sunday November 6 Washington Post even had a headline article speculating on how The Donald would deal with such a crushing defeat. Your polls had all the media pundits, anchor-persons and commentators fooled too, it seems. Now, of course, they’re pointing their fingers at the pollsters, vociferously complaining that they don’t do math, they just report and analyze the results, and it’s completely your fault that the numbers they were nattering about for months were complete horse hockey.”
“I know,” Minkukel choked out, fighting back against what would be a decidedly unmanly and unprofessional surge of tears, “and as Chairman of the APOI, my name is Mudd.”
“Perhaps we should begin,” I ventured, “with an exploration of how you and your colleagues in the public opinion poll sector managed to get everything so spectacularly wrong for so astoundingly long.”
“Make no mistake about it,” he muttered. “We’ve been burning the midnight oil every minute since nine PM Tuesday trying to figure that out.”
“And what results have you obtained so far?” I inquired.
“Actually,” he replied, “the general consensus among pollsters at the moment is that we were right, but everybody else misinterpreted what we were saying.”
“Interesting,” I dryly responded. “Could you explain the reasoning behind that conclusion?”
“Yes, um,” he stammered, “my colleagues explained to me that ah… well, by the week before the election, the standard error limit for the presidential polls had expanded to a value larger than the separation of the estimated vote percentages for the two major presidential candidates.”
“By what amounts?” I pressed.
“Uh… well,” he related, “while the difference in the estimates for Clinton and Trump differed by two percentage points, the um… the error in the measurements totaled plus or minus seven percent instead of the usual three percent.”
“Thus,” I sought to verify, “rendering the presidential poll numbers useless for prediction of the election outcome.”
“In fact,” he elaborated, “even if the error had only been three percent either way, under those conditions, the poll numbers would still have been useless to predict the outcome.”
“And you expect everybody to believe a load of statistical malarkey like that?” I asked.
“The APOI thinks it would be mighty nice if they did,” he answered.
“But of course,” I noted, “if you pollsters go around telling the world that you can’t actually measure public opinion accurately enough to predict something as important as the outcome of a presidential election, how are you going to get anyone to pay you money to determine, say, for instance, whether the public favors margarine over butter?”
“We can’t,” he sighed. “And that’s why I’m trying to discourage my colleagues from explaining it like that. The trouble is, not all of them are listening to me.”
“Oh, really?” I said. “Who isn’t?”
“First of all,” he ruefully revealed, “there’s Sam Wang over at Princeton. He told me he’s going to use that explanation when he goes on CNN to fulfill his pledge to eat a bug if Trump won the election.”
“So,” I surmised, “Dr. Sam Wang is going to go on cable news, present an argument that tells Americans they’re idiots because they don’t understand statistical error analysis, and then demonstrate his intellectual superiority by eating a bug?”
“I hear,” Minkukel added, “that afterward, he intends to encourage everyone to focus on serious issues instead of complaining about silly things like the accuracy of public opinion polls.”
“Have you and your colleagues managed,” I wondered, “to come up with any less asinine explanations for your profession’s massive failure in the 2016 election?”
“One possibility we’ve been discussing,” he admitted, “is that maybe some people were lying about who they were going to vote for.”
“Lying about voting for Trump, presumably?” I remarked.
“Well, yes,” he nodded, “I mean, it’s hard to imagine people lying about whether or not they were going to vote for Hillary Clinton. But with all the… um… provocative things… that Trump said during the campaign, we think it’s entirely plausible that a significant fraction of the electorate might have been predisposed to prevaricate about their intentions, particularly if asked in a context that wasn’t explicitly anonymous.”
“Such as being called at home on the telephone,” I suggested.
“I suppose these days,” he agreed, “in such a situation, most folks would be concerned about maintenance of their privacy.”
“Concerned,” I speculated, “that if Hillary Clinton won the election, as everyone expected, they might face some sort of… reprisals… for saying they were going to vote for Donald Trump?”
“I don’t know,” he declared, throwing up his hands dismissively, “yeah, sure, something like that. Scared their boss might fire them, or their relatives would stop speaking to them, or their neighbors would throw toilet paper all over their house in the middle of the night or something like that. I mean, hell, look at the way the Clinton supporters have behaved since Tuesday. They’re out in the streets burning Trump in effigy and shouting obscenities at the guests in Trump hotels. Holy smokes, Tom, God help the poor schnook who tells his wife now that he voted for Trump after she voted for Clinton. You know that guy’s not getting any nookie from her for the next four years, at least!”
“Okay, then,” I allowed, “let’s suppose the reason you pollsters all looked like a bunch of incompetent cretins in 2016 is because a lot of people you polled lied to you. What measures have you identified to deal with that phenomenon in the future?”
“Um… yes… well, that’s a real problem there, Tom,” he averred. “You see, the entire concept of public opinion polling sort of depends on the assumption that if you ask someone, would they purchase green tea ice cream if it was available in the freezer section of their local supermarket selling at the same price as vanilla, chocolate or strawberry, they’re going to tell you the truth.”
“But aren’t there ways,” I asked, “of determining whether a person tends to answer falsely?”
“There are,” he affirmed, “but you have to ask the subject a lot of questions and examine the answers relative to a baseline, and even then, you can’t tell which specific answers are lies. All you can really do is say their responses establish a pattern that justifies disregarding all of their responses.”
“So there’s no practical way,” I concluded, “to identify the voters who will be lying to you in the next election.”
“Not really,” he confessed. “I guess what we could do next time is first conduct surveys with questions aimed at determining how embarrassing it would be to say you’re voting for each of the candidates and then use the results to estimate what percentage of the voting population will be lying about who they will be voting for.”
“Sounds pretty convoluted,” I opined.
“No doubt about that,” he concurred. “It would probably take several elections to properly calibrate the correction parametrics.”
“Any other ideas?” I persisted.
“Well,” he mused, “it’s always possible that we weren’t adequately sampling the correct populations.”
“Meaning what?” I queried.
“It’s just that, uh, say, in the 1948 election for example,” he explained, “the pollsters all had the Republican, Thomas Dewey, ahead of Harry S Truman by a landslide. But Truman won by a very respectable number of votes. Afterward, the reason for the polling mistake was discovered – the pollsters had conducted their polling almost exclusively by telephone, and in 1948, Republicans who owned telephones outnumbered Democrats who did by a substantial margin.”
“So you’re thinking,” I guessed, “that in 2016, you guys should have been using more Internet, more Web, more social media, more smart phones… that sort of thing?”
“Or less,” he groaned with an air of frustration. “Who knows?”
“Suppose the average Trump voter was more likely to respond truthfully during an in-person interview?” I suggested.
At that, a hideous scowl crawled across Minkukel’s face.“Jesus Christ, Tom, do you have any idea how much more expensive collecting accurate data would be under those circumstances? We might as well hire droves of temps at minimum wage to go out and count yard signs!”
“If you had done that this time,” I noted, “you would have seen that Trump was going to win.”
Minkukel slumped forward, burying his face in his hands. “It seems as though the world has gone completely mad,” he moaned.
“You certainly don’t have to be a pollster,” I assured him, “to feel like that.”
“What can we pollsters do,” he implored, “to rescue our profession’s reputation after this kind of ridiculous debacle?”
After pondering his plea for a few moments, I gave Minkukel my recommendation. “How about this – henceforward, every public opinion poll where the limit of error on the estimated percentage for any one of two or more alternatives exceeds the difference between any two values; and / or where the values of any one of those measurements cannot be conclusively demonstrated to be unaffected by the false replies of survey respondents, must be accompanied with a statement, similar to those on tobacco and alcohol products, which reads, ‘APOI Warning: The results presented in this public opinion poll are not statistically significant and should be used for entertainment purposes only.’ Think that might do the trick?”
Minkukel sat bolt upright and glared at me. “Public opinion polls are not a form of entertainment!” he bristled.
“Correction!” I exclaimed. “What you mean is, before 2016 they weren’t!”