By Dr. Philip Castillo
With the heightened interest in the upcoming elections in the USA, attention is on the many polls being conducted to predict its likely outcome. Basically, a poll is a small scale representative subset of the voting population and the aim is to study the subset and then make valid inferences about the wider population. That is the theory, but the practice has been challenging and in some instances, made intentionally so.
With a national population exceeding 350 million, the USA has over 150 million registered voters. However, voter turnout remains relatively low despite major elections occurring every two years. Therefore, pollsters focus not on registered voters, but on “likely voters”—those most likely to actually cast a ballot on election day. To determine who qualifies as a likely voter, pollsters use complex statistical models.
That is the first challenge faced by the pollster since any representative sample of voters will include multiple registered but not likely voters. In the USA, it is known that certain demographics display a higher propensity to vote than others. With that knowledge, pollsters weigh their sample and by so doing seek an enhanced resemblance between their sample and the actual voting population. This weighting process involves a level of subjectivity.
There are many other challenges. The USA presidential voting system is such that the winner is decided not by the popular vote but by obtaining a majority in the Electoral College. So, then, if you vote in New York, for example, the winner in that state obtains all the electoral college votes from that state. With this system, the candidate who obtains the most votes nationally may not necessarily win the elections. That was how Donald Trump became President. While national polls have their benefits, the real focus is on state polls, most notably in states deemed to be competitive – so called swing states.
That poses additional challenges but more so in the USA, where there are hundreds of polling firms and polls are generated daily. Because of these many polls, some of which show quite contrasting and contradictory results, what is increasingly being done is aggregating polls and averaging results from these aggregated polls. The obvious subjective decisions that arise become what polls to include and what weighting to assign to those polls so included.
It is in this regard that there are many other deliberately and maliciously induced challenges for the polling analyst. The billions of dollars involved in the American presidential elections means that there are many partisan pollsters who produce many polls in favor of their preferred candidate. These results when aggregated skew the averages in favor of one candidate. That is their likely aim because the bandwagon effect thereby created is intended to sway undecided voters, who are more likely to gravitate towards a winner.
The above challenges notwithstanding, polling seeks to quantify uncertainty. Indeed, the many acknowledged challenges with polling both as an art and a science exist and persist, but polling continues to evolve to address these problems and the objective pollster applies lessons learnt from the past to improve the next poll. A pollster is only as credible as his last poll.
But there is another type of analyst in the USA. These types eschew polls. They look at what they regard as economic fundamentals, candidate charisma, voter enthusiasm, number of donors and donations, number of campaign volunteers and the quality of the get out the vote ground game. Many of these variables are included in what is referred to as the 13 keys to the White House. This is a subjective prediction system for determining the winner of the US presidential elections. This perspective argues that American voters select their next president according to how well the country was governed in the preceding four years and that election campaigns have little, if any, meaningful effect on American voters.
Based on the quality polling being undertaken, the race is essentially tied. A statistical tie means that the results are within what is regarded as the margin of error. For those not well versed in statistical jargon, a simple explanation for the margin of error is that as long as you are analyzing a sample drawn from a population, there will be always be a discrepancy because, try as you might, the sample – though a subset of the population – will not exactly mirror the wider population in all aspects. The discrepancy between the sample and the population can be objectively measured and calculated and that is your margin of error.
Importantly, in election polls, that margin of error applies to each candidate’s vote share. So, if a poll with a margin of error of ±4 points shows Vice President Kamala Harris at 50 percent and former President Trump at 45 percent, Ms. Harris’ true level of support could be as low as 46 or as high as 54 percent, and Trump’s could be anywhere between 41 and 49 percent. In other words, Harris could be leading by as much as 13 points, or Trump could be leading by as much as 3. It’s still more likely that Harris is ahead in that case, but her lead is within the margin of error.
As a pollster keenly following the published results of the quality polls being done in the USA presidential elections, AND also concurring with the legitimizing narratives of the non-polling analysis, my unsolicited perspective is that Vice President Kamala Harris will win the 2024 Presidential Elections. And it won’t even be close.
Send comments to pjcastillobz@gmail.com