For Whom the Vote Polls

Every national election, especially those which include a presidential election, has a distinct character in the demographics of its voters that contrasts with previous national elections.

Analyses of these changing demographics are most accurately performed after an election when actual voting numbers are available. But if changing voting patterns could be reliably estimated before an election, that information would be a very useful tool beyond the traditional polling techniques.

In those instances when the dynamics of a changing electorate are especially significant, traditional polls tend not to capture the full scale of the shift taking place. We need to look to other indicators for a hint of future results.

The 2024 elections are just such a cycle.

In the past, the most commonly used poll was a national poll of voters asking their choice of candidate, party or issue. In regard to a presidential election, this increasingly has proved to be a crude and even misleading measurement because, even if accurate, these polls measure national popular vote, and not necessarily the vote in the individual states whose electoral votes actually determine the outcome of the election. In any reasonably close election, which most recent ones have been, this is critical.

In other words, a Democrat might lead a Republican presidential nominee by as much as 3 or 4 points in a national poll and still lose the election by the electoral votes from the states. In recent years, large states such as California, New York and Illinois have huge majorities of voters who are Democrats, and any national poll reflects this bias. In fact, Republicans have recently won two presidential elections while losing the popular vote, in 2000 and 2016.

The only polls, then, which are truly useful are state polls. But even these face the same technical problems which all polls face today — statistical accuracy in the face of growing public unwillingness or availability to be polled. More and more voters decline to answer or are not available to polls. There is also significant variation in statistical accuracy between polls conducted in person, by telephone, or by the internet. This is further complicated by the reluctance of more conservatives than liberals to trust and participate in polling.

At the end of each poll, the pollster lists a margin of error, plus or minus a number of points. In the old days, when voters readily cooperated with pollsters, this was reasonably accurate. But today, with all the problems mentioned above, the margin of error is often not what the pollster claims, but in reality, in the double digits — which means such polls are of little or no use.

Further, the raw numbers of most polls are adjusted using “models”, usually based on the pollster’s subjective assumptions about the voters, in an attempt to achieve a more accurate approximation of the electorate than is reflected by whomever happens to respond to the poll. The results are thus “weighted” using demographic data, and in recent years, this has often critically distorted poll results.

In 2024, there is considerable evidence, much of it anecdotal, but some of it from polling, that there is a significant shift in voters’ party and/or candidate preference by demographic or identity groups which make up each party’s base of voters. This means that “models” and “weighting” based on past elections may be especially difficult to formulate this year.

Polls of subgroups (such as by race, religion, or ethnicity) are not always reliable, so the paradox remains: polls are necessary but not dependable. It is true that the larger the poll sample, the more accurate it is likely to be, but most mass media pollsters often cannot afford such large samples.

A key further refinement between likely voters, registered voters, and all voters is crucial, but also drives up the cost for the pollsters. Only polls of likely voters are useful, especially later in the campaign.

Most campaigns hire pollsters to gauge voter feelings about certain issues and see how they are doing against their opponents at various stages of the campaign. These polls, which are rarely made public, are usually more accurate than public media polls because the pollster is paid more for them, and so has the resources to contact the number of voters to achieve more accuracy.

In each cycle, a few pollsters innovate in their polling method and often have better results.

All of this has been well-known for several recent cycles, but is even more relevant in the very idiosyncratic 2024 cycle. Many voters are, this cycle, in open revolt — faced not only with controversial ballot choices, but also living with economic inflation, open-border crime and terrorism, international uncertainty, and numerous domestic pressures.

The only objective measure of the voters’ temperature so far is the limited results from the early state primary elections. Primary voters traditionally are activists as well as those motivated by close contests. A dearth of the latter has limited what we know about this year’s general election voter, yet there has been some reinforcement of the assumption that certain voter groups are shifting.

But how much are they shifting?

This is the question which remains unanswered by conventional polls. The answer to that question would arguably reveal more about who will win and who will lose when the votes are counted than polls which claim to tell us Donald Trump’s and Joe Biden’s current popular vote numbers.

Because black, Hispanic and Jewish voters have recently voted overwhelmingly for Democrats, only a relatively small percentage shift by these voters could have a decisive impact on the 2024 election.

That is because they are concentrated in the urban areas that give Democrats enough votes to overcome Republican majorities in the rural and other areas of battleground and purple states. Conversely, a small shift of suburban women in these areas could turn the tide to Democrats.

Many pollsters are a step or two behind in asking the questions which might tell us about the current national election. Voters need to know that many polls are tools of manipulation rather than tools for understanding.

In 2024, as was also especially true in 2016, those who want to understand the electorate and what it will do in the actual election will need to look beyond the polls for the anecdotal evidence that is more dependably telling us what will actually happen.