In the 2022 U.S. midterm election, Republicans ended the Democratic majority in the House of Representatives by a thin margin, after waiting in anxiety for an agonizingly long period. The advantage of Republicans fell quite short of the expectations of the election polls.
A Senate runoff in Georgia in December will not decide which party controls the chamber this time, as Democrats have already secured 50 seats. Even if Republicans win the runoff, Democrats still maintain their slim majority, as Kamala Harris can break the tie if necessary. Liberal media celebrated the results as Democrats successfully staving off the “red wave,” because the easy Republican victory predicted by pre-election polls did not materialize. President Joe Biden boasted that red wave “didn’t happen” and he was “incredibly pleased.”
Republicans, on the other hand, were quite disappointed. According to Statista, the polling in the days before the 2022 United States midterm election, by more than 40 news agencies and political research institutes, a generic congressional vote between the Democratic and Republican parties reflected a lead for the Republicans, who led Democrats by 2.5 percentage points. Why did the hugely anticipated red wave not pan out? Why were the election polls off the mark?
From the political perspective, some analysts pointed out that Republicans did not live up to expectation because some of their candidates were far-right minded or endorsed by Donald Trump. Although these candidates performed quite well during the primaries, they did not appeal to nationwide independents and moderates in the midterm.
But we also need to take a closer look at election polling itself. Election polling can be thought of as the vote before the vote, a prediction of the upcoming actual results based on opinions of selected voters. It is a gig originating from the media industry to attract and boost readership and viewership.
In 1920, starting with the election of President Warren Harding, a weekly magazine called The Literary Digest correctly picked the winner of each subsequent presidential election up until 1936. The Literary Digest conducted the election polling by soliciting opinions from its huge base of subscribers, one out of four voters, representing every county in the United States, as claimed by the magazine.
In 1936, The Literary Digest polling showed that Republican Alfred Landon was beating incumbent Franklin Roosevelt 57% to 43%, a forecast that totally missed the point. In the most lopsided general election in U.S. history, Roosevelt won all but two states with 60% of the popular vote. Poor sampling was to blame for this polling embarrassment. The subscribers of the magazine were mostly middle to upper class, totally unrepresentative of the general voters. Despite the tallies of more than 2 million subscribers, the sampling was skewed towards wealthier Americans of the time.
George Gallup, with the first scientific political survey ever conducted, predicted the winning of his mother-in-law in Iowa in 1932. His breakthrough moment came in 1936 when Gallup picked Franklin Roosevelt in stark contrast to The Literary Digest. By systematically sampling the public to success, Gallup quickly gained credibility, rising to the top of the new field of public opinion prognosticators. Gallup also had his share of debacle in the presidential polling of Truman versus Dewey in 1948. But his math-based polling approach continued to be honed by statistics until it began to utilize random sampling techniques. Random sampling aims to collect pertinent data without any conscious or unconscious bias: Each sample has an equal probability of being chosen.
When used in election polling, random sampling, despite its smaller sample size, turned out to be a more accurate predictor. Randomly selected voters, from the streets or a phone book for instance, can be pretty representative of the general voting public.
Random sampling has been widely employed for election polling and other polling purposes. In the early days, election pollsters collected opinions from random passers-by on the streets, and later mostly through telephone calls. Election polling through telephone worked quite well until recent years.
Although the sampling has been meticulously designed, today’s pollsters have difficulty getting through to the call recipients they want. With the help of caller ID, people are screening calls more frequently to avoid telemarketers, conmen, advertisers and others not in their contact lists. Younger generations prefer messaging, emojis and other new ways of interactions on various social apps.
Therefore, pollsters are no longer getting the desired random sampling. People actually responding to polling calls are no longer collectively a random sampling: They can be biased towards a certain age, education, income or other inclinations, thus making the polling not as random or scientific as originally engineered. Math and science will advance to make election polling more accurate down the road, but right now people are justified in distrusting election polls. It would be wise for politicians not to give too much credence to election polls.
(The author is an independent financial investor.)