With a US Presidential election now less than three weeks away, some in Israel took note when a recent news article in an Israeli news outlet (here) reported that US Presidential candidate, Joe Biden, led US President Donald Trump by 14 points in the latest nbc/wsj pre-election poll. Since most Israelis appear to favor US President Donald Trump over Biden (here), such a US poll result means something in Israel.
Since that article, more US polls have begun to show a widening lead for Biden. The realclearpolitics US poll average now (as of October 14, 2020) shows Biden with a full 10% average lead--across all polls in the US. This suggests that Donald Trump's re-election hopes are probably doomed. Is this true?
One reader of that Israeli article cited above didn't believe the nbc/wsj poll results. In a comment to the article, this reader wrote that the poll was biased because it showed that 46% of poll respondents were Democrats, while only 35% respondents were Republicans. This 11-point gap in respondent Party affiliation suggested that, even before poll phone-callers began dialing to respondents, Biden appeared to have a built-in advantage.
Why? Because in America today, voters are extraordinarily polarized. Something like 84%-86% of Democrats appear to be hard-core Biden 'advocates', and a similar percent of Republicans appear to be hard-core Trump supporters.
Such polarization predisposes Democrat respondents to prefer Biden exclusively. It also suggests that Republicans can be expected to remain equally loyal to Trump.
There appears to be no middle road for voters in today's America. There could be very few voters of either Party who will next month choose to vote for the candidate of the other Party. At least, that's what pundits and politicians believe.
Therefore, given the extreme polarization that exists between Democrats and Republicans in the US during this current election cycle, an 11-point poll oversampling (in favor of Democrats) could indeed suggest a wide gap in voter preference. Not surprisingly, this is exactly what the poll showed.
But was that Israeli reader (who had pointed out this 11-point Party affiliation gap) really correct? Did nbc/wsj pollsters really tilt respondents chosen for this poll so heavily towards Democrats as to set the stage for a heavily pro-Biden result?
We got an answer to this question from nbcnews itself. Some 6 days after the nbc/wsj poll came out, nbcnews confirmed that there had indeed been a weighting bias in this poll. The bias favored Biden. But that bias was not, as the Israeli reader had written, an 11-point Biden advantage. This nbcnews article called that bias a 9-point advantage.
From what most US poll readers knew from past poll reports, this 9-point advantage was greater than the more "normal" US polling tradition of showing a 5-6 point poll oversampling in favor of Democrats. Then, just to give further confirmation to this oversampling 'incident', another article (here) declared that the oversampling of Democrats in this poll had been 8%.
To understand how serious such a poll oversampling is, consider that, within Gallup pre-US election polls between early March, 2020 and September 13, 2020, the differences between Democrat-Republican Party affiliations fluctuated between 0% to 7% (favoring Democrats), depending upon the poll (here). Most Gallup polls during this 5.5 -month period carried a 3%-5% advantage--to Democrats. Compared to these numbers, a suddenly appearing 8% or 9% advantage is, shall we say, unusual.
The differences may appear small. But they aren't. If, say, Republicans represent 26% of poll respondents, and Democrats represent 34% of total respondents, this 8-point difference represents a 30+ percent advantage for Democrats (that is, an 8-point advantage for Democrats translates into a 30 percent advantage for Democrats--because the "34" number for Democrats is 30% more than the Republican's "26" number). A 9 point advantage (between, say, 26% and 35%) represents a 34+ percent advantage for Democrats (that "35" number is 34% bigger than "26").
With potentially 130 million Americans voting in November 2020, what would a 30%-34% pre-election poll advantage mean for an actual election result? Well, a poll is supposed to give one a 'sense' of how final voting results will end up, right? So, an 8-9 point poll advantage seems to suggest 30%-34% more votes for Biden.
If Biden really did get 30%-34% more votes than Trump in the November 2020 election, that would mean anywhere between 25-to-40 million more votes for Biden. That kind of election result would indeed be termed a "landslide" for Biden.
So far, no one in America's mainstream media claims a landslide for Biden. The fact that no mainstream news outlet right now is calling for a Biden landslide suggests that these outlandish poll numbers might in fact be more propaganda than fact.
Some in America believe that Biden is winning the hearts-and-minds of voters only in the media--and in faked "polls" (here). Others say something's missing from these polls (here).
What do inaccurate polls in the US suggest? That voters are being lied to?
More important for us in Israel, the American possibility that polls are more propaganda than fact begs an Israeli poll question: how inaccurate are Israel's polls?
For example, how much oversampling of respondents do Israeli polls use? What percent of Israeli poll respondents are committed Netanyahu-haters? How often do Israeli polls print results as suspect as the nbc/wsj poll above?
No one in Israel knows the answer to these questions.
Do Israeli polls attempt to ferret out oversampling biases? Do Israeli polls attempt to test for respondent hatred toward one or another candidate? Do Israeli polls attempt to make sure there's a balance between respondent Party affiliations?
No one in Israel knows. Or, if they do know, they aren't talking.
These questions are important because there is arguably as much hate spewed by media elites and politicians against Benjamin Netanyahu in Israel as against Trump in America. As we see in America, Israel is also extraordinarily polarized (here).
With so much hate driving voter passions, how can voters in either country get an accurate sense of a candidate's electability if polls in both countries don't 'correct' for hatred--or, at the very least, account for poll biases and oversampling?
Shouldn't voters in both countries be told if a poll contains an oversampling--or any kind of respondent bias? Shouldn't voters in both countries be told if a poll has used unusual (and perhaps flawed) assumptions/methodologies specifically to support one candidate over another (here)?
If polls never offer such information, polls risk losing credibility. Especially during times when intense passions drive an election, citizens in both Israel and the US need more than simple poll results. We need information about poll respondent Party affiliation. We need information about respondent biases. We need information about poll biases. We need information about poll assumptions and methodologies. We need information about what a poll might be missing (here). Otherwise, who could trust any poll?
Voters in both countries deserve better information. Just don't bet on them getting that information.
No comments:
Post a Comment