Tag: Trump vs. Biden

Trump vs. Biden: 14 (Another Way to Read the Polls)

In 2016, Clinton’s lead over Trump in the final seven days of polling averaged 2.5 percentage points (or those polls covered by RealClearPolitics (RCP). Clinton’s edge in the nationwide tally of popular votes was 2.1 percentage points. Despite her (meaningless) nationwide edge, she lost to Trump because he won Michigan, Pennsylvania, and Wisconsin — each by less than 1 percentage point. Those states’ 46 electoral votes gave Trump the win over Clinton.

In 2020, Biden’s lead over Trump in the final seven days of polling averaged 7.9 percentage points for those polls covered by RCP. Biden’s edge in the nationwide tally of popular votes was 4.5 percentage points; that is, slightly below the bottom of the 95-percent confidence interval around the apparent 7.9-point lead. Despite his (meaningless) 4.5-point lead in the popular vote, Biden won the election only because he edged out Trump in both Georgia and Wisconsin by less that 1 percentage point and in Pennsylvania by a little more than 1 percentage point. Those states’ electoral votes gave Biden his win over Trump.

Hypothesis: Clinton and Biden underperformed at the ballot box vs. their poll numbers because the polls (on average) were biased somewhat toward Clinton in 2016 and much more so toward Biden in 2020. This hypothesis is supported by a report issued by the American Association for Public Opinion Research (https://aapor.org/wp-content/uploads/2022/11/AAPOR-Task-Force-on-2020-Pre-Election-Polling_Report-FNL.pdf):

● The 2020 polls featured polling error of an unusual magnitude: It was the highest in 40 years for the national popular vote….

● The polling error was much more likely to favor Biden over Trump. Among polls conducted in the last two weeks before the election, the average signed error on the vote margin was too favorable for Biden by 3.9 percentage points in the national polls and by 4.3 percentage points in statewide presidential polls.

● The polling error for the presidential election was stable throughout the campaign. The average error matched closely for polls conducted in the last two weeks, in the final week, and even in the final three days. The challenges polls faced in 2020 did not diminish as Election Day approached.

● Beyond the margin, the average topline support for Trump in the polls understated Trump’s share in the certified vote by 3.3 percentage points and overstated Biden’s share in the certified vote by 1.0 percentage point. When undecided voters are excluded from the base, the two-candidate support in the polls understated Trump’s certified vote share by 1.4 percentage points and overstated Biden’s vote share by 3.1 percentage points.

The report also concludes that the bias wasn’t due to the massive surge in early voting and mail-in voting in 2020.

The foregoing suggests that if the polls remain strongly biased toward Biden in 2024, and if he polls much worse than he did in 2020, his candidacy is doomed.

With that background, I refer you to the following graph, which compares results of polls reported by RCP for the elections of 2016, 2020, and 2024. The results represent polls conducted between June 1 and election day of each election year. We’re a long way from election day 2024, but the results to date are ominous for Biden; he is running behind Clinton’s losing pace of 2016. Stay tuned.

Trump vs. Biden: 13 (A Glimmer of Hope)

If you have read “Election 2024: The Bottom Line” or “Election 2024 in Perspective” you will understand why I use “glimmer of hope” to refer to a possible Trump victory in November. The glimmer of hope that I see is Trump’s standing in the polls this time around compared with his standing in the polls at this point in 2016 and 2020. Specific

I collected all of the two-way polling results reported at realclearpolitics.com for the elections of 2016 and 2020. I weeded out all of the polls that were conducted before June of each election year and plotted the numbers for the rest: the Democrat candidates’ polling lead or deficit vs. number of days before the election. I have also begun to plot similar numbers for this year’s election. Here are the results as of today:

The plot points represent the Democrat candidates’ lead or deficit vs. Trump. Two polls with the same average date (June 4) gave Biden and average lead of 0.5 point. A third poll with an average date gave Trump a narrow lead of 1 point, which converts to a Biden deficit of 1 point.

Clearly, Trump is doing a lot better in the polls this year than he was doing at this time in 2016 and 2020. But a lot can happen between now and when voting starts. A lot did happen to Clinton (2016) and Biden (2020) — and it wasn’t good for them.

Clinton’s polling was all over the place, but she ended up 2.9 points ahead of Trump for polls conducted in the seven days before the election. She “won” the nationwide popular vote by 2.1 percentage points (plotted on the right axis). But she lost the electoral vote because of Trump’s narrow wins in a few key States.

Biden never relinquished his lead in the polls. He ended up 7.6 points ahead of Trump for polls conducted in the seven days before the election. He “won” the nationwide popular vote by 4.6 percentage points (plotted on the right axis). But he won the electoral vote only because of narrow wins in a few key States.

If there’s a pattern, it’s this: Polls, in the aggregate, overstate Democrats’ shares of the nationwide popular vote — at least when they’re up against Trump. Further, because of the electoral college, it takes a large margin in the final polls to be certain of victory — if you’re a Democrat. The necessary margin is greater than final Biden’s 7.6-point lead in the 2020 polls.

Conclusion: Biden is in deep trouble, as of now. But election day is five months away. Thus: a glimmer of hope.

Trump vs. Biden: 12a (Rethinking the “Battleground” States)

There’s a new spate of articles about Trump’s lead in the polls for so-called battleground States: Arizona, Georgia, Michigan, Nevada, North Carolina, Pennsylvania, and Wisconsin. Yes, Trump is still leading in all seven States, if you take an average of polls reported at RealClearPolitics. Here’s how the five-poll averages look for Trump:

  • Arizona – up by 5.2 points and rising
  • Georgia – up by 5.0 points and rising, but below earlier peak
  • Michigan – up by 0.6 point and falling, well below earlier peak
  • Nevada – up by 5.6 points and rising
  • North Carolina – up by 5.8 points and falling, somewhat below earlier peak
  • Pennsylvania – up by 2.6 points and rising, but below earlier peak
  • Wisconsin – up by 0.6 point and falling, somewhat below earlier peak.

Only the leads in Arizona, Georgia, Nevada, and North Carolina are statistically significant (lower bound of 95-percent confidence interval is greater than zero).

Adjustments for pollsters’ political biases — which I haven’t made — might make things look better for Trump. But the real problem with the “battleground” polls is their paucity. This can be seen by contrasting a metric I devised for nationwide polls with similar metrics for the “battleground” polls.

The metric is the change in each pollster’s results from poll to poll. For example, in the Morning Consult poll that was conducted May 3-May 5, Trump was up by 1 point. He was up by 1 point in the next Morning Consult poll which was conducted May 10-May 12. That counts as zero gain on the average date of the later poll: May 11. The full tally for all polls reported at RealClearPolitics since August 2023 looks like this:

Here’s a similar graph for Pennsylvania, which the most heavily polled of the “battlegrounds”:

That’s not much to go on, is it?

What to do? I’m inclined to ignore the polls for individual States and keep my eye on the nationwide polls. But I will be more demanding of myself when I declare that Trump might win with a small lead or deficit in those polls.

As I say in the updated version of “Trump vs. Biden: 2“,

The statistical relationship in the graph [below] is meaningless. What can be meaningful is a narrow margin of victory (or loss) in a few States. This underlines the lesson from “How Good Are the Presidential Polls?“: Even a large lead in nationwide polls doesn’t signify victory in the Electoral College.

Well, the relationship isn’t quite meaningless. Here’s how it looks with a 90-percent confidence interval (which happens to encompass 100 percent of the data because the underlying distribution isn’t normal):

What this means is that I will be confident of a Trump victory (270 or more electoral votes) only if it looks like he will get 53 percent (or more) of the two-party popular vote, nationwide. A tally of at least 54 percent (a margin of at least 8 percentage points) would be convincing. (That’s close to my finger-to-the-wind estimate of 9 percentage points in “How Good Are the Presidential Polls?“, wherein I assessed the accuracy to the nationwide polls for the presidential elections of 2004-2020.)

If Trump doesn’t rack up a big margin, the Dems will be able to manufacture enough votes in key States to steal the election — again.

Trump vs. Biden: 12 (Further Adjustments and Updated Results)

UPDATED 04/30/24

The latest results for the so-called battleground States are positive for Trump. I’ll get to those after I review the national polls. I’ll then remind you how Trump can win the election even if he “loses” the meaningless nationwide tally of popular votes.

NATIONAL POLLS

I introduced a new metric in Trump vs. Biden 11: the change in each pollster’s results from poll to poll. For example, Trump was tied with Biden in the Quinnipiac poll that was conducted on April 18-22, and was down by 3 points in the Quinnipiac poll of March 21-25. That counts as +3 on the average date of the later poll: April 20. The full tally for all polls reported at RealClearPolitics since August 2023 looks like this:

Keep your eye on the red line. It spent a lot of time in negative territory from the middle of March to late April. But it’s now back in positive territory.

Now, for the overall polls. I am dropping the “poll of polls”, which aggregates all national polls reported by RCP, regardless of bias and frequency. It is too easy for polls often-published polls with a strong bias (usually pro-Democrat) to make things look worse for Trump than they actually are. (Demoralization of Trump voters is a main objective of such polls.) From here on out, I’m sticking with two, more selective, bias-adjusted measures of the nationwide balance between Trump and Biden.

First is a collection of all polls report by RCP, with two adjustments. I use only the latest results from each pollster, to avoid overrepresentation. I adjust the results by the average of the anti-GOP bias reported by RCP for 14 nationwide polls. I add that value (3.3 percentage points) to Trump’s 5-poll average. Here are numbers, which include the results of all nationwide polls reported by RCP as of today:

 

Trump slipped somewhat from the middle of March to early April. The trend since then is positive.

The other collection of nationwide polls that I am keeping tabs on is a smaller set, in which each poll is adjusted for the amount of bias shown by the same pollster in 2020:

In this case, Trump’s rebound is clear and convincing.

“BATTLEGROUND” POLLS

The other bit of good news for Trump is that he is doing better in the “battleground” States than he was when I last wrote about them (here). All of the polls underlying my previous report were conducted in March. There is now a new round of polls, conducted in April. In general, it looks better for Trump:

THE BOTTOM LINE

Can Trump win the election without “winning” the meaningless nationwide tally of popular votes? Of course he can, thanks to the Electoral College. He did it 2016, when he “lost” to Hillary Clinton by 2.1 percentage points. He almost did it in 2020, when he “lost” to Joe Biden by 4.5 percentage points. And he would have won in 2020 except for “election interference” by the Democrats on a massive scale. You can read all about that here. I’ll save you the trouble of wading through more than 200 links. Here’s the punch line:

Through a combination of information control, partisan management of election processes, outright fraud, ballot harvesting, and failure to apply election laws on the books, the presidential election of 2020 was stolen by a cabal of super-rich elites, crooked politicians, crooked lawyers, and judges who either didn’t want to hear the truth or were blinded to it by partisan considerations. For a systematic treatment of much of the chicanery mentioned in the preceding sentence, see Mollie Ziegler Hemingway’s Rigged: How the Media, Big Tech and the Democrats Seized Our Elections. Regarding outright fraud, see “Last Thoughts on Voter Fraud” (The Adventures of Shylock Holmes, December 11, 2020) and “The Most Secure Election in American History?” (Gatestone Institute website) for thorough wrap-ups.

Let’s hope that an army of better-armed poll-watchers and lawyers working for the GOP can keep it from happening again.

In any event, it’s my view that if the average of the final ten (unadjusted) polls gives Trump a lead of 1 or more points, he will win the election.

Stay tuned.

Trump vs. Biden: 11 (A New Metric)

In “Trump vs. Biden: 10” I highlighted the tightening of the race (according to nationwide polls). Changes in the 15 days since that post have been slightly in Trump’s favor. Below are results, which I have parsed in three ways:

  • all polls, unadjusted for bias (with polynomial fits for smoothing)
  • the latest version of each pollster’s result, unadjusted for bias
  • the latest version of each pollster’s result, adjusted for bias (using each pollster’s bias in 2020)

Here’s a fourth way to look at the results:

The values represent the change in Trump’s lead (or deficit) since the preceding poll by the same pollster. The typical margin of error for a poll (statistical estimate of the accuracy of the poll) is plus or minus 3 percentage points. Most of the movements are within that range. Therefore, what all of the preceding graphs point to is a race that is now essentially tied.

Well, it’s tied if the polls aren’t picking up “hidden” Trump voters — Trump voters who aren’t telling pollsters their true intentions. If there are enough such voters to swing the election to Trump, it won’t become evident until the votes are counted.

Until then, I can only tell you what the polls are saying.

Trump vs. Biden: 10 (The Race Is Tightening)

Contrary to cherry-picked results presented on some right-wing sites (and I don’t use “right-wing” as a smear phrase), Trump is sliding relative to Biden.

Let’s start with the small picture, Trump’s standing in the so-called battleground States: Arizona, Georgia, Michigan, Pennsylvania, Nevada, North Carolina, and Wisconsin. Here’s are the trends in each of those States, going back to October 2023:







I used 3-poll averages because of the paucity of polls in some States. Only in Pennsylvania is Trump clearly improving his edge over Biden.

The results in most of the battleground States mirror the national results, which I have parsed in three ways:

  • all polls, unadjusted for bias (with polynomial fits for smoothing)
  • the latest version of each pollster’s result, unadjusted for bias
  • the latest version of each pollster’s result, adjusted for bias (using the pollsters’ bias in 2020)

The results, in that order, are below. Because of the greater number of nationwide polls, I have used 5-poll averages.

 

These aren’t pretty pictures if you don’t want Biden to win.

Finally, there are the betting odds. Trump had a 16 percentage-point lead over Biden in the immediate aftermath of the Hur report, which depicted Biden (accurately) as a befuddled old man with a poor memory. That edge has shrunk to 5 percentage points — the smallest it has been since mid-November.

All of this is bad news if you are (as I am) fearful of the outright repression that would follow Biden’s re-election. What do I mean by that? Stay tuned.

Trump vs. Biden: 9 (A Fresh Look at the Numbers)

UPDATED 03/18/24 – 03/21/24

Something has been nagging me about the poll numbers that I’ve compiled and reported in several previous posts (e.g., here). That something is the lack of an adjustment for political bias. I’ve been reporting the statistical margin of error, but that’s not the same thing.

Every poll, biased or not, has a margin of error, which is a statistical measure related to sample size. But what the pollsters don’t tell you is that a margin of error, properly understood, is just an estimate of the accuracy of the polling results — assuming a random (unbiased) sample. That’s not what you get from most polls, which are designed (often deliberately) to slant results toward the candidates of a certain party — usually the Democrat Party. A margin of error doesn’t account for bias.

In 2008, Bush’s average margin for polls conducted in the seven days before election day (with no double-counting of pollsters) understated Bush’s margin in the nationwide popular-vote tally by 1.0 percentage point. Similar analysis for succeeding elections yielded these result: Obama 2008, overstated by 0.1 percentage point (effectively equal); Obama 2012, understated by 3.2 percentage points; Clinton 2016, overstated by 1.6 percentage points; Biden 2020, overstated by 3.2 percentage points. With the exception of 2012, when Obama’s cheap (taxpayer-funded) “heroics” in the aftermath of Hurrican Sandy coincided with the final runup to election day, the bias favoring Democrats has grown.

With that background out of the way, I hereby introduce a new metric that I will use for polling trends. First, I will use a 5-poll average. I had been using 10-poll average, which can span polls conducted over a two-week period, or longer. The 10-poll average also (often) includes more than one poll by the same pollster. I then switched to a 3-poll average, with no double-counting of any pollster’s results. But a 3-poll average is too “jumpy” (unstable). A 5-poll average (with no double-counting) seems like a good compromise.

Second, after computing the 5-poll average, I adjust it by adding 3.2 percentage points to Trump’s margin relative to Biden. That value (the final polling bias against Trump in 2020) corrects for growing pro-Democrat bias, which is probably greater than 3.2 percentage points this year. The result, I believe, is closer to the truth than the polls (on average) would have it because of the prevailing bias toward Biden (and Democrats, generally).

Without further ado, here’s how things look:

Trump’s margin actually peaked last fall, dipped significantly, rebounded somewhat, and dipped again. We’ll see if it rebounds again — and if my revised method yields an estimate that’s close to the actual outcome of Election 2024.

Trump vs. Biden: 8 (New Numbers)

UPDATED 03/03/24

My unvarnished view about Trump’s candidacy to the contrary notwithstanding, it is better for Trump to be ahead (rather than behind) in the polls. Polls do have some (limited) value as a predictor of presidential election outcomes — though it is far too early to use them for that purpose.

In any event, with the release of four new polls yesterday and today, the latest numbers look like this:


Source: Polls reported by RealClearPolitics.

To avoid the over-representation of polling organizations that report frequently, the values plotted above include only the latest poll released by each organization. (Polls without margins of error are excluded.) The three-poll span affords a better indication of trends than the 10-poll average that I had been using.

CAVEATS: The average date of the three most recent polls is February 27 — five days before the publication of this post. The numbers shouldn’t be taken as predictive of the outcome of the election. They simply reflect the stated preferences of respondents when the polls were conducted. Also (and positively), a GOP candidate can win even when he is behind in the (meaningless) nationwide popular vote (see this).

Trump vs. Biden: 7 (My Unvarnished Perspective)

The polls, on average, favor Trump. Although his lead isn’t statistically greater than zero, that’s okay for a Republican.

The betting odds are going against Biden. Bettors see him as much less likely to win than they did a week ago, mainly because of this passage from Special Prosecutor Robert Hur’s report on Biden’s “mishandling” of classified documents:

[A]t trial, Mr. Eiden would likely present himself to a jury, as he did during our interview of him, as a sympathetic, well-meaning, elderly man with a poor memory. Based on our direct interactions with and observations of him, he is someone for whom many jurors will want to identify reasonable doubt. It would be difficult to convince a jury that they should convict him-by then a former president well into his eighties-of a serious felony that requires a mental state of willfulness.

Regardless of the polls and betting odds, I believe (today) that Trump will lose the election. It will happen because of his baggage (e.g., openly vindictive personality and resulting feuds, indictments, possible convictions), which will begin to weigh more heavily as voters actually decide which way to go.

With narrow wins in a few key States, Trump could win the election even if the overall popular vote goes against him by a couple of percentage points (just as he did in 2016). That’s what happened in 2016. I attribute that win to luck and surprise. Clinton’s lead in the polls and betting markets was so large that she seemed to be a shoo-in. As a result, the Democrats didn’t gear up to manufacture enough votes to win the close races that went Trump’s way and gave him an edge in the Electoral College.

Democrats geared up with a vengeance in 2020, with a lot of help from the Covid pandemic which boosted mail-in voting — the happy hunting ground of electoral fraudsters. Democrats will build on their successes of four years ago, while the GOP will try to play catch-up ball. But the GOP will fall short because (for the most part) it will confine its vote-generating operations to legal methods. Democrats (being leftists) will pull out all the stops and officials and judges (mostly Democrats and never-Trumpers) will cover for them.

There are people like me who will vote for Trump if he’s the GOP nominee only because we can’t abide what the Democrats are doing to the country. But there aren’t enough of us, I believe, to overcome Trump’s baggage. Moreover, Democrats still have plenty of time in which to recover from last week’s Biden fiasco — and they find a way to do it, with the enthusiastic aid of most of the media. In the end Biden — or whomever the Democrats nominate — will win in a rerun of 2020.

That’s how it looks from here — as of today. Who knows what will happen in the next several months, or how it will affect the outcome of the election? I don’t.

Trump vs. Biden: 6 (Trump Resurgent)

RealClearPolitics maintains a running tally of presidential election polls. I construct a moving average of the results, where the average represents Trump’s lead (or deficit) for the 10 most recent polls. I also construct a 95-percent confidence interval around the moving average, using the margins of error reported by pollsters. (Having done this for a while and observed some erratic and incredible results from Rasmussen Reports, I removed the results published by that formerly reliable polling organization from my calculations.)

Here’s the trend since late July 2023:

Back-to-back wins in Iowa and New Hampshire boosted Trump considerably. His margin is now (barely) back in positive territory (i.e., the lower band of the 95-percent confidence interval is just above zero).

CAVEAT: This isn’t a prediction of the outcome of the election. The figures simply reflect the stated preferences of respondents when the polls were conducted. Also (and positively), a GOP candidate can win even when he is behind in the (meaningless) nationwide popular vote (see this).

Trump vs. Biden: 4

UPDATED 11/21/23

RealClearPolitics maintains a running tally of presidential election polls (among many others). RCP has also assessed the pro-Democrat or pro-Republican bias of the final presidential-election polls issued by major pollsters in 2016 and 2020. On average, the polls were biased toward the Democrat nominee by 2.3 percentage points.

As the pollsters release their results, I adjust Trump’s lead/deficit for bias. I then construct a moving average of the adjusted results, where the average represents Trump’s adjusted margin for the 10 most recent polls (taking the mid-point of each polling period as the date of each poll).

I then convert that 10-poll average to an estimate of Trump’s share of the two-party popular vote. For example, an average margin of +4 indicates a 52-48 split of the popular vote, that is, Trump gets 52 percent of the popular vote.

Finally, I apply my algorithm for the relationship between the GOP candidate’s share of the electoral vote and his share of two-party popular vote.

The estimates of popular-vote and electoral-vote shares don’t account for the margin of error in pollsters’ findings or the margin of error in my estimate of electoral votes. But the movement of the estimates may be taken as indication of the movement in voters’ preferences between Trump and Biden (or whoever might become their parties’ nominees). It is that movement which I will report from time to time.

Here is the first report, which begins with a clutch of polls that were completed in mid-August, when polling season seems to have begun in earnest: