Why are Trump's approval numbers up?

This week...

Reuters/Ipsos: 42 (+1)
Economist/YouGov: 46 (+1)
Rasmussen Reports: 49 (+4)​

30 days ago...

Reuters/Ipsos: 41
Economist/YouGov: 45
Rasmussen Reports: 45 ​

Shouldn't they be going down with all this 24/7 anti-Trump impeachment inquisition crap?

It looks like the left's attempt to diminish Trump isn't working.

 

Just wait til this farce hits the Senate, with Graham running it, if it makes it that far.

 

Lets look at all the polls instead of the ones you cherry picked.
Trump's RCP approval average for current polls is 43.3%. On Nov 4 it was 43.5%. Virtually unchanged.
https://www.realclearpolitics.com/epolls/other/president_trump_job_approval-6179.html

Lets look at the FiveThirtyEight polling average. Currently its 41.6%. On Nov 4 it was 41.3%. So that is unchanged as well.
https://projects.fivethirtyeight.com/trump-approval-ratings/?ex_cid=rrpromo

So what did you get wrong? Rasmussen is biased and pro-GOP. They always poll Trump's numbers far above anyone else, even Fox News. They also were the only major polling source that incorrectly predicted the Republicans would win in 2018 and keep the house. You should at least consider them along side all the other polls.

But you did consider a couple other polls. Reuters and Economist/YouGov. Problem is that the change in those polls were only 1%, which is far below the margin of error. It is statistically meaningless. You also left out a bunch of polls that didn't support you agenda.

 

Rasmussen has a methodology, just like every pollster has a methodology. Their methodology will yield different results because they have different techniques. It neither makes them bias nor objective. As long as the polling is representative, it has achieved its objective (which is to capture the population mean).

Besides, Rasmussen predicted the 2016 Election winner so they are more accurate than most.

This is not what the Margin of Error signifies... It has nothing to do with the percentage change in the polling data. The margin of error just tells you the range the true population mean (the percentage of people support/opposed) falls within.

For example, a margin of error of 3% and a sample statistic of 46% implies that 46% either support or oppose a candidate, with the upper and lower bounds being 43% and 49% respectively. This never changes regardless of whether or not a poll falls within the MOE.

 

Polls can be very inaccurate as proven in the past over and over again. Too many polls have become a propaganda organ. What I see is the large crowds generated at a Trump rally and the mini crowds generated at the dems rallies. That could change when the single dem candidate is chosen.

 

A couple of things, I think.

One thing could be what I'll call "Democrat fatigue". I think a lot of voters are coming around to the fact that congressional Democrats have simply opted out of doing anything to participate in governing the country, and that's not what their constituents elected them to do. They want them to participate and get things done, rather than only working to unseat this president. Congressional Democrats have been working toward that goal for 3 years to the exclusion of all else, and moderates and independents are worn out with it. The purely partisan nature of the impeachment effort contributes to the fatigue as well.

Secondly, the economy is humming.

We are not at war. That's not to say we don't have people in harm's way in Afghanistan and elsewhere, but our missions are mostly train and support missions and special forces raids like the one that got Baghdadi in Syria.

"Democratic fatigue", a strong economy, and peace are working together to improve Trump's job approval, in my opinion.

 

But their predictions haven't been so good as of late. They got 2018 very wrong. So don't put all your eggs in Rasmussen especially when their numbers are just so different than everyone else's. Consider all the polls.

Lets say the poll had Trump at 45%. And then that number went up to 46%. And the margin of error on both polls was 3%. That means the first poll really is 42%-48% while the second poll is 43% - 49% when considering the margin of error. It could have been that his approval went from 48% down to 43% or his approval went from 42% up to 49% or somewhere in between. When the difference in the polls is smaller than the margin of error, it is statistically meaningless since it could just be a result of the margin of error itself.

 

Rasmussen has a right bias and only mediocre results.

 

Oh look. Fake News.'

Trump approval numbers are actually DOWN

https://www.realclearpolitics.com/epolls/other/president_trump_job_approval-6179.html

 

Democrats have passed almost 400 bills that have been sent to the senate that a single republican has refused to even allow a vote on. In the mean time, what is the senate doing besides confirming massive amounts of partisan and unqualified judges?

Manufacturing just had its fourth straight month of recession. China talks have been delayed till after the election. Debt levels not seen outside of a recession. Massive bailouts to damaged industry. Stagnate wages.

If everything you posted was true — even with trump being the vile human that he is — he would be in the low 60’s to high 50’s but he has never even hit 50% approval.

 

Yup

 

First of all, those consist of polls of the most recent races compiled by fivethirtyreight (most likely the 2018 midterms), which Rasmussen was less than consistent. That doesn't really have any bearing on the fact they predicted the 2016 correctly when most did not.

Secondly, Rasmussen accurately predicted 78% of the races within the last 12 months. That is hardly mediocre by any predictive analytical or statistical standard.

Finally, as for your claim of right bias, there is no data to support such an assertion. They merely use statistics to predict an outcome; and statistics is about what you can prove using data.

 

Is 78% really that impressive? 50% is a coin toss and I'd argue that if you are at 69% then you are pretty unreliable. Most races are in districts where one party is dominant so I would expect an accuracy of over 78%, maybe more like 85-90%.

 

There were many races that Rasmussen predicted inaccurate in 2018 and nationally, they incorrectly predicted the Republicans would retain the House. However, for the most part, the pollster is consistent. Majority of the races within the last 12 months they predicted accurately.

None of this means what you think it means, I'm afraid. The MOE is simply the bounds of where the true population mean could be, given a confidence interval.

And what I mean by "true population mean," I mean the piece of data that you could use to generalize an entire population. Because, after all, you cannot meaningfully poll an entire population. It takes too much time and expense. As a result, pollsters rely on random samples that are intended to be representative of the population. Because polls randomly sample from within a population, there will always be some amount of uncertainty, or variable error (variance), associated with their results. The MOE is merely a measurement of that uncertainty.

Also, the MOE may vary depending on when the poll was taken, but it may also be consistent. This would dependent on the sample size and the confidence interval. Regardless, each poll has its own MOE. So a poll with a point estimator of 46% (Trumps approval rating) with an MOE of +/- 3% doesn't mean that his approval rating could move from 48% to 43%. It means when you consider the variability between the population and the sample size, Trumps approval rating could be anywhere between 43% to 48%, based on the confidence interval, which is pretty good statistically. Obviously, the smaller the MOE the better because you want your analysis to be statistically meaningful.

The confidence interval is also important because it also determines how often (based on the number of times a poll is conducted) that a sample statistic would fall within the MOE. For example, Gallup had Romney winning the election at 49% with Obama at 49%, with a 95% confidence interval and an MOE of 2%. The REAL results of the election showed Obama at 51% and Romney at 47%, which was actually outside the MOE.

Simply put the MOE doesn't show where the results would/should be. It only explains what sort of variability we can expect when we factor the difference between our sample and the population.

 

Unchanged?

Shouldn't they be going down with all this 24/7 anti-Trump impeachment inquisition crap?

It looks like the left's attempt to diminish Trump isn't working.

 

They were soundly in the middle in 2016 as well.
7 out of 11 of the polls taken in the last week of 2016 were within the MOE
A rating of C is average.
Any good poll will sample according to their methodology. This almost always includes weighting by party among other things.

So the assertion Rasmussen has any magic is BS. They are mediocre.

 

Not from 30 days ago. They are actually up.

 

Not from 30 days ago. They are actually up.

 

There are some pollsters who do no better than filling a coin. There a handful of pollsters who do much better than 78%, but there are also a handful who do much worse. I'm not in the pollster business, but part of what I do for a living involves forecasting and I can tell you from personal experience establishing a methodology to predict events is very hard.

I didn't have time to average the data from 538, but in IMO I believe 78% is decent. Certainly not mediocre compared many others, which have 0% accuracy.

 

MOE is the results you might expect if you fielded the same poll multiple times.
So if you do the same poll 100 times and you have a 95% confidence level the results will be within the MOE 95 out of 100 times.

 

Look at the link you quoted.

30 days ago they were at a low. They rose briefly and are trending down

 

They are higher today than they were 30 days ago. Thanks for confirming my assertion.

 

78% accuracy in the last 12 months just isn't very much. The vast majority of races are in districts where one party dominates and the incumbent usually receives over 60% of the vote. If you remove those obvious races, I wonder what their accuracy is then? Maybe more like 50-50? A coin toss? 78% just isn't accurate enough for me to have exclusive faith in their abilities and I want to see the data from other pollsters as well.

You misunderstand. The first poll has him at 45%, which means it can vary between 42% and 48% with a 95% probability of being between these two numbers. The second poll has him at 46% with the MOE range of 43% and 49%. So it is possible that the actual population result for the first poll is 48%, on the high end of the MOE, while the actual population result for the second poll is 43%, on the low end of the MOE poll. Or it could be the reverse. Now the probability that this happened is pretty low because it is more likely the population result is in the center of the MOE range than at the edges, but we can't be certain the population totally matches the poll and is in the center of the MOE range. When the difference between the polls is only 1% and the MOE is 3% in either direction, you can't really say very much about what direction the polls are going.

One problem is that fewer and fewer people have landline phones and fewer are taking calls from pollsters. So the samples are becoming more and more self-selecting and less and less random. They also tend to have more of certain population demographics like older people who have landlines or politically active people who love responding to pollsters.

So pollsters have to adjust their results with a bunch of math and assumptions about what the voting population will be like. It kind of works but different pollsters have different demographic assumptions. Also, they are at the mercy of the sample of people who choose to answer the poll and weirdness in the response population can really throw off their polls.

This is why polls with 3% MOEs from different organizations can sometimes be 10% apart, and why 2016 was such a surprise. I'd say that the real MOE on polls is more like 6% in either direction, and its best the average as any polls as you can to get anything with any chance of accuracy.

 

Because people have already made up their minds about Trump. Some will hate him no matter what, while others will only listen to right-wing news and won't consider the attacks on Trump. But still I was surprised that his numbers were so unchanged from the impeachment stuff. I was also surprised when the Mueller report came out vindicating him of working with the Russians, and his numbers were virtually unchanged as well.

 

Rasmussen simple error average is 5.3, which is around the statistical average of 5.7. 538 rates the pollster with a C average simply because of the herding effect in their methodology, which the 538 model penalizes.

None that any of this means anything to the average person who only cares about the accuracy of the pollster, which is well above average.

You basically said nothing here. The only important factor when it comes to polling is sampling methodology, the questions they ask, how they filter prospects and how often they conduct the samples.

Other than that, there isn't anything complicated about calculating the confidence interval of a randomly generated sample. This is something we still each in elementary statistics.

The data suggest otherwise. Their accuracy is 78% in the last 12 months. The average is 63%. If you're ignoring mathematics, I can't really help you. You might as well just admit that you're a science denier.