Ok, so I should be combining this Turbo graph with the 538 graph to “get it”? I just don’t see how the graph linked shows anything other than the average of all pollsters is more stable than one pollster.
I think it shows that Rasmussen may be an outlier (i.e. they’re the only one showing a consistent decline over the last ~40 days or so). To test this, we might have to see similar graphs for all pollsters with multiple polls over that period – i.e. graphs for Harris vs all-non-Harris-polls, Monmouth vs all-non-Monmouth-polls, etc., which is likely too much to ask for right now. But if it does indeed show that Rasmussen is the only outfit showing a decline right now, then that’s interesting and potentially useful information for evaluating whether there is indeed an actual polling decline going on.
I’m not claiming anything. I’m showing you what’s in the data. My point is that this is what it looks like when you plot the difference between Rasmussen and the average of all other pollsters.
Here’s code to grab the 538 data set and pull it into a dataframe. Feel free to copy it and adapt to answer the questions that you feel are important.
import pandas as pd
import datetime
url = 'https://projects.fivethirtyeight.com/trump-approval-data/approval_polllist.csv'
df_raw = pd.read_csv(filepath_or_buffer = url, parse_dates = ['startdate', 'enddate'])
# Drop columns we don't need
df = df_raw.loc[:,['pollster', 'startdate', 'enddate', 'approve', 'disapprove']]
Looking at RealClear approval poll tracker Reuters has Trump at -14 on 8/26 down from -12 on 8/19. Politico has him at -15 on 8/23 down from -13 on 8/16. So Rasmussen is certainly not the only pollster showing dip at the moment.
Of course Rasmussen is not the only pollster showing dip at the moment. Nobody claimed that.
Lance Turbo, interesting chart and I appreciate the effort, and what it says about Rasmussen. But I would say, this alone does not show that “Almost all of the variation in the last 90 days or so has been due to Rasmussen.” That may indeed be the case, but should you not also show each individual poll other than Rasmussen in a similar plot? If by chance other poll(s) are close to Rasmussen, then your conclusion isn’t valid.
I think I’ve found where folks are getting hung up. When I said, “Almost all of the variation in the last 90 days or so has been due to Rasmussen,” I meant in my chart, not in the 538 chart. In my chart, there is a mostly flat line, a line with big swings, a line that is the difference between the two. Almost all the variation in the difference line is due to the wiggly line, not the flat line.
I was accurately describing my graph. It seemed relevant to the topic of the thread but was not meant as the final answer to the question posed in the thread title.
I wasn’t making a scientific statement. I was describing a picture.
But wouldn’t any given poll show a greater variation than the average of all other polls?
To be clear, I think the big swing in Rasmussen is at least partially responsible for what we’re seeing in 538’s tracker, but since we don’t have the full details of 538’s algo we can’t tell for sure.
But mousing over 538 we see a recent peak on 7/22 at 43.0% and a fairly steady decline to now on 9/3 at 41.4%. A decline of 1.6% or roughly a net approval decline of around 3.2% (-3.2 net net).
Now when I look at the Rasmussen and non-Rasmussen 14 day centered rolling averages for that period.
Rasmussen net 7/22: -1.6
Rasmussen net 9/3: -5.8
Net net: -3.2
non-Ras net 7/22: -10.1
non-Ras net 9/3: -9.1
Net net: +1.0
I don’t calculate my averages the same way 538 does, and we don’t know exactly how 538 does what it does, but I strongly suspect that in the alternate universe where Nate didn’t include Rasmussen polls, Hermitian would not have had a reason to start this thread.
Maybe. Probably. Almost certainly.
However, Rasmussen is special here because it adds a data point every single day. In the last 14 days we have 14 Rasmussen data points compared to 49 for all other pollsters. So the added uncertainty of smaller sample size is still there for Rasmussen but not as big as for others (e.g. Morning Consult with 2 polls in the last 14 days, or Emerson with 1). Rasmussen always has a data point less than 24 hours old while most other pollsters do not.
I’m not just talking about noise though. There is basically no trend one way or the other with all non-Rasmussen polls, and there is enough data to show a large downward trend in Rasmussen polls.
He’s been erratic, such as the time when he looked to the heavens and claimed he was “the chosen one” to deal with China. I also think that he has pretty much lost center-right independents who switched between Obama and Trump, but that’s not really eating into his base. I think what might be hurting him even more is that more and more members of the billionaires club are nervous that he’s going to fuck it all up.
I don’t care if a politician has a base: erratic rhetoric and behavior makes a leader look weak. And I think the Wall St Republicans are increasingly out of explanations and defenses for his ‘deal making acumen’.
Now having said that, he could realize tomorrow that he’s going to lose the trade war and strike a “deal” with China. The reason why I’m not sure he’ll do that anymore is that he’s going to have a recession no matter what, and he’ll look even weaker if he gives in. So in a sense, he’s screwed if he goes to the mat, and he’s screwed if he doesn’t. President Xi may be beginning to sense this, and if his own economy weren’t in the dumps right now, I wouldn’t put it past him to block Trump’s off-ramp - but Xi needs a deal, too, so there will be some space for negotiation on his part. It’s Trump who is now confronted with an increasingly complex set of circumstances that are of his own making.
I would bet good money on this. Not my money, of course. But yours. I’d bet all of your good money on it. 
Only kind of money I know about is “gone”.
I’m sorry if this seems obtuse, and maybe I am missing something, but what relevance do you think it has?
A single pollster has much noisier data than the aggregation of all the others. Okay. Pretty much something everyone knows.
But I’m looking at your graph and seeing both Rasmussen and non-Rasmussen dropping from 7/8 to 7/29, and Rasmussen rising fairly steeply while non-Rasmussen dropped from 8/12 to 8/22.
During that ten day time of net approval dropping on the non-Rasmussen aggregate while rising on the Rasmussen one 538’s tracker DROPPED, from -11.3 to -13.5.
This is very inconsistent with any hypothesis that any changes in the 538 tracker are due to Rasmussen results. It looks like instead Rasmussen has little impact on 538’s tracker.
And expecting C+ rated Rasmussen to have an outsized impact on 538’s tracker is inconsistent with what we do know of their algorithm. Not full details but enough.
Silver knows data analysis better than to have a tracker system that would result in a C+ rated house with a known house effect have outsized impact because they report every day.
I won’t state with confidence that the tracker moving away from the highest numbers of his range (-9.7) in July (everyone going on about how high they were for him), to the portions more in the middle of his range now (-12.7), is not more than background farts … it could be the start of a real collapse and in retrospect be real and not just noise. But I doubt it as much as I doubted the significance of those July numbers in the other direction.
These are quantum froth stochastic variations. Happy to see the random walk to this direction but way too small to read any meaning into it.
Get that net to a consistent under -15 (not just a few days) and you have real change to talk about. I hope we have that conversation!
Like how YouGov does? This 8/31 to 9/1 polling.
38% strongly disapprove
12% somewhat disapprove
19% somewhat approve
25% strongly approve
6% not sure
So roughly 3/4s of those who disapprove pretty much hate his guts. And fewer, but still a majority, about 57%, of those who approve pretty much love him to bits.
That was my first thought as well, of course.
Sorry, that sure looks like a lot of tap dancing to be on every side of the argument.
I accept your apology.
Well, I am somewhat sorry I wasted time on this but that wouldn’t generally be considered an apology.
I ordered one of those at Starbucks. Tasty, but a bit overpriced.