_______
On the side bar, I have a spreadsheet that includes virtually have nearly every poll I can find over the past thirty days (that is not of the volunteer online variety). At this point that's a total of 22 polls. Many of these pollsters have been updating their numbers weekly, and many others are updating their numbers monthly. For the most part this allows me to keep up the latest poll from the lion's share of pollsters (I only include the latest poll for any pollster in the average). So basically... as long as a pollster updates at least once a month, they remain on the list. From this list I calculate a non-weighted average.
Now because I don't allow pollsters (for the most part) to continuously pop up and then drop off my polling spreadsheet, I avoid the large swings that could happen when a pollster with a particular bias replaces another pollster with the opposite bias (because of timing). As it stands right now, well over 95% of the new polls added, are replacing the old poll of the same pollster. This adds to consistency and avoids large swings. For example... the Real Clear Politics average lets some of the polls (including a 12 point June lead for Hillary in the ABC/Washington Post poll) fall off their averages. The new average was limited to a handful of polls (which happened to be favorable pollsters for Trump) making it appear that he made a substantial jump. But when ABC/Washington Post released a new poll, it became part of the new RCP average, as did CNN and NBC/WSJ. At about the same time they dropped two polls more favorable to Trump (Gravis and PPD) off their polling list. The replacement of apples with oranges actually showed significant movement in favor of Clinton, even as she actually didn't improve on average with the three new polls.
On my spreadsheet, the thee new polls from ABC CNN and ABC replaced themselves, while Gravis and PPD remained. My averages showed a slight improvement for Trump, which would make sense since he improved more than he fell in the three new polls. But overall, my spreadsheet did not show either the big swing to Trump or the swing back to Clinton. As both were nothing more than an issue with the timing of when certain polls came out. Overall, only changes within the individual pollsters will create any real movement. As it should be.
_______
I also have a second spreadsheet that tracks the demographic breakdown of these same pollsters (where those cross tabs can be found). Right now, I have twelve different pollsters creating a second spreadsheet that I use to calculate what I show as my "projected result". This second spreadsheet basically ignores top line results, and only considers how the demographic breakdown is voting. In other words, this spreadsheet will tell me who is gaining or losing ground with their own Party, with Independents, and which demographic group has the most undecided voters. Based on the Demographic breakdown, I plug those numbers into a formula where the amount of Republicans, Democrats and Independents remains constant. That formula is based on an average of the past four Presidential elections and some other considerations. It may be a correct assumption, it may be an incorrect assumption. But the formula remains consistent and will not change moving forward.
 |
| Click on me! |
Similarly, the pollsters I have added to this second spreadsheet are all consistently releasing either weekly or monthly pollsters, which eliminates those before mentioned swings because pollster "A" replaces pollster "B". This spreadsheet makes up my "projected result" which I quite honestly feel will be a better reflection of how the race is really going.
_______
So what have these over abundance of spreadsheet numbers been showing? Well the overall polling average (of the 22 pollsters) has shown a very small incremental increase in support for Donald Trump, with a slightly larger (but still incremental) decrease in support for Hillary. This has been consistently moving the average "spread" down. It was, for quite some time over five points. Then it spent several weeks over four points. As of right now it sits at 3.68 points.
Similarly, the projection spreadsheet has been consistently moving in Trump's favor. In fact, it has been moving a bit more than the unweighted polling averages has been. It also originally sat around five points, and has slowly but surely kept coming down, to where it is currently sitting today, at 2.3%.
Because of the sheer number of pollsters and the time it takes for each of them to replace older polls, my spreadsheets will not show wild changes in support, as will some of your other averages. What I am hoping it tracks is the more stable underlying trends that will not simply change with the latest headline (or because pollster A replaced pollster B).
14 comments:
Morning Consult: Going into conventions, this race is pretty much a tie
Taking Morning Consult’s numbers, we’re within the margin of error but Clinton retains a slight edge with her own convention yet to come. So how close is that to reality? First of all, we’re into the stage where the major polling outfits are sampling likely voters rather than just adults, so we may be seeing something a bit closer to reality. But what about the party breakdown in the sampling? While not as bad as some that we’ve seen, this survey still raises some questions. Looking at the crosstabs, MC has sampled 37% Democrats, 34% independents and 29% Republicans.
Does that strike anyone else as a bit off the mark? As of Gallup’s June survey, they’ve got the electorate pegged at 28% Republican (which is pretty close to MC’s sample), 39% independent and 31% Democrat. Mind you, I’m just doing this on the back of a cocktail napkin with my morning coffee, but if you adjust the Morning Consult numbers based on that scale, the results flip and Trump comes out leading by around two percent. It’s still basically a tie, but given the shape of this race there’s little to look at besides the leaners.
http://hotair.com/archives/2016/07/18/morning-consult-going-into-conventions-this-race-is-pretty-much-a-tie/
i've never heard of these morning consult guys before. anyone?
New Monmouth poll shows trump out to a nine point lead among independent voters.
The bottom line is, Hillary is losing it as she always does. In everything, Clinton fails yet profits by corrupt practices, the position in which she failed yielding opportunity for influence and special favors.
This is a race to the bottom: her inevitable failure, against the waning power of the corrupt Clinton political machine. Donald Trump is not untainted by corruption, and is a bigoted buffoon. But no more so than is Clinton, and I believe to a much lesser extent on all counts. Voters are perhaps awakening to that reality, and if so it only gets worse for Clinton.
that sheet is ridiculous. Each state has the numbers from only ONE county. I bet those counties were chosen because they have historically voted Republican
This is nuts.
Meh. Another interpretation is that polling has returned to where it's been most of the time since around Columbus Day. Clinton enjoyed two bump periods (April & early May, mid-June) and Trump has pulled even three times for only a few days at a time.
Both of these clods were well-known and well-loathed long before the nomination process begun. It's not too surprising it's been a pretty stable horse race so far. I'm not going to focus on week to week movement for another 6-7 weeks.
HOW DID WE FIND THE BATTLEGROUND COUNTIES?
Beginning in early 2016, Axiom Strategies set out to find a select group of counties that have historically been indicators for statewide presidential election results. We analyzed individual county election results dating back to 2000 in order to select areas that best fit our criteria. The search began by identifying counties that had correctly predicted the statewide result in each of the last four presidential elections. We then chose the counties that consistently matched the overall statewide result within a few percentage points. Seven counties that reside in seven battleground states fit our criteria and were selected to be Axiom’s Battleground Counties.
http://axiomstrategies.com/abc/#polling
Indy - as pointed out... by using 20 plus polls and replacing old polls with new ones, my averages have not really fluctuated like many other polling average.
At the end of the day, you are probably correct. There has been marginal overall movement (what there has been has been slow marginal movement in Trump's favor) and the bumps have evened themselves out...
Just because RCP shows an arrow up or an arrow down, doesn't really mean anything substantial.
I haven't been viewing your averages, or ever noticed an arrow on RCP.
This race could move either way, and quickly, since there are so many people disgusted with both candidates. In my view, Hillary has made a boatload of mistakes since the Dallas shootings, and unless she wises up and adjusts soon she may be looking up at Donald in the polls.
When you do state wide polls, as reported on RCP differs with your fantasy .
Pennsylvania 45.5 42.3 Clinton +3.2
Florida 44 43.4 Clinton +0.6
Ohio 43.4 41.6 Clinton +1.8Trending Up
Wisconsin 44.3 38.7 Clinton +5.6
Iowa 42.3 39.3 Clinton +3.0Trending Down
Virginia 43.8 39 Clinton +4.8Trending Down
North Carolina 44 42 Clinton +2.0
Georgia 40.8 45 Trump +4.2
New Hampshire 44.5 41.8 Clinton +2.7
Roger - I just explained how my two polling spreadsheets work, and where they have been going. Nothing more, nothing less. The facts are the facts.
Your insistence that suddenly Hillary is on the rise is simply not true. Three new polls that came out today are testament to that. Morning Consult came in at a two point race. Franklin Pierce (new) came in at three points. Monmouth came in at two points (had been six).
I let the numbers argue themselves. It was over five, it was over four, and it's now at 3.5.
Of the oldest four polls on my spreadsheet, three of them sit at 9,9 and 11. Do you believe the next ARG, Pew, or Dem Corps come in around double digits today?
Indy - Hillary has a history of snapping defeat from the jaws of victory.
I read you post on how the counties were good indicators. Without checking their accuracy, I thought it makes sense to take the statewide numbers than from a site that gives you numbers you want.
Yes, she has lost ground, but not as much as you want it took be.
Roger - It's hard enough to trust national polls at this point in time, much less state polls, much less county polls.
What all of them are showing, is a slow decline in support for Hillary.
RCP Average 7/2 - 7/18 Clinton: 43.8 Trump: 41.1 Clinton +2.7
Razor thin.
Post a Comment