#5: Analyzing Trump’s Convention Bounce
(As of July 28th, the historic peak of his popularity)
Today we’re going to be analyzing the Real Clear Politics graph found on their General Election tracker. Trump recently had a convention and his numbers jumped, making the two week chart look like this:
That’s a pretty clear trend. Trump was losing, now he’s winning. The biggest boost came between Thursday and Monday, just after his convention and his well-received speech. Now let’s compare the bounce to the last year of data.
Ceilings and Floors
An interesting trend emerges when you look at the data this way. Trump’s ceiling until his convention is roughly aligned with Clinton’s floor. In this chart I’ve called out all the peaks and valleys for both Trump and Clinton:
At the beginning, Trump is low (34%) and Clinton is high (53%). Then Trump rises as Hillary falls and they’ve spent much of the year between 42% and 46%. Trump has had three stretches recently where he’s gone a fair amount lower, and Hillary’s numbers have gone up alongside.
But Trump has only barely gotten above Hillary a few times, and until now his peaks have lasted less than a week before going into a much deeper slide. So that’s interesting to note. Trump is appearing have trouble getting past 44.3%. In fact, this is the first time he ever has.
So Trump’s all time high is 44.3% and Clinton’s low is 43.1%. Let’s figure out what that means in relation to the bounce he just experienced.
The Bounce
Look again at the last two weeks and see what we can learn. Here it is again, slightly stylized so I can play with the visuals a bit:
Let’s overlay Trump’s ceiling in light red and Hillary’s floor in light blue. That will show the contested ground between Hillary’s low and Trump’s high and give us some context for the bounce he just received.
There you have it! Trump’s numbers shot through the ceiling of his support and yesterday he recorded his highest numbers ever. But there’s another data point here I hadn’t noticed at first. Hillary was at her low on the 14th and 15, a few days before the convention. And then she only went up from there. In fact, she hit her peak as Trump’s numbers were rising!
Here’s an overlay showing when the convention was going on:
And this shows the date range of the Democratic convention:
The Two Ways To Rise
If Trump’s numbers had gone up while Clinton’s numbers went down, we could assume the convention did a good job drawing voters away from her. But instead, Clinton’s numbers went up alongside Trumps, though slower. Then, after four days of extraordinary growth for Trump over the weekend, Clinton actually went up in three of those four days. And even the dip was quickly erased the next day.
So Trump did score a victory by increasing his own numbers, but he didn’t get the bonus of also pulling Clinton down. Either way, he’s still ahead as of today. But will he stay there?
What’s Next for Trump? Down. Probably.
I think Trump’s numbers are going down. First, it’s likely that Clinton will get a bounce that takes from Trump’s lead. Second, he’s above his all time high, and he historically drops after hitting 44%.
In the stock market, they talk about “support levels” for a stock. If you would normally pay $5 for Ben & Jerry’s ice cream, but after partying all night you’re willing to pay $10, you’re likely to return back to thinking the Ben & Jerry’s is only worth $5 once you sober up. I suspect Trump’s stock is experiencing a similar effect and will return to earth.
Inside Baseball on the Numbers
If you go to the Real Clear Politics website and look at the polling data, you’re taken to this view:
They do a moving average, and there some important details to understand about how they get their numbers. First, they only do one poll per organization at a time. See how Rasmussen is the top row but then it’s the first non-grey row? That’s how polls “fall off” the chart.
So if LA Times/USC issues a poll tomorrow and it finds that Trump is up by just +1, that’s actually bad news for moving average because it replaces the +7 that’s currently baked into the estimate. The same is the case for the Economist/YouGov poll that shows an outlier in Clinton’s favor. Notice how two of these bars are not like the others:
So if that big +5 poll gets replaced with a poll that only has Clinton up by +1, or shows Trump with a lead, it can really change the trajectory of the chart in a major way. But let’s say LA Times and Economist don’t update for a few more days. What’s the next poll to fall off? One showing Clinton with a +1 lead. So if the next poll to arrive has her up by at least +1, her polling average will improve. And then after three more polls, Trump’s big +7 lead will fall off as well.
Meaning Trump had better come up with some big +7 level polls shortly, or else his average is going to straight down again. At this point no one can really know if that will happen. But I know what I’d bet on.
