Bad News for Cruz (and Trump)

The polling picture is getting worse for Trump and Cruz

Late last week, Twitter user Hope (@RealPoliticFact) commented on a general election polling chart I’d put together in late March.

Here’s a full-size recreation of the chart on which Hope commented:

Now, I’m aware of all the limitations of general election polling, etc. But these are pretty stark numbers. Trump is running below Cruz, and fading faster, but neither appears to be in a strong position to challenge Clinton in the general election.

Here’s what Hope had to say:

“Just starting to compare/contrast” is an interesting claim well worth examining. After all, the GOP primary has been absorbing the lion’s share of coverage, and the candidates have mostly been fighting with one another. Clinton and Sanders may have escalated their mutual attacks lately, but the GOP primary has been indisputably nastier than its Democratic counterpart.

About a month has passed since I first ran that chart, so now is a good time to revisit it, see what has changed, and see if Hope’s theory holds (at least right now).

At first blush the answer is a resounding no.

Trump and Cruz alike continue to struggle vis-a-vis Clinton in general election head-to-head polling. However, Cruz’s negative trajectory has actually accelerated while Trump’s has decelerated. Trump is still losing ground faster than Cruz is losing ground, but they’re starting to look more and more alike. Cruz hasn’t polled higher than 45% in one of these polls this month. Yikes.

But let’s make things interesting. First, I’m going to recreate these two images, but with their standard errors mapped onto the charts. That way we’ll be able to see whether or not we’re getting more or less precise estimates with our LOESS regression lines.

First the data ending a month ago.

And now the same chart including polling conducted since March 23.

As you can see, if you look at their standard errors, Trump and Cruz are now meaningfully closer together than they were a month ago. And they’re both worse off. So if you’re a Cruz supporter (or a Republican in general), it has to be admitted that the general election picture is bad and getting worse.

However, that’s not the whole story. The great thing about local regression is that we can increase or decrease how sensitive our estimates are. In other words, I can make these charts move more or less based on the polling that’s closest in time to the estimate lines. Note: this isn’t manipulating the data, rather it’s functionally assigning weights to which data points are more or less important, and how much more or less so.

Ok. Let’s really increase our sensitivity levels and rerun the charts. First, as before, the data from March 23 and earlier:

When we make our lines more responsive, March looks like a really good month for Cruz and a more or less flat span of time for Trump.

Yet when we rerun the highly sensitive estimates again with data up to the present, that big Cruz upsurge reverses. Instead, we get an even bigger Cruz crash, followed by nearly parallel upticks for Cruz and Trump.

These tell different stories, but only modestly so. A month ago, both of our estimates had Cruz above 40%. Today, both of our estimates have him below 40%. His popularity is still in flux, but there’s very little indication that he’s improved his position vis-a-vis Clinton since 2016 began. The contrast Hope thinks will change this situation needs to happen soon and it needs to be overwhelmingly favorable to Cruz. I’m not saying that can’t happen, but it requires a leap of faith to assume that it will because we don’t have much evidence for it right now.

Trump, by contrast, has loitered in the mid-to-high 30's for a couple of months. Indeed, neither estimate has put him above 40% for two months. His numbers may not have stabilized per se, but they’re showing less movement than Cruz’s. So if Trump is going to turn his deficit around, it’s going to be a bigger lift than Cruz’s already daunting task.

To close, I’ll add a chart of all candidate unfavorable shares in nationwide polling since 2015 began. I’ll throw in Rubio and Bush too, and I’ll make the estimates sensitive.

Trump has always been the most unpopular candidate in the race, at least since his favorable and unfavorable ratings started getting measured in late spring/early summer 2015. For Cruz fans, there may be reason for confidence. Cruz had passed Clinton in unfavorable ratings at the beginning of the month, but appears to be recovering.

Still, the graph above shows an extremely sensitive model. If we relax the sensitivity a bit, we have less reason for confidence.

So what do we know? Cruz has some persistent issues with his favorable ratings, and he’s still well behind Clinton in general election head-to-heads. He’s doing better than Trump for sure, but that’s not saying a whole lot.

Lastly, as regards Hope’s contrast theory, time will tell. Clinton’s negatives appear to be growing slowly, but they’re largely stable. She’s universally known and not well liked. The question is, how flexible are Cruz’s? The jury is out, but I’m not optimistic.

UPDATE: A reader asked me how Kasich stacks up in this story. The answer is ambiguous, but not great. Here’s the simplest presentation:

Note a few things. First, other than a handful of polls conducted shortly after his announcement, Kasich didn’t get heavy poll traffic until after his second place finish in New Hampshire. Hence Kasich’s standard errors are huge.

When we plot all three candidates together, we see that his overall numbers are definitely better than Cruz’s and Trump’s. Kasich has taken a less abuse and media scrutiny than the other two candidates to be sure. But regardless, Kasich is experiencing the same downward trajectory as Cruz and Trump.

All told, there’s not a lot to be optimistic about here either.


If you like what you see, sign up for my email newsletter and follow me on Twitter. As ever and always, if you’d like to send me angry emails about this essay, you can do so here.

One clap, two clap, three clap, forty?

By clapping more or less, you can signal to us which stories really stand out.