Understanding the USC - LAT Poll Tracker

I laugh every time I see or hear someone (Donald included) point to the USC - LAT poll tracker, as proof of Donald's competitiveness. There are only two pollsters whose tracking polls appear to show high volatility, despite the fact that there is very low volatility in this year's election -- that should be a clue that the problem lies in house bias.

The two important components of the USC - LAT poll tracker are the voter preference and the voter expectation numbers. When you look at them, one -- the voter preference -- appears to be volatile while the other -- voter expectation -- appears to be fairly stable.

USC - LAT voter preference tracking

USC - LAT voter expectation tracking

Looks can be deceiving. The reason why the voter preference looks volatile is because it appears that the two are changing leads. So what happens when you vertically adjust (align) voter preference and voter expectation for each candidate?

Realignment of USC - LAT trackers for Hillary

Realignment of USC - LAT trackers for Donald

In the two charts above, the left side is tied to voter preference while the right side is tied to voter expectation. In essence, the right side shows how much I moved the line up or down; that's the house bias effect.

The voter preference line generally matches the voter expectation, right? When the expectation line tracks lower than the voter preference it roughly means that there is lower enthusiasm, whereas when it is tracking higher it conversely means that there is higher enthusiasm.

Rather than spend even more time attempting to recalculate the numbers based on their separate enthusiasm (intention to vote) tracking number, this simple graphical alignment highlights that the USC - LAT tracking poll has a house bias of 10 points total (5 points upward for Hillary and 5 points downward for Donald). To get to this crude extrapolation, you subtract the difference between the right side and the left side for each chart.

So, when you add 5 points to Hillary's number (43.6 + 5 = 48.6%) and subtract 5 from Donald's (46.2 - 5 = 41.2%), for a total delta of 10 points, you come up with something that actually matches the aggregate of national polls in Pollster's 2-person race, Hillary = 48.1% / Donald = 41.7%.

It's an extremely crude method of removing the house bias, but this is what it looks like when you superimpose Pollster's tracker over the adjusted chart.

Superimposed USC - LAT adjusted poll tracker over Pollster aggregate

Feel free to ignore it if you want, but the USC - LAT tracking poll, once adjusted (crudely) for house bias, looks like it's just a reflection of the poll aggregation.

Without running the numbers to check for correlation, you can visually see that the uniformly adjusted values of the USC - LAT tracking polls match extremely well to poll aggregation.

And that's why I laugh when someone talks smack about how Donald is winning in some of the polls.