Do states with big trends tend to have reversals in the next election?

[Mr. Morden]

This seems like a very basic question that someone must have looked into at some point in time, but I have no idea what the answer is:

If you have a state that trends strongly towards one party in a given presidential election, then in the *next* election, is it more likely to keep trending in that direction, stay roughly where it is, or have a reversal, and trend back towards the other party?

I’m asking because people are asking about the big trends we saw in 2016, and some are saying that things like Iowa trending so strongly R are flukes, and you should look at the recent electoral history of the state, not just the most recent election.  Others are saying no, this is the new normal.

But we’ve had plenty of elections before, so someone must have studied this before.  If I wanted to predict the ordering of the states in the next election, should I just use the most recent election as the new default, as if that’s the “new normal”?  Or should I average over the last five elections, or what?

[Skill and Chance]

I would say that shifts between the states tend to happen in 1 or 2 massive swings, not in a slow linear drift toward one party or the other.  I would be most inclined to take the >5% and especially >10% trends seriously as an indication of the future and to treat minor trends as statistical noise.  The obvious exception is if we are talking about a candidate's home state.

[mencken]

Looking at cases of >13% trends (what Iowa did in 2016) since 1980, and omitting obvious favorite son effects (Georgia 1984, Arkansas 1992 and 2000, Kansas 1996, Hawaii 2008, Utah 2012), the trend of the state in the next election is pretty normally distributed. So, it could swing back, it could swing even further, but most likely it will stay pretty close to its new normal.

[Xing]

I'd say it depends on why that trend occurred. Sometimes, it's because of a certain issue or candidate resonating with voters in that state, in which case the trend often reverses itself in the following election (think Indiana 2008). In other cases, it is due to population changes, or voters aligning with a different party (West Virginia 1996-2016, Virginia 2004-2016, etc.)

[Mr. Morden]

I'd say it depends on why that trend occurred. Sometimes, it's because of a certain issue or candidate resonating with voters in that state, in which case the trend often reverses itself in the following election (think Indiana 2008). In other cases, it is due to population changes, or voters aligning with a different party (West Virginia 1996-2016, Virginia 2004-2016, etc.)

Right, different cases are different.  But what would be great is if someone had done the math at some point, to work out the statistics of how often there's a reversal and how often there isn't.

[Adam Griffin]

Looking at Atlas trend from 2000-2016, there are 88 examples of states trending more than 5 points. Of those 88, 68 can be compared (we don't know yet whether the 20 states that trended by more than 5 points in 2016 are going to continue their trend, reverse, etc).

Of those 68 that can be compared:

• 11 (16%) trended by more than 5 points toward the opposite party in the next election
• ◼️◼️◼️◼️◼️◼️◼️◼️
• 14 (21%) trended by more than 5 points toward the same party in the next election
• ◼️◼️◼️◼️◼️◼️◼️◼️◼️◼️◼️
• 17 (25%) trended by less than 5 points toward the same party in the next election
• ◼️◼️◼️◼️◼️◼️◼️◼️◼️◼️◼️◼️
• 26 (38%) trended by less than 5 points toward the opposite party in the next election
• ◼️◼️◼️◼️◼️◼️◼️◼️◼️◼️◼️◼️◼️◼️◼️◼️◼️◼️◼️

So if we were to draw any conclusions from this limited sample size, it might be that a state that has a fairly large trend toward one party in a given presidential election has a larger chance of experiencing a smaller trend in the opposite direction in the next election than anything else.