What State Polls Can Tell Us About The National Race

0
7
What State Polls Can Tell Us About The National Race

Earlier this week, I was working on an article claiming Joe Biden had better results in state polls than national polls.

Then on Wednesday the inevitable happened. Biden had some great results on national surveys, including 10-point leads in high-quality national surveys from Quinnipiac University and Marquette Law School, which increased his margin on our national survey average to 7.3 percentage points. And he’s had some pretty bad polls. ABC News / Washington Post polls showed him just behind President Trump in Florida and ArizonaFor example, two states where he usually held leads. (All numbers in this article are for late Wednesday afternoon.)

However, it is worth thinking about what government surveys are giving us about National Run. For example, suppose there were no national polls and instead we had to guess how the national referendum would come from government polls. Actually, it wouldn’t be that difficult!

Because we know how states vote in relation to the nation, we can use state surveys to estimate national results. Florida, for example, is typically about 3 points more Republican than the average state. If Biden had a 2 point lead in Florida, it would mean he would have a 5 point lead nationally.

Why bother when we can only look directly at the national surveys?

One good reason is that, when used correctly, government polls can actually provide a better projection of the national referendum than national polls themselves! In 2012, for example, our model – which is mostly used then and now Status Polls to predict the national referendum – showed the then President Barack Obama Win the referendum with around 3 points, even though His lead in national polls was only about 1 point. Indeed, the state polls gave a better estimate of the national referendum than the national polls. (Obama won the referendum that year by around 4 points.)

Remember there is one amount of information contained in these government surveys. They can gauge the preferences of populations more common in their states – such as Cuban Americans in Florida or Mormons in Utah – better than national polls ever can. And for at least this year, pollers who conduct government surveys tend to be rated higher than those who conduct national surveys, especially among pollsters like New York Times / Siena College and our colleagues at ABC News are doing more surveys this year than in the past.

Let me show you a simplified version of how our model uses these status polls. It’s similar to the Florida example above. For each state, our model calculates the national margin based on a state’s partisan bias plus Biden’s current lead or deficit on our survey average. Here’s that calculation for a broad group of purple states (or congressional district in the case of Nebraska) – anything from purple-red (maroon?) States like Texas to purple-blue (indigo?) States like New Mexico. I’ll show you the national margin based on both how a state voted in 2016 and 2012 – in some cases there are significant differences.

What purple state polls say about the national race

Biden’s survey lead or deficit in narrow states and what that means for its national margin, based on 2016 and 2012 results

Biden National Margin
StatusBiden’s current lead or deficit *Based on 2016 resultsBased on 2012 results
Texas-0.7+10.4+18.9
Georgia-1.0+6.2+10.7
Iowa-1.0+10.5-3.0
Ohio-1.0+9.2-0.1
Arizona+3.8+9.4+16.7
Nebraska 2nd District+4.3+8.6+15.3
North Carolina+1.2+6.9+7.1
Florida+1.7+5.0+4.7
Pennsylvania+4.6+7.4+3.1
Wisconsin+6.8+9.7+3.7
New Hampshire+6.9+8.6+5.2
Michigan+7.5+9.8+1.9
Minnesota+9.2+9.8+5.4
Nevada+5.8+5.5+3.0
Virginia+11.3+8.1+11.3
Colorado+10.2+7.4+8.7
Maine+13.9+13.0+2.5
New Mexico+12.9+6.8+6.6
Average (weighted according to voter turnout)+3.6+8.1+7.2

* Based on averages from FiveThirtyEight as of 4:10 p.m. Eastern on September 23. States (or district in the case of Nebraska) will only be included if they have at least five total surveys or surveys from at least three respondents. Sorted from red to bluish.

For example, on late Wednesday afternoon, Biden led our poll average in Wisconsin with 6.8 percentage points. It’s clearly good news for Biden to be the leader in Wisconsin, but what that means for the national race will depend on which year you compare it to. For example, in 2016, Wisconsin was about 2.9 points more Republican than the nation. A 6.8 point lead would mean that Biden had a huge 9.7 point lead nationally. But in 2012, Wisconsin was 3.1 points more Democratically than the country as a whole, a 6.8 point lead would only mean a national lead of 3.7 points.

In fact, you can see in the graph where Biden is doing particularly well or comparatively poorly compared to the way the state usually votes. The fact that Biden is almost linked to Trump in Texas is extremely impressive to Biden, for example. But it’s not a good sign for him that his lead in Florida is so narrow. In states like Iowa, Ohio, Michigan, and Maine, which moved heavily against Hillary Clinton in 2016, Biden’s performance is very impressive compared to four years ago, but not as much as it was in 2012.

However, if you averaged each state’s implied national margin and weighted it based on each state’s turnout, Biden ranks 8.1 points nationally using 2016 as a base, or 7.2 points using 2012 instead. In this case, that’s pretty similar to his current 7.3 point lead in national polls.

But what about polls in deep red and deep blue states? These states may not create much tension in the electoral college, but they count towards the referendum. First, Biden’s national leeway in red states where sufficient surveys are conducted:

What red government polls say about the national race

Biden’s survey lead or deficit in traditionally republican states and what that means for its national margin, based on 2016 and 2012 results

Biden National Margin
StatusBiden’s current lead or deficit *Based on 2016 resultsBased on 2012 results
Oklahoma-23.6+14.9+13.8
Kentucky-18.6+13.3+7.9
Alabama-14.9+14.9+11.1
Utah-12.4+7.6+39.3
Tennessee-12.3+15.8+11.9
Kansas-9.3+13.2+16.1
Louisiana-10.7+11.0+10.4
Montana-8.0+14.3+9.5
Indiana-14.4+6.7-0.3
Mississippi-11.7+8.2+3.7
Missouri-6.7+13.9+6.5
South carolina-6.8+9.6+7.5
Average (weighted according to voter turnout)-12.4+12.0+9.8

* Based on averages from FiveThirtyEight as of 4:10 p.m. Eastern on September 23. States are only included if they have at least five polls or polls by at least three pollsters. Sorted from red to bluish.

Biden has comparatively good polls in many of these red states. He isn’t much of a threat to winning them over, but quite a few of them have single digits deficits when Democrats typically lose them by double digits. However, whether this is real or reflects dubious polls is harder to say.

In very blue and very red states, it’s not uncommon for polls to underestimate the winning candidate’s profit margin. But maybe Biden, who is not as easy to typify as a liberal like Democrats like Clinton, Obama and John Kerry, has a slightly redder pull than other Democratic candidates of late. Overall, the red state polls imply that Biden has a national lead of 12 points (!) If you use 2016 as a base, or a lead of 9.8 if you use 2012 instead.

How about polls in deep blue states? Most of these states weren’t polled very often, in fact, but of those that did, California stands out:

What blue government polls say about the national race

Biden’s survey lead or deficit in traditionally democratic states and what that means for its national margin, based on 2016 and 2012 results

Biden National Margin
StatusBiden’s current lead or deficit *Based on 2016 resultsBased on 2012 results
Connecticut+20.7+9.2+7.2
New Jersey+19.1+7.2+5.2
Washington+24.4+10.8+13.5
new York+26.8+6.4+2.5
Massachusetts+34.1+9.0+14.8
California+29.3+1.4+10.1
Average (weighted according to voter turnout)+27.1+5.2+8.5

* Based on averages from FiveThirtyEight as of 4:10 p.m. Eastern on September 23. States are only included if they have at least five polls or polls by at least three pollsters. Sorted from red to bluish.

Biden leads there with “only” 29.3 percentage points, which is of course a very large lead, but it is actually a little below Clinton’s 30.1 point win in 2016. It makes sense, however. Biden’s strength over Clinton lies in older white voters, while California voters are majority and young. On the other hand, that 29.3 point margin in 2012 – or 2008 – would still be higher than Obama’s margins. Overall, the blue states imply a national lead of 5.2 points for Biden using 2016 as the base or 8.5 points using 2012.

Add it all up and the government surveys are a tiny slightly better for Biden than his national polls show – but just a tiny bit.

The results from the states of red, purple, and blue combined mean that Biden is 8.0 percentage points ahead at the national level (you get the same results using 2012 or 2016 as a base) compared to his current 7.3 lead Points in national surveys.

Perhaps that’s the broader teaching here We don’t really need national polls for all of that.

In fact, our model uses them quite sparingly. The election is carried out at the state level. While knowing where the national referendum stands can be useful for a number of reasons, you don’t need national polls to predict it. You can make pretty good (often better, in fact) estimates of the national referendum by using government polls instead.

LEAVE A REPLY

Please enter your comment!
Please enter your name here