Why is Bernie putting resources into northern, not southern states? Cuz it's smart delegate strategy

You may have seen this going around, about ad spending in the Democratic primary:

Hillary Clinton and her allies are spending $4.1 million on ads in 11Super Tuesday states, while Bernie Sanders is advertising in just five states at $3.3 million, according to ad-spending data from SMG Delta through Feb. 28.

Clinton and her Super PAC are up with ads in Alabama ($416,000), Arkansas ($43,000), Colorado ($540,000, Georgia ($295,000), Massachusetts ($543,000), Minnesota ($386,000), Oklahoma ($378,000), Tennessee ($421,000), Texas ($586,000), Vermont ($7,000) and Virginia ($452,000).

By contrast, Sanders is up in Colorado ($1.2 million), Massachusetts ($650,000) Minnesota ($680,000), Oklahoma ($690,000) and Texas ($32,000)

There has been a lot of speculation as to why the Sanders campaign is targeting larger volumes of television spending on certain states, while the Clinton campaign is spreading its a larger number of states.

Some have argued that it means Sanders can’t win and should give up. You may notice that the people who argue this are the same people who have argued for the past 6 months or so that Sanders can't win and should give up.

Others have argued that Sanders is making a big mistake, focusing on “winning” a decent number of states rather than on delegates, and ignoring southern states where Clinton will win blowouts. At this point in the argument it is usually presumed, without actually doing anything to check whether that is true, that if Clinton wins blowouts in the south, that it will translate in equal measure to delegates. And it is presumed that if Clinton can just expand the size of her presumptive southern wins, that she'll get even more delegates. This is all done without paying attention to the specifics of how delegates are actually allocated.

It is presumed that if Clinton is placing ad buys in states like Tennessee and Alabama, then it will help her run up the delegate score.

South Carolina

Some people have criticized Sanders for not camping out only just in South Carolina, to try and "win” it. What's that crazy guy doing, jetting off to Amherst and Norfolk in between stops in South Carolina, they ask?

But at least in terms of delegates, there is not much reason for Sanders to focus on South Carolina. My model currently has Clinton winning by about 61.7% — 38.3%, or 31 delegates to 22. But even if Sanders did much better, the model only thinks there is a single additional delegate (the At Large delegate to hold Clinton to a 6-5 split) within range. It is not likely to make very much difference at all, in terms of actual delegates, whether Clinton wins South Carolina 62-38 (as my model projects at the moment) or 55-45.

If you're wondering why Sanders was campaigning in Greenville South Carolina recently, the reason is probably that Greenberg is in the 4th Congressional District. That's one of the few places where a delegate is close (according to the model), and so it’s one of the few places that it makes sense for him to campaign in in South Carolina. For the same reason, we should a good chunk of any additional Sanders campaigning in South Carolina will be concentrated in SC-06 — because that's the only other place besides SC-04 where Clinton has a good shot of winning another delegate.

Let's look at the Super Tuesday states next. First, let's look at the states where Sanders is not buying TV ads, to understand why he isn't doing so.

The reason why is because if he improves his support by realistic amounts in those states, he wouldn't get much more delegates.

Then we'll look at the states that Sanders is putting more resources into.

The reason why is because there are a lot of additional delegates within his reach in those states.

Super Tuesday States Sanders is not targeting as heavily Alabama

It also doesn't make very much sense for Sanders to use too many of his resources in Alabama. The reason, again, is that Only a single additional Congressional District delegate (from AL-04) is possibly within his grasp. He'd need to get 50% in AL-04 for a 2-1 split. And if he improved by 4.3% statewide, he could get a single additional At Large delegate.

Now, you could argue that Sanders should put resources into Alabama to ensure that he doesn't do a bit worse than the model says in AL-05 and AL-07, to avoid Clinton winning 3-1 and 7-2 delegate splits. But to throw in a football analogy, if you are behind late in the fourth quarter and need a touchdown, that’s when you call passing plays. Sanders has to operate on the assumption that he'll meet those sorts of minimum thresholds and that he’ll gain a bit more ground in the national polls — because if he doesn’t, he won't win anyway.

Arkansas

Similarly, in Arkansas, there is only one delegate (AR-04) that Sanders could gain by improving by less than 5%. He could also possibly win another delegate in AR-02 by improving 5.5%. But voters in Arkansas have known Clinton for a long time, and are less likely to be persuadable than voters elsewhere. He does need to pass minimum thresholds to stop Clinton from getting more than 4-3 and 2-2 delegate splits for the At Large and PLEO delegates, and more than a 2-2 split in AR-03.

Georgia

Georgia is the one state that goes against my point, given my model's current projection. The model currently projects Clinton to win Georgia 67.9% to 32.1%. This is driven heavily by a few recent polls, particularly from the GOP pollster Landmark, which have shown very bad results for Sanders. This would be projected to translate into

If those polls — and the model's projection — are correct, and Sanders’ support is indeed that low in Georgia, then it would make a lot of sense for Sanders to focus resources in Georgia. In fact, it would be malpractice by the Sanders campaign not to do so. If Sanders could just improve by 3 or 4 percent (to 36% or so statewide), that would be enough for Sanders to win up to 10 additional delegates. That is a very large delegate swing for a small change in support.

On the other hand, if those polls are understating Sanders' support and he is pretty sure to cross that crucial threshold of about 36% support statewide, then it doesn't make sense for Sanders to focus resources on Georgia. If Sanders does get those 10 additional delegates, then Clinton . So Sanders could get as much as 40% of Georgia’s delegates while winning as little as 36% of the statewide vote because of the way the delegate math works out (at least if the model is roughly accurate).

My guess is that the actual situation in Georgia is probably not right by this threshold. Because neither the Sanders campaign nor the Clinton campaign is stupid, and if it were, both would be throwing much more of their resources into Georgia.

Tennessee

With Tennessee, it's the same story. There are not very many additional delegates in realistic contention. If Sanders improves statewide by 5%, he’d only get 1 additional delegate (from TN-05). It simply doesn't make any sense in terms of delegates to put many resources into TN. As long as he passes basic, very low thresholds, he’ll get all the Congressional district level delegates shown here.

Texas

In Texas, there are at least some delegates within reach for Sanders, just not very many in proportion to the enormous size of the state. The reason for that is that so many of the State Senate districts have 4 delegates which will almost certainly split 2-2. And many of the others that have 5 delegates are probably just a bit too far out of reach for Sanders. Other delegates elsewhere are much easier to swing.

While it makes sense to try to swing particular delegates in particular State Senate districts, it doesn't make sense to indiscriminately run TV ads on Texas’ 20 or so media markets. Outreach is better handled in other ways, like on the phone, on the doors, and through the mail. In fact, I can attest that Sanders has ongoing phonebanking efforts in Texas — which are not limited to calling people in places like Austin.

Note: Texas allocates its delegates on the basis of State Senate districts rather than Congressional Districts. I just added this to my model. These projections do not take into account variations in urban/rural/suburban neighborhood distribution for technical data reasons.

Virginia

Note: Virginia's Congressional District boundaries have changed due to a redistricting lawsuit. These projections are based on the demographics of the old districts, since I don't have up to date data, so they will be less reliable.

In Virginia, there are also not many delegates realistically within reach. Improving by about 3 percent could get him 2 more at large and PLEO delegates, but to get many more he would need to improve by 7 or 8 percent (enough to win many CDs outright and get 3-2 delegate splits). That is probably a bit out of reach.

Sanders was just campaigning in Norfolk, so that suggests that maybe the changed demographics of the redrawn districts make the Sanders campaign think that they can swing an extra delegate out of VA-03. So generally VA is not the best state for Sanders to focus on, but it may make sense for him to target a particular Congressional District or two.

Super Tuesday States Sanders is targeting heavily

On the other hand, in the states Sanders is targeting more heavily, and is running TV ads in, the delegates look very different. There are many more delegates that Sanders can win by improving his numbers in those states.

Colorado

Colorado has a caucus, which is a big part of the reason why Sanders is focusing on it. Since caucuses have lower turnout, every additional voter a campaign can turn out has a greater impact. You get more bang for your buck by putting more resources into caucus states.

Currently the model has him at about 53%. .6% more statewide gets him another At Large delegate. If he improves by 2.7%, that is enough for a 5-3 delegate split in CO-01. another 4.9% and 5.6% in CO-03 gets another 2 delegates. In 2008, Obama got 67% in Colorado. While Sanders probably won't do as well as that, it is conceivable that he could get into the low 60s. If he does that, he could win 4 additional delegates (PLEO, CO-02, and CO-06, and another At Large delegate for a 9-5 split). That would be the best case scenario for Sanders, and although that is a reach, it is possible, and it is what Sanders would like to be the result.

So in Colorado, unlike in the states we have looked at previously, there is at least the possibility for Sanders to pick up a lot of delegates. That is why Sanders is making a big push there.

Massachusetts

And why is Sanders putting resources into Massachusetts? Not just to get a "win,” but to get delegates. MA-01 and MA-02 are both in Western Massachusetts, and Sanders could get 4-2 delegate splits in both of them. Just a day or two ago, Sanders was in Amherst. Why? Because it is in MA-02. Even before the New Hampshire primary, Sanders also campaigned in Worcestor. Why? Again, because his campaign was playing the long game, shooting for that 4-2 delegate split in MA-02. Another 4-2 delegate split is at least conceivable (if a reach) in MA-03, which borders New Hampshire. Another 4-2 delegate split is possible (if a reach) in MA-04.

In the eastern part of Massachusetts from Boston to cape cod (MA-05 through MA-09), every single one of these districts have 7 delegates. That means whoever wins each of those delegates gets a 4-3 split. While 5-2 splits are probably a bridge too far, if Sanders ensures that he gets 50% + 1 in each of those districts, that is 5 delegates.

Finally, if Sanders improves statewide by another 1.5%, he gets another PLEO delegate. If he improves by another 4.8%, he gets another At Large delegate.

So the reason to compete in Massachusetts is not just to say that he has "won” a state. It is because MA could net Sanders a good number of delegates.

Minnesota

And what about Minnesota — why is he putting so many resources there? Again, it is not just to “win” a state. It is to win delegates.

Firstly, MN has a caucus like Colorado where resources can go farther.

The second thing to realize is that the model's estimate that Sanders only has 51.7% support in MN may well be an understatement. The most recent poll from MN is a Mason Dixon poll that had Clinton up 59-25 (from January). At least personally, I am highly skeptical that is anywhere close to accurate. Based on demographics, MN should be very friendly to Sanders. Given the vote in New Hampshire, and given that Obama beat Clinton 66-32 in MN in 2008, at least I personally wouldn’t rule out Sanders getting nearly or even in excess of 60% in MN.

Additional At Large and PLEO delegates are all within reach, and additional delegates in MN-02, MN-03, MN-04, MN-05, and MN-08 are all possible if indeed it is realistic to think Sanders could get 60% statewide.

Oklahoma

Finally, there has been some question as to why Sanders is targeting Oklahoma. There has been some suggestion that it is just to "win another state” so that he can say that he won another state. There is some reason to want to win a state simply for the sake of winning a state. The media does focus overly much on how many states presidential primary candidates “win,” so Sanders should want to win a state like Oklahoma.

But if you look at how delegates are allocated, you can see that's not the whole story. For a small state with cheap media markets, Sanders can swing an additional 3, or possibly 4 delegates without much additional improvement. OK-01, OK-02, and OK-04 all have an odd number of delegates (5), and winning 3-2 delegate splits in those goes hand in hand with winning statewide and winning the PLEO delegates 3-2.

Compare that to a state like Tennessee, and you can see why Sanders is focusing resources on Oklahoma rather than on Tennessee — Not just because he could “win” a state, but because it will get him more delegates.

Vermont

Finally, Sanders isn't running ads in Vermont, but he is likely to do very well there. Recent polls from PPP and Castleton have both shown Clinton under 15% in Vermont. If Sanders can indeed hold Clinton under 15%, he will get all 16 delegates. On the one hand, that's much fewer delegates than a state like Virginia with 95 delegates. On the other hand, consider this. The model currently projects Clinton to win by 54 delegates to 41 (a 13 delegate margin). So if the actual results are close to that, little old Vermont with its 1 Congressional District could be more than enough to entirely offset the much larger state of Virginia, with its 11 Congressional Districts. That's not bad for a little old state like Vermont.

The Why’s and Wherefores of Delegate Math

Question:

So why is Bernie focusing television resources on Colorado, Massachusetts, Minnesota, and Oklahoma? Why is he not spending television resources on places like Tennessee, Texas, and Alabama? Why is he crisscrossing around the country campaigning in places like Amherst MA, Greenville SC, Norfolk VA, Denver CO, and Minneapolis MN?

Because his campaign is playing the strategic, disciplined long game — focusing cannily on states and Congressional districts where he has a shot to swing extra delegates. Yes, that could help him win a good number of states on Super Tuesday. And that's not a bad thing. But it’s also the way that Sanders can maximize his delegate count. This are the same reasons that Obama didn’t throw all his cash into states like California on Super Tuesday in 2008.

The real question is why is Hillary Clinton spending ad money in Tennessee and Alabama?

What delegates is she trying to gain? Even if my model is overestimating Sanders in TN, it would have to be doing so by a lot (6% or 7%) in order for her to be on the edge of picking up a reasonable number of delegates. And that would mean Sanders would be at only 33% in TN On the other hand, if it is underestimating Sanders, there are not many additional delegates that Sanders could win anyway. And in Alabama, if she really is in a position where she is trying to do better than the 63-37 that the model currently projects, then she will have won anyway. And if she's trying to stop Sanders from doing better than 37%, she’s wasting her money because Sanders wouldn't get many more delegates by doing better than that anyway.

So what is Clinton doing?

As best as I can tell, it seems that Clinton is trying to win large victories in southern states solely in order to try and drive a media narrative. But in the process, is she once again wasting resources trying to increase her popular vote in places where it won't translate into delegates?

This is a continuing part of an ongoing series using polling data, past exit poll data, census data, and other data sources to analyze the 2016 Democratic Primary.

Previous posts are:

Poll Meta-Analysis: The Bernie and Hillary 2016 Coalitions, and how they compare to 2008 Obama/HRCPoll Data Analysis: The Current State of the Democratic PrimaryHow the delegate math shakes out for Bernie and Hillary down to the Congressional District levelBernie Sanders Did Much Better With Non-Whites In Iowa Than You ThinkDemocratic Primary Model, Feb 8 Update (Pre-NH)Democratic Primary Model, Feb 9 Update (Pre-NH)Democratic Primary Polls for All 50 StatesNevada Caucuses Preview and Democratic Primary Model Projections

For more detail on how delegates are allocated across different states, check out this excellent resource from Torilahure.