Princeton Election Consortium

A first draft of electoral history. Since 2004

And then there were…four?

July 30th, 2014, 4:33pm by Sam Wang


Of the 36 Senate seats up for election, up to nine of them have been worth watching closely. These races will determine who controls the Senate. Over the last month a few seem to have dropped out of consideration. Now, with a new burst of polling, the playing field – at least for now – includes as few as four races. These races, plus a few others, will be crushed by attention.

The new polls, from YouGov/NYT and other organizations, confirm what I’ve said and hinted at: Georgia might be moving out of the competitive range (toward the GOP), and maybe Alaska too (toward the Democrats). That would leave four competitive Senate races. In an election held today, there is an 85% probability that each side would have between 49 and 51 votes; much of that probability is concentrated in a perfect 50-50 split.

Only four races - Kentucky, Iowa, Louisiana, and Colorado – have no clear leader at the moment. If we assign all the other races, that gives 48 Democrats/Independents and 48 Republicans.

Here is where key races stand today. Note the return of Jerseyvotes, which I’ll explain in a moment.

State Median margin Jerseyvotes
Kentucky McConnell +2.0±1.3% 100
Iowa Tie 69
Louisiana Tie 55
Alaska Begich +5.0±6.0% 52
Arkansas Cotton +4.0±1.3% 41
Colorado Udall +1.0±1.8% 28
Montana Daines +7.0±0.7% 1.2
Georgia Perdue +6.0±2.0% 0.8
Michigan Peters +5.5±0.9% 0.4
North Carolina Hagan +7.0±2.7% <0.1

Strictly according to PEC’s method of taking the last 3 polls, there is still statistical uncertainty in Mark Begich’s lead  in Alaska. However, only one poll is fresh, this week’s YouGov result showing him leading Dan Sullivan by an amazing 12 points. The Begich campaign has lately given Sullivan a beating over a serious gaffe – a Sullivan ad shot on top of  a building that Begich got funded. Now we see the payoff.

Some have complained about the fact that YouGov uses Internet-based sampling. YouGov has an excellent record, and Doug Rivers has pioneered the accurate use of their approach to polling. Not only am I happy including the data point, I think it’s our best measure of what is happening in Alaska today. At this point, Begich is the clear favorite.

The remaining four races will be extremely hard-fought. Buckets of money are pouring in – an estimated $2 billion in outside ads, i.e. those that do not come from either candidate. This is a 70% increase from the 2010 campaign. Much of the money will go to those four races, as well as other states.

In the face of that, how can little citizens make their voices and dollars go farthest? That’s where jerseyvotes come in. I introduced this measure in 2004 as a way of quantifying how much power an individual voter has to influence the overall national outcome. To learn more about the concept of individual voter power and how to quantify it, read this explanation. In 2004-2012, I applied this concept to the Presidency. This year, the target of my analysis is Senate control.

The name jerseyvotes comes from the fact that here in New Jersey, we have virtually no influence over the election. Cory Booker (D-NJ) is going to win, period. On the other side of the aisle, Pat Roberts (R-KS) is in exactly the same position. So New Jersey and Kansas voters are of little consequence to the Senate. However, voters in Iowa are another story, where today Bruce Braley (D) and Joni Ernst (R) are tied. An Iowa voter has thousands of times more influence than I do because the race is so much closer (and to a lesser extent because Iowa’s population is smaller than New Jersey’s).

That influence is quantified in Jerseyvotes. Using the amount by which one voter can influence overall Senate-control probability as the yardstick, Kentucky voters are the most influential, at 100 Jerseyvotes each. Iowa voters have 69 jerseyvotes each. Here in New Jersey, my vote is worth far less than 1 jerseyvote. The reason NJ votes are not worth 1 jerseyvote is that our votes are too worthless to be a stable currency – worse than a Weimar Reichsmark. Instead I am normalizing to the most valuable votes – which today are in Kentucky.

If you want to direct your campaign donation wisely, give to one of the top four states I have listed – or to Alaska and Arkansas, the other races with substantially high jerseyvotes.  All races are accessible at the links to the left. Democrats, go to ActBlue. Republicans, go to the NRSC. Libertarians can give to Sean Haugh in NC.

Tags: 2014 Election · Senate

22 Comments so far ↓

  • Jeremy Leader

    If you’re interested in a “stable currency” (so that you can compare Jerseyvote numbers from one election to the next), wouldn’t it be best to normalize it to the state with median voter influence, rather than to a particular state (New Jersey) or the state with highest voter influence?

  • Woo

    I just donated another $120 on your ActBlue page to those four states’ candidates.

    • Sam Wang

      That is great. Conditions may shift a little, but it is likely that those races will still be important come November.

  • Joseph

    I did a little diving into Arkansas, mainly because I’m interested in what effect the Clintons might have there. I noticed via your link to the Huffington Post that the Harstad poll of 7/7-7/10 is showing a large lead by Pryor (unlike the other polls shown). Digging a little deeper, I noticed that the Harstad poll includes numbers for the Libertarian Party (3%) and the Green Party (2%), which most of the other polls don’t show.

    That makes me wonder about the effects of subsuming micro-numbers like these in macro-numbers, especially in a state that might be swayed by two great orators (the two Clintons). I wonder if it might not be worth while to look at these four parties accross these few states, and see if, in these close elections, they might end up being potential “power brokers”, swaying just enough voters in one direction or the other to move the needle from R to D or vice versa.

    • Joseph

      Just to clarify: By “power brokers” I’m referring to the small parties, most importantly the Libertarians and the Greens.

    • Amitabh Lath

      Yes, I cannot imagine tea party voters will meekly pull the lever for Cochran in MS after what they feel was a runoff election “stolen” by illegitimate voters. But since Cochran has >+10 it isn’t likely to sway that election.

      I am skeptical about the likely voter models and demographic weights these polls are using. Two trends I see in news coverage: 1) Obamacare worked better than expected (NY Times has an article today on people who never had health insurance climbing the learning curve) and 2) House GOP voted for an immigration bill that will alienate (and energize) Hispanic voters.

      If both these groups (recently insured, and hispanic voters) turn out in larger numbers than in previous midterm election due, then polls that weight by historical patterns will be off.

    • Sam Wang

      It’s hard to predict in advance how third-party candidacies will do. Generally, support for minor candidates fades in the home stretch. The exception is when the candidate is so strong that his/her supporters think there is a chance for victory.

      My general sense is that the aggregated median of pollsters, whether or not they include minor candidates, comes pretty close to capturing the appropriate amount of support. It is not surprising that the Pryor campaign would include the minor candidates in their survey, in order to get a more favorable result. Strategically, they could get a small advantage by pushing the Libertarian candidate.

    • Joseph

      “Strategically, they could get a small advantage by pushing the Libertarian candidate.” Sometimes even a tiny advantage is enough. The archetype is, of course, Bush/Gore and Florida. To what degree did the Republicans successfully “feed” Ralph Nader’s campaign, and thus tip the balance just enough in favor of George Bush? I’m sure I recall that happening. And why, therefore, should the Democrats be shy about doing the same thing to the Republicans? To my knowledge, “divide and conquer” still works.

  • Amitabh Lath

    According to your own table, Alaska is a tie (5+-6 is basically a wash). Alaska should be added to the list of “No clear leader’ races.

    The CBS poll with Begich +12 seems an outlier. Thankfully, your method of taking the median handles outliers correctly.

  • Bob Grundfest

    The GOP is not helping itself by dithering over immigration and not being able to get things done. The Democrats can make an issue of the Republicans not being able to govern. Perhaps that might sway the close races.

    • ArcticStones

      The Democratic Party’s inability to set the narrative is … stunning!

  • Vaughan

    The overly long span of the current Bull Market indicates a massive correction on its way. This event may well throw everything into the polling blender if it occurs in mid October and result in a ‘grey swan’ moment when all the pollster and pundit predictions are wrong.

  • Pinkybum

    So I read the explanation but my confusion still stands. And my original question is still valid (despite the fact I had not read the explanation in the FAQ). The population of Kentucky is 4.4 million and the population of Iowa is 3.1 million – wouldn’t the power of voting in Kentucky be weaker than in Iowa?

    • Sam Wang

      It’s calculated by (a) moving one state’s poll median by a small amount, (b) measuring the change in Senate control probability, and (c) dividing that difference by state population. This gives a number that is proportional to an individual voter’s power to influence who controls the Senate.

      In this definition, small-state voters have an advantage. So until the YouGov poll drop, Alaskan voters were on top.

  • Pinkybum

    I don’t understand Jersey votes. Why do the voters in Kentucky have anymore impact over Senate control than in Iowa? Is it because there is a weighting factor for Mitch McConnell?

  • Ron Pitts

    Dr. Sam, 36 Seats are up in NOV. Great Analysis. I have been thinking if we win 2 of the 3 (AK LA & NC) we will retain control. So I hope AK and NC stay right where they are. We would only need IA and CO under your scenario and I like that better. (although Braley and Udall are unexciting enough to keep it close)…..Great article as always.

  • Jpell

    Surprised that North Carolina has been removed from the list of competitive races. Seems the lead is about equivalent to Michigan, but with the added uncertainty of the third party candidate, right?

    • Sam Wang

      MI and NC are currently several sigma to the Democratic candidate. it would take a 5-point swing for that to change. Such a large swing is somewhat more likely for NC, where there is 3rd-party support (Haugh, a Libertarian) which is likely to fade in the home stretch.

  • 538 Refugee

    Well, the economy is looking up. Everything else is down. So I guess that is a stalemate in terms of ‘big movers’ for this elections cycle so far?