Welcome to Running Mates, our portal for all of our content related to the American vice presidency including our podcast, the 2020 Vice Presidential Tracker, and featured articles on vice presidents new, old, or never to be.
Latest Running Mates Content
Lars and Michael discuss Richard Nixon’s deliberation over whether or not to keep his vice president, Spiro Agnew, on the ticket and take a deep dive into the bleak prospects for the Democratic nominee George McGovern and his struggle to even find a running mate in the first place
It took a few days, a lot of miscommunication, a lot of errors, and some late nights, but we finally have the results from both the Iowa Democratic caucuses and the New Hampshire Democratic primary. The Iowa caucuses, the 89th Academy Awards of its time, what with a miscalled winner and a lot of confusion,1Good thing the caucuses didn’t accidentally pronounce someone dead, that would have been awkward. seemed to ultimately (if anticlimactically) deliver a split verdict for Vermont Senator Bernie Sanders and former South Bend, Indiana Mayor Pete Buttigieg. The New Hampshire primary, on the other hand, went off more or less without any drama, but reiterated the idea that this is a crowded, close, and increasingly splintered race with Sanders and Buttigieg neck and neck winning around 25% of the vote each, the only other candidate eligible to receive delegates being Minnesota Senator Amy Klobuchar, who came in a close third. After some stunning underperformances by Massachusetts Senator Elizabeth Warren and former Vice President Joe Biden, there’s a big question about how this race shapes out as it moves into larger and more diverse states in the coming weeks before culminating in a make-or-break Super Tuesday.
With the first two major races in, it’s time to take a look back at our Vice Presidential Tracker, which we’ve been updating after each contest. This is our first analysis on how it’s looking since voting has begun, and while Iowa and New Hampshire are important simply because they are the first states in the primary, they are by no means representative of the demographics of the Democratic Party, and represent a very small fraction of the delegates that will ultimately determine the nominee. Nonetheless, because we know this, and our favorite election modelers over at FiveThirtyEight know this, we track our rankings against their candidate’s probabilities of reaching a delegate majority. You may recall (or not, because it’s long and starts to talk about math) in our methodology that the “Rank-Score” for the potential vice presidential nominees runs the VP Score through every single possible nominee weighted by their odds at securing a delegate majority and thus becoming the nominee. So, let’s take a look at where we are as of February 15, the weekend after the New Hampshire primary.
Let’s start with the ostensible front-runners. The odds of either Sanders (36%) or Biden (13%) being the nominee are at just under 50% at the time of publishing this article. These two front-runners are unique in that, ideology aside, they are relatively similar in what they each bring as nominees. They’ve both been in the federal government for over 25 years, they are both white and male, and are from very small, very liberal states. This also means that so long as one of them gains at the expense of the other, the vice presidential standings won’t change that much, since you’re running each potential VP pick through the strengths for Biden and Sanders adjusted by their odds of getting a majority of delegates. This is why Senators Kamala Harris, Tammy Duckworth, Mazie Hirono, and Catherine Cortez Masto, along with former Housing and Urban Development Secretary Julian Castro round out the top five. They are each also individually in the top five strongest picks for both Biden and Sanders. In the senators’ cases, they are each relatively new-on-the-scene (Hirono is the only one so far to have faced reelection as a senator), and are prominent women of color in the Democratic Party. Castro has an advantage in that he’s not up for any kind of reelection and is from a large “swing” state (how contentious Texas will really be in a close presidential year is a matter for debate, see: Beto O’Rourke). Castro and the senators are all of course non-white, which boosts them when we’ve got white men as the frontrunners for the nomination, but another key thing they all share is that they on average have about twice as much non-federal experience as they have federal experience, which balances nicely against Sanders’ and Biden’s decades in federal service.
|Potential VP||Gain from State||Federal XP (years)||Non-Federal XP (years)||Gain from XP Score with Biden||Gain from XP Score with Sanders|
Harris coming out as a heavier front-runner than the rest of these (she’s rank-scored at almost 40 points) is not a surprise. Recall that even though she’s not from a close state (California’s partisan lean in the tracker at this moment is 24 points Democratic), she is from the largest state, which does ever so slightly factor into her score (when in doubt, bigger is better); this also of course boosts Illinois Senator Duckworth and Castro, both of whom are from large states. Senator Cortez Masto gets an added boost in that she is from a close state, Nevada, but loses a good chunk of her score in the coefficient of our calculation due to the risk of her seat being filled by a Republican in a coming election.
Considering she’s been floated as a Biden running mate pretty heavily, I would say that Harris being in the top five is a good indication that the tracker is generating feasible and relatively intelligent results. Considering Castro is also frequently mentioned as a running mate for some of the remaining candidates, his presence in the top five is also encouraging. The top five picks at the moment reflect a good balance of geographic diversity as well, balanced strongly against the eastern frontrunners.
Closing out the top ten in our overall standings we have another notable name: presidential candidate Amy Klobuchar. Klobuchar is buffeted by the competitiveness and medium-size of her state, Minnesota leans Democrat by only 2 points and carries 10 electoral votes.
As we approach the Nevada caucuses, South Carolina primary, and then Super Tuesday, the one potential VP to watch closest is Mazie Hirono, who is in the top five strongest picks for not only Biden and Sanders, but also the former mayors in the field, Pete Buttigieg and Michael Bloomberg. Buttigieg could get traction for squeezing Sanders’ lead in New Hampshire and eking out victory in Iowa, and Bloomberg could start to figure in more heavily what with his unofficial entry into the contests on Super Tuesday. Hirono has received little mention as a possible vice presidential pick, but this looks to be an oversight, as her metrics are good, and she’s been increasingly outspoken about the current administration and the high profile events of the last few years, slowly gaining more name recognition within the party. Her balance of just enough state and federal experience makes her worthy of any shortlist by these septuagenarian frontrunners, and by any up-and-coming mayors who lack much experience at all.
That said, if the Klobucharge (ask my editor, but I stand by I came up with this before the other news outlets did after the New Hampshire primary) proves pervasive and Amy Klobuchar begins gaining traction, or if Elizabeth Warren makes a comeback, we’ll see some new names bumping into the top ten. Tim Ryan and Julian Castro will probably pass over the senators as the highest ranked in this case, and we could see New Jersey Senator Cory Booker and some former Obama administration cabinet members nudge out a few of the governors rounding out the current top ten.
I dropped businessman Andrew Yang off the tracker as he dropped out of the presidential race on the night of the New Hampshire Primary. I have not added him to the potential vice presidential picks, but if one of you Internet people read too much into his ominous tweet, drop me a line and I’ll throw him in there for you.
Due to recent speculation, I added Hillary Clinton as a vice presidential candidate, maybe you’ve heard of her? She is actually a fairly strong running mate for Pete Buttigieg, so keep an eye on him (and her) if Buttigieg continues to do well in the primary.
Now that FiveThirtyEight is tracking specific candidates under it’s “All others” on their primary forecast, we will be using those numbers for each specific potential nominee instead of applying the broad odds for that category at large to each candidate not listed in their overall tracker.
Remember that the state metrics for candidates will all adjust over time as the generic ballot changes. I update this after each primary as well, but it can have a miniscule impact on the competitiveness of various states and, at the margins, can even make a state fall just into our out of the “competitive” range, which could substantially affect a specific VP pick.
Lars and Michael discuss the 1968 running mates in our first full episode. In a divided and tormented nation, Richard Nixon picks Spiro Agnew in an attempt to woo southern voters with a law and order campaign and Hubert Humphrey picks Edmund Muskie to appease anti-war activists, cutting off his chances in key states.
It may feel like the 2020 Democratic primary has been going on forever, but voters will finally have a say in who the eventual nominee will be in under a week at the Iowa caucuses. This first real contest between the candidates will officially begin the onslaught of state contests, delegate counts, and candidate dropouts that will all lead up to the eventual selection of the party’s nominee – either through the candidate security a majority of the delegates via the ballot box, or in a contested convention.
Every website, newspaper, and writer that covers politics will be tracking the delegate counts, the individual state contests, candidate dropouts, endorsements, the debates (don’t worry – there are so many more!), and the nuances therein in the march towards selecting the final nominee. Voting hasn’t even started yet, and there’s already so much available on the content on the candidates, be it through betting markets placing odds on who the nominee will be, poll aggregators, and sites tracking endorsements. We’ve decided to take a more niche approach to the primary and cover something that won’t happen for some time but should be in the back of everyone’s mind throughout this process: who the eventual nominee will choose as their vice presidential running mate.
We’ve covered the vice presidency a couple of times here at The Postrider, and even have a podcast dedicated solely to their electoral implications, but it’s an often overlooked subject and we’ve intentionally spent a lot of time focusing on it for that very reason. With the whole world’s eyes on the Democratic primary this spring, the vice presidential selection is a specific and unique enough lense to view this process through, and it will add additional context to who the Democrats choose as their ticket to represent themselves in the general election this fall. This year, we are covering it primarily through the 2020 Vice Presidential Tracker, an aggregator and tool we’ve spent several months developing and tuning. The tracker allows users to select a potential nominee and provides them with a breakdown of the most effective choices for that candidate’s running mate based on a measure we’ve called the “VP Score”. It also aggregates an overall standing rank, based on who is most likely to be the vice presidential nominee given who is most likely to be the presidential nominee. When the primary votes start coming in and delegates are allocated, this will become more and more rigorous and a more effective prediction given that certain candidates will separate from the field as more likely nominees.
One important caveat, this methodology has been built specifically for the Democratic Party’s VP selection. The same formula, aggregator, and tool will not work (by intention) for the Republican VP selection. The parties have very different voting blocks, backgrounds in past nominations, and preferences for national candidates – not to mention a different delegate allocation structure for the eventual nominee – and the VP tracker takes all of these things into account In the future, we hope to develop a tool that would predict the likely vice president in a competitive Republican Primary as well.
The tracker will update frequently throughout the primary process as more polls and then votes come in, so stay tuned to that page for updates and feel free to play around. If you’re interested in the methodology, keep reading!
Presidential candidates are included if they are actively running for president and have met one of the following two qualifications:
- They qualified for the seventh Democratic debate, so please welcome to the tracker: Joe Biden, Pete Buttigieg, Amy Klobuchar, Bernie Sanders, Tom Steyer, and Elizabeth Warren.
- They are polling at above 3% according to FiveThirtyEight’s national polling average. Good job Andrew Yang and Michael Bloomberg!
Potential vice presidential choices are included if… they are anyone you want them to be! So long as they are eligible to be vice president (they have not served more than six years as president, they are at least 35 years old, are a natural born American citizen, and must have lived in the United States for 14 years), they can be added to the tracker. I worked with the politically-minded The Postrider Editor-in-Chief Michael Lovito to put together a list of just over 50 names who we thought were particularly likely to be on a long list for any potential Democratic presidential nominee, and a couple of fun ones just to test the tool.1Sorry Jon Stewart and Howard Schultz, but at least we know the tool is making good picks. If you’d like us to add a name, shoot us a line at the “Contact Us” at the bottom of our About Us page.
On to the nitty gritty, here’s how an individual potential VP score is determined. It starts with the nominee, for whom the data is their:
- Race (this is a binary, a score of “0” if they are caucasian; a score of “1” if they are any other race including Latinx)
- Years of federal government experience (ex. as a congressman, senator, cabinet secretary, vice president, high level military command, etc.)
- Years of non-federal government experience (ex. as a governor, state representative, general military service, etc.)
Taking that data for a given nominee, it runs through every single potential VP choice, for whom the data is their:
- Their state’s number of electoral votes
- State margin, or the partisan lean of their state plus the generic ballot (positive number is leaning Republican, negative number is leaning Democrat; this is taken from FiveThirtyEight’s open source partisan lean data and is “the average difference between how a state or district votes and how the country votes overall, with 2016 presidential election results weighted 50 percent, 2012 presidential election results weighted 25 percent and results from elections for the state legislature weighted 25 percent.” If they update this for the 2018 midterms and so forth in the coming year we will do an update to the tracker. The generic ballot lean, which is to what degree the national headwinds are tilting towards the Democrats in general on the congressional ballot is then added to this, also courtesy of FiveThirtyEight – seriously, we love them so much. So, for example, if the state has a lean of -2 (2 points more Democratic) and the generic ballot has a lean of -5 (5 points towards the Democrats), then the state margin is -7.)
- Race (this is a binary, a score of “0” if they are caucasian; a score of “1” if they are any other race including Latinx)
- Reelection year (if you are currently in office, this is the next year you are up for election if you are not currently in office, or if your state allows you to run for two offices at once, this number will be “0”)
- Years of federal government experience (ex. as a congressman, senator, cabinet secretary, vice president, high level military command, etc. Note that if they are currently in office this will include all of 2020 as a year, so a representative first elected in 2012 that took office in 2013 and would presumably serve until January of 2021 will have 8 years)
- Years of non-federal government experience (ex. as a governor, state representative, general military service, etc.). Combining years of federal experience with years of non-federal experience results in total experience.
- Minimum viable office (this is a binary, a score of “0” if they do not meet the “minimum viable office” standard of more than 8 years as a member of the House, any service as a Senator, cabinet member, or military command, or as a governor, a score of “1” if they do)
With all the variables, it will run all of the possible vice presidential picks for a given nominee, and determine a “VP Score”, which is the strength of the ticket based on an equation that is easiest to describe by breaking it down into two key parts: the “coefficient” and the “base score”. The coefficient is the values multiplied together that are composed largely of “if this, then reduce the overall score; the base score is what makes up the increases in the score. The base score multiplied by the coefficient gives the final score.
The coefficient has five components all multiplied together:
VPMVO: This is the minimum viable office adjuster. If a VP choice meets the standard, it receives a value of “1” (because this leaves the core score unaffected, this will be the case with all of the variables in the coefficient). If they do not, this value is calculated as (1 + total years of experience) / (19 + total years of experience). This 1/19 base value was selected because out of the 19 vice presidential picks (excluding incumbent vice presidents) since 1968, only one (Geraldine Ferraro, Walter Mondale’s pick in 1984) did not meet this qualification. We use 1968 as the base year throughout this project because it represents the fundamental turning point in how American presidential primaries are carried out in modern times and it lets us start with no incumbent president running at the top of the ticket. Because total experience is added on to the numerator and the denominator, this number can reasonably increase if enough lower-level experience starts to add up, but always remains a detractor.
VPReelectYear: If a potential VP is not up for reelection in 2020, choosing not to run for reelection, or they can run for two offices at once, this has a value of “1”. However, if those do not apply and they’re up for reelection in 2020, then this is calculated as 10 to the power of one over negative the absolute value of the generic ballot score. So if the generic ballot is closer, it becomes far less wise to put the candidate on the ticket at the risk of losing their seat (keep in mind, if the generic ballot is closer, they’re also far less likely to win the presidential election as well). However, if the generic ballot is looking like a blowout, this number approaches 1.
VPState: This one seems redundant, but essentially it boils down to this: your vice presidential pick should not be from the same state you are. Pragmatically because the state of the pick does have a small impact, and electorally because electors from one state are forbidden from voting for two candidates from the same state on their ballot. You’d be forfeiting a given number of electoral votes, possibly ending up with a situation where a president or vice president were elected without the other.2If you love political hypotheticals – this almost happened in the year 2000, both George W. Bush and Dick Cheney were registered in Texas, but Cheney changed his registration to Wyoming before the election. The electoral vote was close enough where this would have deprived Cheney the vice presidency (at least initially) and required the Senate to choose a vice president. If you’re from different states, you get a value of “1” again; if you’re from the same state, then this number will be one over the electoral votes of the VP pick’s state. I do this because it still cuts the score by a lot (the minimum electoral votes a state can have is 3, which still brings it to a total score reduction of two-thirds) but not all states are created equal and technically, having two candidates from California (55 electoral votes) or Texas (38 electoral votes, see footnote above) would be immensely stupid, whereas having two candidates from Rhode Island would be… still pretty stupid, but hey maybe you really, really want those key four electoral votes from the Ocean State… for one of you.
VPKeySeat: Thank God for Ohio Senator Sherrod Brown. He and he alone is responsible for this variable. This attempts to address the strategy in deliberately not picking an otherwise near perfect vice presidential candidate. Brown is not up for reelection in 2020 (unlike Alabama Senator Doug Jones), he’s from a massive, somewhat Republican-leaning state that could go blue and seal the election for the Democrat, and he’s got lots of state and federal experience. So he seems like a perfect pick, right? Wrong! If Brown was on the ticket and Democrats won the 2020 presidential election, Democrats would soon regret their choice, because someone would then have to take Brown’s Senate seat and in an otherwise Republican leaning state, that person would likely be a Republican as well, which would give the Democrats one fewer seat in the all-important Senate.
So, thanks for providing us that niche example, Sherrod, and here’s what we can do about it: if a VP choice’s federal experience is greater than one, their reelect year is not “0” (this means that they are not running for reelection at all), and their state margin (remember this is the partisan lean plus the generic ballot) is between -7.5 and 7.5 (any race that is within this margin would make it at least somewhat competitive), then this is calculated as the square root of the absolute value of the VP state margin over 10. This allows a great pick to be exponentially weaker if their state’s margin is too close, but doesn’t hurt them too drastically if their state’s margin is safer, while still providing a slight disadvantage in picking them. If this situation does not apply, the score is, once again, a simple “1”.
VPRegion: Regional balancing is a thing most presidential campaigns strive for, so that they can claim to be representative of that nation as a whole instead of one particular region. For the sake of this project, we are using the US Bureau of Economic Analysis’ regions, since they are representative of the economic, political, and culturally similar areas of the country, but remain large enough to summate the nation in eight different slices.3These regions are by no means perfect. Alaska has very little in common with California culturally or politically; nor does Florida share a great abundance of economic similarity with West Virginia; but they do pretty effectively capture what the more modern understanding of “regions” seems to be, unlike the US Census Bureau, say, which considers Maryland and Delaware part of the South. Using the BEA regions, we found only two times since 1968 where the president and new vice presidential pick (incumbent tickets are excluded) are from the same region: Richard Nixon (NY) and Spiro Agnew (MD) in 1968, and Bill Clinton (AR) and Al Gore (TN) in 1992. Though I do note both of those campaigns did go on to win the general election, Nixon-Agnew lost a majority of the states in the Mideast region, and Clinton-Gore only got half of the states in the Southeast region. Conventional wisdom favors regional diversity, and thus if the regions are not shared, the value here will be “1”, but if they are shared it is reduced to “0.75” to reflect that a campaign could hope to appeal to at least two of eight regions (or 0.25) leaving at worst six of the other eight (0.75) regions off the table by default.
Multiply all of these together and you’ll get the coefficient half of the equation to determine the VP Score.
The Base Score
The base score has four key components all added together; they constitute the core score that is then reduced by the coefficient. In theory, a perfect candidate would have a base score of 100: 40.63% based on complementary experience (which is a major consideration in selecting a running mate), 26.04% based on the competitiveness and importance of the VP pick’s state (which can help, but doesn’t often make a major difference), 20% based on gender diversity, and 13.33% based on racial diversity. I’ll explain more why these are allocated as such in the individual descriptions.
VPRace and VPGender: If the nominee and the VP choice have the same race value (remember, this is binary), VPRace receives an input of 0. If they are different, it receives a score of 13.33. If the nominee and the VP choice have the same gender value, VPGender receives an input of 0. If they are different, it receives an input of 20.
I’ll be candid, the need for racial and gender balancing on the ticket is relatively new, so there’s not enough data to make much more than an educated guess as to how much it will matter. My inclination is that race matters about two-thirds as much as gender, because women make up a larger constituency of the Democratic Party (60%), and they obviously overlap with other demographics (50% of black voters – who already make up about 20% of Democratic voters – are also women). Because Latinx voters and Asian voters combined also make up about 20% of Democratic voters, and gender issues have been at the forefront of the Democrats’ agenda for the last few years, there will be immense pressure placed on the modern Democratic Party to have a diverse ticket, and I am inclined to agree that a ticket with a woman or person of color on it is more likely and stronger than one without.
However, to make it a bit more objective considering this has only been a consideration in recent years, I pulled from PredictIt’s (a betting site) odds for a Democratic presidential or vice presidential nominee to be a woman from the last 90 days (as of January 8th), and ran a regression using the odds of a woman being the presidential nominee as the independent variable and the odds of a woman being the vice presidential nominee as the dependent variable. With a 95% confidence interval, there was a significant negative relationship, with about 52% of the variance of the odds of a vice presidential nominee being a woman determined by the odds of the presidential nominee being a woman. The equation for the regression line was OddsVPWoman = -0.2037 * OddsNomineeWoman + 0.6863. This indicates that common wisdom assumes about a 69% chance of a woman being the vice presidential nominee if the presidential nominee is not a woman, but for our purposes we’re actually going to use the regression coefficient of -0.2037, which gives us the change given if the nominee is a woman (x = 1) or if the nominee is not a woman (x = 0), giving us a roughly -20% decrease between the two.
We use this 20% as the basis for how much weight we give the requirement of one woman on the ticket because it is the determinant of change between whether there is or is not a woman on the ticket from the regression. An imperfect measure but the best option considering the relative recency of this consideration for the Democratic ticket. As noted earlier, we’ve weighted race by two-thirds as much as gender if for no other reason than there is no predictive data on this, and because women make up 60% of the Democratic electorate and racial minorities make up 40% (there is obviously overlap), so race receives a weight of 13.33.
VPXP: Arguably the most complicated and most important measure, this can make up up to 40.63% of the base score. That’s an oddly specific number, you may note, and that’s because we know that complementary experience and the state margin must add up to two-thirds of the total score; as gender and race account for a third of the score. I allocated these two by looking at every unique vice presidential pick from 1968 and the combined federal experience with the presidential pick (in years) and how many times the VP pick was from a competitive state.4We’re using the same standard as this NPR article, which describes a swing state as any state that had a margin of <5% between the two candidates in any of the four elections chronologically closest (before or after) a particular election year, as well as the election itself. Mike Pence (Indiana), for example, is not a swing state pick, even though there was one election (2008) in which it was won by a margin of less than five percent. The VP pick was from a competitive state nine times out of 19 unique picks (47.37% of the time); and the combined federal experience averaged 25.1053. I ran descriptive statistics on the combined federal experience in years to find a standard deviation of 12.2061, and went one standard deviation on either side of the mean to find an acceptable level of combined federal experience (that would, under the empirical rule, capture about 68% of all presidential tickets, which determined that 14 out of 19 presidential tickets had a combined federal experience within one standard deviation from the mean. This implies that experience is about 1.56 times more important in picking a VP than picking one from a competitive state. Since these both must fit into the remaining 66.67% of the base score, this requires VPXP to be weighted at 40.63% of the base score, and VPStateMargin to be weighted by 26.04% of the base score.
VPXP is calculated in two parts, using the presidential and vice presidential candidates’ combined federal experience and their combined non-federal experience. Recall the average combined federal experience is 25.1053 years, and the standard deviation for that was 12.2061; meaning that under the empirical rule, 68% of values must lie within one standard deviation of the mean. The reason it is not just simply one standard deviation below the mean (and continuing to increase the score the more combined years the ticket has) is because there appears to be diminishing marginal returns in having a ticket with too much experience, otherwise most presidential tickets would have combined federal experiences well into the 40s or 50s. The experience, instead, more closely follows a normal distribution, where there is an increasing benefit up to a point, and then a decreasing benefit. This is also a good proxy for age, as two candidates on the ticket with 40 years of federal experience each would constitute an unusually elderly ticket, something that has received a lot of attention in this election. The formula for determining the experience score is composed of two parts: combined federal experience and combined non-federal experience. Federal experience is weighted twice as much as non-federal experience (meaning it is weighted by 27.0867 compared to non-federal experience’s weight of 13.5433, so that a perfect score in both would result in the 40.63 that the total VPXP value is worth in the base score. The formula for both comes out to:
Where e is the mathematical constant multiplied by 27.0867 in the first half for the federal experience weight; and by 13.5433 for the non-federal experience weight, and federal experience is taken to the power of the candidate’s federal experience (fxp) plus vice presidential candidate’s federal experience (vpfxp) minus the mean squared, over the standard deviation times the sum of the candidate’s federal experience and the vice presidential candidate’s federal experience, all multiplied by negative one. Non-federal experience is to the power of the candidate’s non-federal experience (sxp) plus vice presidential candidate’s non-federal experience (vpsxp) minus half the federal mean (as we’re weighing it by half) squared, over half the federal standard deviation (once again, because we’re weighing it by half) times the sum of the candidate’s non-federal experience and the vice presidential candidate’s non-federal experience, all multiplied by negative one.
VPStateMargin: Last but not least is the VPStateMargin, weighted by 26.04% of the total base score.5The reasons for are discussed above in VPXP, but it boils down to complementary experience being valued by 1.56 times opting to pick a vice presidential candidate from a competitive state. If a VP candidate’s state margin (this accounts for the generic ballot) is between -7.5 and 7.5, then the VPStateMargin is the square root of the vice presidential candidate’s state’s number of electoral votes over one plus the absolute value of their state margin. This is multiplied by the most competitive average electoral vote adjuster, which was determined by taking the total number of electoral votes (538) over the total number of states (51 including DC which has three electoral votes), which is 10.55, over a perfectly competitive state (assuming a state margin of 0), which is a value of 1. The square root of which is 3.2481, which must be multiplied by 8.017 to equal the weight for this factor of 26.04. So 8.017 is the multiplier. This score gets higher if the state they are from has a large electoral vote count and has a very narrow margin. Yes, this technically does let it exceed far over 26.04 in cases where a state with a lot of electoral votes is going to be extraordinarily close, and indeed, if California (with 55 electoral votes) had no partisan lean, picking a vice presidential candidate from California would be a very smart thing to do, so this can give this factor a bit of a step up if something is a no brainer.
If the VPStateMargin is not between -7.5 and 7.5, then the state is not considered competitive for the sake of this value, and it is simply calculated as the VP state’s electoral votes over one plus the absolute value of the state margin. In reality, this is how California falls, with 55 electoral votes over 1 + 30 (California is a D +24 state, and the generic ballot at this time shows a national lean of 6 points Demoratic). Picking a VP from California would increase your score by a tiny 1.77 points.
If you add these four factors together, you’ll have the base score. Multiply the coefficient by the base score to get the VP Score for a given nominee and VP nominee combination. If you choose a nominee it will run this function for every possible VP nominee:
The other side of the tool has a ranker, which is a judgement of how strong a particular vice presidential nominee selection is, based on who the most likely presidential nominee is. What this does is run the VP score for every possible vice presidential pick for every potential presidential nominee and weights them by the odds of a given presidential candidate winning the nomination; then adds them up so they’re reflective of their overall odds to win a majority of delegates in the primary. The presidential candidate’s odds are pulled from FiveThirtyEight’s Democratic primary odds, so that they’ll adjust over time as candidates start to win primaries, and will be updated once after each day a primary occurs. You may note that FiveThirtyEight only has odds for Joe Biden, Bernie Sanders, Elizabeth Warren, Pete Buttigieg, “No One”, and “All Others”. For the candidates we have on our tracker that are not on FiveThirtyEight’s tracker, currently just Amy Klobuchar and Andrew Yang, we are giving them each the odds given to “All Others” (so if “All Others” are at 2%, all three of them in our tracker receive 2%). This slightly inflates the odds of their VPs being picked compared to the odds of the others, considering their low probability, but when combined with the relatively high odds of “No One” winning a majority and a brokered convention, we’re willing to add a bit of additional weight to some far-fetched odds.
So, if for example Biden has a 40% chance of winning the nomination, Sanders has a 20% chance, Warren has a 10% chance, and so on, the VP scores for all possible VPs for Biden will be 40% of their original value, VP scores for Sanders’ VPs will be 20% of their original value, VP scores for Warren’s VP’s will be 10% of their original value, and so on. The ranking here is not in and of itself a probability, it’s more of a “power ranking” of the total potential vice presidents’ scores weighted by the actual odds of a given presidential nominee.
We’re launching a new podcast here at The Postrider as part of our Running Mates project dedicated to the American vice presidency, those who have held it, vied for it, and perhaps are destined for it. The Running Mates podcast will cover every modern presidential election starting with 1968 through the lens of the vice presidential candidates, as Michael and I discuss whether the vice presidential picks made sense, could have been stronger in hindsight, and who might have been a better choice.
Our introductory episode is below to kick us off, so please subscribe and stay tuned for our first episode on the Edmund Muskie versus Spiro Agnew vice presidential election of 1968.
2020 is shaping up to be a big year for the two things I love writing most about at The Postrider: the United States Senate, and the vice presidency. And we’re going to have a swath of content featuring both of these items unveiled over the course of the coming election year. We’ve held off a little bit this year to give you some sense of relief from the seemingly endless campaign, so it might shock you to know that we’re jumping the gun to start looking – in broad terms – at who the Democratic vice presidential nominee might be.
There’s scant evidence to suggest that a vice presidential running mate makes a difference in the general election. They may provide a slim bump in the polls if they’re from a certain state, but notable “swing state” picks like Paul Ryan (Wisconsin), John Edwards (North Carolina), Lloyd Bentsten (Texas), and Geraldine Ferraro (New York) all lost their home state; in fact, Tim Kaine (Virginia) and Al Gore (Tennessee) are the only two running mates from swing states since 1976 to have won their home states.1We’re using the same standard as this NPR article, which describes a swing state as any state that had a margin of <5% between the two candidates in any of the four elections chronologically closest (before or after) a particular election year, as well as the election itself. Mike Pence (Indiana), for example, is not a swing state pick, even though there was one election (2008) in which it was won by a margin of less than five percent. Yes, I too find it humorous to think of a time that Tennessee, Texas, and New York were all once swing states, but rejoice in the fluidity of the American electoral system. Picking someone from a different region and with different political experience than the presidential nominee seems to be a consistent choice by campaigns, though it’s unclear whether or not this actually matters come election day. Nonetheless, for those following campaigns and politics at large, considering potential running mates is high entertainment.
We’ll be unveiling a VP tracker in the next few months to put metrics and actual hard evidence into our analysis of who is likely to be a VP pick for each potential presidential nominee. Over the course of background research for that tracker, I had to face a key question: have we already met the VP candidate? By this I mean, did they run – or are they still running – for president? The Democratic field has had 27 “major” candidates,2We’re relying on the New York Times definition of “major”. with a possible 28th soon to come, a massive number that’s unprecedented in modern presidential campaigns. Many of these names came and went, despite having held prominent offices, and it begs the question if we’ll see one of them eventually resurface as the running mate of an eventual candidate Warren, Biden, Sanders or whoever you believe will grace the DNC stage that fateful night in July.
The easiest way to do this was just to look back at the historical record. I looked at every primary since 1976 that had no incumbent president on the ticket, compared the total number of candidates (this time using FiveThirtyEight’s metric for “noteworthy” candidate, which errs towards over-inclusivity), and looked at how many races featured a running mate who also ran in the primary:
|Year – Party||Nominee||Vice Presidential Nominee||Number of Candidates||VP Ran in Primary?|
|2016 – D||Hillary Clinton||Tim Kaine||5||No|
|2016 – R||Donald Trump||Mike Pence||17||No|
|2012 – R||Mitt Romney||Paul Ryan||12||No|
|2008 – D||Barack Obama||Joe Biden||10||Yes|
|2008 – R||John McCain||Sarah Palin||12||No|
|2004 – D||John Kerry||John Edwards||10||Yes|
|2000 – D||Al Gore||Joe Lieberman||2||No|
|2000 – R||George W. Bush||Dick Cheney||12||No|
|1996 – R||Bob Dole||Jack Kemp||12||No|
|1992 – D||Bill Clinton||Al Gore||8||No|
|1988 – D||Michael Dukakis||Lloyd Bentsen||11||No|
|1988 – R||George H. W. Bush||Dan Quayle||7||No|
|1984 – D||Walter Mondale||Geraldine Ferraro||8||No|
|1980 – R||Ronald Reagan||George H.W. Bush||9||Yes|
|1976 – D||Jimmy Carter||Walter Mondale||16||No|
This is a very small dataset, not enough to do any real statistical analysis, but there are a couple of takeaways that illustrate what to expect of the soon-to-be-nominee picking a running mate in mid-2020:
Choosing a “has-ran” is not as common as you’d think
Considering the media attention you earn by launching a presidential run and the intense focus on the dynamics between candidates, debates, and policies throughout the primary, it’s actually surprisingly rare for the running mate to be chosen from amongst those who ran in the primary. It’s only happened in three of these 15 primaries since 1976. The reason for this is likely tied to political baggage. It’s probably easier to start fresh once you’re the nominee and reset the campaign cycle entirely with someone who has not had attack ad after attack ad run against them. You don’t want the contrasts and contradictions that emerged during the primary between you and your running mate to be focused on or unpacked; projecting unity and a compelling ticket needs to be the emphasis.
It appears to peak in a mid-sized primary
Primaries that are smaller than the average (10.07 candidates) and primaries that are larger than the average do not historically lead to an also-ran candidate being selected as a running mate. That makes sense for primaries that are small, because it means there are fewer candidates who could potentially be considered, but is interesting for famously large fields (like the 1976 Democratic primary or the 2012 or 2016 Republican primaries). In fact it only appears in this sample that those at almost exactly the average (around nine to eleven candidates) saw a running mate selected from the primary field.
Initially I thought this might have something to do with the competitiveness of the primary, how those with more than eleven candidates appear to be less competitive, and those with a small slate would also be less competitive, but the data was inconsistent. The average “competitiveness” (which I took in a loose measure to be the distance between the percentage of the vote that the most-vote-receiving candidate and the second-most-vote-receiving candidate received) was the winning candidate ahead by over 27%. However, both the 2004 Democratic primary and 1980 Republican primary (both in which a running mate was chosen from the other candidates in the primary) were 42% and 36% respectively. Only the 2008 primary, in which Barack Obama chose Joe Biden as his running mate, was the competitiveness extremely high (actually negative 0.7%, as Hillary Clinton received more votes overall). Of all the primaries where a running mate had not also run, they were across the board in terms of levels of competitiveness, ranging from the 2000 Democratic primary, where Al Gore was ahead of second-place Bill Bradley by more than 54% but chose Joe Lieberman (who did not run in the primary) to the 1984 Democratic primary where Walter Mondale beat out Gary Hart by just 2.4% but selected Geraldine Ferraro (who also did not run in the primary).
Competitiveness does not seem to be the determining factor, the reason it has only happened in mid-sized fields is probably a statistical fluke. As noted, there are not enough samples to justify making a conclusion.
The current field may be too big to warrant historical comparison
This being said, the current Democratic primary field has had 26 candidates so far, across a fairly wide spectrum of positions, ideologies, and demographics. The argument for things being a bit different this time is that at some point you’re running out of great picks elsewhere if every up and coming politician is already in the field. Granted, 27 candidates is not that point; I would estimate that the “long list” for any potential Democratic nominee would include around 50 candidates, and that there’s probably a fairly large amount of overlap in those picks between the candidates still in the running for the nomination. There are plenty of Democratic senators, representatives, governors, military leaders, former cabinet secretaries, and other notable party figures or celebrities that are likely to be considered. But the truth is that of those 50 or so, roughly 20 of them have probably already jumped in the presidential race.
If I were to venture a guess, I would still surmise that the eventual Democratic nominee will choose someone who was not in the Democratic primary; as that historically happens about 80% of the time, and this primary has already had some vitriolic moments between the candidates. Even those candidates who we might assume to be angling for the VP slot on a future ticket if they do not win the nomination have gone out of their way to attack the frontrunners. Picking someone without a history of attacking you on the campaign trail and with a solid complementary record seems to be the right move in such a crowded year.