“Most people are interested [in elections] like sports fans. You want to know who’s gonna win, and you’re rooting for your side,” explained Dr. Sam Wang, a professor at the Princeton Neuroscience Institute, from his office littered with books and notepads. “My interests are always different … Informed voters have more leverage than uninformed voters, and it’s my belief that the use of data will help them identify where they have leverage,” Wang continued.
Four years before baseball statistician Nate Silver rose to fame for predicting 2008 presidential election outcomes using his poll aggregator FiveThirtyEight, Wang’s Princeton Election Consortium (PEC) was among the first wave of models to use statistics and probability to predict close election outcomes.
“[Elections] are a game of probabilities,” Wang said. “I thought, ‘I have math abilities. Why don’t I help out with that?’”
Wang is a jack of all trades, master … of them all. At Princeton, he has affiliate appointments in the School of Public and International Affairs’ Program in Law and Public Affairs, the Center for Cognitive Science, the Program in Quantitative and Computational Biology, and the Center for Information Technology Policy.
The PEC is one of several projects Wang has created focused on what he calls “democracy repair.” In 2014, Wang launched the Princeton Gerrymandering Project, an effort to apply science to make redistricting more fair. PGP’s mission then expanded to include other problems in democracy, culminating in a multi-institution nonprofit, the Electoral Innovation Lab (EIL), housed in the municipality of Princeton. Since 2016, Wang has collaborated with History professor Julian Zelizer on a podcast, Politics and Polls. This summer, Wang and the EIL launched the VoteMaximizer, which calculates the power of individuals’ votes in elections all over the country, from the local to national level. All the while, he runs his own lab at PNI, which “investigates how brains learn from sensory experience, in adulthood and development, with relevance for autism.”
“My general idea is to apply science, data science, and statistics to really understand our complicated government and figure out where citizens have the most leverage,” Wang told the Daily Princetonian.
PEC v. FiveThirtyEight
Wang's foray into politics started during a short stint on Capitol Hill. After completing his Ph.D. in neuroscience at Stanford, Wang worked as a postdoctoral fellow for the Senate Committee on Labor and Human Resources.
“I got really interested in that year’s presidential election, Kerry versus Bush. Ever since 2000, we’ve had a string of really close elections, with five of the last six decided with a popular vote of five points or less,” Wang said. “When elections are this close, just a few votes matter, so I thought it would be interesting to aggregate all the data that was available at polling that year, to figure out where voters had the most leverage.”
Wang distinguishes the PEC from other aggregators like FiveThirtyEight, both in intentions and in the math used in the model for poll aggregation itself, creating a workplace rivalry between the two groups. In 2014, Silver tweeted about Wang’s prediction about the Senate race, “I would like to place a large wager against that guy,” to which Wang responded in his articles breaking down his poll analysis on the PEC.
Lucas Manning ’20, who worked on the PEC with Wang while an undergrad, noted that any rivalry or “quips” between the two aggregators are completely academic in nature and stem from different approaches to poll aggregation.
“Sam was more about looking at the polls and just aggregating them, and Silver’s model incorporates a lot of other factors, like inflation in the economy, that have shown to have an effect on the outcome of elections,” Manning said.
While Wang separates the PEC from counterpart FiveThirtyEight’s goal to drive “horse race coverage,” he also acknowledges the role that FiveThirtyEight has played in popularizing poll aggregation, including the PEC. In 2004, the PEC’s analysis of state polls got tens of thousands of hits per day, and since then, the PEC has recorded over five million visits. The PEC’s analysis has been featured on NPR and the WSJ, and in past elections, the New York Times’ Upshot has carried PEC’s work as part of presidential elections coverage.
As of Thursday, Sept. 26, 2024, the Princeton Election Consortium predicted a winning 279 electoral votes for Harris. The PEC also shows Republicans are currently favored to flip the Senate with 51 seats to the Democrats’ 49 seats. Democrats also currently lead Republicans by two percent in the House race.
From horserace to maximizing voter power
In 2012, the PEC correctly predicted the presidential vote outcome in 49 out of 50 states and Obama’s 51.1 percent winning popular vote. The PEC also correctly predicted 10 out of 10 close Senate races. In 2020, the PEC correctly predicted a Biden victory, but overestimated the winning number of electoral votes, both based on then-current state polling and if Trump outperformed the polling margin of three percent.
However, in 2016, the PEC incorrectly predicted a 93 percent chance of a Clinton victory. Wang paid the price, eating a cricket live on CNN after losing a bet that Trump wouldn’t win more than 240 electoral votes.
Manning explained that the PEC’s popularity slightly dipped after the 2016 election.
“In the 2016 news cycle, the PEC was considered one of the major aggregation sites by the New York Times, but I think the site’s reputation was damaged by the 2016 election,” Manning noted. “Compared to FiveThirtyEight’s prediction of a 79 percent chance of a Clinton victory, the odds of a one in three event are much easier to reason than a one in 100 event, so for those reasons, traffic in 2018 and 2020 was definitely smaller.”
Mark Tengi ’16, a computer science undergraduate who helped Wang maintain code for the PEC website during the 2016 election, explained that the PEC’s model of relying solely on the state polls contributed to an inaccurate prediction in 2016 because of a four percent error in the polls themselves. The PEC’s model differs from the methods of other aggregators like FiveThirtyEight, which included Silver’s own electoral projections based on demographics and prior voting patterns, creating a 79 percent chance of a Clinton victory.
“PEC turns polling data into win probabilities, but if that underlying polling data is inaccurate, then the whole prediction is inaccurate,” Tengi summarized.
Despite the post-2016 dip in popularity, Manning observed that this outcome pushed the PEC in a positive direction.
“I think the mission [of maximizing voter leverage] was driven largely by the outcome of the 2016 election — people were getting way too into the horserace of following elections and not really focusing on making a difference,” Manning explained. “[The model] became focused on helping people have more of an impact.”
Wang calls the principle of using math to identify how voters and donors can have the maximum possible leverage the “Democracy Moneyball” principle, a reference to Michael Lewis’s book “Moneyball: The Art of Winning an Unfair Game,” and playing on the public’s sports-like stake in competitive elections.
The “Meta-Analysis,” what Wang calls the math he uses for the Princeton Election Consortium, boils down to normal distribution as a measure for voter power.
Aggregating state polls, he takes an estimate of voter turnout for any given state, then estimates how likely it is that moving a few votes is going to make a difference by calculating the standard deviation, or the bell shaped curve. Tied races sit at the top of the curve, while those that are not tied are at the tails. Using this math, Wang has developed a single formula for per-vote leverage. Additionally, he includes a “meta-margin,” a quantity that corresponds to a virtual lead in a race.
Beyond the Princeton Election Consortium
Applying this same meta-analysis and “democracy moneyball” principle, Wang and others at the Electoral Innovation Lab have created the VoteMaximizer, a tool that identifies races and ballot initiatives from the local to the state level to maximize voters’ power and contributions. The tool identifies states and districts with high per-voter leverage. For instance, Nevada has a voter power of 98 because it is a swing state in presidential elections, has a close Senate race, and has close Congressional races.
“Voters in swing states have higher turnout than other states, and so it’s pretty clear that at the level of the presidential race, people pay attention to where they have more leverage,” Wang explained. “What we’re doing is calculating where that leverage is present at every level.”
Part of the work of the VoteMaximizer is to dismantle a common misconception among voters that votes only matter in presidential swing states. While doing data collection for the VoteMaximizer this summer, Lia Opperman ’25, found ways every voter, even Princeton students, can best leverage their voter power and contributions.
Opperman is the Director of Outreach for the ‘Prince.’
“There are contested elections in basically every state and ballot initiatives in most states,” Opperman said. “There is power that you have, an impact that you can make in your own area.”
In New Jersey’s Congressional race for District 7, the voter power is 65, where Democrat-hopeful Sue Altman aspires to oust incumbent Rep. Tom Kean (R-N.J.).
For the last year, Wang has also been collaborating with Professor of Ecology and Evolutionary Biology Simon Levin on a series of research papers about combating polarization through alternative voting systems, like ranked choice voting. Currently, Levin and Wang are working on a paper analyzing voting patterns to identify the degree to which voters are displayed along a single, one dimensional left-right axis, and the polarization along that axis over time.
Using statistical models to depict this polarization, Levin and Wang are examining how these voters on a left-right spectrum would have voted in alternative voting schemes like ranked choice voting and four primary voting.
“Our role is more to lay out for people what outcomes different voting schemes would have produced, and leave it to people to decide the outcome they prefer,” Levin said.
Levin and Wang both view the U.S. electoral college system elections as flawed.
“There are [currently] only a small number of states in which people’s votes matter … not just at the presidential level, but when you’re electing congresspeople as well,” Levin said. “That doesn’t seem fair to me.”
Part of the duo’s research is not only to uncover the value in alternative voting systems, but how they could feasibly be implemented through reform of local election procedures. Wang argues that voting for a third party candidate in a highly polarized election, like national elections, is an unproductive way to achieve this reform.
“I would love to have a roaring fire in my office. It would be lovely. But if I tried to do that without building a fireplace, then what I have is a burning office. The fire is the third party, and the fireplace is having, you know, rank choice voting,” Wang said. “And so the difficulty is that even though reform is boring, it’s how you get to having a better system.”
Levin echoes Wang’s sentiment that voting reform must be built from the ground up.
“I would say that both Sam and I are very favorable towards some form of ranked choice voting,” Levin said. “In dealing with environmental problems, we always say something like, ‘Think global, act local,’ and that’s what we have to do here: decide what’s the best system and then try and get it instituted in as many places as possible built up from the local level.”
While Wang and Levin work to find innovative solutions for democracy repair, Wang emphasizes the untapped power voters already have within the existing voting system.
“If voters knew how much power they had, it would increase their engagement,” Wang said. “What I’m doing is identifying where all those places are, not only for empowering them this year, but also making democracy stronger for years to come.”
Valentina Moreno is an assistant Features editor for the ‘Prince.’
Read More
‘The heartbeat of our program’: Inside the walk-on experience
Diego Uribe and Tate HutchinsWalk-on athletes play a significant role across the 38 Princeton varsity athletic programs, filling in gaps and adding depth that coaches often cannot fill due to pressure to limit numbers of recruited athletes. Walk-ons come in all shapes and sizes — some communicating with coaches while in high school and some choosing a new sport after coming to campus — but their presence is highly valued by coaches and teammates alike regardless of background.
Walk-on athletes play a significant role across the 38 Princeton varsity athletic programs, filling in gaps and adding depth that coaches often cannot fill due to pressure to limit numbers of recruited athletes. Walk-ons come in all shapes and sizes — some communicating with coaches while in high school and some choosing a new sport after coming to campus — but their presence is highly valued by coaches and teammates alike regardless of background.
Resources Committee considers framework for community feedback, ‘consensus’ on Israeli divestment
Elisabeth StewartTwo months after the Princeton Israeli Apartheid Divest (PIAD) petition submitted its divestment proposal, the CPUC Resources Committee continues to establish a process to consider the proposal and gather community input. Meanwhile, a group of Jewish students and faculty came together to write a petition calling for “NO Consensus on BDS,” which was also submitted to CPUC.
Two months after the Princeton Israeli Apartheid Divest (PIAD) petition submitted its divestment proposal, the CPUC Resources Committee continues to establish a process to consider the proposal and gather community input. Meanwhile, a group of Jewish students and faculty came together to write a petition calling for “NO Consensus on BDS,” which was also submitted to CPUC.
After failed vote in May, graduate students reflect on years-long unionization campaign
Maya Mukherjee and Meghana VeldhuisPrinceton Graduate Students United (PGSU) filed on April 12 and held its vote on May 13 and 14, with just over a third of graduate student voters voting in support of the union. The ‘Prince’ spoke to organizers and critics of unionization about the campaign preceding the election.
Princeton Graduate Students United (PGSU) filed on April 12 and held its vote on May 13 and 14, with just over a third of graduate student voters voting in support of the union. The ‘Prince’ spoke to organizers and critics of unionization about the campaign preceding the election.
