UPDATED: April 20, 2018 at 4:32 p.m.

This article is part one of The Crimson’s five-part series analyzing exit polling data collected during Harvard’s 2018 unionization election. Read part two here, part three here, and part four here.

The results of the election that will decide whether Harvard teaching and research assistants unionize are too close to call, according to exit polling data collected by The Crimson.

Exit poll results adjusted for response bias suggest a slight majority—50.6 percent—of eligible students who cast ballots voted in favor of unionization. But the margin of error—plus or minus 2 percent—means The Crimson cannot be certain the election will result in unionization. The Crimson used a 95 percent confidence interval, the standard for election polling, to calculate the margin of error.

The final result will likely be decided Friday. National Labor Relations Board officials are set to tally the ballots in the NLRB’s regional office in downtown Boston starting 9:30 a.m. Friday.

The unionization election took place over the course of two days, April 18 and 19, at polling sites on segments of Harvard’s campus in Cambridge, Longwood, and Allston. At stake is whether more than 4,000 eligible graduate and undergraduate students will begin collectively bargaining with the University as members of Harvard Graduate Students Union-United Automobile Workers.

If the final results fall in favor of unionization, Harvard will join a small handful of private American university to see a union form without voluntarily granting the union recognition. Columbia University was the first.

The Crimson conducted exit polling of voters at all three voting stations on April 18 and 19, collecting 1,295 responses. A total of 5,046 Harvard affiliates were eligible to vote; it is not yet clear how many eligible voters chose to cast ballots.

Crimson reporters stationed outside voting sites handed voters 23-question surveys asking the students how they voted, various demographic questions, and their views on salient campus issues. At least one reporter stood outside every single site for 78 percent of all hours the sites remained open. The surveys were later electronically processed using open-source software.

Of survey respondents who answered a question asking whether they voted in Harvard’s previous unionization election, held Nov. 2016, 50.4 percent answered yes while 49.6 percent answered no. New voters were 17.4 percent more likely to vote in favor of unionization.

Of respondents who indicated they voted in the last election, 11.7 percent reported changing their minds since Nov. 2016. Of those who indicated they changed their minds, 78.8 percent said they voted in favor of unionization this time around.

The 2016 vote saw a final tally of 1,526 votes cast against against unionization and 1,396 cast in favor. Lawyers for HGSU-UAW challenged the results of that first election, ultimately sparking more than a year of legal battles between Harvard and the would-be union. The NLRB ultimately mandated in Jan. 2018 that the University must hold a second election.

[The Crimson maintained an online article constantly updated with voter interviews throughout the 2018 unionization election. Look back on the two historic days here.]

Of survey respondents who reported their gender, women were 14 percent more likely than men to vote in favor of unionization. Of survey respondents who reported their international status, international students were 9 percent less likely to vote in favor of unionization.

Harvard affiliates pass in and out of a voting station located at Queen's Head Pub on the first day of voting in the University's second unionization election. By Amy Y. Li

Out of all respondents who indicated how they voted, 68.9 percent voted in favor of unionization while 31.1 percent voted against. The Crimson adjusted this result for response bias (see below for methodology), leading to the conclusion a small majority of eligible students may have voted for the union.

Jae Hyeon Lee—a member of Against HGSU-UAW, a Facebook group for anti-union students—wrote in an email that, as someone who “has been vocally against unionization,” he is glad the election cycle has come to an end, regardless of result.

“Looking back, my engagement in this issue has been the most emotionally tiring and contentious chapter of my graduate career,” Jae wrote. “At times it was quite disheartening to see how divisive this issue has been among students.”

“Whatever the result may be, going forward I hope students from both sides can work together towards the common ideal of improving the student experience in a more dispassionate and positive manner,” he added.

HGSU-UAW organizer Andrew B. Donnelly said in an interview Thursday night he thinks the election ran smoothly overall.

“I’m proud of the positive, issues-based campaign we’ve run,” he said. “I’m excited for the count tomorrow.”

METHODOLOGY

After reviewing raw exit poll data, The Crimson identified response bias suggesting that, overall, voters in favor of unionization were more likely to fill out the survey than were voters against unionization, leading to a 32 percent oversampling of pro-union votes.

The Crimson measured this bias by benchmarking data collected April 18 and 19 against results from the previous Nov. 2016 election. This analysis revealed that—of respondents who indicated they voted in both elections—the percentage of respondents who indicated they voted yes in both elections was higher than the actual percentage of voters who voted yes to unionization in 2016.

The Crimson corrected for this apparent pro-unionization skewing by dividing the percentage of respondents who indicated they voted for unionization in 2018 by the increased likelihood that a respondent who favors unionization would fill out the survey.

The Crimson also corrected for possible geographic biases, given the NLRB assigned each eligible voter in the 2018 election to one of the three sites before polling began. This correction was necessary because it is possible the geographic distribution of survey respondents did not match the geographic distribution of actual voters.

Of total survey respondents, 80.6 percent voted in Cambridge, 18.1 percent voted in Longwood, and 1.3 percent voted in Allston. Of total eligible voters, approximately 3,644 voters—or 72.9 percent—were assigned to the Cambridge site, 1,293—or 25.9 percent—were assigned to the Longwood site, and 59—or 1.2 percent—were assigned to the Allston site.

The Crimson applied a geographic bias correction by taking the recorded vote breakdown from each polling site and applying a weight proportional to the total eligible population assigned to that site. In making this calculation, The Crimson assumed there was no statistically significant difference in voter turnout between the three different sites.

Voters could choose to vote in a site other than their NLRB-assigned location—but if a voter chose to do so, that voter had to cast their ballot “under challenge,” meaning the NLRB must review and confirm that individual’s eligibility prior to counting their vote.

The NLRB separated each day’s vote into two segments, keeping polls open from 10:00 a.m. to 2:30 p.m. and from 4:30 p.m. to 8:00 p.m. A Crimson analysis found no statistically significant difference in whether voters were more likely to vote for or against unionization between the two polling periods.

Staff writers Luke W. Vrotsos and Anna Kuritzkes contributed reporting. Contributing writers Paul D. Tamburro, Karina G. Gonzalez-Espinoza, Lainey A. Newman, Aidan B. Carey, and Meena Venkataramanan also contributed reporting to this article.

Crimson editors Brian P. Yu and Phelan Yu conducted data analysis for this story. Questions regarding the survey or methodology can be directed to brian.yu@thecrimson.com and phelan.yu@thecrimson.com.

This article has been updated to reflect the following correction:

CORRECTION: April 20, 2018

A previous version of this article incorrectly indicated that Harvard would be the second-ever private American university to see a union form without voluntarily granting the union recognition. In fact, Harvard would join a small handful of private universities who have seen this happen.

—Staff writer Shera S. Avi-Yonah can be reached at shera.avi-yonah@thecrimson.com. Follow her on Twitter at @saviyonah.

—Staff writer Caroline S. Engelmayer can be reached at caroline.engelmayer@thecrimson.com. Follow her on Twitter @cengelmayer13.

—Staff writer Molly C. McCafferty can be reached at molly.mccafferty@thecrimson.com. Follow her on Twitter at @mollmccaff.