Calculating a win number is one of the first analyses any data team should provide when working with a campaign. Simply put, a win number is the estimated number of votes a candidate needs to win the election. All other analyses, strategies, and tactics should revolve around helping a candidate reach this number of votes.
In primary and general elections, the standard methodology for this calculation is pretty straightforward. NGP VAN, one of the primary technology providers for the Democratic Party, recommends that campaigns calculate their win number using the last three similar elections and the formula below:
While this calculation is based on a lot of assumptions, namely that the last three similar elections will be a good predictor of the number of ballots cast in the upcoming election, it also provides a straightforward framework for campaigns with limited data expertise to set ballpark goals for their voter outreach programs. But what if there haven’t been three similar elections previously, or even any?
This is the predicament my team and I found ourselves in while volunteering for the Dr. Eliz Markowitz campaign for Texas House District 28. After Rep. John Zerwas (R) resigned in September of 2019, a special election was called to fill the vacant seat. We were asked to help the campaign calculate the win number, but there hadn’t been a special election in House District 28 in at least the last decade.
Luckily for us, we did have some data to go off of. As of September of 2019, there had been three other special elections that year across the state of Texas in House Districts 79, 125, and 145. Furthermore, there had been contested elections in these districts over the last three election cycles that we could compare to the special election results.
To estimate the win number for the special election in House District 28, we formed the following hypothesis: the ratio of turnout in the special election to turnout in the past three elections will be the same in House District 28 as it was in the three other House Districts in Texas that have had special elections in 2019.
Because only four out of the last twelve general elections in these House Districts were contested, we decided to base our final estimate on turnout using the primary election data (only one election was uncontested).
Next, we did the math.
And showed our work...
Everything seemed to check out. Our math was sound. Our assumptions were reasonable, but we knew we wouldn’t know whether they were correct until the election. When we crunched the numbers, we estimated that 11,371 votes would be cast in the Texas House District 28 special election. Therefore, we estimated Dr. Eliz’s win number to be 5,913 votes.
However, it was what we didn’t factor into our calculations that proved our estimate to be far from accurate. With the Texas House of Representatives responsible for congressional redistricting following the 2020 US Census, and this particular seat being identified as one of the top flippable seats by the Texas Democratic Party, this race got national attention.
Beto O’Rourke held a rally for Dr. Eliz in Katy, she raised over half a million dollars, and she had national support from the Democratic Party. All of these factors ended up driving turnout far past what we had anticipated. In the end, 29,079 voters cast their ballot in the Texas House District 28 special election yielding a true win-number of 15,121, more than double what we estimated.
Even though Dr. Eliz was the leading vote-getter in the special election, she failed to get over 50% of the votes, forcing a run-off. She ended up losing in the run-off to Gary Gates, the second leading vote-getter, 17,484 votes-to-12,629 votes.
At the end of the day, it’s unclear whether providing a more accurate number would have aided Dr. Eliz in winning the election. In reality, the campaign manager even inflated our estimate to 10,000 votes to try to account for the attention the race was receiving.
Taking a step back, I believe there is a lot to learn from our experience. We learned how important it is to factor in the intangible aspects of a race, and to not just rely on historical data. We also learned the value of using a healthy safety margin when estimating win numbers; it’s better to work towards a more ambitious goal and fall short than to be blindsided on election day.
All in all, our experience is well summarized by one of history’s most quotable characters.
"In theory there is no difference between theory and practice – in practice there is."
- Yogi Berra
You can help advance Bluebonnet’s mission of democratizing data by chipping in $20.21 on ActBlue: https://secure.actblue.com/donate/bluebonnet-data.
About the author: Ben is a Co-Founder of Bluebonnet Data and a Partner & Data Scientist at Moksha Data in Houston, TX. He is passionate about work that lies at the intersection of data science and social impact.