Hey there! My name is Leo Mahe, and I'm a high school senior currently following the 2026 midterm cycle. As an avid follower of U.S. electoral politics, I first became interested in polling while following New York City's 2025 Democratic mayoral primary, where I watched as Democratic socialist Zohran Mamdani skyrocketed from relative obscurity into becoming the decisive challenger to former Governor Andrew Cuomo in the primary. Click here to read more.
For several weeks before New York City’s 2025 Democratic mayoral primary, I would track the New York Times' Latest Polls tab as poll after poll came in, each suggesting that Mamdani had gained momentum but was still far behind Cuomo. Yet as I analyzed the polling data, something felt off. Pollsters selected random participants using weighted averages based on age groups (among other factors), but determined these age groups using data from the previous election, assuming that turnout for 18- to 29-year-olds would stay as low as it had been in 2021. I, along with many others on the left, believed that Mamdani's surging popularity among younger New Yorkers could dramatically shift turnout in the state—a possibility no poll was really factoring in.
Mamdani would go on to win the primary by a significantly larger margin than any poll had predicted: 56.4% to 43.6%. Polls predicted that 18- to 29-year-olds would make up around 10.8% of the electorate, the official rate reported by the New York City Campaign Finance Board for the 2021 mayoral election. However, my own analysis of data from the 2025 primaries* shows that 18- to 29-year-olds made up around 19.3% of the electorate, nearly double the expected rate. This discrepancy in expected young voter turnout limited the accuracy of even the most reputable pollsters, making me realize I wanted a way to analyze poll results while accounting for these possible discrepancies.
The data is already there—people in different communities tell pollsters who they plan to vote for. That data, typically, is transformed into a few clean numbers: support for the Democratic candidate, support for the Republican candidate, and undecided voters. While I will also provide an aggregate tracker based on these figures, my broader goal is to compile polling data that can provide a more comprehensive understanding of what that data means.
While I will make my own predictions, my focus will be on providing context to the polls found on the New York Times poll tracker. Because I am only one person, I will be focusing on three key races in 2026, but will include a simpler tracker for other important races.
*I used data from the New York City Campaign Finance Board to estimate the share of the electorate who were between the ages of 18 and 29. See the full report here. To understand how I developed my own estimate based on this data, click here.
Some of my polling information is sourced from the New York Times poll tracker and 270toWin. My analysis of specific polls is based on the original material published by pollsters, but it will also use information supplied by The New York Times. When discussing previous election cycles, I reference
I first got interested in polling after watching "Let's Talk Elections" on YouTube. Check them out!
For the Maine, Michigan, and Texas Senate races, I provide an in-depth analysis of polls and previous electoral results. I only analyze polls labeled as "select pollsters" by The New York Times to ensure my data remains relatively accurate. Click here to learn how they define select pollsters. I do not include polls taken before August to ensure they reflect recent voter trends.
For each of these Senate races, I do each of the following:
Compare historical polling averages with election results, examining past errors and any methodological improvements.
Analyze current polling data, placing special emphasis on youth turnout rates and voting tendencies.
Assess potential changes in age-based turnout.
Using this information, I will provide my personal prediction for which turnout scenario will be most likely, and—as a result—who I believe will win the election. This prediction is my own opinion and should not be treated as fact.
For all other Senate races I am currently following, I include a polling aggregate that uses the latest polling data to show current voter preferences and election trends. The data is weighted based on recency, and polling from groups labeled as "select pollsters" by The New York Times is also weighted as more significant. Polling results are automatically sourced from The New York Times using Claude. All polling from March onwards will be visible.
I have published multiple articles covering different ongoing dynamics in the 2026 midterm elections. They are available on my Substack or in my school newspaper. Check them out!