In the lead up to the 2023 Chicago municipal elections, Change Research polled eleven races across ten districts, calling the winner correctly in 89% of races with an average absolute error of 5.5 percentage points.
In the lead up to the 2023 Chicago municipal election and runoffs in February and April, Change Research conducted eleven polls of ten different races ranging from Aldermanic races to the citywide Mayoral race. In total, we conducted surveys of the Mayoral and Aldermanic races in nine wards as well as citywide for the Mayoral race.¹
In our final poll of each of these districts, we correctly projected the winner’s final margin of victory to within 5.5 percentage points and called the winner correctly in 89% of the races. Our polls of the Aldermanic and Mayoral races were equally accurate, with both predicting the final margin with an average absolute error of 5.5 percentage points.
These final polls were completed fielding an average of 21 days before election day. As expected, our accuracy increased as the election neared: for the three ward-level polls conducted within the last two weeks, our average absolute error decreased to 4.7 percentage points.
Most traditional pollsters and panel providers are unable to poll in geographies and districts as small as a Chicago ward — many of which have registered voter populations under 30,000 people, and actual municipal voter populations as low as 6,685 in the runoff. However, Change Research is uniquely positioned to provide accurate and in-depth analysis in these areas. With our Dynamic Online Sampling, we are able to reach representative samples even in these small and diverse areas and accurately determine the state of the electorate.
While a poll’s accuracy is a crucial component of its quality, our advice and the actions that our clients take based on our polls are equally important. In the one ward in which we polled twice, we were able to see that our client took our strategic advice and their supported candidate’s lead increased from 5.2 to 13.0 percentage points in the 3 weeks between our two polls. Further, after our final poll, they were able to continue to pull away to win by a final margin of 17.5 percentage points.
¹ We polled one ward twice in the lead up to the runoff election.
When measuring when a poll was completed for this document, we used the final fielding date. For example, we would classify a survey that was fielded from February 15 to 17th, 2023 as being completed 10 days before the February 28th, 2023 general election.
For this analysis, we used 538’s measure of accuracy and calculated our deviation from the final margin of the top two candidates. While giving respondents an option to say that they are undecided in their candidate choice is vital when conducting a survey, when it comes to election day, only the votes of those that vote for one of the candidates are recorded. As such, when calculating the margin in our poll results, we standardize our prediction to include only those respondents who chose a candidate. A hypothetical example of how this works is given below.
| Final Election Result | Poll Result | Standardized Poll Result | |
|---|---|---|---|
| Candidate A | 51.1 | 49.8 | 51.4 |
| Candidate B | 46.9 | 43.8 | 45.2 |
| Candidate C | 2.0 | 3.3 | 3.4 |
| Undecided | — | 3.1 | — |
| Margin | 4.2 | 6.0 | 6.2 |
For the final calculation of error, we use the standardized results that exclude undecided voters. Since the final margin between candidate A and B was 4.2 percentage points and the margin in the normalized poll results was 6.2 percentage points, we would record an absolute error of 2.0 percentage points for this poll.
In each of our ward-level polls, we asked horse race questions about both the Aldermanic and Mayoral races. As the official results for both races are available at the ward-level, we calculate our accuracy for both races in each ward for the purpose of determining our overall accuracy.
For the Mayoral results in the wards, we calculated the winner separately within each ward. For example, in the April 4th runoff, Paul Vallas lost the overall race but beat Brandon Johnson in the tenth ward. As such, we counted Vallas as the winner in the tenth ward and recorded his share as the winner’s share when calculating accuracy in the tenth ward.
Change Research brings the same precision that wins elections to every polling project we take on.