Thanks @benamery21 - what a great connection! The goal is exactly that - rigorous statistical methods applied to rea-time data to detect shifts before they are obvious. I'll have to dig into Hamilton's work.
How far back do you have data on special election results and have you thought about incorporating special election data in a model forecasting seat swing?
Hi Alan, The Downballot has this data back to the end of 2016. I included everything that meets The Downballot's criteria since Trump's first term began - 438 elections to date. I do build data models for Force Multiplier - we are an all-volunteer group that has raised $33 Million+ to date for Democratic candidates and causes. Estimating whether or not we are in a wave type environment is essential for determining how much offense vs. defense is appropriate in any given cycle - so yes, I am using this in forecasting and others could as well.
I’m a little confused by your statement that we haven’t seen a shift in this second term. Do you mean that the Dem’s over performance is likely to hold?
Hi Phillippa, great question. In Trump's first term we saw the Democratic overperformance shift up and down over three distinct periods leading into the 2018 midterms. So far in this term, we've seen one sustained period of strong Democratic overperformance without a statistically significant change. So, in answer to your question - yes, the overperformance has held and we'll keep watching closely as we head into November.
Thanks @MPC — great question. Special elections are one of several indicators I track, and what I love about them is that they are real elections with real motivated voters. It takes awareness and effort to show up for a special election, making them a strong indicator of voter motivation - and with about 40 or more happening across the country every year, they give us a continuous, real-world read on voter motivation over time - one key indicator of a potential wave.
On 2026 vs. 2018 environment: The early evidence is encouraging — more to come. On 2006, I don’t have special election data to evaluate the same way that far back. However, there are some key structural differences and an important caveat on both comparisons: while the energy is strong, the playing field has changed significantly. We're starting with more seats, redistricting has compressed the number of truly competitive districts, and voter suppression efforts may add another layer. Strong wave conditions don't automatically translate to the same seat gains — but they do create the opportunity to go on offense in places we otherwise couldn't. In my work with Force Multiplier, we are doing exactly this.
Very nice. Reminds me of work on automatic early detection of recessions by Jim Hamilton.
https://econbrowser.com/archives/2026/04/a-worried-econ-watcher
Thanks @benamery21 - what a great connection! The goal is exactly that - rigorous statistical methods applied to rea-time data to detect shifts before they are obvious. I'll have to dig into Hamilton's work.
How far back do you have data on special election results and have you thought about incorporating special election data in a model forecasting seat swing?
Alan Abramowitz
Hi Alan, The Downballot has this data back to the end of 2016. I included everything that meets The Downballot's criteria since Trump's first term began - 438 elections to date. I do build data models for Force Multiplier - we are an all-volunteer group that has raised $33 Million+ to date for Democratic candidates and causes. Estimating whether or not we are in a wave type environment is essential for determining how much offense vs. defense is appropriate in any given cycle - so yes, I am using this in forecasting and others could as well.
I’m a little confused by your statement that we haven’t seen a shift in this second term. Do you mean that the Dem’s over performance is likely to hold?
Hi Phillippa, great question. In Trump's first term we saw the Democratic overperformance shift up and down over three distinct periods leading into the 2018 midterms. So far in this term, we've seen one sustained period of strong Democratic overperformance without a statistically significant change. So, in answer to your question - yes, the overperformance has held and we'll keep watching closely as we head into November.
I like the write up, very thorough. I am curious if the data indicates a wave bigger than 2018 and closer to 2006.
Thanks @MPC — great question. Special elections are one of several indicators I track, and what I love about them is that they are real elections with real motivated voters. It takes awareness and effort to show up for a special election, making them a strong indicator of voter motivation - and with about 40 or more happening across the country every year, they give us a continuous, real-world read on voter motivation over time - one key indicator of a potential wave.
On 2026 vs. 2018 environment: The early evidence is encouraging — more to come. On 2006, I don’t have special election data to evaluate the same way that far back. However, there are some key structural differences and an important caveat on both comparisons: while the energy is strong, the playing field has changed significantly. We're starting with more seats, redistricting has compressed the number of truly competitive districts, and voter suppression efforts may add another layer. Strong wave conditions don't automatically translate to the same seat gains — but they do create the opportunity to go on offense in places we otherwise couldn't. In my work with Force Multiplier, we are doing exactly this.
Fascinating overview of processes and methodology. Thank you for sharing this, I look forward to reading more!
more to come!