| We’ve been hard at work since the election wrapped figuring out what happened and how we can better help Democrats win. I’m based in Vermont, and while winter for me means cross-country skiing in sub-zero temperatures, at Deck, it means evaluating our forecasts, adding new features, and telling stories about our amazing campaigns this past cycle. In this newsletter and the ones to come, I’ll spend some time outlining our past work and our plans for the next cycle. |
| 2022 Forecasts + Training Opportunities |
| This cycle, our forecasts performed really well, with an average national error of 2.9 pp, and what I’m so excited about is we had an average bias of 1pp. Significant polling bias against Democrats stoked fears of a red wave. Deck’s forecasts not only much more accurately predicted the final results of elections but most importantly were far less biased in doing so. |  |
| You can see our full breakdown of national forecasts by office type here. You’ll have to be patient for state-specific validation though, we are still waiting on precinct-level election results! But next, I want to dig deeper into our forecasts, bias, and what this all means. First, though, I wanted to share some upcoming events and trainings! – Data for Progress Open Mic, March 22nd, 4 pm ET: Hear from people in the progressive space about interesting new projects that combine data and politics. – Generation Data BigQuery and Looker training April 4th-May 4th: Learn more about how to create data visualizations with SQL. I am a Generation Data alumn and found the trainings incredibly helpful, also the alumni network is amazing! – Analyst Institute Basics of Evidence-Based Electoral Program Training, March 2nd, 2pm ET: This briefing covers how to use evidence to confidently choose the most effective goals, tactics, timing and volume of communication for your campaign or program and to understand how to apply evidence to your specific context. |
| What is Bias? This cycle there was a lot of talk about bias particularly against Democrats. Deck’s forecasts were biased against Democrats as well, but to a much smaller extent. But what is bias and why does that matter? Absolute error is the difference between forecasted vote share and actual vote share, so if we forecasted a Dem won 52% of votes and they actually won 54%, our forecast would have an absolute error of 2%. We talk about our error in terms of median absolute error across all races, but there is variance from one race to another. In some races our absolute error would be 0.5pp and it would be 3pp in others.Bias takes into account the direction of error as well as magnitude. Bias will always be equal to or smaller than absolute error, and it’s always better to have less bias. A forecast with high bias would consistently over or underestimate Democratic vote share in the same way. Though the different forecasts can have the same absolute error, they can have very different levels of bias. Here it’s helpful to think about a full state house chamber. If every forecast is off by 3 points in the same direction, it leads to substantial misallocation of resources to protect seats that turn out to be safe or attempt to win seats that aren’t competitive. If the error is unbiased, targeted seats are much more likely to actually be competitive.Deck’s forecasts, unlike many polling based forecasts, had a much smaller bias. This means you can trust our data to be accurate and to give you the full picture of your race. |
| 2022 Campaign Case Studies Finally, I want to share some case studies from our campaigns last cycle. Read more here about how campaigns used Deck for fundraising texting, digital ads, mail, and field programs. This cycle, voter contact on Deck was up x2 and digital ad spending was up x3 over the last cycle. Campaigns that used Deck did 3% points better than campaigns that didn’t. Campaigns like Kevin Hertle for MI State Senate and Tara Probst for PA State Rep used Deck’s audience to run winning digital campaigns and helped flip their respective chambers. US Senate and Gubernatorial races like Tim Ryan and John Fetterman for US Senate used Deck’s fundraising lists to raise money and identify new supporters. Smaller campaigns like Becky Fast for KS County Commission used Deck to identify and contact persuasion audiences. |
| That is all for now, look out for our next newsletter and more information on trainings coming up! As always, let me know if you have any questions or want to set up a time to chat about our work and plans for the next cycle. I’ll see you on Deck, CG |