GD POLITICS Podcast Summary
Episode: Does Moderation Win Elections? The Nerds Go To WAR
Host: Galen Druke
Guests: Elliot Morris (Strength in Numbers), Laksha Jain (Split Ticket/The Argument)
Date: August 21, 2025
Overview
This episode breaks down the heated debate in political data circles over whether "moderate" candidates have an electoral edge in U.S. House races. The discussion centers on two competing models for assessing candidate overperformance, called "WAR" (Wins Above Replacement), and evaluates the broader ramifications for parties aiming for electoral success, particularly the Democrats. With humor and rigor, Galen moderates an unusually civil face-to-face debate between the two model architects whose analyses have fueled Twitter spats, Substack screeds, and thinkpiece wars.
Key Discussion Points
1. Setting the Stakes: Why Does Moderation Matter?
- Moderation’s value is a live debate for both academic and practical campaign reasons.
- Both guests lean center-left and care about Democrats winning, but their models aim to analyze effectiveness on both sides.
- Small differences—in the range of 1-2%—could decide tight elections and even the presidency.
Quote (Laksha):
"At the end of the day, elections are really close nowadays. And a difference of 1 to 2% is the difference between Kamala Harris being president and Donald Trump." (05:25)
2. Explaining the Models: Two Approaches to WAR
Split Ticket’s Approach (Laksha Jain) [08:29]
- WAR measures how much better (or worse) a candidate does compared to a generic candidate from their party.
- Controls for incumbency (a big advantage) and "lag partisanship" (historical party alignment in a district).
- Not all Trump voters are loyal Republicans; controlling for complex voting history is critical.
- Candidates are assessed vs. actual presidential results in their district.
Examples:
- Jared Golden (ME-2): Overperformed by 6 points over a replacement candidate.
- AOC & Richie Torres: Outperformed Harris by 7-8%, but in right-trending districts so considered slight underperformers.
Conclusion: Blue Dogs (moderates) did about 5 percentage points better than progressives between 2018–2024.
Strength in Numbers (Elliot Morris) [14:18]
- Predicts House candidate performance using pre-election factors: prior presidential lean (last 2 cycles), demographics, candidate experience, fundraising.
- Uses a fully Bayesian model to quantify not just the effects, but the uncertainty around them.
- Tries to compare House candidate performance to that of a hypothetical replacement House candidate, not just to presidential results.
- Moderation yields a 1–1.5 percentage point boost with large uncertainty; difference with Split Ticket partly comes from modeling choices.
3. Fundamental Disagreements & Clarifications
Presidential Baseline vs. House Replacement [18:54, 21:30]
- Laksha: Not using the actual presidential results as a baseline misses crucial context.
- Elliot: Using actual presidential results can artificially inflate the perceived benefit of moderation due to baseline differences.
Role of Uncertainty [24:37]
- Elliot emphasizes massive uncertainty bands; asserts practitioners have traditionally been overconfident.
- Laksha: Cautions against “paralysis by uncertainty” and asserts that, even given uncertainty, consistent patterns—moderates overperforming—are robust.
4. Does Moderation Help? How Much and Why?
Agreement:
Both models show a benefit for moderation; size and confidence vary.
Laksha:
- Blue Dogs overperform, consistently.
- It’s a pattern true on both sides: “extreme” candidates (Greene, Gaetz) do worse among Republicans, too.
- “Moderation” means a mix of policies, a pragmatic/heterodox approach, and being willing to break with the party.
Elliot:
- Yes, but much less than Split Ticket claims, and the confidence intervals are huge.
- "Moderation" is a combination of ideological positions and candidate skill—how voters perceive a candidate (“vibes”) may matter more than measured ideology.
- The DCCC shouldn’t always pick moderates—fundraising, district fit, and other factors may be equally or more important.
Memorable exchange:
A (Galen): "Is it fair to say that you both find that moderation helps? The question is just degree?"
B (Elliot): "Yeah, this question is degree and implications." (49:58)
5. Defining Moderation and Its Role [53:36]
Laksha:
- Moderation means pragmatism, working across the aisle, willingness to diverge from party orthodoxy, and being perceived as relatable to a wider electorate.
Elliot:
- Moderation is often “just vibes”—candidate skill, fundraising, and district fit may explain "overperformance" as much as actual ideology.
- Perceived moderation by voters may matter more than ideology measured by congressional votes.
Laksha retorts:
- It’s not “just vibes”—concrete policy choices distinguish moderates from extremists.
6. Reaction to Critiques & Academic Pushback [42:08]
- Elliot and Laksha discuss public criticisms, notably from Adam Bonica (Stanford) and Jake Grumbach (UC Berkeley), who accused Split Ticket’s model of inherent pro-moderate bias.
- Laksha: Sees their critique as misunderstanding; controlling for lag partisanship/incumbency is good political science, not bias.
- “That is basically just reverse engineering to fit the conclusions you want.” (45:06)
7. Practical Implications and Takeaways [68:01]
- WAR-like models may inform recruitment and campaign strategy but should be one of many tools.
- Moderation "directionally helps," but it's not the only thing that matters (fundraising, message, district fit).
- Lessons from data should be applied with humility about uncertainty and with district/community context in mind.
Laksha: "Our war model number one does say, yes, moderation is important... I think it at least tells you that this theory of winning by just going all in on your base... meaningfully decreases your in-cycle odds of success, which I think is important to note." (68:19)
Elliot: "We don't actually know a lot of the time what is the reason that a candidate won their campaign... That boils down... to uncertainty in elections... [This] pushes us in the direction of doing a lot of due diligence to select skilled candidates." (71:20)
Notable Quotes and Moments
- “Moderation is just vibes.” —Elliot, poking at the idea that “moderation” may mostly be candidate relatability rather than policy positions. (55:13)
- “Paralysis by uncertainty” —Laksha on why some modelers over-weigh error bands and refuse to draw actionable conclusions. (36:17)
- “You can start with the same data and process it in different ways and get very different results... models are not divined by God... they are codes that represent how we view the world.” —Galen reflecting on modeling philosophy. (72:15)
- Both guests stating plainly: “I wouldn’t” when asked if ideological extremism ever helps. (59:02)
Timestamps for Major Segments
- 00:00-04:06 | Episode intro, stakes, ground rules
- 04:43-07:41 | The stakes for Democratic Party and war model motivations
- 08:29-14:05 | Split Ticket’s model: method and findings
- 14:18-17:52 | Strength in Numbers: method and findings
- 18:54-24:37 | Fundamental disagreements, baselines, and uncertainty debate
- 26:22-30:00 | Circle-back on uncertainty, candidate skill, and consistency
- 30:49-34:19 | Model critique: Are we comparing moderates to moderates?
- 42:08-49:44 | Criticisms from academics, model bias, and ideology correlation
- 53:36-58:10 | What is “moderation” and what role does perception play?
- 59:30-60:59 | Non-WAR data: public opinion, issue polling, and forecasting
- 68:01-72:15 | Strategic implications and final thoughts
- 73:58-end | Reflections on modeling, civility, and meta-lessons on political analysis
Final Reflections
Both guests model civil disagreement as well as war above replacement. Their models agree: Moderation helps, on average, and is almost never a liability, but the magnitude, mechanism, and confidence level remain hotly contested. The debate illustrates both the power of modern data modeling and the importance of scrutinizing both methods and interpretation—reminding political junkies that what you get out of a model is shaped as much by your modeling decisions as by “the data” itself.
Elliot (Final Word):
"This was a much more productive conversation than the one that has been happening on Twitter... I've got a lot of respect for him and we're friends in real life, so don't take this out of context." (73:58)Laksha (Final Word):
"Our disagreements are just in design, and those are fundamental disagreements. But again, two smart people... can disagree without going into bad faith attacks." (74:18)
For Further Reading or to Subscribe: gdpolitics.com
