Cues vs. Content: A Proposal to Empirically Distinguish Lewis-Style Coalition Theory from Caplan’s Simplistic Theory

A guest post by Spencer Marlen-Starr

Inspired by my debate with the Lewis brothers, Bet On It reader Spencer Marlen-Starr recently posted this piece on Medium. I’m reposting it here with his permission. He’s looking for your feedback, so comments are especially welcome.

---

Note: I used ChatGPT 5 to help me write this article.
Also, I finally got the sufficient motivation to get off my ass and ask ChatGPT this question last week because of the dramatically increased political tension in this country following the assassination of Charlie Kirk. I think if the Lewis brothers are right, their thesis could be a key to de-escalating tensions in this country right now. However, I also care about the truth, and I am an empiricist, so it still matters to me if their theory or Caplan’s theory is more in line with the relevant data.

I already thought the Lewis brothers’ book “The Myth of Left and Right” was important enough for me, someone who is completely broke and struggling to find employment, to devote the time and effort to write a dual review of it and Caplan’s classic “The Myth of the Rational Voter” last year.

Introduction

In The Myth of Left and Right, the Lewis brothers argue that the concept of there being a real, single dimensional political spectrum with a left wing at one end of it and a right wing at the other end of it that is timeless and spans borders is a myth. They claim that what we call “ideology” in this domain is really a matter of who you belong with — not a coherent commitment to any underlying worldview such as markets or equality or any other single dimensional dichotomy for that matter.

In contrast, Bryan Caplan’s “simplistic theory” claims that leftists are, at bottom, anti-market, whereas rightists are anti-left. This means that the Lewis brothers and Caplan already have the same theory of what defines the right wing if you think about it.

I was completely persuaded by the Lewises’ theory of the political spectrum over Caplan’s after listening to them discuss their book with him and Robin Hanson on YouTube for a while until I realized a lurking potential problem, namely: almost all of their and my (in my head while listening) “counterexamples” are ‘right-wing’ or ‘conservative’ people who dislike markets.

But those cases do not distinguish between the two theories. What would really refute Caplan’s theory in favor of the Lewis brothers’ theory is the existence of multiple ‘left-wing’ individuals with a durable, sincere pro-market stance, ones that persist even under adverse tribal/social cues and are expressed without excessive caveats.

Just to be clear, their book did include at least one counterexample, that would be the change in support for tariffs from 2015 to 2017 among leftists in the United States, but one example is not enough.

So here is a modest proposal for an empirical test, made by an amateur nobody (me!), with the help of ChatGPT 5, for how the dispute can be settled that aims not only to be possible in principle, but also feasible.

The Core Logic: Cue vs. Content

If Caplan is right, then (ceteris paribus) leftists in the absence of social/tribal cues should oppose market-expanding policies (tranched across domains), and cueing should move them only marginally or not at all. By contrast, under the Lewis hypothesis, party/coalitional cues will often override content: people will follow endorsement or opposition signals over their prior economic policy preference intuitions.

To see which dominates, we need to pit content (market vs. state) against (social) cue (endorse/oppose by “Democrats” or “Republicans” or no cue).

Importantly, we must allow opposition cues (e.g. “Republicans oppose”) as well as endorsement cues (e.g. “Democrats support”), because it could be possible that left-wing respondents are more sensitive to out-group opposition than they are to in-group endorsement. Only by including both do we properly distinguish in-group attraction from out-group avoidance.

Experimental Design

Factors (fully crossed):

1. Policy Content (2 levels, within respondent):
   — Market+ (market-expanding reform)
   — State+ (state-expanding or regulatory intervention)

2. Party Cue (5 levels, between respondents):
   — No cue
   — Democrats endorse
   — Republicans endorse
   — Democrats oppose
   — Republicans oppose

3. Policy Domain (4–6 distinct issues, repeated within respondent):
   E.g. up-zoning/ministerial building approvals; occupational licensing reform; tariff reduction; price controls; rent controls; ride-share fare regulation; minimum wages; school‐choice vouchers/ESA.

Each respondent sees all domains in randomized order; only content and domain vary within, cue is assigned between.

Outcomes:

• 7-point support scale (Strongly Oppose through Strongly Support)

• Binary “willing to sign a petition / add your name”

• Manipulation checks:
  • “Which party do you believe supports or opposes this policy?”
  • “Does this rely more on markets or government actions?”

Hypotheses & Contrasts

Hypothesis 1 (Caplan’s content primacy):
Among self-identified liberals, progressives, and/or leftists, in the No cue arm, market-expanding policies will face substantially lower support (i.e. most leftists oppose them most of the time). Cues will shift opinions only modestly because their underlying ideological aversion dominates.

Hypothesis 2 (Lewis’s social cue primacy):
Cues (endorse or oppose) will move support more than content shifts — i.e. a cue of Democrat endorsement or a cue of Republican opposition for a market policy boosts leftist support more than content itself.

Hypothesis 3 (Negativity / asymmetry test):
Opposition cues (especially by out-group) will often have larger magnitude effects than endorsement cues. E.g. “Republicans oppose” will move leftists more than “Democrats endorse.”

Key counterexample test:
If a substantial subset of left-identified respondents begin to support multiple market-expanding policies under Republican opposition cues, that would challenge Caplan’s presumption of leftist anti-market affect and ideology.

Analysis Strategy

We can treat this as a two‐way (Content × Cue) factorial design survey, with repeated measures across domains, analyzed via ANOVA or, better, mixed‐effects models (respondent and domain random effects). I have not done a power analysis to see what the minimum necessary sample size would be for this.

Key interactions to estimate:

• Content × Ideology in the no cue arm (Caplan test)

• Cue × Content × Ideology — does cue flip attitudes over content at all? If so, is it flipped differently by ideology?

• Contrasts: (Dem endorse vs. Rep endorse) on Market+ for leftists; (Rep oppose vs. Dem endorse) as out-group vs in-group effect.

We’ll pre-register all contrasts, multiplicity corrections, exclusion criteria (attention checks, manipulation failures), and robustness checks.

Vignette Examples (Sketch)

Here are a few polished templates. The cue phrase is bracketed and interchangeable.

Up-zoning / Building Approvals (Market+)

“Many urban areas restrict new housing through discretionary review and lengthy permitting. A proposal would allow new apartment construction by-right if basic safety codes are met, limiting discretionary hearings to clear violations. [Cue: Democrats support / Republicans support / Democrats oppose / Republicans oppose / (no cue)]. Do you support or oppose this reform?”

Occupational Licensing Reform (Market+)

“Many jobs require a state license, even when the work is low-risk. A proposal would accept clean out-of-state licenses and reduce required coursework when no safety issue is shown. [Cue]. Support or oppose?”

Tariff Reduction (Market+)

“The U.S. currently imposes tariffs on a wide array of imports. A proposal would phase out many of these tariffs over five years to reduce consumer costs. [Cue]. Support or oppose?”

Price Caps on Essentials (State+)

“During periods of inflation, some states consider emergencies price caps on essentials like groceries. A proposal would authorize a 12-month cap on these prices. [Cue]. Support or oppose?”

Ride-Share Fare Cap (State+)

“A proposal would impose a regulated ceiling on ride-share surge fares during peak demand to protect affordability. [Cue]. Support or oppose?”

Supplement: Panel / Event-Study Component

Beyond the survey experiment, we can exploit real world “cue flips” — times when elite parties or leaders changed stance on market-relevant policies — and see whether mass opinion followed them more than any stable, content-based orientation. Once again, I have not done a power analysis to see what the minimum necessary sample size would be for this.

Candidate domains:

• Trade / Tariff policy (2016 → present): GOP moved from traditional free-trade to protectionism under Trump; Democratic elite signals also shifted.

• Charter schools / school choice: Democratic Party’s internal reorientation over years.

• Regulation / shifting stances on tech / platform markets / price caps.

With panel data (e.g. CCES, ANES, Nationscape etc.), we can run fixed‐effects or event‐study regressions:

If within‐person changes align with elite cue shifts more than with policy content, that supports Lewis more than Caplan.

Why This Matters

This proposal is not just an academic parlor trick. It gets at a deep puzzle: when we see left-right alignment, is it because of principled views (markets vs safety nets) or because people are following coalition cues? If the Lewis brothers are right, political persuasion, polarization, and policy debates look very different — people are following tribes, not making philosophical arguments.

Caplan agrees with them when it comes to people who view themselves on the political right. However, if Caplan is right, there is real antipathy towards markets when it comes to those who view themselves on the left, and cues matter only secondarily.

And crucially, by designing opposition cues and looking especially for left, pro-market respondents, this design directly addresses the counterexample gap identified by Caplan multiple times during his chat with the Lewis brothers. If we find robustly pro market leftists who stick with pro market reforms even under Republican support and Democratic opposition cues, Caplan’s core “left = anti-market” hypothesis starts to crumble.

Objection from Bryan Caplan

I emailed this article to Caplan, and he had a clever objection to the empirical test outlined in this article. Here was his objection in his own words:

Interesting. But since the Lewis brothers say their theory is 100% right, partisan cues should ALWAYS overpower doctrine, right?

It made me realize that in my attempt to keep this outline concise, I failed to explain a crucial justification for this test. Caplan is right that if their theory were pure cues all the way down for every policy belief, then partisan signals should always overpower doctrine.

However, the Lewis brothers explicitly include the key idea of anchor beliefs, i.e. a single or a couple of core policy views that initially pull someone into a tribe.

An example would be a gay man who came of age in 2010, deciding to join the blue team/left tribe because of his strong policy preference for gay marriage legalization. Then, subsequently adopting the rest of that tribe’s policy stances afterward as a matter of socialization into the tribe.

Anchor beliefs are sincerely held, at least initially, and therefore are not dominated by tribalism/social motivations. That means their claim isn’t that cues always override everything, but that once you’ve joined a coalition, cues do override subsequently adopted policy positions, i.e. Caplan’s objection is correct for all non-anchor beliefs.

So, the exception is when a survey item in my proposed research design happens to hit someone’s anchor belief. In that case, you’d expect the anchor to hold fast against the cue.

As a researcher, there’s no way to know in advance which belief is the anchor for any given participant. So, in practice, any experimental design would produce a mix: mostly cue-following behavior, with a minority of stubborn, doctrine-consistent responses that reflect anchor beliefs.

That is why I don’t think the Lewises’ thesis commits them to Caplan’s “always” prediction. It’s “almost always,” with anchors explaining the exceptions. I think this explanation is sufficient to serve as a defense of the research outline proposed in this article, but I will freely admit that this introduces some unfortunate fuzziness in the power of the results of this test to definitively decide the case between their two theories once and for all.

But to change the test so that it requires 100% cue domination 100% of the time would only allow for a more definitive answer between Caplan’s theory and a theory which is not to be found in the Lewis brothers’ book. If anyone reading this can think of a way to tease out anchor beliefs, please let me know!