The Psychology Gap: Why Austin Tips and El Paso Doesn't

Based on the research ofEbert, Obschonka, Aiello, Potter, Rentfrow and Gosling, "Innovation adoption across regions: psychological openness and the early diffusion of digital platforms," Research Policy, 2026

The Variable No Launch Memo Asks About

Every two-sided platform faces what economists David Evans and Richard Schmalensee called the critical-mass hurdle: a marketplace has no value until enough people show up on both sides at once, and whether it clears that hurdle depends on the "nature of network effects, the dynamics of customer behavior, and the distribution of customer tastes" in the place it launches. Market-entry teams have spent two decades operationalizing that hurdle with the same five inputs: population size, household income, housing density, smartphone penetration, and regulatory friendliness. Rank the cities, launch in order, repeat.

A new Research Policy paper by Tobias Ebert, Martin Obschonka, Luca Aiello, Jeff Potter, Jason Rentfrow, and Samuel Gosling adds a sixth input nobody puts in the spreadsheet. The team linked Big Five personality profiles from more than 3.2 million U.S. residents across 349 metropolitan regions to the documented early-adoption histories of three platforms that each required people to do something unfamiliar: Airbnb, Uber, and the early photo-sharing network Flickr. Regions with a higher share of residents scoring high in Openness to Experience, curiosity, receptiveness to new ideas, comfort with the unfamiliar, adopted all three platforms earlier, and the pattern held even after the researchers controlled for economic, demographic, infrastructural, social, and geographic differences between regions. Ebert put it bluntly: "Psychological openness was the only factor reliably associated with adoption across all of the three innovations we studied." The authors call this regional trait an "invisible social infrastructure", present in a city the way broadband or zoning law is present, just impossible to see on a map.

Same Spreadsheet, Opposite Outcomes

The scale of the hidden gap is the part a launch team would never guess from the usual inputs. Across the 349 metros in the study, the share of residents classified as highly open ranged from roughly 21 percent to 48 percent, a more than two-to-one spread sitting underneath cities that can look nearly identical on population, income, and density. The strongest early adopters in the dataset were San Francisco, Portland, and Austin; the weakest were El Paso, Flint, and Jackson. None of those rankings tracks cleanly with city size or median income. It tracks with the proportion of residents who were, on the trait the paper measured, already inclined to try the unfamiliar before their neighbors did.

That is the uncomfortable part for anyone running a launch-city model today. A metro that matches your top-performing market on every demographic input you screen for can still sit at the 24th percentile on openness instead of the 85th, and your model has no column for that difference. You will not find out until you launch there and the thing that worked twice in a row quietly does not work a third time.

Why the Earliest Users Have to Be Weird

The reason openness matters specifically for platforms, rather than for ordinary consumer products, is that platform adoption is not a purchase decision, it is a behavior nobody has normalized yet, and someone has to go first. Airbnb's own company history is a record of exactly how few people that first group was: the founders hosted their first paying guests in October 2007, the site "officially" launched at SXSW in March 2008 and drew two bookings, and only a relaunch timed to the August 2008 Democratic National Convention pushed that to 80. Each of those numbers describes a tiny cohort of San Francisco residents willing to let strangers sleep in their homes before the idea had any social proof behind it. Uber ran the same experiment eighteen months later: it went live in San Francisco in May 2010, asking ordinary drivers to accept rides from strangers through an app, and only reached New York a year later and Paris by December 2011, a sequence of exactly the kind of coastal, culturally experimental cities the Research Policy paper's openness map would predict.

Martin Obschonka, one of the paper's authors, frames the mechanism this way: "Someone has to be willing to try a new behavior before everyone else regards it as normal. This can be especially important for platforms, which need enough local users, hosts, drivers or content creators to generate the network effects that make them useful and viable." That is also the "hard side" problem that investor and writer Andrew Chen built his book The Cold Start Problem around: every network product starts with zero users and has to recruit an "atomic network" willing to show up before there is anything to show up for. Chen's framework treats that recruitment as a product and marketing challenge. The Research Policy data suggests part of it is a standing feature of the city you chose, some populations contain a bigger reservoir of people willing to be first.

The founders were not unusually good at convincing San Franciscans to try something strange. San Francisco contained an unusually large number of people who needed little convincing.

The Diagnosis Trap: Confusing Your City List With Your Strategy

This is where the counterintuitive cost shows up, and it is not that platforms pick the wrong cities, it is that they draw the wrong lessons from the right ones. A platform that happens to launch first in San Francisco, then Austin, then Portland, is launching in a run of unusually high-openness metros almost by accident, because those are the cities where adventurous early investors, employees, and press live. Every operational choice made during that run, the onboarding flow, the pricing, the trust-and-safety design, the referral mechanics, gets credited with the growth that followed. Then the same playbook gets pointed at a city that matches the winning cities on population, income, and density but sits near the bottom of the openness distribution, and it stalls. The natural read inside the company is that the ops team executed worse, or the local competitor got there first, or the market "just isn't ready." The paper's evidence points somewhere else entirely: the pool of residents willing to be an early host, an early driver, or an early public poster was smaller from the start, and no onboarding flow was going to manufacture a reservoir of openness that the metro didn't have.

Building the Psychology Layer Into the Launch Map

None of this argues that population, income, and density stop mattering; the Research Policy authors are explicit that openness is additive to those factors, not a replacement for them, and they are equally explicit that their evidence covers three specific platforms and should not be stretched to predict adoption of, say, AI tools without new data. What it argues is narrower and more actionable: for any product whose value depends on enough strangers doing an unfamiliar thing at once, the regional personality composition of your shortlist cities is a real, measurable, and currently unused predictor of how fast you cross the critical-mass line, and the data to approximate it already exists in published regional-personality research, such as Peter Rentfrow and Samuel Gosling's earlier mapping of psychological regions across the United States, built from more than 1.5 million respondents and showing the same coastal-versus-interior split in openness and creativity that shows up again in this new platform-adoption data.

The operational fix is not complicated to describe, even if it is new to most teams: add a psychological-receptivity estimate to the same shortlist spreadsheet that already carries population, income, and density, weight it the way you weight any other predictive input, and stop assuming that a city which matches your best market on four variables will behave like it on the fifth. The cities that look identical on paper are not always equally ready to go first, and the platforms that forget that tend to discover it only after they have already blamed the wrong team for a launch that was never going to tip.