Does money or character decide whether a business survives?
This is the argument people have at dinner. Did they make it because of who they are, or because of what they had. The study that answers it has been run once, on 653 German founders, and almost nobody cites it.
Correction, 2026.10.15. This page first said I had gone looking for that study and could not find one. That was wrong. An adversarial search run the same day found it in about fifteen minutes. The page has been rewritten around what it actually says. I am leaving this note here because the error is more useful than the tidy version would have been, and because the first draft of this page was built on a hole that turned out not to exist.
What the one study found
Evidence A representative sample of 653 German business founders, followed to 40 months, with personality measures and start-up capital entered in the same model predicting whether the business survived.
The variance in survival that the model explains, split by what explains it. Human capital, meaning education and experience, 35 percent. Business characteristics including start-up capital, 29 percent. Personality, 14 percent. Other personal characteristics, 21 percent.
So the answer to the dinner-table question, from the one study that asks it properly, is that what you brought and what you knew together account for roughly four times what your personality does.
Evidence The individual coefficients are worth seeing because of how small the personality ones are. Starting with 50,000 euros or more carries a coefficient of 0.191, and 10,000 to 50,000 carries 0.133, both significant. On the personality side, openness 0.033, extraversion minus 0.038, and locus of control 0.048. Conscientiousness, agreeableness and neuroticism are not significant at all.
The paper’s own summary of the money finding is one sentence. A higher start-up capital increases survival chances.
Once is not a literature
Observation One study, one country, 653 people, 40 months. The outcome is survival of the same business, which is not the same as the business doing well. And the authors did not set out to settle this argument, so nobody has replicated it, contested it, or run it anywhere else.
A search of all 630 papers citing the field’s own 2018 review, which asked for exactly this analysis, turned up nothing that delivers it. The German team did the work without citing the review. So the comparison exists, and it exists once, and the field that asked for it does not appear to know.
What each side can show on its own
What each side can show on its own
Evidence The largest survival effect anybody has measured belongs to money. Using US Census records on 33,485 firms, businesses starting with $100,000 or more were 23.0 percentage points less likely to close within four years than businesses starting with under $5,000. The best experience variable in that same model was worth 5.1 points.
Something else in that table is worth pausing on. Prior managerial experience raised the chance of closing, by 8.3 points. No story about character predicts that one.
Evidence Personality has a ceiling and it is low. A meta-analysis of 66 samples covering 15,423 people found the Big Five traits explain about ten percent of the variance in how a business performs. Education and experience do worse. A separate meta-analysis of 70 samples and 24,733 people put human capital at a corrected correlation of .098, roughly one percent of variance, with a credibility interval that included zero.
Risk propensity, which is the trait most associated with starting a business, predicts wanting to start one at .40 and predicts how the business performs at minus .02.
Then the careful studies reverse it
Evidence Danish population registers made a natural experiment possible. Researchers found 304 people who started a business only because an unexpected inheritance lifted a money constraint, and matched each against a similar founder who had not been constrained.
At five years, 41.9 percent of the inheritance-enabled founders were still trading, against 50.0 percent of the others. They were 8.85 percentage points more likely to close, and earned about 69,000 Danish kroner a year less.
Money caused the entry. The businesses money bought were worse.
Suggestion A structural model using American survey data points the same way. Removing credit constraints raised the share of people who started businesses from 19 to 22 percent and slightly lowered the average quality of the ideas that reached the market. Removing the role of preferences instead cut entry to 13 percent and raised idea quality by 19 percent.
So money is not a story about suppressed talent. It mostly does not stop good businesses from starting. It limits how big they get, and when it arrives as a windfall it buys entries that should not have happened.
The one table that gets closest
Evidence One study put measured ability and family background in the same equation, on 125,166 person-years of American data. Cognitive ability predicted becoming an incorporated business owner at a coefficient of .576. Then family income and parental education entered the model and it fell to .069 and stopped being statistically significant. Family income survived at .449.
Put both on the same scale and the point is sharper. Moving from the tenth to the ninetieth percentile is worth about a 57 percent change in the odds for family income. For measured ability it is 59 percent without family controls and 6 percent with them.
They look identical in size until they are measured together. Then money keeps its effect and measured ability loses almost all of it.
One more number from that paper. Everything in the model, aptitude, self-esteem, sense of control, risk-taking, race, sex, family income and parental education, explains about three percent of who becomes an incorporated business owner.
Why the comparison is run so rarely
Observation The reason is visible in the papers and it is not about difficulty. Economists measure wealth to the dollar and treat ability as whatever is left over. Psychologists measure traits to two decimal places and put money in a control vector they never report.
One paper carries startup capital in a survival model as a six-category variable and never prints its coefficient. The flagship Big Five and venture survival study has no money variable in it at all, while its abstract claims the traits beat experience, age and gender. That claim is true and it is not the claim a reader takes from it.
Two fields study the same outcome, each controlling badly for the other’s main variable. That looks less like a hard problem nobody solved and more like an artifact of which department owns which measurement. It is a reading rather than a finding, and the German paper shows the work is perfectly doable when somebody bothers.
Neither side wins
If money had won cleanly, the trait literature would be describing a circumstance and calling it a person. That is an argument this site makes elsewhere and it is not available here.
Capital has the biggest raw number and the weakest causal support. Character has a ceiling around ten percent of variance. Neither explains much, and the genre writes as though founder attributes decide the outcome.
The honest summary
Traits predict who opens the door. Money predicts who opens the door. On who is still inside, the one study that measures both puts personality at about 14 percent of the explained variance and what you brought and knew at roughly four times that. The whole model explains 23 percent of the variation in survival, which leaves most of it elsewhere.
Where this stops. Rich countries only, mostly American, Danish, German and Norwegian administrative data. Several key papers are paywalled and I read working-paper versions, including the German one, which can differ from what was published. Two Austrian studies look like they answer the same question and I could only reach their abstracts, so nothing from them is used here.
What I am not saying. That money does not matter or that who you are does not matter. Both predict entry clearly. And I am not saying 14 percent against roughly four times that is the final answer. It is one study, in one country, on 653 people, and it is the only one of its kind I could find after being wrong once about whether it existed at all.