The Cleveland Fed Found Reflexivity and Measured It: Showing People a 14.3% Return Raised Desired Allocation 47% and Purchases 23%
DEEP-DIVE ON THIS CARD. Two cards on this feed cover this study (it appears twice, under macro and news_corporations). It deserves one careful reading, because the design is stronger than the coverage suggests.
WHAT THE STUDY DID.
The Federal Reserve Bank of Cleveland released a working paper on July 14, 2026 titled 'Do You Even Crypto, Bro? Cryptocurrencies in Household Finance', examining the psychology and demographics of US crypto investors.
Two findings, of quite different evidential weight.
The observational finding: differences in expected returns explain more about who buys bitcoin than age, income and education combined. Interesting, but correlational - people who expect high returns buy, which is close to definitional.
The experimental finding is the one that matters. A randomised information experiment gave a treatment group information about bitcoin's recent performance - specifically a 14.3% past-year return. Relative to control, that raised desired crypto allocation by about 47%, and the treated group was roughly 23% more likely to actually purchase bitcoin in subsequent survey waves.
WHY RANDOMISATION CHANGES THE CLAIM.
Almost everything written about crypto sentiment is correlational: prices rose and buying rose, so one presumably caused the other. Reverse causation and common causes are unaddressed.
Randomisation removes that. Treatment was assigned, not chosen. The only systematic difference between groups is the information. So the resulting difference in allocation intent and in realised purchases is caused by exposure to past returns. This is direct causal evidence that showing people bitcoin's past performance makes them buy bitcoin - and critically, it is measured on actual subsequent purchases, not just stated intent, which is where most survey research fails.
WHAT IT IMPLIES MECHANICALLY.
A feedback loop with a measured gain. Price rises, the return figure improves, exposure to that figure raises expected returns, which raises allocation, which raises price. The paper's framing - explaining persistent volatility and how rallies attract new buyers - is exactly this.
Note the asymmetry the experiment implies but does not test: the treatment was a positive return. If the mechanism is symmetric, showing people a negative trailing return should suppress allocation by a comparable magnitude, which would make the loop bidirectional and self-reinforcing in both directions. That is the natural next experiment and it has not been run here, so the downside leg remains an inference rather than a finding.
THE HONEST LIMITATIONS.
It is a working paper, not peer-reviewed or an official Fed position. A 14.3% trailing return is a mild stimulus; effect sizes at extremes are unknown and unlikely to be linear. Survey populations skew toward the already-interested. And a 47% increase on a small base allocation is still a small allocation.
WHY IT IS ACTIONABLE.
It provides a mechanism for something this feed observes constantly. The August rally drew ETF inflows of roughly $1 billion in three days after months of outflows - the study says that pattern is partly caused by the rally itself, not merely coincident with it. Flow-based sentiment indicators are therefore partly a lagging function of price, and should not be read as independent confirmation of it.
Sources (5)
AI Research
Key Takeaway
A Cleveland Fed working paper released July 14, 2026 ('Do You Even Crypto, Bro? Cryptocurrencies in Household Finance') finds expected returns explain bitcoin ownership better than age, income and education combined. In a randomised information experiment, telling people about bitcoin's 14.3% past-year return raised desired crypto allocation ~47% versus control and made them ~23% more likely to actually buy. That is a causal measurement of the price-begets-buying loop.
DEEP-DIVE ON THIS CARD. Two cards on this feed cover this study (it appears twice, under macro and news_corporations). It deserves one careful reading, because the design is stronger than the coverage suggests.
WHAT THE STUDY DID.
The Federal Reserve Bank of Cleveland released a working paper on July 14, 2026 titled 'Do You Even Crypto, Bro? Cryptocurrencies in Household Finance', examining the psychology and demographics of US crypto investors.
Two findings, of quite different evidential weight.
The observational finding: differences in expected returns explain more about who buys bitcoin than age, income and education combined. Interesting, but correlational - people who expect high returns buy, which is close to definitional.
The experimental finding is the one that matters. A randomised information experiment gave a treatment group information about bitcoin's recent performance - specifically a 14.3% past-year return. Relative to control, that raised desired crypto allocation by about 47%, and the treated group was roughly 23% more likely to actually purchase bitcoin in subsequent survey waves.
WHY RANDOMISATION CHANGES THE CLAIM.
Almost everything written about crypto sentiment is correlational: prices rose and buying rose, so one presumably caused the other. Reverse causation and common causes are unaddressed.
Randomisation removes that. Treatment was assigned, not chosen. The only systematic difference between groups is the information. So the resulting difference in allocation intent and in realised purchases is caused by exposure to past returns. This is direct causal evidence that showing people bitcoin's past performance makes them buy bitcoin - and critically, it is measured on actual subsequent purchases, not just stated intent, which is where most survey research fails.
WHAT IT IMPLIES MECHANICALLY.
A feedback loop with a measured gain. Price rises, the return figure improves, exposure to that figure raises expected returns, which raises allocation, which raises price. The paper's framing - explaining persistent volatility and how rallies attract new buyers - is exactly this.
Note the asymmetry the experiment implies but does not test: the treatment was a positive return. If the mechanism is symmetric, showing people a negative trailing return should suppress allocation by a comparable magnitude, which would make the loop bidirectional and self-reinforcing in both directions. That is the natural next experiment and it has not been run here, so the downside leg remains an inference rather than a finding.
THE HONEST LIMITATIONS.
It is a working paper, not peer-reviewed or an official Fed position. A 14.3% trailing return is a mild stimulus; effect sizes at extremes are unknown and unlikely to be linear. Survey populations skew toward the already-interested. And a 47% increase on a small base allocation is still a small allocation.
WHY IT IS ACTIONABLE.
It provides a mechanism for something this feed observes constantly. The August rally drew ETF inflows of roughly $1 billion in three days after months of outflows - the study says that pattern is partly caused by the rally itself, not merely coincident with it. Flow-based sentiment indicators are therefore partly a lagging function of price, and should not be read as independent confirmation of it.