Winning the right to buy a home at a regulated price was followed by a 21% drop in crypto app activity in South Korea. Unemployment was associated with an increase of about 6%, suggesting that financial goals perceived as out of reach may encourage greater risk-taking.
Costas Arkolakis, Spencer Y. Kwon, Dominic Russel, Claire Shi and Hyunjoo Yang examine these links in “Gambling for Goals”, published by the US National Bureau of Economic Research (NBER) in November 2026. Their findings suggest that investment decisions depend not only on income and wealth, but also on whether people believe they can achieve their financial goals.
Someone who cannot save enough to buy a home through regular saving may look for investments offering a chance of much larger gains, despite a substantial risk of loss. The study combines financial records, online activity and survey responses to investigate that possibility.
By the Editorial Team. English adaptation of the Polish original, published on 22 November 2026.
Financial and online records for more than 1.4 million people
The researchers used data from three South Korean companies: credit bureau KCB, telecommunications operator SK Telecom and payment card issuer Shinhan Card. The main sample comprised 1,427,629 people, whose financial circumstances and online activity were examined using data from 2021–2023.
The combined records covered income, wealth, debt, spending, and the use of mobile apps and websites. Individual analyses used smaller samples, depending on data availability.
A survey conducted on 19–20 March 2026 added responses from 2,301 South Korean residents. Participants assessed the importance of their life goals, their prospects of achieving them, expected investment returns and their own willingness to take risks.
The survey deliberately included a larger share of active crypto app users to examine their behaviour in more detail. Statistical weights were then applied to adjust the results.
Hard-to-reach goals were linked to greater risk-taking
Respondents who viewed risky investments as one of the most realistic routes to an important financial goal were more likely to choose an uncertain gain over a guaranteed one. In a hypothetical investment exercise, they chose between a certain 5% increase in wealth and a much larger gain with a correspondingly lower probability. Both options had the same expected value.
People who believed they could not achieve their most important goal without risky investments were 21 percentage points more likely to choose the risky option. The relationship remained after accounting for age, sex, income, education, wealth and employment status.
A one-standard-deviation increase in the average importance assigned to life goals was also associated with a 23% higher probability of owning cryptocurrency. That estimate accounted for factors including expected future returns and respondents’ stated attitudes towards risk.
The goals included buying a home, repaying debt, securing future financial stability and improving financial circumstances. The research therefore extended beyond housing, although home purchases allowed the authors to investigate the relationship between aspirations and investing in greater detail.
App activity is a proxy, not a record of trades
The administrative data did not show the value of participants’ cryptocurrency holdings. The researchers therefore used activity on crypto trading apps and websites as a measure of interest in the market.
About 17% of people in the relevant sample used those services in July 2021. Roughly 7% were classified as intensive users, with at least 30 recorded instances of crypto app or website use in a month.
The survey supported a link between online activity and cryptocurrency ownership. Crypto was held by 25% of respondents who had not used such apps during the measurement period, compared with 45% of moderately active users and 75% of the most active group.
The value of holdings differed too. The most active users reported an average of 22.2 million South Korean won (KRW) invested in cryptocurrency, compared with KRW 5.5 million among moderately active users.
Opening an app does not, however, establish that a trade took place. The telecommunications data could not show whether a user bought or sold cryptocurrency, how much they invested or whether they made a profit.
Crypto app use increased during unemployment
The researchers examined people who had been employed in the first half of 2021 and subsequently experienced unemployment between July 2021 and June 2022. They compared their activity with that of people with similar characteristics and financial circumstances who remained employed. The analysis with additional controls covered 450,990 people.
Among those who became unemployed, the probability of using crypto apps rose by 1.7 percentage points relative to the change in the comparison group. That was an increase of about 6% from the earlier level.
The probability of intensive use increased by 0.86 percentage points, or about 8%. The average monthly number of recorded instances of activity rose by around 0.2, also equivalent to a 6% increase.
The changes were most pronounced among people who had not previously used crypto apps. Most other categories of mobile services showed no comparable increase, suggesting the result did not simply reflect unemployed people spending more time on their phones.
The authors interpret the pattern in terms of deteriorating financial prospects. Losing a job can put a home purchase or a savings target further out of reach, potentially increasing interest in investments that offer a chance to build wealth quickly.
The study cannot establish that unemployed people actually invested larger sums. It observed increased app activity, not an increase in the value of cryptocurrency purchases.
Home purchases were followed by lower crypto activity
The researchers found the opposite pattern among homebuyers. An initial analysis of households’ online activity before and after a property purchase showed crypto app use declining by about 20% within a year. Other app categories, apart from property services, did not show a similarly large fall.
Buying a home can coincide with other changes that also affect financial risk-taking. Starting a family, earning more or becoming financially more secure could all make a home purchase more likely while reducing interest in risky investments.
To examine the relationship more closely, the authors turned to South Korea’s system of lotteries for homes sold at regulated prices. Eligible applicants can seek to buy a new home through a special allocation system. When demand for a development is high, buyers are selected by lottery from appropriately qualified applicants.
A win gives the applicant the right to buy at a fixed price, often below market value. Earlier research cited by the authors put the median discount at about 20%. For someone whose savings were insufficient to buy on the open market, winning could therefore substantially improve the prospect of homeownership.
A 21% decline ten months after the contract date
The researchers identified 1,892 probable housing lottery winners in 2021–2023. They did not have a complete list of participants, so inferred winners’ identities from household records, development locations and subsequent changes of residence.
They compared people who had already won with future winners who had not yet secured the right to buy at the point being analysed. Ten months after the home purchase contract date, monthly crypto app activity was lower by about two recorded days of use on average, a 21% decline relative to the period before the contract.
The probability of belonging to the intensive-user group fell by about 5 percentage points, equivalent to a 35% decline. The largest reduction in crypto interest occurred in the Seoul metropolitan area, where housing is particularly expensive relative to many households’ financial resources. Changes elsewhere were much smaller.
No similar decline appeared in stock trading apps. The authors also investigated whether buyers were reducing crypto activity because they needed to commit savings to the home purchase. The data did not indicate a broad contraction in financial activity or a significant fall in spending after a lottery win.
Nevertheless, the effect of achieving a housing goal cannot be fully separated from the changes in wealth and debt associated with buying property. The lottery awarded the right to purchase; winners typically moved into their new homes only two or three years later.
Crypto ownership peaked among people with some savings
The authors highlight the financial threshold that a household must cross to achieve a particular goal. With sufficient savings and stable income, it can gradually put money aside for a home without relying on high-risk investments.
A person with some capital but earnings too low to accumulate the required sum faces a different calculation. A safe investment may deliver a modest gain without making homeownership attainable, while a much riskier investment offers at least a chance of reaching the target.
This does not mean the poorest people always take the greatest risks. When savings are very small, even an exceptionally high return may be insufficient to buy a home, making the risky investment unattractive as well.
The survey findings were consistent with that pattern. Among non-homeowners who regarded buying a home as highly important, about 30% of respondents with wealth below KRW 50 million held cryptocurrency. Ownership rose to roughly 57% in the KRW 50 million–100 million group, then declined in higher wealth brackets.
Stock ownership followed a different pattern, becoming more common as wealth increased. Cryptocurrency ownership instead peaked among people who already had savings but were still some distance from being able to buy a home.
A model shows how optimism can compound the cost of risk
Using the survey and administrative data, the authors developed an economic model of how important financial goals influence the allocation of wealth between safe and risky investments. In one simulation, they removed the influence of housing aspirations while leaving the other model parameters unchanged.
For people with wealth of KRW 50 million–100 million, predicted cryptocurrency ownership fell from 57% to 23%, a decline of 34 percentage points. This was a simulation, not an observation of actual trades. It illustrates the importance that the model assigned to the desire to accumulate enough money for a home.
The authors then examined overly optimistic expectations of future returns. In the simulation, people in the same wealth bracket who expected high gains and strongly aspired to homeownership allocated about 60% of their wealth to cryptocurrency. Removing the influence of the housing goal reduced that share to about 25%.
Such a concentration in risky assets increased the potential cost of mistaken expectations. In the model, the welfare loss from excessive optimism was about twice as large for people pursuing a difficult financial goal as it was when that goal did not influence investment decisions.
Expressed as a wealth-equivalent cost, excessive optimism amounted to about 15% of savings. This was a simulated economic cost based on investors’ decisions and expected returns, not an actual loss incurred by the study’s participants.
Sources
- Costas Arkolakis, Spencer Y. Kwon, Dominic Russel, Claire Shi and Hyunjoo Yang, “Gambling for Goals”, NBER Working Paper No. 36010, November 2026. DOI: 10.3386/w36010.
- Yale Cowles Foundation, “Gambling for Goals”, Discussion Paper No. 2578, November 2026.
- Arkolakis et al., “Gambling for Goals”, full author manuscript dated 27 October 2026, including tables and methodological appendices.
