Day trading is the practice of buying and selling the same stock within a single trading session. In Learning, Fast or Slow, Barber, Lee, Liu, Odean, and Zhang (2020) analyzed 3.7 billion Taiwan Stock Exchange transactions from 1992 to 2006. Day traders lost an average of 23.9 basis points per day net of fees. Aggregate performance was reliably negative in 14 of the 15 years studied.
What the Study Found
Day traders lost 7 basis points gross per day before costs (t = −10.2); transaction costs more than tripled that figure. Survival rates were 44% at one year, 24% at two years, and 15% at three years. Profitable day traders consistently represented about 5% of all active traders from 1995 to 2006. Unprofitable traders generated 72% of aggregate day trading volume, rising to about 80% in later years. Predictably profitable traders made up less than 3% of all day traders and generated only 9.81% of day trading volume.
Methodology
The dataset covers every transaction on the Taiwan Stock Exchange from 1992 to 2006, totaling 3.7 billion purchases and sales. Individual investors accounted for over 99% of all day traders and 95% of day trading volume in the average month. The authors estimated abnormal returns using daily CAPM regressions and modeled quitting behavior with Kaplan-Meier survival analysis and a Cox proportional hazard rate model. Key controls include the log of past day trading days, log of days since first trade, and log of past trading volume.
Key Statistics
| Metric | Finding | Context |
|---|---|---|
| Gross intraday loss per day | 7 bps (t = −10.2) | Full sample, 1992–2006; before transaction costs |
| Net intraday loss per day | 23.9 bps | Full sample, 1992–2006; 10 bps commission + 30 bps transaction tax |
| Day trader survival at 1 / 2 / 3 years | 44% / 24% / 15% | Kaplan-Meier; traders with 10+ days of experience |
| Profitable traders' share of all active traders | ~5% | Monthly average, 1995–2006 |
| Unprofitable traders' share of volume | 72% overall; ~80% in later years | 1995–2006; traders with 10+ days of experience and net losses |
| Predictably profitable traders | <3% of traders; 9.81% of volume | Positive past returns + 40+ days experience in prior year, 1993–2006 |
| Persistence: unprofitable traders (50+ days) | 95.3% trade again within 12 months | Monthly sorts, 1993–2005 |
| Persistence: profitable traders (50+ days) | 96.4% trade again within 12 months | Monthly sorts, 1993–2005 |
| Annual day trading return skewness | −0.22 | Traders with 10+ days; vs. TSE individual stocks average of 2.36 (1981–2009) |
| Gross intraday return (Eq. A1) | r_gt = Σ(S^L − B^L + S^S − B^S) / Σ(B^L + S^S) | Trade-weighted; open positions marked to closing price |
| Cox proportional hazard model | h(t,x) = h₀(t)·exp(XB) | Models quitting probability as a function of past returns and activity controls |
Why This Matters
Negative aggregate returns undercut the rational "trading to learn" argument: entering a market with negative expected lifetime profits is not consistent with standard Bayesian decision-making. Experienced traders' continuation rates suggest that financial losses alone are insufficient to end day trading careers. Persistent trading in the face of losses points toward overconfidence or non-financial motivations — entertainment, gambling, sensation-seeking — rather than rational skill-testing. Brokerage firms and the government collect commissions and transaction taxes on every trade, profiting from day trading regardless of whether traders succeed.