What types of sampling bias should I know for CFA Level I, and how do they show up in finance research?

Sampling bias is a sneaky topic because the CFA exam tests it through realistic scenarios rather than definitions. Here are the three key types with finance examples:

1. Survivorship Bias

This occurs when your dataset only includes entities that 'survived' to the present, excluding those that failed.

Example: Ashford Research studies hedge fund returns from 2016–2025 using a current database. Funds that blew up or closed between 2016 and 2025 aren't in the database. The reported average return is biased upward because only surviving (usually better-performing) funds are included.

Impact: Average reported hedge fund returns may overstate true performance by 2–4% annually.

2. Look-Ahead Bias

This happens when a study uses information that wasn't available at the time an investment decision would have been made.

Example: Bridgemont Capital backtests a strategy that buys stocks with low P/E ratios based on annual earnings. But annual earnings aren't released until 60–90 days after year-end. If the backtest assumes January 1 knowledge of prior-year earnings, it uses information investors couldn't have had — inflating backtest returns.

3. Time-Period Bias

This occurs when conclusions are drawn from a sample period that isn't representative.

Example: Studying the effectiveness of momentum strategies using only 2009–2021 data (a prolonged bull market) would overstate their efficacy. Extending the sample to include 2000–2002 or 2007–2008 would show very different results.

Exam Strategy: When a vignette describes a backtest or research study, ask yourself: (1) Could some data points be missing because they failed? (2) Was information available when the decision was supposedly made? (3) Is the time period representative of different market conditions?

Deepen your quant skills with our CFA Level I community discussions.