When constructing samples, researchers must be careful not to include just survivors (e.g., surviving companies, mutual funds, or investment newsletters). Since survivors tend to be those that have done well (by skill or chance), funds that have 10-year track records will exhibit performance histories with upward bias—mutual fund companies regularly discontinue funds with poor performance histories or roll their assets into better performing funds. Time period bias occurs when the period chosen is so short that it shows relationships that are unlikely to recur, or so long that it includes fundamental changes in the relationship being observed. A 10-year period typically includes a full economic cycle and is likely to be appropriate for this test. Look-ahead bias is present if the test relates a variable to data that were not available at the points in time when that variables outcomes were observed.