Ohlson O-score
Estimates corporate bankruptcy risk via a logit model fitted to accounting variables.
How it works
Ohlson uses logistic regression to estimate distress from accounting variables such as size, leverage, liquidity and recent losses. The regression first yields log odds. The logistic transform turns those log odds into a probability under the fitted model, for the distress definition and horizon used in estimation.
Worked example
For a given input sheet, suppose the fitted linear predictor is −3. Its implied probability is 1/(1 + exp(3)), about 4.74%. A change of +1 in log odds takes the probability to about 11.92%, an increase of 7.18 percentage points. The unit change is on the log-odds scale, and the probability effect depends on the starting point.
Validation and limitations
Use publication-date accounting data and an out-of-time sample. Evaluate discrimination and calibration separately. Changing the sample prevalence or defining distress differently can shift probabilities even when rankings stay similar. Compare the intended use with the Altman discriminant score.