Using Logistic Regression to Classify Stock Market Crashes

Summary

The article builds a logistic classification model intended to identify crash conditions in stock data. It discusses candidate predictors including company earnings measures, interest rates, and inflation, while explaining that supply-demand data and current events are omitted because of data access and modeling effort. Quarterly price-to-earnings observations are carried forward between releases to align them with the other time series.

For the target label, a script tracks the running high and marks observations as crashes when price has fallen by more than the stated threshold from that high. The article reports a strategy tester accuracy of about 70% for the Apple model and contrasts its result with a weaker Netflix presentation. These results are preliminary: the author explicitly questions the crash-labeling procedure and data collection, and the selected predictors omit potentially important influences. The approach is an educational example, not evidence of a robust crash-warning system.

Key ideas

• Logistic regression can be trained to classify market observations into crash and non-crash categories.
• The example considers macroeconomic indicators and company valuation data as candidate predictors.
• Its target labels flag a drawdown from the running price high that exceeds the article’s crash threshold.
• Quarterly valuation data are carried forward to align with more frequent observations.
• The reported Apple result is preliminary, and the author questions the data and labeling choices.

Tags

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.