Skip to content
Time Series & Forecasting

Anatomy of a Series

Autocorrelation: The Series Talking to Itself

9 min+80 XPStep 1/10

Correlation with your past self

Autocorrelation is ordinary correlation, except the two variables are the same series at different times: yesterday vs. today, last week vs. this week. The lag is the gap — lag 1 compares neighbors, lag 7 compares the same weekday.

Hot today, hot tomorrow: daily temperature has lag-1 autocorrelation near 0.9. That fact is the entire foundation of forecasting — the past genuinely carries information about the future. A series with zero autocorrelation is a series where tomorrow is a coin flip no matter what happened today.

insight

Autocorrelation is why forecasting is possible at all. Without it, history would be trivia — interesting, and useless.