MerQuR+Statistics Glossary

Statistics Glossary → Log return

Log returnconcept

The natural logarithm of the ratio of two successive values in a price series: r_t = ln(P_t / P_(t-1)); it is usually multiplied by 100 to put it on a percentage-like scale.

Price levels are usually non-stationary and their variance grows with the price, so financial models are built on returns rather than levels. The simple return (P_t / P_(t-1) - 1) and the log return are almost identical for small changes: a 1% rise gives a log return of about 0.995 (×100). An important advantage of log returns is that they add up over time: the sum of successive daily log returns equals the log return over the whole period. They are also symmetric: if a price falls by 50% and then returns to its earlier level, the log returns cancel out exactly.

Volatility models such as GARCH model how the variance of a return series changes over time and usually work with log returns multiplied by 100; this scaling improves numerical stability and makes the parameters easier to read. For large changes simple and log returns differ noticeably, so the definition used should be reported. The log return is undefined for series in which the price is zero or negative. When computing portfolio returns, however, simple returns are more appropriate, because the simple returns of assets combine linearly.

An everyday example

A share rises from 100 to 150 and then falls back to 100. The simple returns are +50% and -33%, which add up to +17%, even though the price has returned to where it started. The log returns are about +40.5 and -40.5 (multiplied by 100); they sum to zero and correctly show that the price has not changed.

In MerQur

In GARCH, the 'Transformation (None / Log return × 100 / Percentage change)' parameter converts a price level into returns; the first observation is dropped.

Analyses where this concept appears

GARCH (Volatility Model)

Related concepts

Conditional variance (volatility) and GARCH orders