Scaling laws: a viable alternative to value at risk?

Abstract

Recent research has found a number of scaling law relationships in foreign exchange data. These relationships, estimated using simple ordinary least squares, can be used to forecast losses in foreign exchange time series from as little as one month’s tick data. We compare the loss forecasts from a new scaling law against six parametric Value at Risk models. Compared to these models, the new scaling law is easier to fit, provides more stable forecasts and is very accurate.

Keywords:

• Value at risk
• Extreme value theory
• Financial time series

JEL Classification:

• C14
• C16

Acknowledgments

I am especially grateful to Dr Wing Lon Ng for his consistent support throughout this study. A series of invaluable meetings with Dr Richard Olsen provided a great deal of inspiration for the methodology and were also most enjoyable.

Notes

1 see Glattfelder et al. (2011), table A6, p. 18.