Stats › Volatility and VIX › The HAR model
Three moving averages forecast two thirds of tomorrow's volatility and half of next month's
| Model | One day | One week | One month |
|---|---|---|---|
| Three averages only | 66.7% | 64.8% | 52.4% |
| Adding semivariances | 67.8% | 67.8% | 54.1% |
| Negative semivariance only | 68.8% | 67.6% | 53.9% |
| On bipower variation | 68.0% | 65.6% | 53.0% |
| Adding a signed jump | 68.9% | 67.7% | 54.1% |
| Adding a split signed jump | 69.3% | 67.8% | 54.1% |
Out of sample r squared in per cent for forecasting realised variance of the S&P 500 index tracker, from transaction prices between 23 June 1997 and 31 July 2008. The exact estimation window and forecast start date are not stated in the text. Note the size of the improvements: every refinement adds between one and three percentage points to a base of 66.7.
The standard realised volatility model uses a daily, a weekly and a monthly average and nothing else. Out of sample it reaches 66.7 per cent for the index tracker at one day, 64.8 at one week and 52.4 at one month.
Members only
The rest of this page is for members
Below this point there are 5 sections, 1 table and 4 named sources, roughly 750 words of it. Every figure carries the source it came from and the date the data is from.
You can keep browsing every statistic in the library for free. The intro and the summary are always open.