HAR model explained: Heterogeneous autoregressive volatility (Excel)
HTML-код
- Опубликовано: 30 июл 2024
- Corsi (2009) proposed a very simple and intuitive model for the dynamics of variance that utilises realised variance and can be estimated using OLS. It relies on intraday or high-frequency data and is very useful for variance forecasting. Today we are discussing the concepts behind the Heterogeneous Autoregressive Volatility model (HAR) of Corsi (2009), its implementation in Excel, and its assumptions and limitations.
Don't forget to subscribe to NEDL and give this video a thumbs up for more videos in Econometrics!
Please consider supporting NEDL on Patreon: / nedleducation
You can find the spreadsheets for this video and some additional materials here: drive.google.com/drive/folders/1sP40IW0p0w5IETCgo464uhDFfdyR6rh7
Please consider supporting NEDL on Patreon: www.patreon.com/NEDLeducation
Hey, Nedl I totally love all of your videos! You have a unique way of showing complex mathematics in a very simple and applicable way. I would love to see a a video about the Superior Predictive Ability test. Hansen 2005 and Reality Check test by White 2000. With kind regards !!
Hi Sebastian, and many thanks for such feedback! I was looking through these tests and I am still trying to figure out a video-friendly way of implementing them, but I will definitely incorporate them into a tutorial at some point, appreciate the suggestion!
The intraday example would be nice!
Hi Eduardo, and thanks for the comment! Already in the pipeline, should be released shortly.
@@NEDLeducation THank you!
Hi Sir,
Many thanks for your video, but may I ask a question: can this model be adjusted to be included one or more regressors that suspected to account for that volatility? And may thanks for advance
I read the paper fo Corsi, and he then get a measure of the return using "the highest frequency
volatility component in the cascade (the daily one in this simplified case)". Would you please explain how he does that? Also, I think adding further components would improve the model, like a 2-week, 2-moths component
hey,thanks for your videos! And I want to ask some questions .I am now using R language to reproduce the prediction effect of the HAR-RV model. I first divided it into a test set and a training set, and then I used the training set to regress it to find its coefficient. Then I want to know, for RVt-1, RVt-5, RVt-22, what should I do if the training set does not meet the number of lagged items. For example, for RVt-22, should the regression data start from RV, RVt-1, RVt-5 corresponding to the first item of RVt-22?
I would be great to see your implementatnion of the extension of HAR model as specified in the paper: Forecasting realized volatility: The role of implied volatility, leverage effect, overnight returns, and volatility of realized volatility by Dimos S. Kambouroudis, David G. McMillan, Katerina Tsakou
THANK YOU VERY MUCH! Your videos are very clear and helpful. But I have a question about this method. I have read that HAR is inspired by Heterogeneous Market Hypothesis just like the HARCH methods. What is the difference that makes HAR doesn't concern with conditional heteroscedastic as HARCH does? What conditions should we consider before using one of those two methods?
Can you please consider Adaptive and Generalised models of the HAR - I read about them soewhere...
Hi Arindam, and many thanks for the suggestion! I have released a video on HARQ recently which is a very useful generalisation of HAR: ruclips.net/video/1n0nlrPvE8s/видео.html. Adaptive HAR is just a HAR model that is continuously estimated on a rolling window of data (as far as I have seen the term mentioned in papers).