Measuring market timing: Henriksson-Merton vs Treynor-Mazuy (Excel)

How can one measure how well does the fund manager anticipate overall market movements and how accurate their forecasts are? Two most commonly used approaches for this are Treynor-Mazuy (1966) and Henriksson-Merton (1981) models. Today we are discussing the similarities and differences between them and learn how to apply these models in Excel.
Don't forget to subscribe to NEDL and give this video a thumbs up for more videos in Investment!
Please consider supporting NEDL on Patreon: / nedleducation

[@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

[@MrMahankumar]

Hey man great!!!! This is perhaps the nichest of the niche topics in finance that only serious academic practioners would know. Kudos Man for bringing this topic up!

[@NEDLeducation]

Hi, and many thanks for your comments and continued support of the channel!

[@akshaydamani]

Thank you for uploading this video. This brings in a lot of clarity as to the differences between the 2 models. Market Timing is a very significant skill of a fund manager and hence the models even if theoretical are food for thought. Thank you. Appreciate it.

[@raresblidar7721]

Hi! I was wondering: could the Henriksson-Merton method be generalised to test for multifactor exposure? So instead of using market excess return as explanatory variable, we would use SMB, HML etc. factor returns?

[@vikverreth5571]

Legend!

[@fanfan1184]

Hallo, firstly I must thank you for your amazing videos. The explain complex concepts in such an elegant and understandable way! An hour of intensely studying your videos is probably worth a full university course! I also have a question on this model in particular. The bases is the CAPM, I was wonder, if there was a way to adjust this model to account for a multifactor model? Also is there a way to account for the fund manager’s ability to not just time the market, but time his factor exposure with positive factor-risk premiums? Thank you again for your videos!

[@NEDLeducation]

Hi, and glad you are enjoying the channel! As for your question, it is possible to modify the model to include other risk factors and the general logic of the model would stay the same. Further, you could test whether the manager is timing individual factors (say, size, value, or momentum) using the same procedure. Hope this helps!

[@claytonestey767]

Hi Nedl. Do you have an opinion on extending Henriksson-Merton model to add an asymmetric up-side factor like gamma2*max(r_m - r_f,0)? So it could capture different treatment of up-side vs down-side timing even though what you said was sensible about managers preferring down-side timing.

[@NEDLeducation]

Hi Clayton, and thanks for an excellent question! Yes, you can augment the model this way and test whether upside timing exists and whether it is substantially different from downside timing.

[@ghulamnabi6331]

hi, plz make a vedio on Busse(1999) volatility model

[@NEDLeducation]

Hi Ghulam, and thanks for the suggestion! I will definitely make a video on volatility timing in the future.

[@parvshahhh]

Hey man great video but had some questions about the video so firstly could please tell me what are the other values in the treynor-mazuy table except the coefficients and the standard error and the degree of freedom secondly how has the formula been used which you have mentioned in the side of the sheet (The treynor-Mazuy Model), thirdly why did you subtract 1 while calculating the excess return fourthly what is the linest function and why did you take the full column of excess returns of arkk and put a comma and then put the full spx and spx^2 and then put 1 and then again 1 and lastly how do you know in the full 3 by 5 column the first row is gamma and not beta/alpha like it could be alpha too right? and why is there #N/A in the alpha column. Thank you

[@NEDLeducation]

Hi Parv, and glad you liked the video! Thanks for loads of relevant questions - they are mostly technical and related to Excel fundamentals (not to the concept of the market-timing model itself). The LINEST function allows you to flexibly implement a linear regression model in Excel (here is a video dedicated solely on that and on investigating the outputs of the function: kzbin.info/www/bejne/nZnbcoWlrJhlaac). Long story short, LINEST also provides the R-squared, the F-statistic (to test the statistical significance of the model itself), and ESS and RSS (variability explained by the model and residual variability). The coefficients in LINEST appear in the reverse order of the variables, so this is how one can know the order. Putting 1 and 1 in LINEST is simply to specify whether you want the constant and the additional statistics in your output. Subtracting 1 when calculating returns is to take into account how much is gained on top of the original investment (and not the factor by how much the investment is growing). Say, if you earned 20%, your capital has grown by a factor of 1.2. If you lost 10%, your capital has "grown" by a factor of 0.9.

[@akshaydamani]

In case of Henriksson Model, the formulae that is used and shown is: Rp-RF = Alpha + Beta*(Rm-Rf)+ Gamma*(D*(Rm-Rf)) + Ep.... Why is this different than the one shown in the video which uses : Gamma*(Max(0, Rb-Rm))... Please help regarding difference in the formula. Appreciate your reply on this. It will help clear out a doubt lingering snce 1 year or so in my study. Grateful and gratitute, Akshay

[@NEDLeducation]

Hi Akshay, and thanks for the excellent question! The two functional expressions you mention are mathematically equivalent, these are just different notations to express the same logic (the absolute value of returns below the risk-free rate or zero otherwise). Hope this helps! As a bonus, there are at least four equivalent ways to express it: (Rm-Rf)- (Rf-Rm)+ max(Rf-Rm, 0) min(Rm-Rf, 0)

[@vinaykriplani9947]

@@NEDLeducation if i change the dummy variable to 1 if rm>rf and 0 otherwise would the results change or be the same. i am pretty confused regarding implement the same model in excel as there is no information available elsewhere regarding how to do hm model in excel. thanks