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
The video was much in line with what I was looking for, Thanks for that. Following this video I have two questions. 1) How can we do the regression analysis with just applying the "Regression" Analysis available in Excel 2) For multiples like EV/ Sales or EV/ Ebidta how to address the drivers and come up with Fair value with the same way you did for P/E and P/B
Hi Aarsh, and thanks for the comment, glad the video helped! As for your questions: 1) You can select the same arrays when specifying the regression tool from the data analysis tab, make sure to untick the box for this particular application so the constant term is not included. However, I strongly advise using LINEST as it is more flexible and allows to see where the results originate from. I have also got another video where I discuss the fundamentals of the linear regression and how to treat it as a matrix algebra exercise: th-cam.com/video/mFuOzDv9xbM/w-d-xo.html 2) Enterprise value multiples are not directly translatable into value drivers as these are concerned with the joint value of debt and equity for a particular company. However, you can always introduce an EBITDA-based value driver, for example, by simply calculating EBITDA per share, and then dividing by the price to obtain "EBITDA yield". Hope it helps!
Another great video. Thanks for the regular content. Neat trick for overcoming heteroscedasticity. Couple of comment though, the valuation multiples are not a product of such a regression but rather directly computed using price data and fundamental data from company fillings. Additionally, the main reason why I think quants use inverse of typical fundamental ratios is in order to reduce the chance of division by zero or NA.
Hi Aravind, and many thanks for your feedback! Glad you liked the video. As for your comment, exactly, that is why I call the E/P, B/M, and S/P "inverse valuation multiples". While negative book equity or EPS is the main reason such method is more general, zeroes and N/A values can also be a valid point! I had a video on a more traditional practitioner-inspired approach to valuation multiples (it is actually the very first video I have ever made for this channel, check it out if you are interested: th-cam.com/video/HtH0OnJEbEc/w-d-xo.html).
Hi, and thanks for the question! It is Liu, Nissim, and Thomas (2002) "Equity valuation using multiples" : onlinelibrary.wiley.com/doi/abs/10.1111/1475-679X.00042
I just did it. Have 8 predictors for industry sector based on 50 + stocks. R2 is 0.96 +- 7, F stat =158, df 44 Equation is intercept( -6) + 0.48 *dividend yield + 0.25*price/earnings + 0.02 *book value + 0.29* "52 week low" + 0.71 * "52 week high", + 2.1*10 exp-11 * market capitalisation + 1.04 * price/sales - 0.35* price/book. Regressor Y is price. 52 week high p
14:31 I was curious as to how the forecasts fared after 4 years. Out of 20, only half were hits. Of course, this is an example case: three metrics clearly won’t explain everything. UPS: correct, it dropped. Honeywell: basically correct; it didn’t grow, just stayed at the same level. UNP: incorrect, it grew. Boeing: correct, it dropped. Lockheed: incorrect, it grew significantly. Raytheon: correct, it grew significantly. 3M: correct, it didn’t grow, stayed at the same level. Caterpillar: incorrect, it doubled in value. General Electric: correct, it grew astronomically. Deere & CO: incorrect, it grew significantly. FedEx: correct, it didn’t grow. CSX: incorrect. Illinois Tools: incorrect. Norfolk Southern: correct, it didn’t grow. Northrop Grumman: incorrect, it grew significantly. Waste Management: incorrect, it grew significantly. Eaton Corp: incorrect, it grew significantly. Emerson Electric: incorrect, it grew. General Dynamics: correct, rose slightly Roper Technologies: incorrect, rose
Hi Tan, and thanks for the question! This particular estimation is a weighted least squares estimation where all variables are divided by the price, so the constant should also be scaled accordingly.
Hi Debopriya, not necessarily. Being from the same sector is generally more important than being from the same exchange for this valuation approach. Hope it helps!
Hello Savva! Another awesome video! One practical question or better to say remark about it. The model you have presented us is practically cross-sectional model, not time-series model. To develop this model and apply it in reality we would need to test it on some time-series model, or better to say if we apply exact what you have shown to us we would need to constantly adjust the model because all of those indicators change over time? Am I right?
Hi Ivan, and glad you enjoyed the video! Absolutely, this is how you would back-test the fundamental analysis strategy to determine if it works or not.
I am just starting to go through your videos and I found them most interesting. One question on this one : if the constant leads to conclude that any other criteria would be marginally useful too calculate the fair value, it would be interesting to backtest it. Intuitively, I would have thought that FCF and dividends (which are partially captured by earnings) for this kind of companies would have some significance.
Hi, and glad you are enjoying the channel! As for your question - absolutely, you could interpret the constant in such a value driver model as a sign of some potentially important variables missing from the specification. This framework is quite useful also because it allows to quite naturally test the importance of various factors, from cash flow, as you stated, to something like R&D expenditure, scaling all of them by market cap.
@@NEDLeducation Thanks for the answer. I performed a quick test with the sell/buy recommendations deriving from your analysis by comparing the performance from your spreadsheet prices and yesterday's close; on this very simple analysis the regression is quite conclusive : the average performance is +31% for the buys vs. +17% for the sells. If you don't mind, I would like to exchange with you via e-mails on several topics, particularly (but not exclusively) on options.
Great to hear the strategy performed well! As for your request, please feel free to email me on savvashanaev@yandex.ru if you would rather ask something via email.
Looking back hind sight, many of the companies that were undervalued stayed undervalued because they were poorly managed companies. There is no perfect formula for valuation, every valuation is subjective. Stock prices are just a fundamental mechanism of human behaviors and corp governance. I would never rely on excel spread sheets to make my investment decisions.
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
You make great videos. Congratulations. I couldn't find this spreadsheet, what is the name of the spreadsheet?
The video was much in line with what I was looking for, Thanks for that.
Following this video I have two questions.
1) How can we do the regression analysis with just applying the "Regression" Analysis available in Excel
2) For multiples like EV/ Sales or EV/ Ebidta how to address the drivers and come up with Fair value with the same way you did for P/E and P/B
Hi Aarsh, and thanks for the comment, glad the video helped! As for your questions:
1) You can select the same arrays when specifying the regression tool from the data analysis tab, make sure to untick the box for this particular application so the constant term is not included. However, I strongly advise using LINEST as it is more flexible and allows to see where the results originate from. I have also got another video where I discuss the fundamentals of the linear regression and how to treat it as a matrix algebra exercise: th-cam.com/video/mFuOzDv9xbM/w-d-xo.html
2) Enterprise value multiples are not directly translatable into value drivers as these are concerned with the joint value of debt and equity for a particular company. However, you can always introduce an EBITDA-based value driver, for example, by simply calculating EBITDA per share, and then dividing by the price to obtain "EBITDA yield". Hope it helps!
LOVE IT MAN! Thanks for all you do!
Another great video. Thanks for the regular content. Neat trick for overcoming heteroscedasticity. Couple of comment though, the valuation multiples are not a product of such a regression but rather directly computed using price data and fundamental data from company fillings. Additionally, the main reason why I think quants use inverse of typical fundamental ratios is in order to reduce the chance of division by zero or NA.
Hi Aravind, and many thanks for your feedback! Glad you liked the video. As for your comment, exactly, that is why I call the E/P, B/M, and S/P "inverse valuation multiples". While negative book equity or EPS is the main reason such method is more general, zeroes and N/A values can also be a valid point! I had a video on a more traditional practitioner-inspired approach to valuation multiples (it is actually the very first video I have ever made for this channel, check it out if you are interested: th-cam.com/video/HtH0OnJEbEc/w-d-xo.html).
Clean explaination, well done!
Really nice video, my question is, where can I find the data to perform this analysis?
What is the name of the 2002 paper you mentioned at 3:39
Hi, and thanks for the question! It is Liu, Nissim, and Thomas (2002) "Equity valuation using multiples" :
onlinelibrary.wiley.com/doi/abs/10.1111/1475-679X.00042
I just did it. Have 8 predictors for industry sector based on 50 + stocks. R2 is 0.96 +- 7, F stat =158, df 44 Equation is intercept( -6) + 0.48 *dividend yield + 0.25*price/earnings + 0.02 *book value + 0.29* "52 week low" + 0.71 * "52 week high", + 2.1*10 exp-11 * market capitalisation + 1.04 * price/sales - 0.35* price/book. Regressor Y is price. 52 week high p
Excellent to see you have successfully applied the technique to your own dataset for your own research! Well done!
@@NEDLeducation thank you! Your video helped a lot! 😀
@@toshiro6589 Also watch out though because R^2 mechanically increases when you add more variables so it can seem higher than it is
14:31 I was curious as to how the forecasts fared after 4 years.
Out of 20, only half were hits. Of course, this is an example case: three metrics clearly won’t explain everything.
UPS: correct, it dropped.
Honeywell: basically correct; it didn’t grow, just stayed at the same level.
UNP: incorrect, it grew.
Boeing: correct, it dropped.
Lockheed: incorrect, it grew significantly.
Raytheon: correct, it grew significantly.
3M: correct, it didn’t grow, stayed at the same level.
Caterpillar: incorrect, it doubled in value.
General Electric: correct, it grew astronomically.
Deere & CO: incorrect, it grew significantly.
FedEx: correct, it didn’t grow.
CSX: incorrect.
Illinois Tools: incorrect.
Norfolk Southern: correct, it didn’t grow.
Northrop Grumman: incorrect, it grew significantly.
Waste Management: incorrect, it grew significantly.
Eaton Corp: incorrect, it grew significantly.
Emerson Electric: incorrect, it grew.
General Dynamics: correct, rose slightly
Roper Technologies: incorrect, rose
Why the constant equal to 1/Price? What confuse me is that why or how fix the constant to 1 (before expressing in terms of percentage of price)
Hi Tan, and thanks for the question! This particular estimation is a weighted least squares estimation where all variables are divided by the price, so the constant should also be scaled accordingly.
The comparable companies have to trade in the same index?
Hi Debopriya, not necessarily. Being from the same sector is generally more important than being from the same exchange for this valuation approach. Hope it helps!
Hello Savva! Another awesome video! One practical question or better to say remark about it. The model you have presented us is practically cross-sectional model, not time-series model. To develop this model and apply it in reality we would need to test it on some time-series model, or better to say if we apply exact what you have shown to us we would need to constantly adjust the model because all of those indicators change over time? Am I right?
Hi Ivan, and glad you enjoyed the video! Absolutely, this is how you would back-test the fundamental analysis strategy to determine if it works or not.
I am just starting to go through your videos and I found them most interesting. One question on this one : if the constant leads to conclude that any other criteria would be marginally useful too calculate the fair value, it would be interesting to backtest it. Intuitively, I would have thought that FCF and dividends (which are partially captured by earnings) for this kind of companies would have some significance.
Hi, and glad you are enjoying the channel! As for your question - absolutely, you could interpret the constant in such a value driver model as a sign of some potentially important variables missing from the specification. This framework is quite useful also because it allows to quite naturally test the importance of various factors, from cash flow, as you stated, to something like R&D expenditure, scaling all of them by market cap.
@@NEDLeducation Thanks for the answer. I performed a quick test with the sell/buy recommendations deriving from your analysis by comparing the performance from your spreadsheet prices and yesterday's close; on this very simple analysis the regression is quite conclusive : the average performance is +31% for the buys vs. +17% for the sells.
If you don't mind, I would like to exchange with you via e-mails on several topics, particularly (but not exclusively) on options.
Great to hear the strategy performed well! As for your request, please feel free to email me on savvashanaev@yandex.ru if you would rather ask something via email.
@@marketsandbeyond5584 I'm wondering if most diversified portfolio strategy would further improve returns on buys ( '_')
Looking back hind sight, many of the companies that were undervalued stayed undervalued because they were poorly managed companies. There is no perfect formula for valuation, every valuation is subjective. Stock prices are just a fundamental mechanism of human behaviors and corp governance. I would never rely on excel spread sheets to make my investment decisions.