Friday, 1 May 2020

Using Regression Analysis to Screen for Stocks (Relative Valuation Approach)


1. Data Source: The data sets for this article have been sourced from moneycontrol.com (open to everyone) It looks like moneycontrol.com updates its data every day at 1930 hrs. As the earnings season progresses more numbers will get updated but most of the earnings numbers should belong to 2019. (GIGO - the quality of outputs is subject to the quality of inputs)

2. Consistency and Timing: Multiples approach to valuation is subject to whims and fancies of the analysts. For example, PE ratio which is the most widely used multiple across the global financial markets can have many variations within the rules of the game. Although price is the current market price most of the time, earnings can be previous years earnings or trailing twelve month earnings or forward earnings. An analyst who is bearish on the stock may use past earnings while an analyst who is bullish on the stock may use forward earnings. Even the price used can be the current price, weekly or monthly average or a moving average. To facilitate a fair comparison across companies consistency and timing in the definition of multiples is critical. Both the numerator and the denominator used in the multiple should match each other in consistency and timing. Example, Price to Earnings is a consistent multiple but Price to EBITDA is not a consistent multiple as it scales an equity variable to a firm level variable.

3. Descriptive Stats: Price multiples are skewed distributions as shown in the charts below and subject to selection bias. Negative multiples are not included in the sample as they are not useful in making any meaningful conclusions.





As you can see in the summary stats above, median multiple is a much better representative measure of these data sets than average multiple. These average multiples are affected by very large observations in the right tail. 

 4. Analysis and Application: There are some 5000 stocks listed on BSE and it is very difficult for ordinary unsophisticated retail investors to screen these stocks.

To be fair, it is not a great time to run this regression as many stocks may look under priced due to the impact of Covid-19. (most of the earnings numbers are of 2019 and prices are current prices) Nevertheless the results are presented below for your perusal.

Based on inputs provided by Prof. Damodaran's Relative Valuation a multiple regression has been run on Indian stock multiples and presented in the spreadsheets below.

Price to Earnings Multiple Regression:

https://drive.google.com/open?id=1o2dnsw-5MOUTTCi_0Z8bNTJM5BB1jiOk

Price to Book Value Multiple Regression:

https://drive.google.com/open?id=1r-JDRZB7c3ambWembvxqsfRFgWHVsIeg

EV to EBITDA Multiple Regression:

https://drive.google.com/open?id=1X1q3Oa94XQD_HWvpDlAbD8xBhiT4pnzG

(It is not really difficult to run a regression on an excel sheet and predict multiples and readers may give it a shot as this can be a good starting point to make more informed investment decisions !)

A summary of regression results is presented in the table below. By substituting these variables into the equation we can find a predicted multiple. In effect, we are estimating a multiple which is the dependent variable based on its relationship with the independent variables. 


 The results obtained by using these equations can be very different from the results obtained using discounted cash flow (intrinsic value) techniques. Nevertheless, we have a starting point which we can use to investigate further into the fundamentals of under priced companies to discover undervalued companies.Similar analysis can be performed across sectors or global markets.

(Reference and Acknowledgements Prof. Damodaran)

No comments:

Post a comment

your comments are welcome!