Have you check this website.

http://www.bseindia.com/indices/betavalues.aspx

Thanks for the reply Iftekhar Ahmed. But, the site you suggested contains beta values only for 30 S&P stocks & that too only for 1 financial year. I am interested in beta values of the 50 Nifty stocks over a much longer period so that I can have time varying beta values.

60 months 

Thanks for the answer. Could you please provide the logic for using 60 mths period & not any other period? Thanks in advance.

Hemlata

Monthly.

Thanks for the answer. But could you provide information about the applications / research for which monthly data is used? Thanks in advance.

A practical answer is: try rolling regressions of 36 or 60 months and simply observe how much your Betas vary over time. Then you can decide. You can also try using periods of 1-year of weekly data which I think is what Bloomberg does.

I think frequency should be monthly. Period depends upon availability of the data. The longer the period the more accurate result you will get. 

Thank you, Eduardo Walker & Muhammad Usman Yusuf  for taking out time to answer my query.

I'll be using rolling beta values to find out  time varying betas. But, I was not sure whether it had to be rolling monthly values for a year or rolling yearly values for 3 years which traders use as I was given to understand.

Please bear in mind that beta is a regression of the volatility of stock relative to the market. Thus the further back in history one goes, less relevant data becomes, on estimating future performance of the stock in question. On the other hand, using shorter time frames during period of overall market volatility may underestimate or overestimate beta in some industries. Most IBs use monthly, 12-36 month data.

Thanks a lot for the additional insights.

I feel that if the objective of estimating beta is to make predictions about the future, then a 36 to 60 monthly data may be adequate. This period I think would  accommodate the most short and medium term required to generate a beta required for forecasting purpose. A longer period  may produce beta from Data not too relevant for the future  forecast and planning..

The previous colleagues have put the right perspective in mind. It isn't the objective that is important to consider the period. If you can share the perspective the answers will be more focussed. Databases which provide corporate information and mutual fund websites provide calculated betas. Kindly look into prof. J.R.Verma's website of iim Ahmedabad

apart from time period you should also see that there are no outliers in the data during the select period. you should also  make necessary adjustments to the returns for corporate actions like stock split, bonus shares, dividend declaration etc., 

CMIE Prowess  database provides calculated betas for all the listed companies india. 

Thanks a lot,  Kameshwar Rao Modekurti & Srikanth Potharla for your answers to my query. I also appreciate the suggestion regarding databases for calculated betas. But, I believe that when a statistic is calculated yourself, you have complete faith in the statistic & there is no chance of any error. When you calculate a statistic yourself, you are sure how it has been arrived at, the choice of the data period, the adjustments made to the data as pointed out by you, the outliers, etc. Hence, my effort to calculate it to ensure rigour.

Dear Ms. Hemalatha,

You are true that beta has to be calculated by oneself. In fact I suggest all my students when I give projects to do it themselves instead using the existing ones. I was responding to your querry since you asked for websites. Else beta has to be calculated.

Mr.  Srikanth's suggestion about the outlier treatment is usually more relevant in a cross sectional analysis. In the case of a single company time series analysis all information about a company has to be captured. If Mr. Srikanth meant abnormal returns and abnormal periods in which stock prices are considered then there would be an abnormal beta.

Any way please clarify the objective for which you are attempting the beta calculation? Fund analysis, Fama French Asset Pricing Modelling, Corporate Valuation, Asset Beta Calculation, Equity Valuation, Cost of Capital Calculation. Please appreciate each one requires a different kind of assumption and period and regression modelling and variable construct.

Dear Ms. Hemlata,

Just once again check your addressing of beta as a "statistic". It is usually referred to as a a "parameter" since it is estimated from a sample to comment about the population. Statistic is more related to a sample and is a single measure like Mean, Median, Correlation, Std. Deviation. Beta is a slope coefficient. 

Thanks once again for taking out time to provide a detailed response & initiation of a discussion on my query.

Regarding the points, you raised, my viewpoint:

I had enquired about the web sources for calculated values of beta but when I referred to some sources I was suggested by my colleagues, I realized that calculation is better to have greater understanding & transparency. 

You have suggested that beta is usually referred to as a a "parameter" since it is estimated from a sample to comment about the population. IHow can beta calculated for a short period of time for a few stocks chosen out of Nifty 50 stocks be representative of the entire population? For me, it is proportion of my sample that has high beta values & low beta values. Hence, my referring to it as my sample statistic.

Regarding the objective, I am unable to completely specify the same. However, it is not for the purposes you have mentioned above. It is related to something more practical - doing it as per the requirement of the assigning organisation. Involves back-testing of a strategy to think about the practicality of future adoption.

Your views & suggestions would be welcome.

Hemlata

Dear Ms. Hemlata,

Fine I understand the need for keeping a research objective secret out of the fear of espionage.

I would like to clarify that population according to me is not what you are assuming, - all companies listed in the stock market. According to me the population for a company's stock for beta calculation is the entire set of prices of the stock since its inception. We have chosen a sample period and hence you feel it has to be called a statistic. 

I still maintain due to the popular usage and references in econometrics and books, that the word "beta" as a statistic is only referred when they want to talk about the power of a test, in testing of hypothesis and the context of type I and type II errors. 

What you are referring to is Beta in the financial context which is a sensitivity of stock returns to market returns. In case you are not attempting any paper out of this exercise use of statistic or parameter would not have any impact. Else please re consider the usage.

All the purposes which I mentioned are also done by corporate in practice. Since faintly you have referred to back testing you have to definitely consider the period (whatever it is) in which you would like to back test. Say for instance you have calculated the beta for 2013 - 2017 and fit a model basing on some of your optimisation techniques, then your back testing of your investment strategy should also be during the same time.

Hi,

I never mentioned that my population is all the companies listed in the stock market.  It is what you have assumed when I said I am referring to a proportion of the stocks as high or low beta. For me, it is one of the statistic of our sample, out of the many others, estimated for a defined period of time.

Regarding the treatment of outliers by Mr. Shrikanth, you may have your own opinion, but investment industry definitely considers outliers, as standard procedure.

I am not detailing here whether we would or would not attempt a paper or any other form of write - up from this exercise.

Regarding the purposes which you mentioned were done by corporate in practice, after more than 2 decades with the industry, investment & finance world before academics, I would not enter into any argument about what is used by them in practice or not (Fama French, Carhart, et al.). Since the research staff are well versed with the industry practices, they merely wanted to know from me what academicians consider as an optimal period for beta calculations.

I'm not delving into any further comments about back-testing / optimisation, etc..here.

Dear Ms. Hemlata,

Thanks for the clarification. I understand your definition of statistic which is a proportion of companies not the beta.

I definitely would not be interested in further comments about your project. Please don't misunderstand.

I was genuinely trying to arrive at the appropriate period for calculation of beta after knowing the objective of the study. I am extremely sorry if I probed into the matter a little more than necessity. I showed my typical sincere teacher attitude in a research forum.

In case you and your team have not read the book written by the following professor, then you may consider.

Pablo Fernandez

Professor of Finance. IESE Business School, University of Navarra

It was a nice experience interacting with you

Thanks for your time & sincere effort to answer my query. There is no misunderstanding anywhere & you need not say sorry. This is a research forum for knowledge & discussion. What I’m stressing is that I’m unable to provide greater details on the project & answers to your repetitive questions about it, the terminology, etc.., which was deviating the discussion from the main question. Thanks again for all your goodwill & suggestions. It was nice to interact with you.