# Help with dummy variables

10 stocks, starting Jan 2004 through Dec 2008 (60 monthly observations) info is availabe in Excel. Assignment - Using a dummy variable approach, test whether the stock market up to 2007, behaves differently from the stock market from 2007-2008. That is, test whether there is any shift in the alphas and betas over time. Interpret your results, and plot it. ok - So I understand that we will be needing only 1 dummy variable as a flag to identify the world before 2007 and the world between 2007-2008. What I don’t understand is how do I test if there is any shift in alpha and beta over time?

I don’t think that you can, which is one of the issues that they mention of regression analysis

Part b of the assignment b. Using Excel (or any other regression program) run a regression of each investment’s monthly return against the S&P Index (return) and estimate alpah, beta, R-Squared. Check which coefficients (and regressions) are significantly different from zero, at 5% significance levels. Plot the regression lines. Interpret your results. So after adding dummy variables - we need to see the drift in alphas and betas over time. Any help?

y=a + b*dummy+c(x)+d(x*dummy)+e Test for significance of b and d. If different from zero, there is a difference in alpha and beta.

I did some regression of IBM stock with S&P 500 and got thr following for the best-fit line y = 0.0505x + 30.778 R2 = 0.2381 So beta(IBM) = 0.0505 alpha(IBM) = \$30.778 23.81% of variation in IBM stock is explained by the S&P movement. I think I have completely messed up with the excel, calculation - because, do you think beta for IBM can be as low as 0.0505 Any help appreciated?

wyantjs Wrote: ------------------------------------------------------- > y=a + b*dummy+c(x)+d(x*dummy)+e > > Test for significance of b and d. If different > from zero, there is a difference in alpha and > beta. Do this, Then test the single IV model and compare the R-Squares.

Thanks DanLieb & wyantjs - that dummy variable part comes later in the assignment. I am stuck at the initial step itself (see my post at 02:52PM). I highly doubt that beta for IBM is 0.0505, that means I have messed up my excel (which I am not at all familiar with). I am not from the field and feel bad for asking such saddy-saddy question here. I’ll mail the spreadsheet if anyone wants to look at my mess.

equity book gives IBM Beta as 1.72 in some examples… e.g. pg. 133 maybe you are not measuring these in the same units… so you have an issue because of that.

Are they testing for structural breaks post 2007? If so you should be doing the Chow test. If you don’t mind can you send me an email with detaisl and the data and I see what I can do for you.

Ignore what I just said. I didnt realise the part of dummy variables. Dummy variable method can also be used to test for structural breaks. However my offer is still valid my address is yukiarashi1984@yahoo.com. Incidentally is it a must for you to use excel? Can other regression software be used? I hate the graphs that excel churns out. Bloody ugly.

Thx charu and cpk - I am way off with beta calculation then. Rainbow - I have sent all the details needed, the 5-part assignment and the IBM spreadsheet I created.

swaption i did the initial stages of the qtn. will do the rest later. i sent u an email further idscussing the issue

HydrogenRainbow Wrote: ------------------------------------------------------- > Ignore what I just said. I didnt realise the part > of dummy variables. Dummy variable method can also > be used to test for structural breaks. However my > offer is still valid my address is > yukiarashi1984@yahoo.com. > Incidentally is it a must for you to use excel? > Can other regression software be used? I hate the > graphs that excel churns out. Bloody ugly. I actually agree with you on the Chow test. I only suggest the method above because he wanted to use dummy variables, and this would allow him to distinquish between changes in alpha OR beta. The Chow test would only allow him to tell if they had collectively changed. Both methods are applicable.

Am wondering if anyone has thoughts on this. Since the objective is to calculate both alpha and beta of IBM, the appropriate regression to run will be: R(IBM)-Rf= alpha + beta(market risk premium). I did the regression and the beta I got is 1.10 In one part of the quesiton swap gave me, the following was asked. :Breakdown the total risk into systematic and non-systematic risks, and explain your results. -I don’t think this question makes sense because beta is the systematic risk ?? Idiosyncratic risk has been diversified away, no?

It would be defined as the variance of the residuals.

Makes sense. Thanx!

Hey HydrogenRainbow - thanks a billion. I was not feeling well and had dozed off. I saw a bunch of emails now and will go through them. I don’t need the Chow test as I only need to use dummy variables to see drifts in alpha or beta in the 2 zones (Zone1: 2004-2007 and Zone2: 2007-2008). No test for structural breaks is needed. If Idiosyncratic risk is a sohpisticated term for unsystematic risk, then to answer your question rainbow, It will get diversified away and all that remains is the beta (also the systematic risk) and that’s what the investor gets the compensation for. No pity for holding on to the unsystematic risk. Thanks for helping out.