Page **1** of **1**

### Cointegration: allocation of weight

Posted: **Sat Aug 25, 2012 6:43 am**

by **Abhinav Sisodia**

Hi Professor,

I have read your papers. They are very useful for me as i am working on cointegration optimization process. Thanks for uploading them mam.

Professor, I am facing problem in weight allocation with the help of regression. Will you please tell me how can we allocate weights in index tracking and second problem is which one is best Johansen test or Engle & Granger test for testing cointegration relationship between two or more series.

Thanks in advance professor. I am waiting for your reply.

### Re: Cointegration: allocation of weight

Posted: **Sun Aug 26, 2012 2:26 pm**

by **coalexander**

Hi

I recommend you read Chapter 5 on cointegration in Vol II, where on page 239 the merits of Engle-Granger versus Johansen methods are discussed, and in the next 5 pages the allocation to weights for index tracking is described with an empirical case study.

Best wishes, Carol

### Re: Cointegration: allocation of weight

Posted: **Sun Aug 26, 2012 2:46 pm**

by **Abhinav Sisodia**

Thank you very much professor.

But will you please help me how can i read your Vol. 2.

### Re: Cointegration: allocation of weight

Posted: **Sun Aug 26, 2012 2:54 pm**

by **coalexander**

I have emailed you the relevant pages

### Re: Cointegration: allocation of weight

Posted: **Mon Aug 27, 2012 10:10 am**

by **Abhinav Sisodia**

Once again thanks and i have gotten your mail.

### Re: Cointegration: allocation of weight

Posted: **Tue Aug 28, 2012 9:20 am**

by **Abhinav Sisodia**

Professor, will you please explain the meaning of "normalize the coefficient estimates so that they sum to 1". I am not getting what does it mean. How can we normalize the coefficient is there any method or what?

Sorry, for asking this stupid question. Please help me mam.

### Re: Cointegration: allocation of weight

Posted: **Tue Aug 28, 2012 6:01 pm**

by **coalexander**

When the coefficients are estimated they do not sum to 1. So change the coefficients to be coeff/sum of coefficients, so these normalised coefficients add up to 1.

### Re: Cointegration: allocation of weight

Posted: **Wed Aug 29, 2012 6:10 am**

by **Abhinav Sisodia**

I get it professor. Thank you very much.

### Re: Cointegration: allocation of weight

Posted: **Sat Sep 01, 2012 8:00 am**

by **Abhinav Sisodia**

Hello Professor,

Thanks for helping me and solving all my doubts. Professor, If am wrong please correct me. For finding cointegration between two things we run E&G test. In E&G test first we perform the cointegration regression and save the residuals of that regression and at last run auxiliary regression. Mam, Please will you explain me last step that is auxiliary regression, in this we regress residuals with whom independent variable i.e, stocks or with dependent variable which is our Benchmark.

Thanks in advance.

### Re: Cointegration: allocation of weight

Posted: **Sat Sep 01, 2012 11:34 am**

by **coalexander**

It is the Dicky-Fuller regression which tests whether the residuals are stationary, i.e. I(0). If they are, then we deem the variables to be cointegrated in that Y - beta X is I(0). All this is very carefully explained in Chapter 5 of Volume II, so I suggest you read that. Carol