1/2/13 Gmail - RE: Cointegration and Unit Roots

1/5 https://mail.google.com/mail/ca/u/0/?ui=2&ik=37033dcd3a&v iew=pt&search=inbox&th=12f 2cb148a46f 6f 3

Boris Vaisman <boris.vaisman@gmail.com>
RE: Cointegration and Unit Roots
2 messages
Tallis, Hans C - GCIB NY <hans.tallis@baml.com> Fri, Mar 4, 2011 at 10:26 AM
To: "Vaisman, Boris" <bvaisman12@gsb.columbia.edu>
Cc: "Ballingall, Jamie" <jamie.ballingall@credit-suisse.com>
Boris,

The problem with nonstationarity is depicted fairly well in the attached xls.
1. On Trends sheet, set B5 and B8 to 0
2. F9 a few times, observing the nave regression coefficient in H15. The series X and Y are random, and have
no correlation.
3. Now make both series cointegrated on a third variable. (here, the variable is time): set B5 and B8 to 1. The
regression coefficient is significantly positive now.

The problem is if we use the latter regression model to infer causality. If we shock X upward, we shouldnt
expect any change in Y, but our nave regression model suggests that Y will increase as well. This is a spurious
regression.

This is an example of an omitted variable bias. By including time as a regressor, we can fix the problem. See
the regression coefficient in J15.

In financial series youll often see cointegration from..
Prices that use an inflating currency
Organic quantity growth
seasonality

If you work in log space for your variables, back out seasonality and linear trend, you remove the most common
sources of spurious regression in time series analysis.

--Hans
o: +1-646-855-4707

1/2/13 Gmail - RE: Cointegration and Unit Roots
https://mail.google.com/mail/ca/u/0/?ui=2&ik=37033dcd3a&v iew=pt&search=inbox&th=12f 2cb148a46f 6f 3
From: Vaisman, Boris [mailto:bvaisman12@gsb.columbia.edu]
Sent: Friday, March 04, 2011 9:48 AM
To: jb3240@columbia.edu
Cc: ht2109@columbia.edu
Subject: Cointegration and Unit Roots

Dear Professor Ballingall,

Per our discussion after class I have included some references to cointegation.

I have a very rudimentary understanding of the concept. If the time series is non-stationary you could get spurious
regression. So, need to test for unit roots and if the test is positive the series is cointegrated. Unfortunately, I do
not understand why you run into issues, what are unit roots, what are unit root test, and how cointegration solves
the problem.

I would be very grateful if we could go through cointegation in a review session.

Lecture Notes:
Regression fallacy: http://www.econ.uiuc.edu/~roger/courses/472/lectures/L8.pdf
Introduction to non-stationary time series: http://www.econ.uiuc.edu/~roger/courses/472/lectures/L9.pdf

Cointegration Primers:
http://www-stat.wharton.upenn.edu/~steele/Courses/434/434Context/Co-integration/Murray93DrunkAndDog.pdf
http://www-stat.wharton.upenn.edu/~steele/Courses/434F2005/Context/Co-integration/Primer_Mar_Apr1991.pdf
http://quanttrader.info/public/testForCoint.html

Best regards,
Boris


-----------------------------------------------------
Boris Vaisman
Columbia Business School
MBA 2012
E-mail: BVaisman12@gsb.columbia.edu
Mobile: 860-944-3935
http://www.linkedin.com/in/borisvaisman
1/2/13 Gmail - RE: Cointegration and Unit Roots
3/5 https://mail.google.com/mail/ca/u/0/?ui=2&ik=37033dcd3a&v iew=pt&search=inbox&th=12f 2cb148a46f 6f 3

This message w/attachments (message) is intended solely for the use of the intended recipient(s) and may
contain information that is privileged, confidential or proprietary. If you are not an intended recipient, please notify
the sender, and then please delete and destroy all copies and attachments, and be advised that any review or
dissemination of, or the taking of any action in reliance on, the information contained in or attached to this
message is prohibited.
Unless specifically indicated, this message is not an offer to sell or a solicitation of any investment products or
other financial product or service, an official confirmation of any transaction, or an official statement of Sender.
Subject to applicable law, Sender may intercept, monitor, review and retain e-communications (EC) traveling
through its networks/systems and may produce any such EC to regulators, law enforcement, in litigation and as
required by law.
The laws of the country of each sender/recipient may impact the handling of EC, and EC may be archived,
supervised and produced in countries other than the country in which you are located. This message cannot be
guaranteed to be secure or free of errors or viruses.
References to "Sender" are references to any subsidiary of Bank of America Corporation. Securities and
Insurance Products: * Are Not FDIC Insured * Are Not Bank Guaranteed * May Lose Value * Are Not a Bank
Deposit * Are Not a Condition to Any Banking Service or Activity * Are Not Insured by Any Federal Government
Agency. Attachments that are part of this EC may have additional important disclosures and disclaimers, which
you should read. This message is subject to terms available at the following link:
http://www.bankofamerica.com/emaildisclaimer. By messaging with Sender you consent to the foregoing.
Spurious.xls
243K
Ballingall, Jamie <jamie.ballingall@credit-suisse.com> Wed, Apr 6, 2011 at 5:07 PM
To: "Vaisman, Boris" <bvaisman12@gsb.columbia.edu>
Cc: "ht2109@columbia.edu" <ht2109@columbia.edu>
Boris,

Im conscious that I never gave a completely satisfactory answer on the cointegration question. To help remedy
that, let me add:

1. The day-to-day approach to dealing with cointegration is to find a process that is stationary and mean-
reverting. This is often a difference between two processes. For example, if we were attempting to model WTI and
Brent, we would probably take one (say Brent) as the base process and model the WTI minus Brent process.
That is likely to give a highly mean reverting process, which we can model in the usual way. After the simulation,
we can unwind this trick and get back simulations for Brent and WTI, which are then cointegrated.

2. If we have contextual information, like above, we are best off choosing the differences to take manually
because we can conform the model to economic reality. But if there isnt an obvious base process we can use
Principal Component Analysis to do some of the work for us (see attached or http://www.mathworks.com/help/
toolbox/stats/brkgqnt.html#f75476 or a quick web search. The Wikipedia article is a bit heavy on the technical
details in my view).

3. You might want to consider joining the MathWorks Webinar on cointegration (see attached email). They are
1/2/13
4/5 https://mail.google.com/mail/ca/u/0/?ui=2&ik=37033dcd3a&v iew=pt&search=inbox&th=12f 2cb148a46f 6f 3
trying to sell you a MATLAB license, but if you ignore that its free and should be fairly accessible. Im going to
try to dial in myself.

Jamie
[Quoted text hi dden]
==============================================================================
Please access the attached hyperlink for an important electronic communications disclaimer:
http://www.credit-suisse.com/legal/en/disclaimer_email_ib.html
==============================================================================
---------- Forwarded message ----------
From: MathWorks <webinars@em.mathworks.com>
To: "Ballingall, Jamie" <jamie.ballingall@credit-suisse.com>
Cc:
Date: Tue, 5 Apr 2011 09:41:54 -0400
Subject: Explore Cointegration for Forecasting and Trading with MathWorks
To view this email as a web page, go here.
To ensure our e-mails reach your inbox, add the domain @em.mathworks.com to your saf e sender list. Tell me more.
Jamie Ballingall
Join a new MathWorks webinar:
Cointegration and Pairs Trading with Econometrics Toolbox
14 Apr 2011
9:00 AM EDT
2:00 PM EDT
Do you need to create better time-series models and forecasts?
In this webinar we introduce new capabilities for the R2011a release of Econometrics Toolbox. Through
product demonstrations, attendees will see how the Econometrics Toolbox is used for:
Cointegration testing
Vector-error-correcting (VEC) and vector-autoregressive (VAR) modeling
Pairs trading strategy development
9:00 AM EDT

Register for this webinar.
Gmail - RE: Cointegration and Unit Roots
5/5 https://mail.google.com/mail/ca/u/0/?ui=2&ik=37033dcd3a&v iew=pt&search=inbox&th=12f 2cb148a46f 6f 3
2:00 PM EDT

Register for this webinar.
We look forward to seeing you and your colleagues online on April 14. Please contact us with any
questions.
Sincerely,
Stuart Kozola
MathWorks
508-647-7000
webinars@mathworks.com
Sincerely,
William Mueller
MathWorks
508-647-7000
webinars@mathworks.com
To update your address, send e-mail to service@mathworks.com
2011 The MathWorks, Inc.
MATLAB and Simulink are registered trademarks of The MathWorks, Inc. See
www.mathworks.com/trademarks for a list of additional trademarks. Other product or brand names may be
trademarks or registered trademarks of their respective holders.
OPT-OUT Click the link to remove this e-mail address from MathWorks commercial electronic mailing list.
The MathWorks, Inc. - 3 Apple Hill Drive, Natick, MA 01760 - 508-647-7000
principal_components.pdf
118K