Order Description

Using cointegration and distance method to find 10 pairs, 5 from same industry and 5 from different industry, using data from time period 01/01/09 to 31/05/16.

My entire project is 5000 words in total, excluding references, tables and appendices using Harvard referencing. To use eviews for any testing and to add these to the appendix.

Trading and formation periods

Using 12 months as a formation period followed by 6 months trading period (as in Gatevs paper) to work out when to open/close positions, many papers will discuss this concept. The paper’Return and Risk Exposure in Pairs Trading Evidence from the German Equity Market’ (attachment called Fabian_berg_og_joerge) discussed the method of taking normalised prices (by taking the difference between current price and average price divided by standard deviation of price in formation period) and then to take the spread, section 5.2 discusses the pair formation in more detail section 6.2 discusses the trading period and section 6.4 discusses the trading rules followed by a graph showing the trading rules, I was hoping if this could be done please, Gatevs paper shows something similar, (section 2.2 Trading period). This is also discussed very well in the paper: The Evolution of Pairs Trading Lasse Ravn Lassen (called ‘main’ attached) under section 3.2.2 trading rules with a graph showing points of entering and exiting positions and finally an excellent paper ‘A new approach to Pairs Trading Using fundamental data to find optimal portfolios Author: Erik Jakobsson’ (attachment FULLTEXT01.), section 4 methodology.

To show ‘results of distance approach: Sum of Squared Deviations/differences’ To measure the co-movement between two stocks based on their normalised price series and to square them to make numbers positive.

I also found this I was not sure if it helps with anything: Calculating the Sum of Squared Deviations between two Normalized Price Series

Divide the time series of the share prices by the price on the first day of the pairs identification period and then subtract the two new time series from each other, square these differences and sum the result. The smaller the value, the more likely the two stocks are to be a good pair.

Returns (this section often follows the formation and trading period done in most literature)

To calculate returns, to check if returns are significant (pre and post transactions costs); checking to see if when taking transaction costs into account that returns are still positive. A good example is in the paper: Return and Risk Exposure in Pairs Trading Evidence from the German Equity Market (attachment called Fabian_berg_og_joerge) section 7.5 intra-industry pairs which is a very interesting section to read as this is similar to what I am aiming to look at. They also describe how to calculate excess returns and sharpe ratios.
The paper The Evolution of Pairs Trading Lasse Ravn Lassen (called ‘main’ attached)transaction costs section is very good.

Sharpe ratios
Additionally to work out the sharpe ratios please, this is also described well in the paper (refer to attached paper ‘main’) section 3.3.2.1 sharpe ratio.

Fama French 3 model:
Check returns on the different pairs and regress them on Fama French 3 Factor model to run regression (and to have factors picked).
Eg. Call return on pair i: rpairi, then regress:
rpairi= β0 + β1 (rm – rf) + β2 (SMB) + β3(HML) + β4 (DUMMY: same industry/no) + ε
if dummy significant: i.e if come from same industry or not; same=1 different= 0

Run regression on all pairs in one go and then to run it a second time (again on all pairs) with an addition of cointegration value. This is to check if the pairs cointegration makes a difference and/or if same or different industry pairs makes a difference, there may be a chance they are interlinked. For example it may become more positive as value of cointegration is more positive.

To check whether pairs from same vs. different industries yield a better return and control for levels of cointegration/correlation (depending on what you used to form pairs) between the variables. This is to possibly see if stocks from same industries perform better just because their prices are more correlated, not because they are from the same industry. After checking that the dummy variable for “same industry” is significant, to run a second regression adding also the variable cointegration or correlation. It may be that this decon variable will become the statistically significant one showing that the same/different industry actually does not matter.’

Also if alpha is required for FF3 model could this be added please if it is relevant.

The paper named thesis_schmidt attached is a very good paper looking at the cointegration approach in detail.

I also require raw data that may be used please.

I will attach the papers I have referred to above.