Selecting peers with as many similarities as possible is obvious. A few things to consider in your specific application might include some of the following:

• size of the firm (or maybe the percentage of the industry's output)
• private vs public (and maybe closely held)
• single product vs multi product --- or even conglomerate (i.e., multi-industry)
• a history of previous mergers or acquisitions
• historical stability of the leadership
• a dummy variable to indicate previous stock splits might be useful
• many others....

I hope helps and encourages others to offer suggestions. There are so many others that might help.

Rob

Thank you very much Respected Sir. Indeed your answer is too comprehensive..

I suggest you consider clusters in the industry, i.e. economic group (union) of various industrial companies in thematic clusters, with total financial goals. I recommend Michael Porter.                                                                                                                  Regards, Snezhinka

thank you Doctor. its very good and useful 

Size alone is not sufficient to classify firms within the same category. Firms with the same size can have dramatic differences in their capital structure leading one to consider a stock split while the other not. The key elements to consider to put a set of firms in a given category are: (1) Publicly trade firms or privately held firms. A publicly traded firm which is subject to the discipline of capital markets cannot be in the same category as a privately held firm no matter how similar they are in terms of assets, liabilities and products mix; (2) Firms in the same category must have similar asset structure; (3) same liability structure; (4) same business risk; (5) same financial risk; and finally (6) similar products mix, similar sales volume and similar sales growth rate.

I agree with most of suggestions you have already received, I would then only suggest to consider relative dimensions and correlation of cash flows with the state of the economy. As in MM theory this is a rather simple element to consider, but quite robust in clustering peer companies.

My inclination is to start the easy way and use a database like Factiva which lists peer companies for any company you choose.  I typed in Air New Zealand and Factiva gave me 10 other airlines.  It provides details of peer sales, number of employees etc. and these can provide useful control variables if the first testing of your hypothesis with Factiva peers isn't satisfactory.

My bet is that there will be a country effect and most premium will be in markets with highest levels of corruption.  Use corruption index as control variable.

Best of luck with this interesting idea.