Seeking Beta - Calculating an Accurate Beta

I am following Aswad Damodaran's Valuation Course online. Mr Damodaran very generously publishes all of his lectures online with supporting materials. His lectures are clear, logical and very interesting. My only complaint is that the sound levels in some of the lectures are a little low which can make them hard to follow on my iPod. Fortunately I can throw technology at the problem and the tiny freestanding speakers at home makes it all possible.

I have just been looking at Betas. Betas are intended to give you an indication of the market risk that your company is exposed to (that is the risk that can not be diversified away from). The basic idea is that you can get a Beta for your company by regression of the stock price over the index as a whole. The problem with this is that what you will calculate will typically contain a huge amount of error - simply because the amount of data to get the error down to a reasonable level is unattainable. Additionally you could want a beta for an unlisted company - here you have no data to look at.

The steps from this are really logical - if you can't get a Beta for your company then the best thing that you could do is get a Beta of a company just like it. Of course to level the playing field you want an unlevered-Beta - that is you want to take away the effect of leverage. That way you can look at like with like. Of course doing this you are stuck with a small sample size, and the possible addition error of the company being different.

The next step is to say, well, what if there are a whole load of similar companies we could use all of those and average them. This means that with care you can get a Beta that is representative of the market into which you sell into by looking at all the players in that market. You have two choices when calculating this - you can either look at the companies that sell in that market, or in some cases, the companies that buy from that market. Your aim is to find as large a representative sample of companies as possible from which to calculate your Beta as this will give you the most accurate Beta as the larger your sample the smaller your statistical error. Calculating a Beta this way can give you a far more accurate Beta than than using the historic share price of your target company. Why? its a far larger sample.

At the end of the day what is Beta? Beta is a measure of the elasticity of your market. Low beta - inelastic market (essentials), high beta elastic market (discretionary purchases). What you are trying to do is calculate the risk that your company faces from the elasticity of the market it sells into. Or so, at least, is my understanding of things.