Life can only be understood backwards; but it must be lived forwards. - Søren Kierkegaard

There are two main schools of thought with regard to risk analysis.

One method, probably the more common and definitely the more conventional is technical in nature, using volatility as a means of assessing short term risk. The other approach uses fundamental analysis and is nothing more than an application of value investment and seeking out true blue chip companies.

Statistical risk analysis should come under the banner of technical analysis, as it is based on finding statistical standard deviations of prices, or in other words, looking at how choppy the pricing of a stock is. Don't go calling it technical analysis to a traditionalist financial advisor or fund manager though, the same people that use statistical risk analysis are often the ones that sternly advise against the stupidity of charting and stress diversification and fundamental research above all else.

Just like all technical analysis techniques, statistical risk analysis is an examination of historical price data. The idea has merit as a very short term estimator of risk, as long as you define risk as the probability of getting your money back in a few weeks if you suddenly change your mind, assuming that the economy and the stock market are in an essentially steady state and nothing important is going to happen.

These techniques again are all derivative of the same thinking that led to the efficient market hypothesis. All companies are perfectly researched, stocks move in random walks, there is no reason why the market booms and busts, except as a random phenomenon that is absolutely unpredictable and has nothing to do with any of these silly cyclical theories.

The key to statistical risk management is to try to minimise volatility in a portfolio. Techniques have evolved that are highly effective at reducing volatility and the results of such management are steady and unspectacular gains in line with average growth in investments world-wide.

Stocks are analysed for their correlations. This is a statistical concept that determines which investments trend toward growth and yet perform the best at different times in the economy. For example gold stocks are generally considered high risk because of the volatility of their prices, and yet have a role in risk reduction of the overall portfolio because generally gold performs the best at times of economic uncertainty. While gold stocks won't necessarily perform well most of the time, when an international crisis such as a war or major catastrophe occurs investors will rush into gold as a currency. The result is a great appreciation in gold stock values that helps to smooth out the value of your portfolio when the rest takes a beating.

As always, Warren Buffett ridicules conventional thinking, insisting like he always does that the market is neither efficient nor random. Warren Buffett's approach seeks risk management by buying stocks when they are cheap, and investing in very well run companies with competent management capable of expanding the business in even difficult economic times.

Warren Buffett does not believe in the benefits of diversification, citing the words of the great economist (and legendary investor) John Maynard Keynes, who wrote in 1934: "It is a mistake to think one limits one's risk by spreading too much between enterprises about which one knows little and has no reason for special confidence.... One's knowledge and experience are definitely limited and there are seldom more than two or three enterprises at any given time in which I personally feel myself entitled to put full confidence."

Value and growth investors have no need for statistical risk analysis, such concepts only get in the way of researching a company to find a true gem. Indeed by the standards of statistical analysis many of these portfolios are considered very high risk.

The troubles with statistical analysis include the complete ignorance of fundamental valuations and economic cycles. Like many forms of technical analysis, statistical risk analysis usually instructs you to buy high and sell low, by showing little risk buying in a bull market and large volatility (hence large risk) after a fall.

Just before the crash of October 1987 statistical risk analysis would have said very little in the way of predicting a fall, in fact the whole rationale of the technique simply would have urged you to invest in whatever would make the returns more average. A couple of weeks after the fall, which of course value investors well know was the absolute best time to buy stocks, the crash would have registered as a big jump in volatility and hence would have advised that risk has increased, which would lead to attempts to reduce the holdings in shares to get into something less volatile, such as bonds. So value investors would have ploughed money into the stock market to take advantage of the lowest prices that they may ever encounter, while portfolio theorists would have been getting out in large numbers.

The other problem with using historical prices is well known to many who try to generate predictive systems. Analysts find counter-trending investments by computer analysis, historic trends in correlations are valid only in history. Who is to say that the unique combination of economic and company-specific factors that led to the calculated correlations will lead to the same correlations in the future? Currency movements may well have an opposite effect on importers vs exporters, however this ignores the influence of tariffs and international trade agreements that can have a big effect on one or the other, or the same effect on both. A historic correlation between the two is meaningless without good estimations of the future factors that will influence this correlation.

In order to play the risk-reward efficient frontiers and tally growth to risk, risk analysts look at another technical number related to the volatility. Called a beta value, the number beta indicates the relative movement of a portfolio with respect to the market it represents. A portfolio with a beta of 2 will quadruple in price when the market doubles, and fall by a half when the market falls by a quarter. A high beta portfolio is not entirely different to a leveraged portfolio of low beta stocks. The trouble with beta values is the same as other forms of risk analysis of this type, they rely exclusively on historical prices and have no real predictive value.

Statistical risk analysis does have its place, as a short term measure of the likelihood of getting your money back if you suddenly want to pull out of the market they are definitely valuable. This is why these techniques are taught even today to financial advisers and stockbrokers. In the short term volatility does make a difference, it is perfectly possible and possibly reasonable to give a numerical probability of getting a certain return in a certain time, which is important advice to a client.

You have to invest differently when you are dealing with someone else's money. Only a very small number of managed funds have been genuine focus funds, they include the Sequoia Fund and the Legg Mason Focus Trust, both highly successful investment vehicles. Although over time their performance has beaten the market, by a substantial margin, volatility has been higher than that of conventional funds and would have provided a very worrying ride for unsophisticated investors who nervously watched their fortunes waxing and waning. On the other hand, however, in almost every year except the one just passed Warren Buffett's Berkshire Hathaway company has produced returns that are both higher, and lower in volatility than the general market, even though the explicit reduction in volatility has little to do with Buffett's approach.