Financial issues: loans, mortgage, investment

Although the inflation-adjusted long-term distributions show that prices spend much more time at lower levels, there is a tendency to consider the average price of a market as the midpoint between the highest and lowest prices. Using the gold example, the midpoint for 1979 to 1993 was $451. The average of all monthly prices will give a reasonable approximation of a normal price; however, the best measure is the median, or middle, value when all monthly prices are sorted from high to low. The median value for gold over the same period was $381. If we look back at earlier posts we find that skewness is measured as the difference between the mean and the median, as a percentage of the standard deviation. In this case, the difference of $70 is a very large value, indicating a distribution with a peak far to the left of center. For price distributions, the median is a much more useful value than the average, although not as convenient to calculate. The median naturally adjusts for the skewness in the price patterns.

The data used to determine the standard deviation is very important. Because it is a statistical measure, it is most accurate when a large amount of data is applied. For example. you might find that 1 standard deviation of the crude oil daily price move is only SO.25 per barrel when measured over the past 10 years, but during the 6 months of the Gulf War the same measurement yielded $.50, twice as large.
Most trading applications using the standard deviation tend to apply short data intervals to its calculation, such as 20 days. This short period is not likely to represent the same price distribution as a 10-year calculation; therefore, the probabilities given by the resulting standard deviation value must be interpreted differently. While it is less likely that the price will make a move of 3 standard deviations compared with 1 standard deviation, the probabilities can be misleading. Statistics tell us that there is only a 1% chance that prices will move a distance of 3 standard deviations higher or lower; however. that value is reliable only when measured over a long data period. If you selected 20 days of unusually low volatility, the chance of a 3standard deviation move Would be very high.
The frequency distribution is another very practical approach to measuring price distributions. It has the advantage of having a much clearer visual interpretation. While the standard deviation gives us what appears to be a highly mathematical probability, the large error factor that is caused by small amounts of data may make its usefulness about the same as the frequency distribution.