Process

• The investment performance is charted using the investor's acceptable return on investment as the mean.NA/AbleStock.com/Getty Images
  
  When measuring semivariance, the investor must set a target level for return on his investment that he feels is acceptable. The dispersion of all the observable returns below that target level is measured according to the percentage of decline from the target level when calculating the semivariance.
  
  

Formula

• The target level of return is used as the mean when calculating semivariance. Semivariance equals the average of the square of the deviations of returns below the mean. Semivariance investing is based on the premise that the lower the semivariance, the lower the risk that a substantial loss will occur.
  
  

Purpose

• When making investment decisions, there is an attempt to quantify uncertainties in terms of risk based on probabilities. Investors seek to formulize uncertainties using statistical theory that claims a future outcome will be within the average of past outcomes. Semivariance is used to measure the average decline an investment has realized in the past in order to predict the potential downside risk of the investment in the future. Semivariance is similar to variance, except that semivariance only considers negative fluctuations because it is only concerned with projecting the potential downside, not the potential upside.
  
  

Semivariance and Semideviation

• Semivariance and semideviation both attempt to calibrate the potential downside risk of an investment below an investor's acceptable level of return. However semideviation is simply the average of the deviations below the mean. If the semivariance is known, the semideviation can be determined by taking the square root of the semivariance.
  
  

Risk

• Probability is used to quantify the outcome of a single event. Strict semivariance investing assumes that return on investment (ROI) is based on normal distribution according to statistical models. This is often the case with financial data formulated over a relatively short period. However, underlying uncertainty may be asymmetric, so semivariance models should be adjusted by the "degree of surprise," a calculation of the variance between the expected return and the actual return -- that is, each past return used in the calculation must be individually examined to determine whether it falls within the expected semivariance calculation. This is done by removing one return at a time and computing the semivariance. The result is the amount that would be expected for that one return, given the formula of calculation. If the actual amount of that return varies from the expected amount, the percentage of difference is equal to the degree of surprise for that return. Take the average of all the degrees of surprise to determine the overall degree of surprise used to adjust the overall semivariance.