When choosing bonds to invest in, it is critical to determine how much money you are willing to lose. Whether you're investing your own money or somebody else's, losing more than you expect can be a devastating blow to your lifestyle or career. Therefore, you should always be sure to use rigorous metrics to determine how much of a loss you might sustain. One common metric used by risk analysis is the "Value at Risk" or "VaR" of a portfolio--a measure of the amount of money likely to be lost on it during a particular period of time.

Determine the period you want to use for the VaR. The VaR is always calculated with respect to a particular period--usually one day, but sometimes a week or more--and it reflects the amount of loss that may occur in that time period.

Determine the confidence you want to use for the VaR. The confidence represents the percent chance that the loss over a given period will exceed the VaR. Usually, analysis use 5% or 1% for this figure.

Use your interest rate model to determine the 99th percentile (or 95th percentile if you chose 5% in Step 2) interest rate increase that could occur in your market within the period you chose in Step 1. If your interest rate model is bell-curve shaped (as most are), then this will be the value 2.33 standard deviations to the right of the mean (1.65 standard deviations if you chose 5% in Step 2).

Recalculate the bond value based on the increased interest rate from Step 3. Do this by multiplying the original price of the bond by the original interest rate and dividing by the new interest rate.

Subtract the new price from Step 4 from the original price of the bond. This amount is the Value at Risk of the bond, given the period and confidence chosen in Steps 1 and 2.

#### Warnings

Keep in mind that there is always a chance that your models will be wrong or that you could lose much more than predicted by the VaR calculation. There is always a chance--though it is usually very small--that the payer of the bond could go bankrupt and the bond could become valueless.