Methodology

Each Bond Type Curve is created with a consistent design and calculation methodology that enables meaningful comparison to each other and over time. That methodology includes the following steps:

1. The Bond Types that comprise each individual curve are defined. The curves are grouped into general categories: quality grades, source of payment, state interest taxation, bank qualification, and specific indices for California bonds.
   
   
2. We identify trades that match the specific Bond Type attributes associated with each curve. MSRB trade records include a field that indicates whether the trade was dealer-to-dealer, sale-to-customer, or purchase-from-customer. We use all three types of transactions in our calculation. We exclude trades for securities with zero coupons as they represent a fundamentally different type of debt investment than bonds which pay interest over the life of the security.
   
   
3. We then group the trades by “years to effective maturity”, which is either the stated maturity or the date of an announced redemption.
   
   
4. We then employ a par weighting formula for both the yield and the effective maturity. From this we produce a scatter plot graph and then calculate a “best fit curve”*. We then add all trades to the original scatter plot graph. For each trade we calculate the deviation from the “best fit curve”. For those trades that are greater than +/- 2.58 standard deviations (a factor that our analysis shows to be statistically valid) we exclude from consideration as outliers.
   
   
5. With the remaining trades we recalculate the “best fit curve” following the same methodology employed with the original group of trades described in step two. This results in a curve derived from par weighted trade data using commonly accepted statistical methodology.
   

As a final step we identify the trade “metrics” that went into calculation of each curve- the number of trades, par value traded, and the “R-squared” value** for the best fit curve.

Taken as a whole, we believe that this transparent methodology, visual display of the raw data and the calculated results provide a highly meaningful level of information and context that allows a user unprecedented insight into actual yield and price levels as determined by contemporaneous trades.

*	Best Fit Curve- we employ a third order polynomial equation to determine a graphical curve that represents the individual data points of a scatter plot graph accurately.
**	R-squared value- also known as the “coefficient of determination”, is a statistic which indicates the strength of fit between two variables implied by a particular value of the sample correlation coefficient r. The highest possible value would be “1”, indicating a perfect positive correlation.