Home Page

Products

Papers

Contacts

About MFC

Links

  Economic Scenario Generator

The Economic Scenario Generator was developed to provide features and benefits critical to the simulation of yield curve scenarios for asset liability simulation.

The Mathematical Finance Company (MFC) Economic Scenario Generator™ (ESG™) is a software program for producing scenarios of yield curves, equity indices and dividend yields. It is based on the Double Mean Reverting Process™ (DMRP™).

 
The model allows for multiple equity-like indices. These can be used to model multiple equity portfolios, but also asset classes such as real estate or a foreign asset portfolio that is modeled on a return basis. By including a dividend series for these types of asset classes the system can produce both an income and total return value. The equity like indices can be correlated to each other and to the interest rate variables of the DMRP model.

Interest Rate Model Description

The DMRP interest rate model is a 2 factor interest rate model for the stochastic process on the yield curve. The DMRP models the logarithm of the short term rate directly. It models this logarithm as being subject to a random shock but headed towards a target rate. This target rate is itself moving towards a long term target rate, but is also subject to a random shock.

The random shock on the target rate and on the logarithm of the short term rate are correlated. The parameters of the model are the volatilities of the two shocks, the correlation of the two shocks, the two rates of mean reversion and the ultimate target rate that the target rate is headed towards.

The DMRP model has been calibrated to historical yield curves in both the U.S. economy and to the Canadian, Japanese, and Swiss economies. The model has been found to be robust over these multiple economies without dramatic alterations in its parameter values.

The underlying logic of these values is that the rate of mean reversion of the logarithm of the short term rate is high towards the target rate, but the rate of mean reversion of the target rate to the ultimate target rate is low. Because of this, the target rate determines a trading range or a regime of the economy.

The regime changes slowly towards the long run tendency of the economy. Thus long episodes of high, medium or low interest rates can occur in the model. The volatility of the target rate is lower than that of the logarithm of the short term interest rate so that a regime can persist. However, within a regime there can still be variation in rates.

The yield curve is determined at any point in time by the so-called risk neutral values of the parameters. In the risk neutral process, the target rate is higher than in the normal, or real probability process. Also the ultimate target rate is higher.

These higher values reflect the presence of risk aversion against higher rates. The risk neutral process incorporates risk aversion by making higher rates more likely to occur. In addition, the rate of mean reversion of the logarithm of the short term rate towards the target rate is higher in the risk neutral process, while the rate of mean reversion of the target rate towards the ultimate target rate is lower. This reflects the historical shapes of yield curves observed.

The higher rate of mean reversion of the logarithm of the short term rate to the target rate means that the target rate has a bigger impact on the shape of the yield curve. This allows the yield curve to be more steeply slowed. The lower rate of mean reversion of the target rate to the ultimate target rate results in more allowed variation in the value of long term yields. The two volatility and the correlation parameters are the same from the risk neutral to the real probability.

By adjusting the parameter values, the different major industrial economies can be modeled. For example, Swiss interest rates have had a smaller variation in range historically than those in the U.S.. This fact can be accommodated by primarily adjusting the rates of mean reversion.

In particular, the rate of mean reversion of the target rate to the ultimate target rate is made higher. This keeps the range of rates more compact. Because interest rates are lower in the model, the absolute volatility automatically becomes lower with the lower absolute level of rates. This is because the logarithm of the interest rate is modeled, not the absolute level itself. Thus a given numerical value of the random shock results in a smaller change in interest rates when the absolute level of rates is smaller.

Similar adjustments allow the modeling of the Canadian, Japanese and UK interest rate history. In addition, historical sub-periods can be modeled as well. For example, the last 10 years have been a period of lower volatility. This primarily shows up in a lower volatility for the shock to the logarithm of the short term interest rate.

Inflation Modeling

The rate of inflation is closely related to the level and shape of the yield curve. When the 3 month yield is high, then inflation is high. But also when the yield curve is flat to inverted, inflation is high. The generator can be combined with a simple model of inflation in terms of the yield curve plus an additional mean reverting deviation from this implied inflation rate to generate inflation as well with the other variables of the model. The system as currently configured generates this inflation residual as well as supplying the model that goes from the yield curve variables to the inflation rate.

Equity Indices Modeling

Regime switching on the parameters is an option for equity index modeling. The software allows for multiple equity indices to be modeled. These can be correlated to the two stochastic variables, the logarithm of the short rate, and the target rate that are the drivers of the interest rate model.

The equity index can be interpreted as a price index, such as the S & P 500, or as a total return index. The dividend yield is the dividend yield on the price index. When treating the index as a total return index, the dividend yield is still available, but to compute the price index, the total return series and the dividend series must be used to calculate the price index outside the system.

Averaging of Equity Index Module

An additional module to the system allows averaging of equity indices as is necessary for equity indexed annuities (EIA) or equity indexed products (EIP). This module is not supported by a graphical user interface but is available through a text file interface.

What the system produces

The system produces two scenario files. The merged yield file contains scenarios of the two state variables of the interest rate model and yield curves. The other file contains the two state variables of the interest rate model, the equity indices and the dividend yield. These two files are ASCII files with spaces between the variables. They are in the following format.
scenario index, time index, short rate, target rate, y_1, y_2.,...,y_n
where y_1 to y_n are the yields of bonds in the merged yield file, or are the value of equity indices and the associated dividend yields in the other file.
To see an example of the format of the output files, click these links: Merged yield file and Equity indices and dividend yield.

User-provided input

The user provides the following input:
  • Points on the yield curve that are simulated and fit.
  • Starting value of the treasury yield curve for these points.
  • Number of scenarios.
  • Time interval in scenarios.
These values are entered in a screen interface. The user then launches the calculation from the interface after having selected these values.

In addition, the user can select different parameter files from the interface corresponding to different parameter sets for the interest rate model parameters. Examples are the real process parameters, the risk neutral and the parameters for different economies or different sub-periods. Parameters for different economies must be purchased separately at a modest fee above the base fee for the system.

Parameter Sets Available: U.S., U.S. 1950's and 1960's, US Low Interest, US Low Volatility, Canadian, Japanese, Swiss and UK. The Economic Scenario Generator is supported by MFC through annual upgrades of both the software and the licensed parameter sets.