Forecasting the Yield Curve

Dear Miguel,

I do not know of any GAUSS code implementing the yield code prediction methodology presented by Diebold and Li (2006). However, the simple example and procedure below will allow you to estimate the factor loadings β_1t, β_2t, and β_3t from yield curve data. The procedure is based on the least squares methodology outlined in section 3.2 of the paper cited in your original question.

//Data should be stored as a K x T matrix
//Each column represents a month and each row represents
//yield for a given maturity at that month
new;
cls;

//User specified lambda
lambda = .0609;

//Vector of observed maturities
TimeToMat = {3,6,9,24,36,60,84,120,240,360};


//Read data from .xls file
yields = xlsReadM("examples\\dailyTreasuryYields.xlsx","b2",1,"");
yields=yields';

//Call procedure to estimated betas
bta = findBetaDieboldLi(yields/100,lambda,TimeToMat);

//This defines the procedure which estimates the factor loading
proc(1) = findBetaDieboldLi(y,lambda,timeToMaturity);
   local x,bta,numMaturities,numMonths;

   numMaturities = rows(y);
   numMonths = cols(y);

   x = ones(numMaturities,1)~(1-exp(-lambda*timeToMaturity)./(lambda*time   ToMaturity))~((1-exp(-lambda*timeToMaturity)./(lambda*timeToMaturity))   -exp(-lambda*timeToMaturity));

   bta = y/x;

   retp(bta);
endp;