Term Structure Estimation in Low-Frequency Transaction Markets: A Kalman Filter Approach with Incomplete Panel-Data

There are two issues that are of central importance in term structure analysis. One is the modeling and estimation of the current term structure of spot rates. The second is the modeling and estimation of the dynamics of the term structure. These two issues have been addressed independently in the literature. The methods that have been proposed assume a sufficiently complete price data set and are generally implemented separately. However, when the methods are applied to markets with sparse bond price, results are unsatisfactory. We develop a method for jointly estimating the current term structure and its dynamics for markets with low-frequency transactions. We propose solving both issues by using a dynamic term structure model estimated from incomplete panel data. To achieve this, we modify the standard Kalman filter approach to deal with the missing-observation problem. In this way, we can use historic price data in a dynamic model to estimate the current term structure. With this approach we are able to obtain an estimate of the current term structure even for days with an arbitrary low number of price observations. The proposed methodology can be applied to a broad class of continuous-time term-structure models with any number of stochastic factors. To show the implementation of the approach, we estimate a three-factor generalized-Vasicek model using Chilean government bond price data. The approach, however, may be used in any market with low-frequency transactions, a common characteristic of many emerging markets.