5 Estimating Prepayments

5.1 The Danish Prepayment Data

Before we jump into the actual estimation of the model, we will have a look at the characteristics of the data used in the estimation procedure. In contrast to the mortgage market in the USA, the Danish market consists of rather few but very large mortgage pools. The large mortgage pools are a result of the demand for high liquidity in the underlying mortgage bonds. The Danish data are by law \(^{4}\) published through the exchange and contains information on the borrower composition (CK92), preliminary prepayments (CK93), cash flows (CK94) and the final prepayments (CK95). The data is therefore very detailed and in practice, it is often seen that prepayment models are extended to take all kind of aspects of the data into account. In this thesis we will only consider the CK95 data, but one could for example extent the model to discriminate borrowers by loan size using the CK92 data. Figure 9 gives an example of the CK95 data for a \(5 \% 30\) year callable mortgage bond issued by Nykredit Realkredit. From the figure it is clear how the prepayments increase during times of falling interest rates as the mortgage liabilities increase during these periods. We also see a tendency that some borrowers are more quickly prepaying compared to others. This tendency is consistent with borrowers having different costs associated with prepaying, which will imply that interest rates can stay in a falling trend while prepayments keep occurring. In the next subsection we will look into how these prepayment rates may be estimated.