Authors
Advisor(s)
Abstract(s)
Like any other asset, it is important for investors to value accurately their mortgage portfolio. This
valuation will not be precise without available information on the underlying cash flows. To evaluate
mortgages properly, they need to have reliable information not only on the scheduled payments, but
also on all payments not expected to occur over the duration of the mortgage, either partial
prepayments (curtailments) or total prepayments.
Currently, in Portugal, the information used is based on international studies. It was not possible to
find evidence about Portuguese prepayment studies. The experience of working with mortgages
indicated that levels of prepayments referred in most literature (about 6% of the outstanding
principal) would be lower in Portugal.
The information about prepayments is relevant to several intervenients in mortgages: financial
institutions that lend and need to assess their portfolio; lenders who use scoring systems / pricing
and need to incorporate information from expected levels of prepayment; investors in Mortgage
Backed Securities (MBS) and rating agencies that evaluate institutions and portfolios.
This work involved a first major challenge of getting data (in significant numbers) from the financial
institutions about total and partial prepayments in order to have a reliable data base. This
information is not easily available and this goal was achieved only with the support of one of the
largest credit institutions in the country (which prefers to remain anonymous due to confidentiality
reasons). After clarification meetings about the study and its relevance, COO and Administration
became interested in the study, approving data delivery. This institution provided prepayments data
from the year 2000 until 2012.
This allowed us to gather about 747.000 records of information. Only knowledge about mortgages,
about the institution, its database and the extraordinary collaboration of this institution’s staff
enabled all data treatment and validation.
Historical prepayments and seasoned prepayments were calculated as a percentage of the total
prepayments in each period (month or maturity) over the total mortgage amount. This analysis
revealed low levels of prepayment. The average prepayment level of 1,21% was well below the
usually referred 6% level.
Then, the data was associated with some macro-economic variables (mainly factors referred to in
literature) in order to characterize how prepayments were correlated to those indicators. This
analysis allowed us to understand which variables were most strongly correlated with prepayments.
During the analysis period a Decree-Law (DL) was enacted that limited the level of penalties on
prepayments. We assayed the DL’s effect at the prepayments level.
In order to get a predictive tool, we regressed prepayments as a function of most significant variables
and formulated a predictive model, with a 12 months lag. This model revealed to have a considerable
level of significance.