Mortgage Loan Portfolio Optimization

Mads Gelting Østergaard Bech

Student thesis: Master thesis


In this thesis I investigate the possibility to improve current mortgage advisory. I find that a borrower of a 3,000,000 DKK loan, who had applied an ex-ante optimal strategy during the historical period 2010-2018, would have refinanced her loan eight times, and saved loan costs of 1,113,289 DKK relative to an otherwise identical borrower who didn’t refinance at all. Had she instead applied a strategy based on the rules of thumb commonly used by mortgage bank advisers, she would have only been able to achieve 48,349 DKK or about 5% of the potential savings, and this reveals that there is room for improvement in current mortgage advisory. I investigate the possibility to unlock a larger portion of the potential gains, by implementing stochastic programming models, which takes into into account the individual preferences of a borrower. More specifically i test a model proposed by Rasmussen, Madsen & Poulsen (2013)[26], in which decision making is based on 200 alternative scenarios of the future development in interest rates and bond prices, that are simulated in a stochastic framework. The objective of the strategies suggested by the model is to minimize a mix of average costs and risk across scenarios, in which the relative importance depends on the risk aversion level of the borrower. I find that a borrower who had applied a low risk strategy designed for risk-averse borrowers during the 2010-2018 period, would have refinanced her loan four times, and achieved loan costs savings of 294,387 DKK, by combining fixed rate- and adjustable rate loans in a portfolio. Even if she had been unwilling to originate adjustable loans, she would have refinanced five times, and achieved costs savings of 107,641 DKK, i.e. about twice as large as those of the rules of thumb strategy. As the excellent performance of this strategy might have been a consequence of the specific development in interest rates and bond prices in the historic scenario, results are tested out-of-sample by simulating 250 alternative sets of historic data, and repeating previous analyses for each of them. The low risk strategy still achieves the largest savings, and is moreover less risky than the rules of thumb strategy. These results imply a great opportunity to improve mortgage advisory, by introducing model-based optimization models

EducationsMSc in Finance and Investments, (Graduate Programme) Final Thesis
Publication date2018
Number of pages87
SupervisorsRolf Poulsen