Real estate investments are characterized by heteroskedastic variance, liquidity risk and high transaction costs and, thus separating them from the assumption of independent and identically distributed data and the efficient market hypothesis. The thesis investigates the pros, cons and differences between the classic mean variance portfolio optimization theory by Harry Markowitz and the more recent theory by Cheng, Lin & Liu in order to determine the validity of the underlying assumptions as well as estimate an optimal real estate allocation in a mixed-asset portfolio. The empirical data represents the Danish market from 1990 to 2014. The thesis uses a neo positivistic perspective in order to generate objective and unbiased results, create new knowledge and add value to the theoretical field of real estate investments. By applying desmoothing theory by David Geltner the thesis discovers appraisal bias to cause a significant risk reduction in the periodic real estate return as well as autocorrelation within the periodic returns which is a clear violation of the I.I.D. assumption. Hence de-smoothing is essential to all real estate indices regardless of the chosen optimization approach. Without de-smoothing the ex post real estate volatility is underestimated by 77% relative to the de-smoothed estimates. Neglecting the ex-ante volatility implied by liquidity risk causes the mean variance approach to underestimate the total risk of real estate investments compared to Cheng, Lin & Liu. The estimated deviations are most significant with shorter holding periods and tend to decline as the investment horizon increases. For the one-year holding period, the ex-ante risk caused by illiquidity consists of up to 57% of the total risk. The heteroskedastic variance causes investment horizon to influence portfolio optimization due to yearly increases in the average periodic risk. Using the Cheng, Lin & Liu approach the optimal investment horizon varies from two to seven years depending on the expected time-on-market and level of transactions costs. The corresponding real estate allocation varies from 8,94%, when the time-on-market and transaction costs are low, to 3,22% when the time-on-market and transaction costs are high. The mean variance approach estimates an optimal allocation of 22,12% given the same level of risk aversion. Based on Cheng, Lin & Liu’s ability to adjust for real estate characteristics the thesis recommends an allocation in line with those findings.
|Educations||MSc in Finance and Accounting, (Graduate Programme) Final Thesis|
|Number of pages||130|