Out-of-sample performanceCompared with in-sample performance, people are more concerned about out-of-sample model performance, because the latter canbe used to infer how well the model will perform in the practice. Therefore, we test for the out-of-sample performance of stock-bondportfolio constructed by our modified DCC-MIDAS model and compare it with other popular portfolio strategies such as DCC-MIDASþRCand DCC-MIDAS þ Investment Sentiment models. We divide the whole sample period into two subperiods: the in-sample periodfor modeling and parameter estimating, which extends from January 3, 1986 to December 31, 2009 (5744 data points) and the out-ofsampleperiod for evaluating portfolio performance, which is from January 1, 2010 to September 30, 2015 (1376 data points). Weestimate the model parameters using the in-sample data and then use the parameter estimates to forecast the future stock-bond ratio.Next we consider the optimal portfolio of stocks and bonds, that is, test for the out-of-sample performance of different strategies andcompare them. The return on a stock-bond portfolio can be denoted as: