Sector-Level Out-of-Sample Performance of the Naive and Sharpe Portfolios Using a Covid-Correction Breakpoint

Recent research has demonstrated that many mean-variance and shortfall-based optimal portfolio selection fail to out-perform the Naive (1/n) Portfolio in out-of-sample testing. This paper revisits this line of inquiry by applying the Naive and Sharpe Portfolios to 1100 sector-specific S&P 500 re-sampled data sets from the 2007-2021 time frame. Using April 2020 as the baseline train-test split break point, the Naive Portfolio delivers statistically significantly superior Sharpe Ratios in the test data in ten of the eleven sectors. However, the Sharpe Portfolio delivers statistically significantly superior shortfall values in all eleven sectors in the test data. Using March 2020 and May 2020 as alternative breakpoints gave similar results to the baseline analysis. Interestingly, when the data set was truncated at February 2020 (i.e., before the Covid correction) the Sharpe Portfolio returned statistically significantly better Sharpe Ratios than the Naïve Portfolio in the test data in all but the Energy sector; as in the baseline analysis, the Sharpe Portfolio returned statistically significantly superior shortfall values for all eleven sectors. Thus, the Sharpe Portfolio can deliver acceptable out-of-sample performance, but the conditions for success appear to vary by sector and test data erraticism.