As guidance for successfully relying on dynamic asset allocation, the authors analyze the impact of rebalancing frequency on portfolio performance. The rebalancing is guided by predictive regression forecasts of stock, bond, and bill returns and conditioned on potential patterns of transaction costs.
Dynamic asset allocation (DAA) can offer investors potentially appealing returns, provided that its required portfolio rebalancing is based on reliable out-of-sample return forecasts for the involved asset classes. The authors test whether profitable portfolios can be constructed and rebalanced at various frequencies, relying on predictive regressions of the constituent assets’ returns and conditioning across a range of potential transaction costs. They conclude that DAA can be profitably implemented to offer investors the greatest rewards via monthly rebalancing if unit transaction costs are less than 50 bps. For higher costs, when extending much higher than 400 bps, worthwhile gains remain attainable under annual rebalancing.
How Is This Research Useful to Practitioners?
For wealth managers advising clients with suitable risk tolerance, identifying appropriate applications for DAA would seem to make active management a plausible alternative to indexing. The authors strive to construct models grounded on underlying market dynamics to shape return forecasts derived from principal component analysis pertaining to a variety of fundamental, macroeconomic, and technical factors. This approach yields portfolio asset weights that are significantly linked to business cycle fluctuations, enhancing their reliability and thus offering the prospect of effectively exploiting time-varying risk premiums. They do not test the theoretically maximum profit possibilities in which the timing and degree of rebalancing is conditioned on evolving economic circumstances as opposed to fixed frequency adjustment (e.g., with the start of a new year). But by focusing on commonly adopted strategies of invariant and time-dependent rebalancing, their approach cannot be criticized for overlooking typical investor behavior.
The authors’ frame of reference for test comparison—the benchmark balanced portfolio of 60% stocks, 35% bonds, and 5% bills at a constant expected return baseline—is a plausible target. Their finding that DAA portfolios outperform this benchmark at all rebalancing frequencies (assuming no transaction costs) with their advantages magnified during times of extreme economic stress, such as during severe recessions, definitely addresses the perennial investor need of asset protection. The study may have been even more useful to wealth managers if it had considered taxes in addition to transaction costs. Overall, the study appears to sensibly address investment themes of significance.
How Did the Authors Conduct This Research?
The authors analyze the performance of DAA portfolios constructed from monthly, quarterly, semiannual, and annual out-of-sample forecasts of US stock, bond, and bill returns. The forecasts are from January 1965 to December 2012. They construct DAA portfolios in two steps, first using a modified version of the Black–Litterman model to establish investors’ active views of expected returns. The derived posterior return covariance matrix is then transformed to calculate DAA portfolio weights. Performance is compared against the balanced benchmark portfolio with an assumed constant expected return (random walk with drift). The DAA and benchmark portfolios are rebalanced at the same frequency as the forecast horizon, with the DAA’s asset weights showing a strong link to business cycle fluctuations.
Average returns for the DAA portfolios are shown to be at least 60 bps higher across all rebalancing frequencies, with somewhat higher standard deviations than their benchmark counterparts. To determine how this result can be affected by transaction costs, information ratios and certainty equivalent returns during the 1965–2012 study period are computed to generate several transaction costs/rebalancing frontiers that illustrate the rebalancing frequency that has the greatest outperformance for any given unit transaction cost. If low transaction costs are available, more frequent (i.e., monthly) rebalancing is desirable to exploit changes in expected returns. But as costs rise, the advantages of greater reliability of signals provided from annual return forecasts and lower portfolio turnover point to less frequent (i.e., annual) rebalancing. The study ignored incremental taxes that could result from DAA.
The authors seem to have carefully studied relevant academic literature to construct their models with a likely sensitivity to sound methodological principles. But it also must be recognized that two of the authors are professionals in the wealth management unit of a major Wall Street firm and the remaining co-author is an academic with a consulting agreement to the other authors’ employer. So, there is a possibility that analytical objectivity is at least unconsciously affected by a bias toward dynamic investment strategies suited to their firm’s self-interest (i.e., fee generation). Client acceptance will likely prove the ultimate judge of their ideas.