Strategic asset allocation research

Question

I am currently trying to form an overall asset allocation strategy which combines base strategic allocation and tactical shifts. My model already incorporates the tactical shifts using various factors like momentum,carry etc. But i am using a base 1/N strategic allocation for all my asset classes (equity, bonds and commodities). I would like to improve this base allocation using some strategic allocation model.

I have already read the following papers regarding this:

Strategic Asset Allocation
Karl Eychenne, Stéphane Martinetti, Thierry Roncalli, 2011

To implement strategic asset allocation, we must determine risk and return expectations for the various asset classes. Starting from the paradigm that long-run asset returns are determined by the long-run fundamentals of the economy, a fair value approach to building expectations is crucial. This paper proposes to formalize a quantitative and systematic methodology for optimizing portfolios, from the determination of long-run fundamental pillars through the modeling of asset returns and the assessment of market risks. We apply forecasting models and build in the specific of the main asset classes (equities, bonds and alternative investments) depending on the uncertainties they represent for the risk-averse investor. Our resulting allocations within the equity asset class, and with regard to the place of alternative investments, question the choices of long-term institutional investors such as pension funds that have shifted their long-run allocations in response to the recent financial crisis.

https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2154021

and

Dynamic Strategic Asset Allocation: Risk and Return Across Economic Regimes
David Blitz, Pim Van Vliet, 2011

We propose a practical investment framework for dynamic asset allocation across different economic regimes, which we illustrate using a sample of U.S. data from 1948 to 2007. We identify four regimes in the economic cycle and find that these regimes capture pronounced time-variation in the risk and return properties of asset classes. Time-variation is also observed in the risk of a traditional, static strategic asset allocation portfolio. In order to stabilize risk across the economic cycle we propose a dynamic strategic asset allocation approach, which has the potential to enhance expected return as well. The proposed approach is found to be robust to variations in the variable composition of the regime model and can easily be extended with different economic variables and/or additional assets.

https://papers.ssrn.com/sol3/papers.cfm?abstract_id=1343063

They use forecasting of long run return using key fundamental and economic pillars.

I would appreciate if anyone has any relevant paper links to strategic allocation( recent research is preferred).

A little update on exactly what kind of research I am looking at:

I am not looking for research related to risk parity or MV optimized portfolios but on asset allocation strategies that combine views about the economy of different countries and how that will affect the expected returns across broad asset classes. Which i can use to fix the strategic allocations for a period of next 5-10 years and overlay monthly tactical shifts (models for which I have already developed) Two such kind of paper I have attached.

I am not sure why I am getting downvotes on a reference question, maybe my research objective is not clear. Here is another paper from BlackRock regarding this, ideally I would like to replicate this but again they have disclosed too little, so I need similar "academic research".

Building resilience: a framework for strategic asset allocation

Designing a suite of models that explicitly reflect cross-asset and macroeconomic linkages. These models are the key input to inform our views on returns across asset classes

Link to paper

Answer 1

One thing I personally believe is that SAA should minimize forecasting the future. Ideally (though not possible), SAA is the optimal asset allocation when you have no views about the future at all. A weaker version of that statement is that SAA is the portfolio that's mostly likely to meet your risk/return-objectives (subject to constraints), based on assumptions that are as reliable as possible. Accordingly, I generally do not like introducing economic views and/or incorporating alphas into an SAA program. This is just a personal belief and I fully acknowledge other approaches might produce better results, but I thought I'd lay this out at the beginning to provide some context for the comments below.

SAA at its core involves just two steps:

1. Setting Capital Market Assumptions ("CMA") for the asset classes/factors you plan to use in your SAA.
2. Calculate the weights for these assets to meet your long-term return objective subject to constraints (such as risk/spending/illiquidity).

There is a large literature on CMA out there. This answer summarizes the different techniques and provides some well-known practitioner examples. Because it was more fixed income oriented, some updates are needed:

1. AQR has been publishing their CMAs for major asset classes for a few years ago. They use a building-block approach that's very popular in the industry. I'm a fan – the methodology is theoretically sound and makes relatively few assumptions.
2. Research Affiliates also publishes extensive documentations on their website. The methodology is based on building blocks as well, but it incorporates additional assumptions such as valuation reversion. Unsurprisingly, it would have better in-sample forecasting performance than AQR's approach (certainly for equities). It can be argued that valuation reversion cannot be depended on, which precludes it from entering the SAA process (again, just an opinion).
3. Simple risk parity ideas can be used to generate forward-looking assumptions as well. A quick example – let's assume that all assets over the long run will achieve the same Sharpe ratio of 0.3 (roughly what was realized and the equality assumption ensures investors won't favor one assets over another on an ex-ante basis), and let our forwad-looking volatility assumption for stocks be 15%, then the forward-looking excess return assumption is simply 0.3 * 15% = 4.5%. You can then add this to an expected cash return assumption to get the forward-looking total return. I like this approach too because of its simplicity. The principle assumption is that all assets are roughly the same on a risk-adjusted basis, which again fits well with the belief that SAA should have no strong views.
4. Many other shops provide similar reports (e.g., JPMorgan). They all end up being variations of similar ideas.

Additional searches for "Capital Market Assumptions" will return many relevant answers, but as I mentioned in the CMA answer, empirical evidence suggests that simple approaches are remarkably difficult to beat. For example, starting yield levels are more than sufficient for forecasting next 10-year government bond returns. In equities, John Bogle's formula for forecasting long-term equity return is surprisingly effective (and the methods cited above can be considered its variations). I urge you to explore these simpler techniques before delving into complex econometric models. Keep in mind that the forecast horizon is very long and the range of outcomes is enormously wide, so any number will likely be wrong. I'd also comment that your desire to link returns to growth/inflation is understandable. But I hope I'm conveying that 1) the added complexity is not necessary, and 2) literature will disappoint (e.g., long-term growth and equity returns have poor linkage, think China...).

With regard to the second step, as others have pointed out, the traditional approach, such as MVO, remains the workhorse. I do think this is the step where extra work could yield dividends. For example, the shortcomings of MVO are well known, and volatility is a poor measure of risk in the long-run. Alternatives include resampling optimization and CVaR optimization. I recommend JPMorgan's excellent paper "Non-normality of market returns" for inspirations.

Answer 2

Markowity remains the stonghold, thus you must start reading that in my view. Afterwards, you can have a look at risk parity. Recent developments are based on hierachical clustering and neural Networks. I suggest reading Lopez de Prado on hierarchical clustering asset allocation.