# How to evaluate Asset Allocation skill?

There have been studies that show that Asset Allocation can explain 90% of the variance of returns on a portfolio. If true and Asset Allocation is the primary driver of return risk, how can you assess the skill of Asset Allocators? Is there a benchmark to evaluate the success or failure of an Asset Allocation process?

Given the broad range of Capital Market Assumptions, with their prediction errors and concomitant prediction errors of various Asset Allocation decisions, how can one attribute the success of a program and which part of the Asset Allocation process is successful or not?

Not necessarily an answer but was too log for a comment.

I guess a starting point will be the papers outlined here.

That said, if you think of it intuitively, it's not surprising that the choice of your assets matters most.

I am a big sceptic when it comes to skills of asset allocators so I am not surprised if that number is 90% or higher. Especially since the superior performance of the stock market over other asset classes ultimately rests on the shoulders of a select few.

In essence, if you read the referenced papers in the first link, total value added is the difference between the actual portfolio return and the benchmark return. $$R_A = \sum_{j=1}^M w_{P,j}*R_{P,j} - \sum_{j=1}^M w_{B,J}*R_{B,J}$$ The first summation has both portfolio weights and and the return on actively managed portfolios, denoted by subscript $$P$$. The second summation has benchmark weights and benchmark returns, denoted by $$B$$. Subscript $$j$$ to $$M$$ is a counter for asset classes. This can be rewritten as the sum of active asset allocation decisions and the weighted sum of the value added from security selection, $$R_{A,j} = R_{P,j}-R_{B,j}$$ within each asset class:

$$R_A = \sum_{j=1}^M \Delta w_{P,j}*R_{B,j} - \sum_{j=1}^M w_{B,J}*R_{A,J}$$

If you think of just stocks and bonds you get a simplified formula that looks like this.

Deviations from portfolio benchmark weights drive the value added by active portfolio management (or value lost in most cases). That is the last computation in the link. Ignore the question of the person who asks this. $$\Delta$$ is the difference to benchmark weights what they call "active allocation weights". The A (also in my syntax) in the second term means actual active weights.

There are two primary asset allocation decisions:

1. Strategic asset allocation: This tends to be a long-term, passive portfolio mix that an institution would hold if there are no active views. In principle, measuring success/failure is easy – does the SAA process generate the necessary return profile to support an institution's needs/missions? For example, most endowments want the value of their endowment to keep pace with inflation, after making an annual distribution of 5%. In this context, a natural benchmark is CPI+5%. Most of the common benchmarks you'd see in this circle – going from something as simple as 70% equity + 30% bonds, to more complex "policy portfolios" – all share the basic goal of achieving the 5% real return target. Aside from these "market-based" benchmarks, you could also be benchmarked against your peers. Being in the top-decile is probably a good sign. Of course, if you're a pension, your SAA needs will be very difficult and a different set of benchmarks will be appropriate.

2. Tactical asset allocation: TAA can completely decouple from SAA, but I'm more referring to tactical "tilts" here. For example, your SAA may call for a 50% allocation to global equities, but your current market view is that equities will outperform other assets in your SAA mix, so you increase global equities to above 50%. In these cases, the value-add of the process can be quantified using traditional performance attribution framework (such as the one proposed by @Akdemy).

P.S. I'm reasonably sure that the research you're thinking about was talking about the variances in returns can largely be attributed to asset allocation decisions.

• Thanks very much for your thoughtful response (corrected the question to variance). I am concerned more about SAA. Success or Failure to meet the returns is one metric (and the most important). But how can you tell if the asset allocator has any skill and this was not just dumb luck? Was is CMAs (return and distritbution estimates), optimization, etc or any other part of the process that led to the success? Apr 30, 2021 at 14:52
• "how can you tell if the asset allocator has any skill and this was not just dumb luck?" - you could use a benchmark for comparison. I don't know what the benchmark is for asset allocation; perhaps someone more knowledgeable can chime in. Apr 30, 2021 at 20:07
• @AlRacoon Super interesting question. I don't think there's an easy answer, given how different these funds are managed. It's particularly challenging because we really need a long history to assess SAA. For example, ANY diversified portfolio underperformed a simple 70/30 or 100/0 portfolio in the last decade. That doesn't mean diversification is bad; just the past decade is not favorable. Anyways I looked at our research archives and good starting points might be "Alpha and Performance Efficiency of Ivy League Endowments: Evidence from Dynamic Exposures" & "Alpha, Beta & the Endowment Model." May 1, 2021 at 19:40