User:Stocksurfer/Risk parity portfolio

From Bogleheads
Jump to navigation Jump to search

Also: All-Weather Portfolio, Bridgewater All-Weather, Ray Dalio All-Seasons Portfolio

Risk parity is a methodology to derive an asset allocation through numerical means that focus on balancing the risks across asset categories, and more specifically, across the expected performance of those assets in four different economic environments. It is geared to portfolios that are broadly diversified beyond stocks and bonds, for example when a portfolio contains 15 or more asset classes.

Risk parity portfolios were popularized by Bridgewater Associates and Ray Dalio in particular.[1] They used the term Risk Parity to describe the All Weather Portfolio -- a specific commercial offering -- as well as the All Weather strategy.[2] Tony Robbins popularized a simplified version of the All Weather portfolio called All-Seasons portfolio in his book MONEY Master the Game: 7 Simple Steps to Financial Freedom. [3], but while this portfolio was produced in conjunction with Ray Dalio many people disagree that it appropriately reflects the All-Weather strategy.[4] [5]

To quote Ray Dalio: "Risk parity is the means of adjusting the expected risks and returns of assets to make them more comparable. This is done for the purpose of creating a better diversified portfolio that will have a better return-risk ratio than would otherwise be possible. Once the better diversified portfolio is created and the return-risk ratio is improved, the portfolio can be geared (i.e. leveraged) to the desired level of risk and return." [6]

There are 3 conceptual parts to a Risk Parity All-Weather portfolio:

  1. The notion of Risk Parity, an investment strategy that balances risk across a number of asset classes.
  2. Recognizing four economic environments defined by the four possible combinations of: inflation is above or below expectations priced into the market, and economic growth is above or below expectations.
  3. An implementation of a portfolio that (a) uses broad diversification across asset classes to cover all four economic environments, that (b) selects an asset allocation based on equalizing the risk across the asset classes and economic environments, and then (c) leverage the portfolio to achieve the desired level of rains vs. risk.

The role of leverage

A simple risk parity portfolio ends up having a rather low expected mean return. For this reason, Bridgewater's All-Weather portfolio uses leverage to increase the return. The use of special assets as well as the use of leverage makes it difficult for a retail investor to replicate the all-weather portfolio:

  1. the management complexity of balancing risk exposures
  2. the operational complexity of leveraging low-return assets: one needs to manage a portfolio of rolling futures contracts and/or access the repo (repurchase) market.

A possible simple solution is to increase allocations on long duration nominal bonds (EDV), long duration inflation linked bonds (LTPZ), gold (IAU), and commodities (GSG).[7]

A possible complex solution is to open an Interactive Brokers account and gain the appropriate risk exposure through equity, interest rate, and commodity futures. This is expensive analytically and far beyond the capabilities of typical retail investors.[7]

Problems with Risk Parity

As Risk Parity gained popularity after Bridgewater's first publication a number of criticisms began to surface. Jonathan Cooper describes two of them well in The Problem With Risk Parity and some more formal problems are analyzed by JP Morgan Asset Management in Improving on risk parity.

Cooper criticizes that Risk Parity uses estimates or forecasts of asset risks (typically expressed as volatility over a number of recent months) but ignores similar estimates or forecasts about recent/expected returns and that as a result Risk Parity portfolios provide poor returns. In addition, Risk Parity calculation typically operate on compound assets (such as index funds) which causes the calculations to be sensitive about how the primary assets were grouped into those compound funds. He gives an example where grouping the primary assets (stock and bonds comprising the US market) into two different pairs of funds yields very different portfolios.

JP Morgan Asset Management's criticism is similarly about not using asset class performance forecasts at all. Their report is cited here less for the specific criticism than as an example of identifying a problem with Risk Parity and then proceeding to improve on it, which rapidly devolves into rather complex financial theory. At some level it appears that this has become a blueprint for asset management firms: use "risk parity" as an advertisement banner, then identify a flaw in the methodology, and finally propose a solution and turn it into a sellable strategy.

On the academic front there are also a number of examinations and improvements to Risk Parity: Equal-Weighted Risk Parity and Clustered Risk Parity are two examples.

Economic Environments

To be written...

Naive risk parity example

The example in this subsection is meant to convey the basics of how Risk Parity works, it is not meant as a how-to example to actually follow and implement in real life. The background for naive risk parity comes from Dynamic Asset Allocation for Practioners:Part 4: Naive risk parity [8] or Risk Parity for Dummies.[9] Naive risk parity ignores correlation between assets.

This example uses Azanon's portfolio from the Improving the Dalio/Robbins All-Seasons Portfolio thread:

  • 25% Vanguard Mid-Cap Value ETF (VOE)
  • 10% Market Vectors Emerging Mkts Local ETF (EMLC)
  • 20% Vanguard Extended Duration Treasury ETF (EDV)
  • 30% PIMCO 15+ Year US TIPS ETF (LTPZ)
  • 7.5% ETFS Bloomberg All Commodity Strategy (BCI)
  • 7.5% iShares Gold Trust (IAU)

First retrieve data about the assets from Portfolio Visualizer's Asset Correlations calculator. Using this link returns a complete correlation matrix out of which we only need the monthly standard deviation column to calculate asset weights:

Ticker stddev 1/stddev Kt weight
VOE 4.04% 24.75 12.77%
EMLC 2.92% 34.25 17.67%
EDV 4.52% 22.12 11.41%
LTPZ 2.73% 36.63 18.90%
BCI 2.37% 42.19 21.77%
IAU 2.95% 33.90 17.49%
0.005159 sum: 100.00%

where Kt=1/sum(stddev) and weight=(Kt/stddev)

Note that due to the fact that BCI was created in April 2017 the data for the above calculation does not go far back. A useful next step would be to substitute an older fund that invests similarly, for example DJP.

Azanon's improved all-seasons portfolio

Perhaps this section should be moved onto its own page, keeping it here until there's more clarity...

Late 2016 forum member azanon set out to improve the Tony Robbins / Ray Dalio All Seasons portfolio. He started an Improving the Dalio/Robbins All-Seasons Portfolio thread and stated the following issues with the All-Seasons portfolio:

  • it seems to over-allocate to defending against deflation, and under-allocate to defending against inflation
  • it isn't adequately designed to generate enough return, especially considering that it isn't leveraged.

Azanon's is interesting from a BogleHead perspective because it attempts to capture the benefits of the All-Weather strategy without diving head-first into the complex financial theory that underlies most (all?) so-called Improved Risk Parity strategies. Instead, it blends the strategy and asset allocation provided by Robbins with newer information, with ideas about using internally leveraged funds, and with a small dose of Risk Parity calculations. The result is a portfolio that could be used in BogleHead buy-hold-rebalance style by those that are interested in a Risk Parity style portfolio.

The following subsections retrace Azanon's steps based on a December 2019 post.

The subsections below need rephrasing and more details

Initial asset classes

Pick the asset classes based primarily on those mentioned/referenced in Bridgewater's All Weather Story since that's what I was trying to replicate (Here: ... -story.pdf). I then gave some consideration (less than All Weather Story though) to what was doing (asset classes used and weights) since those guys are former Bridgewater employees who are trying to replicate the same strategy. And then finally I considered what Ray said to Robbins, and gave this the lowest priority since I highly suspect Ray intentionally gave Robbins a watered-down version of what Bridgewater does. 4. I have to mention "Tag" again too (former Bridgewater employee) who also helped here....

Apply Risk Parity to the economic environments

I sorted the asset classes based upon which of the 4 categories/quadrants they went in from the All-Weather Story 4-box grid linked above (I believe all of them are 50/50 to two quadrants), and then weighted their quantity based on their annualized standard deviation (using portfolio visualizer "asset correlation"). As a quick example, asset A with 10% annual SD is needed twice as much as asset B with 20% SD, for them to be risk balanced. Unfortunately, I don't think I kept clean notes of which quadrants each assets fit into and their respective weights, but in the end, at least mathematically, each of the 4 quadrants had as close to as exactly the same risk-adjusted weights as the other. And of course each box had 2-4 assets per box. Again, I used mathematics to make sure they were still the same risk, in aggregate.

Tune asset classes

Also on this same portfoliovisualizer screen (Asset Correlations), I considered how "uncorrelated" each potential asset class (ETF) was for inclusion in the portfolio to all of the other assets, with lower correlations being better. For instance, I was able to confirm that local currency EM bonds (EMLC) was lowly correlated to almost all of the other asset classes, while also being individually volatile (a good thing in risk parity), so it easily made the cut. It was this screen that also let me determine that "mid-cap Value" was the most uncorrelated of all the possible 9 morning star boxes vs. the other prospective asset classes (e.g. LC Growth has higher correlation to the other assets, collectively, so it was rejected. MC Value, in contrast, had the lowest while being very volatile, and I later was able to approximate with a proxy that VFVA (value factor) was even lower correlation (and more volatile) than regular VG MC Value index).

Note about expected returns

No consideration was given whether an individual asset class had a high return on its own. So, I didn't penalize commodities, for instance, because they average no real return by themselves. Likewise, stocks weren't increased in quantity for the same reason either. The portfolio is designed exclusively to be risk balanced for all possible economic scenarios - agnostic to any climate - so it provides whatever return it provides. It is similar to the permanent portfolio in this viewpoint. In summary, a powerful thought to consider is that, at least in theory, you largely if not exclusively don't care what's coming next - growth, slowdown, recession, .... whatever. You're ready for it regardless.

Resulting Portfolio

  • 25% Vanguard Mid-Cap Value ETF (VOE)
  • 10% Market Vectors Emerging Mkts Local ETF (EMLC)
  • 20% Vanguard Extended Duration Treasury ETF (EDV)
  • 30% PIMCO 15+ Year US TIPS ETF (LTPZ)
  • 7.5% ETFS Bloomberg All Commodity Strategy (BCI)
  • 7.5% iShares Gold Trust (IAU)


Forum member Klaus14 described a variation on Azanon's portfolio where he splits the stock allocation to use a fund that holds leveraged equities as well as a multi-factor fund. He also splits the emerging markets bonds into local currency as well as USD-denominated bonds. With some further changes his resulting portfolio is:

  • 10% WisdomTree 90/60 U.S. Balanced Fund (NTSX) 0.20%
  • 10% Vanguard US Multifactor ETF (VFMF) 0.18%
  • 5% iShares Core MSCI Emerging Markets ETF (IEMG) 0.14%
  • 5% iShares EM Local Currency Bond ETF (LEMB) 0.30%
  • 5% Vanguard Emerging Markets Government Bond ETF (VWOB) 0.30%
  • 20% Vanguard Extended Duration Treasury ETF (EDV) 0.07%
  • 30% PIMCO 15+ Year US TIPS ETF (LTPZ) 0.20%
  • 7.5% ETFS Bloomberg All Commodity Strategy (BCI) 0.29%
  • 7.5% SPDR Gold MiniShares Trust (GLDM) 0.18%

See also



Further reading

Introductory material, e.g., why risk parity and all-weather:

More detailed discussions:

More specific how-to and calculations:

External links