Академический Документы
Профессиональный Документы
Культура Документы
Abstract
A broad diversified portfolio that is designed to collect risk premiums across liquid
assets throughout various economic environments has a more stable performance and
less volatility over time than traditional equity-concentrated portfolios. By allocating equal
risk to each economic environment, surprises in the underlying economy have less
impact on portfolio performance. The portfolio has low drawdowns under stress
scenarios, such as the euro area crisis in 2011, but is vulnerable when risk asset
correlations drift towards the one that occurred after the Lehman Brothers collapse. Our
research suggests that active asset allocation after controlling for trading costs in the
period from 2001 to 2013 does not generate significant excess return measured against a
static asset allocation with rebalancing. Furthermore, frequent data (weekly) controlling
for trading costs increases the Sharpe ratio, but fails on drawdown and the Sortino ratio
when measured against less frequent data (monthly).
Head of Equity Strategy, Trading Strategy Team, Saxo Bank | Philip Heymans All 15, 2900 Hellerup | pg@saxobank.com
ii
Head of Macro Strategy, Trading Strategy Team, Saxo Bank | Philip Heymans All 15, 2900 Hellerup | mdk@saxobank.com
Portfolio basics
The framework for the Permanent Portfolio is built on the idea of four economic environments: prosperity,
inflation, deflation and recession. Based on the premise that the future is difficult to predict, Harry Browne
assigned an equal weight (probability) to each economic environment. This methodology of assigning equal
weight to model factors lies squarely within the thoughts of Daniel Kahneman, the winner of the 2002 Nobel
Memorial Prize in Economic Sciences, which he elaborates on in his book Thinking, Fast and Slow (2011).
The All Weather framework uses a slightly different terminology. In the world of Bridgewater, the economy goes
through four different economic environments: rising growth, falling growth, rising inflation and falling inflation, all
relative to prevalent expectations (see table below). Each economic environment benefits different asset classes
and by balancing the portfolios risk across all environments in equal weights, the portfolio earns the asset class
risk premiums while lowering the exposure to any one environment. As a further diversification step, which sets
the All Weather apart from the Permanent Portfolio, each environment bet is done with a mix of assets. For
example, when inflation rises faster than expectations inflation-linked bonds, commodities, emerging-market
credit and gold are expected to benefit (see figure below).
Bridgewater uses slightly different assets. They use corporate spreads and emerging-market debt spreads
instead of corporate credit and emerging-market credit. The reason why we changed the assets was because
debt spreads are not accessible through cheap and liquid ETFs which are the basis for implementing the asset
allocation. Furthermore, we have included gold as a separate asset in the rising inflation environment similar to
the approach in the Permanent Portfolio.
The next step is to calculate the assets target weights, which is done by multiplying the economic environment
weight with the assets weights, which we set to equal. The weights are presented in the table below.
The combined asset weights are illustrated in the pie chart below.
Portfolio implementation
The seven assets in the Balanced Portfolio are tracked by seven liquid ETFs traded on US exchanges (see table
below). As the total expense ratio shows, ETFs have become a cheap way to gain broad market exposure. The
total expense ratio for the Balanced Portfolio is approximately 31 basis points (0.31 percent) calculated as the
weighted average of expense ratios.
All ETFs are quoted in USD and as a result, the portfolio is traded on a USD account with no currency hedging.
Investors with another base currency than USD have at least two options. The first option is to trade it without
currency hedging, which may lead to higher volatility relative to the USD-implementation. A second option is to
dynamically hedge the currency exposure through the spot exchange rate between the investors base currency
and the USD on a monthly basis.
To keep trading costs as low as possible, the portfolios weight will not be rebalanced when deviations from the
target weights are small. Instead, the assets have target weights with a fixed band that is defined as +/- 20
percent (see table below). If weights stay inside the bands, no rebalancing will occur.
Backtesting results
The backtesting has been performed on monthly data on total return benchmark indices associated with each
asset starting January 2001 and ending March 2013. The total number of observations is 148. Rebalancing is
conducted with trading costs of 0.2 percent to capture slippage and commissions. Since the backtesting utilises
the underlying total return benchmark indices, the ETF expense ratio for acquiring these has not been taken into
account. This inflates the annualised performance figures. However, the weighted expense ratio is approximately
31 basis points on an annual basis, so the drag is minimal.
Part of backtesting the Balanced Portfolio was also to see if more frequent data increases performance. The table
below shows the performance statistics between weekly and monthly data. Although the weekly portfolio beats
the monthly on annualised return and the Sharpe ratio, it fails on the most crucial point: the Sortino ratio and
maximum drawdown. The Sortino ratio does not penalise upside volatility and thus is a better risk-adjusted
performance measurement than the popular Sharpe ratio. The maximum drawdown is 2.4 percentage points
lower for the monthly portfolio. Overall, these figures combined with the simplicity of monthly rebalancing have
made us choose to implement the Balanced Portfolio on monthly data. In the backtesting the portfolio was
rebalanced on 24 occasions.
The chart below shows the monthly Balanced Portfolios backtesting performance since January 2001. It clearly
shows the benefits of diversification leading to a smooth cumulative return profile with small drawdowns,
excluding the meltdown following the bankruptcy of Lehman Brothers.
The chart below shows the drawdown profile for the Balanced Portfolio. Despite a balanced and diversified asset
allocation, the portfolio still experienced a moderate drawdown in late 2008. Excluding this drawdown, the other
drawdowns are all less than 5 percent, highlighting the robust framework.
Another way to look at the 2008 drawdown is in the chart below. It shows that the two largest monthly declines
happened consecutively and thus amplified the drawdown. The chart also shows that those two monthly returns
were significantly non-normal, adding evidence to the well-proven fact that financial markets do not follow a
normal distribution.
Sep. 08
Oct. 08
During the escalation of the euro area crisis in August 2011, the Balanced Portfolio experiences a 3.7 percent
drawdown and it only lasts for two months; again, proof of the portfolios resilience. The average drawdown length
excluding the financial crisis of 2008-2009 is only four months (see table below).
The table below provides an overview of monthly and yearly returns. The Balanced Portfolio has only had one
year of negative return out of more than 12 years of backtesting. The Balanced Portfolio did worse in 2008,
returning -6.3 percent compared with -0.7 percent for the Permanent Portfolio. This highlights the unique
character of the meltdown of 2008. The reason for the Permanent Portfolios outperformance in 2008 was due to
its 25 percent weight in cash, which also ended up being the only safe haven as risk asset correlations drifted
towards one and diluted the effects of broad asset diversification.
Future expectations
Past returns are no guarantee of future results. The backtesting covers a unique period (2001-2013) for financial
markets with historical large volatility in stocks and rapidly declining interest rates, and thus a few comments
seem prudent. The portfolio allocates more than half of its total capital to various forms of credit, which has greatly
outperformed equities in the backtesting period. Furthermore, the annualised return of 9.0 percent in the
backtesting does not account for ETF costs, which are currently 31 basis points for this portfolio (May, 2013).
These costs will be a drag on live performance. Furthermore, slippage and commissions related to rebalancing
might be larger than 0.2 percent used in the backtesting.