Risk Management through Costless Collars
January 5, 2010
by Geoff Considine
Nassim Taleb and Zvi Bodie are among those who advocate a wealth management strategy that includes options. Despite their evangelism, though, options are rarely a part of retirement portfolios.
The risk-limiting properties of options will be increasingly valuable as investors react to the amplified volatility of the last two years. The costless collar, a straightforward options strategy, gives investors the upside of an asset class (such as equities) while absolutely limiting the downside risk.
Costless collars are created by selling call options against a security or an index, and using the proceeds to purchase an equivalent amount of put options. The costs of the option transactions offset one another, and establish a maximum return and minimum loss for the asset.
My research suggests a growing role for options in wealth management, and collars in particular will have considerable appeal to investors. Among its practitioners is the $2.8 billion Gateway Fund (GATEX), which uses collars as part of a tactical asset allocation strategy. A recent startup fund, the Collar Fund (COLLX), applies this strategy in a more mechanical fashion and provides a good illustration of the value costless collars offer.
Lets explore the research behind costless collars, how well they work under todays market conditions, and their appropriate role in a wealth management strategy.
Research confirms the value of the collar strategy
Academic research supports the idea that a costless collar can enhance the risk-adjusted returns of a portfolio. A 2008 article by Louis DAntonio, Equity Collars as an Alternative to Asset Allocation, looked at how a mechanical strategy of repeatedly executing rolling costless collars with a one-year period to expiration for the S&P500 would have altered the risk and return of a portfolio versus a straightforward investment in the S&P500 for the period from 1926-2005. The collar structure sells a call with a strike that is 20% above the current level of the S&P500 and buys a put with a strike that is 10% below the current level of the S&P500. The portfolio receives the first 20% in annual gains and is protected against any losses beyond -10% over each one-year period. I will discuss the reasonableness of the assumptions underlying this strategy shortly.
If such a collar could be executed, what would happen? The average (arithmetic) annual return of the S&P500 over the 79-year period was 12.3% with a standard deviation of 20.2%. The average annual return of the collar strategy was 8.9% with a standard deviation of 12.1%.
Unsurprisingly, the collar strategy lowered both risk and return.
As a direct benchmark, the analysis also shows that a portfolio that is made up of 52% S&P500 and 48% five-year Treasury bonds had an average return of 8.8% with a standard deviation of 10.9%, almost identical in average return to the collar strategy, but with less risk. The big difference between the collar and the 52/48 portfolio was in the extreme downside. The worst one-year return for the collar strategy over the period was -10%, as compared to -23.7% for the 52 /48 portfolio. If the standard deviations are the same, but the probabilities of extreme events are higher, this suggests that the simple stock/bond portfolio has much higher kurtosis (i.e. fatter tails) than the collar portfolio and this is exactly what the author found. Extreme upside deviations further confirmed that the collar portfolio had lower kurtosis.
Exploring costless collars with Monte Carlo simulation
To further illustrate these concepts, I looked at three alternative portfolios using Monte Carlo Simulation (Quantext Portfolio Planner, in this case):
- 100% investment in S&P500 ETF (SPY)
- 50% investment in SPY and 50% investment in bond ETF (AGG)
- Investment in SPY with costless collar executed on SPY
I simulated returns for a one-year time horizon for an S&P500 ETF (SPY) and for a bond ETF (AGG). From there it is simple to model the costless collar portfolio. The probabilities of the range of possible returns for a one-year time horizon are summarized in the chart below (a percentile chart).