Can You Beat the Market? Part V. Collars Outperform Buy and Hold

Parts: one, two, three, four

University of Massachusetts Professor Schneeweis and doctoral student Szado conducted a study, sponsored in part by the Options Industry Council (OIC).  The results are first being announced today, 9/23/2009, in a paper entitled: "Loosening Your Collar: Alternative Implementations of QQQ Collars."

Using 10 years of data, the study compared the performance of QQQQ, the PowerShares exchange traded fund when collared and when unhedged (owning QQQQ only, with no options).  As with the CBOE S&P 500 CLL 95-100 Collar Index, this study involved the purchase of 6-month puts and the consecutive sale of one-month call options. This time both the puts and calls were 2% OTM when traded.

[NOTE:  If the same nomenclature were used for this collar as for the CBOE collar, it would be called the OIC PowerShares 100, 98-102 Collar Index].

The study included two different collars: passive (with a set of fixed rules) and active (the rules vary according to changing economic conditions).

The study shows that the collared QQQQ portfolio significantly outperformed a portfolio that simply owned QQQQ from April 1999 through May 2009.  And if that's not enough of an eye-opener, risk was reduced by 65% over the 122 month study.

These results are spectacular, especially when compared with the performance of the CLL 95-110 Collar Index.  I'll comment on that later, but for now, let's concentrate on the findings.


Author's Conclusions

The 6-month put purchase is better than buying either one- or 3-month puts.

Active collar more effective than passive.  But, this is not the place to be concerned with the active collar.  The original paper can be accessed for that information.  UPDATED:  link to that paper.

The authors state that the collar was less effective during the steady up years of 2002 to 2007.  That's as expected.


My comments:

The graph says it all.  The QQQQ got hammered during the bursting of the technology bubble, and that's understandable because QQQQ consists of the 100 largest capitalized, non-financial stocks that trade in the NASDAQ market.  That means this ETF was loaded with technology stocks.

But, something bothers me – and I'm publishing this post before discovering the answer.  Collars are slightly bullish strategies. If you recall from previous discussions, a collar is equivalent to selling a put spread.  These positions are protected against large losses, but how did the collared QQQQ increase in value as the underlying asset was plunging in 2001? 

Is it possible that the premium collected by selling the call was so much higher than the premium paid for the put that it offset losses in the decline of the ETF?  That's just doesn't sound reasonable.  When markets are falling, IV is high, but is skewed so that lower strike prices (the puts bought) is higher than that of higher strike prices (the calls sold).

The only way I can see the collars increasing in value occurs if the strategy is not a true collar but includes more than one put per 100 shares of QQQQ.  Yet, the collar in this paper is defined as one put and one call per each 100 share.  I'll go through the details when available, and report back.  But right now, I find these results very unusual.

Addendum @ 6 minutes later.  Apparently when puts were significantly ITM, there was no longer any downside risk.  Thus, the sale of 2% OTM call options resulted in a profit month after month, as the calls expired worthless in a declining market.  These feels 'wrong' to me.  It's essentially selling (almost) naked calls.  Who would do that in the real world?


These results are very unexpected.  The normal expectation for collars is that they will do an excellent job and protect your portfolio when the market declines.  But, to gain the benefits of that downside protection, the collar owner must accept a limit on upside profits.

And that's the problem.  During the  years of this study, the market was down.  The annualized return for QQQQ over the 122 months of this study is -3.57%, and that's a significant decline.  In other words, there was no upside to sacrifice.  But there was plenty of downside – specifically the bursting of the technology bubble in 2000 – 2001 and the 2008- early 2009 bear market.  Under these conditions, collars always perform well.

Don't misunderstand.  As someone who believes strongly that millions of individual investors should seriously consider using collars on all, or part, of their holdings, I like the results of this study.  But I have no plans to use it out of context.  I want to see how this strategy worked from 1995 through 1998 and May 2009 through the present time.  I want to see the data for this strategy at its worst, not only at its best.  Then we would all be able to see just how well collars perform when compared with buy and hold.  I know this study was performed in academia, and that the numbers can be trusted, but I wonder whether the time period chosen was cherry-picked to produce these outstanding results.  More data please.

From the performance of the CLL 95-110 Collar index (see part IV), we can see severe underperformance for the collar strategy in the most bullish years.  Yet, collars were very valuable during bear markets.  Overall, the performance of this specific collar was disappointing. 

But, in defense of collars, the CBOE produced a collar index that is not as valuable as they might have chosen.  Writing call options that are 10% OTM may be suitable to the bullish investor, but to someone who wants to use collars for both capital gains and protection, I believe the more useful Index is the 95-100 CLL Index, which includes the sale of ATM calls.  I urge the CBOE to publish the CBOE S&P 500 95-100 collar index, which would allow a direct comparison with its BXM (Buy-Write Index).  

There is a lot of data, but it's costly and time consuming to put it all together.  And there are so many possible collars that it's difficult to know which to produce.  But I would love to see the CBOE continue to produce indexes that pull all the data together.  And I'm encouraged that the OIC took part in this study, and I encourage them to support additional studies.

Bottom Line: This paper is very welcome and shows that collars are not always too expensive to pay for themselves and are capable of producing market out-performing returns.  I just wish we had more data.


8 Responses to Can You Beat the Market? Part V. Collars Outperform Buy and Hold

  1. 09/23/2009 at 12:52 PM #

    Hi Mark: Let’s say I bought an option on a $100,000 house for $2,000. If the house increases in value to $120,000 and the purchaser of the option decides to purchase the house, which of the following is the gain due to having purchased the option:
    1) $20,000
    2) $18,000 – $20,000 minus the cost of the $2,000 option
    3) It could go either way depending on how the contract was written. If he only owed $98,000 more because of the $2,000 option, the answer is $20,000. If he owes the total $100,000 to purchase the house, the gain is $18,000.
    In option contracts, which of the above solutions would be most common?

  2. Mark Wolfinger 09/23/2009 at 1:13 PM #

    # 2 The option becomes worth $20,000. Subtract the cost to get the profit.
    # 3 NEVER applies. The cost of an options is NOT a down payment. Thus, in your example, the option owner does not owe an additional $98,000. He/she owes the full strike price, or $100,000.
    The option gives it’s owner the right to buy at the strike price. Not at a discounted price.

  3. RS 09/23/2009 at 1:17 PM #

    Hi Mark,
    I agree that the data during the bubble burst looks suspect, and await the link to the paper to see their methodology.
    They certainly did cherry-pick the data, although it matches the inception of the PowerShares QQQ. If you look at the NDX for the previous 10 years, say beginning of 1989 to beginning of 1999, the index gained over 900%. Because of the long-term tendency for stock indexes to go up, I’m not convinced that a passive collar strategy would out-perform a buy and hold strategy (although I don’t know on a risk-adjusted basis). But it certainly decreases risk and volatility, and should be considered as an alternative.

  4. Mark Wolfinger 09/23/2009 at 1:46 PM #

    I’ll look for that link shortly.
    But I have the methodology now. I just spoke with someone in the institutional department at the OIC, which was behind this study.
    The methodology is set in stone. And that’s usually a good diea for such a study. The put is purchased and held until expiration. That’s six months. When the markets fell, the put delta moved near 100, and that means no downside losses.
    But, they continued to write 2% OTM calls every month – and with a high implied volatility, they received dcent premium. Those sales represented the profits.
    It’s not right for a collar to be profitable during down markts, but with this methodology (details in the full paper) it’s obvious why they profited during the decline: They were ssentially naked short calls.
    There’s nothing to be done about the data range. For PowerShares QQQ, 1999 was, as you say, as far back as they could go.
    If you look at parts III & IV of this series, you will see the data that supports your view. The NDX increased so rapidly that no collar strategy could come close to matching the performance.
    I agree: An alternative. That’s the idea. Alternative risk management techniques are needed and the collar seems to be a good place to begin.
    This study is promising, but is flawed by the methodology. The question is: how to produce the perfect collar index?

  5. RS 09/23/2009 at 2:56 PM #

    I went to the CBOE website to look at the CLL paper. It was interesting to see the Risk-Return Tradeoff chart for the various Collars (1986-2007). The closer the call got to 100, the worse the return but with less volatility. The further away the put got from 100, the better the return but with more risk. The closest they got to a 95-100 was a 95-102, which was the most stable but achieved relatively low returns. Also, I would have liked to have seen a 1-month Put comparison.
    I don’t think there is a perfect collar index, but the chart was useful for comparison. I think an active collar approach would perform better, but that’s an entirely different story.

  6. Mark Wolfinger 09/23/2009 at 4:11 PM #

    Agree. Active management requires a diffrent mindset, plus the ability to spend the necessary time.
    On a personal level, I’m working on getting passive investors to see the benefits of using collars. Not an easy task.

  7. Mark McCracken 09/23/2009 at 7:29 PM #

    Hi Mark –
    I found your article and the paper to be quite interesting. As a quick check, I attempted to replicate their returns for the Sep07-May09 period using ThinkOrSwim’s thinkBack feature (their QQQQ data only goes back to 2005, so unfortunately I couldn’t look at the 1999-2002 timeframe). I found TOS’s P/L calculations to be confusing, so I ended up using it for price discovery and simply tallying up the cash flows in Excel. Sure enough, using a slightly modified (purely for convenience) strategy of selling 2-strike OTM 1-month calls and buying 2-3-strike OTM 6-month puts on QQQQ whilst staying long the underlying, repurchasing after assignment as necessary, I verified about -1% annual return (vs -21% for buy/hold). To save time, all my transactions were done with closing bid/ask prices on the options and closing last prices on QQQQ. All calls were written and puts were bought at market close on the last day of trading for the previous month (for example: Jun08 call was written at the final bid price on 5/16/08).
    The naked-call aspect of the 1999-2002 time period is clearly visible on the graph from the angle of the various put expiries. A 1-month put purchased monthly for 3 years is going to cost a lot more TV than a 6-month put purchased semi-annually for the same time period. I would expect if they had cranked the put dials out to a year or more we would see even clearer positive correlation between put expiry time and 1999-2002 returns. The bigger the disparity between the short call expiry and the long put expiry, the more “free” premium collected during falling markets.
    It seems to me that a collar strategy is just a synthetic short put (short call + long stock) plus a long put – so couldn’t we replicate the strategy with put diagonals and avoid the trading costs of owning stock in addition to writing calls? I confess I don’t have a good sense of the risk profiles of diagonals, never having studied them. In this particular case, we would need to re-leg into the near short put monthly, leaving the long far put alone, right?
    I’d love to see:
    * their actual trades – this is the ultimate dataset needed to verify their results
    * information on why they chose QQQQ – why not some other ETF?
    * similar studies using verticals or diagonals – it could go farther back if we broke from ETFs and traded SPX or something
    * your thoughts on the relative merits of verticals vs collared stock positions vs diagonals
    Thanks for the day-of-release post. This was really interesting and timely.
    – Mark

  8. Mark Wolfinger 09/23/2009 at 8:19 PM #

    Thanks Mark,
    1) Nice contribution. Verification is always a good idea. But as you say, we must see the actual trades.
    2) You are correct about the synthetic equivalent of the collar. For a standard collar, with both options expiring in the same month, the equivalent position is short one put spread.
    Beacuse of the diagonal nature of the collar used in the study, the synthetic equivalent is the diagonal put spread in which we sell the 2% ITM front-month put and buy the 2% OTM 6-month put. Yes, that can be replicated.
    [As an aside, I want to discuss collars again and again – from different persepectives. I hope to encourge their use by the average individual investor. Only later would I introduce them to the equivalency of selling put spreads. I just know that trying to convince people who are unfamiliar with options – that selling put spreads – would be a non-starter.]
    3) One can always ask the question: why did you choose this stock or ETF for any study, and because I have not had the patience to read the entire paper yet, I don’t know if that question was answered.
    4) When you know the answers in advance, you can choose the study that gives the answer you want to see. That’s a big problem. I trust the integrity of the OIC and there is no reason to doubt the seriousness of the professor and his student. But, who knows what was in their minds when they began?
    5) As you said, in falling markets the 6-month put is better than shorter-term. In the CBOE study that produced the CLL 95-110 index, they showed that the 3-month put performed better than the 6-month put when the markets rose (6/86 thru 12/07). Those findings are consistent.
    Maybe we need to see the plain vanilla collar – one month calls, one month puts. Otherwise, there is a market bias built into any collar: Longer term puts for the bears and short-er term puts for the bulls. And (from my standpoint) if the reason to adopt the collar is to gain protection – in other words, fear of the downside – which put should be bought?
    There is much that can be learned from the data.
    6) Right now I have one big problem with this new study, and that’s the outstanding performance of the collar strategy when the marekts were falling sharply. In theory, the collar is supposed to minimize losses, not show profits. Yet this study shows very positive results during a servere decline – FOR A STRATEGY THAT IS SLIGHTLY BULLISH. Those profits make me uncomfortable. From whence came those profits: selling calls. Who would sell ATM calls when protection is a synthetic call that’s substantialy OTM. Not an average investor. That would be a bearish and risky trade. And collar traders seek to avoid risk.
    Those profitable call sales severely limit the usefulness of this study, IMHO. But it does show that collars are not always too costly to consider as portfolio insurance.
    I anticipate much dicussion to come. Thanks for sharing your ideas.