What I Learned From Shorting

The recent market downturn is the first downturn I’ve known since I developed an active interest in investing. I was around for the dot-com bubble burst in 2001, and I lived in Brazil during the currency crises of the late 1990s, but at that time I was recently out of graduate school and didn’t have anything to invest myself. I watched it from the sidelines as others went through it. After missing so much of the dot-com bubble’s price rise, I confess it was a bit of a relief not to be in for the prices falling. So the recent rough winds are the first opportunities I’ve had where I decided to exercise the short end of the macro investing stick.

One reads in texts that taking a short position on a security is not just the opposite as going long, but it really does take experience to get a feel for the differences. Part of the problem is that oftentimes one’s shorting recommendations are simply results of mean-variance or other optimization methods, which produces a list of optimal weightings, and the short suggestions are simply negative. From there, it’s a simple matter of multiplying your Net Asset Value (NAV) to get your target dollar exposure, and then dividing by share or security prices to figure out the quantity to go long or short. At some point, you rebalance, move back to your target weights, and the process starts over again. This system works reasonably well, but it hides some of the subtleties of short positions. Short positions really do feel just like “negative long positions” when you’re just following what the optimizer tells you to do.

The Problem

In my own portfolio, I was comfortable from my macro analysis that my stock and fund picks would outperform the market in the coming months, but I wasn’t confident that an overall market decline wouldn’t take everything down with it. Essentially, I wanted to be market neutral; I sensed rough waters ahead, and I wanted to sleep better at night, knowing I was reasonably hedged against a general decline, even if I thought my bets would do better than most. The solution: short a market index ETF to bring my portfolio beta to zero (in theory I could have done this with a future, but my account was not set up for futures). That way, no matter what panic came from Wall Street, my expected return should be zero, and if my picks did better than the market as a whole, the portfolio would grow.

But how much to short? This became a surprisingly complex problem. The mathematical answer was to short the number of shares sufficient to bring the portfolio beta to zero (or close). Those mechanics were fairly simple and could be done in Excel by multiplying exposure weights by asset betas and solving for the weight that would bring beta closest to zero. Fair enough. But the dollar exposure that came up was about two thirds of my total NAV.

I considered three options: 1) buy shares of a “short ETF” indexed to the S&P 500 whose share values rise when the market values fall; 2) short a long ETF indexed to the S&P; and 3) short an “Ultra Long” ETF that is levered to produce 2x the percentage change in the S&P. Each option had interesting features.

Buying a “Short ETF”

Because I was concerned about my shorting experience and knew there were extra complexities to shorting, I considered the short ETF version first (ProShares Short S&P 500; ticker: SH). This ETF got me the exposure I desired, and it would behave pretty much like a long position, except that it would increase in value when the market was falling. The one problem is that I needed enough cash to purchase the right number of shares and didn’t have enough dry powder available to buy the right amount of the short ETF.

[extra analysis: the Short ETF is an excellent option for those who, for whatever reason, don’t have the ability to short, but need short exposure. The main drawback is that it freezes up your capital faster, and that the expense ratio tends to be a little higher. The first drawback is not really a drawback, as we will see later.]

Shorting an index ETF

I next considered shorting a regular S&P index ETF (iShares S&P 500 index; ticker: IVV). This had a low expense ratio, and I could short the required number of shares with virtually no cash outlay. This sounded great, but surely there must be some hidden risk here. If it was too expensive for me to purchase SH, why was it essentially free for me to take a short position on IVV? Whereas many traders and investors might have simply said “great, it’s almost costless, let’s just do that,” I felt there was some risk I was missing here, and I needed to understand the difference.

Shorting a levered index ETF

Finally I looked at shorting an UltraLong Index (ProShares Ultra S&P 500; ticker SSO). This fund is levered to give a beta of 2, so I could get market neutral with only 1/2 as much dollar exposure, and it was also essentially costless to enter, like shoring the unlevered fund. This dollar exposure looked like something I could handle better, and it met my goals, so I ultimately went with SSO because I felt I couldn’t wait any longer.

Results

Given how things worked out, it was good that I made that decision when I did, because it would have been very bad to wait much longer. However, it turns out that there isn’t much difference in the risk of 2x dollar exposure in an unlevered ETF and 1/2 the dollar exposure in a 2x levered ETF, and neither required any cash outlay. I instinctively felt that something like that was true, but hadn’t yet found the right way to think about it clearly. In fact, given that the levered fund might have greater tracking error, and that leverage often risks strange unintended consequences, it might have been marginally more sensible to go with shorting the unlevered ETF, with its lower expense ratio and tighter index tracking.

Lessons from the Short Side in Risk

The reason that it is essentially costless to enter a short position (other than transaction costs) is because at the moment of sale, one still has the proceeds of the short sale to cover one’s position. This money is collateral that secures your loan and your ability to buy back shares when you cover. If you wanted to reverse your position right away, all you’d need is to pay for another set of transaction costs, and whatever (hopefully) small price movement had happened in the meantime. This is why entering a short position does not *appear* to eat into your NAV. In fact, you might wonder: if it doesn’t cost anything net to enter a short position, why do short-only hedge funds - assuming they are skilled - even need to seek capital?

The answer is that your capital base should not be thought of as determining how much return you can grab, but how much risk you can take. Obviously, risk and return are linked in markets, but this principle emphasizes the oft-made point that one’s allocation decisions really need to start from the risk perspective. And this is why short-only hedge funds need to accumulate capital: so they can take a larger number or size of [positive expected outcome] risks.

This need for this principle is much clearer in the case of shorting than in taking long positions. In a long-only portfolio, when you enter a position, you have clearly “used” a portion of your NAV to express a view, and there is obviously less left over to use to enter other positions. The long-only portfolio has a (partially) self-enforcing check to assist in allocating wisely. But when you enter a short position, none of your NAV appears to have been “used up.” Indeed, in theory, one could use the short proceeds of one position to enter a long position in another, with no net cash outlay (most brokerages won’t let you do this without substantial collateral, however). As a result, other than the brokerage’s margin requirements, it is not intuitively obvious what the limits of sensible short exposure are, since there are no self-checking mechanisms (hence the need for margin on short positions that you wouldn’t need for comparable unlevered long positions). It was this lack of a check that signaled that something was very different here.

The need for the risk approach becomes quickly apparent, however, when positions move against you. In this case, long and short positions have very similar effects on NAV: a wrong bet reduces your NAV by (exposure)x(price_change), whether it cost you anything to enter the position or not. In both cases, your fund NAV indicates either 1) how long you can hold out against an adverse price move, should you decide to hold to your views, or 2) how many losses you can afford to endure, if you decide to get out. This interpretation holds whether your position is long or short. And this risk-taking-ability is what gets “used up” when you enter a position, whether long or short.

“It’s the volatility, stupid...”

The cost of entering a position turns out to be (almost) irrelevant, except to the extent that transaction costs eat away at total return. What really matters is how much risk your position contributes relative to your NAV, and whether that risk is a) within your tolerance and b) the best expected return available for that risk. There are two ways of looking at this risk: 1) on an asset-by-asset basis, and 2) on the portfolio-wide basis. Ultimately, the portfolio-wide basis is most important, but the asset-by-asset approach is informative too, and this is what I had been missing until now. The portfolio-wide risk includes correlation analysis and is typically managed by optimization techniques, but the way that individual assets contribute to that risk, and what risks are sensible or not sensible to take are obscured by the optimization process. That can lead investors who use them to blindly follow their optimizer wherever it goes.

Budgeting one’s risk taking ability in relationship to NAV means that your position size needs to be linked to your investment’s volatility, and not simply the initial cash outlay. This is why shorting half the position size of a 2x levered index ETF turns out to be identical to a full short position in the unlevered ETF. The unlevered ETF requires twice the dollar exposure, but the levered ETF has twice the volatility, so at the end of the day, you are either getting twice the volatility with SSO or twice the exposure with IVV, and the impact of a market move on the portfolio should be the same either way. [In truth, SSO wavers more from its benchmark and has a higher expense ratio, so IVV would have been a marginally better decision]. It’s also true that the risk effect of going short one of these indices is about the same as the effect of going long the inverse S&P ETF after all, except for the potential interest costs of buying SH on margin.

Was I being dumb? No, but now I know why not.

So it turns out that shorting a 2x levered index ETF exposed me to about the same risk that I would have had if I had taken a long position in the unlevered inverse index ETF that I had initially thought I couldn’t afford. Was I wrong to do the short, or was I wrong to conclude I shouldn’t do the long inverse ETF.

Given my objectives, and taken in a portfolio context, shorting was a smart thing to do, because it reduced portfolio risk substantially. The reason it reduced portfolio risk is because there was strong negative correlation between the short position’s returns and the returns on other assets in the portfolio. Just as the costless-ness of entering the short position was a bad indicator of how much I could take, the costliness of the long inverse ETF position was also a bad indicator of a position that should in fact be taken, given macro views, objectives, and correlations between assets.

Risk control without optimizers

As fond as I am of computers (and I am a big technophile), I also have great respect for the clever application of low-technology solutions to the types of problems high technology is commonly used to approach (for one of my favorites, see the Ancient Inca “Khipu”). Jack Schwager in his popular book, The New Market Wizards, points out one method of risk control used by several successful traders: they size their positions in proportion to their total trading capital (i.e. NAV). When NAV goes up, they take larger positions, and when NAV goes down, they take smaller positions. Effectively, each position represents a roughly constant proportion of NAV, say 2% or so. This meant that as a trader or investor rode through bad periods, the size of their (new) bets were automatically reduced. In fact, optimization techniques also use this method, in that optimizers produce portfolio weights as outputs, which are then multiplied by portfolio size to generate positions. By contrast, amateur and professional traders who use constant dollar position sizes would be placing a larger proportion of their capital at risk when they are on a losing streak than when they are winning. This has undesirable properties both from a psychological perspective (you may panic and think less clearly when you are losing at the same time that you are risking more), it also increases the chance of being wiped out, because single bad turns start to have a larger and larger effect on your NAV.

Another solution, revealed by former Turtle Trader Curtiss Faith in his book The Way of the Turtle, is to tune position sizes to NAV and the volatility of the instrument you are trading. Optimizers do this too, but it is harder to get an intuitive feel for how. Tuning position size to individual asset volatility is an even cleverer way to manage risk, and applies as much to long positions as short positions. The idea here is that positions are sized in “units” so that the each unit in a position contributes approximately the same percentage to portfolio’s volatility. The advantage of doing this is that it keeps the investor conscious of how positions contribute overall portfolio risk, and helps create evaluate position sizing in ways that do not depend on acquisition costs.

Additional considerations on Shorting

One of the differences in working from the short side is that when a position moves against you, you become more exposed to it. This means that after a bad day with a short, a second bad day of equal magnitude is going to be more than twice as bad for your NAV, other things equal. The inverse is also true. If a short position makes you money, it will now form a smaller part of your portfolio than before. Handling short positions, therefore, can feel a bit like rotating a gyroscope in your hands - it sort of behaves the way you expect, but not completely. Portfolios that have long and short positions balanced to achieve certain objectives (like a market beta = 0) have to be watched more carefully, because the changing position sizes can throw off the balance. For example, if you have a portfolio balanced so that beta=0 and your long positions have advanced to become a greater part of your portfolio while your short positions have declined (generating profits for you) and become a smaller part of your portfolio, this actually means that your beta is likely to have crept upwards because the short positions are now a smaller part of your portfolio. This means that portfolios with shorts need to be watched and rebalanced more closely, and that large market movements are likely to do stranger things as hedges move out of alignment.

What I Learned From Shorting

The recent market downturn is the first downturn I’ve known since I developed an active interest in investing. I was around for the dot-com bubble burst in 2001, and I lived in Brazil during the currency crises of the late 1990s, but at that time I was recently out of graduate school and didn’t have anything to invest myself. I watched it from the sidelines as others went through it. After missing so much of the dot-com bubble’s price rise, I confess it was a bit of a relief not to be in for the prices falling. So the recent rough winds are the first opportunities I’ve had where I decided to exercise the short end of the macro investing stick.

One reads in texts that taking a short position on a security is not just the opposite as going long, but it really does take experience to get a feel for the differences. Part of the problem is that oftentimes one’s shorting recommendations are simply results of mean-variance or other optimization methods, which produces a list of optimal weightings, and the short suggestions are simply negative. From there, it’s a simple matter of multiplying your Net Asset Value (NAV) to get your target dollar exposure, and then dividing by share or security prices to figure out the quantity to go long or short. At some point, you rebalance, move back to your target weights, and the process starts over again. This system works reasonably well, but it hides some of the subtleties of short positions. Short positions really do feel just like “negative long positions” when you’re just following what the optimizer tells you to do.

The Problem

In my own portfolio, I was comfortable from my macro analysis that my stock and fund picks would outperform the market in the coming months, but I wasn’t confident that an overall market decline wouldn’t take everything down with it. Essentially, I wanted to be market neutral; I sensed rough waters ahead, and I wanted to sleep better at night, knowing I was reasonably hedged against a general decline, even if I thought my bets would do better than most. The solution: short a market index ETF to bring my portfolio beta to zero (in theory I could have done this with a future, but my account was not set up for futures). That way, no matter what panic came from Wall Street, my expected return should be zero, and if my picks did better than the market as a whole, the portfolio would grow.

But how much to short? This became a surprisingly complex problem. The mathematical answer was to short the number of shares sufficient to bring the portfolio beta to zero (or close). Those mechanics were fairly simple and could be done in Excel by multiplying exposure weights by asset betas and solving for the weight that would bring beta closest to zero. Fair enough. But the dollar exposure that came up was about two thirds of my total NAV.

I considered three options: 1) buy shares of a “short ETF” indexed to the S&P 500 whose share values rise when the market values fall; 2) short a long ETF indexed to the S&P; and 3) short an “Ultra Long” ETF that is levered to produce 2x the percentage change in the S&P. Each option had interesting features.

Buying a “Short ETF”

Because I was concerned about my shorting experience and knew there were extra complexities to shorting, I considered the short ETF version first (ProShares Short S&P 500; ticker: SH). This ETF got me the exposure I desired, and it would behave pretty much like a long position, except that it would increase in value when the market was falling. The one problem is that I needed enough cash to purchase the right number of shares and didn’t have enough dry powder available to buy the right amount of the short ETF.

[extra analysis: the Short ETF is an excellent option for those who, for whatever reason, don’t have the ability to short, but need short exposure. The main drawback is that it freezes up your capital faster, and that the expense ratio tends to be a little higher. The first drawback is not really a drawback, as we will see later.]

Shorting an index ETF

I next considered shorting a regular S&P index ETF (iShares S&P 500 index; ticker: IVV). This had a low expense ratio, and I could short the required number of shares with virtually no cash outlay. This sounded great, but surely there must be some hidden risk here. If it was too expensive for me to purchase SH, why was it essentially free for me to take a short position on IVV? Whereas many traders and investors might have simply said “great, it’s almost costless, let’s just do that,” I felt there was some risk I was missing here, and I needed to understand the difference.

Shorting a levered index ETF

Finally I looked at shorting an UltraLong Index (ProShares Ultra S&P 500; ticker SSO). This fund is levered to give a beta of 2, so I could get market neutral with only 1/2 as much dollar exposure, and it was also essentially costless to enter, like shoring the unlevered fund. This dollar exposure looked like something I could handle better, and it met my goals, so I ultimately went with SSO because I felt I couldn’t wait any longer.

Results

Given how things worked out, it was good that I made that decision when I did, because it would have been very bad to wait much longer. However, it turns out that there isn’t much difference in the risk of 2x dollar exposure in an unlevered ETF and 1/2 the dollar exposure in a 2x levered ETF, and neither required any cash outlay. I instinctively felt that something like that was true, but hadn’t yet found the right way to think about it clearly. In fact, given that the levered fund might have greater tracking error, and that leverage often risks strange unintended consequences, it might have been marginally more sensible to go with shorting the unlevered ETF, with its lower expense ratio and tighter index tracking.

Lessons from the Short Side in Risk

The reason that it is essentially costless to enter a short position (other than transaction costs) is because at the moment of sale, one still has the proceeds of the short sale to cover one’s position. This money is collateral that secures your loan and your ability to buy back shares when you cover. If you wanted to reverse your position right away, all you’d need is to pay for another set of transaction costs, and whatever (hopefully) small price movement had happened in the meantime. This is why entering a short position does not *appear* to eat into your NAV. In fact, you might wonder: if it doesn’t cost anything net to enter a short position, why do short-only hedge funds - assuming they are skilled - even need to seek capital?

The answer is that your capital base should not be thought of as determining how much return you can grab, but how much risk you can take. Obviously, risk and return are linked in markets, but this principle emphasizes the oft-made point that one’s allocation decisions really need to start from the risk perspective. And this is why short-only hedge funds need to accumulate capital: so they can take a larger number or size of [positive expected outcome] risks.

This need for this principle is much clearer in the case of shorting than in taking long positions. In a long-only portfolio, when you enter a position, you have clearly “used” a portion of your NAV to express a view, and there is obviously less left over to use to enter other positions. The long-only portfolio has a (partially) self-enforcing check to assist in allocating wisely. But when you enter a short position, none of your NAV appears to have been “used up.” Indeed, in theory, one could use the short proceeds of one position to enter a long position in another, with no net cash outlay (most brokerages won’t let you do this without substantial collateral, however). As a result, other than the brokerage’s margin requirements, it is not intuitively obvious what the limits of sensible short exposure are, since there are no self-checking mechanisms (hence the need for margin on short positions that you wouldn’t need for comparable unlevered long positions). It was this lack of a check that signaled that something was very different here.

The need for the risk approach becomes quickly apparent, however, when positions move against you. In this case, long and short positions have very similar effects on NAV: a wrong bet reduces your NAV by (exposure)x(price_change), whether it cost you anything to enter the position or not. In both cases, your fund NAV indicates either 1) how long you can hold out against an adverse price move, should you decide to hold to your views, or 2) how many losses you can afford to endure, if you decide to get out. This interpretation holds whether your position is long or short. And this risk-taking-ability is what gets “used up” when you enter a position, whether long or short.

“It’s the volatility, stupid...”

The cost of entering a position turns out to be (almost) irrelevant, except to the extent that transaction costs eat away at total return. What really matters is how much risk your position contributes relative to your NAV, and whether that risk is a) within your tolerance and b) the best expected return available for that risk. There are two ways of looking at this risk: 1) on an asset-by-asset basis, and 2) on the portfolio-wide basis. Ultimately, the portfolio-wide basis is most important, but the asset-by-asset approach is informative too, and this is what I had been missing until now. The portfolio-wide risk includes correlation analysis and is typically managed by optimization techniques, but the way that individual assets contribute to that risk, and what risks are sensible or not sensible to take are obscured by the optimization process. That can lead investors who use them to blindly follow their optimizer wherever it goes.

Budgeting one’s risk taking ability in relationship to NAV means that your position size needs to be linked to your investment’s volatility, and not simply the initial cash outlay. This is why shorting half the position size of a 2x levered index ETF turns out to be identical to a full short position in the unlevered ETF. The unlevered ETF requires twice the dollar exposure, but the levered ETF has twice the volatility, so at the end of the day, you are either getting twice the volatility with SSO or twice the exposure with IVV, and the impact of a market move on the portfolio should be the same either way. [In truth, SSO wavers more from its benchmark and has a higher expense ratio, so IVV would have been a marginally better decision]. It’s also true that the risk effect of going short one of these indices is about the same as the effect of going long the inverse S&P ETF after all, except for the potential interest costs of buying SH on margin.

Was I being dumb? No, but now I know why not.

So it turns out that shorting a 2x levered index ETF exposed me to about the same risk that I would have had if I had taken a long position in the unlevered inverse index ETF that I had initially thought I couldn’t afford. Was I wrong to do the short, or was I wrong to conclude I shouldn’t do the long inverse ETF.

Given my objectives, and taken in a portfolio context, shorting was a smart thing to do, because it reduced portfolio risk substantially. The reason it reduced portfolio risk is because there was strong negative correlation between the short position’s returns and the returns on other assets in the portfolio. Just as the costless-ness of entering the short position was a bad indicator of how much I could take, the costliness of the long inverse ETF position was also a bad indicator of a position that should in fact be taken, given macro views, objectives, and correlations between assets.

Risk control without optimizers

As fond as I am of computers (and I am a big technophile), I also have great respect for the clever application of low-technology solutions to the types of problems high technology is commonly used to approach (for one of my favorites, see the Ancient Inca “Khipu”). Jack Schwager in his popular book, The New Market Wizards, points out one method of risk control used by several successful traders: they size their positions in proportion to their total trading capital (i.e. NAV). When NAV goes up, they take larger positions, and when NAV goes down, they take smaller positions. Effectively, each position represents a roughly constant proportion of NAV, say 2% or so. This meant that as a trader or investor rode through bad periods, the size of their (new) bets were automatically reduced. In fact, optimization techniques also use this method, in that optimizers produce portfolio weights as outputs, which are then multiplied by portfolio size to generate positions. By contrast, amateur and professional traders who use constant dollar position sizes would be placing a larger proportion of their capital at risk when they are on a losing streak than when they are winning. This has undesirable properties both from a psychological perspective (you may panic and think less clearly when you are losing at the same time that you are risking more), it also increases the chance of being wiped out, because single bad turns start to have a larger and larger effect on your NAV.

Another solution, revealed by former Turtle Trader Curtiss Faith in his book The Way of the Turtle, is to tune position sizes to NAV and the volatility of the instrument you are trading. Optimizers do this too, but it is harder to get an intuitive feel for how. Tuning position size to individual asset volatility is an even cleverer way to manage risk, and applies as much to long positions as short positions. The idea here is that positions are sized in “units” so that the each unit in a position contributes approximately the same percentage to portfolio’s volatility. The advantage of doing this is that it keeps the investor conscious of how positions contribute overall portfolio risk, and helps create evaluate position sizing in ways that do not depend on acquisition costs.

Additional considerations on Shorting

One of the differences in working from the short side is that when a position moves against you, you become more exposed to it. This means that after a bad day with a short, a second bad day of equal magnitude is going to be more than twice as bad for your NAV, other things equal. The inverse is also true. If a short position makes you money, it will now form a smaller part of your portfolio than before. Handling short positions, therefore, can feel a bit like rotating a gyroscope in your hands - it sort of behaves the way you expect, but not completely. Portfolios that have long and short positions balanced to achieve certain objectives (like a market beta = 0) have to be watched more carefully, because the changing position sizes can throw off the balance. For example, if you have a portfolio balanced so that beta=0 and your long positions have advanced to become a greater part of your portfolio while your short positions have declined (generating profits for you) and become a smaller part of your portfolio, this actually means that your beta is likely to have crept upwards because the short positions are now a smaller part of your portfolio. This means that portfolios with shorts need to be watched and rebalanced more closely, and that large market movements are likely to do stranger things as hedges move out of alignment.