In this presentation, I want to first separate risk

Transcription

1 Utilizing Downside Risk Measures Michelle McCarthy Managing Director and Head of Risk Management Nuveen Investments Chicago Investment advisers and fund managers could better outperform relevant benchmarks on a risk-adjusted basis by analyzing differently their current and prospective client portfolios. This improved performance can be achieved by focusing primarily on downside risk measures and understanding whether portfolios exhibit asymmetrical profiles with fat tails. In addition, certain classic risk management strategies are useful, such as the Sharpe ratio and the information ratio, whereas other measures, such as the Sortino ratio and semi-standard deviation, may be misleading during up-market cycles. In this presentation, I want to first separate risk measures from performance measures because they are often confused in people s minds. Then I will discuss classic measures, such as the Sharpe ratio, the information ratio, and downside capture, because they are considered the key risk-adjusted performance measures to use to assess the past s of a portfolio or fund. Next, I will discuss two conditions that are important in assessing downside risk volatility and drawdown. Finally, I will discuss how to implement a risk management strategy for clients portfolios, particularly focusing on value at risk (VaR) and ex ante tracking error. Performance vs. Risk Measures: Looking Backward and Looking Forward Measures that are created by analyzing the past performance of a portfolio can be used to assess what may happen in the future for that portfolio unless strategies or markets change in meaningful ways. When meaningful change occurs, the backwardlooking measures are not really helpful. There are significant differences between performance measures and risk measures. Performance Measures. If what is being measured has already happened, then it is performance, not risk. If a measure is conducted on today s holdings, it represents a forward-looking performance measure. But if it is conducted on past s of a fund, then it represents a backwardlooking performance measure. For example, past performance of a fund is typically a good guide This presentation comes from the Wealth Management Conference held in Garden Grove, California, on February 2014 in partnership with CFA Society Orange County. to understanding the future, but it will not alert an investor when the strategy of the portfolio has changed in a meaningful way or the manager has tried something new. In those situations, past performance is not going to help in understanding the future. If the markets change in meaningful ways from historical patterns, then backward-looking performance measures are not strong indicators for understanding future risks. Measures applied to the past series of a fund are what I would consider performance variability measures rather than risk measures. The distinction between performance measures and risk measures is important for several reasons. First, performance variability measures can show portfolio manager skill, whereas risk measures do not. If managers are able to buy lower and sell higher than their peers, their portfolios will have a lower volatility and better standard deviation of s. Second, when portfolio composition changes significantly, performance measures will reflect the change in portfolio composition slowly, whereas the risk measures will reflect the change in portfolio composition instantly. To protect equity portfolios from future downside, risk measures are more appropriate. Risk measures will also help determine whether the manager has changed the holdings in some particular way from the past. But if portfolio strategies do not change, performance and risk measures are similar. The classic performance variability measures are standard deviation, ex post tracking error (i.e., the historical variability of performance relative to the benchmark), the Sharpe ratio, the information ratio, such variants on the Sharpe ratio as the Treynor ratio and Jensen s alpha, downside measures (e.g., the Sortino ratio and downside capture), and most beta measures. Beta calculations 42 Third Quarter CFA Institute cfapubs.org

2 Utilizing Downside Risk Measures can represent either performance or risk measures depending on how they are calculated. For example, backward-looking beta calculations based on the historical portfolio s s versus the market s past s represent performance measures. Forward-looking beta calculations based on the prior s of the securities currently held in the portfolio versus the market represent risk measures. Risk Measures. Classic names for risk measures include VaR, ex ante tracking error, and on risk-adjusted capital. VaR is a statistical technique that, when used by banks, is designed to quantify how much they could lose as a probable worst outcome in 1 trading day or 10 trading days. When used by asset managers, however, it can instead be used to quantify how much a portfolio could lose over, for instance, an ordinary year. Ex ante tracking error is going to be the potential underperformance of the benchmark. The on risk-adjusted capital is defined as net income divided by the allocated risk capital and is typically used by banks or leveraged market participants rather than by traditional asset managers. The allocated risk capital represents a firm s portfolio capital adjusted for a maximum potential loss based on the probability of future s or volatility of earnings. Risk measures provide an early warning when portfolios have changed significantly and before the change is crystallized in performance. Risk measures are often bound by the same market history as performance variability measures, with the exception of a family of measures called scenario risk measures. Investment analysts often cannot predict whether a portfolio will gain or lose, but they can show which portfolios have wider ranges of potential s and thus a greater potential for loss. Performance measures are more readily available; risk measures are more difficult to come by but are becoming increasingly available. Understanding Key Risk-Adjusted Performance Measures Now I want to discuss classic performance variability measures and how to use them for managing risk within client portfolios. I will focus on the Sharpe ratio and information ratio. Sharpe Ratio. The Sharpe ratio is defined as the annual portfolio in excess of the risk-free rate divided by its variability during the period. Its variability is quantified in terms of the standard deviation of s measured daily, weekly, or monthly over the same past period. The formula for the Sharpe ratio is Portfolio Risk-free rate Standard deviation of portfolio. The Sharpe ratio provides a measure of the quality of absolute performance. It is beneficial to compare with the same statistic shown for the peer benchmark. The goal is to maintain a Sharpe ratio of greater than 1.0. This measure is sometimes hard to achieve, but based on the statistics available for peer sets, greater than one is considered good. Information Ratio. The information ratio is defined as how much a portfolio ed in excess of its benchmark divided by how much it could have underperformed (i.e., the ex post tracking error). The formula for the information ratio is Portfolio Benchmark Standard deviation of benchmark relative. The information ratio provides a measure of the quality of relative performance. The goal is an information ratio of greater than In some efficient or competitive markets for example, large-cap US equities a ratio greater than the range may be excellent. For absolute strategies, the Sharpe ratio and information ratio are equivalent. In the industry, managers typically use the 0.5 level (as opposed to 1.0) as the standard rule of thumb for a good ratio, even though their benchmark is cash. Downside Capture. Sometimes managers provide downside capture, which measures a portfolio s when market benchmark s are less than zero. Downside capture helps quantify what percentage of down months for the benchmark are down months for the portfolio. Downside capture is calculated by dividing the cumulative of the portfolio in the period when the corresponding benchmark is less than zero by the cumulative of the benchmark in periods when the is less than zero. A less risky portfolio should have less downside capture than its benchmark. So, if a portfolio has a of 9% and the benchmark s is 11%, the downside capture is 82% because the portfolio had less loss during the down periods than the benchmark. Other Measures. The Treynor ratio, Jensen s alpha, and semi-standard deviation focus on subsets of volatility, such as only downside volatility. They may also allow a manager to be more precise about which risk-free rate and which portion of volatility (systematic versus total) is used in the computation. But I am skeptical about the use of measures that focus only on the down periods. I believe that both up and down periods provide a 2014 CFA Institute cfapubs.org Third Quarter

3 CFA Institute Conference Proceedings Quarterly more objective view, and a rapid increase in fund s can also be indicative of risk. This view was particularly true during the period of the internet bubble in when the equity markets, particularly internet stocks, rocketed upward but showed few downward movements until the end of the bubble. I would argue that semi-standard deviation was misused to understate the risk of this upward-trending market but that standard deviation correctly showed a high level of variability in s during the period. Volatility and Drawdown For illustration purposes, Figure 1 shows two hypothetical funds performance relative to a benchmark. Fund A mirrors the benchmark fairly well and actually slightly exceeds the benchmark toward the end of the time horizon. Fund B has much greater volatility but ends at the same value as Fund A at the end of the time horizon. If an investor purchased Fund B with the expectation that it would have markedly different performance than the benchmark and have the opportunity for outperformance, then Fund B has those characteristics; it is not objectively a poor investment. But if the investment period had been shortened, Fund B could have realized a large loss relative to the benchmark. Volatility matters, particularly when investors need to withdraw funds or sell a fund holding before the long-term expected is achieved. Table 1 gives the annual, standard deviation of, and Sharpe ratio for the two funds and benchmark shown in Figure 1. Fund A, which tracked closely with the benchmark, outperformed the benchmark annual and had a higher Sharpe ratio of 1.37 compared with the benchmark s Sharpe ratio of Fund A was actually more moderate than the benchmark, as its lower standard deviation of shows, but still managed to outperform it. Fund B had more than three times the volatility of the benchmark but outperformed the benchmark annual. Fund B s Sharpe ratio was 0.35 compared with the Table 1. Return and Volatility Information for Hypothetical Funds and a Benchmark Fund A Fund B Benchmark C Annual 6.0% 6.0% 5.2% Standard deviation 4.2% 16.3% 5.0% of Sharpe ratio Benchmark relative 0.65% 0.65% Standard deviation 3.1% 16.0% of benchmark relative Information ratio Figure 1. Performance of Hypothetical Funds and a Benchmark Value Time Fund A Fund B Benchmark C 44 Third Quarter CFA Institute cfapubs.org

4 Utilizing Downside Risk Measures benchmark s ratio of So, both funds outperformed by 65 bps relative to the benchmark, but their standard deviations relative to the benchmark are very different. The ex post tracking error, which is the standard deviation of the benchmark relative, equals 3.1% for Fund A and 16% for Fund B. In other words, Fund B had the volatility to either outperform or underperform the benchmark by 16%. Their information ratios are also both positive: Fund A is 0.21 and Fund B is To illustrate the importance of the endpoint, I changed the hypothetical example only slightly so that Fund A s performance goes down to slightly below the benchmark right at the end of the time series. Table 2 shows the updated information. Despite mostly tracking the benchmark during the period, Fund A underperformed by 12 bps over the full period. So, in the final month, Fund A forfeits a small amount of its benchmark relative performance and thus underperforms for the full period, and its information ratio changes to negative. It is important to remember that these risk measures are highly dependent on the endpoint and best paired with the full series for the greatest insight. In addition, smaller Sharpe and information ratios have poor discriminatory power. If these ratios are close to the significant levels mentioned earlier (e.g., a Sharpe ratio of 1.0 or an information ratio of ), then they should reasonably indicate strong performance, but levels around zero may represent more noise rather than provide an accurate signal. Another issue for analysts is to beware of measures that average extreme periods, such as , with dull periods. The results are often misleading because of offsetting s within the periods. For example, risk calculations that include the financial crisis period during 2008 will likely give an extreme statistic, especially if it includes both 2008 and Analysts may not realize that the downside Table 2. Return and Volatility Information for Hypothetical Funds and a Benchmark Fund A Fund B Benchmark C Annual 5.1% 6.0% 5.2% Standard deviation 4.2% 16.3% 5.0% of Sharpe ratio Benchmark relative 0.12% 0.65% Standard deviation 3.2% 16.0% of benchmark relative Information ratio and upside were so extreme because they offset one another. Finally, as small numbers converge toward zero, they have less explanatory power for client portfolio calculations. Small numbers are indicating less and are more likely representing noise in the statistics. But a strong and large positive number has more explanatory power than small numbers that vary around zero. For example, the small negative information ratio for Fund A in Table 2 does not tell the whole story. Volatility is meaningful, but it is not the only metric for assessing past performance. When assets have a symmetrical distribution, a normal distribution can be used to understand the possible range of their s. To estimate the worst-case scenarios and to be prepared for the effects of a bad day, a manager can multiply the standard deviation by a known Z-score. But when an asset has an asymmetrical profile, also called fat tails, managers cannot estimate its potential losses by knowing its standard deviation alone. Thus, it is important to understand which assets have asymmetrical distributions. Asymmetrical Returns. Asymmetrical s come from such elements that can increase downside risk as credit default risk, selling (writing) options, long short or hedging strategies that rely on constant correlation, illiquid markets, overconcentrated positions, and positions too large for the market liquidity. Credit default risk increases the s on many fixed-income funds, but it also causes a fat left tail and makes distributions asymmetrical. Selling options increases the left side of the tail s (losses) and does not increase the right side of positive distributions (gains) beyond the premium income earned for writing the options. Long short or hedging strategies that rely on constant correlation do not mathematically create a bigger left tail the way sold options do. These strategies combine pairs of trades or one hedge after another, and if the correlations between the pairs change, it can create unusual losses or gains. For example, hedging strategies when the underlying asset is not hedged effectively can result in larger realized losses. Illiquid markets and over-concentrated positions that are too large for the current market liquidity can be difficult to identify. For some assets, the size of the position relative to daily average trading volume is used as an indicator. For example, if a large fund purchases so much of a popular stock such that its holding represents 90% of the daily average trading volume, the historical standard deviation of the stock could be measured. But if the 2014 CFA Institute cfapubs.org Third Quarter

5 CFA Institute Conference Proceedings Quarterly fund sold that entire stock position in one day, it would not experience that historical standard deviation. It would create a loss because of its outsized position. On the equity side, the rule of thumb is that market participants do not like to represent more than 10% of the daily average trading volume. It is important to know whether a position is too large relative to the market liquidity because it will affect the downside risk when it is time to sell. When liquidity is materially damaged, as it was during the financial crisis, the normal statistics do not apply, and liquidity is the hardest to track. Those in risk management spend a lot of time trying to obtain good metrics about the number of market makers, bid offer spreads, and the daily average trading volume. This information is sought for learning purposes going into the next period of history. Liquidity is like a sold option. Poor liquidity during relatively healthy markets will mean very bad liquidity during a liquidity crisis. If investors purchase an asset that is not liquid, or they are over-concentrated in one that is not liquid, the asset performance will never be better than historical measures of risk because the measures were based on whatever holding sizes are normal for that market, but the asset performance can be a lot worse. Thus, for strategies with asymmetry or fat tails, a metric beyond standard deviation is needed to fully understand downside risk and to describe past s to investors. Pairing Maximum Drawdown with Volatility. When advisers hear about a strategy that combines low standard deviation and high s, then it is time to look for the elements of asymmetrical risk just described to determine whether they are in acceptable proportions or whether the proportions are distorted. A good metric for uncovering the downside risk elements in a fund is to look at the maximum drawdown. Unlike standard deviation, which averages performance over a given period, maximum drawdown highlights the worst period (e.g., day, month, or quarter) of performance for a strategy in a given period. The length of period selected should be whatever is relevant for a manager s client. To fully understand the worst-case scenario for a fund or portfolio, pair volatility (e.g., standard deviation) with an additional maximum drawdown statistic. Using the same hypothetical funds and benchmark from Figure 1 and Table 1, Table 3 shows the maximum one-month drawdown for each. The more volatile strategy, Fund B, has a dramatic down month that should not happen if it was supposed to be a safe fixed-income-type holding. Similarly, relative to the benchmark, it markedly underperformed. Managers and fund families understand drawdown because they know that it is somewhat disingenuous to only disclose standard deviation. Investors should ask, What was the worst single month (or whatever period the investor prefers) for the strategy in the backtests? The maximum drawdown statistic is more revealing for strategies that appear to have a low standard deviation, high s, and asymmetrical profiles, which the funds in Table 3 did not contain. Long-trending bubbles (e.g., the internet stock boom from 1995 to 2000 or the housing bubble from 2003 to 2007) can still defy this measure until they collapse. For asymmetrical strategies, it is good practice to pair a maximum drawdown statistic with standard deviation when describing past performance and simulating future performance. Implementing a Risk Management Strategy Now I will turn from using performance variability numbers to using forward-looking risk measures. VaR and ex ante tracking error are two key forwardlooking risk measures that can be used, along with consideration of client objectives, to implement a risk management strategy in client portfolios. Value at Risk. VaR is the potential loss in net asset value for a given holding period (e.g., 1 day, 10 days, or 1 year), at some confidence interval (e.g., 99%, 95%, 1 downside standard deviation, or 84.15%). For example, banks use short holding periods and high confidence intervals (e.g., 10 days at 99%), whereas asset managers tend to use longer holding periods and lower confidence Table 3. Return, Maximum Drawdown, and Volatility Information for Hypothetical Funds and a Benchmark Fund A Fund B Benchmark C Annual 6.0% 6.0% 5.2% Standard deviation 4.2% 16.3% 5.0% of Maximum onemonth 1.8% 6.1% 2.3% drawdown Sharpe ratio Benchmark relative 0.65% 0.65% Standard deviation 3.1% 16.0% of benchmark relative Maximum 1.9% 7.5% one-month underperformance Information ratio Third Quarter CFA Institute cfapubs.org

6 Utilizing Downside Risk Measures intervals (e.g., 1 year at 84.15% confidence). So, in one use of the VaR number, banks measure how much they could lose in 10 days with 99% confidence and have several multiples of that amount set aside, as mandated by the Basel regime. In asset management, funds are usually provided upfront capital by investors; VaR does not need to be used to measure how much capital is needed. The asset manager s main concern is quantifying potential future losses for a given strategy. Some given number of days of history is used to measure the volatility of market risk factors and the correlations between them; sometimes exponential smoothing is applied. VaR is a flexible number, and it can provide a good forecast of how much a fund could lose in an ordinary year, month, or week. The one-year, 84.15% measure, which is equal to one downside standard deviation, indicates the kinds of losses that are possible at the center of the probability distribution. Expected income can be added to this loss measure to estimate how much a portfolio could lose in an ordinary year; for example, if the strategy could lose 10% in market value according to the VaR measure but is regularly expected to earn 3% in income, the total potential loss in an ordinary year is about 7%. Managers can also approximately double the measure (multiplying a one standard deviation measure by a Z-score of ) to estimate a 99% worst-case year except if the strategy is asymmetrical. Once the VaR is calculated, the next step is to compare it with other portfolios to determine whether the risk is relatively high. Managers can compare the VaR with the same measure for other portfolios of a similar strategy, if available; a portfolio s realized standard deviation of past s; a peer set s standard deviation of past s; or past losses for this or similar portfolios that were deemed unacceptable. For example, how does the VaR calculated based on a one-year, 84.15% confidence interval differ from the standard deviation of realized performance? This measure uses the risk factors in the current portfolio (new portfolio, old markets) rather than the historical portfolio s realized s (old portfolio, old markets). Knowing what is possible in an ordinary year helps advisers begin to understand potential downside and place their clients in investments that have an acceptable range of risk of loss. Ex Ante Tracking Error. Ex ante tracking error is also referred to as relative VaR. It is defined as the potential underperformance of a given portfolio strategy compared with the benchmark, and it is usually calculated based on a one-year period and one standard deviation confidence interval. Ex ante tracking error stands alone as a measure of potential underperformance. Unlike VaR, for which expected portfolio income is subtracted from the measure, there is no regularly expected amount of benchmark outperformance that is subtracted from ex ante tracking error. To determine whether ex ante tracking error is relatively high or whether active risk is relatively low, managers compare the measure with the same measure for comparable portfolios, if available; any standard for the portfolio for example, an unacceptable level of underperformance; the portfolio s ex post realized tracking error; or a peer set s ex post realized tracking error. Ex ante tracking error is different from ex post tracking error in that it uses the risk factors in the current portfolio rather than the realized s of the historical portfolio. Implementing a Risk Management Strategy. To implement a risk management strategy for client portfolios, the most important step is to understand the liabilities or the objectives of each client. The liabilities or objectives of the client should be understood relative to the timing of the required cash flows and what market factors can increase those liabilities. Then it is important to understand a client s tolerance for both annual volatility and shorter-term drawdowns. After these steps are completed, construct a portfolio in which investment values increase when liabilities increase and stay within the volatility and drawdown tolerances. In terms of the clients expectations, ask what they need money for and when they need it. For pension funds, advisers should look at actuarial studies of the timing and magnitude of their liabilities and how variable the outcomes can be. Advisers model their clients liabilities (e.g., liquidity requirements) in terms of timing and match those timing requirements with assets. The sooner the liabilities are required, the less variability the portfolio can tolerate, and the assets required to meet those liabilities will likely be short-term highquality fixed-income securities. The assets required for funding the timespecific liquidity requirements can include the entire spectrum of asset classes. More common assets will be fixed-income securities and equities. Liabilities will represent actual cash flow 2014 CFA Institute cfapubs.org Third Quarter

7 CFA Institute Conference Proceedings Quarterly requirements to be met with the asset classes that the advisers are investing in. An important factor is a high correlation between the investments and the stream of expected payments to ensure that there is a high probability of meeting the client s objectives. For example, if the liability is short term, then it will correlate well with short-term fixed-income investments. But if the liability is long term, it will require the adviser to consider each asset class separately relative to the risk tolerance of the client, as well as to construct a portfolio of investments that can maintain real value. For long-term liabilities, equities, longterm fixed-income securities, and real assets are common asset classes to use. Conclusion I hope I have convinced you to consider using different downside risk measures as part of a risk management strategy. Critically analyze current and prospective portfolios for asymmetrical profiles that exhibit fat tails, and then show both standard deviation and maximum drawdown measures when describing past performance. Also, be sure to understand the limitations of the Sortino ratio and semi-standard deviation when applied during up-market cycles. These practices will enable investment managers to better minimize future losses and optimize risk-adjusted s. This article qualifies for 0.5 CE credit. 48 Third Quarter CFA Institute cfapubs.org

8 Question and Answer Session Michelle McCarthy Q&A: McCarthy Question: Can you provide more detail about implementing a risk management strategy? How useful are derivative overlays? McCarthy: If there is too much risk for the client in the current portfolio, the first step is to assess the asset allocation, which drives most downside risk. There are times when derivative overlays are perfect for portfolio rebalancing. If a long-term fixedincome manager is concerned about interest rates rising and does not want to sell investment positions in the portfolio, then it is possible to use derivatives to hedge the interest rate exposure (e.g., reduce the duration). Sometimes clients or managers prefer to use a derivative overlay to change a portfolio s risk profile without having to sell assets and generate capital gains. To accomplish this, managers may end up having to use large derivative positions, and the overlay will show a lot of profit or loss in those cases. For example, if rates rise, certain derivative contracts pay off, but if rates decrease, then derivative positions will result in a realized loss. Although these fluctuations will usually be offset in the portfolio being hedged, the tax treatment can be different, and that is something to consider. There are times when hedging is optimal, but the principle is to determine whether the client s liabilities or liquidity requirements are being managed correctly. Question: Does Bloomberg or some other third-party vendor have the toolkit to enable an adviser to do the analysis efficiently? McCarthy: Bloomberg has many series to compute historical volatility, which is similar to the standard deviation measures I discussed. The COMP function compares the investment with its relevant benchmark, and the PORT function provides many risk statistics. FactSet and other vendors also provide standard deviations and the information ratios for various funds. Bloomberg provides historical datasets on many different funds, but if they do not have the datasets available, then customers can request additional coverage. Question: How do you measure the risk for VIX-based strategies? McCarthy: The VIX (Volatility Index) is a contract on implied volatility. The near-term VIX is an indicator of the stock market s expected movement (up or down) over the next 30 days. It represents equity market volatility and increases when the options that are used to buy or sell equities, and the level of volatility implied in their prices, increases. VIX contracts are used by managers to hedge the variability in investment portfolios or for diversification purposes. Managers quantify risk for VIX-based strategies using the methods I discussed for portfolios with asymmetrical profiles. Question: Can you explain why long short strategies cause asymmetrical distributions? McCarthy: Long short strategies by themselves might not lead to asymmetrical s. Strategies that rely on tight hedging, or long short strategies, involve buying undervalued stocks and selling overvalued stocks. These strategies might appear to have low risk, but if the correlation between the long side and the short side changes in ways not recently seen in history, it can generate surprising performance. Certain funds use hedging to actively manage duration risk. But funds that are purchasing high-yield bonds and using hedging contracts based on government securities could incur higher realized losses if the correlations diverge. For example, during 2008, many high-yield funds incurred realized losses as market participants demanded higher-quality risk-free investments. Although the bundle of investments in high-yield funds that have been hedged to reduce interest rate risk may appear to have low risk, the slightest decoupling of correlation can cause losses if the market suddenly demands higher quality and if yields are rising for low-quality assets while they are falling for the high-quality assets that are the basis of the hedge. In these cases, both parts of the portfolio can show losses instead of hedging one another. Although long short equity funds may claim a beta of 0, a lack of correlation can cause their beta to depart from 0. For the most part, long short equity funds are not meant to be hedged perfectly but to have risk on both the long and short sides. When both sides move to the same degree, they can seem perfectly hedged, but part of the strategy is that both sides should move somewhat differently, ideally generating alpha but sometimes leading to risk CFA Institute cfapubs.org Third Quarter

9 CFA Institute Conference Proceedings Quarterly Question: Have you found Monte Carlo simulation to be useful in downside risk measures? McCarthy: When using risk measures, such as VaR, managers also need simulation-based models to test outcomes with strategies that use options. Options are embedded in certain fixed-income instruments, such as call and put options and prepayment options relevant for mortgage-backed securities. To determine the downside of these instruments, it is best to re-price them at various market points. Models based on Monte Carlo simulation are good to use in a VaR calculation because they reprice all the factors in the portfolio for various market conditions and show at what point options move in and out of the money. Question: Do you believe the liquidity risks related to risk parity strategies are similar to the portfolio insurance strategies of the 1980s? McCarthy: Risk parity strategies invest in a range of uncorrelated assets to keep risk levels or dollar exposure constant. Portfolio insurance strategies rely on selling as the market decreases and buying as the market increases, which is what banks do when they write an option. That option has negative convexity or negative gamma. If the market gaps, the bank loses more than expected because it is not able to instantly transact in little increments as the market increases or decreases. Large market movements up and down may trigger losses for a portfolio insurance strategy, which is intended to create s in stable markets, although it may generate higher losses during volatile market conditions. The portfolio insurance strategy ingredients exacerbate gap risk and exhibit more risk during a financial crisis. Question: What is your view of the efficiency of concentrated positions in publicly traded stocks, and are there hidden risks? For example, how would an investor with $100 million of a stock put a collar on that position? McCarthy: One approach is to sell covered calls on the position and cap the upside; it will earn a good premium as part of that. The problem is if it really goes up, assets have to be liquidated to pay out under that option. If the goal is to not lose control of the stock, then selling covered calls will not work if the market moves up and other assets cannot be liquidated to pay the unrealized gains of the stock. Other liquid assets have to be available to cover the entire position if selling the stock is not an option. Question: Why do some investors in certain strategies with asymmetrical risk remove the upside volatility while retaining the downside risk to the extent that the loss would exceed the premium? McCarthy: There are a number of covered call funds available for investors, and generally, investors who buy options tend to slightly overpay for them; selling options is usually a better deal over the long term. Covered call funds normally have lower volatility and higher s than other funds they might be benchmarked to. So, it is not too ridiculous to consider covered calls as a risk management strategy. But it is important to remember that it is not insurance. I think of it this way: If I had a bus, I could buy an insurance contract that costs $1,000, or I could decide that the 10 cents people pay me to ride the bus will be sufficient over time to pay for any bus crashes I might have. That second option is not an insurance strategy, and that is what selling calls does. Investors get a little bit of premium, and might even earn a little bit more than the eventual payoff of the options, but it is not insurance. Insurance is buying a put for the downside that pays off if the market crashes; it is expensive but actually protects you in a downturn. Selling a call provides a small amount of extra income that helps investors a very small amount when the markets are careening down and stops them from enjoying the upside. Over the long term, though, the options pay off less (and prevent the enjoyment of market upside less) than the investor is getting paid for them, and that is why it is a popular strategy. Question: As structured products have become more mainstream, how would you screen the wide variety of products that do not present on term sheets as being a product with an asymmetrical? McCarthy: Providers of these instruments are obligated to disclose downside risk, behavior in various market environments, upside risk, and the relevant benchmarks. Some investors may buy these instruments because they have strategies that align with these instruments that have asymmetrical s. Managers need to ask various questions: Do my clients need this structure in their portfolio? Is there some reason why my clients require an option embedded in a bond? Could I simply have sold put options on my own and replicated the embedded derivative? It is important to understand the distinct components that are combined in the structured product. Few investments in the business have zero risk, so always think about the client s risk tolerance. 50 Third Quarter CFA Institute cfapubs.org