Precisely Forecasting Price Ranges with Volatility

Updated: Nov 16th, 2018 | Vance Harwood | @6_Figure_Invest

Using a tool like Bollinger Bands® to forecast future price ranges is a time-honored technique but its calculations are simplified and in some situations flawed. Incorporating the log-normal nature of stock prices into the calculations gives better answers.

One greed inducing aspect of volatility is that it enables us to make theoretically sound forecasts about the future.  It doesn’t matter whether it’s stock prices, annual rainfall, or deaths by shark attack, we can often use past volatility as a way of quantifying the probability of future values.  Volatility doesn’t enable us to predict whether average values are going to go up or down (sometimes we can use long-term averages for that) but it does allow us to make statistically valid predictions of the plus/minus ranges we can expect.  Of course, there are assumptions (e.g., that the level of volatility remains stable) but typically those are reasonable baselines.

Techniques like Bollinger Bands®, Probability Cones, and Z-scores use simplified calculations for computing price ranges.  For low levels of volatility and relatively short time periods their errors are small but as time and volatility increase your results can be significantly wrong or even nonsensical.  Given that institutional money is at stake, option pricing techniques such as the Black-Scholes equation and binomial tree calculations don’t take such shortcuts.

The incorrect assumption of the simplistic calculations is that since plus and minus returns are typically symmetrical, price ranges are symmetrical too.  As a simple counterexample, consider the end results for two symmetrical sequences of returns: three +2% days in a row, versus three -2% days in a row.  In the uptrend case, the final price increase factor would be (1+.02)3 = 1.0612, 6.12% higher than the starting price, and the downtrend price would be (1-.02)3 = 0.9412, 5.88% lower.  Not symmetrical!


Enter the Lognormal

As a result of this asymmetry, values of tangible things like stock prices, rainfall, and shark attacks that can’t go below zero often end up having what’s called a log-normal distribution.  These distributions are skewed to the right with a fatter right tail and a truncated left tail.

The chart below is a histogram showing the distribution of annual rainfall in Boulder Colorado for the last 123 years.

The horizontal axis has one-inch bins for annual rainfall (e.g., 12.3 and 12.8 would both go into the “12” bin) and the vertical axis is the number of occurrences of annual rainfall that landed in that bin).

The red line calculation uses the volatility of annual rainfall along with log-normal statistics to predict Boulder’s annual rainfall distribution—not a bad match to the historical data.

Historical price distributions of stock prices generally don’t have a log-normal shape because there’s another confounding factor—long-term growth.  The average annual rainfall in Boulder hasn’t changed much in the last 123 years but stock prices are usually different—the stocks that stick around tend to grow. Persistent growth smears out the price distributions over time.  Illustrating that, the chart below shows the simple histogram of S&P 500 prices since 1950—no log-normal in evidence.


We can remove the impact of growth by adjusting the price data with the time-series’ average growth rate—this is a calculation similar to using historic inflation rates to adjust for the buying power of a currency.

In the case of the S&P 500 the compounded growth since 1950 has been 0.0295% per day.  If we compound that value daily we obtain a factor that adjusts the price for the constant growth component of the S&P 500.  For example, the April 5th, 2018 S&P index closed at 2662.84.  We can remove the cumulative average growth from that value by dividing it by the compounded average growth factor (1+0.000295)17175 = 158.52, where the 17175 is the number of trading days since January 3rd, 1950.  The adjusted price is thus $16.80.  Applying this adjustment technique to the entire S&P 500 daily time series we get the price distribution chart below:

Not pretty—but recognizably log-normal.

When trying to predict future price ranges the analysis needs to take into account that stock prices are log-normally distributed—not Gaussian.


Transforming the Log-normal

Mathematically, dealing directly with log-normal distributions is ugly.  To simplify things the standard approach uses a mathematical transform.  A good transform converts a problem that’s tough to solve with default techniques into a different framework where the problem is easier to solve.  For example, it’s tough to physically copy a shirt directly (absent a suitable 3-D printer) but if the shirt is transformed—taken apart at the seams, the resulting pieces of cloth can be used to make a two-dimensional pattern.  With a pattern, it’s easy to make copies.  In the case of log-normal distributions, the logarithm function or “log” for short is the transformational equivalent of seam ripping.

If you take the log (usually base e, known as the natural logarithm LN) of each data point in a price history, you transform the distribution from a log-normal one into a normal one—we are now in “Log World”.  Standard statistical tools can now be used with this normal distribution.

However, this first step doesn’t solve the problem of growth smearing the data—we resolve that by utilizing the period to period differences (returns) rather than the log prices. You can write this step as ln(Pn) – ln(Pn-1), where n equals the period number. Mathematically these log and return calculations can be combined into one compact operation: ln (Pn /Pn-1). This formula is often called the “log returns” calculation.  When we convert stock price histories into log returns we get nice symmetric distributions suitable for forecasting.  The chart below shows the log returns distribution for the S&P 500.

Now, in preparation for doing our price projections, we can compute the statistical average growth rate (mean) and the volatility of this distribution using standard tools applicable to Gaussian distributions.  The average of the log returns is the geometric mean (the continuously compounded version), which I’ll call GMCC. The volatility of the returns is computed by taking the sample standard deviation of the log returns.  Be aware that both of these numbers were computed in Log World so they don’t directly relate to prices.  Also be aware that volatility is usually quoted in annualized terms—which enables apples-to-apples comparisons between different securities or time period lengths.  However for the calculations that follow we must keep the time period of the mean and standard deviation the same as the data set being used (typically daily).


Projecting Some Prices—the Median Forecast

Now that we have the key parameters we can move on to predicting some prices.

The simplest prediction is the median price.  Statistically 50% of the time the actual closing price will be higher than the forecasted median price and 50% of the time the actual price will be lower.  The formula for the forecasted median price is:

Pn = Ps* e(GMcc*n)


  • Pn = median price n periods in the future
  • Ps = starting price
  • e = Euler’s number ( approximately 2.71828) , in Excel use the EXP() function
  • GMcc = Geometric Mean (continuously compounded version)
  • n = desired number of trading periods in the future for the forecast

So, for example, using historic parameters for the S&P 500, if the S&P is at 2900 and you want to estimate the median price point 21 trading days in the future the answer would be:

P21 =   2900 * e(.000295*21)   =  2918                                        (+0.62%)

The “e” in the “e(GMcc*n) “ part of the equation converts things from “Log World” back to log-normally distributed prices.  It’s the equivalent of what a sewing machine does in my clothing analogy—transforming pieces of cloth back into real-world clothing.

The key assumptions are that the log returns are normally distributed and that the geometric mean stays stable.  If the average growth rate of the stock starts trending up or down then all bets are off.

The trading period can be any consistent period of time—days, minutes, months, etc.  The only restriction is that the geometric mean and standard deviation need to be computed using data with the same period length.


Enter Volatility

For a median price estimate, we only need the geometric mean but traders and investors are often interested in probabilities other than 50-50.  To make those estimates, we also need to know the estimated volatility of the underlying security.  The volatility is a key component because it quantifies how widely a price is likely to diverge from its current price —for example, the odds of a low volatility utility stock increasing 10% in the next month might be quite low but a volatile tech stock might easily move that much or more.

The probabilities associated with volatility have a fixed relationship to the standard deviation.  For example, if returns are normally distributed the probability of a stock’s one-day percentage up move exceeding one standard deviation (+1 sigma) is ~16% and the odds of it exceeding a two standard deviation move (+2 sigma) is ~2.23%.  The generalized equation, which adds a volatility term, predicts upside price points over time:

Pn = Ps * e(GMcc*n+k*stdev*square root (n))


  • stdev = sample standard deviation of the log returns (in Excel: STDEV.S())
  • k = multiplication factor, where 1 = 1 sigma, 2 = 2 sigma

Yes, that square root of the number of periods in the added volatility term is weird.  See Volatility and the Square Root of Time for more on that topic

For the sake of convenience, we usually use integer values of k but if you need different probabilities (e.g., quartiles) it’s straightforward to compute them.  For example, if you need to know the price that will not be exceeded 75% of the time k is +0.67, for 90% the k factor is +1.28.

If we want to want to compute a S&P index value that should statistically only be exceeded 16% of the time, (+1 standard deviation or one sigma) 21 days from now, we can compute:

P21+16%  = 2900 * e(0.000295*21 + 1* 0.01*square root (21))    =   3054.85    (+5.34%)

If we want to compute the upside S&P price 21 days from now, that only has a 2.23% chance (+2 sigma) of being breached, we compute:

 P21+2.5%  = 2900 * e(.000295*21 + 2* 0.01*square root (21))    =  3198.10    (+10.28%)


The Downside—Similar but not Symmetrical

The downside equation is similar but instead of adding the volatility term we subtract it.

Pn = Ps * e(GMcc*N – k* stdev*square root (n))

If we want to want to compute a downside S&P index value that should statistically only be breached 16% of the time (-1 standard deviation or minus one sigma) 21 days from now we compute:

P21-16%  = 2900 * e(0.000295*21 – 1* 0.01*square root (21))    =  2787.3   (-3.89%)

Because of the log-normal nature of stock prices, the downside percentage for a -1 sigma move is less than the 1 sigma upside move.

The key assumptions for these equations are the same as the median prediction with the additional restriction that the volatility of the stock stays consistent.  If volatility really picks up or fades then all bets are off.


Making Things Simpler—but Sometimes Too Simple

Popular and useful tools like Bollinger Bands®, Probability Cones, and Z-scores use a simpler approach.  Usually, they assume the median price forward is today’s price and leave out the exponential function. Their approach to projecting price ranges is to increase it from the starting price by +k*stdev* square root(n) and decrease by –k*stdev* square root (n).  These simplier formulas are reasonable if durations are short (e.g., 2 months or less) and volatilities, as well as sigma levels are low.  For the S&P 500, the typical error on the median price for 21 trading days would be around -0.5% and the 1 sigma price bands errors will be around 2.2%.

On the other hand, if you are using higher sigma levels (e.g., Bollinger Bands default to 2), longer periods of time, larger geometric means, and/or securities with high volatilities (e.g., TSLA, NFLX, VXX, UVXY, TVIX, SVXY) then the errors become significant. For example, the 2X leveraged volatility fund, VelocityShares’ TVIX—typically has a large geometric mean (e.g., -0.5%), and a high standard deviation (e.g., 4%).  With those characteristics, the error using the simplified calculation after 21 days on the median is -5%, the one sigma high/low band errors are -3% and 15%, and the 2 sigma high/low ranges have errors of -12% and +25%.


Monte Carlo Simulation Validation

As a cross-check of this analysis, I ran a Monte Carlo simulation of a test stock similar to Google (GOOG) with a daily geometric mean (.07%) and standard deviation (2%).  The simulation models the “random walk” behavior of a stock with the specified characteristics.

The smooth solid lines for high and low ranges are my two sigma theoretical projections.  The Monte Carlo lines show the first high and low simulated prices that are at the two sigma probability ranges (2.23%) for each period.  The dotted lines are the simplified estimates for high and low ranges using a Bollinger Band style analysis.

As you can see, the simplified analysis significantly underestimates the statistically expected ranges—after a year the predicted high range is 33% low and the low range is 28% low.



When the accuracy impact is low, it’s reasonable to simplify calculations to make them less intimidating or reduce the computational burden but it’s important to understand when those simplifications aren’t appropriate.  When habitually using these oversimplifications not only do all the results become suspect, it also hinders our understanding of what’s really going on.

For anyone looking into the finer points of volatility based price forecasts the equations incorporating the log-normal distribution of prices should be used.  Don’t create unnecessary distortions in your crystal ball.

Volatility White Papers and Presentations

Updated: Sep 16th, 2018 | Vance Harwood | @6_Figure_Invest

Below I’ve collected links to some of my favorite white papers and presentations on volatility. I’ve organized them in the following categories:

  • Volatility Concepts & Volatility Trading
  • Probability Distributions—Normal and Otherwise
  • The VIX and VIX Futures
  • Volatility Contagion—Will Short Volatility Destroy the World?
  • Variance Swaps—the Technology That Underlies VIX & VIX Futures

For an index of my 60+ posts on volatility see here.


Volatility Concepts & Volatility Trading

  • Volatility: A New Return Driver?” by Greggory Flinn & Roger Schreiner
    • A good non-mathematical overview of volatility, volatility products including futures and a couple example trading strategies using volatility Exchange Traded Products
  • Easy Volatility Investing by Tony Cooper
    • Available via free download on the SSRN repository, this paper provides a good non-mathematical overview of volatility investing. It includes a good discussion on the Volatility Risk Premium (VRP) which is an important concept.  It also provides detailed analysis of several volatility based trading schemes
  • Volatility Trading: Trading Volatility, Correlation, Term Structure and Skew” Bennett & Gil
    •  Over 200 pages of wide-ranging information—from covered calls to exotic options, to links between CDS spreads and implied volatility.  Something for everyone.

Probability Distributions—Normal and Otherwise 

  • Tales of the Unexpected by Andrew Haldane
    • This accessible paper (only one equation) is the best that I’ve ever read on the differences between processes accurately modeled by Gaussian/normal distributions and those better matched by power law distributions. I have seen this distinction made many times, but this paper provided examples and reasoning that really helped me internalize the differences.   Most of our stock market computations (including Black & Scholes for option pricing) and risk management formulas assume normal (or log-normal) distributions but this paper lays out a compelling case for why power law distributions are often a better match.
  • The normal distribution is the log-normal distribution by Werner Stahel & Eckhard Limpert
    • This presentation does a very nice job of distinguishing between the normal and log-normal distribution and providing guidelines for when they should be used. Bottom line, for stock price distributions we should use the log-normal distribution.

Read More

Using the VIX Futures Term Structure to Predict Volatility ETP Prices

Updated: Oct 4th, 2018 | Vance Harwood | @6_Figure_Invest

Status quo forecasting is sometimes very easy to do.  For example, if you predict that tomorrow’s high temperature will be the same as today’s high, your estimate will be close to the actual high much of the time.  Predicting volatility Exchange Traded Products (ETP) prices is not so straightforward.

The VIX futures that volatility ETPs like VXX, SVXY, UVXY, and TVIX track are similar to stock options in that they have a time value that usually decaying.  Generally the longer the VIX future has until expiration the higher its price.  If you plot VIX futures prices versus time until expiration the chart often looks like the one below from VIX Central.  This curve is called the VIX Futures Term Structure.

The term structure curve can be relatively stable for significant periods of time—which raises the question of whether we can use the term structure to predict volatility ETP prices.

Even if the price vs time curve of the VIX Futures stays exactly the same, several underlying factors that impact the prices of the volatility ETPs are in a state of change.  For example:

  • The individual VIX future’s prices change as they approach expiration
  • The mix of VIX futures that determines the ETP values changes based on their time to expiration and their prices
  • The position size of VIX Futures held by the leveraged ETPs (e.g., TVIX, UVXY, SVXY, VMIN, ZIV) changes on a daily basis based on the previous day’s percentage moves

Assuming the VIX futures term structure is stable (including the Cboe’s VIX spot price) allows us to project how much decay/gain is “built-in” to the prices of the long/inverse volatility ETPs. This information can help us set strike prices for option strategies, set limit prices, and determine risk/reward parameters.  More than 80% of the time, the VIX Future Term Structure is in a configuration called contango, where futures with more time until expiration are priced higher than the “spot” VIX price.  While in contango, decay factors on long volatility funds like VXX, UVXY, and TVIX can be considerable as can the boost factors on inverse funds like SVXY, VMIN, and ZIV.

Read More

Volatility ETP Price Projection Service

Updated: Nov 5th, 2018 | Vance Harwood | @6_Figure_Invest

I am offering a Volatility ETP Projection Service that calculates future volatility Exchange Traded Product (ETP) prices assuming the current VIX futures term structure is stable.
In my post Using the VIX Futures Term Structure to Predict Volatility ETP Prices, I show how this approach can be used to produce statistically valid ETP price projections and ranges.

This forecast does not attempt to predict upcoming volatility spikes or slumps—it’s totally focused on the price trends that would occur with a static VIX futures term structure.

With a stable term structure (and a stable spot VIX), the VIX futures prices that underlie the volatility ETPs like VXX, VXXB, UVXY, and SVXY do change but they precisely follow the price/days-til-expiration curve.  The VIX Central chart below shows the closing VIX futures prices for August 23, 2018.  If the term structure is stable then the curve at the end of the 24th would have the identical price vs time shape but the blue data points, representing futures values, would all be shifted slightly down and to the left.

The VIX futures that underlie the volatility ETP are volatile creatures—tomorrow’s values can be dramatically different than today.  I’m not trying to predict those sorts of changes.  What I am computing is the decay or boost that the volatility ETPs experience if the term structure stays in a stable contango or backwardation configuration.  This calculation is not an easy problem—there are a lot of moving parts even when the market is stable.

Historically the VIX futures term structure has been in a contango configuration 80%+ of the time.  Contango fuels a situation where the long volatility ETPs like VXX, UVXY, or TVIX suffer from high decay factors.  Anyone that’s looked at their long-term charts will see the massive impact of those decays over the long run.

Because of the typical decay in long volatility products, short volatility trades are popular but the possibility of volatility spikes makes risk management an important concern.  By estimating median prices and +-1 sigma ranges traders have some analytical results that can help quantify payoffs and establish appropriate risk management thresholds.

The chart below shows a typical SVXY projection when the term structure has been in contango for a while.

Read More

How Does VelocityShares’ ZIV Work?

Updated: Jul 29th, 2018 | Vance Harwood | @6_Figure_Invest

Just about anyone who’s looked at a multi-year chart for a long volatility fund like Barclays’ VXX has thought about taking the short side side of that trade. VelocityShares’ ZIV is an Exchange Traded Product (ETP) that allows you to hold a short volatility position while avoiding some of the issues associated with a direct short position in VXX.  Because ZIV is tied to VIX futures with at least 4 months until expiration its daily percentage moves are considerably smaller than the moves of funds (e.g., VXX, UVXY, TVIX) that are tied to shorter term, more volatile VIX futures.

To have a good understanding of how ZIV works (full name: VelocityShares Daily Inverse VIX Medium-Term ETN) you need to know how it trades, how its value is established, its characteristics, its risks, and how VelocityShares (and the issuer— Credit Suisse) make money running it.

How does ZIV trade?

  • ZIV trades like a stock.  It can be bought, sold, or sold short anytime the market is open, including pre-market and after-market time periods.  With an average daily volume of 120 thousand shares, ZIV’s liquidity is good.  Its bid/ask spread tends to run around 10 cents, which is on the high side, but as a percentage of its trading value that’s ~0.15% so it’s not a big economic penalty.
  • Unfortunately, ZIV does not have options available for it.  However, both of its closest Exchange Traded Fund (ETF) equivalents, REX ETF’s VMIN and ProShares’ SVXY -0.5X short term ETF do have options available.
  • Like a stock, ZIV’s shares can be split or reverse split—but unlike VXX (with 5 reverse splits since inception) ZIV has only split once, a 1:8 split in June 2011 that took its price from  $129 down to $16. Unlike Barclays VXX, ZIV is not on a hell-ride to zero.
  • ZIV can be traded in most IRAs / Roth IRAs, although your broker will likely require you to electronically sign a waiver that documents the various risks with this security.  Shorting of any security is not allowed in an IRA.

How is ZIV’s value established?

  • Unlike stocks, owning ZIV does not give you a share of a corporation.  There are no sales, no quarterly reports, no profit/loss, no PE ratio, and no prospect of ever getting dividends.  Forget about doing fundamental style analysis on ZIV.  While you’re at it forget about technical style analysis too, the major price moves of ZIV are not driven by supply and demand for ZIV itself but rather by the moves of the large, liquid VIX futures market.
  • Ultimately ZIV value is tied to the daily resetting inverse of an index (S&P VIX Medium-Term Futurestm) that specifies a hypothetical portfolio of VIX futures with 4 through 7 months until expiration.  Every day the index specifies a new mix of VIX futures in that portfolio. On any given day one-third of ZIV’s assets are allocated to VIX futures with 5 months till expiration, another third is allocated to 6th-month futures, and the final third is split between 4th and 7th-month futures. This mix of VIX futures gives ZIV the approximate performance of a VIX future with 153 days until expiration.
  • The index ZIV tracks, SPVXMP, is maintained by the S&P Dow Jones Indices.  The theoretical value of ZIV, if it were perfectly tracking the inverse of the index, is published every 15 seconds during market hours as the “intraday indicative” (IV) value.  Yahoo Finance publishes this quote using the ^ZIV-IV ticker. Because ZIV’s day end value is set by the settlement prices of VIX futures the closing IV value of ZIV is established around 4:15 PM ET not at the 4 PM NYSE close.
  • Wholesalers called “Authorized Participants” (APs) will at times intervene in the market if the trading value of ZIV diverges too much from its IV value.  If ZIV is trading sufficiently below the index they start buying large blocks of ZIV—which tends to drive the price up, and if it’s trading above they will short ZIV.  The APs have an agreement with Credit Suisse that allows them to do these restorative maneuvers at a profit, so they are highly motivated to keep ZIV’s tracking in good shape.  According to ZIV’s median tracking error relative to its index is -0.04%.

How Does ZIV Behave?

  • Almost all the time the medium-term VIX futures that underlie ZIV are in a configuration called contango where the longer dated futures are more expensive than the ones closer to expiration.  Persistent contango sets up an attractive short trade because as long as contango persists the VIX futures shorted by ZIV will tend to go down in value over time.  Contango does not guarantee profits for the short seller because if volatility spikes the medium-term futures tend to go up in unison but historically around 75% of the time ZIV is increasing in value.
  • This situation sounds like a short sellers dream, but VIX futures occasionally go on a tear, turning the short volatility trader’s profits into losses very quickly. While not as volatile as the short-term volatility funds ZIV can drop dramatically.  Its record one day drop so far was -26% on February 5th, 2018 and one day drawdowns of over 10% are fairly common.
  • The chart below shows ZIV from 2004 using simulated values.

  • ZIV does not implement a true short of its tracking index.  Instead, it tracks the -1X inverse of the index on a daily basis and then rebalances investments at the end of each day.  For a detailed example of what this rebalancing looks like see “How do Leveraged and Inverse ETFs Work?
  • There are some very good reasons for this rebalancing, for example, a true short can only produce at most a 100% gain and the leverage of a true short is rarely -1X (for more on this see “Ten Questions About Short Selling”.  ZIV, on the other hand, is up almost 500% since its inception on November 29th, 2010, and it faithfully delivers a daily percentage move very close to -1X of its index.
  • Detractors of the daily reset approach correctly note that ZIV and funds like it can suffer from volatility drag.  If the index moves around a lot and then ends up in the same place ZIV will lose value, whereas a true short would not, but as I mentioned earlier, true shorts have other problems.  Surprisingly, if the underlying index is trending down, daily resetting ETPs can deliver better than their stated leverage performance.  For more see “A Hat Trick for Inverse / Leveraged Volatility Funds
  • Historically ZIV has median moves of -0.21X compared to the CBOE’s VIX index.  If the VIX moves up 10% you can expect ZIV on average to move down 2.1%.  However, this relationship is not cast in stone.  At times ZIV and VIX will even move in the same direction.
  • Another important statistical ZIV relationship is its typical moves relative to the short-term volatility index which big volatility funds like VXX, UVXY, and TVIX are tied to.  Its median beta, the ratio of ZIV percentage moves to VXX moves, is around -0.44.  This ratio varies but as the chart below shows the variation since VXX started trading in January 2009 has remained between -0.2X and -0.72X.


What are the Risks?

  • Along with their impressive upsides, inverse volatility funds like ZIV carry considerable risks. The risks include the inevitability of volatility spiking up during market scares, corrections, or bear markets. Since its inception in 2010, ZIV has experienced 30 single day drawdowns of -5% or more, including the previously mentioned -24% crash. As scary as this is the other inverse volatility funds -0.5X SVXY and VMIN carry even more drawdown risk with worse case one-day drawdowns of -48% and -38% respectively. Buying and holding these securities is not for the faint of heart.
  • Another risk is termination. ZIV’s prospectus states that if ZIV drops 80% or more in a single day it will likely terminate. Before February 2018 there was a lively debate on whether this was a credible risk even for the higher leveraged former -1X short term inverse funds XIV and SVXY—which were much more likely to terminate than ZIV.
  •  On February 5th, 2018 both XIV and SVXY dropped more than 90%.  XIV was subsequently terminated by Credit Suisse and SVXY was deleveraged by ProShares down to -0.5X.  Based on that day’s behavior it would take a one-day VIX spike of over 300% to put ZIV in risk of termination (the 5-Feb-2018 VIX spike was 115%).  If VIX futures became even more reactive than they were on the 5th it might require a lower VIX jump than that but the bottom line is that it would likely take an event equivalent or bigger than the October 1987 crash to terminate ZIV.


How do VelocityShares and Credit Suisse make money on ZIV?

  • Credit Suisse collects a daily investor fee on ZIV’s assets—on an annualized basis it’s 1.35%.  With current assets at $120 million, this fee brings in around $1.6 million per year.  That should be enough to cover Credit Suisse’ ZIV costs and be profitable.  My understanding is that a portion of this fee is passed onto to VelocityShares for their technical and marketing activities.
  • Unlike an Exchange Traded Fund (ETF), ZIV’s Exchange Traded Note structure does not require Credit Suisse to report what they are doing with the cash it receives for creating shares.  The note is carried as senior debt on Credit Suisse’s balance sheet but they don’t pay interest on this debt.  Instead, they promise to redeem shares that the APs return to them based on ZIV’s daily closing indicative value.
  • Credit Suisse could hedge their liabilities by shorting VIX futures in the appropriate amounts, but they almost certainly don’t because there are cheaper ways (e.g., over-the-counter swaps) to accomplish that hedge.

With XIV delisted and SVXY deleveraged ZIV has a comparable leverage factor with the remaining inverse volatility funds (VMIN, and -0.5X SVXY).  Historically it has declined less than its competitors on the really high volatility days (5-Feb-2018 and Brexit shocks).  ZIV can’t guarantee that advantage in the future but it is comforting to see a track record of smaller drawdowns during historic VIX spikes.  In the post-February 2018 volatility landscape, ZIV is an attractive choice for shorting volatility.

 For more information