Evaluating Liquidity of Low Volume Exchange Traded Funds 

Updated: Aug 1st, 2015 | Vance Harwood

It’s disconcerting when the Exchange Traded Fund (ETF) you’re interested in has low trading volumes.  Days can go by with no volume at all.  If it was a small cap stock you wouldn’t consider buying it.  But low volume in itself shouldn’t prevent you from considering an ETF—some have the liquidity of a large-cap stock.


ETF Liquidity

Unlike stocks, ETFs have two levels of liquidity.  The first level is the same as a stock—the pool of people / institutions that hold, or are short the security.  The second level of liquidity is the domain of institutions.  They monitor the difference between the ETF’s trading price and the price of the underlying securities.  If that gap gets big enough they will buy or sell the ETF as appropriate and trade an offsetting transaction in the underlying.

For example, let’s say there was a low volume ETF that tracks Apple, Amazon, and Google—I’ll call it AAG, which holds one third of its assets in each stock and rebalances at the end of each day.  During the day you’d expect AAG’s intraday value to be the market prices of AAPL, AMZN, and GOOG times the number shares it’s holding in each.

But what if AAG is trading lower than its asset value?   If you can buy AAG at 55.03, and it has 200,000 shares outstanding the market is valuing the fund at $11,006,000, but what if you could sell AAG’s stock holdings for $11,050,000 –a 0.5% premium?

The second tier liquidity providers know exactly what to do; they buy undervalued shares of AGG, and short the appropriate number of Apple, Amazon, and Google shares to lock in their profit.  At day’s end they go to AAG’s issuer and sell back (redeem) the AAG shares they bought in exchange for shares of AAPL, AMZN, and GOOG. These ratios are fixed for the day so they get exactly the number of shares they need to cover their short positions.  They’ve just made 0.5% on their invested capital with zero risk.  It doesn’t matter what happens to AAG, AAPL, AMZN, and GOOG’s prices after their trade executions—their position is perfectly hedged.  This is an example of riskless arbitrage.

So how do the transactions of the 2nd tier providers (called Authorized Participants, or APs) impact AAG’s price?  Well, if AAG is underpriced they are buying it, driving up the price, and if it is overpriced they are selling AAG—exactly the actions that are needed to bring AAG’s value closer to its ideal level.   Because these transactions are riskless to the APs they are happy to continue them until the price difference between the market and the ideal level becomes too small to be profitable given transaction costs, bid/ask spreads, etc.

In the case of AAG the liquidity provided by the APs makes AAG no riskier for the investor than owning Apple, Amazon, and Google.  You can buy or sell large quantities of AAG without impacting its price.

If the underlying securities of an ETF are liquid then the ETF itself will be liquid too.

On the other hand, the liquidity picture for an ETF that tracks small cap Chinese stocks (e.g., ECNS) is significantly different.  The underlying stocks aren’t even trading when the USA markets are open because there’s no overlap in trading hours with the Chinese exchanges.  Now instead of riskless arbitrage the 2nd tier liquidity providers take significant risks if they buy or sell ECNS shares.  They will do these transactions, but not without requiring a significant premium to sell you shares or a discount in buying shares from you.


What About Exchange Traded Notes?

 Although the term ETF is used generically to reference all exchange traded products, a subset of funds called Exchange Traded Notes (ETNs) handle things differently than the example above.  While ETFs explicitly hold the securities specified in the index they track, ETNs link to their index without revealing how they hedge their exposure to market moves.  In practice ETN issuers do hedge their positions, but usually with derivatives like swaps or options that are less expensive than the actual underlying securities.

When an ETN’s shares are redeemed issuers pay out cash according to the Net Asset Value (NAV) per share rather than exchanging securities.  If the AAG example above was an ETN the Authorized Participant would cover their short positions themselves by buying Apple, Amazon, and Google shares at day end rather than obtaining the offsetting securities from the issuer.  If the underlying securities are liquid the arbitrage transactions by 2nd tier liquidity providers are just as effective for ETNs as they are for ETFs.


Evaluating Market Depth

 As a liquidity cross-check for ETFs and ETNs you can look at their Level 2 market depth.  These reports tell you not only how many shares are being offered at the National Best Bid and Offer (NBBO) quotes, but also can give you an indication of how much additional liquidity is available if you are willing to pay a premium on a buy or accept a discount on a sale.

The Level 2 Schwab StreetSmart Pro screen shot below of market depth is for Direxion’s Lightly Leveraged  Daily S&P 500® Bull 1.25x Shares (LLSP), an ETF with low daily volume.  Its underlying is the S&P 500— probably the most liquid security in the world.


The numbers on the far right indicate the number of shares various market participants are willing to sell at the prices listed in the middle.  The share amounts shown are minimum amounts—other players like High Frequency Traders (HFT) and dark pools will likely jump in with additional liquidity if a buyer appears.   The top two depth lines are essentially a duplicate, showing that the quote provided by the “arca” exchange is the current overall best bid.

A market order to sell 1000 shares would likely sell 600 shares at 26.64 and the remaining 400 at 26.63—a mere 0.04% less than the initial bid value.

A limit order to sell 1000 shares would probably fill entirely at 26.64 within a few minutes.

The next screenshot shows the level 2 quotes for another low volume ETF— iShares’ MSCI China Small-Cap (ECNS).  It does not have good liquidity even though it holds more than 20X the assets of LLSP.

bid ECNS-liq

A market order (not recommended!) to sell 1000 shares of ECNS would likely clear the book down to the 47.45 level—a big 2.3% drop from the starting bid price of 48.57.  A limit order at 48.57 would enable you to sell at least 300 shares at that price, but there is a good chance the rest of order would not fill because the next official bids are 0.5% lower.  The illiquidity of the underlying stocks makes it unprofitable for the tier 2 liquidity providers to offer better prices.


Some Investigation Required

Unlike a small cap stock you can’t judge the liquidity of an ETF or ETN by its daily volume or assets under management.  A quick look at the securities underlying the fund will spell out the real liquidity story—it’s dumb to use small cap rules of thumb.

Related Posts

VQTS: A Large Cap Investment That Protects Itself

Updated: Nov 9th, 2015 | Vance Harwood

VQTS is an exchange-traded product that that attempts to solve the essential conundrum of the stock market: how to run with the bulls without getting mauled by the bears.

The algorithms in VQTS are tuned to outperform the other hybrid volatility funds like Barclays’ VQT and VelocityShares’ SPXH during good times and to be competitive during the bad times.

We don’t have much trade data since UBS only introduced VQTS on December 3rd, 2014.  It stumbled in its first month of trading, but then recovered during the first half of 2015—but after 8 months it is only lagging SPX by 0.5% since its inception.

Hybrid-since-VQTS incept


All of the hybrid volatility funds dynamically allocate assets into the S&P 500 (SPX), VIX futures, and cash depending on market conditions.  VQTS is the first fund of this type that invests essentially all of its assets into the S&P 500 when the market’s overall volatility is low.  This avoids the costs of hedging when the market is least likely to go down—during the long upward stretches of bull markets.  In comparison, two similar funds, Barclays’ VQT and PowerShares’ PHDG both have at least 2.5% of their assets allocated to long volatility—and that small amount significantly drags down their returns during bull markets.

The chart below compares the simulated performance of VQTS compared to SPX, VQT, and SPXH from early 2006.



VQTS defines low volatility as being historical volatility less than 10% using the higher of two exponentially weighted moving averages on volatility.  Before shifting allocations VQTS requires that its historical volatility metric move by more than 10% from the target volatility used at its last rebalancing.  Once the target volatility climbs above 10% VQTS starts holding cash and volatility securities (2/3rds cash, 1/3rd volatility) in addition to the large cap S&P 500.  The table below shows the range of asset allocations as realized volatility increases.


VQTS Asset Allocations

Realized Volatility
(exponentially weighted)
Equity % Volatility % Cash %
0% to 10% 100% 0% 0%
10% to 20% 100% to 50% 0% to 16.7% 0% to 33.3%
20% to 30% 50% to 33.33% 16.7% to 22.22% 33.3% to 44.44%
30% to 40% 33.33% to 25% 22.22% to 25% 44.44% to 50%
40% to 50% 25% to 20% 25% to 26.67% 50% to 53.33%
50% to 60% 20% to 16.67% 26.67% to 27.78% 53.33% to 55.56%
60% to 70% 16.67% to 14.29% 27.78% to 28.57% 55.56% to 57.14%
70% to 80% 14.29% to 12.5% 28.57% to 29.17% 57.14% to 58.33%
80% to 90% 12.5% to 11.11% 29.17% to 29.63% 58.33% to 59.26%

In October 2008 the realized volatility as computed by VQTS’ algorithms peaked at 82%


When not fully allocated to equities VQTS takes a long or short position in VIX futures depending on the curvature of the futures’ term structure.  The term structure is the curve that’s formed if you plot VIX futures’ price vs time to expiration.  The decision to go long or short is determined by the comparison between the slope of the two nearest to expiration futures (1st and 2nd) and the slope of the 4th and 7th month futures.  I’ve marked up the chart below from vixcentral.com to illustrate the calculation.



VQTS’ curvature calculation is similar, but not identical to the more familiar designations of contango and backwardation for futures’ term structures.  In general VQTS will go long volatility if the term structure is in backwardation (futures prices less than spot), and short volatility if the curve is in contango.  This approach would have worked well in the past, profiting from the fast rise of volatility as a crash / big correction develops, and then switching to be short volatility as volatility mean reverts.  The chart below shows the simulated performance of VQTS during the 2008/2009 crash.


VQTS experienced around a 20% drawdown in the fall of 2008 before rallying to a year end gain of +13%.

Many strategies that backtest well on historical data do not perform well once they go live.  As I noted at the beginning of this post VQTS has already shown that it can approximate the S&P 500 during periods of low volatility—the condition the market is in 75% of the time.  VQTS’ drawdowns during crashes and corrections are likely to be significant, but VQTS’ strategy of going long volatility during panicky periods and short volatility during the recovery should continue to work well as a way to power through severe downturns.  The real challenge for it will be choppy markets with fast declines and V-shaped recoveries.

UBS did think about Black Swan events when constructing this fund.  If VQTS was allocated to short volatility when a major disaster (e.g., terrorist attack) occurred the fund could experience very heavy losses.  If VQTS is down 60% or more intra-day from the previous close or the indicative value drops below $5 the fund is terminated.  This is called an acceleration.  The shareholder receives a payout that is equal to the liquidated assets of the fund—which may be higher or lower than the acceleration threshold levels.

VQTS is still a small fund with only $25 million in assets, so investors are hesitant to invest in it, but since the S&P 500 and VIX futures are its underlying securities its liquidity is excellent.  Bid/ask spreads have been reasonable—in the 6 to 7 cent range (0.3%).

Everyone knows this bull market will end, the tough part is guessing when. With VQTS you can ride the bull and be prepared for the inevitable bad ending—the primary question being how well it handles the inevitable corrections that happen inbetween.

Related Posts

How Does VXUP / VXDN’s Corrective Distribution Work?

Updated: Jun 26th, 2015 | Vance Harwood

It’s now clear that the VXUP/VXDN naysayers were right—the actual behavior of these funds is nowhere close to Accushares’ claim that “VXUP and VXDN are the first securities to offer direct “spot” exposure to the CBOE Volatility Index® (VIX).

Accushares has said nothing publicly about the poor behavior of the funds.  Their only response has been to move up the “go live” date of their Corrective Distribution process.

This is a desperation move.  Apparently not understanding what has happened to them they are rushing to use the last weapon at their disposal to fix their horrible tracking error (as high as 18%) relative to the VIX®.   This move will reduce their tracking problem for less than a day and add yet another complexity to these already complicated and broken funds.

Accushares’ Corrective Distribution (CD) is intended to reduce any ongoing differences between VXUP / VXDN’s Net Asset Value (NAV) and its market price.   I don’t know what specific scenarios they were targeted with this process, but I’m guessing they were worried about a slow, progressive creep in the market prices versus the NAV.

The CD is triggered if there are 3 consecutive days where the tracking error (difference between NAV and closing price) is 10% or greater.  To be a valid closing price the last trade must occur within 30 minutes of market close.   When the CD is triggered it doesn’t occur immediately, it’s scheduled to accompany the next monthly Regular Distribution or a Special Distribution if that occurs first.  A Special Distribution is triggered by a greater than 75% rise in the VIX compared to the reference VIX value established at the beginning of the monthly cycle.

Accushares did not expect frequent CDs to be required, in the prospectus they state: “The Sponsor expects that Corrective Distributions will be infrequent, and may never occur.”

It’s likely they will occur on a near monthly basis.    When there is a significant gap between the VIX’s value and VIX futures prices (which is most of the time) the VXUP/VXDN tracking error will be large.  See this post for a near real time accounting of these errors.

With a combined Regular Distribution and Corrective Distribution three things occur:

  1. The NAV of the higher valued fund is set to equal the NAV of the lower valued fund.
  2. A dividend is issued that compensates the holders of the higher valued fund for the drop in NAV value. The dividend is either in cash or an equal number of VXUP/DN shares with a net value equal to the cash dividend.
  3. Accushares issues a new complementary share to every shareholder. If you have VXUP you will get VXDN shares and vice versa.  This is the Corrective Distribution mechanism.

Since Accushares can’t create assets out of thin air they must compensate for the newly doubled number of shares outstanding by dropping their value by half (or do a 2:1 reverse stock split).

The effect of the Corrective Distribution is not obvious.  Working through an example is a good way to understand it.   Imagine that a CD has been triggered and that a Regular Distribution is about to occur.  You own 1000 shares of VXUP.  Let’s assume that the market price of VXUP is $27.5, the VXUP NAV is $25, and that the VXDN NAV is $22 at market close right before the distribution.

You would receive a dividend of $3/share due to the resetting of VXUP’s $25 NAV value down to VXDN’s ending cycle value of $22.   You would also receive 1000 shares of VXDN.  The new NAV value would be $22/2 = $11 / share because Accushares doubled the number of shares outstanding.

Before the CD the VXUP shares in your account were worth $27.5 X 1000 = $27.5K.   After the Regular / Corrective Distribution your account has:

Your net account value drops to $25K—your $2.5K premium over NAV has disappeared.   Any premium over the closing NAV value is wiped out by the CD.  The next day VXUP will likely trade at a multiple percentage points over NAV, but your VXDN shares will like be trading at a symmetrical discount from NAV, so the net value of your shares will remain at around $22K.   Accushares has essentially cashed out your account at the NAV price—no premium for you…

No diligent shareholder will willingly take this sort of loss, nor short seller pass up this opportunity for profit; the market will ensure the value of these funds converges near to NAV the eve of the distribution.

From an entertainment value perspective there will be a couple of things to watch once the CD mechanism becomes effective:

  1. Will traders attempt to prevent CD’s from happening? Imagine a scenario where some groups are trying to prevent a CD from happening by selling, or short selling shares, while others hoping to profit from a CD are buying shares hoping to keep the tracking error above 10%.
  2. How low will the tracking errors go before the CD date? Short sellers would tend to drive the tracking errors to zero, but the about to expire VIX futures values will be decaying rapidly at that point, so significant intra-day profits might still be available to arbitrageurs on the last days of trading.

The net effect of the CD will be to complicate and disrupt an already difficult situation.   It won’t fix the funds.  What Accushares should do is eliminate the Corrective Distribution.

Once broken is better than twice broken.

Related Posts

Did Accushares Create a Product That is Linearizing the VIX Futures Term Structure?

Updated: Nov 14th, 2015 | Vance Harwood

Eli Mintz, the publisher of vixcentral.com, has published an analysis that shows the arbitrage relationship between Accushares’ VXUP shares and VIX Futures.  It shows that investors can use VXUP shares to lock in a low risk position for any VIX Future expiration month.

What struck me about this analysis was the incentive to sell VIX futures that had a higher daily roll yield, regardless of what month expiration they are.    This would tend to linearize the term structure of VIX Futures, and with a whopping 3 days of VXUP trading under our belts, this seems to be happening.   Normally I would expect very heavy contango between the first and second month futures as we bump up against all time highs in the S&P, but instead the contango has decreased from last week.


This might just be an effect from the May futures expiration that occurred during this transition, but I think  contango normally stays pretty consistent across expiration boundaries.

It could be that Accushares has unintentionally created a VIX futures term structure linearizer…

Related Posts

Tracking Information and Status Update on Accushares’ VXUP and VXDN

Updated: Nov 14th, 2015 | Vance Harwood

I’ve taken 20 minute delayed information to calculate tracking information and other items of interest using Google Sheets and Yahoo Finance’s on-line resources.   The day-end NAV values for these funds can be found here and here.

The last two lines in the table above show an experiment in progress.  It’s likely that the VXUP shares will track fairly close to the VIX futures that expires a couple days after VXUP / VXDN’s monthly cycle.  Currently I just multiply the current VIX future value by the Accurshares’ period multiplier factor (shown in the table below) and then apply the daily adjustment factor of 0.15% (only applies when starting VIX is <= 30) to the IV start value.


Status Update  16-June-2015

Nothing like a little real data to shake things up…  With 20 days of operational data to look at we can say some things about AccuShares’ VXUP and VXDN funds:

  1. These funds are NOT behaving the way described in the prospectus, in the marketing materials, and in the interviews provided by the supporters.  Tracking errors have matched what the detractors predicted–horrible.  Day end errors have been more than +-10% several days, and the funds are tracking VIX futures percentage movements rather than the VIX.  In contrast UVXY, an aggressive 2X short term volatility fund has a median tracking error of 0.08%.
  2. However, these funds are not dead in the water.
    • Volumes have been significant, especially VXUP.  One day had volume over 190,000 which is very impressive for a new fund.
    • Outstanding shares have quadrupled to 400K shares for each fund, giving a combined market cap of  $20 million after a week, again very impressive.
    • Bid / Ask spreads for both funds have been good, often dropping below 1%, very acceptable
    • Intraday Indicative Values (IV) are currently being published by Yahoo Finance under ^VXUP-IV and ^VXDN-IV
  3. Some implications of the numbers above:
    • The market makers for these funds were well prepared,  they were not surprised by the tracking error.  This is doubly impressive because of the novel nature of these funds.  VXUP / VXDN come in paired sets from Accushare, which I suspect broke a lot of software.
    • The Authorized Participants, the so called primary market for ETFs, were able to do share creation transactions with Accushares in spite of the wide tracking errors.  They needed to be able to short VXUP/VXDN shares at a combined price that was higher than the combined price of the NAV / IV prices that the funds were supposed to be trading close to.   The APs do risk free arbitrage operations for ETFs by shorting shares if they get too high above the IV values and doing share creations with the issuer (Accushares in this case) at the end of day to close out their short position at a profit.
    • I think both the market makers and the APs are using VIX futures to hedge their positions, otherwise I don’t see any way they would offer the small bid / ask spreads, or be willing to short VXUP/VXDN intraday and then wait for the end of day creation process.
  4. Puzzles and upcoming excitement
    • VXDN volume has been consistently lower than VXUP.  300K VXDN shares have been created since inception, and I’m pretty sure the total buy/sell volume for VXDN has been quite a bit less than that.  The market makers are apparently holding them, and I suspect are loaning them out for short selling.  Are the market makers willing to hold onto (“wear” in their vernacular) an indeterminate number of VXDN shares?
    • In the July 2015 cycle the Corrective Distribution (CD) process (see 11 Things You Should know about VXUP/VXDN) comes into play.  Its intent was to correct slowly accreting tracking errors (I’m guessing), but in actuality it looks like it will generate a truly minding bending challenge for the arbitrageurs that are trading these funds now.   If the the tracking error goes beyond +-10% for 3 consecutive days then a CD will be scheduled for the next distribution.  The CD will change the payout to the detriment of the arbitrageurs.   This creates a Prisoners Dilemma situation where everyone is punished if one trader gets greedy and pushes the fund past a 10% when there have been two consecutive closes with greater than 10%. For more see this post by Jay Wolberg on Trading Volatility.   Who knew how much entertainment value these funds would create!
    • The VIX term structure has been unusually linear since these funds came out.   In particular the price of the 2nd month future appears to have been brought down a bit.   This might cast a little light on what I think is a very murky arbitrage picture between VIX futures and the SPX options that theoretically provide their underpinnings.
  5. As near as I can tell AccuShares itself has gone dark except for accelerating the CD schedule by one month.   I think it’s highly likely that its filings for additional products like VXUP/VXDN will go on indefinite hold.    I’m hoping they will talk to some people that can help them understand what happened and chart a course forward that lets them be successful.  These funds are gathering assets, that is something a lot of startup ETFs have failed to do.


Related Posts

Responsive Menu Clicked Image