Saturday, December 11, 2010

Perspectives on Exchange-Traded Funds

Are exchange-traded funds good or bad for the market?

That was the title of a lively and interesting session at Markets Media's third annual Global Markets Summit last Thursday. The session was organized as an old-fashioned debate between two teams. On one side were David Weild and Harold Bradley (joined later by Robert Litan on video), who argued that heavily traded funds composed of relatively illiquid small-cap stocks were responsible, in part, for the sharp decline in initial public offerings over the past decade, with devastating consequences for capital formation and job creation.

Responding to these claims were Bruce Lavine, Adam Patti and Robert Holderith, all representing major sponsors of funds (WisdomTree, IndexIQ and EGShares respectively). The sponsors argued that they are marketing a product that is vastly superior to the traditional open-end fund, provides investors with significant liquidity, transparency and tax advantages, and is rapidly gaining market share precisely because of these benefits. From their perspective, it makes as little sense to blame exchange-traded funds for declining initial public offerings and the sluggish rate of job creation as it does to blame them for hurricanes or influenza epidemics.

So who is right?

Bradley and Litan have previously argued their position in a lengthy and data-filled report, and Wield has testified on the issue before the joint CFTC-SEC committee on emerging regulatory issues. Their argument, in a nutshell, is this: The prices of thinly traded stocks can become much more volatile as a result of inclusion in a heavily traded fund as a consequence of the creation and redemption mechanism. For instance, a rise in the price of shares in the fund relative to net asset value induces authorized participants to create new shares while simultaneously buying all underlying securities regardless of the relation between their current prices and any assessment of fundamental value. Similarly a fall in the fund price relative to net asset value can trigger simultaneous sales of a broad range of securities, resulting in significant price declines for relatively illiquid stocks. This process results not only in greater volatility but also in a sharply increased correlation of returns on individual stocks. The scope for risk-reduction through diversification is accordingly reduced, which in turn influences the asset allocation decisions of long term investors. The result is a reduction in the flow of capital to the smaller, more innovative segments of the market, with predictably dire consequences for job creation.

The sponsors do not deny the possibility of these effects, but argue that any mispricing in the markets for individual stocks represents a profit opportunity for alert fundamental traders, and that this should prevent prolonged or major departures of prices from fundamentals. But this is too sanguine an assessment. Fundamental research is costly and its profitability depends not only on the scale of mispricing that is uncovered but also on the size of the positions that can be taken in order to profit from it. Furthermore, since a significant proportion of trades are driven by the arbitrage activities of authorized participants, mispricing need not be quickly or reliably corrected. Both illiquidity and high volatility serve as a deterrent to fundamental research in such markets.

The problem, in other words, is real. But what I find puzzling about Bradley's position on this issue is that he seems unable (or unwilling) to recognize that precisely the same effects can be generated by high-frequency trading. As was apparent in an earlier session at the conference, he remains among the most vocal and fervent defenders of the new market makers. His justification for this is that spreads have declined dramatically, lowering the costs of trading for all market participants, including long term investors.

There is no doubt the costs of trading are a fraction of what they used to be, but a single-minded focus on spreads misses the big picture. It is worth bringing to mind John Bogle's wise words:
It is the iron law of the markets, the undefiable rules of arithmetic: Gross return in the market, less the costs of financial intermediation, equals the net return actually delivered to market participants.  
If spreads and costs per trade decline, but holding periods shrink to such a degree that overall trading expenditures rise (due to significantly increased volume), the net return to long term investors as a group must fall. Furthermore, if increases in volatility and correlation induce shifts in asset allocation that have the effect of reducing financing for small companies with high growth potential, then even gross returns could decline.

I have been arguing for a while now that the stability of an asset market depends on the composition of trading strategies and, in particular, that one needs a large enough share of information trading to ensure that prices track fundamentals reasonably well. But changes in technology and regulation have allowed technical strategies to proliferate, and high frequency trading is a significant part of this phenomenon. The predictable result is a secular increase in asset price volatity and an increased frequency of bubbles and crashes.

The flash crash of May 6 was just a symptom of this. Viewed in isolation, it was a minor event: prices fell (or rose, in some cases) to patently absurd levels, then snapped back within a matter of minutes. But the crash was the canary in the proverbial coal mine -- it was important precisely because it made visible what is ordinarily concealed from view. Departures of prices from fundamentals are routine events that, especially on the upside, are not quickly corrected. Some of the proposed responses to the crash that were favored at the conference -- such as trading halts followed by call auctions -- are cosmetic changes. They will have the effect of silencing the canary while doing nothing to lower toxicity in the mine.

It is the unremarkable, invisible, gradually accumulating departures of prices from fundamentals that are the real problem. These show up in the magnitude and clustering of asset price volatility and, through their effects on the composition of portfolios, leave their mark on the path of capital allocation, employment, and economic growth.


I am grateful to Terry Flanagan of Markets Media Magazine for the invitation to attend the summit.

I would also like to mention that the Kauffman report contains a number of assertions with which I disagree. For instance, Bradley and Litan endorse the claims of Bogan, Connor and Bogan that an exchange-traded fund with significant short interest could collapse with some investors unable to redeem their shares. This has been refuted very effectively by Steve Waldman in his comments on the Bogan post, and by Kid Dynamite. It is unfortunate that most responses to the report have focused on this dubious claim, rather than the more legitimate arguments that are advanced there.


Update (12/11). David Weild writes in to say:
I think we are seeing capital leave the microcap markets for a variety of reasons including:
  • Loss of liquidity providers
  • Emergence of ETFs (they don't buy IPOs and most don't buy follow-on offerings)
  • Indexing displacing fundamental investing (again, when this occurs, the funds stop investing in IPOs)
  • Loss of the retail broker as a stock seller.
If you don't have access to sufficient capital then capital formation, innovation and economic growth will suffer. That is clearly where we are.
I have also heard from someone who was once active in convincing the SEC to expand approval of ETF applications (and prefers to remain anonymous). He asserts that "the effects now being debated were certainly not an anticipated consequence. I can't remember a single conversation externally or internally at the SEC about whether the creation and redemption mechanism would increase correlations."

In hindsight it seems obvious that returns would become more highly correlated, but the fact that it was completely unanticipated at the time illustrates the enormous challenge of regulatory adaptation to financial innovation.


  1. I think ETFs are good for the market:

    1. Small-cap ETFs provide diversified low fee access for long term investors, this should lower the cost of equity for small cap issuers.
    2. If short term trading in ETFs creates volatility, well-timed IPOs may capture the upside thus enhancing the capital formation.
    3. Fundamental research is not required to correct mispricing created by ETFs. Computerized technical trading strategies should be successful in identifying uninformed selling/buying pressure driven by the creation/redemption mechanisms.

    A case can be made that the correlation of returns on individual stocks was too low previously. Volatility of fundamentals is lower than the volatility of stocks, and the suppression of excess idiosyncratic volatility may result in the higher correlation of returns.

    ETFs facilitate the transmission of AD shocks to all corners of the investable universe, this increases the efficiency.

    I don't believe that increased correlation of returns reduces the benefits of diversification for long term investors, at least if measured by the final-wealth risk.

  2. 123, regarding your second point, IPOs have fallen dramatically over the past decade and a half. Even if you don't think that changes in market microstructure are responsible for this, the idea that ETF's create opportunities for well-timed IPOs seems inconsistent with the evidence.

    I'm sure that technical strategies to exploit mispricing due to creation and redemption exist and are in use, and such strategies should dampen volatility and lower correlation. But as long as they are technical, I believe that this effect will only be partial. Again, this is what the evidence seems to suggest.

  3. Rajiv,

    I my point is that excess upside volatility creates the opportunity for well-timed IPOs. For example, stock market was very inefficient in 1999, and correlation was too low as growth stocks were severely overvalued, but small cap value stocks were severely undervalued. The result was a dotcom and telco IPO boom, but I don't think that it was a good thing, as the cost of capital provided to such companies was absurdly low.

    This time we are seeing high correlations together with the absence of severely overvalued or undervalued market segments. I take it as a sign of increased market efficiency, and as a result the number of IPOs is lower than before, but this is driven by the absence of inefficient IPOs.

    Another reason correlation increased is that the poor monetary and fiscal policy has generated the increased volatility of AD. By the way, increased correlation is a strong argument against the RBC interpretation of the current crisis.

    Changed legal environment has also create the lack of IPOs. A good example is Facebook - a company that is a very good fit for IPO. Instead IPO we have the OTC market in Facebook derivatives.

    I believe that the technical trading strategies should be sufficient to stop the excess volatility caused by the creation/redemption mechanisms. Such volatility is not fully suppressed because creation/redemption mechanisms transmit the fundamental information about AD shocks.

  4. I was active in the IPO market all year long. I think I had a piece of nearly every deal that got done. A very good number were winners this year.

    I can't imagine how anyone could claim that IPO and secondary activity has been impaired by ETFs. The market worked perfectly. Hundreds of deals got done. There were very few periods during the year where the calendar backed up and deals had to be pulled.

    There were many highlights, including the mega GM and deal and don't forget that Treasury sold all the Citi stock.

    What ever arguments there might be about ETFs, IPOs and syndicate stuff is not a problem.

    I am surprised that Rajiv does not pound on the table that ETFs are just a derivative. Rajiv does not like those based on his prior missives.

    Folks, ETF are one GIANT derivative. They walk like a duck, swim like a duck and trade like a duck. They are a duck.

    I am betting that we have a problem with this in the next year or so. One day too many people will try to squeeze threw a liquidity window and we will have another flash crash. But this time it will be different. It will go down by 10% and surprise, surprise it will stay there. When and if that happens it will will upset the ETF applecart.

    These things trade like water, but push come to shove they have no liquidity.

  5. Rajiv

    It's hard to disagree that the problem is excessive homogeneity of trading strategies. But I'm skeptical that diversity can somehow be imposed via regulatory fiat. Bradley and Litan overemphasize just how "diverse" mutual funds were before the rise of indexation and ETFs. In fact, one of the reasons why index funds became popular is because so many actively managed funds were closet indexers. ETFs are really just the logical conclusion of this evolution from a market where micro fundamental analysis dominates to one where macro asset allocation dominates. By the way, if homogeneity is the problem, then even index derivatives are culpable. Many professional investors use equities in their asset allocation primarily through such derivatives. In my opinion, if investors want to avoid firm-specific risks, they will find a way to do so no matter what the regulations.

    I spent some time working in index funds and we never held all the stocks in the index, just enough to ensure that our "tracking error" (deviation of performance from index) was within acceptable limits. As correlation goes up, the cheapest portfolio that mimics the index performance within the acceptable tracking error has fewer and fewer of the smaller names in the index. Even if I wanted to hold all the stocks in the index, my clients would start screaming at my higher costs compared to my competition who seemed to be achieving the same result at a lower cost. My thesis, albeit speculative, is that the initial trigger was a fall in the real returns and diversification from focusing on new firms/small cap which led to a reduction in assets allocated towards them. At some point, this rise in homogeneity became self-fulfilling and started to feedback onto the real economy. Hysteresis means that if we want to turn back the clock, then the "real" situation needs to improve a lot more than we think before we get a positive self-fulfilling increase in diversity.

    On a different note, I'd also point out that the fall in IPOs is partly driven by the simple fact that many startups spend a much longer time in the private venture capital-funded stage than they used to. In the late 90s I could reasonably hope to find tech stocks with a potential return of 10-20x but now that's much harder. Many startups either get big and then IPO or die in the much longer pre-IPO stage.

    And I tend to agree with Bruce that part of the solution is simply to enforce losses on investors in the next inevitable flash crash. As well as removing the bank financing constraints on new firms by tackling the oligopoly/TBTF problem in the banking sector.

  6. 123, Bruce, thanks for your comments.

    If you look at the Weild/Kim report (available here) you'll see that even at the height of the tech bubble IPOs were lower than in the early 1990s. The peak was 1996. In 2000 they collapsed completely and have remained very low throughout the decade. This does not mean that new issues were unsuccessful, it just means that there were far fewer of them. Part of it is Sarbanes-Oxley but that cannot be the whole story, it doesn't fit the timing.

    Regarding correlation, please look at the data from the JP Morgan Delta One team (available here). Return correlations right now are at levels that are completely inconsistent with earnings correlations. The reason is that the less liquid securities are being batted around by AP arbitrage, rising and falling in unison.

    Bruce - I'm not against derivatives, they are critical for hedging and this is an economically important function. I just think that the use of derivatives for two-sided speculative bets diverts collateral from more productive uses and has become excessive.

  7. On a broader note, I find that much of debates like this are dominated by two viewpoints: one, that markets are somehow perfect and never reflexive. And the other that markets are reflexive therefore always prone to fragility. My view which we've discussed before is that reflexivity rarely kicks in unless there is some underlying loss of resilience. The problem of course is what we can do once it has kicked in. I'm not ideologically averse to regulatory solutions but my experiences in banking have taught me that regulations that try to ban a certain kind of behaviour get arbitraged very easily. I tend to prefer incentive-compatible solutions but getting out of such a cycle is incredibly hard. In the language of the JPM report, earnings correlations probably need to be driven down a lot before returns correlations start falling. Of course there's always collapse as the cure for fragility, a solution I hope we won't have to go through.

  8. Ashwin, thanks for your comments. Regarding enforcing losses I agree completely, and have been critical of broken trades ever since my May 7 post on the flash crash.

    Regarding index derivatives, again I agree - they have very similar effects to ETFs and HFT, as the JPM report makes clear.

    Your comments on indexing are interesting. Weild actually made a similar point about closet indexing during the debate.

    I did not mean to suggest that the problem could be solved by regulatory fiat. I think that our positions on this are not too far apart. Certainly as far as TBTF we are on the same page and your posts on this have been excellent.

  9. Rajiv,

    Early 90s were a period of credit strains, so IPOs were a substitute for the slow growth of credit.

    Sarbanes-Oxley is one of the factors that prevented the revival of IPOs after the tech bubble crash, but of course there are lots of other very important issues, such as credit arbitrage that worked in the direction opposite to the IPOs (see the massive growth in the private equity industry).

    I don't think that a comparison of return and earning correlations is useful. If QE works through Tobin's Q, AD shocks can do little to change the correlation of quarterly earnings, but they strongly impact the cost of equity all across the board.

  10. 123, I still don't understand your point. Correlations rise whenever there are large unanticipated shocks to aggregate demand, or major unexpected policy announcements, or bubbles, or crashes. But they rise temporarily.

    What's different about the current situation is that correlation remains at historical highs although volatility is well below crisis levels, and there have been no major unexpected demand or policy events.

    The JPM report (linked above) explains why: index derivatives, ETFs and HFT. The mechanism is straightforward and completely plausible. What am I missing?

  11. Rajiv,

    The JPM report is based on the premise that the line in the correlation/volatility space (Fig. 3) represents some lost golden age of market efficiency. During the tech bubble when the market was extremely inefficient the correlation was below the line, but JPM are not willing to formulate a hypothesis that the market is more efficient than average when correlation is above the line.

    During the time when correlation/volatility relationship was normal there was always at least one obviously overvalued stock market segment, and at least one obviously undervalued market segment, so I presume that when the correlation/volatility relationship is normal, some market segments drift above fundamental values, and some drift below. This is not the case at this time, I don't see any obviously undervalued or overvalued market segment now.

    Instead of blaming ETFs and futures for the "Correlation bubble", I would like to suggest the possibility that the "normal" relationship between correlation and volatility was driven by the excess profits of dumb 3-18 month price momentum and earnings momentum strategies. For some reason such dumb strategies are less active now. It is likely that such strategies will become successful again, and the correlation will drop.

    You said:
    "What's different about the current situation is that correlation remains at historical highs although volatility is well below crisis levels, and there have been no major unexpected demand or policy events. "

    The monetary policy is constrained by the zero interest rate bound, and fiscal policy is constrained by political considerations. It is a good thing that the policy is powerful enough to prevent sharp shocks to confidence (this explains the absence of high volatility), but the policy is not powerful enough to fix AD at the levels of Great Moderation, and AD expectations are drifting up and down, this may explain high correlation.

  12. I guess we'll just have to agree to disagree on this. I just don't see enough volatility in AD expectations to cause such high levels of correlation, while I do see the index related activity. Sometimes the most obvious explanation really is the right one. In any case, thanks for your comments.

  13. Rajiv,
    I'll just add in support of the AD thesis that the correlation between S&P500 and 5 year breakeven inflation expectations is very high this year.

    I think you need additional proof to show that index activity that has displaced momentum strategies based on individual sectors has increased the market inefficiency.

    Thank you for very useful discussion, it has increased my understanding of weak points in my position. My position is tentative, and I am ready to revise it when we get all the information generated by the coming crash of the "correlation bubble".