Are ETFs Causing Market Volatility?

New research suggests the current process of creating and redeeming ETF shares could have detrimental effect on long-term prices

Reviewed by: Farah Khalique
Edited by: Farah Khalique

< >Do exchange-traded funds (ETFs) cause higher volatility in their underlying stocks? Are they to blame for the August 24 flash crash? Much of the existing research on ETF liquidity is empirical, based on data from ETFs and their underlying basket of securities.

< >But Semyon Malamud, professor of finance at the Ecole Polytechnique Fédérale de Lausanne and Swiss Finance Institute, has taken a theoretical approach in his recently published paper to analyse intereactions between ETFs and the securities they track.

< >He found that the creation/redemption mechanism used by Authorized Participants (APs) can sometimes have a detrimental effect on market stability – as APs react to short surprise events, say, a quick rise in demand for shares, that can have a long-term effect on future prices.

< >Malamud's research was funded by the ETF Research Academy, which was created in 2014 by the Paris-Dauphine House of Finance, in association with ETF provider Lyxor Asset Management.

< > Explain the so-called 'shock propagation channel', which you mention when it comes to APs creating and redeeming shares. What does that mean and why is this important?

< >Semyon Malamud (SM): Think about the flash crash on 24th August. Imagine there is a big demand shock in the ETF price: if there is a big demand the price goes up. And ideally, you'd like this to die out quite quickly.

< >But what happens if there is a big demand shock? if I'm an AP then I am also a market maker, and the way I will respond that evening is go to the primary market and create a lot of shares [to meet demand]. But these shares are going to be trading tomorrow, and therefore the effect will only be seen in the market tomorrow. The shock that was just today and should disappear will now have an impact on tomorrow's prices.

< >That's exactly what we do not want.

< >The creation/redemption mechanism does not let it die out because the shock will now propagate in a way to tomorrow's prices. This may amplify volatility, and lead to a drop in liquidity.

< > What solution do you suggest?

< >SM: One thing the ETF sponsor could do is widen the spreads – we might be talking about increasing the spread by one basis point. It may sound weird and counterintuitive: it's a controversial theory. Increasing the spread might be one solution; it's probably not feasible because investors won't like it.

< >[Editor’s notes: If spreads widen, it would be more expensive for APs to create and redeem fund shares. If there are less shares being created, there would be less primary market activity and therefore less volatility in the underlying shares, according to the paper].

< > Do ETFs cause volatility in their underlying securities?


< >SM: This is not always the case. The right way to formulate this question is to ask, is it true that if I introduce an ETF on a certain basket of stocks the volatility will go up in the underlying?

< >There is a “substitution effect”. It is definitely the case that demand shocks to a new ETF will channel into the basket [of underlying securities] and that will introduce a new level of volatility [to those securities]. But that demand shock for a new fund will steal demand from existing ETFs, and that will reduce volatility.

< >But this scenario depends on whether the demand for the two ETFs – the new and the existing one – are positively or negatively related. If they are negatively related, then demand increases for the new ETF and demand is decreased for the existing ETF – the total effect is that volatility will go down.

< > Can we have too many ETFs? Is there a tipping point?

< >SM: The formula I get to is immensely complicated. Numerically, there does seem to be a capacity. When you start introducing ETFs, volatility initially drops. But then at some point it starts increasing again.

< >I do not know what the market capacity is exactly, but it seems in numerical experiments that the number of ETFs should be less than half the number of existing securities.