## Two Takes on Volatility and High-Frequency Trading

### How the rise in high-frequency trading has impacted volatility … and how it hasn't

by Adam Warner 9/9/2013 7:31 AMIn my **write-up of high-frequency trading** last week, I said, "It's unclear whether **[#HFT]** increases/decreases volatility."

Well, perhaps that's not true. I got this response from Eric Scott Hunsader of Nanex, "Looks pretty clear to us." Included was a link to **this post**:

Since 2005, there were 3.9 billion quotes and 579 million trades in SPY (an ETF that tracks the S&P 500). One billion of those quotes affected the NBBO (National Best Bid and Ask) resulting in 44 million NBBO price changes. We then use those NBBO price changes to compute a ratio called the Relative Intraday Volatility or RIV by dividing the number of NBBO price changes in a day by that day's price range ((high - low)/close).

The charts below [on his site] show the RIV (and its 20 period moving average) for each trading day from January 2005 through August 30, 2012. Notice how the RIV was quite steady up to the start of 2007 which coincided with the final roll-out of Reg NMS and the birth of High Frequency Trading (HFT)as we define it. Since then, the average intraday volatility in SPY has more than doubled, and was nearly six times higher in August 2011, and the peak intraday volatility in August 2011 was 10 times higher than it was in 2006. The second chart is scaled to the 20 period moving average to show detail. Note that the RIV already includes a volatility component (the daily trading range).

You need to click through to see the charts, but I can summarize it for you. The ratio of RIV to the day's range explodes in basically every name you look at.

Does this make my statement wrong? Well, kind of. But that's not the volatility I was referring to, or even remotely contemplating. I've never seen volatility measured in such a way, and it's certainly interesting.

I was referring to basic implied and realized volatility calculated on a daily basis. Realized volatility (RV) is essentially the denominator in his equation. Neither implied volatility (IV) nor RV has changed much when viewed through the long-term lens. Here's a graph of **SPDR S&P 500 ETF Trust (SPY)** volatility back to 1998:

*Chart courtesy of IVolatility*

Both clearly spike at times (see 2008), but the long-term mean of the **CBOE Volatility Index**, for example, was about 20 before 2005 and it's still about 20, even with the 2008 unrest. Look no further than now for example. We're basically mired on a low-volatility regime where a run above 20 VIX would feel like a high-volatility market.

It's highly likely that in a couple of years, this **low-volatility regime** will give way to a higher-volatility regime, and 20 VIX will seem more like a floor than a ceiling. Same as it ever was. To date, there's no particular evidence that HFT has disrupted *longer-term volatility cycles*.

Shorter-term, though, Mr. Hunsader clearly makes an excellent and irrefutable point. The rise of HFT has clearly led to a lift in tick-by-tick vol.

I'll take a pretty strong guess as to what's going on. The machine executes two virtually simultaneous trades and locks on pennies or fractions of pennies. Those translate to a price change in his work, which in turn leads to a spike in the ratio he calculates.

You and I aren't feeling that as volatility, though. To our naked eyes -- which can't process the tape to the micro-milli-nano second -- the price of the underlying has barely budged (if it's budged at all). What it is, though, is a de-facto transaction cost. It's pennies and half pennies and whatever the machines are essentially gobbling up ahead of us. It's someone essentially front-running your order a nano-second ahead of you and then flipping back to you for a tiny profit over and over again.

The higher his ratio gets, the more money leaves our hands.

So I do think it's insidious and all, but I don't believe it materially increases volatility as we understand and process it. I do believe it impacts volatility on the *margins*, with the Flash Crash on May 2010 as the prime example, and other mini-flash-crashes (like on Friday) as other examples.

*Disclaimer: The views represented on this blog are those of the individual author only, and do not necessarily represent the views of Schaeffer's Investment Research.*

permanent link