Every cloud has a silver lining: Fast trading, microwave connectivity and trading costs Andriy Shkilko & Konstantin Sokolov Wilfrid Laurier University Central Bank Workshop on the Microstructure of Financial Markets, Paris 2016

Are these differentials good for liquidity? What we do, in a nutshell A speed race in modern markets leads to substantial speed differentials among traders Are these differentials good for liquidity? Our answer: No

Speed differentials: theory The effect of speed differentials on market quality may be positive or negative Positive: Hoffmann (2014) Jovanovic and Menkveld (2015) Roşu (2015) Negative: Biais, Foucault and Moinas (2015) Budish, Cramton and Shim (2015) Foucault, Hombert and Roşu (2016) Menkveld and Zoican (2016)

Which order types do traders use? Informed (and, recently, fast informed) traders often use limit orders: Hasbrouck (1991) Griffiths, Smith, Turnbull and White (2000) Bloomfield, O’Hara and Saar (2005) … O’Hara (2015) Brogaard, Hendershott and Riordan (2016) Chordia, Green and Kottimukkalur (2016) Liquidity suppliers try to stay on top of the latest technology to maintain a speed advantage Brogaard, Hagströmer, Nordén and Riordan (2015)

Information transmission between Chicago and New York

The race to zero in the Chicago-New York corridor Legacy fiber-optic cable: 7-8 milliseconds (ms) Spread Networks cable: 6.5 ms Microwave networks: 4.5 ms Speed of light: 4 ms

Microwave networks (MWNs)

MWN characteristics In 2011-2012, accessible by a small number of trading firms limited number of FСС licenses low bandwidth Relatively easily disrupted rain snow

Fast, but not always reliable: rain fade

What we find When speed differentials are eliminated, in the short run, due to precipitation adverse selection declines trading costs decline, in part due to the emergence of latent liquidity volatility declines in the long run, … Marketable orders play an important role in futures-equities arbitrage The results are more pronounced in assets with binding tick sizes, in which informed liquidity supply is more difficult

Democratization circa 2013 In early 2013, one of the technology providers begins selling futures pricing information to everyone on a subscription basis effectively democratizing information transmission

What we find When speed differentials are eliminated, in the short run, due to precipitation adverse selection declines trading costs decline, in part due to the emergence of latent liquidity volatility declines in the long run, … Marketable orders play an important role in futures-equities arbitrage The results are more pronounced in assets with binding tick sizes, in which informed liquidity supply is more difficult

What we find When speed differentials are eliminated, in the short run, due to precipitation adverse selection declines trading costs decline, in part due to the emergence of latent liquidity volatility declines in the long run, due to technological democratization all of the above trading volume increases Marketable orders play an important role in futures-equities arbitrage The results are more pronounced in assets with binding tick sizes, in which informed liquidity supply is more difficult

Data and samples Sample period I: 2011-2012 Sample period II: 2013-2014 Millisecond trade and quote data for equities (DTAQ) Millisecond order book data for select futures from the CME Precipitation data from the National Oceanic and Atmospheric Administration (NOAA) Two samples: Small: 5 ETFs Large: 100 ETFs

Precipitation along the MWN paths (www.noaa.gov)

Data and samples Sample period I: 2011-2012 Sample period II: 2013-2014 Millisecond trade and quote data for equities (DTAQ) Millisecond order book data from the CME Precipitation data from the National Oceanic and Atmospheric Administration (NOAA) Two samples: Small: 5 ETFs and corresponding futures contracts Large: 100 ETFs

Price impacts Price impact are 30-40% of effective spreads Chakrabarty et al. (2016) find a similar share for a recent sample of equities

Price impacts during MWN disruptions Price impacts decline by 0.047 standard deviations (or 7.05%) during significant precipitation episodes This effect is most pronounced for ETFs, in which the tick size is the most binding

Costs and revenues during MWN disruptions Effective and realized spreads decline by, respectively, 7.2% and 5.3%

Latent liquidity Under favourable conditions, latent liquidity may emerge Chakrabarty et al. (2016)

Volatility declines by 5.8%

What happens on the other side of the corridor? Hasbrouck (1995) and Westerlund, Reese and Narayan (2014) methodologies suggest that the futures market leads price discovery Consequently, asymmetric information and trading costs in futures do not change when the MWNs are disrupted

Event study: 2013 democratization In early 2013, a data provider begins selling latest price information at both ends of the Chicago-New York corridor this move effectively removes advantages of the fastest traders consequently, we find no precipitation effects in 2013-14 A quasi-DID analysis of the Quincy move shows declines in adverse selection, trading costs, market maker revenues and volatility Notably, trade increases

Robustness

Conclusions Liquidity providers do not always rush to adopt the latest technology and prefer to transfer increased adverse selection risk to liquidity demanders The fastest traders do not always assume the market making role. Rather, they often take liquidity Consistent with theory, speed differentials lead to higher adverse selection trading costs volatility Thank you

Trading activity and volatility The number of trades declines by 17.8%. Expectedly, this decline is observed only in the most constrained ETFs Volatility declines by 5.8%

Intraday patterns

A weather front