Budish, Cramton Too Shim On High Frequency Trading
Today I taught a truly overnice newspaper to my MBA class, "The High-Frequency Trading Arms Race" by Eric Budish, Peter Cramton together with John Shim. I've been fascinated past times high frequency trading for a acre (Some previous posts inward the novel "trading" label on the right.)
Eric, Peter together with John expect at the arbitrage betwixt the Chicago S&P500 e-mini futurity together with the New York S&P500 SPDR. This is a overnice case, because at that topographic point are no fancy statistical strategies involved: high speed traders only merchandise on short-run deviations betwixt these 2 essentially identical securities. Some cool graphs capture the basic message.
First, nosotros larn to expect at the quantum-mechanical limits of property pricing. At a 1 sixty minutes frequency, the 2 securities are perfectly correlated.
But equally nosotros expect at finer together with finer fourth dimension intervals, toll changes croak less together with less correlated. If the ES rises inward Chicago, someone has to shipping a purchase message to New York. We write downward Brownian motions for convenience, but when yous truly expect at real high frequency they interruption down.
It's non obvious this action "adds liquidity." If yous larn out a SPY bound monastic say standing, together with thence the fast traders volition choice yous off when they run across the ES ascent earlier yous do. The authors call this "sniping."
You tin run across the to a greater extent than jagged ES price. According to authors, it seems that Chicago is where the "price discovery" happens, equally the Chicago prices Pb New York. The New York SPY includes dividends, which the Chicago futures create not, thence New York is the natural domicile of "long term" traders. Once again, nosotros run across the interesting pattern of 1 marketplace position for "price discovery," together with and thence that toll communicates to exactly about other market.
How has high frequency trading affected the market? On the left, they plot correlation equally a business office of fourth dimension interval over time. In 2005, the correlation of toll changes was notwithstanding null at 100 ms. By 2011, 100ms toll changes rose to 0.5 correlation, together with the correlation stayed pretty expert downward to forty ms. The boundary of high correlation got shorter together with shorter.
Next, they calculate potential arbitrage opportunities: times when the toll difference exceeds the bid-ask spread. They too calculate how long those opportunities last. As yous tin run across below, the upshot of high frequency trading has been to dramatically bring down the duration of arbitrage opportunities. Once the prices diverge past times to a greater extent than than the bid-ask spread, inward 2005 that divergence could final 100 ms. Now, that divergence seldom lasts to a greater extent than than 10 ms.
You mightiness shout back that the profitability of arbitrage has declined. In 1 feel it has not:
These are the profits per opportunity, essentially a mensurate of how broad the toll spreads are independently of how long they last. The toll dispersion seems to live the same equally ever, it exactly goes away much to a greater extent than speedily than it used to.
In sum, nosotros larn hither a real construct clean representative of what high frequency trading does together with how it affects prices inward 1 market. It is lovely to run across the upshot of "arbitrageurs" making markets "more efficient."
But is this efficiency truly worth it? Does lodge truly gain plenty from having New York SPDR prices reverberate Chicago futurity prices 100 ms sooner, to justify laying always faster cable betwixt the 2 places? Does high frequency trading brand markets "more liquid" or exactly "more efficient?"
The theory portion of the newspaper examines the "arms race" of high frequency trading. That race is peculiarly clear here. If others merchandise this chance at 10 ms, together with yous tin larn at that topographic point inward ix ms, yous larn to choice off all the stale quotes together with larn out the other traders nothing. In turn, this arms race results because both markets are bound monastic say books inward which yous larn everything if yous house an monastic say 1 nanosecond earlier the other guy, yet prices must live discrete. Classic economic science predicts an overinvestment inward speed inward that game.
The theory portion of the newspaper explores this arms race game together with a natural proposal: Why non bring an auction in 1 lawsuit per second? You submit anonymous bids, together with in 1 lawsuit per 2nd render equals demand.
This agency that people submitting the same toll may bring to part fulfillment of the order, together with hold off a 2nd if they desire to purchase more. I'm all for efficient markets, but mayhap 1 2nd is efficient enough.
Put exactly about other way, if it is advantageous to specify a minimum tick size, thence prices croak discrete, mayhap it is advantageous to specify a minimum fourth dimension interval equally well. Computers function on a "clock" thence that all the signals settle downward earlier data is transferred. That mightiness live a expert blueprint for markets equally well. The brain enquiry I tin run across is how this impacts simultaneous orders seat inward dissimilar places.
Hopefully, I tin summarize the theory inward a futurity post.
First, nosotros larn to expect at the quantum-mechanical limits of property pricing. At a 1 sixty minutes frequency, the 2 securities are perfectly correlated.
It's non obvious this action "adds liquidity." If yous larn out a SPY bound monastic say standing, together with thence the fast traders volition choice yous off when they run across the ES ascent earlier yous do. The authors call this "sniping."
You tin run across the to a greater extent than jagged ES price. According to authors, it seems that Chicago is where the "price discovery" happens, equally the Chicago prices Pb New York. The New York SPY includes dividends, which the Chicago futures create not, thence New York is the natural domicile of "long term" traders. Once again, nosotros run across the interesting pattern of 1 marketplace position for "price discovery," together with and thence that toll communicates to exactly about other market.
Next, they calculate potential arbitrage opportunities: times when the toll difference exceeds the bid-ask spread. They too calculate how long those opportunities last. As yous tin run across below, the upshot of high frequency trading has been to dramatically bring down the duration of arbitrage opportunities. Once the prices diverge past times to a greater extent than than the bid-ask spread, inward 2005 that divergence could final 100 ms. Now, that divergence seldom lasts to a greater extent than than 10 ms.
In sum, nosotros larn hither a real construct clean representative of what high frequency trading does together with how it affects prices inward 1 market. It is lovely to run across the upshot of "arbitrageurs" making markets "more efficient."
But is this efficiency truly worth it? Does lodge truly gain plenty from having New York SPDR prices reverberate Chicago futurity prices 100 ms sooner, to justify laying always faster cable betwixt the 2 places? Does high frequency trading brand markets "more liquid" or exactly "more efficient?"
The theory portion of the newspaper examines the "arms race" of high frequency trading. That race is peculiarly clear here. If others merchandise this chance at 10 ms, together with yous tin larn at that topographic point inward ix ms, yous larn to choice off all the stale quotes together with larn out the other traders nothing. In turn, this arms race results because both markets are bound monastic say books inward which yous larn everything if yous house an monastic say 1 nanosecond earlier the other guy, yet prices must live discrete. Classic economic science predicts an overinvestment inward speed inward that game.
The theory portion of the newspaper explores this arms race game together with a natural proposal: Why non bring an auction in 1 lawsuit per second? You submit anonymous bids, together with in 1 lawsuit per 2nd render equals demand.
This agency that people submitting the same toll may bring to part fulfillment of the order, together with hold off a 2nd if they desire to purchase more. I'm all for efficient markets, but mayhap 1 2nd is efficient enough.
Put exactly about other way, if it is advantageous to specify a minimum tick size, thence prices croak discrete, mayhap it is advantageous to specify a minimum fourth dimension interval equally well. Computers function on a "clock" thence that all the signals settle downward earlier data is transferred. That mightiness live a expert blueprint for markets equally well. The brain enquiry I tin run across is how this impacts simultaneous orders seat inward dissimilar places.
Hopefully, I tin summarize the theory inward a futurity post.
No comments