What do you think of the big HFT study? It’s this big HFT study that CFTC chief economist Andrei Kirilenko conducted on S&P 500 e-mini futures at the CME, and it’s already inspired a metaphor from CFTC commissioner and all-purpose spinner of metaphors Bart Chilton:
Mr. Chilton said that the study would make it easier for regulators “to put forth regulations in a streamlined fashion. It’s a key step in the process and it should fuel-inject the regulatory effort going forward.”
Not his best effort, fine. Anyway, the study: I’m not sure I’ve earned the right to have an opinion, both because (1) that, generally, and (2) my model of high frequency traders as micro-mini-market-makers is a bit upended by the fact that the bulk of the HFTs in this study seem to be taking, rather than providing, liquidity.1 It’s possible that the e-mini market is not the best place to measure the overall effects of HFT, either for fundamental reasons (its use for hedging etc.) or more crassly because it lacks the liquidity rebates that drive a lot of HFT in other markets.
That said what I like about this study is that instead of measuring transaction costs in naive ways like “bid/ask spread,” it measures transaction costs in sensible ways like “in a series of zero-sum transactions, how much money do HFTs suck out in profits.” Though the measure of profitability is sort of kooky:
The profits calculated in Table 3 are the implied short-term profits: we calculate the marked-to-market profits of each trader on 1 minute frequencies2 and reset the inventory position of each trader to zero after each of these 1 minute intervals. Then, we sum up all the 1 minute interval profits to get a measure of daily profits. Therefore, we capture the short-term profits of traders and not gains and losses from longer-term holdings.
What this means is that if your model of market-making is “buy at 99.9, wait five minutes, and sell at 100.1,” then your profits might end up showing as 0.2 or -0.2 or zero or something else on that calculation.3 So regular old market making may look bad, while HFT market making – designed to move quickly – looks much better. And so you get this table:
This shows daily profits of each class of trader – active, mixed, and passive HFTs (defined by how often they take liquidity vs. providing it; active HFTs take liquidity >40% of the time); non-HFT market makers, large fundamental traders, small retail traders, and “opportunistic” traders, meaning those who don’t fall into any other category. And in fact non-HFT market making looks crummy, though not as bad as being a retail investor, which, obviously. Here is the same data by cost per (~$70,000 notional4) contract:
You could draw a couple of tentative conclusions from this, but here’s one from the paper:
Since Aggressive HFTs collect $2.04 per contract (Column 9, Row 2) over 15.2% of the volume (Table 1), while everyone else loses while trading 84.8% (=100% – 15.2%) of the volume, the effective Aggressive HFT imposed transaction cost on all other traders is $2.04*(0.152/0.848) = $0.36 per contract. Scaling this per-contract transaction cost by $50,000 the approximate price of a contract yields an estimated HFT-imposed transaction cost of 0.0007%.
Or you can take an upper bound which is the $5.05 per ($70,000) contract that retail traders apparently lose to passive high-frequency traders and conclude that HFT costs small investors as much as three-quarters of one basis point of returns.
Here’s the author’s comment:
Mr. Kirilenko warned that the smaller traders might leave the futures markets if their profits were drained away, opting instead to operate in less transparent markets where high-speed traders would not get in the way.
That seems extreme for less than a basis point! Especially when you consider that small traders lose money to everyone, and they lose more to non-HFT market makers (i.e., whoever would be selling to them in less transparent markets) than they do to active HFTs (i.e. the traders who make up the bulk of the transparent, HFT-infested markets).5 “Screw you guys, I’m going to my broker’s dark pool,” say the retail investors, and lose twice as much. Or something.
Anyway, the paper is worth reading. There’s good stuff on the consistency of HFT profits and what it says about how HFT firms make their money. But I wouldn’t count on it to fuel-inject efforts to shut down high frequency trading. This paper doesn’t seem to me to suggest that high frequency trading costs “regular” investors all that much, and it certainly doesn’t suggest that it costs them more than the alternative.
Matthew Baron, Jonathan Brogaard & Andrei Kirilenko: The Trading Profits of High Frequency Traders [NBER]
High-Speed Trades Hurt Investors, a Study Says [NYT]
High Frequency Trading Arms Race Has Plenty of Drawbacks [MarketBeat]
1. I am not, however, alone in being puzzled by that:
Those findings run counter to a frequent claim by high-frequency traders, or HFTs, that they are largely “passive” investors that provide rivers of liquidity to the market, helping regular investors move in and out of stocks more easily. Instead, the firms—at least those most profitable—appear to be sucking up vats of liquidity.
Mr. Kirilenko says the finding was among the most surprising of the study. “It is in direct conflict with the statement that HFTs are providers of liquidity,” he said in an interview.
2. Table 3 says 10-second intervals. It’s a draft.
3. As I read it. I find it a little puzzling but for instance if you buy at 10:00:05.00 at 99.9, and then at 10:01:00.00 the mark-to-market is 99.9, then you have zero profit (or if it’s 99.8 you have -0.1, etc.). If at 10:01:30.00 the price goes up to 100.0, you have no profit on this measure, because your inventory was reset to zero at 10:01:00.00, though in reality (1) your inventory is one contract and (2) you have a 0.1 actual profit. If you then sell out at 10:05:05.00 at 100.1, then this gives you a short position and you have to look at the price at 10:06:00.00 to see if you actually have a profit; if it’s back down to 99.9 you show a 0.2 loss. Even though your timing was actually impeccable.
4. The e-mini is $50 x the S&P 500 level, which makes it worth $70,000-ish now; the study is August 2010 – August 2012 so notional starts at around $55,000. The authors use a $50,000 approximate contract price.
5. Though they pay non-HFT market makers less than they do to passive HFTs. This is the most puzzling result of all to me, actually: that passive HFTs (that is, classic high-frequency market makers) make $5.05 a trade off of retail investors, while (1) classic non-HFT market makers make less and (2) passive HFTs lose $0.62 per trade to large fundamental investors. Everything there goes against my first intuitions:
- HFT market makers should make less than non-HFT market makers off liquidity takers, since they should be faster and lower-cost and able to undercut non-HFT market makers.
- HFT market makers should make less off retail than they do off fundamental investors, since they should offer price improvement to get small retail orders while backing away from large informed orders.
The answer may lie in measurement: the fundamental traders could just be better at buying contracts (at a wider bid/ask spread) and seeing their price rise in the one-minute interval that is measured in this paper. The authors say:
The empirical results support the first hypothesis that Fundamental (institutional) traders are generally informed traders able to evade leaving a detectable pattern in their trading activity from which HFTs glean information. The results also support the hypothesis that Small (retail) traders are noise traders who incur the largest effective transaction costs per contract.