Do Fast Traders Help Prices Find the Truth? Brogaard, Hendershott and Riordan

There is a version of the high frequency trading debate that is really about information.

Forget for a moment whether HFTs tighten spreads or provide fleeting liquidity. Ask instead: when a fast trader buys, are they right? Are they pushing the price toward the value it will eventually settle at, or are they shoving it around temporarily and profiting from the noise they create?

That is the question of price discovery, and it is arguably the most important one, because a market's core social function is to produce a price that reflects what things are worth. Brogaard, Hendershott and Riordan had a dataset that could answer it.

The problem: the good and bad stories are observationally similar

Both stories predict that HFT trades are followed by price moves in the same direction.

The good story: HFTs process public information fast. Earnings hit the wire, they read it before anyone else, they buy, the price moves up to the correct new level, and everybody's price is more accurate sooner. Under this story HFTs are the mechanism by which markets become efficient.

The bad story: HFTs detect a large institutional order being worked, they buy ahead of it, the price rises because of the institution's own buying, and the HFT sells back into it. The price move is real but it is not information, it is pressure. Under this story HFTs are a tax on real investors.

In both cases: HFT buys, price goes up. You need a way to separate them.

The key idea via analogy: does the move stick?

The separator is permanence, an idea we owe to Hasbrouck.

A price move driven by genuine information is permanent. The price goes up and stays up, because the world really did learn that the asset is worth more. A price move driven by temporary pressure is transitory. The price goes up, the pressure abates, and it comes back down.

So the test is beautifully clean. Look at HFT trades, and ask: do they trade in the direction of the permanent component of price changes, or in the direction of the transitory component?

Think of a boat on a choppy lake. The lake has a genuine, permanent tilt, its water level rising slowly. It also has waves, which go up and down and mean nothing. A trader who buys on the tilt is trading on information. A trader who buys on the crest of a wave is trading on noise.

Brogaard, Hendershott and Riordan used Nasdaq data in which trades are flagged as coming from high frequency firms, which is what made the whole exercise possible. And they found:

HFTs trade in the direction of permanent price changes. When they buy, the price tends to go up and stay up. They are, on the whole, right about where value is going.

HFTs trade against transitory pricing errors. When the price is temporarily too high, they sell into it. When it is temporarily too low, they buy. They lean against the noise, which is exactly what a market maker or an arbitrageur is supposed to do, and it makes prices more efficient, not less.

Put those two findings together and the verdict is: HFTs facilitate price discovery. They push prices toward fundamental value and away from temporary distortion.

There is a further wrinkle, and it is the honest and uncomfortable part. The authors examined HFTs' aggressive trades (where they cross the spread and take liquidity) and their passive trades (where they post quotes and get hit). The price discovery contribution comes disproportionately from the aggressive side. In other words: HFTs move prices toward truth largely by taking liquidity from slower traders who had posted stale quotes.

That is genuinely good for price efficiency and genuinely bad for whoever posted the stale quote. Both things are true simultaneously, and the paper does not pretend otherwise.

Why it mattered

• It answered the central question with actual identification. Because the data flags HFT trades, this is not inference from proxies. It is direct measurement. That makes it one of the most credible pieces of evidence in the whole HFT debate.
• It supported HFTs on efficiency while conceding the distributional cost. The finding is nuanced in the right way: markets are more informative because of fast traders, and the mechanism by which that happens involves picking off the slow. Efficiency and fairness point in different directions here, and pretending otherwise is what makes most HFT commentary useless.
• It clarified what HFTs are trading on. The evidence suggests fast traders are largely reacting to public information faster, not to private information nobody else has. That distinction matters enormously for policy: a firm that reads the same news as you but reads it in a microsecond is doing something quite different from a firm with inside information.
• It gave regulators a real cost-benefit frame. If you slow the market down, you get less adverse selection for slow quoters, and you also get less efficient prices. This paper is why that tradeoff is now the standard way the debate is framed, and it is the empirical backdrop against which proposals like frequent batch auctions and speed bumps are argued.

The honest limitations

• Efficient prices are not the same as a good market. The paper shows HFTs make prices more accurate. It does not show that this accuracy is worth what society pays for it. Budish, Cramton and Shim's argument is precisely that an enormous arms race in speed produces an accuracy improvement measured in milliseconds that essentially nobody needs, funded by a tax on liquidity. Both papers can be right.
• The winners and losers are real. "HFTs improve price discovery by picking off stale quotes" is a sentence with a victim in it. The victim is the liquidity provider who was slower, and ultimately, arguably, the investors on whose behalf that liquidity was being provided.
• It is one venue, one period, one set of HFT firms. Nasdaq, a specific window, a specific set of flagged firms. HFTs are heterogeneous, and the aggregate result may mask both benign market makers and predatory strategies netting out.
• The HFT flag is imperfect. The classification identifies firms that are predominantly high frequency, but large banks and brokers also run HFT strategies inside their own flow and would not be flagged. The measured HFT group is a subset, not the whole population.
• Normal times only. The dataset covers ordinary trading. The gravest concerns about HFT concern behaviour in stress, and this analysis cannot speak to it.

The one-line takeaway

Brogaard, Hendershott and Riordan showed that high frequency traders push prices toward fundamental value and lean against temporary mispricing, making markets genuinely more efficient, but that they do it largely by aggressively picking off the stale quotes of slower traders, which is a benefit and a cost that land on different people.