What is high-frequency trading doing to markets?

The short answer

High-frequency trading firms—Citadel Securities, Virtu, Jump, Hudson River Trading, Tower Research, DRW—operate algorithmic systems that submit and cancel thousands of orders per second across hundreds of stocks. The economic role most relevant today is market-making: standing on both sides of the bid-ask spread to capture small profits per trade across enormous volumes.

The popular debate has moved on from “are HFTs front-running everyone.” The empirical record shows that HFT market-makers narrow bid-ask spreads in normal times, providing real efficiency gains for retail and institutional traders alike. The economically relevant question shifted around 2015 toward what HFTs do in stress.

The answer is that they retreat. When volatility spikes, HFT market-makers widen quotes and reduce size or step out entirely—rationally, given their thin per-trade margins. The flash crash of May 6, 2010 and the August 24, 2015 ETF dislocation both displayed this dynamic in compressed form.

→ New to algorithmic trading? Microstructure framework

What the data shows

The HFT industry has matured substantially since its early-2010s peak, with consolidating market share and declining profitability per share traded.

The empirical picture (TABB Group, SEC, academic research, 2009-2024):

• HFT share of US equity volume: peaked near 61% in 2009 (TABB Group), declined to approximately 50% post-2010, broadly stable since
• Concentration: a 2009 SEC analysis found 2% of trading firms accounted for 73% of order volume; the top 5-7 HFT firms now likely capture an even larger share
• Futures markets: HFT share approximately 60%+ in major US futures (S&P E-mini, Treasury futures) since 2014
• Per-share profitability: industry estimates of HFT revenue per share traded have declined from approximately $0.001 in 2009 to roughly $0.0001 in recent years—a 90% compression
• Capital expenditure: top HFT firms reportedly invest $100-500 million annually in low-latency infrastructure (microwave links, colocation, FPGA hardware)

The exception is event-driven episodes: HFT volume share spikes in periods of high volatility (correctly measured) as short-term directional strategies activate, not as market-making expands.

→ Dataset: VIX dataset

Why it happens — the macro mechanism

The mechanism that ties HFT to market quality—and to market fragility—operates through three interconnected channels.

Spread compression and depth provision in calm regimes. HFT market-makers compete on quote prices and sizes. The result, well documented across academic studies (Hendershott, Jones and Menkveld, 2011; Brogaard, Hendershott and Riordan, 2014), is narrower spreads, smaller market impact and faster price discovery in normal conditions. This is the efficiency case for algorithmic dominance.

Cancellation discipline and information signal extraction. HFTs cancel a high share of submitted orders—often above 90%—as part of their inventory and information-management algorithms. Critics see this as quote stuffing; defenders see it as standard market-maker behavior at high speed. The angle most coverage misses: the modern HFT debate is no longer about adversarial speed races—those have been largely commoditized—but about what happens to liquidity provision when many HFT firms hit similar risk limits at the same time.

The 2010 flash crash and the August 2015 ETF dislocation both featured exactly this synchronization.

Risk-budget homogeneity. Most HFT firms run quantitatively similar inventory limits, P&L stops, and signal frameworks. When a stress event triggers risk-budget breach simultaneously across firms, all of them widen quotes or step out at once—and the market loses its normal liquidity floor. The more HFT supplies the marginal bid, the more synchronized the withdrawal becomes.

Synthesis by regime. In low-volatility regimes (most of 2010-2019, 2021), HFT compresses spreads, deepens books and lowers transaction costs—a clear net positive. In stress regimes (May 2010 flash crash, August 2015 ETF dislocation, March 2020 dash-for-cash), the same firms tighten or vacate, leaving abnormal volatility and wide spreads precisely when liquidity is most needed. The pivot is typically a sharp realized-volatility move that breaks risk-budget thresholds across multiple firms; markets recover when volatility normalizes.

HFT shrank the spread and shrank its own profit pool to do it; the modern question is no longer who is fastest but what disappears the moment everyone agrees they are too long.

→ Framework: Market microstructure & price formation

What it means for different economic actors

Retail traders. Net beneficiaries in normal regimes via tighter spreads, faster execution and price improvement on internalized orders. The “average retail trade” experience improved measurably through the 2010s as HFT compressed costs.

Institutional traders. Mixed picture. HFT competition narrows visible spreads but can detect and trade against institutional iceberg orders, raising implementation shortfall. Most large institutions use sophisticated execution algorithms (VWAP, TWAP, implementation shortfall) and dark pools to mitigate this.

Exchanges and venues. HFT generates a large share of exchange volume and rebate revenue (maker-taker fee structures). Exchange profitability depends materially on retaining HFT participation; this creates a tension when regulators consider speed-bump or order-handling reforms (IEX speed bump 2016, Cboe periodic auctions).

A common error is to assume that HFT is monolithic. The industry includes pure market-makers (Citadel Securities, Virtu, Jump), statistical arbitrage firms (Hudson River, DRW, XTX), and proprietary directional traders—each with different risk profiles and stress responses.

Practical observation

Go deeper

Related questions

• Risks of algorithmic trading dominance?
• How do dark pools change price discovery?
• How does payment for order flow work?
• Why is operational resilience now a macro concern?

Frequently asked questions

Is HFT really 50% of US trading?

That figure is the broadly cited industry estimate, with HFT having accounted for approximately half of US equity volume since the early 2010s. The exact share is hard to measure precisely because HFT is a strategy classification, not a venue-tagged activity—different studies use different definitions (latency, holding period, cancellation ratios). The 2009 SEC analysis identified that 2% of trading firms were responsible for 73% of order volume, which gives a sense of the concentration; subsequent industry estimates have placed HFT volume share around 50% with some range above and below depending on year and methodology.

Have HFT profits really collapsed?

Yes—dramatically. The industry’s revenue per share traded has declined from roughly $0.001 in 2009-2010 (a peak many cite) to estimated levels around $0.0001 in recent years. The competition for speed advantages has been substantial—submarine cables, microwave towers between Chicago and New Jersey, custom FPGA hardware—and the marginal speed gains became progressively more expensive while spreads compressed. The result is consolidation: smaller HFTs have exited or been acquired, and the surviving Big-5 (Citadel Securities, Virtu, Jump, Hudson River, Tower Research) now dominate the residual profit pool.

Did HFT cause the 2010 flash crash?

The SEC-CFTC joint report (September 30, 2010) found that the proximate trigger was a large algorithmic sell order in E-mini futures by a fundamental seller (Waddell & Reed), but that HFT firms’ responses—withdrawal of liquidity provision combined with hot-potato trading among themselves—amplified the dislocation. The episode established the modern framing: HFT is not the cause of stress events, but the homogeneity of HFT responses can amplify a triggering shock significantly, especially when liquidity providers face simultaneous risk-limit breaches.

Last updated — 23 July 2026