High-Frequency Trading: Speed, Spreads, and Market Quality

Key Takeaway

High-frequency trading has fundamentally reshaped equity market microstructure over the past two decades. HFT firms now account for roughly half of all equity volume in the United States and a growing share in Europe and Asia. The academic evidence is nuanced: HFT has narrowed bid-ask spreads, improved price discovery in normal conditions, and lowered explicit trading costs for most participants. Yet the speed arms race imposes real costs on the financial system, liquidity can vanish precisely when it is needed most, and episodic flash crashes reveal structural fragilities that regulators continue to grapple with. Understanding both sides of this debate is essential for any systematic investor operating in modern electronic markets.

The Rise of High-Frequency Trading

The transformation began in the early 2000s, when exchanges shifted from floor-based trading to fully electronic order books. Regulation NMS in the United States (2005) and MiFID in Europe (2007) fragmented liquidity across multiple venues, creating opportunities for fast traders to arbitrage price discrepancies. By 2010, HFT firms accounted for over 50% of U.S. equity volume, up from roughly 10% in 2004.

The defining characteristic of HFT is speed. Modern HFT firms co-locate their servers in exchange data centers, use microwave and millimeter-wave transmission networks, and optimize code at the nanosecond level. The latency between Chicago and New York, a critical arbitrage corridor, has been compressed from roughly 16 milliseconds (fiber optic) to under 4 milliseconds (microwave), with firms spending hundreds of millions of dollars to shave off microseconds.

This speed advantage enables several strategies: electronic market making, statistical arbitrage across correlated securities, latency arbitrage (trading on stale quotes at slower venues), and event-driven trading around news releases and data publications.

HFT as Modern Market Maker: Menkveld (2013)

Menkveld (2013) provided one of the first rigorous empirical portraits of a single large HFT market maker operating on the Dutch equity market. The study exploited the entry of Chi-X as a competing venue to Euronext in 2007, which coincided with a dramatic increase in HFT activity.

Menkveld found that this HFT firm behaved remarkably like a traditional market maker, continuously posting bid and ask quotes and rapidly mean-reverting its inventory. The key difference was speed and efficiency: the HFT market maker turned over its inventory far more rapidly than traditional specialists, holding positions for seconds rather than minutes or hours.

The entry of HFT market making on Chi-X was associated with a significant reduction in bid-ask spreads on both Chi-X and the incumbent Euronext. Effective spreads fell by roughly 30% for large-cap Dutch stocks during the study period. The HFT market maker earned a modest average profit per trade (roughly 0.5 basis points), consistent with a competitive market-making business rather than predatory extraction.

This study established a foundational narrative: HFT, at least in its market-making form, could serve a socially useful function by providing continuous liquidity at tighter spreads than traditional intermediaries.

HFT and Price Discovery: Brogaard, Hendershott, and Riordan (2014)

Brogaard, Hendershott, and Riordan (2014) used NASDAQ data with HFT firm identifiers to examine whether high-frequency traders improve or impair price efficiency. Their dataset covered 120 randomly selected NASDAQ-listed stocks over a 2-year period, with trades classified by whether HFT firms were on one or both sides.

The central finding was that HFT activity improves price discovery. When HFT firms traded aggressively (taking liquidity), their trades were more informed on average than non-HFT trades, pushing prices toward fundamental values. HFT firms were net suppliers of liquidity in calm markets and tended to demand liquidity during periods of high volatility, consistent with informed trading around temporary mispricings.

Importantly, the study found that HFT firms contributed positively to price efficiency in the vast majority of stocks examined. The improvement was particularly pronounced for smaller, less liquid stocks where price discovery is typically slower. This challenged the narrative that HFT only profits at the expense of other traders; instead, the evidence suggested that HFT accelerates the incorporation of information into prices.

However, the authors noted a critical caveat: during extreme market stress, HFT firms reduced their liquidity provision. This asymmetry between calm and stressed conditions would become a recurring theme in the literature.

The Arms Race Problem: Budish, Cramton, and Shim (2015)

While Menkveld and Brogaard et al. focused on the benefits of HFT, Budish, Cramton, and Shim (2015) fundamentally reframed the debate by arguing that the speed arms race itself is socially wasteful.

Their theoretical model showed that continuous-time trading creates a mechanical arbitrage opportunity: when a correlated asset (such as an ETF or a futures contract) moves, there is a brief window during which the quotes of individual stocks become stale. The fastest trader captures this profit; everyone else loses. The result is an arms race in which firms invest enormous resources to be marginally faster, but the aggregate social benefit is approximately zero because the price discrepancy would resolve itself within milliseconds regardless.

Budish et al. estimated that latency arbitrage profits in the E-mini S&P 500 futures market alone amounted to roughly $75 million per year. Across all correlated securities, the total rent extraction from the speed arms race was substantially larger. These costs are ultimately borne by end investors through wider effective spreads than would prevail under an alternative market design.

Their proposed solution was radical: replace continuous limit order books with frequent batch auctions, in which orders are collected over discrete intervals (e.g., every 100 milliseconds) and executed at a single clearing price. This would eliminate the advantage of being microseconds faster, redirect competitive investment toward price discovery rather than speed, and potentially narrow spreads further.

Quantifying the Arms Race: Aquilina, Budish, and O'Neill (2022)

Aquilina, Budish, and O'Neill (2022) provided the first large-scale empirical test of the Budish et al. framework using message-level data from the London Stock Exchange. Their dataset covered all FTSE 350 stocks over a multi-year period, allowing them to identify and measure latency arbitrage races in real time.

The findings were striking. They documented approximately 20,000 latency arbitrage races per day in FTSE 350 stocks alone. These races lasted a median of 5-10 microseconds and involved 2-3 firms competing to be first. The annual cost of latency arbitrage to the London market was estimated at roughly $60 million for FTSE 100 stocks.

Crucially, the study showed that latency arbitrage directly increases the cost of liquidity provision. Market makers who lose latency races to faster arbitrageurs must widen their spreads to compensate for the adverse selection. Aquilina et al. estimated that eliminating latency arbitrage through batch auctions could reduce bid-ask spreads by approximately 17% for the most liquid stocks.

This empirical validation of the Budish et al. theory strengthened the case for market design reform and influenced regulatory discussions at the SEC, the FCA, and other global regulators.

High-Frequency Market Microstructure: O'Hara (2015)

O'Hara (2015) synthesized the emerging literature into a coherent framework for understanding how HFT changes market microstructure. She argued that traditional market microstructure models, built around the assumption of sequential trade at human timescales, are inadequate for analyzing markets where trades occur in microseconds.

O'Hara identified several key structural changes wrought by HFT. First, the nature of information in markets has changed: speed itself has become a form of informational advantage, distinct from the fundamental analysis that traditional models assume. Second, the distinction between market makers and informed traders has blurred, as HFT firms alternate between providing and consuming liquidity within the same second. Third, market fragmentation across venues has created complex dynamics in which latency differences between exchanges generate arbitrage opportunities that did not exist in centralized markets.

She also highlighted the problem of phantom liquidity: quotes that appear in order books but are canceled before slower traders can access them. This creates a gap between displayed and actual liquidity that is invisible in traditional metrics. O'Hara argued that new analytical tools are needed to measure true market quality in the HFT era.

The Flash Crash of May 6, 2010

The most dramatic demonstration of HFT-era fragility was the Flash Crash of May 6, 2010, when the Dow Jones Industrial Average plunged nearly 1,000 points (roughly 9%) in 36 minutes before largely recovering. Individual stocks experienced even more extreme dislocations: Accenture briefly traded at $0.01, while Apple traded at $100,000.

The SEC/CFTC investigation identified a large sell order in E-mini S&P 500 futures by a mutual fund (later identified as Waddell & Reed) as the initial trigger. However, the cascading dynamics that turned a large order into a near-collapse of market structure were driven by algorithmic and HFT interactions. As prices fell, HFT market makers rapidly pulled their quotes, evaporating displayed liquidity. Algorithms trading against each other in the absence of fundamental buyers created a feedback loop that pushed prices to absurd levels.

The Flash Crash revealed a fundamental asymmetry in HFT liquidity: it is abundant in calm conditions but can disappear in seconds during stress. This is qualitatively different from traditional market making, where designated specialists had affirmative obligations to maintain orderly markets. HFT firms, as voluntary participants, have no such obligation.

Regulatory responses included the introduction of single-stock circuit breakers (later replaced by the Limit Up-Limit Down mechanism) and enhanced requirements for large trader reporting.

The August 24, 2015 Market Break

The dynamics of the Flash Crash repeated on August 24, 2015, when concerns about Chinese economic growth triggered a wave of selling at the U.S. market open. ETF prices diverged sharply from their net asset values, with some broad-market ETFs trading at discounts of 20-30%. The iShares Core S&P 500 ETF (IVV) briefly traded at a 35% discount to its underlying assets.

The cause was again the withdrawal of HFT liquidity at the market open, combined with the inability of authorized participants to efficiently arbitrage ETF discrepancies when underlying stocks were halted or opening late. Over 1,200 individual trading halts were triggered across NYSE-listed securities.

This event reinforced a key lesson: the HFT-driven market structure works well in steady-state conditions but can malfunction precisely when resilience matters most.

The IEX Speed Bump: A Market Design Response

The Investors Exchange (IEX), launched in 2016, represented a practical market design response to concerns about predatory HFT. IEX introduced a 350-microsecond speed bump, implemented via a coiled fiber optic cable, that delays all incoming orders equally. This delay is sufficient to neutralize the latency advantages that HFT firms use for speed-based strategies while being imperceptible to human traders and most institutional algorithms.

IEX's design philosophy draws directly from the Budish et al. framework: by removing the speed advantage, competitive investment is redirected from infrastructure toward price improvement and execution quality. Empirical evidence suggests that IEX achieves midpoint execution rates significantly higher than other exchanges, consistent with reduced adverse selection.

The SEC approved IEX as a national securities exchange in 2016, and its market share has grown modestly but steadily. Several other exchanges have since proposed or implemented their own speed bumps or asymmetric delays, though none have achieved the same structural simplicity as IEX.

Spread Reduction Over Time

The following table summarizes the evolution of bid-ask spreads and HFT market share in U.S. equities:

The data shows a dramatic compression of spreads coinciding with the rise of HFT and electronic trading. However, the causal relationship is not straightforward: decimalization (2001), increased competition among venues, and improvements in exchange technology all contributed independently. Disentangling the HFT-specific contribution from these parallel forces remains methodologically challenging.

It is also important to note that effective spreads may understate the true cost of trading. Phantom liquidity, quote fading, and adverse selection by faster traders can impose costs that are not captured in standard spread metrics. O'Hara (2015) argued that new measures of market quality are needed to account for these HFT-era phenomena.

The Ongoing Debate

The academic and policy debate over HFT remains unresolved along several dimensions.

Liquidity quality versus quantity. HFT has increased the volume of displayed quotes and narrowed spreads in normal conditions. But critics argue that this liquidity is lower quality: it disappears during stress, is often canceled before execution (phantom liquidity), and may front-run slower institutional orders. Empirical evidence supports both sides, suggesting the truth depends on the specific HFT strategy and market conditions.

Social value of speed. The marginal social return to being 1 microsecond faster is essentially zero, as Budish et al. argued. Yet the cumulative effect of electronic trading and competition among fast traders has driven down spreads and improved execution for retail investors. The challenge is separating the benefits of electronic trading (large and real) from the costs of the incremental speed arms race (also real but harder to measure).

Regulatory approach. Regulators have generally adopted a cautious approach, implementing guardrails (circuit breakers, market-wide halts) rather than fundamental structural reforms like batch auctions. The SEC's 2023 market structure proposals, including tick size reform and order competition requirements, represent the most significant potential changes since Regulation NMS, though their ultimate form remains uncertain.

Equity versus other markets. Most HFT research focuses on equity markets, but HFT is increasingly prominent in foreign exchange, Treasury markets, and commodity futures. The dynamics may differ in OTC-structured markets with different transparency and settlement characteristics.

Limitations

This review focuses primarily on equity markets in the United States and Europe, where the most rigorous academic studies have been conducted. HFT dynamics in Asian markets, cryptocurrency exchanges, and fixed-income markets may differ substantially. The empirical literature is also subject to selection bias: studies that find significant effects are more likely to be published. Flash crash dynamics are inherently difficult to study systematically because they are rare events with unique triggers. Market structure continues to evolve rapidly, and findings from studies using 2010-era data may not fully reflect current conditions.