Empirical Identification

Low-latency activity can be identified in public market data through patterns such as "strategic runs" of linked messages, revealing effective latencies on the order of 2–3 milliseconds during the early 2010s on NASDAQ, which is the hallmark of proprietary HFT behavior.

Three Structural Developments

Historically, HFT rose alongside three key developments in the 2000s:

• 1.
  The migration from floor-based to electronic, automated limit-order books
• 2.
  Regulatory changes that enhanced intermarket competition and fragmentation
• 3.
  Technological advances in networks, matching engines, and colocation

Infrastructure

HFT firms minimize both network and computational latency: they colocate servers in exchange data centers; subscribe to direct data feeds; and engineer deterministic, low-jitter software stacks that can process and act on order-book changes in microseconds. The academic literature tracks this via order-level timestamps and message dynamics, showing the fastest cohorts reacting within a few milliseconds in early samples.

Routing and Fragmentation

Because U.S. equities trade across many venues, HFT systems must route orders to capture displayed liquidity, avoid adverse selection, and manage queue position. Regulation NMS's order-protection and access rules, formalized in 2005, institutionalized this inter-venue game and increased the value of speed in smart-order routing.

Speed Bumps and Market Design Responses

Concerns about "latency arbitrage" led to design countermeasures. IEX, for example, implemented a 350-microsecond, symmetric delay (a "speed bump") to blunt race-to-be-first advantages. As IEX became an exchange in 2016–2017, empirical and policy analyses explored effects on price discovery and routing; evidence suggests speed bumps can reduce certain arms-race dynamics while leaving aggregate market quality largely intact, though magnitudes vary by design and market share.

Economic Channels

Case studies and follow-on literature indicate the primary revenue source for many HFTs is the spread capture from providing liquidity, net of exchange fees/rebates and inventory costs. Profits are highly sensitive to transaction costs, adverse selection, and the cost of capital; the business model relies on turning over tiny per-trade margins at scale.

Why Speed Matters

In continuous markets with fragmented liquidity, being first to update or cancel a quote reduces adverse selection, and being first to a stale quote captures micro-arbitrages. The "arms race" literature argues that a portion of private spending on speed is rent-seeking—valuable privately but not socially—because the gain from beating a rival by microseconds mostly comes from redistribution among fast traders rather than better allocative outcomes.

Design Responses

Speed bumps (e.g., IEX's 350 µs delay, 2016–2018) and proposals like batch auctions are attempts to re-optimize market rules to retain the benefits of automation while reducing socially wasteful latency races. Evidence following IEX's transition to an exchange suggests protected markets with symmetric speed bumps can improve certain aspects of price discovery with limited side-effects, though impacts depend on the exchange's market share and precise mechanics.

For Retail and Institutional Investors

Narrower spreads and faster quote adjustments, when they occur, reduce trading costs for liquidity takers. Liquidity-providing HFTs can also "smooth" liquidity by supplying when it is scarce and demanding when it is cheap, stabilizing short-run order-book conditions (evidence from algorithmic-trading studies and HFT-identified datasets).

For Issuers and the Real Economy

Improved secondary-market liquidity (tighter spreads, deeper books) can lower the cost of capital by making primary-market securities more attractive to investors. While this link is mediated by many forces, microstructure research emphasizes the centrality of displayed limit orders and efficient price discovery—elements that regulation has repeatedly recognized as "critically important" (SEC, 2005; and updates in 2024 revisiting tick sizes and display).