Latency
Latency refers to the time delay between when a trading signal or order instruction is generated and when it is received and processed by an exchange or trading venue. In electronic trading, latency is typically measured in microseconds or even nanoseconds and represents a key competitive dimension among high-frequency trading firms and other market participants.
Key takeaways
- Latency has three primary sources: network latency (time for data to travel between locations), processing latency (time for hardware and software to process data), and exchange latency (time for the venue to match and confirm the order).
- Firms with lower latency than competitors can act on market information before those competitors, enabling strategies like latency arbitrage.
- Co-location services, offered by exchanges, allow trading firms to house their servers in the same data center as the exchange's matching engine, dramatically reducing network latency.
- Ultra-low-latency market data feeds (direct feeds) provide faster price information than consolidated feeds, giving subscribers earlier awareness of price changes.
- Slippage in execution—the difference between the decision price and the fill price—is partially driven by latency; faster execution reduces the window during which prices can move adversely.
Explanation
Latency in electronic financial markets encompasses every delay in the end-to-end trading process: from the moment a price update arrives at a trading system, through the decision logic, order generation, network transmission, exchange receipt, matching engine processing, and confirmation return. This total round-trip latency determines how quickly a firm can react to changing market conditions. In competitive electronic markets, even microseconds of difference can determine whether an order executes at the intended price or suffers adverse selection.
The sources of latency are well-categorized in market microstructure literature. Network latency arises from the physical limitations of data transmission: electrical signals in copper wire travel at roughly 67% the speed of light, while fiber optic signals travel at approximately 69% the speed of light in glass. Microwave and millimeter-wave communication links can travel at near the speed of light through air, which is why firms have installed microwave relay towers between financial centers. The speed-of-light distance between New York and Chicago is approximately 2.5 milliseconds; adding processing and equipment delays, the fastest networks achieve round-trip latency of under 4 milliseconds on this route. Processing latency is reduced by using field-programmable gate arrays (FPGAs) that execute trading logic in hardware rather than software, achieving nanosecond-scale processing times.
Exchanges and trading venues compete on latency as a feature, offering co-location services that allow trading firms to house their servers in the same data center as the exchange's matching engine, typically reducing one-way network latency to hundreds of nanoseconds. Exchanges also offer low-latency direct market data feeds, which deliver price and trade data faster than the consolidated SIP (Securities Information Processor) feed that aggregates data from all exchanges but introduces additional delay.
For most institutional investors—pension funds, mutual funds, and many hedge funds—latency in the microsecond or even millisecond range is irrelevant because their investment strategies operate on much longer time horizons. However, latency becomes significant for any institution using algorithmic execution, because adversely selected fills (where the market moves against the order between decision and execution) accumulate into material slippage costs over large numbers of trades. This is particularly true for strategies that trade large numbers of smaller positions at high frequency, where execution cost is a larger component of total return.
Regulatory attention to latency has focused on whether differential access to low-latency infrastructure creates an unfair advantage. The SEC's Market Access Rule (Rule 15c3-5) requires broker-dealers to have risk controls in place before orders reach the exchange, while MiFID II in Europe requires synchronized clocks and timestamping of orders to sub-microsecond precision to enable surveillance of latency-sensitive trading. The debate over whether co-location is inherently unfair or simply reflects competitive investment in infrastructure continues in regulatory and academic circles.
Formula
Round-Trip Latency = Network Latency (×2) + Processing Latency + Exchange Matching Latency
Example
A market-making hedge fund places its trading servers in co-location at the NYSE data center in Mahwah, New Jersey. Its round-trip latency to the NYSE matching engine is 95 nanoseconds. A rival firm without co-location has a round-trip latency of 850 microseconds. When a large institutional order arrives at NYSE and moves the price of Apple stock by one cent, the co-located firm can update its bid/offer quotes in 200 nanoseconds, while the non-co-located firm's quotes are stale for 850 microseconds—a window during which it faces adverse selection. Over millions of quotes per day, this latency disadvantage translates to meaningful losses for the slower firm through being picked off on stale quotes, while the faster firm captures the bid-ask spread with minimal adverse selection risk.
Related terms
Artificial Price Bid Ask Spread Co Location Electronic Trading Exchange Ginzy Trading Hedge Fund High Frequency Trading Mifid Ii Pre Trade Transparency Price Discovery Slippage