TL;DR
This paper reveals a "three-tiered ecosystem" of stock exchange order types driven by financial regulations and trader sophistication. It finds that simple market and limit orders consistently lose money, while complex, non-routable orders—designed to bypass regulatory routing to avoid fees and gain millisecond-level speed advantages—dominate trading volume and short-term price discovery.
What Problem Does This Paper Address?
In the modern high-frequency trading (HFT) era, stock exchanges offer a dizzying array of order types beyond the "market" and "limit" orders found in textbooks. The authors investigate why this proliferation occurred and who uses these complex tools.
A central focus is Regulation National Market System (Reg NMS). While Reg NMS was designed to ensure investors get the "best price" by forcing exchanges to route orders to whichever venue has the best quote, the authors argue this creates a "gross vs. net" price conflict. Because the regulation ignores exchange fees and rebates, the "best" gross price often results in a worse net price for the trader. The paper explores how sophisticated traders use specific order types to navigate (or circumvent) these rules to capture profits.
Methods
The researchers provide the first "anatomy" of exchange order types by decomposing them into three functional categories:
- Price Conditions: (e.g., Market, Limit, Pegged, Stop)
- Time-in-Force (TIF): (e.g., Immediate-or-Cancel, Day, Good-Till-Cancel)
- Special Instructions: Specifically focusing on routing decisions (Routable vs. Non-routable like "Do-Not-Ship" or Intermarket Sweep Orders) and display decisions (Reserve/Hidden orders).
They then analyze the profitability of these types across different time horizons—from one second to 30 days—to infer the "clientele" (the type of trader) using each order.
Datasets
The study utilizes a massive, proprietary dataset:
- NYSE System Order Database (SOD): Contains 3.2 billion message-level records (additions, updates, cancellations, executions) from January 2010 to March 2011.
- Sample Size: A stratified sample of 109 stocks, resulting in 152 million transaction records.
- Granularity: Unlike standard public data, this includes "route-out" indicators, "make/take" fee indicators, and specific order instructions that are usually invisible to the public.
Key Insights and Ideas
- The Three-Tiered Ecosystem:
- Tier 1 (Naive): Simple market and limit orders account for <10% of volume and lose money (e.g., limit orders lose ~50 bps over 30 days).
- Tier 2 (Short-term Sophisticated): Non-routable orders (ISOs and DNS) win speed races to "snipe" stale quotes for small, immediate profits but hold no long-term informational edge.
- Tier 3 (Long-term Informed): Orders like "Buy-Minus-Zero-Plus" (BMZP) are used by firms for share repurchases; they are routable and patient, earning massive long-term returns.
- Bypassing the "Best Price": 57% of continuous trading volume comes from non-routable orders. Traders refuse to let the exchange route their orders because the 0.30 cent/share routing fee often outweighs the price improvement gained by moving to another exchange.
- Speed as a Defense: Non-routable "Do-Not-Ship" (DNS) orders are used to cancel stale quotes 42.75% of the time before they can be sniped by HFTs, whereas regular routable orders are almost always "picked off."
- Information Substitution: Sophistication in short-term trading (speed) and long-term investment (fundamental info) appear to be substitutes. The fastest traders aren't the best long-term pickers, and the best long-term pickers don't care about millisecond execution speeds.
Key Results
- Price Discovery: Intermarket Sweep Orders (ISOs) contribute more than 90% of price discovery on a trade-by-trade basis.
- The BMZP Alpha: BMZP orders, designed to comply with SEC share repurchase "safe harbor" rules, realize a staggering 30-day return of 706 bps (7.06%).
- Routing Inefficiency: 77.49% of routed limit orders are sent to other exchanges only to receive the same gross price or a worse net price after fees.
- Speed Advantage: DNS limit orders earn 1.02 bps higher returns than routable limit orders just 30 seconds after execution due to better queue positioning and avoiding adverse selection.
Limitations and Future Work
- Data Recency: The primary dataset is from 2010-2011. While the fundamental economic drivers (fees and Reg NMS) remain, the market has become even faster, and some specific order types mentioned (like DNS) have since been superseded by newer variants like "Add-Liquidity-Only" (ALO) or "Slide" orders.
- Anonymity: The data does not include trader IDs. The authors must infer trader intent and sophistication based on the order types used, which they acknowledge provides a "noisy signal."
- Single Exchange Focus: While the logic applies broadly, the empirical results are specific to the NYSE. Future work could compare these dynamics across different exchange models (e.g., inverted fee structures).
Summary updated