Trade Types and Order Flow

Prerequisites

• trading-fundamentals — spreads, market makers, inventory vs information models

Not every trade hits the order book. Before a single share matches in a CLOB (Central Limit Order Book — see order-books), someone decided how to route it. A retail buy on Robinhood takes a completely different path from a pension fund selling 2 million shares.

This note covers the five major trade types — agency, principal, block, algorithmic, and retail/PFOF — and explains why each exists and what tradeoffs it carries.

Agency vs principal execution

The first question in any trade: whose capital is at risk?

Agency execution means the broker acts on your behalf — finding a counterparty and matching your order without ever owning the shares themselves. The broker earns a commission and bears no market risk. Your trade confirmation shows the commission (SEC Rule 10b-10). The broker’s obligation is best execution (FINRA Rule 5310): they must seek the best reasonably available price, but their pay doesn’t depend on how good that price actually is. The incentive is regulatory, not economic.

Principal execution means the dealer trades against you from their own inventory. You want to sell 10,000 shares? The dealer buys them from you directly, taking on the position and the risk that comes with it. The dealer’s revenue isn’t a commission — it’s the markup baked into the price (for bonds, FINRA Rule 2232 requires this markup to be disclosed in basis points). The dealer profits from the spread, which creates a tension: wider spreads are better for the firm, worse for you.

Riskless principal is the hybrid that dominates modern equity markets. The broker finds a fill on the other side of the market first, then books an offsetting trade to you — so the firm never holds a net position. Economically it behaves like agency (no directional risk), but legally the trade prints as principal because the firm momentarily stands between the two sides. PFOF wholesalers (see below) operate as riskless principals for most retail flow.

	Agency	Principal	Riskless principal
Revenue	Commission	Markup / spread	Small markup or commission
Market risk	None	Full inventory risk	Transient (seconds)
Best execution	Required	Required (tension with spread)	Required

Reference

Harris, Trading and Exchanges, Ch. 6 (Brokers) and Ch. 13 (Dealers) cover agency and principal execution in depth.

Block trades and upstairs trading

A block trade is a transaction large enough to move the market if exposed to the displayed book. Traditional thresholds: NYSE defined blocks as 10,000+ shares or $200,000+ notional; CBOE options use 250+ contracts.

Why blocks can’t just hit the order book

The displayed book has finite depth at any price level. Dropping a 500,000-share sell order into a book with 50,000 shares at the best bid would:

1. Exhaust visible liquidity across multiple price levels, causing severe price impact (see kyle-lambda for the theory of how order size maps to price movement)
2. Signal selling pressure to every participant watching the tape, causing them to pull bids and widen spreads before the order finishes filling
3. Invite front-running by high-frequency firms that detect the large order’s footprint and trade ahead of it

The upstairs market

Block traders at broker-dealers solve this by working off-exchange. The process:

1. Canvassing: the block desk contacts known institutional holders (“naturals”) — pension funds, mutual funds, insurance companies — to find contra-interest without revealing the full order size. The desk might say “we have interest in the name” rather than “we’re selling 500,000 shares.”
2. Negotiation: price is agreed bilaterally, often at a discount to the mid-price to compensate the contra-side for providing liquidity without the benefit of public price discovery.
3. Capital commitment: sometimes no natural contra-side exists. The broker commits its own capital — buying the block as principal and assuming inventory risk — then works out of the position over time.
4. Tape print: the completed trade is reported to the TRF (Trade Reporting Facility — the FINRA-operated system that publishes off-exchange trades to the consolidated tape) so it appears in the public record.

Modern electronic equivalents

The upstairs market has partially migrated to electronic dark pools — trading-venues covers the venue taxonomy. Key platforms for block discovery include Liquidnet (peer-to-peer institutional crossing) and MS Pool (Morgan Stanley’s dark ATS). These platforms automate the canvassing process while preserving anonymity, but the economics are identical: find a natural contra-side before information leaks.

Reference

Harris, Trading and Exchanges, Ch. 15 (Block Traders).

Program and basket trades

Block trades move one name at a time. A program trade (also called a basket trade) moves an entire portfolio at once — historically defined by the NYSE as 15 or more names totaling $1 million or more in notional.

Who uses them and why

User	Use case
Index funds	Rebalancing after index reconstitution (additions/deletions)
Pension funds	Transition management — moving an entire portfolio from one manager to another
ETF authorized participants (APs)	Creation/redemption baskets — exchanging a basket of underlying securities for ETF shares (or vice versa)
Quantitative funds	Executing factor-model portfolio changes where individual stock alphas are small but the basket has a statistical edge
Risk arbitrage desks	M&A arb positions requiring simultaneous long/short across multiple names

Why trade as a package

Three reasons to bundle rather than execute name-by-name:

1. Risk transfer: the broker can offer a guaranteed close price (or a risk bid for the whole basket), absorbing execution risk for a fee. The client locks in a known cost.
2. Cost efficiency: trading costs scale sub-linearly with basket size because correlated names partially offset. A basket that is long tech and short energy has less net market risk than the sum of its parts.
3. Avoiding information leakage: executing 200 names simultaneously is harder to reverse-engineer than dripping them out one by one over hours.

Historical significance

Program trading played a central role in the 1987 crash (Black Monday). Portfolio insurance strategies — designed to replicate a protective put by mechanically selling index futures as the market fell — generated self-reinforcing selling pressure. As futures fell, the strategy demanded more selling, which pushed futures lower, which triggered more selling. The feedback loop between programmatic selling and market impact was a key amplifier, though not the sole cause.

Algorithmic execution

Institutional orders are rarely sent as a single market order. A pension fund buying 2 million shares of Apple can’t dump that into the order book without cratering the price. Instead, execution algorithms slice the parent order into hundreds or thousands of smaller child orders, dripping them into the market over minutes or hours.

Every execution algorithm navigates the same tradeoff:

$\text{Trade fast}\to \text{high market impact}\text{Trade slow}\to \text{high timing risk}$

Market impact is the cost of demanding liquidity now — your own buying pushes the price up against you (see kyle-lambda for the formal model). Timing risk (also called drift risk) is the danger that the price moves against you while you’re still only half done — and higher volatility makes this worse. Trade too fast and you pay through impact; trade too slow and the market drifts away from you.

Each algorithm is named after the benchmark it tries to match — the reference price the trader uses to evaluate whether the execution was good or bad.

VWAP — Volume-Weighted Average Price

VWAP is first and foremost a benchmark: the average price at which a stock traded during the day, weighted by volume at each price. If 60% of the day’s volume traded at $100 and 40% at $101, the VWAP is $100.40.

The VWAP algorithm tries to match this benchmark by dividing the parent order into time slices sized proportional to the stock’s historical volume profile. US equities typically follow a U-shaped intraday pattern — high volume at the open (9:30–10:00 AM) and close (3:00–4:00 PM), quiet midday. The algorithm sends more child orders during the busy periods and fewer during the lull, so the trader’s average fill price tracks the day’s VWAP.

When to use: the default institutional benchmark for orders with no urgency. A portfolio manager rebalancing across 50 names doesn’t care about getting the best possible price on each — they care about not paying more than the market average.

Weakness: the algorithm uses yesterday’s volume curve to predict today’s. It ignores price direction entirely: it will keep buying at the same pace into a rallying market, missing the opportunity to buy more when prices dip.

TWAP — Time-Weighted Average Price

TWAP is the simplest benchmark: the average price over the execution window, with each moment weighted equally regardless of how much volume traded.

The TWAP algorithm slices the order into equal-sized pieces spaced evenly over time. If you’re selling 120,000 shares over 2 hours, you sell 1,000 shares every minute.

When to use: stocks where the historical volume profile is unreliable — newly listed names, low-volume securities, or markets where the U-shaped pattern doesn’t hold (some Asian markets have a midday close).

Weakness: equal-sized slices hit harder during thin periods. Selling 1,000 shares at noon when only 5,000 shares trade per minute has 10x the relative impact of selling 1,000 shares at the open when 50,000 shares trade per minute.

Implementation Shortfall (IS)

The first two algorithms are mechanical — they follow a fixed schedule and ignore what happens to the price along the way. Implementation shortfall takes the opposite approach: it adapts.

Perold (1988) defined the IS benchmark as the gap between the paper return (what you would have earned trading instantly at the decision price) and the actual return after execution costs, impact, and drift. The decision price is typically the mid-price at the moment the trader submits the order — so the benchmark measures everything you lost by not being able to teleport the trade into existence.

Almgren and Chriss (2000) turned this into an optimization: given a model of how your trading impacts the price, find the trading trajectory (how much to trade at each moment) that minimizes:

${\min }_{\text{trajectory}}E[\text{cost}]+\lambda \cdot \text{Var}[\text{cost}]$

Notation

• $\lambda$ — risk aversion parameter. Higher $\lambda$ means the trader would rather pay known impact costs than gamble on how much the price drifts while waiting.

The solution is intuitive:

• Risk-averse trader ($\lambda$ high): trade aggressively upfront. You pay more impact but eliminate drift risk early. The trajectory is front-loaded.
• Patient trader ($\lambda$ low): spread execution over the full window. You minimize impact but accept that the price might move against you. The trajectory is back-loaded.

IS algorithms are the most sophisticated — but they require a calibrated model of how your orders move the price. A bad impact model produces a bad trajectory.

Comparison

	VWAP	TWAP	IS
Benchmark	Day’s volume-weighted average price	Simple time-weighted average price	Mid-price at order submission
Schedule	Proportional to historical volume	Equal slices over time	Optimized for impact vs. drift
Adapts to price?	No	No	Yes
Best for	Large orders, no urgency, liquid names	Low-volume or newly listed stocks	Orders where timing matters (alpha decay, event-driven)
Main weakness	Ignores price; uses stale volume profile	Equal slices hit harder in thin periods	Requires calibrated impact model

Reference

Harris, Trading and Exchanges, Ch. 14 (Algorithmic and Automated Trading Strategies) and Ch. 21 (Execution Costs). Almgren and Chriss (2000), “Optimal Execution of Portfolio Transactions.”

Retail order flow — PFOF and Rule 606

Everything above applies to institutional flow — pension funds, mutual funds, hedge funds. Retail orders take a different path entirely, one built around a simple economic fact: retail traders almost never have private information about where prices are heading.

The PFOF pipeline

PFOF (Payment for Order Flow) is the practice where a retail broker routes customer orders to a wholesaler — a high-speed market-making firm — in exchange for a per-share rebate.

Retail customer
    │
    ▼
Retail broker (Robinhood, Schwab, Fidelity)
    │  routes order + receives PFOF rebate
    ▼
Wholesaler (Citadel Securities, Virtu, Susquehanna, Two Sigma, Jane Street)
    │  internalizes: fills order at or inside NBBO
    ▼
Trade reported to TRF → [[trading-fundamentals|consolidated tape]]

NBBO (National Best Bid and Offer) is the best available bid and ask across all lit exchanges, consolidated by the SIP (Securities Information Processor — see trading-fundamentals). Wholesalers fill retail orders at prices equal to or better than the NBBO — the difference is called price improvement.

Why wholesalers pay for retail flow

Retail flow is uninformed — in the language of the Glosten-Milgrom model, retail traders are the uninformed participants whose orders carry no signal about the true value. A retail buy is equally likely to precede a price increase or decrease. This makes retail flow systematically profitable to fill: the wholesaler earns the spread without the adverse selection cost that dominates institutional flow.

This is classic flow segmentation: wholesalers cream-skim the benign (uninformed) flow, leaving the toxic (informed) flow — hedge funds, prop desks, algorithmic strategies — to interact on lit exchanges where market makers must price adverse selection into wider spreads.

Regulatory framework

Rule	What it requires	Effective
SEC Rule 606(a)	Quarterly public reports on order routing — which venues received orders, PFOF received, net payment per share	Ongoing
SEC Rule 606(b)(3)	Individual customer routing information on request — where your orders went	Ongoing
FINRA Rule 6151	Centralizes Rule 606 reports for public access	Ongoing
SEC Rule 605 (amended March 2024)	Monthly execution-quality statistics: effective spread, realized spread, price improvement, fill rate, speed. Now applies to broker-dealers with 100,000+ customer accounts, not just market centers	Compliance phasing in

The 2022–2025 reform saga

SEC Chair Gensler proposed a four-rule package targeting equity market structure:

Proposal	Status (as of early 2026)
Rule 605 amendments — expanded execution-quality disclosure	Adopted March 2024
Tick size / access fee reforms — half-penny ticks ($0.005)fortick-constrainedstocks(thosequotingpredominantlyattheminimum$0.01 increment); access fee cap reduced from $0.003\to$0.001 per share	Adopted September 2024
Reg Best Execution — codifying best execution as an SEC rule (currently only a FINRA/common-law obligation)	Proposed, not adopted
Order Competition Rule — requiring certain retail orders to be exposed to an auction before internalization	Proposed, not adopted

The tick size reform matters because many liquid stocks are “tick-constrained” — their natural spread would be narrower than $0.01, but the minimum tick prevents further tightening. Half-penny ticks allow these stocks to quote tighter spreads, potentially reducing the profitability of internalization (since the price improvement wholesalers offer is bounded by the tick size).

Source summary

Source	Coverage in this note
Harris, Trading and Exchanges (2003)	Agency/principal (Ch. 6, 13), block trading (Ch. 15), algorithms (Ch. 14, 21)
Perold, “The Implementation Shortfall” (1988)	IS benchmark definition
Almgren & Chriss, “Optimal Execution of Portfolio Transactions” (2000)	IS optimization framework
SEC Rule 10b-10	Trade confirmation disclosure
FINRA Rule 5310	Best execution obligation
FINRA Rule 2232	Bond markup disclosure
SEC Rules 605, 606	Execution quality and routing disclosure
SEC Release 34-99679 (March 2024)	Rule 605 amendments
SEC Release 34-101070 (Sept 2024)	Tick size and access fee reforms

See also

• trading-fundamentals — spreads, market makers, and the fundamental problem of trading
• order-books — CLOB mechanics, order types, and matching priority
• trading-venues — dark pools, RFQ, and the full venue taxonomy
• kyle-lambda — price impact theory underlying algorithmic execution
• ho-stoll-inventory-model — inventory risk that principals and block desks absorb
• spread-decomposition — decomposing the spread into adverse selection, inventory, and order processing components
• retail-trading-access — how retail accesses non-equity asset classes (Treasuries, corporate bonds, spot FX, options, ETFs)