Transaction Cost Accounting — the arithmetic that separates a real edge from a paper one

Every gross return this wiki reports has to survive the arithmetic of actually trading it. This page gives the components of realistic trading cost (spread, market impact, commissions, borrow), documents how academic papers systematically mis-state them, and gives the numbers on which specific strategies these costs have killed — the load-bearing check behind What Counts as an Edge Here: The Evidence Bar This Wiki Applies to Every Technique's "does it survive costs" requirement.

Bid-ask spread

the gap between highest bid and lowest ask, computed as (Ask − Bid) ÷ Midpoint × 100, or in basis points as (Ask − Bid) ÷ Ask × 10,000 (quantt.co.uk). The effective spread — what a trade actually pays, since orders can execute inside the quote — is 2 × |Trade Price − Midpoint|, not the posted spread. Concretely: a fund trading £10 million notional in a FTSE 100 stock at a 0.01% spread pays roughly £1,000 in spread cost per round trip (quantt.co.uk). Quoted spreads were far wider historically — 20-60 bps through the 1980s-90s — before compressing after US decimalization (2001), electronic trading, and venue competition (WebSearch, unverified at primary source).

Market impact

decomposed into *permanent* impact (a lasting price shift reflecting the information content of the trade) and *temporary* impact (the immediate cost of demanding liquidity, which reverts after the order completes) — the Almgren-Chriss framework (smallake.kr; emergentmind.com). Almgren's 2005 model fit impact against roughly 700,000 Citigroup US equity trade orders (Dec 2001-Jun 2003), parameterizing impact as a function of daily volatility, shares outstanding, average daily volume, and trade size, using power-law relationships between trade size and price impact (unverified, WebSearch snippet only). For medium and large institutional traders, market impact — not spread or commission — is the dominant cost (WebSearch, unverified).

Commissions

near-zero explicit fees at modern retail brokers; institutional commissions run roughly 0.25-1+ bps depending on negotiation and order size (WebSearch, unverified). Damodaran (NYU Stern) states commissions "do not constitute a large fraction of the total cost of trading except in small-size trades" — spread and impact dominate for anything larger.

Short-selling borrow cost

stock loan fees range from ~0.25%/year (easy-to-borrow) to 100%+/year (hard-to-borrow), set by supply and demand. As of 2024, the value-weighted average fee across liquid "General Collateral" US equities was about 80 bps annualized (WebSearch, unverified at primary source). Acadian Asset Management reports financing costs can add "50 to 200 bps of expense, and sometimes more" to a portfolio's bottom line depending on the securities traded. Before shorting, SEC Regulation SHO's "Locate Requirement" obliges a broker-dealer to borrow the security or have a bona fide belief shares can be borrowed by settlement (sec.gov, primary source) — an operational constraint, not just a cost.

How Academic Papers Under/Overstate Costs

Papers frequently ignore transaction costs entirely, understate them, or bury them in a footnote; per one practitioner source, "the gap between backtest performance and live trading exists almost entirely in transaction costs" (hedgefundalpha.com, WebSearch summary — treat as practitioner-attested, not peer-reviewed). A specific, common backtest error: filling trades at the midpoint of the bid-ask spread, when real execution requires crossing the spread (WebSearch, unverified). Academic factor returns are typically computed pre-cost on hypothetical portfolios that may include illiquid names, whereas live implementation must absorb spread, impact, commissions, and taxes (WebSearch, unverified).

Pulling the other way: Frazzini, Kabiller & Pedersen (referenced here as "Frazzini et al. 2012") analyzed $1.7 trillion of live trade-execution data from a large institutional manager across 21 developed equity markets over 19 years and reported actual trading costs "an order of magnitude smaller than previous studies suggest" (WebSearch snippet, not read at primary source — unverified). This does not generalize to every strategy or segment; it is evidence that some academic cost estimates were inflated relative to what large, careful institutional executors actually pay, not that costs are negligible everywhere.

A specific gap in the papers: permanent price impact from information revealed by large orders is rarely modeled — most papers model only the temporary spread cost — even though for medium/large trades the permanent component can dominate (emergentmind.com).

Cost Thresholds That Kill Strategies — Documented Numbers

These are the concrete numbers that connect this page to specific strategy verdicts elsewhere in the wiki (unverified at primary source unless noted; all via WebSearch snippet):

- Momentum (Novy-Marx & Velikov 2016): average transaction cost of 48.4 bps/month over 1963-2013, versus 5.7 bps/month for size and 5.5 bps/month for value — momentum's cost burden is roughly 8-9x the other two factors. Documented strategy capacities: $5B for momentum, $170B for size, $50B for value. See Cross-Sectional Momentum in Equities — the strongest documented anomaly, and how much of it survives costs. - Turnover threshold: per Novy-Marx & Velikov, anomalies with one-sided monthly turnover below ~50% still generate significant net-of-cost spreads when cost-mitigated; few strategies above that turnover survive. - Buy/hold spreads — letting existing positions ride rather than actively trading out — is reported as "the single most effective simple cost mitigation strategy" (same source). - Institutional round-trip costs, historical: median 6.24 bps per rebalance for NYSE/Nasdaq stocks, Aug 1998-Sep 2013 (Keim & Madhavan, via WebSearch). Earlier period (Jan 1995-Apr 1997), average round-trip cost ~27 bps, with round-trip cost by strategy type: value traders 0.45%, index traders 1.09%, momentum traders 2.04% — a >4x cost gap between value and momentum implementation. - Recent retail (2025 study): round-trip execution costs ranged 7-46 bps of dollar value (WebSearch, unverified — source paper not identified beyond the snippet). - Factor premium erosion, general: gross factor premia of roughly 6-8%/year get "substantially" reduced by costs and capacity constraints; Garg et al. (2019, Financial Analysts Journal, syzygyassetmanagement.com) state implicit market-impact costs "may substantially erode a strategy's expected excess returns" while remaining largely invisible to investors. - Idiosyncratic momentum (Blitz, Hanauer, Vidojevic): reported to produce "only slightly lower [gross] returns but greatly reduce crash risk," nearly doubling the Sharpe ratio, by cutting turnover and implementation cost relative to standard momentum (WebSearch, unverified at primary source).

(unverified) All Novy-Marx/Velikov, Frazzini et al., Keim/Madhavan, and Blitz/Hanauer/Vidojevic figures above come from WebSearch snippets summarizing the papers, not from reading the papers directly — treat the exact figures as reported-by-search, not independently confirmed.

Rules of Thumb for Turnover-Adjusted Net Returns

Net return net of turnover cost:

$$r_{t+1} = \left(1 - \kappa \sum_i |w_{i,t} - w_{i,(t-1)}|\right) (w_t)' r_{t+1}$$

where $w_{i,(t-1)}$ is the pre-rebalance weight, $w_{i,t}$ the post-rebalance weight, $\kappa$ the proportional transaction cost (bps/10,000), and $r_{t+1}$ the return vector (WebSearch, unverified at primary source). $\kappa$ scales with the sum of absolute weight changes across all positions — i.e., cost is a direct, linear function of turnover, which is why turnover appears everywhere in this section as the key strategy-killing variable.

A related framing: a strategy's expected value has two parts — expected alpha skill (positive) and implementation cost (negative), where the cost term depends on turnover and execution skill (WebSearch, unverified). A simple breakeven rule derived across multiple turnover-literature sources (not a single quoted formula):

Breakeven cost = Gross alpha ÷ Turnover ratio.

This is the formula operationalized step-by-step in Computing a Strategy's Transaction-Cost Threshold — the four-step check that decides whether a documented edge is tradable, which works a numeric example against a specific strategy documented in this wiki.

Why this page matters for the wiki's objective

This wiki's bar for "a real edge" requires stating whether an effect survives realistic transaction costs and the cost threshold where it disappears. This page is the reference that makes that requirement checkable: it supplies the actual cost components (spread, impact, commission, borrow), the documented failure points (momentum's 48.4 bps/month versus its ~5.5-5.7 bps/month peers, the $5B momentum capacity ceiling, the >4x value-vs-momentum round-trip cost gap), and the arithmetic (turnover-adjusted return, breakeven-cost formula) that every strategy page in this wiki should be checked against before it is called verified.

What does NOT work

Backtests filled at the midpoint of the bid-ask spread systematically overstate net returns, because live execution must cross the spread — this is flagged repeatedly in the practitioner sources above as a beginner mistake, not a subtle one. Cost estimates taken from a single academic paper without checking whether the sample period predates decimalization (pre-2001) or reflects institutional-scale execution should be treated skeptically: the historical spread data above show 20-60 bps quoted spreads in the 1980s-90s versus far tighter spreads after 2001, and Frazzini et al.'s finding that live institutional costs ran an order of magnitude below prior academic estimates shows the two can diverge sharply.

Related

- What Counts as an Edge Here: The Evidence Bar This Wiki Applies to Every Technique — states the four-part evidence bar (effect size, proof regime, cost survival, out-of-sample persistence); this page is the working reference for the "cost survival" leg of that bar. - Computing a Strategy's Transaction-Cost Threshold — the four-step check that decides whether a documented edge is tradable — takes the breakeven-cost formula and turnover-adjusted return formula given here and walks a step-by-step worked numeric example against a real strategy. - Cross-Sectional Momentum in Equities — the strongest documented anomaly, and how much of it survives costs — the strategy whose cost numbers (48.4 bps/month, $5B capacity) are the clearest documented case here of costs constraining a real, positive-gross-return effect. - Short-Term (1-Week/1-Month) Reversal — Large in 1962-1986, Roughly Zero Unconditionally After 2000 — a strategy whose short holding period makes it especially exposed to the spread and bid-ask-bounce costs detailed in this page's "Components" section.