Quantitative Trading

Execution Research

Cost models, trading algorithms and transaction cost analysis for every order the firm places, from a Treasury note to a basket of two thousand stocks.

Purpose

Costs are certain. Returns are not.

A forecast of return is uncertain; the cost of acting on it is not. For high-turnover strategies, trading costs can be as large as the edge they are trying to capture.

The execution research group sits in Chicago with the quantitative trading teams, but it serves the whole firm. Fixed income, asset management and systematic portfolio managers all route orders through the same cost models and receive the same post-trade reports.

Orders analyzed per year
About 1.9 million
Brokers and venues reviewed
Quarterly
Asset classes covered
Equities, options, futures, Treasuries, credit
Research team
9 researchers and engineers
Implementation shortfall of a hypothetical stock purchase, in basis points A waterfall chart. Starting from the price when the decision was made, delay adds 3 basis points of cost, the bid-ask spread 2, market impact 9, fees and commissions 1, and the cost of shares left unfilled 2, for a total shortfall of 17 basis points. 01020 Cost, basis points DelaySpreadImpactFeesUnfilledTotal 3291217
Hypothetical example for a purchase of 200,000 shares; not Lenox results.

Implementation shortfall

The gap between the decision and the fill.

Implementation shortfall, a measure introduced by André Perold in 1988, compares a portfolio’s actual result with a paper portfolio that traded instantly at the price when the decision was made. Every cost the paper portfolio avoided shows up in the gap.

Market impact is usually the largest piece, and the hardest to see. Our models, like most in the field, find that impact grows roughly with the square root of order size relative to daily volume: doubling an order raises its cost per share by about 40%, not 100%.

Trading slowly reduces impact but increases delay and the chance that the price moves away. Choosing that trade-off for each order, given its urgency and size, is the core of the work.

Before, during, after

Every order is measured three times.

  1. Pre-trade

    Estimate

    Before an order is released, the model estimates its cost under several trading speeds. The portfolio manager sees the estimate and can resize or delay the trade. For bonds, the system suggests how many dealers to ask for quotes.

  2. In-trade

    Monitor

    Algorithms trade within participation limits, typically no more than 10% to 15% of volume. Traders are alerted when an order drifts outside its expected cost range.

  3. Post-trade

    Review

    Each fill is compared with arrival price, interval benchmarks and later prices. Results by broker, algorithm and venue feed back into the cost model each month.

Governance

Best execution is a duty, not a preference.

As an investment adviser, Lenox owes clients a duty to seek the most favorable overall terms for their trades. Lowest commission is only one part of that.

Two colleagues working at desks by a window
  • Best Execution Committee

    Portfolio managers, traders, the Chief Compliance Officer and the risk team meet each quarter to review broker, algorithm and venue results and approve changes to the broker list.

  • Research paid by the firm

    Lenox pays for third-party research from its own resources. Client commissions are used only to pay for execution.

  • Fair allocation

    When orders for several client accounts are combined, fills are allocated by a written procedure decided before the trade, at the same average price.

  • Errors corrected at our cost

    If a trading error harms a client account, Lenox restores the account. Gains from errors stay with the client.

Metrics

What we measure, and what each number hides.

Transaction cost analysis measures
MeasureWhat it comparesLimitation
Implementation shortfallActual result against the price when the decision was madeNoisy for single orders; meaningful only across many
Arrival price slippageAverage fill price against the price when the order reached the traderIgnores delay between decision and release
VWAP slippageAverage fill price against the volume-weighted average price over the order’s lifeA large order moves VWAP itself, so it can flatter poor execution
MarkoutsPrice movement in the seconds to days after each fillSeparates temporary from lasting impact, but needs large samples
Participation rateOrder volume as a share of market volume while it tradedA control, not a cost; high rates usually mean higher impact
Dealer cover (bonds)Winning quote against the second-best quote in a request for quoteShows competition among dealers, not the true market level

Strategy descriptions are not an offer or solicitation. Investing involves risk, including possible loss of principal.