Skip to content
All library documents

Estimating Stock Locate Fees Without Historical Data

Article Quant Q&A · Author: David
Read the original

Summary

The document asks how to estimate stock borrow locate fees for a short-selling strategy when a historical fee database is unavailable. It identifies potentially useful predictors, including liquidity, trading volume, short interest, and sector or industry, and notes that fee estimates must be weighed against a model’s expected return when deciding which trades to take.

It provides no proposed model, empirical results, or tested proxies; it is a request for ideas rather than a demonstrated method. Any estimate would therefore need validation against observed fees as they become available, and the listed stock attributes alone do not establish how accurately fees can be predicted. The practical objective is to balance expected trade value with participation, while accounting for an uncertain borrowing cost.

Key ideas

  • Locate fees can vary across stocks and over time, but the author lacks comprehensive historical fee data.
  • Liquidity, volume, short interest, and industry classification are suggested as possible fee predictors.
  • A short-selling decision should compare expected trade return with the estimated locate cost.
  • The document poses the modeling problem but offers no tested method or evidence.

Tags

EquitiesExecutionStatisticsRisk management
Full text
# Seeking a Model to Extrapolate Locate Fees for Short Selling in Absence of Historical Data


# Seeking a Model to Extrapolate Locate Fees for Short Selling in Absence of Historical Data












I'm in the process of developing an automated stock trading algorithm, with short selling being a significant part of the strategy. A key factor in deciding whether to short a stock is the associated "locate fee". This fee can vary significantly across stocks and over time.

While I understand that the most accurate way to predict these fees would be through historical data, I currently don't have access to a comprehensive database of past locate fees. As such, I'm looking for suggestions on constructing a mathematical or statistical model to extrapolate or estimate the locate fees based on other available stock attributes or market data.

Some considerations and information:

- The trades are driven by a predictive model that provides an expected return for each trade.

- I'm looking for a balance between maximizing the expected value of the model and maximizing the number of trades.

- I have access to other stock data, including liquidity measures, trading volume, short interest, sector/industry classifications, etc.

Has anyone faced a similar challenge or can provide insights on how to approach this problem? Are there specific models, statistical techniques, or proxies that have proven effective in estimating locate fees in the absence of direct historical data?

Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.