The world of securities finance is entering a transformative era driven by advances in artificial intelligence and increasing regulatory transparency. A groundbreaking proof-of-concept (POC) study conducted by Advanced Securities Consulting (ASC) has demonstrated how deep learning (DL) models significantly outperform traditional ARIMA (AutoRegressive Integrated Moving Average) models in forecasting key market variables. This milestone could reshape how securities lenders optimize their returns and collateral strategies, particularly as they prepare to implement FINRA’s SLATE disclosures.
ASC’s Breakthrough in Predictive Analytics
At the IMN Global Securities Finance Conference in Ft. Lauderdale on February 4, 2024, ASC presented the findings of its first-phase Deep Learning Proof-of-Concept (POC I). The study provided compelling evidence that deep learning transformer models (DL models) can forecast daily average rebates for cash-collateralized equity loans more accurately than ARIMA models.
Unlike ARIMA, which relies on linear assumptions and historical trends, DL models dynamically adjust to complex and nonlinear market interactions. This makes them particularly well-suited to handle the volatility of securities lending markets. By leveraging publicly available data—specifically, those data points included in the forthcoming SLATE disclosures—ASC’s POC I demonstrated that financial institutions can optimize lending fees and rebates without requiring access to proprietary lending datasets.
Goldilocks Rate Model: A New Approach to Lending Fees
One of the most innovative contributions of ASC’s POC I is the development of the “Goldilocks Rate” model. This model determines an equilibrium fee/rebate rate that maximizes lender profitability while maintaining borrower demand. Unlike traditional models, which often impose rigid assumptions on pricing elasticity, the Goldilocks Rate ensures a balanced approach to cash and collateral management.
By incorporating multiple interdependent variables—such as borrower demand, loan volume, and collateral types—the DL model allows market participants to set more precise lending fees. This optimization is critical as market participants prepare to navigate the increased transparency introduced by FINRA’s SLATE disclosures.
Regulatory Alignment: SEC Rule 10c-1 and SLATE’s Role
The SEC’s Rule 10c-1a aims to enhance transparency in securities lending, requiring lenders to disclose critical data related to loan pricing, volume, and counterparty risk. Historically, the securities lending market has been characterized by opacity, allowing intermediaries to maintain inefficiencies that increase borrowing costs.
ASC’s POC I directly supports the SEC’s regulatory intent by demonstrating that reliable forecasting can be achieved solely using the variables that will be publicly available through SLATE. This validation has two key implications:
- Democratizing Market Access: Securities lenders and borrowers can make accurate rate forecasts without proprietary data, leveling the playing field across market participants.
- Reducing Systemic Risk: Increased transparency aligns with the post-2008 regulatory push to mitigate shadow banking risks, as the Dodd-Frank Act envisioned.
Implications for Data Vendors and Market Participants
The SEC recognized that the transparency requirements of Rule 10c-1a could pose challenges for data vendors by providing more comprehensive public data, which may decrease the demand for their services. However, data vendors could address this challenge by enhancing their analytical products.
The Commission preliminarily believes that the proposed Rule may cause a loss in revenue for the commercial vendors of securities lending data. . . .The Commission preliminarily believes that a potential mitigating factor that could reduce the severity of this loss in revenue would be that commercial data vendors could offset some of the impact of lowered demand for their data by enhancing their related businesses using the data in the proposed Rule. [86 FR 69802, 69839]
As FINRA’s SLATE disclosures become operational, major data vendors will gain access to granular loan-level data, allowing them to refine their predictive models. The ASC POC I indicates that these vendors could enhance their analytical capabilities further by integrating deep learning techniques. The result would be a more efficient and transparent securities lending market, with improved risk assessments and pricing strategies.
A strong synergy exists with widely adopted industry platforms like FIS’s Lending Pit and Loanet solutions, which numerous financial institutions depend on for their securities lending workflows. By incorporating advanced AI-driven analytics into these well-established offerings, firms can streamline their trade lifecycle processes, reduce operational risks, and obtain real-time insights across various lending activities.
Moreover, financial institutions that leverage AI-driven forecasting will be able to navigate the upcoming regulatory landscape more effectively. Those who continue relying on traditional models may find themselves at a competitive disadvantage as market conditions evolve.
Conclusion: A Turning Point for Securities Lending
ASC’s POC I findings represent a pivotal moment for the securities lending industry. By proving the superiority of deep learning models over ARIMA for rebate forecasting—and validating the sufficiency of SLATE disclosures for predictive analytics—ASC has demonstrated how AI can enhance market efficiency and regulatory compliance.
As a firm affiliated with CSFME, ASC’s research aligns with the Center’s mission to improve transparency, reduce risk, and support sound financial regulation. The intersection of advanced AI modeling and regulatory oversight has the potential to reshape securities lending, creating a more accessible and equitable marketplace for all participants.
With SLATE disclosures on the horizon, financial institutions should start integrating deep learning methodologies into their forecasting strategies—or risk falling behind in an increasingly data-driven world.