Paper: arXiv 2404.07225

Authors: Anoop Kumar, Suresh Dodda, Navin Kamuni, Rajeev Kumar Arora

Abstract

This study examines the effects of macroeconomic policies on financial markets using a novel approach that combines Machine Learning (ML) techniques and causal inference. It focuses on the effect of interest rate changes made by the US Federal Reserve System (FRS) on the returns of fixed income and equity funds between January 1986 and December 2021. The analysis makes a distinction between actively and passively managed funds, hypothesizing that the latter are less susceptible to changes in interest rates. The study contrasts gradient boosting and linear regression models using the Double Machine Learning (DML) framework, which supports a variety of statistical learning techniques. Results indicate that gradient boosting is a useful tool for predicting fund returns; for example, a 1% increase in interest rates causes an actively managed fund’s return to decrease by -11.97%. This understanding of the relationship between interest rates and fund performance provides opportunities for additional research and insightful, data-driven advice for fund managers and investors

Complexity vs Empirical Score

  • Math Complexity: 6.5/10
  • Empirical Rigor: 8.0/10
  • Quadrant: Holy Grail — high math complexity, high empirical rigor

Why this score: The paper employs advanced causal inference techniques (Double Machine Learning, DAGs) and complex ML models (gradient boosting, VAR), indicating high mathematical sophistication, while its large-scale real-world dataset (7,000+ funds) and detailed methodological description demonstrate strong empirical rigor.

Research Flowchart

  flowchart TD
  A["Research Question: Impact of US FRS Interest Rates on Fixed Income & Equity Funds"] --> B["Data & Framework"]
  B --> C["Methodology: Double Machine Learning"]
  C --> D["Computation: Gradient Boosting vs. Linear Regression"]
  D --> E["Key Findings & Outcomes"]
  
  subgraph B ["Data & Framework"]
      B1["Period: Jan 1986 - Dec 2021"]
      B2["Target: Actively vs. Passively Managed Funds"]
      B3["Input: Interest Rate Changes (US FRS)"]
  end
  
  subgraph C ["Methodology: Double Machine Learning"]
      C1["Causal Inference Approach"]
      C2["Separates Prediction from Treatment Effect"]
  end
  
  subgraph D ["Computation: Gradient Boosting vs. Linear Regression"]
      D1["Gradient Boosting Outperforms Linear Regression"]
      D2["Predicts Fund Returns Accurately"]
  end
  
  subgraph E ["Key Findings & Outcomes"]
      E1["Result: 1% Interest Rate Increase ⟶ -11.97% Return Drop for Actively Managed Funds"]
      E2["Finding: Passively Managed Funds Less Sensitive to Rate Changes"]
      E3["Implication: Data-Driven Strategy for Investors & Fund Managers"]
  end

Evidence Card

What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card

Proposes

The paper proposes a framework that combines Double Machine Learning with gradient boosting and linear regression to analyze the causal impact of Federal Reserve interest rate changes on fund returns.

Source sentence
The study contrasts gradient boosting and linear regression models using the Double Machine Learning (DML) framework, which supports a variety of statistical learning techniques.

Assumes 6

  • The analysis assumes a one‑way causal impact of Federal Reserve interest rate policies on market behavior, with no feedback loops. market
    Source sentence
    the study concentrates on the one-way impact of interest rate policies set by the US FRS on market behaviors.
  • Directed Acyclic Graphs (DAGs) are used to represent causal assumptions, implying no feedback loops in the causal structure. statistical
    Source sentence
    With no feedback loops, Directed Acyclic Graphs (DAGs) provide a visual depiction of causal assumptions by indicating the direction of influence among variables
  • After detrending, the majority of variables are assumed to be stationary for analysis. data
    Source sentence
    successfully decreased variable correlation and made the majority of variables stationary.
  • A seven‑month lag is assumed to be the appropriate lag structure for the VAR models used to select variables. model
    Source sentence
    Based on AIC scores, the study settled on a seven-month lag.
Show 2 more
  • Funds with insufficient historical observations are excluded to ensure uniform time‑series length. data
    Source sentence
    funds with inadequate historical data were excluded, guaranteeing uniformity throughout the time series study.
  • Double Machine Learning reduces bias and variance in parameter estimation, addressing confounding and high‑dimensional data issues. model
    Source sentence
    By enabling any ML technique to estimate functions and so lowering bias and variation in parameter estimations, DML tackles the problems of confounding and high-dimensional data.

Data

  • Bloomberg fund dataset · Bloomberg fundamentals commercial
    December 1985 to December 2021 · U.S. traded mutual funds (14,816 funds, reduced to 1,948 passive and 5,000 active)
    Source sentence
    the final dataset included 1,948 passively managed funds and 5,000 actively managed funds.
  • Federal Reserve Economic Data (FRED) · FRED macro public
    Source sentence
    The Federal Reserve Economic Data (FRED) was the primary source of information for the macroeconomic variables used in the study

Costs

Not applicable The paper reports no trading or execution result.

Finds 5

  • A 1% increase in interest rates leads to an actively managed fund's return decreasing by about 11.97%. negative
    Source sentence
    a 1% increase in interest rates causes an actively managed fund's return to decrease by -11.97%.
  • Gradient boosting captures the interest‑rate effect mainly for actively managed funds, while linear regression shows a substantial negative effect for both active and passive funds. statistical
    Source sentence
    While gradient boosting reveals this effect mainly in actively managed funds, linear regression models show a substantial negative causal effect on both fund types.
  • The DML models may be over‑fitted, suggesting caution in interpreting the large adverse effects observed for active funds. other
    Source sentence
    the examination indicated that the DML models might have been overfitted, which calls for a cautious interpretation of the noteworthy adverse effects observed in actively managed funds.
  • Residual analysis shows no patterns indicating non‑linearity or heteroskedasticity, supporting model robustness. performance
    Source sentence
    The absence of patterns in residual plots (refer to Fig. 3) that would have suggested nonlinearity or heteroskedasticity in the models suggests their robustness.
  • Passive funds show no significant impact from interest‑rate changes, possibly due to data collection and classification challenges. negative
    Source sentence
    the absence of a meaningful correlation with passively managed funds raises questions and points to the need for more research

Checked by others

No other paper in the archive has been found to support or contradict these findings (compared with 29 related papers). Coverage is partial, so this is neither replication nor refutation.

Code

No repository by the authors found in the paper.

Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.

Cited References (23)

  1. The Factors Influencing Given Investment Choices of Individuals (2015) — link
  2. Which Variables Predict and Forecast Stock Market Returns? (2016) — link
  3. Causal inference in statistics : a primer (2016)
  4. The Turbulent Market of Modern Debt-overleveraged and Promoter-dominated Corporations (2018) — link
  5. Financial Markets Prediction with Deep Learning (2018) — link
  6. Double machine learning with gradient boosting and its application to the Big N audit quality effect (2020) — link
  7. On the relationship of machine learning with causal inference (2019) — link
  8. Double/debiased machine learning for difference-in-differences models (2020) — link
and 15 more
  1. Double machine learning-based programme evaluation under unconfoundedness (2022) — link
  2. Causal Inference Meets Machine Learning (2020) — link
  3. Impact of Tax Incentives on Foreign Direct Investment: Evidence from Africa (2021) — link
  4. Using Machine Learning to Quantify the Multimedia Risk Due to Fuzzing (2021) — link
  5. An analysis of the robustness of UAV agriculture field coverage using multi-agent reinforcement learning (2023) — link
  6. Roulette-Wheel Selection-Based PSO Algorithm for Solving the Vehicle Routing Problem with Time Windows (2023) — link
  7. Predicting the Infection Level of COVID-19 Virus Using Normal Distribution-Based Approximation Model and PSO (2023) — link
  8. Stock Price Prediction Using Artificial Intelligence Based on LSTM–Deep Learning Model (2023) — link
  9. Detection of a facemask in realtime using deep learning methods (2024) — link
  10. Enhancing the predictive capability of a mathematical model for pseudomonas aeruginosa through artificial neural networks (2024) — link
  11. Advancing Audio Fingerprinting Accuracy with AI and ML: Addressing Background Noise and Distortion Challenges (2024) — link
  12. Enhancing End-to-End Multi-Task Dialogue Systems: A Study on Intrinsic Motivation Reinforcement Learning Algorithms for Improved Training and Adaptability (2024) — link
  13. Prediction of suicidal risk using machine learning models (2024) — link
  14. Alternative agriculture land-use transformation pathways by partial-equilibrium agricultural sector model: a mathematical approach (2024) — link
  15. From simulations to reality: enhancing multi-robot exploration for urban search and rescue (2025) — link

0 of 23 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.