USA Tariffs Effect: Machine Learning Insights into the Stock Market

TL;DR

The paper investigates how tariff announcements and the 2025 implementation affected the Australian stock market, focusing on the S&P/ASX 200 within a 2025 window. It adopts an integrated workflow combining Exploratory Data Analysis and machine learning-based regression (Linear Regression, SVR, kNN, Random Forest) with lagged and rolling features to forecast weekly closing prices and volatility. Across the analysis, Random Forest regression delivers the strongest predictive performance ($R^2=0.877$, $MSE=7047.39$, $MAE=67.64$), while nonlinear SVR underperforms due to parameter sensitivity. The study also documents a positive cross-market correlation between US and Australian indices, highlighting global spillovers and the potential value of cross-market features for policy-shock forecasting.

Abstract

The imposition of tariffs by President Trump during his second term had far-reaching consequences for global markets, including Australia. This study investigates how both the announcement and subsequent implementation of these tariffs, specifically on 02-Apr-2025, affected the Australian stock market, focusing on the S\&P/ASX 200 index over the period from 21-Jan-2025 to 25-Jul-2025. To accurately capture the significance and behavior of market fluctuations, the exploratory data analysis (EDA) techniques are applied. Furthermore, the impact of tariffs on stock performance is evaluated using machine learning-based regression models. A comparative assessment of these models is conducted to determine their predictive accuracy and robustness in capturing tariff-related market responses.

Figures (7)

• Figure 1: Daily closing prices of the Australian stock market over the same period. The price movement pattern reflects market behaviour that can be compared with the USA trends.
• Figure 2: Daily closing prices of the USA stock market over the observed period. The plot captures short-term fluctuations and long-term trends based on historical data.
• Figure 3: Correlation heatmap depicting the linear relationships between key stock market variables (Open, High, Low, Close) from the USA and Australian (AUS) financial datasets.
• Figure 4: Comparison of actual and predicted stock market values using different models. The plot illustrates the performance of each model in tracking observed data over time, highlighting variations in predictive accuracy.
• Figure 5: $R^2$ values and error metrics illustrating the predictive performance of different models.