Lithium, Uranium and Tech : A Quantitative Dive into ETF Time Series

• KWEB - KraneShares CSI China Internet ETF
• LIT - Global X Lithium & Battery Tech ETF
• URA - Global X Uranium ETF

Objectives

• Conduct a time series analysis of three sector-specific ETFs over a 10-year horizon.
• Evaluate price dynamics, log-returns, and distributional properties.
• Test for stationarity and assess short-term dynamics using ARIMA models.
• Investigate long-term equilibrium relationships using the Engle–Granger cointegration test.

Methods & Models

Descriptive Analysis

• Weekly log-returns
• Mean, standard deviation
• Skewness, kurtosis
• Histogram & QQ-plots
• Correlation matrix
• ACF (Autocorrelation Function)

Univariate Time Series Modeling

• ARIMA(6,1,0) estimation per ETF
• Residual diagnostics
• Impulse Response Functions (IRF)

Stationarity Testing

• Augmented Dickey-Fuller (ADF)
• KPSS Test

Cointegration Analysis

• Engle–Granger 2-step test (KWEB vs LIT)

Data

• Source: Yahoo Finance
• Frequency: Weekly
• Period: January 2015 - January 2025

Tool

• Python (Jupyter Notebook)
• Report written in LaTeX

Key Findings

• All three ETFs exhibit non-normal return distributions with excess kurtosis and skewness.
• ARIMA models showed no statistically significant autoregressive lags at the 5% level.
• Shocks observed in log-prices tend to dissipate rapidly (low persistence).
• ADF and KPSS tests confirm non-stationarity of log-price series.
• No cointegration found between KWEB and LIT, suggesting no long-term equilibrium relationship.

---

Thank you for exploring this project. I am continuously learning and developing my quantitative finance skills, and I welcome all feedback, suggestions, or ideas for improvement.

Gianni Marchetti
📧 [email protected]