A sensitivity analysis of non-fungible tokens (NFTs) and comparative assets using time series forecasting

Abstract

Financial assets are central to economic stability, yet the macroeconomic sensitivity and predictability of emerging digital assets, particularly non-fungible tokens (NFTs), remain unclear. This study evaluates the responsiveness of NFTs, cryptocurrencies, and traditional assets to interest rate and inflation fluctuations using time series forecasting and sensitivity analysis. ARIMAX, Partial Least Squares, Ridge Regression, and Long Short-Term Memory (LSTM) models are employed to capture linear and nonlinear dynamics across asset classes. Using daily data from July 2017 to November 2024, results indicate that LSTM achieves superior predictive accuracy for highly volatile and nonlinear assets, although forecast reliability is limited by structural breaks and thin trading. Traditional assets such as bonds and gold display stable sensitivities to macroeconomic variables, reinforcing their hedging role. In contrast, digital assets exhibit higher volatility and weaker, less stable macroeconomic linkages. NFTs show low correlations with traditional assets, suggesting diversification potential, but low forecast error variance does not imply low risk. Cryptocurrencies demonstrate stronger macroeconomic sensitivity alongside greater instability. Overall, the findings reveal a trade-off between diversification benefits and forecast reliability when integrating digital assets into portfolios.