Forecasting Value-at-Risk and Expected Shortfall for Bitcoin and Ethereum after Spot ETFs

Published: 29 October 2026| Version 1 | DOI: 10.17632/zfsbm2tnbk.1
Contributor:
sheeba kapil

Description

This dataset contains the data, code and results needed to reproduce every table and figure in the paper “Lower Risk, Clustered Losses: Forecasting Value-at-Risk and Expected Shortfall for Bitcoin and Ethereum after Spot ETFs” (Suman, Kapil and Kapil, 2026). The paper studies how the launch of U.S. spot exchange-traded funds for Bitcoin (11 January 2024) and Ether (23 July 2024) changed the level of tail risk and the performance of tail-risk forecasting models. It produces one-day 95% and 99% Value-at-Risk (VaR) and Expected Shortfall (ES) forecasts from seven models: historical simulation, Normal, Student-t, GARCH-t filtered historical simulation, conditional extreme value theory, ES-CAViaR and a one-factor GAS model. It evaluates them with coverage, independence and dynamic quantile tests, three ES backtests (McNeil–Frey, Acerbi–Székely, Du–Escanciano), the Nolde–Ziegel traffic light and the Model Confidence Set. Contents: data/raw: daily BTC-USD and ETH-USD prices from Yahoo Finance, 1 January 2018 to 26 September 2026. code: Python scripts. data_prep.py downloads and prepares the data, including Coin Metrics community prices for 2016–2017 (used only to start the long estimation windows), seven control cryptocurrencies and the Cboe VIX. forecast.py produces the forecasts, evaluate.py runs all backtests, s2_returns.py and s4_pretrend.py run the attribution analysis, figs.py draws the figure, and run_all.py runs everything in order. output: daily VaR and ES forecasts for each model (forecasts_btc.csv, forecasts_eth.csv), all backtest results (evaluation.json), simulated null distributions for the Acerbi–Székely test, the attribution and placebo results, and Figure 1. README.md explains how to run the code. CHANGES_FROM_ORIGINAL_CODE.md lists the corrections made to an earlier version of the analysis.

Files

Institutions

Categories

Finance, Cryptocurrency

Licence