We elicit reservation wage premia for relocating to two Ukrainian cities, using a household survey conducted in mid-April to mid-July 2024 during the Russian invasion of Ukraine: high-risk Kharkiv (near the frontline) and moderate-risk Kyiv. Risk tolerance is a strong predictor of willingness to move to Kharkiv - the most risk-averse have roughly half the odds of the most risk-tolerant - but matters much less for Kyiv. This asymmetry is difficult to reconcile with the hypothesis that risk tolerance merely proxies for general mobility preferences. Separately estimating the elasticity of intertemporal substitution (EIS~0.04), we find that including it renders risk tolerance insignificant for Kyiv but not for Kharkiv - a pattern illuminated by the Epstein-Zin separation of risk aversion and the EIS: risk aversion adds predictive power only when danger is high, while the EIS operates equally for both cities as a common relocation-cost channel. The very low EIS implies that relocation incentives structured as future benefits may be ineffective; frontloaded subsidies are more likely to influence behavior.