In this paper, we investigate supply and demand shocks in the US natural gas market, focusing on how the effects of these shocks have changed over time. Using a sign-identified structural vector autoregression (SVAR) model that allows for both time-varying parameters and stochastic volatility, we decompose the real price of natural gas into supply shocks, aggregate demand shocks driven by changes in the US economic activity, precautionary inventory demand in anticipation of changes in future demand-and-supply conditions, and residual demand shocks not otherwise accounted for by the previous three shocks. Using quarterly data from 1976 to 2015, we find that an unanticipated supply disruption raises natural gas prices, reduces the aggregate economic demand, and lowers the precautionary inventory demand, while negative aggregate demand shocks, on the other hand, depress natural gas prices, reduce natural gas production, and increase precautionary inventory demand. We also find that following a negative precautionary inventory demand shock, aggregate demand driven by real economic activity declines marginally, and the marketed natural gas production and real prices decrease as well. Our results further suggest that such impact responses have evolved considerably over time with changing market conditions.