We develop a sufficient-statistic test to determine when observed asset-price run-ups are welfare-reducing bubbles rather than privately efficient risk-sharing or information-driven revaluations. Embedding heterogeneous beliefs, financial frictions, and pecuniary externalities in a dynamic general-equilibrium model, we show that the wedge between the private and social value of capital equals the product of (i) the elasticity of future financial conditions to the current price and (ii) the covariance between the asset’s price and future aggregate consumption.

Using U.S. stock and housing data from 1929 to 2023, we estimate this wedge in real time and find that episodes exceeding a threshold of 0.4% of GDP are followed by output losses whose present value exceeds the ex-ante capital gain, indicating inefficient bubbles.

Counterfactual policy experiments reveal that a state-contingent macro-prudential tax on leveraged purchases set to the estimated wedge eliminates 70–85% of the subsequent welfare losses while raising average consumption by 0.6%. Because the tax is triggered only when the sufficient statistic crosses the threshold, it avoids the average 1.2% consumption cost of unconditional leaning-against-the-wind rules.

Optimal interventions are larger and more prolonged when:

• the bubble is concentrated in housing
• bank leverage is high, and
• monetary policy is constrained by the zero lower bound.

Our findings imply that policy intervention is not universally necessary, but it is welfare-enhancing precisely when private agents underestimate the systemic spillovers of their collective exuberance.