I investigate the effect of financial innovation on portfolio risks when traders have belief disagreements. I decompose traders’ average portfolio risks into two components: the uninsurable variance, defined as portfolio risks that would obtain without belief disagreements, and the speculative variance, defined as portfolio risks that result from speculation. My main result shows that financial innovation always increases the speculative variance through two distinct channels: by generating new bets and by amplifying traders’ existing bets. When disagreements are large, these effects are sufficiently strong that financial innovation increases average portfolio risks, decreases average portfolio comovements, and generates greater speculative trading volume relative to risk-sharing volume. Moreover, a profit-seeking market maker endogenously introduces speculative assets that increase average portfolio risks.

We provide a continuous-time “risk-centric” representation of the New Keynesian model, which we use to analyze the interactions between asset prices, financial speculation, and macroeconomic outcomes when output is determined by aggregate demand. In principle, interest rate policy is highly effective in dealing with shocks to asset valuations. However, in practice monetary policy faces a wide range of constraints. If these constraints are severe, a decline in risky asset valuations generates a demand recession. This reduces earnings and generates a negative feedback loop between asset prices and aggregate demand. In the recession phase, average beliefs matter because they not only affect asset valuations but also determine the strength of the amplification mechanism. In the ex ante boom phase, belief disagreements (or heterogeneous asset valuations) matter because they induce investors to speculate. This speculation exacerbates the crash by reducing high-valuation investors’ wealth when the economy transitions to recession, which depresses (wealth-weighted) average beliefs. Macroprudential policy that restricts speculation in the boom can Pareto improve welfare by increasing asset prices and aggregate demand in the recession.

This article proposes a welfare criterion for economies in which agents have heterogeneously distorted beliefs. Instead of taking a stand on whose belief is correct, our criterion asserts that an allocation is belief-neutral efficient (inefficient) if it is efficient (inefficient) under any convex combination of agents’ beliefs. Although this criterion gives an incomplete ranking of social allocations, it can identify positive- and negative-sum speculation driven by conflicting beliefs in a broad range of economic environments.