I examine the possibility of information‐based trading in a multiperiod consumption setting. I develop a necessary and sufficient condition for trade to occur. Intertemporal substitution introduces a desire to correlate current consumption with future aggregate shocks. When agents have heterogeneous time‐inseparable preferences, information differentially affects relative preferences for current and future consumption, making information‐based trading mutually acceptable. The no‐trade result continues to hold if there is no aggregate shock, or if agents have either homogeneous or time‐separable preferences.

Persistent performance in venture capital is routinely interpreted as evidence for skill. We present a dynamic model of delegated investment with endogenous fund heterogeneity and deal flow, which generates performance persistence without skill differences and predicts mean reversion in long-term performance. Investors working with multiple funds use contingent payments and tiered contracts to induce proper project nurturing and managerial effort. Successful funds receive continuation contracts that tolerate investment failure and encourage innovation, and subsequently finance entrepreneurs through a path-dependent assortative matching favoring incumbents. Recent empirical findings corroborate the model’s general implications, and the economic mechanisms are robust to short-term contracting, endogenous bargaining, and double moral hazard issues.

Economic interactions often involve sequential actions, observational learning, and contingent project implementation. We incorporate all‐or‐nothing thresholds in a canonical model of information cascades. Early supporters effectively delegate their decisions to a “gatekeeper,” resulting in unidirectional cascades without herding on rejections. Project proposers can consequently charge higher prices. Proposal feasibility, project selection, and information aggregation all improve, even when agents can wait. Equilibrium outcomes depend on crowd size, and project implementation and information aggregation achieve efficiency in the large‐crowd limit. Our key insights hold under thresholds in dollar amounts and alternative equilibrium selection, among other model extensions.