Observational Learning in Sequential Fundraising

יום רביעי 25.11 11:30 - 12:30

Abstract: We characterize equilibrium behavior in a sequential fundraising game and study what its outcome reveals about the state of the world. Investors move in a fixed order, observing earlier decisions and receiving private signals about whether the project is good or bad. The project proceeds only if a prescribed number invest; contributors gain from a good project and lose from a bad one, while a failed raise and declining to invest both yield zero. We give a recursive characterization of Markov perfect equilibrium, determining investment thresholds and state-contingent completion probabilities, with a unique pure equilibrium at regular states under the stated refinement. The solution identifies partial delegation, in which investors lower their evidence thresholds because later investors screen the project, and full delegation, in which an investor commits regardless of her signal even outside a cascade. Building on this characterization, we show that completion is more likely for a good project than for a bad one, strictly outside cascades, under the stated regularity condition where required. With bounded signals, however, learning remains limited: posterior odds following completion from an even prior have a ceiling independent of the number of investors, and completion near an optimistic cascade boundary reveals arbitrarily little. A voting extension shows how changing who bears the outcome changes equilibrium behavior: when one affirmative vote suffices, early voters demand weakly stronger evidence than a myopic decision-maker.

Speaker

Moran Koren

BGU