Does context matter more for hypothetical than for actual contributions? Evidence from a natural field experiment