In a sharing game the players' choices yield a revenue, each player's choice carries a cost, and a player's payoff is a portion of the revenue minus the player's cost. Such games are appealingly simple devices for partially aligning individual incentives with organizational goals, but their equilibria may be inefficient, i.e., at an equilibrium the surplus (revenue minus the sum of the costs) may not be maximal. Sharing games as a general class have not been well studied. We start a general theory of sharing games by going beyond the common economic setting, where strategy sets are continua and cost and revenue functions are smooth. We include games in which some of a player's strategies are equally costly, and revenue changes when he switches from one of them to another. We consider several large classes of reward functions, including nondecreasing residual (NDR) functions, in which residual (revenue minus rewards) does not drop when revenue increases. That class includes budget-balancing functions, where residual is always zero. To focus the discussion, we examine a ``Folk Claim", which asserts that at every inefficient equilibrium shirking, in some sense, occurs. We show that in NDR games a complementarity condition indeed insures that no one squanders at equilibrium (spends more than at an efficient profile). But when we drop complementarity, the situation changes sharply, and there are games with compelling equilibria, at which some players squander. The shirking/squandering distinction is particularly important in tracing the effect of technical improvement on the surplus shortfall at a sharing game's equilibrium. The paper also obtains conditions for existence of (pure-strategy) equilibria and finds, in particular, that every finite game in which rewards are linearly related has an equilibrium.