A note on Markov perfect equilibria in a class of non-stationary stochastic bequest games (Q2401830)

In a stochastic paternalistic bequest game, an infinite sequence of generations labeled by \(t=1,2,\dots\) is considered. From its ancestor, generation \(t\) gets the endowment \(s_t \in (0,+\infty)\) which is segmented into consumption \(a_t\) and investment \(y_t=s_t-a_t\) according to some strategy \(c_t: a_t=c_t(s_t)\). The investment \(y_t\) determines the endowment of the successor according to some transition probability \(q_t(d s_{t+1}|y_t)\). Expected utility of generation \(t\) depends on its own consumption and consumption of its successor and is given by an expression \[ P_t(a,c_{t+1})(s)=u_t(a)+\int_{S} v_t(c_{t+1}(s'))q_t(ds'|s-a) \] with \(a=a_t\), \(s=s_t\), \(s'=s_{t+1}\). Here \(u_t(a_t)\), \(v_t(a_{t+1})\) describe utilities of consumptions. Markov perfect equilibrium is given by a condition \(\sup_a P_t(a,c^*_{t+1})(s)=P_t(c^*_{t}(s),c^*_{t+1})(s)\). In the article, the existence of a Markov perfect equilibrium in a non-stationary version of game is proved in both the stochastic and deterministic cases. Unlike some previous results, the proposed method allows to study models with unbounded state space and unbounded utility functions.