Annuitizing at a bounded, absolutely continuous rate to minimize the probability of lifetime ruin (Q6072267)

An individual has wealth \(W(t)\) at time \(t\) and possesses \(A(t)\) annuities. An annuity pays at rate \(1\) as long as the individual is alive; that is, the individual has an income at rate \(A(s)\) at time \(s\) from annuities. It is not possible to sell annuities, but the individual can buy new annuities at rate \(q(s) \in [0,\bar q]\). Interest at a constant rate \(r\) is paid for the wealth. The individual consumes the wealth at a constant rate \(c\). Thus the processes considered fulfil \begin{eqnarray*} d W(s) &=& (r W(s) - c + A(s))\;d s - \bar a(s) q(s)\; d s\;,\\ d A(s) &=& q(s)\; d s\;, \end{eqnarray*} where \(\bar a(s)\) is the present value of an annuity paying one unit continuously. The goal is then to minimise the probability that there is not enough wealth to cover the consumption during the lifetime. It is then shown that it is optimal either to buy new annuities at the maximal rate \(\bar q\) or not to buy further annuities. Two boundaries characterise the optimal strategy: if the wealth is below a buy-bound, no annuities are bought. There is a stop boundary, such that once above this boundary the consumption can be paid through the interest and the annuities held. Between the boundaries, annuities are bought at the maximal rate until the stop boundary is met. There is a further boundary between the two bounds such that above this boundary, the stop boundary will be reached before life-time ruin occurs. That is, above this boundary the optimal strategy has ruin probability 0. The result is proved by showing basic properties of the model, by solving a HJB equation, and proving a verification theorem. Letting the maximal rate \(\bar q\) tending to infinity, a known result is recovered that either a lump sum purchase to reach the stop boundary or no purchase is optimal.