We analyze the payout decision of a financially-constrained
firm that cannot raise external funds. Exogenous cash flows are generated
by a two-state Markov regime-switching process and are positive in one re
gime and negative in the other regime. The firm is motivated to pay out di
vidends to impatient shareholders but is also motivated to accumulate cash
within the firm to make required payments when cash flow is negative. If
the cash on hand is insufficient to make these payments\, the firm termina
tes\, thereby losing its claim on future cash flows. The optimal payout po
licy can be described as a form of precautionary saving. However\, contrar
y to conventional wisdom about precautionary saving\, we find that such sa
ving falls in response to a mean-preserving increase in the variance of ca
sh flows.

\n

\nWe extend the model to include a capital investment
decision. Instead of smooth and convex adjustment costs\, we introduce an
upper bound on the investment-capital ratio\, which leads to a bang-bang
solution for investment. Thus optimal investment and dividends are each go
verned by trigger policies. We derive and interpret analytic solutions for
these triggers. The trigger for optimal investment equates marginal q to
the "static cost of funds\," which is technically the marginal valuation o
f a dollar of cash within the firm. Despite the linear homogeneity of the
value function in the state variables (capital and cash on hand)\, average
q and marginal q are not equal. At the optimal trigger for investment\, a
broader measure of average q\, equal to the value of the firm divided by
the sum of the replacement cost of its capital and its cash on hand\, equa
ls marginal q. Finally\, we analyze a particular myopic value of a unit of
capital\, and show that if this myopic value of capital is sufficiently h
igh\, specifically when the myopic value exceeds one\, the firm can ignore
the financing constraint when making its investment decision. Otherwise\,
the firm will invest in capital only if its cash on hand is greater than
or equal to an optimally-derived trigger.