This externality, which combines what we shall call a consumer-surplus effect (during the
life of the new product) and what Aghion and Howitt term an intertemporal spillover efSect
(extending over the lives of later products), measures (log A)/p in terms of the utility
metric in (1) and (2).
Second, a successful innovator generates a negative externality for extant industry
leaders. In effect, the innovator "destroys" the producer surplus of the firm it displaces.
Aghion and Howitt term this the business-stealing effect. This adverse effect is compounded
by a multiplier, as the loss of income suffered by owners of the displaced firm means less
demand and less profits for all remaining industry leaders. In all, the innovation causes
instantaneous profits of others to fall by A - 1. This flow must be discounted at rate L +p,
the effective discount rate for profits, which takes into account the expected rate of arrival
of the next innovation in each industry. So the total negative externality imposed by the
innovator equals (A - 1 ) / ( ~+p). For low or high values of A the adverse effect is larger,
while the combined consumer-surplus and intertemporal-spillover effects dominate when
A takes on an intermediate value."
The optimal growth rate is more likely to exceed the equilibrium rate when Lla, is
large; i.e. in large economies as measured in units of (R&D) efficiency labour. Then the
optimal rate of innovation is great. Larger values of L* reduce the size of the businessstealing
effect per unit output without changing the size of the consumer-surplus effect
per unit output.
The optimum can be decentralized here by means of a tax or subsidy on R&D outlays.
Let T be the multiple (or fraction) of R&D costs borne by the firm, with T > 1 for a tax
and T< 1 for a subsidy. With such a policy in effect the no-arbitrage condition (11) is
replaced by one with the left-hand side divided by T. An increase in T shifts the l7l-I
curve upward in Figure 1 and so generates an equilibrium with greater spending and less innovation. A decrease in T has the opposite effects on resource allocation. By varying
T, the government can achieve any point along LL, including of course the optimum.