preference and technology parameters. Our model shares the basic quantitative char-
acteristics of monopolistic models that have been studied by, for example, Gali (1994)
or Rotemberg and Woodford (1995). In the following we concentrate on the investment
fluctuation and its effect on production and consumption.
We use the following utility specification:
U(ct, ht) = c1-σ/(1 - σ) - h,1+ν/(1 + ν) (24)
where σ ≥ 0 and ν ≥ 0. This simple specification allows us to obtain some analytical
insight as we see later, although the labor hour will not be stationary in the balanced
growth path in this specification. We set the technological growth rate at g = 1 and
inflate the depreciation rate δj by the observed productivity growth rate so that the
simulated sectoral oscillations continue to match the oscillations in the data. From the
utility specification we obtain the equilibrium price conditions immediately:
wt = ctσhtν (25)
rt = (ct+ι∕ct)σ /β (26)
A contemporaneous equilibrium (yt, ct, ht, wt) given kt, it, rt is determined by (25) and:
yt = ((1 — 1∕ξ )(1 — α)∕wt)(1-0*αNkt (27)
wtht = (1 — 1∕ξ )(1 — α)y (28)
yt = ct + it (29)
The first equation is derived by aggregating the optimal production level when the
capital is given. The second equation is obtained by aggregating the optimal employ-
ment given capital. It shows that the labor share is equal to (1 — 1∕ξ)(1 — α). The
third equation is a product market equilibrium condition. Given these equilibrium re-
lations, the equilibrium path (kt , it, rt) is determined by the capital accumulation (2),
the equilibrium interest rate (26), and the selection algorithm for it with the optimal
threshold rule (13).
We resort to numerical simulations to solve the equilibrium path. In the simulation,
we assume that the representative household and monopolists have a static expectation
on future investment. Namely, the expected future investment is set at the time average
level ∑j∙ δjkj. Computational difficulty is the reason we do not solve for a perfect
foresight equilibrium. Since the investment crucially depends on the details of the
configuration of producers capital positions, solving the perfect foresight path requires
prohibiting computational loads. Also, it is not realistic to suppose that the agents are
able to form a perfect foresight. Besides the computational problem, the agents would
16