the indeterminacy of capital level in the smooth economy. When the wage and interest
rate are not fixed, the aggregate capital does have a unique time average level. How-
ever, the attraction power of the time average in the dynamics of aggregate capital is
vanishingly small as the wage and interest rate bacomes insensitive to production.
4 Conclusion
This paper explores a mechanism of investment propagation as a fundamental shock to
the business cycle fluctuations. We consider industrial sectors which are characterized
by constant returns to scale technology and monopolistic pricing. Demand for inter-
mediate inputs forms a positive feedback of capital adjustment in the interindustrial
relations. We suppose that the sectoral capital exhibits an intermittent adjustment
where a large investment occurs occasionally. Under this environment, we derive the
distribution function of the propagation size. The propagation size has a large variance
when the real wage and real interest rate do not respond sensitively to the production
level. In the limit case when the wage and interest rate are fixed, the variance of capital
growth rates does not depend on the level of dissaggregation.
Simulations show that the investment propagation mechanism ab ove can explain
the aggregate fluctuations of the U.S. economy quantitatively. We specify the represen-
tative household’s utility as a separable function in leisure and consumption and solve
for the equilibrium paths. The results show that the standard deviation, the correla-
tions between production and investment and consumption, and the autocorrelation of
production, investment, and consumption match the U.S. postwar business cycles well.
Thus we show that, given the magnitude of oscillations that a manufacturing sector
exhibits, the sectoral oscillations can add up to the aggregate fluctuations through the
investment propagation mechanism with the correct second moments of the business
cycle variables.
The paper leaves three points for further explorations. First, the simulation shows
that the correlations between two demand components and production are not strong
enough simultaneously. It stems from that the consumption responds weakly to income
when capital level is fixed. The behavior of representative household needs to be modi-
fied in such a way that the income effect becomes strong, for example by incorporating
the liquidity constraint. Secondly, the intertemporal substitution of consumption in
the simulation is set larger than the evidence for the U.S. economy suggests. Thirdly,
the deterministic oscillation of sectoral capital is assumed. It is no doubt an over-
simplification that a sectoral capital jumps in one period and depreciate capital over
many years. Incorporating the time-to-build of capital would make the sectoral oscil-
25