AN ALGORITHM FOR DYNAMIC PROGRAMMING OF ECONOMIC GROWTH.
CALIFORNIA UNIV BERKELEY CENTER FOR RESEARCH IN MANAGEMENT SCIENCE
Pagination or Media Count:
The paper studies the computational properties of an algorithm for optimal resource allocation over time, and applies the algorithm to a model of the U.S. economy. The algorithm is applicable to multisector models with differentiable production and objective functions, and uses a method of successive approximation by linear-logarithmic functions, in which the recursive features of dynamic programming are combined with exact formulas for optimal solutions in the linear-logarithmic case. Memory and computing requirements go up approximately linearly with the number of state variables, and experience indicates that the algorithm can handle a model with 10 sectors and 50 time periods in a few minutes on a machine like the IBM 7094. The algorithm is applied to a four-sector empirical model of the U.S. economy, and various optimal paths are compared with the observed path of the economy from 1910 to the present. The sensitivity of the optimal solution to the parameters of the system is studied in some detail. A horizon of 50 gives a good approximation to the solution for an infinite horizon this can be improved by suitable approximations for the value of final stocks. Author
- Economics and Cost Analysis
- Operations Research