The present study has been based on a neoclassical, aggregative model of economic growth in which the bias of technological progress is determined through entrepreneurs' maximization behavior. The innovation possibility function, which relates the proportionate rates of increase in the efficiency of factor inputs in the production process, is assumed to be smooth, concave to the origin, and bounded from above. Under these assumptions the following points have been demonstrated.
1. If technological progress is factor-augmenting every-where, and if the elasticity of substitution remains either less than, equal to, or greater than unity, then there exists a unique steady-state growth process. If the elasticity of substitution remains equal to unity, factor-augmenting technological progress must always exhibit Hicks-neutrality, and hence Harrod -neutrality either. The steady-state growth process is globally stable, and the pattern of income distribution between wages and profits is determined quite exogenously. If the elasticity of substitution (not necessarily being constant) remains positive but less than unity, the steady-state growth process is also stable in the large. The character of technological progress may vary continuously over time within the Harrodian trichotomy, but it tends to be neutral as the economy converges to the steady-state growth. path. Ifthe elasticity of substitution always exceeds unity, however, the unique steady-state growth process has a saddle-point instability.
2. When the elasticity of substitution is a given constant, the model can be extended to cover the more general type of technological progress. If the elasticity of substitution is either less than, or equal to unity, the above conclusions concerning the existence, uniqueness, and stability of the steady-state growth process also apply. If the elasticity of substitution is greater than unity, on the other hand, there may be no steady-state growth path for a sufficiently large value of the elasticity of substitution. In such a case the rate of growth of capital stock will ultimately tend to be larger and larger, and the share of profits will tend to unity. For the value of the elasticity of substitution not far from unity, there may be several steady-state growth processes. When there are at least two such processes, they are alternatively stable and unstable. The steady-state growth process associated with the smallest rate of growth is stable, and the one associated with the largest rate of growth is unstable. According to most of the recent empirical studies, it would be quite unlikely that the value of the elasticity of substitution between labor and capital in an economy actually exceeds unity.10) If these empirical estimates are valid, the present study can give some justification for a supposition that technological progress tends to become Harrod-neutral in the long-run.
Nevertheless, we must admit that our method of approach has several limitations. It should be noted, among other points, that technological progress has been taken as occurring at a given and foreseeable rate. This means that our analysis can be applicable to a small, though seemingly becoming more important, part of the entire innovational activity. Furthermore, it may be quite unrealistic to suppose that the position of the innovation frontier is independent of entrepreneurs' behavior and technological progress is costless. The rate of technological progress, if taken as a result of research and development activities, must somehow be related to the amount of expenditure devoted to such activities. The effect of a random and exogenous change in the position of the innovation frontier may be analysed even in the present framework by applying the method of comparative dynamics. A more general theory of innovation will have to be developed, however, in order to explain the extent, as well as the character, of technological progress.

View full abstract