INTRODUCTION

Investment decisions occupy a central role among the determinants of growth. As empirical studies such as Levine and Renelt (1992) have revealed, fixed investment as a share of gross domestic product (GDP) is the most robust explanatory variable of a country's growth. DeLong and Summers (1991) also provides evidence emphasizing the correlation of investment in equipment and machinery with growth. Investment is also the most variable component of GDP, and therefore an understanding of its determinants may shed light on the source of cyclical fluctuations. Policy-makers are typically concerned about the ultimate impact of alternative policy measures on investment and its variability. Several theories of investment have emerged since the 1960s in an attempt to explain the determinants of investment. The most notable of these have been the neoclassical model of investment, the cost-of-adjustment-Q-theory model, the time-to-build model, the irreversibility model under uncertainty and the fixed-cost (S, s) model of lumpy investment.

Beginning with the neoclassical model developed by Jorgenson and his collaborators (see for example, Hall and Jorgenson 1967; Jorgenson 1963), investment theory distinguishes between the actual capital stock and the desired or optimal capital stock, where the latter is determined by factors such as output and input prices, technology and interest rates. In the neoclassical model of investment, an exogenous partial adjustment mechanism is postulated to yield a gradual adjustment of the actual capital stock to its desired level as is observed in the data. An alternative way of obtaining a determinate rate of investment is to assume the existence of convex costs of adjustment, as has been proposed by Eisner and Strotz (1963), Gould (1967), Lucas (1976) and Treadway (1968). Abel (1979) and Hayashi (1982) have shown that the standard adjustment cost model leads to a Tobin's Q-theory of investment under perfect competition and a constant returns to scale production technology. A complementary explanation assumes that it takes time to build productive capacity. (See Altug 1989, 1993; Kydland and Prescott 1982.)

A number of authors have emphasised irreversibility and uncertainty as important factors underlying the gradual adjustment of the capital stock. The notion of irreversibility, which can be traced back toMarschak (1949) and subsequently to Arrow (1968), was initially applied in the context of environmental economics where economic decisions, such as the destruction of a rain forest to build a highway, often entail actions that cannot be ‘undone.’