We calibrate and simulate a neoclassical growth model with a variable elasticity of substitution production function and three types of technological change: labour-augmenting, capital-augmenting and investmentspecific. In this framework, we find that the decline in US labour share was caused by a large decline in capital efficiency, which led to a decrease in the ratio of effective capital to effective labour in a context in which capital and labour are gross complements. Moreover, the decline in the relative price of investment contributed to reducing the fall in US labour share, while the increase in the economic depreciation rate of US fixed assets accounted for a small reduction in US labour share. 636 ECONOMICA [OCTOBER implicitly recognizes it when he writes: 'Until the laws of thermodynamics are repealed, I shall continue to relate outputs to inputs-i.e. to believe in production functions'. Both the intensity and direction of the changes in the relative factor shares will depend on the shape of the production function and the bias of technological change.According to neoclassical theory, if markets are frictionless, and perfectly competitive, then gross labour share equals the output elasticity for labour, which is an increasing (resp. decreasing) function of the ratio of effective capital to effective labour if and only if the elasticity of substitution between capital and labour is lower (resp. higher) than 1. Moreover, the lower the elasticity of substitution between capital and labour, the higher the fall in gross labour share which is caused by the decrease in the ratio of effective capital to effective labour. 4 Therefore the technological forces causing a fall in the ratio of effective capital to effective labour lead to a fall in gross labour share if capital and labour are gross complements, while they lead to a rise in gross labour share if capital and labour are gross substitutes.Although there is some debate, empirical evidence increasingly indicates that the elasticity of substitution between capital and labour is lower than 1. Our own estimations support this hypothesis. However, Piketty (2014) estimates an elasticity of substitution between 1.3 and 1.6, while Karabarbounis and Neiman (2014) obtain an average estimate of 1.25. More recently, Koh et al. (2017) estimate that the elasticity of substitution between a constant elasticity of substitution (CES) aggregator of two types of capital (intellectual property and equipment plus structures) and labour is 1.13. Many studies estimate elasticities of substitution lower than 1. In their pioneering study, Arrow et al. (1961) estimate an elasticity of substitution of 0.57 using data on 24 manufacturing industries in a sample of 19 countries. David and van de Klundert (1965) and Kalt (1978) estimate elasticities of substitution equal to 0.32 and 0.76, respectively. Based on a literature survey, Hamermesh (1993) concludes that the elasticity of substitution is likely to be in the range 0.3-0.7. Chirinko et al. (2004), using an extensive panel of 1,...