Most macroeconomic forecasters underestimated global investment during the late 1990s. One potential reason was that the models they were using were insufficiently disaggregated. In this paper, an empirical model is estimated whose out-of-sample forecasts largely predicted the global investment boom of the late 1990s. The main factor behind the improved model performance is the distinction between investment in ICT assets and investment in other assets, using disaggregated investment data provided by the OECD. In line with previous studies on US and UK investment performance, ICT investment is estimated to be much more responsive to changes in the real user cost of capital. In particular, panel and seemingly unrelated regression (SUR) estimates suggest very strong relative price effects on ICT investment for all G7 countries and Australia. The data also allow an examination of the effects of possible deflator mismeasurement; but within our framework, the measurement of investment using harmonised, rather than national, deflators is not found to have a material impact on forecasting performance.
