Aggregate investment is the most volatile component of GNP. Furthermore, investment decisions are at the basis of the process of capital accumulation and, therefore, of growth and future prosperity. Understanding the dynamics and the determinants of investment is crucial for economists and policy makers alike. Our project will be based on and will develop two different streams of the most recent literature; the most recent theory of investment with non-convex adjustment costs and more generally on models with non-convex adjustment costs (consumption, labour demand); and the theory and practice of aggregation of individual behaviour. We pursue two main goals: first, to estimate a flexible and realistic (S,s) model for firm investment. Particular care will be devoted to identify the effect of economic fundamentals on the desired stock of capital and on the time series properties of the shocks of interest. Second, we will study how the dynamics of microeconomic adjustment is translated into the dynamics of aggregate data. In particular, we will identify which moments of the cross sectional distributions are sufficient to describe the dynamics of aggregate investment. We will use a new dataset: the ARD, a yearly survey of about 15,000 plants in the UK gathered by the ONS since 1980.