In this paper a model of aggregate investment is derived which incorporates fixed investment costs and capital market imperfections on the micro-level. Aggregate investment reacts nonlinearily with respect to aggregate shocks to productivity and liquidity of firms. Employing non-parametric kernel estimation methods to analyze a sample of annual account data of UK companies, these nonlinearities also show up empirically. Furthermore a difference in strength between the long- and the short-run effect of liquidity on investment is found, which is inconsistent with models that explain the empirical correlation of investment and liquidity solely as the result of some long-run relationship like liquidity-dependent costs-of-capital. Moreover, as a side-result when imposing a linear structure, we find, as reported in previous studies, that financially constrained firms appear to be less influenced by liquidity than unconstrained ones. However, this finding disappears when the investment function is estimated without any restrictions on the functional form and thus has to be attributed to a misspecification error, which is avoided by the estimation strategy of this paper.