Do all types of demand have the same effect on output? To answer this question, I estimate a cointegrated vector autoregressive (VAR) model of consumption, investment, and government spending on US data, 1955US data, -2007. I find that: (1) economic growth can be decomposed into a short-run (transitory) cycle gravitating around a long-run (permanent) trend made of consumption shocks and government spending; (2) the estimated fluctuations are investment dominated, they coincide remarkably with the business cycle, and they are highly correlated with capacity utilization in both labor and capital; and (3) the long-run multipliers point to a large induced-investment phenomenon and to a smaller, but still significantly positive, government spending multiplier, around 1.5. The results cover a lot of theoretical ground: Paul Samuelson's accelerator principle, John Kenneth Galbraith's stress on consumption and government spending, Jan Tinbergen's investment-driven business cycle, and Robert Eisner's inquiries on the investment function. The results are particularly useful to distinguish between economic policies for the short and long runs, albeit no attempt is made at this point to inquire into the effectiveness of specific economic policies.