We estimate the macroeconomic effects of government spending by constructing a new series of defense spending surprises using a high-frequency identification approach. We identify events in the U.S. federal budget process that reveal news about defense spending, focusing on dates when different versions of defense appropriations bills are passed in the House or the Senate. We then quantify the defense spending surprise associated with each event using daily stock returns of large U.S. defense contractors. A positive defense spending surprise predicts a persistent increase in future defense spending that is entirely passed through to total government spending. On average, the identified spending increases are financed with a combination of taxes and debt and are met with a muted monetary policy response. We estimate the dynamic causal effects of defense spending and find sizable increases in output and consumption. Our estimates imply a cumulative fiscal multiplier---the cumulative output response divided by the cumulative spending response---of 1.2 over a five-year horizon. Our results indicate that government spending stimulates (``crowds in'') economic activity in the private sector. A standard heterogeneous-agent New Keynesian model matches our empirical evidence.