Abstract
This article uses Bi-directional Reiterative Truncated Projected Least Squares (BD-RTPLS) to estimate annual dGDP/dG (GDP, Gross Domestic Product and G, government spending) multipliers for the USA between 1930 and 2008. The analysis is redone with quarterly data from 1947 to 2008. To account for the influence of omitted variables, BD-RTPLS produces a separate dGPP/dG estimate for every observation in the data set. I find that whenever the US government increases government spending by an unusually large amount in a given year or quarter, the resulting government spending multiplier plummets. This is not good news for the current US government which is hoping that a huge fiscal stimulus package will rescue the USA from the current crisis.