Using non-linear methods, we argue that existing estimates of government spending multipliers in expansion and recession may yield biased results by ignoring whether government spending is increasing or decreasing. In the case of OECD countries, the problem originates in the fact that, contrary...