Why is the Government Spending Multiplier Larger at the Zero Lower Bound ? Not (Only) Because of the Zero Lower Bound

I develop a New Keynesian model with search and matching frictions, in which the government buys goods produced by the firms. I solve the non-linear model globally and examine the magnitude of government spending multipliers in and out of the ZLB. I distinguish the cases of Nash-bargained and rigid real wages. The model with Nash-bargained wages gives results that do not differ qualitatively from the existing literature. It generates a high multiplier at the ZLB due to a higher elasticity of real marginal cost to aggregate demand in those conditions—which is at odds with empirical evidence. The model with rigid real wages exhibits job rationing (Michaillat (2012a)) and also delivers a higher multiplier at ZLB than in normal times. In this setup however, the multiplier is not larger because real marginal cost is more responsive to aggregate demand, but just the opposite. With a slack labor market, it is easy to recruit and real marginal cost is less responsive to aggregate demand. This is in line with available empirical evidence