The usual mechanism through which government spending can be effective in increasing output in a liquidity trap emphasizes the role of aggregate demand. Higher government spending generates a lower expected real interest rate through a higher real wage, which translates into higher inflation. This lower real rate increases aggregate demand. I present new evidence that casts doubt on the empirical relevance of this mechanism. In particular, liquidity traps occur exclusively in recessions, which appear to be situations where higher government spending generates less inflation than in expansion times. To rationalize this, I build a New Keynesian model with search and matching frictions in the labor market and a downward rigid nominal wage. When solved using global methods, this model generates asymmetric fluctuations of recruiting costs along the business cycle. This permits the model to generate (i) a higher government spending multiplier in recessions vs expansions and (ii) a significantly higher multiplier at the zero lower bound without relying on a counterfactually large reaction of wages and inflation, both of which are in line with empirical evidence. Decomposing the contributions of recession versus liquidity trap dynamics, I find that the latter explain only around 37% for a multiplier of 1.5.