This paper uses historical data since mid-19th century to test the validity of Wagner's Law for the Italian economy. Unlike the previous studies, we accommodate possible nonlinear asymmetric effects of total gov- erment spending and GDP toward their long-run equilibrium. Our results show the presence of a threshold cointegrating relationship between the two variables with significantly different error correction adjustments in normal and extreme regimes. Particularly, we find the validity of Wagner's Law from 1862 to 2009, only when we take into account strong nonlinear responses of government spending during the WWI and WWII period. Robustness checks clearly recognize nonlinear behaviour of government expenditure driven by military spending.