Empirical studies analyzing the determinants of US presidential popularity have delivered quite inconclusive results concerning the role of economic variables by assuming linear relationships. We employ penalized spline smoothing in the context of semiparametric additive mixed models and allow for flexible functional forms and thus possible nonlinear effects for the economic determinants. By controlling for the well-known politically motivated covariables, we find strong evidence for nonlinear and negative effects of unemployment, inflation and government consumption on presidential approval. Additionally, we present new results in favor of nonparametric trivariate interaction effects between the macroeconomic covariables. Copyright © 2014 John Wiley & Sons, Ltd.