Many of the recent attempts to find evidence of downward nominal wage rigidity in micro data have suffered from a number of problems, including composition bias and the effects of measurement error. In order to avoid these problems we explicitly model the determinants of wage changes and the measurement process that leads to observable earnings changes, thereby directly tackling the question of whether and to which extent downward nominal wage rigidity exists in German micro data. We find a high degree of downward nominal wage rigidity, which for rates of inflation lower than three percent is shown to lead to higher equilibrium unemployment via the generated real wage wedge.