The conventional practice in estimating DSGE models is to rely on seasonally adjusted data. While convenient, this approach distorts the microeconomic foundations of the model. An alternative is to model seasonality explicitly, but this often introduces severe misspecification. This paper proposes a middle ground: using year-over-year growth rates instead of quarter-over-quarter growth rates, which allows the model to endogenously determine the seasonal adjustment. This approach greatly improves forecast accuracy by more than 20% while keeping the internal consistency of the model. Moreover, we show that model misspecification and seasonal adjustment can offset each other, implying that seasonality should be treated as model-specific rather than imposed exogenously. Empirical results for U.S. and Russian data confirm that structural seasonality improves forecasting performance, and model fit relative to conventional seasonal adjustment methods.