This paper estimates a Heterogeneous Expectations New Keynesian model to assess its ability to capture the dispersion of expectations observed in surveys. We use realized macroeconomic data and survey data on expectations in the estimation, exploiting both first moments (mean forecasts) and second moments (the cross-sectional forecast dispersion) as observables. The results reveal reasonable correlations between the model-implied and observed dispersion series. However, the model significantly overestimates the magnitude and volatility of disagreement, particularly for output, compared with their levels in survey forecasts.