Seismic piezocone (SCPTu) data compiled from loess soil sites in the greater Christchurch, New Zealand area are used with multiple linear regression to develop an empirical correlation for predicting shear wave velocity (Vs) from cone penetration test (CPT) data. The performance of the model is assessed through analysis of the variation in prediction bias with different CPT parameters and through comparisons of measured and predicted Vs profiles. Comparisons with the recently-developed Christchurch-specific general soil CPT-Vs correlation show that this general soil correlation (based on alluvial, marine, estuarine, and peat/swamp soils) significantly underpredicts the Vs of the loess soils, likely due to the cemented nature of these deposits, thus demonstrating the need for the loess-specific correlation presented here.