The risk of the presence of haloacetic acids in drinking water as chlorination by-products and the shortage of experimental mutagenicity data for most of them requires a research work. This paper describes a QSAR model to predict direct mutagenicity for these chemicals. The model, able to describe more than 90% of the variance in the experimental activity, was developed with the use of the spectral moment descriptors. The model, using these descriptors with multiplicative effects provides better results than other linear descriptors models based on Geometrical, RDF, WHIM, eigenvalue-based indices, 2D-autocorrelation ones, and information descriptors, taking into account the statistical parameters of the model and the cross-validation results. The structural alerts and the mutagenicity-predicted values from the model output are in agreement with references from other authors. The mutagenicity predicted values for the three haloacetic acids, which have available experimental data (TCAA-Trichloroacetic acid, BDCAA-Bromodichloroacetic acid, and TBAA-Tribromoacetic acid), are reasonably close to their experimental values, specially for the latest two.