An estimation method for the dipole moment is proposed in this work. Polarity is caused by the effect of different neighboring groups in a molecule. A second-order group contribution approach has been chosen due to its ability to account for this effect. It could be shown that the use of second-order groups can cause a significant increase in prediction quality compared to the conventional first-order approach. The prediction quality of the proposed method is good, even for higher values of the dipole moment. Those higher than about 3.5 D are however underestimated. The model also enables the prediction of the small but finite dipole moments of hydrocarbons. Adequate estimations are not limited to compounds from the training set, but also compounds from a test set show good prediction results. Hence, it can be concluded that this method is able to give quick and reliable predictions for not yet measured dipole moments.