Ratio of samples with positive labels for each subgroup in the protect class in the Adult, COMPAS and COMPAS (fair) datasets.

We compare percentages present in the true labels of the real data and the predicted labels. Analogously, we measure the ratio of samples with positive label present in the synthetic generated data and predicted labels for datasets generated using distinct synthesizer techniques. Predictions(R) represents ratio of positive prediction labels of an experiment where model trained on synthetic data was evaluated on real data, and Predictions(S) ratio of positive prediction labels of an experiment where model trained on synthetic data was evaluated on synthetic data.