Spearman's statistic is essentially constructed by taking the correlation between the ranks. The tests' results presented preceding are very questionable: Pearson's coefficient can always be calculated, but in order to derive confidence bands such as the ones that we see here, the data needs to be normal and linear (both assumptions are violated in this case). Nevertheless, Spearman does not require any distributional assumptions.

Because we take correlations just for descriptive purposes, and because we usually want to see how much linear relationship exists in the data, we usually prefer Pearson's coefficient. However, if we want to see how monotone the relationship is, we should use Spearman's rank statistic.