Neither. Both measure the average impact of all previous articles published in a journal in a given time frame. Given the skewed distribution of article impact, this is a poor predictor of the quality of any individual article currently published in the journal.