In the non-experimental methods, there is not any manipulation of antecedents like there are in experiments, Non-experiments also lack the random assignment of treatment conditions. In experiments, we manipulate an independent variable to see what would happen to the dependent variable.
Essentially, we are looking for a cause and effect in experiments, non-experimental methods such as correlations do not measure what A would do to B. Correlational designs look at how A relates to B (Hansen and Myers, 2012). Experiments offer data that is more concrete than what non-experiments do.
Correlations, for example, are used to give us an insight into what might be happening, experiments give us an insight into what will happen, assuming that in this case, our hypothesis is correct.
Correlations would be a good starting point to an experiment but they cannot be used to determine something that is absolute. Experiments do a better job at assessing causality because there isn’t a third variable problem like there are in non-experimental designs like correlations.