Observational studies do not include experimenter directed treatments. In these kinds of studies the subjects, knowingly or not, choose their own treatment. There are many situations where it is just not practical to "assign" subjects to different treatments. Furthermore, if one is willing to just "observe" differences and outcomes we have whole histories of people doing things over decades that we can carefully follow and analyze. You would hardly want to subject children to "lead-based" paints, or to certain "fogged mosquito" insecticides and then wait forty years to see if there any long term effects of such exposure. On the other hand, we have whole populations of people who were exposed to such "treatments" forty years ago. We could observe such people, along with others of a similar age and background who were not so exposed to the treatments, to see if we can detect differences between the groups.

In a similar fashion, we could observe the injuries caused in car crashes among those people wearing or not wearing seat belts. Again, we could "experiment" in such situations using crash dummies, but we are not about to use real people as subjects in such experiments. However, since there are numerous well documented, accidents involving both those using and those not using seat belts, we can "observe" differences without doing an actual experiment.

This latter example gives us a chance to compare an "experiment" with and "observational study". Using crash dummies, we can have identical dummies moving at identical speeds, sitting in identical car models, and hitting identical objects in our experiment. The only difference between experimental runs will be whether or not the crash dummy is bucked with a seat belt. In our observational study we will never find identical people driving at identical speeds, in identical vehicles, and hitting identical objects. Any and all of the differences attributable to those non-identical aspects of situations may have a bearing on the amount of damage that the crash victims sustain.