Surveys are possibly the most commonly undertaken form of research. They are rightly popular because they are extremely flexible and can be constructed and modified to answer a wide variety of research questions.
They are good at showing 'association' but not 'causality'. For example, a survey could be undertaken of nursing students asking them about their use of cannabis and their mood. It has been suggested that recent forms of high-strength cannabis, such as 'skunk' may precipitate mental health changes.
It might be found that students who smoke cannabis have a lower mood state than those who do not, but this does not prove that cannabis caused this. It might instead indicate that students with a low mood state are likely to seek out cannabis for relief. The association between the two may be shown, but not proof that one causes the other.
If cause and effect must be demonstrated, an experimental structure is required. However, in this example, no ethical committee would agree to a research project where large groups of nursing students were given cannabis and told to smoke it. In this sort of situation, a survey may be the only suitable research design available.