27.03.2018

This can obviously have serious consequences for the reliability of the estimates obtained from RCTs, and it is not unusual for studies not to describe their randomisation process accurately, or in much depth.

Furthermore, many RCTs do not address the issue of attrition appropriately. Attrition refers to individuals that have been assigned to either treatment or control groups but have then decided not to proceed with the experiment. It is often not clear why those individuals drop out, and this behaviour can have adverse effects on the results of the experiment. It is frequently argued that individuals dropping out would have been worse off than the ones remaining and hence a risk of overstating impact estimates exists, but the opposite can also be true. Drop-outs change the composition of treatment and control groups thereby influencing the results of the experiment since their outcomes cannot be observed. It is possible to track the individuals that drop out, and thereby one can address any side effects of attrition, but this is a costly undertaking.