Randomization is the process used to assign the potential TUP beneficiaries to one of two groups: the group receiving the program or the group that does not receive the program. The development terminology for these groups is treatment and control. Randomization can be done on the computer using Excel to generate random number assignment or it can be done through a public lottery.
Why randomizations? Assigning beneficiaries to treatment and control accomplishes several goals. First, it creates two statistically identical groups. Since choosing randomly is inherently uncorrelated with any characteristic in the sample, each characteristic has the same probably of appearing in the treatment group and the control group. Second, having statistically identical treatment and control groups allows intrepid researchers to answer the question "What would have happened in the absence of the intervention?" This question is called the counterfactual. Once again employing the powers of Microsoft paint, the graph below visually demonstrates this concept:
The red line left of the vertical line is the entire sample prior to the intervention. The gray line is the start of the intervention and the division of the sample into treatment and control groups. Since randomization created identical treatment and control groups, the only difference between the groups is one receives the intervention and the other does not. Thus, any difference in outcome between the groups can be attributed to the program. This is essentially the concept used in medical research applied to answering social science questions.
Looking the graph, we see in this case that the treatment group income increased at a higher rate than the control group income after the intervention. Again, because the randomization created identical treatment and control groups, we can say with confidence that this change is attributable to the intervention. The green control group has served to answer the counterfactual.
"But this is not fair!" you cry. "Some people get the program while others don't!"
Not so. Randomization is actually the most fair method to decide who receives the intervention. It gives each beneficiary an equal chance of getting the intervention. Further, as is the case with most development programs, there is often not enough resources to serve everyone so randomization allows for meaningful impact evaluation and fair assignment to treatment. Finally, what if the program actually doesn't help incomes but hurts them? While programs hopefully do not have that outcome, randomization again has given each participant an equal chance at receiving or not receiving the treatment of ambiguous impact.
Randomized Control Trials, or RCTs, is the methodology of choice for researchers and aid agencies wanting to maximize the use of resources.
Read more at:
www.poverty-action.org
www.povertyactionlab.org
Or watch:
http://www.ted.com/talks/lang/eng/esther_duflo_social_experiments_to_fight_poverty.html
No comments:
Post a Comment