One of the most widely used tools in in modern empirical economic research is the randomized experiment. In this blog post, we seek to help explain the basic idea of randomized control trials (RCTs) to the Everydata reader.
The idea behind this approach is to mimic the medical experiment framework. Most people are familiar with the medical experiment approach intuitively, but the idea is that a group of potential subjects are randomly assigned to one of two groups--a group that would take a treatment (say a new blood pressure medication) and another that would take a placebo.
By comparing the blood pressure of both groups before and after the medication (or placebo) is taken, the researcher can isolate the impact of the pill on blood pressure. The reason "random assignment" is so important is because you want a baseline that can control for any other factors that might also affect blood pressure. In other words, the "randomization" means that the experience of the control (placebo) group mimics what would have happened to the treatment group if they had not taken the medication.
In the medical world, it is easy to administer a treatment, but how does that work in social sciences? In Chapter 5 of our upcoming book Everydata, we introduce the concept of randomization and natural experiments. But, for this blog, I just wanted to highlight some examples from social science research where random assignment has been used to attempt to isolate statistical effects:
Do flip charts in Kenya help improve educational outcomes? (article)
Does offering cash incentives to high schoolers for test performance improve long run outcomes? (article)
Does actively seeking employment discouraged by unemployment insurance? (article)
How does saving behavior change for low-income workers when provided with matching IRA contributions? (article)
Randomization will come up repeatedly in studies you read in the newspaper and hear in the media, so a basic understanding can be important. In future blogs, we'll explore additional topics with randomized control experiments.