Because of excess demand for the program, the NUSAF evaluation is being done as a randomized trial. In February we ran a baseline survey of the participants. There are a couple of good reasons to do a baseline:
1. From the programming side, the baseline tells us who we have reached, and so also tells us who we have not. Preliminary results from the baseline data suggest that our recipients are probably not the most vulnerable, which is counter to our basic goals. Without a baseline, we could never know for sure.
2. A baseline tells us something about our recipients today so we have something to compare to in the future. We can then subtract today from tomorrow and find the effect of tomorrow.
3. If there are important differences between the treatment and control groups, we know them and can control for those differences (there don’t appear to be any for the NUSAF evaluation).
So what is the advantage of a randomized trial? If we get so much from the baseline, why do we need a randomized treatment and control?
A control group gives us something to compare. People will undoubtedly improve over the next few years. Without a control, we can’t know if it’s because of the program, or because of general improvements. But of course we could do the same with any control group, and again control for the differences using the baseline.
A randomized treatment and control helps us to ensure very little differences in the unobservables, i.e. all of those things a researcher can’t see, like entrepreneurial ability, eagerness to work, etc. Of course, we can’t ever be sure we captured these things, but a randomized selection is the most likely method to ensure this.
Saturday, October 4, 2008
Subscribe to:
Post Comments (Atom)
No comments:
Post a Comment