Imagine a sprint. We decide to study the winners. We note the winners have shoes and the losers don't. Imagine there is perfect correlation. We conclude that if we give everyone shoes then everyone can win. Replace shoes with credentials.
Evan actually captures two problems here.
The first is correlation as compared to causation, the second is the importance of relative impacts.
Obviously many factors come into play in athletic performance. However, a focus on one apparent important statistical connection can act to conceal others. Further, the giving of shoes to all may improve the performance of those without shoes as compared to those with shoes, but does not affect overall relativities since these are primarily based on athletic ability, not shoes.
Assume now that the Government cannot afford to give shoes to all, or at least not comparable shoes. Some people get cheap shoes, some good shoes, some no shoes at all. In these circumstances, the National Shoe Program will create new relativities; some people (those who don't have shoes) may actually be worse off.
The statistical averages used to evaluate the National Shoe Program may show an overall increase in athletic performance since those previously without shoes do a bit better and pull the statistical average up. The Government may then decide to extend the Program. Further, since inequities are now clear, special programs are introduced to address them.
The Government now faces a difficulty in that the national benchmarks used to evaluate athletic performance show no improvement in performance. In fact, the country's performance appears to be falling behind. New steps need to be taken.
A national shoe design committee is appointed to set national standards for running shoes. New quality control and reporting procedures are mandated for all shoe suppliers. And so it goes on!