Trying to estimate the returns to education has been one of the greatest challenges for education economists in the previous years. Common sense would tell you that people with more education do indeed earn more, on average, than people with less education. However, for the purpose we are aiming, this fact is… irrelevant. For instance, if individuals with higher earning capacity (or ability) choose to obtain more schooling, then we would naturally expect a positive correlation between education and wages, but this would not imply necessarily that there is a causal relation between the two. Thus, in order to disentangle these effects, the question we should be posing is: do people with more education earn more, on average, than they would, had they acquired less schooling? And this creates obvious problems, as we cannot have the same individual with two different levels of education at the same time.
A possible approach to overcome this ability bias is by using natural experiments, i.e., finding a way in which two seemingly identical people, because of some quirk of nature or fate, end up getting different levels of education. 
Perhaps the most famous experiments of such kind are those with twins that, by hazard, ended up following different educational paths. A less known, but equally fascinating one, is the Vietnam draft lottery, carried out by the U.S. government, in 1969, to establish the priority of call to military service in the Vietnam War.
Back then, the 366 days of the year were written on slips of paper, which were placed in separate blue plastic capsules that were mixed and then dumped into a glass jar. Capsules were then drawn from the jar, one at a time, and the order of exit would indicate the order of draft. Men with low lottery numbers were the first to be drafted, whereas men with high numbers were almost sure to escape it.
There was, however, a way to be exempted from military service, even if you had a low number, which was to remain in school and obtain an educational deferment. Thus, the draft lottery encouraged men with low lottery numbers to get a few more years of education to avoid military service, whereas men with high numbers, who were already safe from war and didn’t have to go to college, actually chose not to go. The resulting wage gap between workers with low lottery numbers and high schooling levels and workers with high lottery numbers and low schooling levels can therefore be seen as a measure of the true rate of return to schooling, under the reasonable assumption that there are no ability differences between the two groups.
By controlling for ability bias in this way, economists have concluded that an extra year of schooling is associated with a 6.6% increase in weekly earnings, which is perhaps one of the best estimates of the returns to education that we have, but still far from perfect or unbiased.
In fact, it can be claimed that even though this approach has controlled for the ability bias, it may still not represent the true return to education. On one hand, the higher wage that these individuals ended up earning is not necessarily synonym of a higher utility. It is possible, for instance, that these men, who had chose not to go to college in other situation, had to suffer a lot to remain at school and it is at least reasonable to assume that they could have been happier living a life where they earned less money. On the other hand, higher education may have had other benefits not reflected in the higher wages and thus not accounted for, as a better health or more stable jobs.
This simple experiment shows us that, no matter how fascinating or well constructed an approach seems to be, until science develops to a point where cloning is possible – which may never happen -, the true returns to education may never be known for sure.
Ana Lemos Gomes
 Steven Levitt.
 The first number drawn was 258 (September 14), so all registrants with that birthday were assigned lottery number 1. The second number drawn corresponded to April 24, and so forth.
 That is no reason to believe that men born on the days assigned a low lottery number have a different average ability than men born on days assigned high lottery numbers, given the random nature of the lottery.