FEW ideas in education are more controversial than vouchers——letting parents choose to educate their children wherever they wish at the taxpayer's expense. First suggested by Milton Friedman, an economist, in 1955, the principle is compelling simple. The state pays; parents choose; schools compete; standards rise; everybody gains.
Simple, perhaps, but it has aroused predictable——and often fatal——opposition from the educational establishment. Letting parents choose where to educate their children is a silly idea; professionals know best. Cooperation, not competition, is the way to improve education for all. Vouchers would increase inequality because children who are hardest to teach would be left behind.
But these arguments are now succumbing to sheer weight of evidence. Voucher schemes are running in several different countries without ill-effects for social cohesion; those that use a lottery to hand out vouchers offer proof that recipients get a better education than those that do not.
Harry Patrinos, an education economist at the World Bank, cites a Colombian program to broaden access to secondary schooling, known as PACES, a 1990s initiative that provided over 125,000 poor children with vouchers worth around half the cost of private secondary school. Crucially, there were more applicants than vouchers. The programme, which selected children by lottery, provided researchers with an almost perfect experiment, akin to the pill-placebo studies used to judge the efficacy of new medicines. The subsequent results show that the children who received vouchers were 15—20% more likely to finish secondary education, five percentage points less likely to repeat a grade, scorced a bit better on scholastic tests and were much more likely to take college entrance exams.
Vouchers programmes in several American states have been run along similar lines. Greg Forster, a statistician at the Friedman Foundation, a charity advocating universal vouchers, says there have been eight similar studies in America： seven showed statistically significant positive results but was not designed well enough to count.