Voucher Program Long-Run Outcomes

What the evidence says

The evidence on this page covers three different voucher programs: experimental and Section 8 housing vouchers, free or subsidized health-product vouchers, and simulated school vouchers. The housing evidence comes from randomized trials. The school evidence comes from a model, not from observed outcomes. Their results do not transfer across programs.

Housing vouchers show the clearest long-run result, and it depends on the child's age at the move. Children whose families took up an experimental voucher to move to a lower-poverty area before age 13 have annual income $3,477 (31%) higher in their mid-twenties, against a control mean of $11,270. Their college attendance between ages 18 and 20 rises by 2.5 percentage points (16%) from a control mean of 16.5%. As adults they also live in lower-poverty neighborhoods, and young females are less likely to be single parents. The sources do not give a size for those last two results. The source ties the age pattern to duration of exposure to a better environment during childhood. Gains fall as the age at the move rises.

Older children and adults do not show the same gains. For children who moved at ages 13 to 18, the experimental-voucher estimate is -$967 (intention to treat) and about -$2,000 (treatment on the treated). Neither estimate can be told apart from zero. For adults whose families moved, the treatment-on-the-treated estimate for individual earnings is -$734 (4.7%), and the 95% confidence interval includes a 12% increase. The sources report little or no long-term economic gain for adults.

Targeting appears to matter. A Section 8 voucher, which carries no low-poverty requirement, raises individual income by $1,723 for children under 13. That is 50% of the experimental-voucher estimate of $3,477. The sources do not say whether the gap comes from smaller neighborhood-poverty reductions or from the counseling and choice restrictions in the experimental design.

For health-product vouchers in Dupas (2014), a free or highly subsidized net in year one raises take-up a year later. Take-up in the "1st LLIN Free" group is 6.1 percentage points (41%) higher than in the non-free groups. The high-subsidy group in phase 1 also redeems more in phase 2, but that result holds only at the 10 percent level. Households that used the first net show a larger effect: an instrumental-variable estimate of +13 percentage points, about a 90% increase on a 15% mean in the non-free group. The source reads this as learning by doing. A direct test finds no anchoring at zero price among households that received a free net, and the learning effect dominates any anchoring effect. This is the only evidence here on whether a subsidy lowers later willingness to pay. It covers one product in one setting.

The school-voucher results come from a simulation (Ferreyra 2007), so they show what the model implies, not what happened. In the model, universal vouchers raise private school enrollment, and most households gain school quality at vouchers of at least $3,000. Low-income households gain the most in welfare. Some voucher users move to neighborhoods with lower tax-inclusive housing prices and send their children to private schools, which weakens residential stratification. Restricting vouchers to non-sectarian schools lowers Catholic school enrollment as the voucher rises. Households that prefer Catholic schools benefit most from universal vouchers and lose most from non-sectarian ones. The sources give no magnitudes for most of these results. No claim on this page measures long-run academic or earnings outcomes for school vouchers.

Where the evidence splits

The housing-voucher results split by age at the move. Children under 13 gain in adult earnings and college attendance. Children aged 13 to 18 show estimates at or below zero, and adults show no gain. The source describes this as consistent with exposure duration being a key determinant of long-term outcomes. The sources on this page do not separate that explanation from adolescent disruption costs, so the cause of the split is unexplained here.

Thresholds and dose-response

• Housing vouchers: the sources group children as under 13 and over 13 at the move. Gains are large and positive in the under-13 group, and the point estimates are at or below zero in the 13-to-18 group. The sources say the gains fall with age at the move. They do not establish whether a critical-period cutoff exists.
• School vouchers (simulation): most households gain school quality at vouchers of at least $3,000. Households with about $35,000 in wealth gain the most quality in both modeled programs.
• Health-product vouchers: the "high-subsidy" group in phase 1 is defined as a price of 50 Ksh or less, and it shows higher phase 2 redemption at the 10 percent level.

Open empirical questions

• Do the earnings gains for children who moved before age 13 persist past the mid-twenties, and do they reflect sustained labor market success?
• Is the age profile of housing-voucher gains linear, or does a critical-period cutoff exist? Does adolescent disruption explain the older-child estimates?
• Do observed school-voucher programs produce the enrollment, school-quality and stratification changes the simulation implies, and what are the long-run academic or earnings outcomes?
• Does the learning-by-doing effect on later take-up appear for products other than the one studied, and over longer than one year?