Valence-Romans' CNAF “Optimal Fair Matching” Algorithm for Daycare Place Allocation

Public daycare (crèche) places are scarce in France — nationally there are fewer than two places for every ten children under three — and allocation has traditionally relied on a municipal committee scoring applications against criteria such as parental employment status, family location and specific circumstances like illness or professional-integration needs. Researchers from CNAF (France's national family-welfare agency), working through its ISAJE programme studying the impact of daycare access on child development, partnered with the Valence Romans Agglo municipality (Valence, roughly 100km south of Lyon) to convert these existing criteria into a computerised “student-optimal fair matching” algorithm — a variant of the Gale-Shapley deferred-acceptance mechanism.

The explicit aim was to reduce discretion and cronyism in a process long suspected of favouring well-connected families, guaranteeing that no lower-priority family could take a place from a higher-priority family who preferred it.

Because the pandemic prevented the human allocation committee from meeting in person, the May 2020 round was run entirely by the algorithm for the first time in France, satisfying 400 of 1,014 applications (39.4%) submitted for that cycle. The algorithm's proposed distribution was then reviewed and formally validated by the agglomeration's allocation commission before places were confirmed.

Coverage of the pilot notes it addressed only one allocation round in one municipality, with no published before/after comparison against the discretionary system it replaced, and some local officials elsewhere in France voiced concern about losing decision-making discretion — one Paris official was quoted asking, “if computers sort out the demands, what are we going to do?”

The case is a modest but concrete example of applying matching-market theory to a genuine administrative bottleneck in early-childhood education access, with real applicant numbers and an explicit anti-cronyism rationale, though its evidence base remains a single pilot round rather than a rigorous evaluation.

Read the full analysis: https://algorithmwatch.org/en/algorithms-to-fight-cronyism-in-french-daycare/