Daycare Matching with Siblings: Social Implementation and Welfare Evaluation

Imagine you are a parent trying to find a spot for your children in a popular daycare center. You have two kids, and you really want them to go to the same place. Why? Because if they go to different centers, your life becomes a logistical nightmare. You have to drive to two different locations, manage two different schedules, and deal with two different sets of teachers and rules. It's like trying to juggle two balls while riding a unicycle on a tightrope.

This paper is about a team of researchers who helped a city in Japan (Koriyama City) figure out how to fix a broken system where families with multiple kids were getting left behind.

Here is the story of their discovery, explained simply:

1. The Problem: The "Split" Penalty

Before 2024, the city assigned kids to daycares based on a simple score. If you had two kids, you applied for them separately. The problem? The system didn't care if your two kids ended up at different schools.

The researchers realized that standard math models were missing a huge piece of the puzzle. They assumed parents just cared about distance. But in reality, parents care about togetherness.

The researchers discovered that being forced to send your kids to different daycares is like a massive "tax" on your happiness. They calculated that this "split penalty" feels just as bad as having to drive an extra 4.8 kilometers (about 3 miles) every single day. That's more than double the average commute! It's not just the gas money; it's the stress, the lost time, and the chaos of managing two different worlds.

2. The Experiment: The "Sibling Bonus"

The city wanted to fix this. They asked the researchers: "How much extra 'points' should we give families with siblings so they get assigned to the same place?"

The researchers built a computer simulation—a digital twin of the city's daycare system. They tested thousands of different scenarios:

Scenario A: Give siblings no extra help. (Result: Siblings get split up, parents are miserable).
Scenario B: Give siblings a huge bonus. (Result: Siblings are happy, but families with only one child get pushed out).
Scenario C: The "Goldilocks" Reform (The one they actually implemented in 2024).

3. The Results: A Delicate Balancing Act

The reform they designed was a success, but it came with a catch. Think of it like a see-saw between two goals:

Efficiency (Happiness): Making sure families get what they need.
Equity (Fairness): Making sure no group is treated unfairly.

What happened with the reform?

The Good News: Families with siblings got a huge boost. Their happiness (welfare) went up by 6.4%. The "split penalty" was largely removed.
The Trade-off: To make room for the siblings, some families with only one child got slightly lower chances of getting a spot. The gap between the two groups narrowed, but it didn't disappear completely.

The researchers found a clear rule: If you try to make the system 100% perfect for everyone's happiness, you inevitably make it less fair for some. It's like trying to fill a bucket with water; if you tilt it to fill the front, the back spills over.

4. The Big Lesson: Don't Ignore the "Bundle"

The most important takeaway is about how we do math in the real world.

Imagine you are buying a pizza.

Old Way of Thinking: You estimate how much you like the cheese, then how much you like the pepperoni, and add them up.
New Way of Thinking (This Paper): You realize that a pizza with both cheese and pepperoni is worth way more than the sum of its parts. If you get them separately, it's not the same experience.

The researchers showed that if you use the "Old Way" (ignoring the sibling connection), you completely miss how much parents hate being split up. If you ignore this, you think the reform is only a small improvement. But when you count the "togetherness" factor, the reform is a huge win.

Summary

In short, this paper is about teaching computers to understand that families are teams, not just a collection of individuals.

By giving a little extra help to families with siblings, the city made life much less stressful for them. The researchers proved that while you can't make everyone 100% happy without making someone else slightly less happy, acknowledging the special needs of families with multiple kids leads to a much better, more humane system for everyone.

1. Problem Statement

In centralized assignment mechanisms (e.g., school choice, residency matching, daycare allocation), agents often have preferences over joint assignments rather than individual ones. A prominent example is families with multiple children (siblings) who prefer all children to be assigned to the same facility to minimize logistical burdens (commuting, coordination).

The Gap: Standard preference estimation methods in market design typically model agents as having preferences over individual objects (schools/daycares) independently. They fail to account for demand complementarities (the utility gain/loss from joint vs. split assignments).
The Trade-off: Mechanisms that prioritize joint assignments (e.g., sibling priority) improve efficiency for families with siblings but raise equity concerns for single-child families. Existing literature lacks an empirical framework to quantify the magnitude of the "split assignment" disutility, making it difficult to design optimal priority structures that balance efficiency and equity.

2. Methodology

The authors develop an empirical framework to estimate family preferences while explicitly incorporating sibling complementarities.

A. Data Context

Setting: Koriyama City, Fukushima Prefecture, Japan.
Mechanism: A centralized, priority-based assignment system for licensed daycare centers.
Reform: In 2024, the city implemented a reform introducing sibling priority points to address systematically lower assignment rates for families with multiple children.
Data: Anonymized application data from 2022–2025, including household demographics, rank-ordered lists (ROL), priority scores, and assignment outcomes.

B. Structural Model

The authors extend the standard discrete choice model (following Fack, Grenet, and He) to handle joint assignments.

Utility Function: Family utility is defined over a tuple of assignments $\delta = (d_1, d_2, \dots)$ $δ = (d_{1}, d_{2}, \dots)$ .
$U_{f,\delta} = \sum_{c \in C(f)} u_{c, \delta(c)} - \Gamma_f \cdot \mathbb{I}(\text{split}) - \kappa \cdot d_{f,\delta} + \epsilon_{f,\delta}$
- Flow Utility ( $u_{c,d}$ ): Depends on facility fixed effects, child age, and household covariates (e.g., maternal employment).
- Commuting Cost ( $\kappa \cdot d$ ): Linear penalty based on trip-chain distance (home $\to$ facility 1 $\to$ facility 2).
- Split Assignment Penalty ( $\Gamma_f$ ): A fixed non-distance disutility incurred if siblings are assigned to different facilities. This captures coordination costs (different schedules, supplies, communication methods).
- Stability: The model utilizes the "extended stability" concept from Sun et al. [2024], which allows for preferences over tuples, to define the set of feasible assignments ( $\Delta_f$ ) for likelihood estimation.

C. Estimation Strategy

Likelihood Function: The observed matching is assumed to be stable with respect to true preferences. The authors transform the stability condition into a discrete choice problem where families choose the utility-maximizing tuple from their feasible set.
Maximum Likelihood Estimation (MLE): Parameters ( $\alpha, \beta, \gamma, \kappa$ ) are estimated by maximizing the likelihood of the observed assignments given the priority scores and reported preferences.
Counterfactual Simulation: Using the estimated parameters, the authors simulate assignment outcomes under various priority point schemes (varying weights for simultaneous and incumbent sibling applicants) to evaluate welfare and inequality.

3. Key Contributions

Novel Estimation Framework: First to explicitly estimate demand complementarities in centralized matching markets, moving beyond the assumption of independent individual preferences.
Quantification of Split Disutility: Provides the first empirical evidence quantifying the massive welfare loss associated with splitting siblings, separating it from pure commuting costs.
Policy Evaluation & Design: Offers a rigorous welfare evaluation of a real-world policy reform (Koriyama's 2024 sibling priority), demonstrating how ignoring complementarities leads to significant underestimation of policy benefits.
Efficiency–Equity Frontier: Maps the trade-off between mean social welfare and inequality in assignment rates across sibling groups, providing a quantitative basis for priority design.

4. Key Results

A. Preference Estimates

Split Assignment Penalty: The fixed disutility of splitting siblings ( $\Gamma$ $Γ$ ) is estimated at 3.485 (in utility units).
- Converted to distance equivalents (dividing by the distance coefficient $\kappa \approx 0.72$ ), this penalty is equivalent to an additional 4.8 km of daily commuting.
- This is more than twice the average commuting distance (2.26 km) in the sample.
Model Comparison: A conventional model (ignoring complementarities) underestimates the distance disutility ( $\kappa = 0.67$ vs. $0.72$) and the mean utility of daycare. It incorrectly interprets families choosing distant daycares to avoid splits as having "low distance sensitivity," rather than high split-aversion.

B. Welfare Effects of the 2024 Reform

Welfare Gain: The implemented reform increased overall mean welfare by 6.4%.
Source of Gains: The gains are driven almost entirely by families with siblings (specifically simultaneous applicants), who avoid the massive split-assignment penalty.
- Simultaneous applicants saw a welfare increase of 22.9% (0.565 km equivalent).
- Single-child families saw negligible changes (<1%).
Bias in Conventional Models: A model ignoring complementarities estimates the welfare gain at only 4.0%, substantially understating the reform's success because it fails to capture the avoided split-assignment disutility.

C. Efficiency–Equity Trade-off

The Frontier: There is a clear trade-off. Increasing mean welfare by 100 meters (in distance-equivalent utility) is associated with an increase in assignment-rate inequality (standard deviation across groups) of approximately 1.7 percentage points.
Optimal vs. Implemented Policy:
- The welfare-maximizing policy (theoretical optimum) would assign even higher priority to siblings, increasing total welfare by 17.4% but worsening inequality by displacing more single-child households.
- The implemented reform (160 points) successfully reduced inequality (from 0.111 to 0.099) while achieving a significant portion of the potential welfare gains, striking a balance between efficiency and equity.

5. Significance

Policy Relevance: The paper demonstrates that ignoring sibling complementarities leads to flawed policy analysis. Standard models would suggest the reform is less beneficial than it actually is and might recommend different priority weights.
Market Design: It provides a blueprint for designing priority mechanisms in markets with interdependent agents (couples, siblings, teams). It shows that "fairness" (treating all children equally) can be inefficient if it ignores the high coordination costs of splitting families.
Generalizability: The methodology is applicable to other centralized matching markets (e.g., medical residency for couples, university housing for roommates) where joint preferences are critical but often unmodeled.

In conclusion, the paper establishes that the "cost" of splitting siblings is a dominant factor in daycare choice, far exceeding simple commuting costs. Incorporating this into structural estimation is essential for accurately evaluating the welfare impacts of priority reforms and designing mechanisms that balance efficiency with equitable access.