📈 economics

The Matching Function: A Unified Look into the Black Box

This paper utilizes network theory to unify various matching function forms, demonstrating that match efficacy is primarily determined by the distribution of search intensities, where greater inequality in these intensities negatively impacts the matching process.

Original authors: Georgios Angelis, Yann Bramoullé

Published 2026-05-12

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: Georgios Angelis, Yann Bramoullé

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine the job market not as a giant, invisible machine, but as a massive, chaotic game of "connect the dots."

For decades, economists have used a tool called the Matching Function to predict how many unemployed people will find jobs and how many open positions will get filled. Think of this function as a "black box." You put numbers in (like the number of job seekers and the number of open jobs), and a number pops out (the number of hires). But for forty years, no one really knew how the box worked inside. They didn't know what was happening with the connections between the people looking for work and the companies looking to hire.

This paper, "The Matching Function: A Unified Look into the Black Box," opens that box up. The authors, Georgios Angelis and Yann Bramoullé, use network theory—the math of how things are connected—to explain the magic inside the box.

Here is the story of their findings, explained simply.

1. The Game of "Who Knows Whom"

Imagine a giant room with two groups: Applicants (people looking for jobs) and Vacancies (open jobs).

In the old view, everyone was assumed to be the same. Everyone applied to the same number of jobs, and every job got the same number of applications. It was like a perfectly organized dance where everyone had the same partner.

The authors say: "No, that's not how it works." In reality, the connections are random and messy.

Some people are "super-connectors." They know a lot of people, have great skills, or are really good at finding job ads. They apply to many jobs.
Others are "shy connectors." They only apply to one or two jobs.
Some jobs are "popular magnets" (advertised everywhere), while others are "hidden gems" (only known to a few).

The authors model this as a network. A "link" is formed when an applicant applies to a job. The paper asks: How does the shape of this web of connections change the number of people who actually get hired?

2. The "Black Box" Revealed

The authors show that this messy, random network actually explains almost every famous formula economists have used for the last 40 years.

Think of it like a Swiss Army Knife. The network model is the main tool, and all the different "specialized tools" (the old formulas like CES, Urn-Ball, etc.) are just specific ways of holding the handle.

If everyone applies to every job, you get one specific formula.
If everyone applies randomly but equally, you get another.
If people apply with different intensities, you get a new, more accurate formula.

The paper proves that if you look at the network closely enough, you can derive all these old formulas as special cases. It unifies the whole field under one roof.

3. The "Inequality" Problem (The Key Finding)

This is the most surprising part of the paper. The authors discovered that inequality in how hard people look for jobs is bad for the whole system.

Imagine a relay race.

Scenario A: Everyone runs at a steady, medium pace. The team finishes efficiently.
Scenario B: One person runs at a super-fast sprint, while everyone else walks slowly.

The authors found that Scenario B is worse for the team.

In the job market, if some people are "super-searchers" (applying to 50 jobs) and others are "lazy-searchers" (applying to 1 job), the system becomes less efficient.

The "super-searchers" end up competing with each other for the same popular jobs, creating traffic jams.
The "lazy-searchers" miss out on jobs that the super-searchers didn't even see.
Result: Fewer total matches happen. The "match efficacy" (how well the system works) drops.

The Takeaway: It's not just about how many people are looking or how many jobs are open. It's about how evenly the effort is spread. If the effort is too uneven (high inequality in search intensity), the market works worse.

4. The "Inverted-U" Surprise

Usually, economists think: "If people try harder (search more), more people get jobs." It's a straight line up.

The authors found a twist. If the "harder searching" comes with more inequality, the line can actually go down.

Imagine a party where everyone is trying to talk to the host.

If everyone tries a little bit, the host talks to everyone.
If everyone tries a lot, but a few people are screaming and hogging the host's attention while others are ignored, the host gets overwhelmed and misses some people.

The paper shows that if the average effort goes up, but the gap between the "super-searchers" and the "lazy-searchers" gets wider, the total number of jobs filled can actually decrease. This creates a "hump" shape: effort helps at first, but too much inequality kills the efficiency.

5. What About "Locations"?

The authors also looked at what happens if jobs are grouped into "locations" (like cities or industries).

If all jobs are in one big city, the network is very connected.
If jobs are scattered across many small towns, the network is fragmented.

They found that inequality is bad here too. If some towns are packed with jobs and others are empty, and people don't move between them, the matching process suffers. It's better if the "density" of jobs is spread out more evenly.

Summary

This paper takes the "black box" of the job market and replaces it with a clear, transparent model based on networks.

The Old View: The market is a simple machine where more effort always equals more hires.
The New View: The market is a complex web.
- Uniformity is good: When everyone searches with similar intensity, the system works best.
- Inequality is bad: When a few people do all the searching and others do none, the system clogs up, and fewer people get hired.
- More isn't always better: If the "more searching" comes with more inequality, the market can actually get worse.

The authors didn't just guess this; they used math to prove that this "network" view explains all the old formulas and provides a new, testable way to understand why some economies match workers to jobs better than others.

Technical Summary: The Matching Function: A Unified Look into the Black Box

Problem and Motivation
The matching function serves as the linchpin for models incorporating search frictions across labor, credit, goods, and international trade markets. Despite its ubiquity, the function has remained a "black box" for nearly forty years, treated as a reduced-form object for tractability without a systematic analysis of the underlying frictions (information limitations, coordination failures) that generate it. Consequently, the relationship between the micro-structure of search frictions and the macro-properties of the matching function remains poorly understood. This paper aims to open this black box by tracing the properties of the matching function to the granular structure of connections between applicants and vacancies.

Methodology
The authors propose a network-based model that links micro-level search behavior to macro-level matching outcomes. The primitive of the environment is a bipartite random network connecting a set of applicants ( $U$ ) and a set of vacancies ( $V$ ).

Network Structure: Links $ij$ are independent Bernoulli random variables with formation probabilities $p_{ij}$ . The matrix $(p_{ij})$ captures the granular search frictions (informational, geographic, skill-mismatch) implicitly assumed in the literature.
Protocol: An applicant applies to a vacancy with probability $p_{ij}$ . If a vacancy receives multiple applications, the job is offered to the most qualified applicant, where qualifications are drawn independently from a common distribution. This implies a vacancy offers a job to any specific applicant with probability $1/d_j$ , where $d_j$ is the number of applicants.
Search Intensity: Heterogeneity is introduced via search intensities. In the benchmark case, link-formation probabilities are applicant-specific ( $p_{ij} = p_i$ ), interpreted as the mean number of links an applicant forms. The analysis focuses on the large market limit ( $U, V \to \infty$ ), where degrees converge to Poisson distributions.
Mathematical Tools: The analysis utilizes tools from network theory, specifically the properties of Poisson binomial and mixed Poisson distributions, and moment-generating functions (MGF). The job-finding probability is derived as a closed-form expression involving the MGF of the search intensity distribution.

Key Contributions and Results

1. Unification of Matching Technologies
The paper derives a general, closed-form expression for the mean job-finding probability $f(\theta; G)$ , where $\theta$ is market tightness and $G$ is the distribution of applicant search intensities. This single framework recovers multiple existing functional forms as special cases:

Urn-Ball: Recovered when search intensities are degenerate (all applicants apply everywhere) or in the dense network limit.
Kiyotaki-Wright (1993) & Lagos (2000): Recovered under specific distributions (e.g., exponential) and limits.
Shimer (2007) & Mortensen (2009): Recovered when vacancies are aggregated into "locations" with specific distributions.
CES (Constant Elasticity of Substitution): The paper provides a condition under which the matching process in any network from the analyzed class can be approximated "as if" it comes from a CES function (up to a first-order approximation). This condition relates mean search intensity to market tightness via a specific scaling equation. Unlike previous derivations relying on specific queuing primitives, this result applies to a broad family of networks and is directly testable using application data.

2. Theory of Match Efficacy and Search Inequality
The paper establishes a theory of match efficacy, identifying inequality in search intensities as a key determinant of matching efficiency.

Dispersion is Detrimental: A robust finding is that dispersion in search intensities (modeled as a mean-preserving spread or second-order stochastic dominance shift in the distribution $G$ ) unambiguously reduces the matching probability. This holds for heterogeneity on the applicant side, the vacancy side (advertising intensity), and when vacancies are aggregated into locations (market tightness dispersion).
Ambiguity of Mean Search Intensity: While standard matching functions assume monotonicity in mean search intensity, this paper shows the effect is generally ambiguous. If an increase in mean search intensity is associated with a rise in the Gini coefficient (increased inequality), the job-finding probability can exhibit an inverted-U shape. This result is novel for macro matching functions, which typically assume monotonic increases.

3. Aggregation and "Locations"
The analysis extends to scenarios where vacancies are aggregated into "locations" (e.g., firms or regions), reducing coordination frictions within locations but maintaining them across them. The results on the negative impact of dispersion generalize to this setting, confirming that dispersion in the distribution of vacancies across locations (and thus in local market tightness) negatively affects match efficacy.

Significance and Claims
The paper claims to provide the first systematic study of how the structure of underlying search frictions affects the emergent matching function's functional form and efficacy. By treating the applicant-vacancy network as an intermediate, granular object, the authors:

Unify Disparate Literature: They demonstrate that seemingly disparate matching technologies (Kiyotaki-Wright, urn-ball, CES, etc.) are special cases of a single network-based mechanism.
Provide Testable Conditions: The condition for the CES approximation is derived as a testable relationship between observable network statistics (link probabilities) and market tightness.
Refine Theoretical Foundations: The paper offers sharper theoretical underpinnings for match efficacy, moving beyond the standard assumption that only the mean search intensity matters. It highlights that the distribution of search efforts is critical, with inequality acting as a drag on the matching process.
Contextual Generality: While framed within the labor market, the authors note the analysis is not context-specific and applies to any market where search frictions play a role (e.g., credit, goods, assets).

The authors remain modest regarding the scope, acknowledging that their setup abstracts from dynamic offer/rejection processes, vertical heterogeneity in applicant quality, and network clustering (transitivity). They posit that their random network framework serves as a minimal, novel, and useful formalization of "search" to uncover the implications of search patterns for the matching function, rather than capturing every dimension of reality.