Heterogeneous Firms, Homogeneous Ambition: Why Melitz's 2003 Trade Model Still Explains Who Goes Global

The puzzle Melitz set out to solve

Before 2003, the workhorse models of international trade treated firms within an industry as broadly interchangeable. Countries traded because they differed in endowments or technology, and a representative firm stood in for the whole sector. That framework had a stubborn empirical problem: even inside the same narrow industry, in the same country, exporting is rare and concentrated. Most firms never sell abroad at all, and the ones that do are systematically different from the ones that do not.

Marc Melitz built a model in which firms are explicitly heterogeneous in productivity. Each firm draws its own productivity level and then decides, given that level, whether serving a foreign market is worth it. The result reframed the question. Trade is no longer only about why nations exchange goods; it is about which firms inside a nation get to participate, and why the answer is so uneven.

The model in one moving part: a productivity threshold

The engine of Melitz's model is a single, powerful idea. Selling into a foreign market requires paying a fixed cost that domestic sales do not: learning the market, meeting its standards, building distribution, establishing a presence a foreign buyer will trust. Much of that cost is sunk, spent whether or not the venture succeeds.

A fixed cost changes the arithmetic of participation. A more productive firm earns higher operating profit on every unit it sells, so it can cover the fixed cost of entry and still come out ahead. A less productive firm cannot, and so it rationally stays home. This produces a clean cutoff: there is a productivity level above which exporting pays and below which it does not. Firms sort themselves around it. Melitz called this self-selection into export markets, and it is the reason exporters, as a group, look more productive than non-exporters before they ever ship a single order abroad.

Self-selection, not transformation

The direction of causation matters, and it is where the model corrects a common intuition. It is tempting to assume that exporting makes a firm stronger, that going global is a growth tonic. Melitz's mechanism runs the other way: productive firms become exporters because they were already productive enough to clear the threshold. The sorting happens first.

This is a slightly deflating point. It means that opening a border, or opening a new digital channel, does not by itself lift the firms below the cutoff. It rewards the firms already positioned to cross it and can even reallocate demand away from the least productive, who lose ground as trade exposes them to stronger competition. The model predicts winners and losers inside the same industry, not a rising tide that carries every small firm along.

Why small firms export: what the US data shows

The pattern Melitz's model predicts is visible in the plainest US export statistics. Small firms are overwhelmingly the majority of exporters by count and a distinct minority of exports by value.

According to the U.S. Small Business Administration's Office of Advocacy, 97.2 percent of US exporting firms in 2023 were small, some 270,014 of all identified exporters, yet their 588.4 billion dollars in exports amounted to only 33.0 percent of total identified-firm export value. Read through the heterogeneous firms lens, this is exactly what the theory expects. Many small firms clear the threshold to export at all, but they cluster near it, shipping modest volumes, while a small number of large, high-productivity firms account for the bulk of the value. Participation is broad; scale is narrow.

The context underneath those numbers sharpens the point. The SBA counts 34.8 million small businesses, responsible for 45.9 percent of private-sector employment and 43.5 percent of GDP, operating on thin margins: between March 2023 and March 2024 small firms recorded 982,940 closings against roughly 1.1 million openings, a near one-to-one churn. For a firm living that close to the line, the fixed cost of reaching a new market is not an abstraction. It is precisely the barrier the model says it is.

Homogeneous ambition, heterogeneous capacity

The title of this piece is not a rhetorical flourish; it names the asymmetry at the heart of the theory. Ambition is widely and fairly evenly held. Owner-operators across verticals want to grow, to reach beyond their block, their city, their state. What is unevenly held is capacity: the productivity, the operating margin, the organizational slack required to absorb a fixed cost of entry and survive the wait for it to pay off.

Melitz's model is, in this sense, a theory about the gap between wanting to go global and being able to. It does not moralize about the firms below the threshold. It simply observes that the cutoff exists and that firms sort around it whether or not anyone intends them to. Homogeneous ambition meets heterogeneous capacity, and the threshold is where the two are reconciled.

The threshold re-forms at every new layer of borderless commerce

Each wave of globalization has been sold as the wave that finally erases the threshold. Cheaper shipping, then the open internet, then cross-border e-commerce, and now agentic commerce were each expected to let any small firm reach any buyer. The consistent pattern is that the legal and technical door opens before the practical one does, and a new fixed cost of entry re-forms behind it.

The policy layer illustrates the split. The World Trade Organization frames its e-commerce moratorium and its Trade Facilitation Agreement provisions, e-signatures, paperless trading, e-invoicing, as small-business enablers. Yet the WTO's own handbook for policymakers names the binding constraints in the same breath: small firms shoulder disproportionate burdens from non-digital customs procedures, lack familiarity with e-commerce mechanics, and face persistent transport and logistics gaps. The rules are being fixed faster than the operational capability to use them.

The newest layer shows the same shape while it is still forming. OpenAI's Instant Checkout inside ChatGPT opened to all US users in February 2026, promising agent-mediated purchase to any participating merchant; by that same month Forrester's Emily Pfeiffer reported only around thirty Shopify merchants were actually live on it, and OpenAI moved in early March 2026 toward a merchant-controlled checkout model. The distribution surface exists; the practical ability of small merchants to be present and visible inside it does not yet, which is the threshold re-forming in real time.

A counter-thesis worth testing: the Long Tail

Hold the model up against its strongest rival before accepting it. Chris Anderson's Long Tail thesis argues the opposite of a hardening threshold: when distribution and shelf-space costs fall toward zero, niche and small-catalog participants can aggregate into a market as large as the head, and small players win precisely because the cost of reaching a scattered audience collapses.

The two views are not simply reconcilable, and the evidence to date is mixed rather than decisive. Where the fixed cost of entry genuinely approaches zero, the Long Tail's optimism has force. Where a new layer of commerce quietly reintroduces a fixed cost, a machine-legible catalog, a trusted entity, presence inside a curated answer, Melitz's sorting reasserts itself. Treat the democratization claim as a hypothesis to test against each specific surface, not a law to assume. The useful question is never whether small firms can go global in principle, but what the fixed cost of entry actually is on the surface in front of them.

What this means for a small US firm today

The most consequential fixed cost of entry for a small US firm is no longer only shipping or customs. It is machine-legibility: being structured, consistent, and credible enough that the systems now mediating discovery, classic search, the local map pack, generative answer engines, and the agentic checkouts forming behind them, can find you, trust you, and name you. A firm that is invisible to those systems has, in Melitz's terms, failed to clear the threshold, regardless of how good its actual product is.

This is the quietly optimistic corollary of an otherwise sorting theory. A fixed cost of entry is, by definition, payable. The productivity threshold is not a moral verdict on a business; it is a barrier that can be engineered down. The firms that will be visible across borders and across surfaces are not necessarily the largest. They are the ones that treat discoverability as the fixed cost it now is, and pay it deliberately rather than hoping the door was already open.