The Lewisian Model of Unequal Exchange

The catastrophic failure of the law of one price

The 2024 Nature Communications paper by Hickel, Hanbury Lemos, and Barbour presents a stark empirical finding: the contemporary world economy is organized around a massive, persistent net appropriation of labor from the Global South to the Global North through unequal exchange. The scale is staggering. In 2021, the South’s net transfer of embodied labor—valued at Northern wage rates—was equivalent to €16.9 trillion, or about 826 billion hours of work. Valued as total value added (wages plus profits), the transfer rises to roughly €30.5 trillion, a figure on par with the GDP of the entire Global North. These transfers occur despite decades of industrialization in the South, and they persist even when comparing like for like: within the same sectors and skill categories.

The underlying reason is very simple. Market wage rates in the Global South are a small fraction of those in the Global North. Southern wages are a tenth of Northern wages for the same work.

We find Southern wages are 87–95% lower than Northern wages for work of equal skill. While Southern workers contribute 90% of the labour that powers the world economy, they receive only 21% of global income.

Hickel et al. (2024). Emphasis mine.

The disparity is especially pronounced for low-skill work. But it extends across the skill spectrum.

Southern wages are 87% less for high-skill labour, 93% less for medium-skill labour, and 95% less for low-skill labour. The disparity is so extreme that high-skill labour in the global South receives 68% less than low-skill labour in the global North. Stated otherwise, for every hour of work at a given skill level, Northern workers are able to consume 8–19x more of the global product than Southern workers.

Hickel et al. (2024).

In an integrated global economy under competitive conditions, the law of one price would ensure that equivalent labor commands equivalent remuneration, adjusting for productivity and transaction costs. The data demonstrate the opposite: a catastrophic failure of the law of one price for labor. Mechanically, this is due to the fact that labor markets are contained within national silos by low rates of migration, further enforced by a global regime of border policing to keep Southern migrants out of the Global North.

Yet, despite the low rates of migration, how could it be that exchange rates and terms of trade don’t adjust to equalize the price of equivalent work across countries over time? Is this not what globalization is supposed to do?

Many economists believe that, because catch-up growth is faster, we should expect conditional convergence. Why don’t we see convergence then? How can a wage gap of 10X persist decade after decade? Contrary to theoretical expectations, we see not convergence to the law of one price, but massive divergence.

The average Southern wage has increased from €0.46 to €1.62 per hour (an increase of €1.16), while the average Northern wage has increased from €12.60 to €24.95 per hour (an increase of €12.35). Northern wages have increased 11x more than Southern wages. There is no “catch-up” occurring; on the contrary, it is a pattern of dramatic divergence.

Hickel et al. (2024).

W. Arthur Lewis offered an interesting explanation, that the authors report in the supplement. Lewis argued that in economies with a large subsistence sector, wages in the modern sector are not determined by productivity in that sector, but rather by the wage floor set in the subsistence economy. As long as surplus labor remains, productivity improvements in the modern sector will lower the prices of exports rather than raise domestic wages.

Lewis argued that national wages are set in the “subsistence sector” rather than in the export sector, so improvements in productivity in the latter do not result in proportionately higher wages. Therefore, the benefits are accumulated by the Northern importers, who are able to buy more of the product for the same price.

Hickel et al. (2024). Supplementary Information.

This Lewisian mechanism, if correct, would explain both the persistence of low wages for highly productive Southern labor and the structural nature of unequal exchange. Our aim here is to test the predictions of the Lewisian model. Does it provide an efficient explanation of this pattern at the largest of scales?

If the Lewisian explanation for the pattern documented in the Nature paper is correct, we should expect subsistence wages to predict wages in other sectors, even after controlling for productivity. In what follows, we test this hypothesis.

---

Methods

We construct a harmonized panel dataset of earnings and productivity data from the ILO. The dependent variable in all specifications is the log of sectoral wages in US dollars at market exchange rates. The key explanatory variable is the log of agricultural wages, interpreted as the subsistence wage floor in the Lewisian sense. Sectoral productivity is also expressed in US dollars per worker at market exchange rates.

In all-sectors specifications, productivity is split into tradables and non-tradables by interacting productivity with a tradability indicator. The idea is to test whether the same pattern holds for tradable and non-tradable sector—the theoretical expectation being a stronger relationship with productivity in tradable sectors.

We run four main specifications. The first two are in levels: (i) all sectors, with separate productivity terms for tradables and non-tradables, and (ii) industry only, with a single productivity term. The second two are in first differences: (iii) all sectors, and (iv) industry only. The levels specifications address the substantive Lewisian question directly, since the model speaks to equilibrium wage levels; the first-difference specifications are robustness checks that control more aggressively for common trends, at the cost of discarding low-frequency information.

We use PanelOLS for estimation with two-way fixed effects—entity (country×sector) and year for all-sectors specifications—and country and year for industry-only specifications. Standard errors are clustered by country. To ensure comparability across variables, all regressors are standardized (z-scored) within the estimation sample. The response is not standardized, so that the coefficient estimate correspond to semi-elasticities: b = 1.0 means that +1 standard deviation higher value of the regressor implies 100% higher wages.

---

Results

The updated estimates confirm the robustness of the Lewisian mechanism across all four specifications. In the levels models, the coefficient on the subsistence wage floor is large—with semi-elasticities close to 0.78—and highly statistically significant. This means that a one-standard-deviation increase in the subsistence wage is associated with 78% higher sectoral wages, holding productivity and fixed effects constant. The within R² is 0.76 in the all-sectors model and an exceptionally high 0.91 in the industry.

The first-difference models yield smaller coefficients, as expected when focusing on year-to-year changes, but the effect remains both statistically and economically significant: 0.48 in the all-sectors model and 0.50 in the industry-only model, with within R² values of 0.31 and 0.57 respectively.

The levels models yield large, precisely estimated coefficients, consistent with the long-run structural role of the subsistence wage floor, as in the Lewisian theory. The first-difference models confirm that the relationship holds dynamically, albeit at smaller magnitudes, reflecting short-run adjustments. The consistency of the estimates across specification changes strongly supports the Lewisian hypothesis.

Remarkably, we find that in all specifications, the subsistence wage is a stronger conditioner of wages in other sectors than sectoral productivity. We do not find a systematic difference between tradable and non-tradable sectors, which is surprising and may reflect the coarse nature of our dataset.

The upshot is that the Lewisian explanation of the persistent gap in wage levels between the Global North and the Global South is consistent with the empirical evidence.

---

Discussion

Consistent with the Lewisian model, the subsistence wage floor—here proxied by agricultural wages—exerts a powerful influence over the wage structure as a whole. Improvements in the subsistence sector translate into higher wages in other sectors, while productivity gains in modern sectors, absent such improvements, fail to raise wages and instead manifest as lower export prices, to the benefit of the Global North.

This finding reframes the birth lottery in structural terms: which country one is born in determines not just initial endowments and institutions, but the wage floor against which all other earnings are set. For two workers of identical skill and productivity, one in the Global North and one in the Global South, the difference in lifetime earnings is overwhelmingly determined by this large-scale structure of the global economy.

The persistence of such disparities represents not just a violation of the law of one price, but a systematic inversion of its logic. In a genuinely integrated market, high productivity in the Global South would lead to wage convergence with the Global North. In our siloed world, high productivity in the Global South fails to lead to convergence in wages because of the downward pull exerted by the existence of surplus labor and barriers to migration. Our empirical findings, consistent across model specifications, suggest that unequal exchange—and therefore global polarization—will persist until most of the surplus population of the Global South is absorbed into more productive work.