The empirical record for unconditional convergence at the aggregate level is mixed. While GDP per capita convergence across countries is elusive (Johnson and Papageorgiou 2019, 2020), certain sectors behave differently. Rodrik (2013) documented unconditional convergence across manufacturing sectors internationally in 1965–2005; Rodrik (2011) extended this argument to stress manufacturing’s role as an ‘escalator’ sector. In a new paper (Klein et al. 2025), we push the insight back to the late 19th century and within a single, highly integrated national economy.
Crucially, our data overcome a major limitation in international comparisons: as Herrendorf et al. (2022) stress, manufacturing data are often not comparable across countries due to sectoral price differences and inconsistent purchasing power parity (PPP) adjustments. In the US, long-standing goods-market integration (Harrison 2023) means our measures of value added per worker are more directly comparable across states and decades, eliminating a key source of bias in cross-country analyses.
By harmonising historical industry classifications (Klein and Crafts 2020, Crafts and Klein 2021), we track productivity at the SIC three-digit level for 14 benchmark years between 1880 and 2007. This allows us to examine convergence across distinct technological regimes – First Industrial Revolution (steam power, textiles), Second Industrial Revolution (electricity, internal combustion), and post-war information and communications technology (ICT) industries.
The speed of convergence
Across the full sample, the annual unconditional convergence rate is 7.6%. In practical terms, the productivity gap between high- and low-performing states halved every nine years. Figure 1 plots the estimated convergence coefficients by period. Rates were stable (~6.5%) before and after WWII but surged in the 1940s – from 5.5% in 1930-40 to 10.6% in 1940-47 and 9.4% in 1947-58 – before declining below 8%. This spike aligns with major structural changes in the South: federal labour market reforms, reduced educational barriers (Caselli and Coleman 2001), and adoption of Northern manufacturing technologies.
Figure 1 Unconditional convergence rates in US manufacturing by decade
Notes: Rates from regressions of productivity growth on initial productivity, with industry-by-time fixed effects. Source: US Census of Manufactures, authors’ calculations.
Convergence across technological eras
Convergence occurred across all sectors, but speeds varied. Table 1 summarises rates for the post-1958 period, when all three technology categories are observed. ICT industries converged fastest (8.3% per year), followed by the Second Industrial Revolution (7.0%) and the First Industrial Revolution (6.3%) industries. This pattern supports the idea that newer general-purpose technologies offer greater catch-up potential when adoption barriers are low, but even mature sectors can exhibit strong convergence under conducive conditions.
Table 1 Convergence rates by technological era, 1958–2007
Notes: Industries grouped following Mowery and Rosenberg (1999).
Interpreting the results
The convergence rates exceed predictions from a closed-economy neoclassical model (Barro et al. 1995) but are below the instantaneous convergence of a frictionless open-economy model. A more plausible mechanism is technology diffusion (Nelson and Phelps 1966), where the speed of adoption depends on both the technology gap and absorptive capacity – itself shaped by education, institutions, and infrastructure. The Southern case highlights the complementarity between market integration and institutional reform. Before WWII, Southern manufacturing lagged substantially. Low human capital and a segmented low-wage labour market (Wright 1986, Connolly 2004) limited technology adoption. The 1930s-40s brought a turning point. As emphasised by Caselli and Coleman (2001), reforms leading to lower barriers to education paved the way for labour to move to manufacturing and operate better technologies from frontier manufacturers in the North. These events accelerated industrialisation in the South and coincided with an increase in manufacturing employment shares in the South and a decline in the North in the following decades. This mirrors concerns in current debates about ‘premature deindustrialisation’ (Rodrik 2015, 2016). Although the South reached peak manufacturing at lower shares than the North, it did so at similar levels of income but decades later.
Lessons for today
Our results suggest three conditions for rapid convergence: integrated markets, accessible frontier technologies, and adequate human capital. These were historically rare and took decades to develop in the South. For present-day economies, the US South’s trajectory offers both hope and caution: structural change requires complementary policies that lift barriers to sectoral mobility. Manufacturing may still play a unique role, but as Crafts and Klein (2021) show for the 20th century US, the geography and composition of industry evolve over time.
References
Barro, R (2012), “Convergence and Modernization Revisited”, NBER Working Paper 18295.
Barro, R and X Sala-i-Martin (1992), “Convergence”, Journal of Political Economy 100(2): 223–251.
Barro, R, N G Mankiw and X Sala-i-Martin (1995), “Capital Mobility in Neoclassical Models of Growth”, American Economic Review 85(1): 103–115.
Caselli, F and W Coleman (2001), “The US Structural Transformation and Regional Convergence”, Journal of Political Economy 109(3): 584–616.
Crafts, N and A Klein (2021), “Spatial concentration of manufacturing industries in the United States: re-examination of long-run trends”, European Review of Economic History 25(2): 223-246.
Connolly, M (2004), “Human Capital and Growth in the Postbellum South”, Journal of Economic History 64(2): 363–399.
Gruss, B and M Celasun (2018), “The declining share of manufacturing jobs”, VoxEU.org, 25 May.
Harrison, J M (2023), “Exploring 200 Years of US Commodity Market Integration”, Explorations in Economic History 88.
Herrendorf, B, R Rogerson and Á Valentinyi (2022), “New Evidence on Sectoral Labor Productivity”, NBER Working Paper 29834.
Johnson, P and C Papageorgiou (2019), “It’s too soon for optimism about convergence”, VoxEU.org, 16 April.
Johnson, P and C Papageorgiou (2020), “What Remains of Cross-Country Convergence?”, Journal of Economic Literature 58(1): 129–175.
Klein, A and N Crafts (2020), “Agglomeration Externalities and Productivity Growth: US Cities, 1880–1930”, Economic History Review 73(1): 209–232.
Klein, A, M León-Ledesma and N Crafts (2025), “A Long-Run Perspective on Unconditional Convergence in Manufacturing”, CEPR Discussion Paper No. 20488.
Mowery, D C and N Rosenberg (1999), Paths of Innovation, Cambridge University Press.
Nelson, R and E Phelps (1966), “Investment in Humans, Technological Diffusion, and Economic Growth”, American Economic Review 56(1): 69–75.
Rodrik, D (2011), “Manufacturing is special”, VoxEU.org, 9 November.
Rodrik, D (2013), “Unconditional Convergence in Manufacturing”, Quarterly Journal of Economics 128(1): 165–204.
Rodrik, D (2015), “Premature deindustrialisation in the developing world”, VoxEU.org, 12 February.
Rodrik, D (2016), “Premature Deindustrialization”, Journal of Economic Growth 21(1): 1-33.
Wright, G (1986), Old South, New South, Basic Books.