Should governments invest heavily in place-based industrial policies such as the EU’s Smart Specialisation Strategy? A growing body of research documents the positive economic effects of large-scale industrialisation programmes in the US, Finland, the Soviet Union, and China (Fan and Zou 2021, Schweiger et al. 2022, Mitrunen 2025, Garin and Rothbaum 2025). Through agglomeration economies and increasing returns to scale, locations targeted by place-based industrial policies are typically found to sustain higher levels of economic activity in the short and medium run. A particularly influential contribution, Greenstone et al. (2010), exploits the allocation of “Million-Dollar Plants” across candidate US counties and identifies positive spillovers generated by such industrialisation programmes on host local economies.

In a new paper (Heblich et al. 2026), we revisit this question through a unique historical experiment: an unprecedented industrial policy that involved the most comprehensive technology transfer in modern industrial history and laid the foundations of China’s industrialisation (see also Giorcelli and Li 2022a, 2022b). Under the Sino-Soviet Treaty of Friendship, Alliance and Mutual Assistance, the USSR assisted China in constructing 149 large industrial facilities – so-called “Million-Rouble Plants” – in the 1950s. We track the locations that benefited from these initial investments and study the dynamics of the industrial clusters that emerged around them. We provide causal evidence that these investments initially generated positive spillovers, creating a boom in employment and triggering structural transformation – consistent with the existing literature, which focuses on short- and medium-run effects. Yet, although the Million-Rouble Plants themselves continued to thrive in the long run, they eventually generated substantial negative production spillovers, leading to a bust. This decline is driven by the performance of other local establishments, which exhibit low productivity, limited innovation, and high markups. As the industrial clusters matured, local economies became increasingly specialised (Henderson et al. 1995). Such specialisation appears to have weakened knowledge spillovers and constrained the emergence of new clusters and local entrepreneurship, consistent with the mechanisms discussed by Duranton and Puga (2001) and Faggio et al. (2017).

The Million-Rouble Plants: A place-based industrial policy from another era

Our study builds on a fascinating historical experiment that is little known outside China but occupies a prominent place in Chinese economic history under the name of the “156 Program”. In 1949, both the world and China were undergoing profound geopolitical transformations. The aftermath of WWII was crystallising the division between two geopolitical blocs, while China, emerging from the Sino-Japanese War and the Chinese Civil War, faced the challenge of reconstructing its economy and positioning itself within this new international order.

For both ideological and geopolitical reasons, the Chinese government chose to pursue close economic cooperation with the Soviet Union. This collaboration culminated in the Sino-Soviet Treaty of Friendship, Alliance and Mutual Assistance of 1950 and Soviet involvement in China’s First Five-Year Plan. As part of this plan (1953–1957), the Soviet Union agreed to assist China in the construction of state-of-the-art industrial facilities. The programme involved extensive transfers of technology, human capital, and industrial knowhow, alongside financial assistance and equipment provision. At the height of the cooperation, approximately 20,000 Soviet scientific, industrial, and technical experts worked in China, helping design factories and organise production.

Following Stalin’s death, however, the alliance gradually deteriorated as ideological and geopolitical differences emerged between the two countries, leading to a formal Sino-Soviet split in 1961. The abrupt termination of industrial collaboration materialised in the sudden withdrawal of experts and engineers from China, the repatriation of Chinese students from the USSR, and the cancellation of ongoing industrial projects, leaving about 150 state-of-the-art industrial sites from the initially planned 156 factories.

A rise-and-fall pattern

A key challenge in identifying the spillovers generated by Million-Rouble Plants is isolating exogenous variation in the allocation of these investments. We exploit the fact that policymakers not only considered important economic criteria such as market access and access to natural resources but also responded to an uncertain geopolitical environment that induced some randomness. In particular, factories were preferentially located in areas protected from potential aerial attacks, taking advantage of the security umbrella provided by nearby Soviet-allied military bases. This criterion ceased to be relevant after the Sino-Soviet split, when strategic considerations were shaped by a very different geography of vulnerability. This shift is reflected in the subsequent Third Front Movement, which directed industrial investment towards China’s interior under the principle of locating facilities “close to the mountains, dispersed, and hidden in caves”.

Our empirical strategy adapts the idea of Greenstone et al. (2010) to this historical setting. We first identify a set of counterfactual locations that resemble the eventual host sites on observable economic characteristics (Figure 1). We then exploit variation in exposure to potential aerial attacks – which influenced the selection of host locations but became irrelevant after the Sino-Soviet split – as a source of exogenous variation in the selection process among these sites (Figure 2). 

Figure 1 Treated counties (red) and the group of control counties (blue)

Figure 2 Exposure to potential aerial attacks in 1953 (left) and in 1964 (right)

We find that the causal effects of hosting a Million-Rouble Plant were initially large and positive. GDP per capita was approximately 80% higher in treated counties, while the industrial employment share increased by 22 percentage points. These effects are an order of magnitude larger than what would be implied by the direct output of the average Million-Rouble Plant alone. However, treated and control counties experienced more than complete convergence between 1982 and 2010, as illustrated in Figure 3 for the industrial employment share. By 2010, treated counties were roughly 20% less productive than their control counterparts, even though the Million-Rouble Plants themselves remained productive and innovative. The decline instead reflects the performance of other manufacturing establishments in treated counties, which were less productive, less competitive, and less innovative than establishments in control counties.

Figure 3 The dynamics of counties hosting Million-Rouble Plants

What explains the poor performance of other establishments?

Two factors explain the poor performance of other establishments in treated regions. First, treated counties became increasingly specialised around the production chains of the Million-Rouble Plants. A disproportionate share of local firms operates either upstream or downstream of the Million-Rouble Plants. Yet these firms are less productive, less competitive, and less innovative than comparable firms elsewhere. Outside of the Million-Rouble Plants production network, economic activity remains fragmented across small and disconnected clusters. 

Figure 4 The missing entrepreneurs

The second factor is that potential entrepreneurs – defined as working-age individuals with a college degree – are less likely to occupy entrepreneurial leadership positions in treated counties. This pattern is partly explained by the fact that prospective entrepreneurs export their skills to other locations, echoing mechanisms discussed by Chinitz (1961) and Glaeser et al. (2015). As illustrated in Figure 4, college-educated cohorts born between 1955 and 1975 – and therefore at prime entrepreneurial ages in 2015 – are substantially less likely to hold top entrepreneurial positions in treated counties than in control counties.

This pattern does not reflect weaker human capital accumulation. If anything, treated counties are better educated than their control counterparts. Rather, we find that many of the potential entrepreneurs produced by these locations choose to export their skills and manage firms in other locations. The problem is therefore not the production of entrepreneurial talent, but its retention.

Overall, our findings confirm the established insight that industrial policies help concentrate business activities and, in a developing country context, can help trigger structural change. However, these medium-run gains may not persist in the long run. If clusters emerge around one or a few dominant firms, the production structure may become overly specialised, giving rise to close production relationships that are mutually reinforcing but come at the expense of competition from new entrants. Reduced competitive pressure can stifle innovation and limit the emergence of new growth opportunities, leaving dominant firms highly productive but with limited scope to generate the next wave of economic growth.

Any lessons for place-based industrial policies?

Place-based industrial policies remain a central tool of economic development. In advanced economies, a prominent example is the EU’s Smart Specialisation Strategy (S3), while developing countries have pursued similar approaches through initiatives such as Vietnam’s high-tech industrial clusters, Ethiopia’s Hawassa Industrial Park for textiles and apparel, India’s Production Linked Incentive (PLI) schemes for advanced chemistry cell batteries, solar photovoltaic modules, and drones, and Indonesia’s industrial down-streaming (‘hilirisasi’) policy for nickel, which combines special economic zones with investments in nickel processing and electric vehicle battery manufacturing.

China’s Million-Rouble Plants offer a rare opportunity to understand the long-run consequences of policies that simultaneously target specific sectors and specific locations. Our findings suggest that excessive industrial concentration can generate persistent costs. Places receiving highly specialised investments experienced weaker long-run economic performance. This implies that policymakers may mitigate the downside risks of place-based industrial policy by deliberately engineering diversification at the point of initial investment.

Place-based industrial policy can be effective, but its design matters. Rather than creating narrowly specialised industrial enclaves, policymakers may achieve more durable gains by fostering diversified local production structures that encourage knowledge spillovers, labour market resilience, and adaptation to technological and market shocks. In practice, this argues for combining strategic sectoral priorities with sufficient industrial variety within places, reducing the risk that local economies become locked into a single trajectory. As governments around the world devote large public resources to industrial policy, the historical experience of the Million-Rouble Plants programme serves as a reminder that diversification is not merely a complement to industrial strategy – it may be one of its key determinants of long-run success.

References

Atalay, E, A Hortacsu, M Runyun, C Syverson and M F Ulu (2023), “The impact of place-based policies on regional inequality”, VoxEU.org, 19 July.

Barone, G and G de Blasio (2023a), “Place-based policies in the Italian case, part 1: A lot of money for little or no growth”, VoxEU.org, 31 January.

Barone, G and G de Blasio (2023b), “Place-based policies in the Italian case, part 2: Mind the negative side effects”, VoxEU.org, 1 February.

Chinitz, B (1961), “Contrasts in Agglomeration: New York and Pittsburgh”, American Economic Review 51(2): 279–289.

Duranton, G and D Puga (2001), “Nursery cities: Urban diversity, process innovation, and the life cycle of products”, American Economic Review 91(5): 1454–1477.

Faggio, G, O Silva and W C Strange (2017), “Heterogeneous agglomeration”, Review of Economics and Statistics99(1): 80–94.

Fajgelbaum, P D and C Gaubert (2020), “Optimal spatial policies, geography, and sorting”, The Quarterly Journal of Economics 135(2): 959–1036.

Fan, J and B Zou (2021), “Industrialization from Scratch: The ‘Construction of Third Front’ and Local Economic Development in China’s Hinterland”, Journal of Development Economics 152.

Garin, A and J Rothbaum (2025), “The long-run impacts of public industrial investment on local development and economic mobility: Evidence from World War II”, The Quarterly Journal of Economics 140(1): 459–520.

Giorcelli, M and B Li (2022a), “Technology Transfer and Early Industrial Development: Evidence from the Sino-Soviet Alliance”, NBER Working Paper 29455.

Giorcelli, M and B Li (2022b), “Technology transfer and early industrial development: The case of the Sino-Soviet Alliance”, VoxEU.org, 10 January.

Glaeser, E L, H D Kallal, J A Scheinkman and A Shleifer (1992), “Growth in cities”, Journal of Political Economy100(6): 1126–1152.

Glaeser, E L, S P Kerr and W R Kerr (2015), “Entrepreneurship and urban growth: An empirical assessment with historical mines”, Review of Economics and Statistics 97(2): 498–520.

Greenstone, M, R Hornbeck and E Moretti (2010), “Identifying Agglomeration Spillovers: Evidence from Winners and Losers of Large Plant Openings”, Journal of Political Economy 118(3): 536–598.

Heblich, S, M Seror, H Xu and Y Zylberberg (2026), “Industrial Clusters in the Long Run: Evidence from Million-Rouble Plants in China”, CEPR Discussion Paper 21527.

Henderson, V, A Kuncoro and M Turner (1995), “Industrial Development in Cities”, Journal of Political Economy103(5): 1067–1090.

Mitrunen, M (2025), “War reparations, structural change, and intergenerational mobility”, The Quarterly Journal of Economics 140(1): 521–584.

Neumark, D and H D Simpson (2015), “Place-Based Policies”, in G Duranton, J V Henderson, and W Strange (eds.), The Handbook of Regional and Urban Economics, Elsevier/Academic Press.

Schweiger, H, A Stepanov and P Zacchia (2022), “The Long-Run Effects of R&D Place-Based Policies: Evidence from Russian Science Cities”, American Economic Journal: Economic Policy 14(3): 322–51.