EU barriers to scaling up: The case of fragmented product markets

It is by now well established that EU firms underperform their American counterparts when it comes to productivity. The main issue is a lack of scale East of the Atlantic. In the US, the average mature firm (above 25 years old) employs eight times as many workers as an average young firm (below the age of three). In the EU, that scaling up factor is a meagre two (IMF 2024). Absent scale, EU firms struggle to absorb the fixed costs of innovation, perpetuating the productivity gap with the US.

Recent IMF work has highlighted a few critical factors behind this staggering scaling-up deficiency: domestic structural policy gaps (Budina et al. 2025), barriers to cross-border mobility of workers and capital (Arnold et al. 2025, IMF 2025), and product markets that remain fragmente…

Among these factors, the presence of intra-EU trade barriers, and their estimated size, have taken on a prominent role in the European policy debate of the past year. In this column, based on Cerdeiro and Rotunno (forthcoming), we present evidence of remaining barriers to product market integration in the EU. Given that it is widely accepted that domestic services sectors remain protected through high barriers, the focus is on goods markets.

What barriers to trade?

The principle of the free movement of goods within the EU is enshrined in Article 26 and Articles 28-37 of the Treaty on the Functioning of the European Union (TFEU). This includes the principle established in the 1979 “Cassis de Dijon” ruling that, with some exceptions, products sold lawfully in one member state can also be sold in another.

These legal underpinnings have made the EU the most successful integration agreement in the world by far (Hofmann et al. 2019). With such solid legal foundations, one could reasonably think that EU goods markets are fully integrated. Here we discuss two sets of evidence suggesting that there is still scope for more integration.

Qualitative and sectoral evidence

There is ample qualitative evidence that helps illustrate where untapped potential lies. 1

Diverging national labelling and packaging requirements provide costly market entry costs to EU-wide distribution. For example, the use of the ‘Green Dot’ symbol is mandatory in some member states but penalised in others, forcing manufacturers to develop country-specific packaging (see Garicano et al. 2025 for further examples).
Territorial supply constraints (Letta 2024) mean that retailers wishing to buy certain goods face compulsory referral to the national branch of their manufacturers. This practice prevents low-price products from other member states from being bought through channels other than those established by the brand owner.
Goods trade is slowed by a fragmented transportation sector. Felbermayr and Tarasov (2022) document that road distances between two points that cross a national border are about 22% longer than within-country connections, and underinvestment in cross-border infrastructure can explain about 20% of the border effect in a standard gravity regression.
Government procurement is biased, despite EU Directive 2014/14 aiming to ensure non-discrimination and equal treatment. Foreign goods represent around one-quarter of goods purchases by firms and households in France and Spain, for example, compared to just around 2% for government purchases (Garcia-Santana and Santamaria 2024).

What do firms say?

Surveys can help gauge more systematically what firms have to say. 2 According to the 2025 EIB Investment Survey, 62% of EU firms report have varying requirements and standards for their main product across different member states (Bending et al. 2025). Among small and medium-sized enterprises (SMEs) already exporting to other EU member states in 2025, the European Commission (2025) reports difficulties in scaling up operations related mainly to taxation and VAT (30% of firms), permitting and authorisations (28%), and the posting of workers (17%).

Unearthing barriers from trade data

Some background

Following the seminal work of McCallum (1995) on the US-Canada border (see also, inter alia, Anderson and Yotov 2010), scholars have used ‘border effect’ estimates to assess the size of barriers to cross-country trade by comparing the extent of cross-border to within country trade (e.g. Head and Mayer 2000).

Identification of the level of intra-EU trade barriers using a gravity approach requires specifications without country-pair fixed effects – normally included to identify changes in trade barriers – since these would absorb the border effect. In lieu of the fixed effects, specifications that identify level estimates control for observable bilateral determinants of trade, such as distance, which do not directly respond to policies.

What is a reasonable benchmark?

Focusing just on intra-EU trade data, IMF (2024) and Adilbish et al. (2025) estimate a level of trade costs within the EU of 44% for manufacturing and 110% for services, on average. Bernasconi et al. (2026) estimated these at 67% and 95%, respectively. 3 As we emphasised in IMF (2024) and Adilbish et al. (2025), however, it is difficult to control for all bilateral determinants of trade. Notably, preferences for locally produced goods are not easily measurable and would therefore be loaded on gravity-based estimates of barriers. But there are ways around this problem.

If internal US trade is largely free from man-made barriers, internal US barrier estimates greater than zero should reflect this bias toward locally produced goods. As done in the seminal work by Head and Mayer (2021), benchmarking to internal US barriers can thus be helpful as a way to size up the policy scope for intra-EU barrier reductions. In IMF (2025), we therefore expand on Head and Mayer (2021) and estimate intra-EU and intra-US barriers jointly.

Focusing on the difference between internal US and intra-EU trade barriers can both over- and underestimate the full scope for intra-EU liberalisation. On the one hand, higher estimated intra-EU barriers can partly reflect preferences for locally produced goods that arguably are stronger across countries than across US states. Note, though, that the gravity-based approach does strip out any home bias that is correlated with bilateral observables, such as different languages. In other words, while the approach would fail to strip out, for example, any Austrian preference for Austrian beer over German beer, it could capture a possible preference of French consumers for French over Italian wine. On the other hand, the US is unlikely to be a fully frictionless benchmark.

Datasets of intra-EU (and intra-US) trade

Three different types of data that provide both cross-border and domestic trade can be leveraged to carry out a gravity-based estimation of internal trade barriers in the EU:

Class A: Trade and gross output databases. These databases rely on standard sources for international trade flows (e.g. COMTRADE) and impute domestic sales as the difference between gross output and total exports. Prominent examples are the ITPD-E (Larch et al. 2025) and TradeProd (Mayer et al. 2023) databases. The dataset assembled by Head and Mayer (2021) also belongs in this class. The absence of any adjustments due to balancing requirements (ensuring, for instance, that total supply equals total use) makes these datasets the go-to resource to estimate trade policy effects.
Class B: Global input-output (I-O) databases. Global I-O databases map flows of production, consumption and investment within and between countries in a way that is consistent across countries and with national accounts. The OECD TiVA and the Eurostat FIGARO Inter-Country Input-Output (ICIO) databases belong to this class. As we argue below, the balancing requirements lead to data adjustments in the input output tables that are critical for our application.
Class C: Freight surveys. Statistical agencies sometimes compile data on freight flows. The European Road Freight Transport Survey (ERFT) is a vehicle-based survey that is arguably the most comparable to the only available data for internal US trade flows, namely, the US Commodity Flow Survey (CFS). ERFT does not contain trade values; Santamaria et al. (2021, 2023) convert ERFT shipment microdata into annual bilateral trade values between regions.

A key dimension of these datasets for our purposes is how domestic sales compare to cross-border sales. 4 Figure 1 shows the ratio of cross-border (across EU countries or US states) to domestic sales for the different datasets in the year 2017 in manufacturing. The level of cross-border relative to domestic trade is higher in the US than in the EU. The value of trade by truck in the CFS across US states is 1.8 times the value of trade within US states, compared to just one-quarter in EU ERFT data. Relative cross-border trade in the EU remains well below the ratio in the US using other data, but with noticeable differences. Class B data give generally a lower cross-border to domestic ratio within the EU than Class A data (with important differences even within this class).

Figure 1 Ratio of cross-border (within EU and the US) sales to domestic sales

(2017, manufacturing goods; for the US, cross-border refers to cross-state)

Note: “US road” includes only road, thereby being most comparable to the EU’s ERFT survey. See Cerdeiro and Rotunno (forthcoming) for further details.

Sources: Santamaria et al. (2023), OECD, Eurostat, TradeProd, Head and Mayer (2021), ITPD-E.

Why is cross-border EU trade larger (relative to domestic trade) in Class A than in Class B datasets? We can identify two main reasons.

Missing domestic sales. In Class A datasets, aggregation across many sectors counts missing gross output values as zero, leading to an underestimation of domestic sales in the aggregate. For example, in the TradeProd database, the textile and apparel sectors of France and Germany have missing domestic sales in 2017, even though both sectors export around $40 billion. In ITPD-E data, the “wearing apparel” sector of these two countries in 2017 has missing domestic sales, while it reports between $6 and $10 billion worth of exports. While these missing values for domestic trade may lead to underestimation of the corresponding aggregate, they do not affect estimates done at the industry level such as those of Fontagne and Yotov (2025).
Re-exporting. In many countries, large shares of exports to neighbouring countries can be re-exports of products without further transformation. For countries that are major regional trading hubs – such as Belgium (Antwerp), the Netherlands (Rotterdam), and Germany (Hamburg) – these shares can be very high. National supply and use tables show that re-exports account for approximately 45% of total (i.e. not only intra-EU) manufacturing exports by Belgium and the Netherlands in 2017; even in France, which has important ports like Marseille, re-exports account for 11% of total exports. Trade values in I-O tables are adjusted to only reflect imports retained in the importing country coming from exports by the country of origin of the products (Yamano et al. 2023). 5 In COMTRADE, re-exports and re-imports are often poorly reported – for instance, Belgium and the Netherlands have no re-exports and re-imports for 2017. Figure 2 shows that Class A databases report higher intra-EU exports than TiVA and FIGARO, with the difference being largest for EU ‘gateway’ countries. 6 The effect of these re-exports can rise if these countries process more trade over time (e.g. due to deeper integration with China).

The absence of balancing requirements and other modelling adjustments in Class A datasets can be beneficial when assessing the trade effects of trade policies (provided re-exports are not part of the shock), but a shortcoming for estimating the level of internal EU barriers. Lower domestic sales and higher cross-border sales would both contribute in the same direction to lower internal barrier estimates. Note, however, that the two differences have opposing effects on aggregate gross output estimates: while the former leads to underestimating a country’s gross output, the latter would inflate it. Figure 3 shows, for the three largest intra-EU exporters, and Belgium and the Netherlands, that the issue of re-exporting can be large enough to more than offset underreported domestic sales. Relative to EU KLEMS, Class A datasets overestimate manufacturing gross output, especially for Belgium, the Netherlands and France.

Sizing up the scope for further product market integration

Table 1 shows the estimated ad-valorem equivalent of intra-US and intra-EU trade barriers using the different datasets described above, for the year 2017. 7

Consistent with the descriptive evidence in Figure 1, estimated intra-EU trade barriers are significantly higher than trade barriers across US states. Using the ERFT data for intra-EU trade, we find that intra-EU barriers are equivalent to 44%, almost three times the level of intra-US barriers. Using TiVA and FIGARO for intra-EU trade flows and the full CFS data for the US in the second and third columns lowers the gap between intra-EU and intra-US barriers, which nonetheless remains both economically and statistically significant. As in Head and Mayer (2025), the estimated intra-EU barriers are, however, on average lower with Class A datasets. They remain higher than intra-US ones with TRADE PROD data, and go below the US ones with the ITPDE dataset.

Table 1 Estimates of trade barriers across datasets

(ad valorem equivalent)

Notes: All ad-valorem estimates apply a trade elasticity of 5. Standard errors in parentheses for ad-valorem equivalents. The “Test” row reports the chi-squared statistic. See Cerdeiro and Rotunno (forthcoming) for further details.

Closing Europe’s productivity gap

Both qualitative and quantitative evidence suggests that intra-EU trade barriers are high and – once excluding data that likely overstate cross-border trade – higher than what is observed for the US. This means that finding ways to ease commerce across borders in the EU can help close the EU’s productivity gap. For example, IMF (2025) finds that reducing intra-EU trade costs to the level estimated to prevail between the states in the US could lift EU productivity by 5.7%. In a comprehensive reform package that also reduces labour mobility barriers and fully closes domestic structural reform gaps, EU productivity would be higher by about 20%. A still ambitious but more immediately feasible intermediate reform package where these structural gaps are halved would yield still substantial gains of about 9% in the aggregate.

None of these reforms would be easy. They would require, above all, political will. Including, and perhaps most of all, from member states – so that they champion an ambitious reform agenda at home and at the EU level.

Authors’ note: The views expressed herein are those of the author and should not be attributed to the IMF, its Executive Board, or its management.

References

Adilbish, O E, D A Cerdeiro, R A Duval, G H Hong, L Mazzone, L Rotunno, H H Toprak, and M Vaziri (2025), “Europe’s Productivity Weakness: Firm-Level Roots and Remedies”, IMF Working Paper 25/040.

Airaudo, F, F de Soyres, A Gaillard and A M Santacreu (2025), “Recent evolutions in the global trade system: From integration to strategic realignment”, Forum on Central Banking, European Central Bank.

Arnold, N G, A Dizioli, A Fotiou et al. (2025), “Lifting Binding Constraints on Growth in Europe: Actionable Priorities to Deepen the Single Market”, IMF Working Paper 25/113.

Anderson, J E and Y V Yotov (2010), “The Changing Incidence of Geography”, American Economic Review 100(5): 2157–2186.

Bending, T, M Ferrazzi, P Harasztosi, D Revoltella, H Tran and V Žďárek (2025), “Geopolitical shifts, uncertainty, and investment: Evidence from the EIB Investment Survey 2025”, VoxEU.org, 17 October.

Bernasconi, R, N Cordemans, V Gunnella, G Pongetti, and L Quaglietti (2026), “What is the untapped potential of the EU Single Market?”,* ECB Economic Bulletin *8/2025.

Budina, N, O E Adilbish, D A Cerdeiro et al. (2025), “Europe’s National-Level Structural Reform Priorities”, IMF Working Paper 25/104.

Cerdeiro, D A and L Rotunno (Forthcoming), “EU barriers to scaling up: The case of fragmented product markets”, IMF Working Paper.

European Parliament (2020), “Legal obstacles in Member States to Single Market rules”, November.

European Commission (2025), “A simpler Single Market to make companies choose Europe,” 20 May.

Garicano, L, B Holmström, and N Petit (2025), The Constitution of Innovation: A New European Renaissance.

Hammershøj Bentsen, O, M S van Deurs and A Y Zhuang (2025), “Fewer barriers in the EU single market could increase Danish welfare”, Analysis 17, Danmarks Nationalbank.

Head, K and T Mayer. 2000. “Non-Europe: The Magnitude and Causes of Market Fragmentation in the EU,” *Weltwirtschaftliches Archiv *136(2): 284–314.

Head, K and T Mayer (2021), “The United States of Europe: A Gravity Model Evaluation of the Four Freedoms,” *Journal of Economic Perspectives *35(2): 23–48.

Head, K and T Mayer (2025), “No, the EU does not impose a 45% tariff on itself”, VoxEU.org, 13 November.

Hofmann, C, M Ruta, and A Osnago (2019), “Horizontal Depth: A New Database on the Content of Preferential Trade Agreements,” World Trade Review 18(3): 365-398.

IMF – International Monetary Fund (2024), “Regional Economic Outlook for Europe Note 1: Europe’s Declining Productivity Growth: Diagnoses and Remedies”, October.

IMF (2025), “Regional Economic Outlook for Europe Note 2: Making European Reforms a Success on the Ground”, November.

McCallum, J (1995), “National Borders Matter: Canada–U.S. Regional Trade Patterns”, American Economic Review 85(3): 615–623.

Santamaría, M, J Ventura, and U Yeşilbayraktar (2021), “Borders within Europe”, CEPR Discussion Paper No. 15633.

Santamaría, M, J Ventura, and U Yeşilbayraktar (2023), “Exploring European regional trade”, Journal of International Economics 146.

Yamano, N, A Alsamawi, C Webb, A Cimper, C Zürcher, and R C Pechanskyand (2023), “Development of the OECD Inter Country Input-Output Database 2023”, OECD Science, Technology and Industry Working Paper 2023/08.