The debate over deglobalisation, ‘friend‑shoring’, and the weaponisation of finance has pushed cross‑border restrictions back to the centre of the policy agenda. The idea that “tariffs on goods may be a prelude to tariffs on money”
captures a widespread concern: will the next phase of geoeconomic fragmentation be driven as much by controls on capital, payments, and financial institutions as by trade measures?

This is hardly a theoretical question. From the post‑war Bretton Woods system, policymakers have repeatedly revisited how much ‘sand in the wheels’ of international finance is desirable. A large academic literature has tried to measure and assess these policies – from early narrative indicators of capital account regulation such as Quinn (1997) to binary AREAER‑based indexes in Chinn and Ito (2006), and the more granular asset‑by‑asset measures in Fernández et al. (2016).

Yet, despite decades of work, our empirical picture of cross‑border restrictions has remained surprisingly narrow. Most measures focus on capital account items, which are less than one-fifth of total cross-border restrictions in the new data. In addition, they are annual and often compress different types of tools into a single number by measuring the stance, not the action on individual policy changes. This makes it difficult to speak to today’s policy debates about how, when, and why countries use controls as part of broader macro‑financial toolkits. In a new study (Bergant et al. 2026) we revisit this question using modern AI tools to create a daily dataset on cross-border restrictions for all IMF member countries for the last 75 years and make the data publicly available for all researchers.

Letting AI read 14 million words on the history of cross-border controls

The starting point is the IMF’s Annual Report on Exchange Arrangements and Exchange Restrictions (AREAER), published since 1950. A section of the reports records, in narrative form, all restrictions that member countries impose on international payments and transfers, current and capital account transactions, foreign exchange markets, and related financial‑sector regulations. Over time, they have become both richer and more complex: by 2022, the cumulative text reaches about 14 million words, with increasingly technical jargon to describe the growing web of cross-border tools.

Thankfully, the AI doesn’t get tired of even the most complex reading. To turn this unstructured and changing document into systematic data, we fine‑tune a domain‑adapted language model on manually labelled AREAER passages,
and then deploy it to classify every reported policy change since 1950 along several dimensions: category of policy (e.g. foreign exchange markets, capital accounts, etc.), direction of change (tightening or loosening), direction of flow (inflows, outflows, or both), and the type of instrument (price‑based, quantity‑based, administrative).

We then build two complementary indices:

1. An integrated BoP restrictions index in changes, iBoP‑C – a high‑frequency measure of the changes in net tightenings and loosenings of restrictions implemented at the country–category level, available from 1950, and easily aggregated from daily to monthly, quarterly, or annual frequencies.
2. A complementary stance index, iBoP-S – an annual measure of the level of restrictiveness at the extensive margin, constructed from the post‑1995 stance sections in the AREAER using the same LLM method.

Crucially, we validate our AI‑based classifications in two ways. First, we show that the large language model performs on par with human annotations — despite being far less labour-intensive. Second, we benchmark our index against existing indexes of cross-border flow restrictions mentioned above. These strongly correlate with our indexes but not perfectly as our index offers additional information, thanks to the longer time series, broader coverage, and daily frequency.

Three lessons from seven decades of cross‑border restrictions

Lesson 1: Financial liberalisation – but in a non‑linear way

The popular narrative is one of steady financial liberalisation: from heavy post‑war controls to the ‘Great Moderation’ of open capital markets, followed by some post‑global crisis rethinking. Our data reveal a more stop‑go pattern (Figure 1, left panel).

Figure 1 Average cumulative net tightenings: Full sample and by income group

Notes: The left panel shows average cumulative net tightening index (iBoP-C) (negative values indicate net loosening). The right panel shows average cumulative net tightening (iBoP-C) by income group (2024 World Bank Income Group Classification).

After some post‑1950 easing, restrictions tightened sharply around the breakdown of Bretton Woods. Foreign exchange market measures, requirements on export proceeds, and financial‑sector restrictions all saw large waves of tightening.

From the mid‑1980s onwards, most categories exhibit a sustained liberalisation, consistent with broader financial deregulation and globalisation.

Lesson 2: Liberalisation has been uneven across countries and instruments

High‑income economies liberalised earlier and more decisively than middle‑income countries, which followed a similar path with a lag of roughly two decades, particularly for capital account measures. Many low‑income countries tightened restrictions in the 1970s and remain comparatively restrictive today, especially through current‑account measures such as import payment and export‑proceeds regulations (Figure 1, right panel).

The type of instrument also mattered: quantity‑based and administrative restrictions (quotas, bans, licences, approval requirements) show the strongest long‑run liberalisation whereas price‑based measures (taxes, fees, spreads, subsidies) are more persistent and have been relaxed more modestly (Figure 2, left panel). This pattern fits the view that once administrative bottlenecks and quotas are removed, it is relatively easier – politically and practically – to keep some price‑based levers in place.

Figure 2 Average cumulative net tightenings by type of restriction and by direction of flow

Notes: The left panel shows average cumulative net tightening (iBoP-C) by type of restriction (price-based, quantity-based, administrative). The right panel shows average cumulative net tightening (iBoP-C) for inflow and outflow restrictions.

Lesson 3: Outflows have been liberalised more than inflows – but not everywhere

Our classification allows us to distinguish measures targeted at inflows from those aimed at outflows. Both types have been liberalised since the 1980s, but outflow controls have been rolled back faster and further, especially in advanced and larger emerging economies (Figure 2, right panel).

This is consistent with their role as crisis‑management tools: authorities often impose outflow controls during acute stress to prevent capital flight and exchange‑rate collapse, and remove them once conditions stabilise. Inflow controls, by contrast, are used to lean against credit booms and balance‑sheet risks and tend to be more persistent.

Importantly, low‑income countries remain less likely to liberalise outflow restrictions. Even in recent years, low‑ and middle‑income countries are roughly twice – and for capital account items sometimes three times – as restrictive as high‑income economies.

How are these tools used in practice? Bunching, crises, and politics

The high‑frequency nature of the new data also sheds light on how countries deploy cross‑border restrictions.

First, measures are rarely implemented in isolation. About 35% of policy actions occur on days when at least one other measure is adopted on the same day. If we look at a ±30‑day window, more than three‑quarters of measures are accompanied by other restrictions.

Moreover, countries do not just pile up multiple tools within a narrow window, they also mix and match across categories. Episodes of ‘policy packages’ frequently combine capital‑account measures with changes in foreign exchange regulations, current‑account procedures, and financial‑sector rules. This has an important implication for empirical work: single‑policy event studies are likely to misattribute the effects of coordinated packages to one flagship measure.

Second, the number of measures rises markedly during macro‑financial crises. Using the Laeven and Valencia (2018) dataset, we find that in crisis years countries more than double their typical annual use of cross‑border restrictions (two versus one measure in non-crisis years). The increase is driven mainly by tightening measures and by policies targeted at outflows, especially during currency and sovereign debt crises.

Third, the political economy of these tools is non‑trivial. Using PRS institutional‑quality indicators, we document that countries with weaker institutions and higher political–financial–economic risk adopt significantly more restrictions, even after controlling for country and year fixed effects. Election years also see a statistically significant uptick in cross‑border measures, echoing recent findings that macroprudential tools are adjusted opportunistically around elections (Müller 2023).

What motivates capital controls today?

To open the ‘black box’ of policymakers’ motives, we complement the AREAER‑based indices with the IMF’s Taxonomy of Capital Flow Management Measures, which identifies ‘macro‑critical’ capital flow measures since the adoption of the Institutional View in 2012.
For this subset, we hand‑collect official statements and classify the stated motivations.
As can be seen in Figure 3, six main (non‑mutually exclusive) motives emerge.

The most common are fears of disruptive outflows and floating, followed by the fear of overborrowing. About a fifth of measures are motivated by long‑term structural objectives, and roughly a third of motivations are not directly linked to the business cycle. Low‑income countries more often cite outflow and exchange‑rate concerns, while high‑income countries emphasise overborrowing and financial‑stability risks from inflows.

Figure 3 Motivations for changing cross-border flow restrictions

Notes: Categories are not exclusive.

Implications for today’s policy debates

As debates on deglobalisation, sanctions, and financial fragmentation intensify, understanding the full landscape of cross‑border restrictions – not just capital controls narrowly defined – will be critical. Our hope is that the new iBoP measures will provide a common empirical foundation for that conversation.

We also show that the use of large language models allows us to transform a vast, semi‑structured archival record into a living dataset, which can be updated with the click of a button. This opens the door to revisiting classic questions – about the effectiveness, spillovers, and welfare consequences of controls – with richer data and cleaner identification strategies.

Author’s note: The views expressed in this column are those of the authors and do not necessarily reflect those of the International Monetary Fund, its Executive Board or its management.

References

Bergant, K, A Fernández, K Teoh and M Uribe (2026), “Expanding the landscape of cross‑border flow restrictions: Modern tools and historical perspectives”, CEPR Discussion paper No. 21045.

Chinn, M and H Ito (2006), “What matters for financial development? Capital controls, institutions, and interactions”, Journal of Development Economics 81(1): 163–192.

Fernández, A, M W Klein, A Rebucci, M Schindler and M Uribe (2016), “Capital control measures: A new dataset”, IMF Economic Review 64: 548–574.

Laeven, L and F Valencia (2018), “Systemic Banking Crises Revisited”, IMF Working Paper 2018/206.

Magud, N, C M Reinhart and K S Rogoff (2018), “Capital controls: Myth and reality”, Annals of Economics and Finance 19(1): 1–47.

Müller, K (2023), “Electoral Cycles in Macroprudential Regulation”, American Economic Journal: Economic Policy 15(4): 295–322.

Quinn, D (1997), “The correlates of change in international financial regulation”, American Political Science Review 91(3): 531–551.

Quinn, D and A Toyoda (2008), “Does capital account liberalization lead to growth?”, Review of Financial Studies 21(3): 1403–1449.