In international affairs, what a country does depends on its own interests and on what its partners do. This column uses data on political institutions and UN voting to recover the network of geopolitical influence. The network tracks major shifts in international politics since the 1960s and reveals three notions of power: autonomy, control, and the ability to shape the game. Across all countries, the US has the most autonomy but China has the most control. Domestic political change in the US or China moves other countries the most, as the network transmits it furthest.

Geopolitics is a network phenomenon. On 2 March 2022, the UN General Assembly voted on a resolution condemning Russia’s invasion of Ukraine: 141 countries voted in favour, five against, and 35 abstained, India and South Africa among them. India’s abstention may reflect its own interests, the pull of its long-standing ties to Russia, or both. The vote by itself cannot tell us which.

What a country does in international affairs depends not only on its own interests but also on what its partners do. A sanction works only if others enforce it. US controls on exports of advanced chipmaking equipment to China, for example, depend on matching restrictions by the Netherlands and Japan, home to other key suppliers, which agreed to them in January 2023.

This interdependence creates a basic challenge: what countries do is a joint outcome, not a direct expression of what each would choose if it acted alone. The network that generates this pull is itself hard to observe, because influence travels through many channels that can pull in opposite directions. 

In Liu et al. (2026), we develop a framework that separates these forces and recovers the network as a whole. Countries choose geopolitical actions that balance their domestic interests against the actions of the countries they are connected to. Before acting, they spend resources to influence one another, either directly or through third countries: a great power that courts a regional leader also moves the neighbours that follow it.

The central implication is that the network filters disagreement unevenly. Disagreements that run between blocs pass through into actions, whereas disagreements among close partners are dampened by the incentive to coordinate. When two countries with very different political systems nonetheless vote alike, something is pulling them together. Comparing how strongly each dimension of domestic disagreement shows up in actions reveals the blocs, how much each matters, and hence the network itself.

Much existing research examines one channel of influence at a time, such as the economic instruments through which a hegemon exerts power (e.g. Clayton et al. 2026, Liu and Yang 2026) or the link between political alignment and trade exposure (e.g. Kleinman et al. 2024). Because our network is inferred from how domestic interests translate into actions, it aggregates all channels — trade, finance, arms, ideology — including those that are hard to measure directly. It requires only two inputs: a measure of what countries want and a measure of what they do.

For what countries want, we use the Varieties of Democracy (V-Dem) data, aggregating 37 variables that measure domestic interests and political institutions (Coppedge et al. 2011, Teorell et al. 2019). For what they do, we use votes in the UN General Assembly (Voeten 2000, Bailey et al. 2017). We estimate the network separately for each decade from the 1960s to the 2010s.

The network’s leading cleavage — the dimension along which tightly connected blocs agree internally and disagree with one another — tracks major shifts in the structure of international politics (Figure 1). In the 1960s, it aligns closely with the Cold War division between the Western and Soviet blocs. In the 1970s, it shifts to a divide between the major powers — including the US and the Soviet Union — and the developing world. China sits with the major powers, even though Deng Xiaoping had told the General Assembly in 1974 that China belonged to the Third World. The East–West divide returns in the 1980s. Since the 1990s, the US and Western Europe have occupied one pole, China the other, and many countries in the Global South lie in between. None of these divides is imposed on the data.

Figure 1 The leading cleavage by decade

Notes: Leading cleavage by decade. The countries with the darkest red and the darkest blue are those with the most extreme loading on the geopolitical cleavage. The leading cleavage is the second joint principal component of the cross-country covariance matrices of UN General Assembly votes and V-Dem domestic interests, estimated separately for each decade; the first captures consensus shared by all countries, and the top five account for 97% and 96% of the two matrices, respectively. A dimension becomes the leading cleavage because votes diverge along it by more, relative to domestic interests, than along any other, so the familiar East–West split in raw UN votes does not by itself produce the ranking. All maps use Natural Earth’s country-level dataset (version 5.1.1, released in May 2022), with the same geographic boundaries used across all decades. The country boundaries follow the source dataset and do not imply any position by the authors regarding sovereignty, territorial claims, or political status.

The ties we infer line up with information we did not use to construct them: countries with closer ties trade more, are closer geographically, and more often share a religion, a language, or a former coloniser, and the influence our model predicts between two countries tracks how often their leaders visit each other.

The estimated network lets us run counterfactuals that sit at the centre of current policy debates. Would a China with US-style domestic institutions align with the West? Do alliances extend a great power’s reach or tie its hands? Which countries’ absence would most change how the rest of the world behaves? Answering these questions yields three notions of power: autonomy, how much of a change in a country’s interests shows up in its own actions; control, how much other countries’ actions move when its own action moves; and the ability to shape the game, how much everyone else’s actions would change without it in the network. 

A convergence in China’s domestic interests towards those of the US would pass through only partly into its own actions: China’s action moves 38% of the way, because its network partners, such as Russia and India, anchor it. The change that does occur moves partners such as Syria, North Korea, and Iran, whose own interests are unchanged (Figure 2). Syria and North Korea were two of the five countries that voted with Russia in March 2022.

This does not mean that political change in China would be inconsequential. Our counterfactuals hold the network fixed, and a change this large might, over time, rewire it.

Across all countries, the US has the most autonomy (60%), followed by China (38%) and India (30%), which is one reason a single vote, such as India’s abstention in 2022, is a weak guide to what a country wants. China has the most control: each unit of movement in its action moves the average other country by 0.26 units, about twice as much as for India (0.15) or the US (0.13).

This reflects China’s ties rather than its size. The two superpowers have nearly the same economic weight in our estimates, but other countries place more than twice as much weight on China, consistent with the Belt and Road Initiative and with China having overtaken the US as the larger trade and investment partner of many countries.

The difference shows up in UN votes. Across the 766 General Assembly roll calls of 2010–2019, shifting China’s interest on each vote across the middle half of the range of countries’ interests changes 6,088 votes cast by other countries, about eight per roll call. The same shift changes 3,408 votes for the US and 249 for India. For no other country does it change more than five.

In January 2026, after eight European NATO members sent military personnel to a Danish-led exercise in Greenland, President Trump threatened them with tariffs; their leaders warned jointly of ‘a dangerous downward spiral’. The threat was withdrawn days later. Weakening the network ties between the US and its European NATO allies by three quarters raises US autonomy from 60% to about 90%, but lowers its control from 0.13 to about 0.11.

Alliances that constrain the US also extend its reach. European allies pull US actions toward their own interests, and the same ties carry US actions to the rest of the world. France, by contrast, loses almost no control, because its position rests mostly on its ties within the EU.

Weakening the China–Russia tie by the same three quarters raises both countries’ autonomy and lowers their control, but by less than in the transatlantic case, partly because a single tie is a smaller part of the network. It changes no individual UN vote. By these estimates, loosening America’s own alliances would do more to reshape the network than splitting its two main rivals.

India sits between the blocs, in the Quad with the US, Japan, and Australia, and in BRICS and the Shanghai Cooperation Organization with China and Russia. Days after the US raised tariffs on Indian goods to 50% in August 2025, citing India’s purchases of Russian oil, Prime Minister Modi met Xi Jinping and Vladimir Putin at the organisation’s summit in Tianjin.

Tying itself more closely to either side would reduce India’s autonomy, as its actions come to reflect more of its partners’ interests, and raise its control, as new ties carry its influence further. The gain in control is largest when India leans East, reflecting its location and its ties to Asian countries oriented toward China and Russia. This is the trade-off behind India’s pursuit of ‘strategic autonomy’, although neither shift is large enough to change an individual UN vote.

Debates about isolating Russia after its invasion of Ukraine, and about the US withdrawal from 66 international organisations in January 2026, turn on a common question: how would the rest of the world line up without a given country? We remove each country from the network in turn and measure how much everyone else’s actions change (Figure 3).

Removing the US or China changes others’ actions about twice as much as removing India, and 2.4 times as much as removing Japan or Russia. Size accounts for most of a country’s ability to shape the game, but not all: Russia matters as much as Japan despite its smaller economy, and more than half of its influence runs through its partners. Iran also has a large impact relative to its size, Australia a small one.

Figure 2 Change in each country’s geopolitical actions if China’s domestic interests matched those of the US

Notes: Map of the mean absolute change in each country’s equilibrium geopolitical actions when China’s interest is set equal to that of the US, holding all other countries’ interests and all network connections at their 2010–2019 values. Changes are averaged over 1,000 international issues drawn from the estimated distribution of domestic interests. China’s own action moves by 4.3 on average, 38% of the 11.3 it would move under full pass-through. All maps use Natural Earth’s country-level dataset (version 5.1.1, released in May 2022), with the same geographic boundaries used across all decades. The country boundaries follow the source dataset and do not imply any position by the authors regarding sovereignty, territorial claims, or political status.

Figure 3 Size explains most, but not all, of a country’s ability to shape the game

Note: This figure shows the results of excluding each country in turn from the network and recomputing the equilibrium action vector among the remaining countries. Vertical axis shows the change in the equilibrium configuration of actions, measured as the expected Euclidean distance between the remaining countries’ actions with and without the excluded country, over international issues drawn from the 2010–2019 distribution of domestic interests. Horizontal axis shows the excluded country’s structural importance, as estimated using GDP at purchasing power parity. The dashed line shows the linear regression fit; countries above it, such as Russia and Iran, shape the game by more than their size implies.

Per unit of change, shifts in what the US or China wants at home move geopolitical actions about three times as much as the largest network change we consider, the weakening of transatlantic ties, and change more than ten times as many UN votes.

Power is not simply a function of national capabilities or bilateral leverage, but also of where countries are positioned within the network and how influence propagates through it. Whether a sanctions coalition holds, whether export controls bite, and which way a swing state leans depend on that position as well as on what countries want. In our estimates, the domestic politics of Washington and Beijing matter most because the network carries them furthest.

Source : VOXeu