Deglobalization Trends and Communication Variables: A Multifaceted Analysis from 2009 to 2023

Abstract

This paper examines the correlation between rising trade protectionism—an indicator of economic deglobalization—and key communication and social variables from 2009 to 2023. Drawing on data from Global Trade Alert, Nexis Uni, Google searches, and Facebook (via CrowdTangle), we investigate the prevalence of “deglobalization” discourse, language entropy, political polarization, protests, and digital authoritarianism. The analysis is framed by Optimal Information Theory, World Systems Theory, and other social science perspectives to explain how deglobalization may potentially reshape public communication. The results suggest that greater trade protectionism is associated with increased mentions of deglobalization, higher language entropy (i.e., less dominance of English), amplified political polarization, more frequent protest activity, and heightened digital authoritarian measures.

1. Introduction

1.1. Setting the Stage: From Global Financial Crisis to Deglobalization

The 2008–2009 global financial crisis was a watershed moment in international trade and global economic integration dynamics. Before that crisis, the late 20th and early 21st centuries were frequently characterized by accelerating globalization: nations lowered trade barriers, signed multilateral agreements, and embraced free-market principles that promoted cross-border trade, capital flow, and labor mobility [1,2,3,4,5]. The post-World War II regime of gradually expanding global trade, underpinned by institutions such as the General Agreement on Tariffs and Trade (GATT) and its successor, the World Trade Organization (WTO), had ushered in decades of relatively consistent growth in global economic interdependence.

However, the financial turmoil of 2008–2009 profoundly shocked this established order, catalyzing a surge in economic uncertainty and political discontent. As world economies struggled to recover, new waves of populist sentiment and nationalist rhetoric began to gain traction, challenging long-standing orthodoxies about the benefits of free trade [6]. Governments in many regions started adopting more inward-looking policies, including tariffs, subsidies for domestic industries, and other trade barriers [6,7]. This confluence of protectionist measures and nationalist politics has often been labeled “deglobalization”: a reorientation from global commitments toward domestic priorities. While deglobalization may manifest primarily in economic terms [8], there is a growing consensus that its impacts are far broader, encompassing cultural identity, political polarization, protest activity, and governance of digital spaces [9,10]

1.2. Research Problem and Significance

Against this backdrop, the present research investigates how deglobalization—operationalized largely through the lens of trade protectionism—correlates with a series of communication variables: (1) the prevalence of “deglobalization” discourse itself, (2) language entropy, (3) political polarization, (4) protest activity, and (5) digital authoritarian practices. While numerous studies examine deglobalization in economic or geopolitical terms (e.g., rising tariffs and shifts in foreign direct investment), fewer systematically link these trends to transformations in communication patterns and social outcomes. Communication is integral to the social fabric, how individuals and institutions converse, how languages are adopted or marginalized, and how governments regulate the flow of information. These processes may reflect or even exacerbate economic reorientations in many parts of the world.

Understanding these linkages is crucial for both scholars and policymakers. Scholars gain insight into how macro-structural shifts in trade relationships cascade into shifts in language use, social conflict, and digital freedoms. On the other hand, policymakers may benefit from recognizing how protectionist choices can inadvertently fuel domestic polarization or prompt them to clamp down on dissent online. Such an appreciation is increasingly vital in an era when digital communication is central to politics, civil society, and global commerce [11,12].

1.3. Scope

Methodologically, this paper focuses on the period from 2009 to 2023, capturing the years after the global financial crisis. We adopt an annual time interval, which aligns with the typical yearly cycle for implementing trade measures, such as tariffs and subsidies. We compile data from multiple sources, including Global Trade Alert (GTA) for the number of new and ongoing protectionist measures, Nexis Uni and Google for media references, and Facebook’s CrowdTangle for large-scale text data, on which we perform language identification.

1.4. Research Questions

• RQ1: Does increasing trade protectionism correlate with heightened deglobalization discourse and shifts in language usage (i.e., language entropy)?
• RQ2: Are higher levels of trade protectionism associated with intensified political polarization and more frequent public protests?
• RQ3: Does rising trade protectionism predict enhanced digital authoritarianism (e.g., more censorship and internet shutdowns)?

By addressing these questions, we aim to integrate perspectives from communication studies, political economy, and sociology, shedding light on how deglobalization is embodied and enacted in information flows, collective behaviors, and government strategies. Our approach is exploratory and correlational; any patterns observed should not be interpreted as strict causal effects due to the limited sample and interdisciplinary data context.

2. Literature Review

2.1. Conceptualizing Deglobalization

Deglobalization refers broadly to a reversal or deceleration of global integration in economic, political, and social domains. While often understood as declining trade flows or financial interconnectedness, the concept encompasses a broader transformation that includes shifts in public discourse, nationalist policy agendas, technological fragmentation, and cultural retrenchment. Scholars have debated whether deglobalization constitutes a cyclical correction, a structural rupture, or a rhetorical phenomenon [13,14,15,16,17,18]. Rodrik [14] emphasizes the tension between global markets and national democratic governance, suggesting that political backlash against globalization’s domestic consequences drives regulatory divergence and protectionism. Meanwhile, Stiglitz [6] argues that disenchantment with neoliberal globalization—particularly its failure to deliver equitable gains—has fueled nationalist sentiment and eroded public trust in multilateral institutions.

The literature typically distinguishes between economic and political deglobalization, though these are increasingly interconnected. Economic deglobalization includes reduced cross-border capital flows, onshoring of production, and trade protectionism, while political deglobalization involves state withdrawal from global governance, increased nationalism, and growing geopolitical fragmentation [15]. A third dimension—discursive deglobalization—has gained attention recently, emphasizing how political elites, media, and publics frame globalization in more negative, contested, or securitized terms [16].

This study builds on that discursive turn by analyzing how mentions of nationalism, protectionism, protests, and polarization evolve alongside structural indicators such as FDI and global mobility. It treats deglobalization not as a binary state but as a multidimensional process detectable in behavior and language. We also explore how digital authoritarianism and cyber sovereignty emerge as symptoms and accelerants of deglobalization, particularly as states seek to insulate their information spaces from perceived foreign influence.

By treating discourse, protest, and policy as co-evolving features of deglobalization, our approach extends traditional frameworks and adds a communicative layer to economic and political models. This enables a more comprehensive view of how global connectivity erodes, not only through declining trade volumes or retreating multilateralism, but through transformations in how nations talk about and act upon the idea of being globally embedded. These transformations also signal a potential weakening of the global core–periphery hierarchy described by World Systems Theory [19], where shifts in communicative centrality and regional realignments challenge the long-standing dominance of core nations in both material and symbolic global networks.

2.2. Trade Protectionism as an Indicator of Deglobalization

Trade protectionism—encompassing tariffs, quotas, export restrictions, and domestic subsidies—is a salient and quantifiable indicator of deglobalization [6,20,21]. From an economic standpoint, classical liberal theories hold that open trade maximizes efficiency and mutual gains. At the same time, mercantilist or nationalist perspectives argue for shielding domestic producers and industries from foreign competition [4]. In practical terms, measuring the prevalence and severity of new or ongoing protectionist measures provides a useful empirical proxy for how seriously a nation diverges from global trade norms.

Trade protectionism directly impacts international trade volume and flow, a core aspect of economic globalization [22]. Protectionism is the opposite of a “free trade” policy. By imposing barriers to trade, countries reduce their economic interdependence with other nations, resulting in a contraction of global trade networks. This reduction in cross-border trade is a hallmark of deglobalization, as it signifies a retreat from the previously integrated global economy.

Trade protectionism is often rooted in nationalist and populist ideologies prioritizing domestic interests over global cooperation [21]. These ideologies advocate for reducing reliance on international markets and enhancing national sovereignty, which are central themes of deglobalization. Therefore, the rise in such policies indicates a shift towards economic deglobalization.

According to World Systems Theory [19], global economic integration is characterized by the dominance of core countries over peripheral ones through trade and economic exchanges. This network structure promotes globalization, with core countries serving as the primary drivers of trade. Trade protectionism disrupts this integration by creating barriers between nations, dismantling the core–periphery relationships that underpin globalization. Consequently, trade protectionism is a significant indicator of deglobalization within this theoretical framework.

Historical and contemporary data illustrate that periods of increased trade protectionism coincide with phases of deglobalization. For instance, the Great Depression of the 1930s [21] and the 2008 financial crisis [6] both witnessed a surge in protectionist measures, resulting in significant declines in global trade. These empirical patterns reinforce the validity of using trade protectionism as an indicator of deglobalization.

Trade protectionism provides a quantifiable and observable measure of deglobalization. Data on tariffs, quotas, and subsidies are systematically collected and reported by various international organizations, such as the World Trade Organization (WTO) and Global Trade Alert. This data availability allows for rigorous empirical analysis and longitudinal studies to track deglobalization trends.

Empirically, the Global Trade Alert database offers a detailed record of trade interventions flagged as “harmful” or “restrictive”. These range from sector-specific tariffs on steel or agricultural goods to broader policies shaping supply chain decisions [20]. Correlating such data with broader socio-communicative variables is relatively uncharted territory, underscoring the need for interdisciplinary study, precisely what this paper aims to address.

Historical Patterns of Globalization and Deglobalization

Trade openness stands in contrast to trade restrictions, and examining trade openness in the literature helps contextualize the measures of deglobalization that involve trade restrictions. Figure 1 illustrates the pattern of trade openness from 1870 to 2017, based on the sum of the world’s imports and exports divided by global GDP. However, we use a different measure based on trade restrictions. Politically, trade openness reflects the rise in liberal policies favoring global over national interests [23,24,25,26]. Nationalist populist political ideologies, often emerging from right-wing orientations, typically oppose liberal trade policies, instead imposing tariffs and domestic subsidies. Examples of deglobalizing political positions include Trump’s “America First” and Orban’s “Europe First” focus [27].

Following World War II, there was a shift toward increased international trade facilitated by the establishment of the General Agreement on Tariffs and Trade (GATT) in 1947 and the subsequent formation of the World Trade Organization (WTO) in 1995 [3]. Many countries reduced trade barriers and tariffs during this period, significantly increasing global trade and economic growth.

The 2008–2009 financial crisis reversed the trend with a wave of protectionism [6,21]. Deglobalization occurred as the growth in global trade relative to global GDP growth declined [9], further declining in 2018 with the trade war initiated by Trump with China [5,28], which also spread protectionism in many countries. This trend was further exacerbated by the COVID-19 pandemic, with some countries implementing export restrictions on medical supplies and other essential goods, including food [29].

In addition to protectionist policies, there has been a rise in nationalist movements and political leaders prioritizing national sovereignty over globalism, most notably with Trump’s election in 2016 [6]. These populist movements often focus on immigration and trade issues, promoting policies prioritizing their citizens’ interests over those of other countries. Increased nationalism has eroded the global nature of the internet as countries seek greater control over their digital territories [11]. This cyber sovereignty, a term promoted by China [30], is a form of digital protectionism [31]. Data protectionism has also increased [32].

Another feature of recent deglobalization has been a retreat from multilateralism to disembedded unilateralism [33], with some countries opting to withdraw from international organizations or negotiate bilateral trade agreements rather than multilateral ones. This trend toward unilateralism has been driven by a perception that multilateral institutions are ineffective or biased against certain countries. Populism and nationalism encourage these perceptions and unilateral policies. Reacting to this trend, in 2023, the World Trade Organization advocated for increased multilateralism, based on simulations of economic effects [34], in line with its mission to promote, not just study, world trade. Deglobalization has increased geopolitical tensions as countries fiercely compete for resources, influence, and strategic advantage [16]. This heightened competitiveness has increased economic, diplomatic, and kinetic conflicts in many parts of the world, including Africa and Ukraine. In Asia, diplomatic disputes have been prevalent in the East China Sea [35], the South China Sea, and China’s relationships with Taiwan. More broadly, protests and riots in many countries have marked social unrest, and governments have responded with digital authoritarianism measures [36].

Although the economic impacts of trade deglobalization have been extensively studied, research is needed on its relationships to communication events and processes, such as the move to nationalize the internet [11], the polarization of social media [37,38], social unrest [39], and digital authoritarianism [40]. We seek evidence that since 2008, the use of English has declined, while other languages have increased, accompanied by political polarization, protests, and digital authoritarianism. Communication theory offers a possible explanation.

2.3. Communication Outcomes of Deglobalization

Deglobalization is not solely an economic phenomenon; it reverberates through cultural, political, and technological domains. Specifically, we consider the factors below.

2.3.1. Deglobalization Discourse

Public and media commentary on “deglobalization” can shape how populations interpret shifts in trade policy. Carothers and O’Donohue [10] note that narratives about economic sovereignty, foreign competition, and national identity can become self-reinforcing if widely adopted. Tracking the usage of “deglobalization” across diverse media platforms provides insight into whether it is a fringe concept or a widely recognized trend.

2.3.2. Language Entropy

One underexamined effect of deglobalization is on language usage. Globalization has often been associated with the rising dominance of English across business, academia, and online communication [41]. Deglobalization, conversely, may create space for local languages to reassert themselves [42]. Such a shift is captured by the metric of language entropy, based on Shannon’s definition of entropy in information theory [43], which quantifies linguistic diversity. High entropy indicates that language use is more evenly distributed across multiple languages, while low entropy signals dominance by one or a few languages. If cross-border exchange diminishes, we might expect a lessened incentive to use a global lingua franca (English) and a relatively greater use of diverse local languages.

2.3.3. Political Polarization

Economic uncertainties or dislocations associated with protectionist policies can widen ideological rifts [44]. Populist leaders often capitalize on trade grievances to mobilize support, framing globalism as detrimental to national sovereignty [20]. This can result in affective polarization, where individuals become more antagonistic towards opposing political camps and loyal to their in-group [45]. Elevated political polarization can, in turn, hamper democratic governance [10].

2.3.4. Protests

Relative Deprivation Theory posits that abrupt changes in economic fortunes and inequalities—possibly triggered by protectionist measures—can lead to dissatisfaction, group mobilization, and, ultimately, protest [46]. Additionally, social strain theory suggests that when structural pressures accumulate, individuals tend to resort to collective actions to express their grievances [47]. The global wave of protests over the past decade—from the Arab Spring to widespread demonstrations in Asia, Europe, and the Americas—has sometimes intersected with economic frustrations stemming from tariffs, commodity prices, or trade disruptions [48,49].

2.3.5. Digital Authoritarianism

Finally, governments may respond to internal unrest or polarization—fueled partly by trade disruptions—by tightening control over digital spaces [10]. Digital authoritarianism includes tactics such as internet shutdowns, censorship of online platforms, and surveillance of dissidents [11,40]. In regions where populist or nationalist leaders hold sway, the impetus to manage the domestic conversation may intensify, particularly if protectionist policy changes spark a backlash [32]. By measuring references to these practices across multiple data sources, we can assess the extent to which trade protectionism might track with rising digital repression.

2.4. Theoretical Underpinnings

The literature on globalization and deglobalization has covered these economic and political impacts extensively. Despite the abundance of studies on the economic dimensions of deglobalization, few studies have examined how this phenomenon is experienced socially and culturally [49]. The communication-related consequences, particularly in the context of trade protectionism, remain underexplored. Previous studies have suggested that economic downturns and trade restrictions can increase political polarization and social unrest [8,41]. However, the specific mechanisms through which these economic policies impact communication patterns, protests, and digital authoritarianism are not well understood. This study aims to fill this gap by examining how trade protectionism affects these variables, employing a novel approach grounded in Optimal Information Theory [50], which offers the most parsimonious explanation consistent with the concept of countries as Information Societies [51,52]. First, we examine World Systems Theory and the idea of globalization as the foundation for the deglobalization that has emerged since its inception.

2.4.1. World Systems Theory

World Systems Theory, developed by Immanuel Wallerstein in the 1970s, provides a foundational framework for understanding global inequalities, power dynamics, and the structural organization of the capitalist world economy [19]. The theory divides the world into a hierarchical system of core, semi-peripheral, and peripheral states. Core nations dominate in terms of economic power, technological advancement, and control over trade and communication flows, while peripheral states are structurally disadvantaged, often supplying raw materials and labor.

Barnett [53] extended World Systems Theory into the study of international communication, examining transformations in global telecommunication flows as indicators of systemic change. His analysis of inter-country telecommunication traffic from 1978 to 1996 found that globalization increased the centrality of core countries within global networks [54]. Similar patterns were observed in the distribution of internet hyperlinks and domain name registrations, where core nations again assumed dominant positions [55].

However, under conditions of deglobalization, this structure appears to fragment. As global linkages weaken, communication becomes more regionalized and multipolar. Danowski [56] identified such regionalization in studies of Arab information ecologies, and Choi and Danowski [57] found similar patterns in Google group discussions organized around national identities. These trends indicate a reconfiguration of the global communication structure, wherein peripheral and semi-peripheral entities increase internal and regional exchanges [58]. This shift aligns with a discursive form of deglobalization and reflects the breakdown of an integrated global core–periphery structure.

Barnett and colleagues further emphasized the evolving structure of global communication and education networks. For example, Barnett et al. [59] noted the dominance of the U.S., Canada, and Australia as destinations for international students, reflecting a core–periphery dynamic in educational mobility. Moon et al. [60] similarly showed that the U.S. and U.K. dominate global music flows. These cultural and communicative asymmetries exemplify core nations’ symbolic and informational power within the world system. Deutschmann [58] also explored the increasing complexity of cross-border communication and mobility, underscoring the role of information flows in maintaining or disrupting global hierarchies.

In sum, World Systems Theory, with modifications, remains a useful lens for interpreting the effects of deglobalization on global communication networks. Periods of global integration consolidate core states’ dominance in material and symbolic information flows. In contrast, deglobalization weakens these linkages and fosters regional realignments that challenge the systemic stability of the core–periphery structure. As communication flows decentralize, World Systems Theory helps illuminate how discursive and infrastructural fragmentation may reconfigure global power in a multipolar direction.

2.4.2. Optimal Information Theory

Optimal Information Theory (OIT) offers a systems-level framework [61] for understanding how communication networks adapt to external uncertainty. Building on Ashby’s Law of Requisite Variety [62], which states that a system’s internal complexity must match that of its environment to remain stable [63], OIT proposes that rising environmental uncertainty prompts systems to increase the density and interconnectedness of their internal communication structures [42,63,64,65,66,67,68]. These denser networks allow systems to reduce ambiguity, enhance meaning-making, and maintain coherence in the face of complex or volatile inputs.

Danowski’s empirical studies, drawing on organizational data from institutions such as the Department of Defense Office of Civil Preparedness, Chase Manhattan Bank, and Korean villages, found that groups exhibiting over 50% internal ties were better equipped to navigate uncertain environments [64]. These findings resonate with Beniger’s [51] work on the control revolution, which highlighted the central role of information processing in managing societal complexity. Similarly, Castells [2] emphasizes how the architecture of communication networks influences power and cohesion in a globalized world.

OIT also warns of dysfunction when systems are overwhelmed by excessive information (overload) or deprived of it (underload), both of which can induce stress and disorganization [64]. Danowski and Edison-Swift [68] found that under high uncertainty, individuals respond by forming new communicative ties to re-establish clarity. Weick adopted these concepts in articulating sense-making in organizations [69]. This adaptive behavior scales to national systems: under conditions of globalization, the influx of diverse external inputs encourages societies to construct robust, multi-nodal internal networks, granting those who manage or curate information greater influence over public discourse [70].

However, under deglobalization—when external inputs recede—systems face reduced external signals and must compensate by generating their own. This process of internal signal amplification can increase affective polarization (supporting H3), foster perceptions of threat and grievance that motivate protest (H4), and incentivize governments to enact digital authoritarian controls to contain disruptive feedback (H5) [71]. In this sense, internal compression under deglobalization is not merely a coping mechanism but a driver of systemic tension. Conversely, shocks such as regime transitions or trade liberalization may restore external informational flows, decreasing the need for internal compensation and enabling greater stability.

By connecting internal communication patterns with national-level political behaviors, OIT provides a theoretical bridge between global informational structures and domestic sociopolitical dynamics. This study aligns with its broader effort to explain how macro-level processes, such as deglobalization, translate into communication, protest, and governance shifts across societies.

2.4.3. Theorizing the Effects of Trade Protectionism on Social Unrest and Polarization

Trade protectionism can provoke social unrest and deepen polarization through multiple theoretical mechanisms. According to Relative Deprivation Theory [46], when trade barriers lead to higher prices, reduced product availability, and job losses, a gap emerges between what people expect and experience, fueling grievances and unrest. Social strain theory [47,72,73,74,75] further suggests that economic disruptions strain social structures, increasing the likelihood of protests as workers and consumers face instability.

In addition, resource mobilization theory posits that the economic hardships resulting from trade restrictions empower affected groups to pool resources and organize opposition, thereby intensifying social mobilization [76,77]. Similarly, Political Opportunity Structure Theory posits that periods of political vulnerability create openings for groups to challenge the status quo, thereby catalyzing protest and dissent [78,79]. Protest Cycle Theory explains how initial economic discontent can diffuse across regions, creating unrest-related waves [79].

The backlash against globalization also plays a role. Theories on Globalization Backlash [80,81,82] suggest that trade restrictions, although intended to mitigate negative global impacts such as job losses and inequality, may exacerbate these issues, inciting further social discontent. Identity and Social Cohesion Theories note that deglobalization can erode social cohesion by intensifying divisions between those who benefit from protectionism and those who do not, thereby heightening conflict [83,84,85].

These dynamics contribute to rising political polarization. Research indicates that polarization is increasing globally [10,44,45], resulting in legislative gridlock and a decline in trust in government [86,87]. Media environments, shaped by Media Framing and Agenda-Setting Theories [88,89,90], reinforce these divides by selectively presenting trade issues in polarized terms.

Finally, theories such as Economic Self-Interest [91,92], Social Identity [93,94,95,96], Cultural Backlash [97,98], Relative Deprivation [99,100], and System Justification [101] provide additional lenses through which to view how trade restrictions polarize economic interests and group identities, ultimately deepening societal divisions.

Together, these frameworks illustrate the multifaceted impact of trade protectionism on social stability, as it incites unrest and intensifies political polarization.

2.4.4. Theoretical Explanations for Trade Restriction Effects on Digital Authoritarianism

Digital authoritarianism refers to state practices restricting online freedoms through censorship, surveillance, and network disruptions—a concept initially illustrated by China’s Great Firewall and later exemplified by Russia’s online monitoring [11]. Freedom House [12] tracks these practices via indices that include blocking social media, restricting politically sensitive content, manipulating online discourse, and even targeting dissenters through arrests or technical attacks. Recent reports highlight a persistent decline in global internet freedom, with multiple shutdowns occurring worldwide and an increasing trend toward digital controls as states seek cyber sovereignty [12,31,40].

Theoretically, trade restrictions can fuel digital authoritarianism through several interrelated mechanisms. Economic crises triggered by protectionist policies may prompt governments to intensify digital surveillance to prevent social unrest (Economic Crisis and State Control Theory). Similarly, by creating economic uncertainties, trade barriers lead states to control digital information flows to manage public perception, as Resource Dependence Theory explains. As social cohesion weakens under financial strain, governments may use digital authoritarian measures to maintain order (Social Control Theory) and exploit political vulnerabilities to limit opposition (Political Opportunity Structure Theory). Furthermore, to preserve legitimacy amid economic downturns, regimes might deploy digital censorship to shape narratives and secure public trust (Legitimacy and Performance Theory). Finally, technological determinism suggests that advances in surveillance technology empower states to monitor and repress online dissent more effectively, consolidating power through digital means [102,103].

Together, these frameworks illustrate the multifaceted impact of trade protectionism on social stability, as it incites unrest and intensifies political polarization.

2.4.5. Deglobalization and Language

Globalization promotes a lingua franca—primarily English—which streamlines international communication but often diminishes the use of local languages. In contrast, deglobalization tends to revive indigenous languages as nations assert sovereignty and nationalism, thereby altering political dynamics and contributing to polarization, protests, and even digital authoritarian responses. This shift can be quantified using the concept of language entropy—a measure from information theory that gauges the diversity and distribution of languages. Higher entropy indicates a more balanced use of multiple languages (reflecting deglobalization), whereas lower entropy signals dominance by one language (as seen under globalization).

3. Hypotheses

Based on the above synthesis, we derive five interrelated hypotheses (H1–H5), each linking trade protectionism (our key independent variable) to a specific communication or social outcome based on the theories shown in

Table 1. Hypothesis–theory mapping.

Hypothesis	Theoretical Framework(s)
H1	Media Framing Theory; Discourse Theory
H2	Cultural Globalization Theory; Optimal Information Theory (network density)
H3	Affective Polarization; Optimal Information Theory (internal compression)
H4	Relative Deprivation Theory; Resource Mobilization; OIT (signal generation)
H5	Political Opportunity Structure; OIT (information control; regime response)

.

We emphasize that these hypotheses anticipate associations rather than definitive causal effects:

H1: Trade protectionism is positively associated with the discourse of deglobalization.

• Rationale: As nations adopt tariffs and other protectionist measures, public and media discourse shifts to concepts such as “deglobalization”, highlighting the move away from globally integrated markets.

H2: Trade protectionism is positively associated with language entropy.

• Rationale: Constraints on cross-border exchange reduce reliance on a common global language, potentially allowing local or regional languages to proliferate. Higher protectionism may thus correspond to more linguistically diverse communication environments.

H3: Trade protectionism is positively associated with political polarization.

• Rationale: Domestic politics may harden into “for or against” stances on globalization, triggering or exacerbating ideological extremes. We expect periods of high protectionism to coincide with intensified polarization discourse.

H4: Trade protectionism is positively associated with protests.

• Rationale: Protectionist policies can disrupt economies, engendering public dissatisfaction. This may lead to more frequent protests or collective actions as citizens express grievances.

H5: Trade protectionism is positively associated with digital authoritarianism.

• Rationale: Governments anticipating unrest (partly due to economic strains from protectionism) might expand internet censorship and surveillance to pre-empt dissent. Thus, higher protectionism may be associated with more frequent mentions of digital repression tactics.

Each hypothesis effectively connects trade protection, our key independent variable, with a unique communication or social outcome. While correlation does not establish causation, finding consistent patterns across these domains would support the theoretical arguments that economic deglobalization is intertwined with broader societal changes. We will tentatively interpret any such patterns, acknowledging alternative explanations (e.g., a third factor, such as global crises, could drive protectionism and social unrest).

4. Methodology

This section outlines the methodological strategy used to examine the relationships between deglobalization discourse and its social, political, and economic correlates. Drawing on the theoretical frameworks introduced in Section 2 and Section 3—including World Systems Theory, Optimal Information Theory, and discourse-centered approaches—we operationalize deglobalization not only as a structural transformation in global trade and investment but also as a shift in how nations communicate about sovereignty, risk, and identity. The indicators selected—ranging from protectionist policies to language entropy and digital authoritarianism—reflect behavioral outcomes and discursive trends corresponding to the fracturing of global systems described in World Systems Theory. We use content analysis, correlation matrices, and principal component analysis (PCA) of cross-national and cross-temporal data sources from 2008 to 2022 to explore these dynamics.

4.1. Data and Sampling

4.1.1. Time Frame and Justification

We focus on the period from 2009 to 2023, capturing the post-financial crisis era, often marked by surging economic nationalism. Annual observations (i.e., each data point represents a year) align with how trade policies are typically enacted or adjusted yearly [6]. For example, governments normally introduce budgets or trade measures annually, and global indices, such as the number of protectionist measures, are often reported annually. This time frame also corresponds to data availability constraints; notably, Facebook’s CrowdTangle platform, used for social media data, provides accessible archives from 2008–2009.

The resulting sample consists of 14 observations (years), which is small for statistical analysis. Each observation, however, aggregates very large datasets, such as millions of Facebook posts and thousands of news articles per year, as well as comprehensive trade policy records. We thus treat the analysis as exploratory, leveraging the richness of cross-domain data while recognizing the limited statistical power. The interdisciplinary nature of the data, which combines economic indicators with media and social media metrics, introduces both breadth and complexity. This approach requires careful methodological choices (discussed below) and cautious interpretation of results, as combining different domains may violate assumptions or introduce unique biases.

4.1.2. Key Data Sources

To capture both economic and communicative aspects of deglobalization, we drew from several sources, each providing a different lens:

• Global Trade Alert (GTA): This is an independent database that tracks worldwide trade interventions. From GTA, we extracted two metrics: (a) Inceptions, the count of newly introduced restrictive measures each year (e.g., new tariffs enacted that year), and (b) Protections in Force, the total stock of active protectionist measures in effect each year (cumulative count up to that year). These serve as quantitative proxies for the extent of trade protectionism (and by extension, economic deglobalization) at a global level. Units: Both are simple counts of measures; Inceptions reset each year, whereas Protections in Force accumulate over time.
• Facebook (CrowdTangle): We used CrowdTangle, a tool that indexes public posts from Facebook pages and groups. Through keyword queries, we gathered two types of information from Facebook posts: (a) language usage data for computing language entropy (by searching broadly for posts containing the term “communication” and retrieving the language each post is written in), and (b) mention counts for specific terms like “political polarization”, “protest”, “digital authoritarianism”, etc. (by querying posts containing those phrases). Units: For language, we obtained counts of posts per language per year. For mentions, we obtained annual frequencies of posts containing each target term. These are raw counts, which we later log-transformed for analysis.
• Google Search Data: Using Google’s search engine, we logged the approximate number of results (hits) each year for keywords such as “deglobalization”, “political polarization”, “digital authoritarianism”, “protest”, etc. This provides a broad measure of how often these terms appear on the web (across websites, not just news or social media). Units: The raw numbers represent Google’s reported hit counts, which are rough estimates of indexed pages containing the term. We collected these data for each year (e.g., by using time-bound search filters when possible or by recording annual values if available from Google Trends or similar sources). Because Google’s reporting can be inconsistent, we treat these as indicative volume measures and apply log transformations for analysis.
• Nexis Uni (News Articles): Nexis Uni aggregates content from major newspapers worldwide, including dedicated tracking of key sources such as The New York Times. From Nexis Uni, we retrieved (a) the annual count of news articles that mention our keywords (“deglobalization”, “political polarization”, etc.), and (b) full-text corpora of those articles for qualitative context (though our quantitative analysis primarily uses the counts). We specifically included English-language major newspapers across different regions to capture mainstream media discourse. Units: The counts represent the number of articles per year that contain the term across a defined set of “major newspaper” sources. In some cases, we separated The New York Times (NYT) as a benchmark U.S. outlet from other global sources, known as “Major Newspapers”, to determine whether trends were global or driven by specific media.
• Global Internet Activity (Google Page Count): As a control variable, we obtained an estimate of the total number of webpages indexed by Google each year, serving as a proxy for overall internet content growth. This serves as a covariate to account for the fact that mentions of anything tend to increase over time simply because more content is put online each year. Units: We used global estimates from Google statements and independent research sources. In our dataset, this is represented as the annual Google Pages (total) count, which we log-transformed. For example, if Google indexed ~50 billion pages in 2010 and ~200 billion in 2020 (hypothetical figures), this growth could inflate all search-based counts; controlling for it helps isolate more specific trends.

In summary, our dataset encompasses multiple domains, including official trade measures, social media content, web presence, and news media coverage. This interdisciplinarity is a strength, as it allows cross-validation of trends (e.g., if both news and social media show a rise in “deglobalization” talk, confidence in the trend increases). However, it also introduces methodological challenges: each source has different biases and units. We mitigated some issues by standardizing through transformations (next section) and combining related metrics via PCA.

4.1.3. Justifying Mentions as Proxies for Actual Behaviors

In this study, digital mentions—references to concepts such as “protests”, “political polarization”, or “deglobalization”—are used as proxies for corresponding real-world behaviors. This approach is grounded in both empirical pragmatism and theoretical insight into the role of discourse in shaping social reality.

• Mentions as Reflective Signals of Behavior

Mentions often reflect changes in actual behavior. When protest activity increases, nationalist sentiment rises, or trade protectionism expands, these shifts tend to generate corresponding increases in media and public discourse. This relationship stems from the following facts:

• News media and digital platforms respond to observable changes in society, covering them in real time;
• Publics engage with issues that are salient or experienced directly, such as attending a protest, reacting to new tariffs, or navigating border restrictions;
• Empirical studies confirm strong correlations between digital discourse and on-the-ground events (e.g., correlations between Google Trends data and consumer behavior or between Twitter posts and protest turnout).

Therefore, an increase in mentions is likely to coincide with, and be partially driven by, real shifts in the behavior or condition being referenced.

2.

Mentions as Constitutive Forces in Social Dynamics

Equally important, mentions do not merely reflect social behaviors; they also constitute them. A growing body of studies supports this view in the literature in communication, sociology, and political science that recognize the performative and agenda-setting power of discourse:

• Framing theory shows how issues that are discussed shape how they are understood and acted upon.
• Discourse analysis highlights that repeated mention of particular narratives (e.g., “deglobalization”, “national sovereignty”, and “foreign threat”) can create shared mental models that influence perception and guide behavior.
• Mentions in digital networks contribute to the viral spread of ideas, norm activation, and even mobilization, such as through hashtags that help coordinate protest activity or signal ideological alignment.

In this sense, mentions are not passive indicators; they are active elements of the sociopolitical process, reinforcing sentiments, triggering emotions, and shaping expectations. They contribute to the affective climate and cognitive map within which behaviors unfold.

3.

Strategic Utility and Theoretical Fit

Given the interdisciplinary framework of the study—spanning political communication, institutional theory, and globalization studies—using mentions as indicators aligns with both of the following:

• The Optimal Information Theory lens views increases in communicative activity as responses to rising uncertainty and as mechanisms of system adaptation.
• The discursive institutionalist perspective treats language, narrative, and symbolism as central to institutional change and political action.

Moreover, in contexts where official data may be sparse, delayed, or selectively reported (as in protests, soft power dynamics, or discursive shifts), mentions offer a timely and scalable alternative measure of social change.

In sum, mentions are both responsive to behavior and constitutive of it. By capturing what is salient, contested, and circulated in public discourse, they offer a valid and meaningful proxy for social behaviors in the digital age, especially when triangulated with broader theoretical and empirical insights. This dual role justifies their use in exploratory, interdisciplinary analyses of macro-level dynamics such as deglobalization and ideological polarization.

4.2. Variables

We constructed the following variables for analysis:

• Trade Protectionism: Inceptions (annual count of new protectionist measures) and Protections in Force (cumulative count of active measures). These two indicators capture different dynamics: Inceptions highlight yearly surges or dips in protectionism, while Protections in Force reflect the accumulated trade barriers at a given time. In our tables, we sometimes denote Inceptions as PROTECT (short for protectionism in that year) and Protections in Force as PROTECTIONS (plural, indicating total ongoing measures). Both are measured as raw counts (continuous, potentially large numbers), which we later log-transform due to high skew.
• Deglobalization Discourse: To quantify the prominence of “deglobalization” in public conversation, we aggregated mention volumes from multiple sources: Google hits, Nexis Uni articles, and Facebook posts referencing “deglobalization”. Because these measures are on different scales and exhibit highly collinear trends, we employed principal component analysis (PCA) to combine them into a single composite index. Specifically, we obtained the annual counts from DEGLOBAL_GOOGLE, DEGLOBAL_NEWS (major newspapers), DEGLOBAL_NYT (specifically, the NYT), and DEGLOBAL_FACEBOOK (FB posts), each of which was log-transformed. PCA yielded a first principal component that explained the majority of variance across these four indicators (we report it as DEGLOBAL_FACTOR). This factor (a continuous variable, with a mean of zero by construction) represents the overall intensity of deglobalization discourse per year. Units: It is unitless (standardized score), but higher values indicate years with collectively higher “deglobalization” mentions across all sources. (For reference, PCA factor loadings were all > 0.85 on this component, indicating strong contributions from each source.)
• Language Entropy: To measure linguistic diversity, we computed the Shannon entropy (H) of language use in Facebook posts about “communication”. Using the language detected for each post (via FastText’s language ID for 157 possible languages), we calculated the following Equation (1):

  $$H=-{\sum }_{i=1}^N{p}_i{\log }_2{p}_i$$

  (1)

  where p_i is the proportion of posts in language i for a given year, and N is the number of distinct languages observed that year (bounded by 157) (units: bits (since log base 2)). Higher ENTROPY means a more even spread among languages (no single language dominates), whereas lower entropy means one or few languages account for most content. For example, if, in 2009, English accounted for 80% of posts and other languages made up the rest, entropy might be low (~1 bit); if, by 2020, English was 60% and many others share the remainder, entropy rises (~2 bits or more), indicating a more balanced distribution. This variable captures the concept of shifts in English versus local language usage—an indirect proxy for cultural deglobalization.
• Political Polarization: We measured this by the frequency of the term “political polarization” (and its close variants) in our sources. We summed logged counts from Google, Nexis Uni (news), and CrowdTangle (Facebook) for mentions of “political polarization” per year. We treat POLARIZATION as a composite count (in practice, because these sources correlated strongly, one could also use any single source; we opted to sum and then log) (units: effectively, the log of the count of mentions per year). This reflects the prominence of polarization in discourse, which we interpret as a proxy for the actual intensity of polarization in society, acknowledging its imperfection.
• Protests: Similar to polarization, we tracked the frequency of “protest” or “protests” in Google results, news articles, and Facebook posts annually. PROTESTS is our aggregated, logged count of protest-related mentions (units: log count per year of mentions). This highlights the salience of protests in public communication, which often aligns with actual protest events; spikes in media discussion of protests typically coincide with real protest waves, even though not every mention corresponds to an actual protest.
• Digital Authoritarianism: We developed an index for digital repression narratives, utilizing terms such as “digital authoritarianism”, “internet shutdown”, “online censorship”, and “social media ban”. Specifically, we counted mentions of a set of related keywords across Google, news, and Facebook. These counts were combined (summed after log transformation for each term) to form an overall DIGITAL_AUTH variable (units: log count of mentions per year (across sources, multiple terms)). We included multiple terms to capture various facets of the concept; for instance, some articles may not use the term “digital authoritarianism” but instead report that “the government blocked Facebook”. Our composite aims to reflect the general level of discourse about governments controlling the internet or digital repression.
• Control—Global Content Growth: GOOGLE_PAGES is our control variable, representing the annual log of Google-indexed pages. By including this, we attempt to account for the secular growth of content and connectivity over time. Without this control, correlations among our variables might be inflated by the fact that all variables increased over 2009–2023 (e.g., more trade measures, more internet content, and more mentions of any word, simply due to global growth). This variable has a mean of approximately 0 and a small variance after log transformation and normalization (since it was centered) and is used in partial correlation analyses.

All variables above (except the control) align with our hypotheses H1–H5, respectively: deglobalization factor (H1), entropy (H2), polarization (H3), protests (H4), digital auth. (H5). To ensure clarity,

Table 2. Descriptive statistics.

Variables	N	Min	Max	Mean	Std. Dev.
POLARIZATION (log hits)	14	8.86	14.33	10.76	1.66
DIGITAL_AUTH (log hits)	14	10.22	17.09	13.11	2.83
GOOGLE_PAGES (log, bil.)	13	−1.43	0.63	−0.14	0.74
PROTESTS (log hits)	14	11.41	18.77	14.4	2.3
DEGLOBAL_NYT (log count)	14	2.4	6.95	4.5	1.37
DEGLOBAL_FACEBK (log count)	14	0.69	7.77	4.1	2.32
DEGLOBAL_NEWS (log count)	14	2.2	6.87	4.3	1.37
DEGLOBAL_FACTOR (PCA score)	14	−0.59	2.84	0	1
ENTROPY (bits)	14	1.12	2.29	1.65	0.41
PROTECTIONS (count)	14	1609	25,066	12,794.21	7819.87

Note: POLARIZATION, DIGITAL_AUTH, and PROTESTS are aggregated logged mention counts (unitless after log). GOOGLE_PAGES is a log of the total indexed pages (approximately centered around 0). DEGLOBAL_NYT/FACEBK/NEWS are log counts of “deglobalization” mentions in those sources; DEGLOBAL_FACTOR is the PCA composite (mean 0; SD 1); ENTROPY is Shannon entropy in bits; and PROTECTIONS is the count of protectionist measures in force (not logged here for interpretability). N = 13 for Google Pages because 2023 data were unavailable (we used 2009–2021 for that series; this does not affect other variables, which had 2009–2022 or 2009–2023). Inceptions (new measures per year) are not listed separately here. However, in 2009, Inceptions = 1609 (the same as Protections’ minimum, since the cumulative start year equals that year’s new measures). This number rises to a peak of ~5000+ in some years, which can be inferred from annual increments in Protections.

(“Descriptive statistics”) presents each variable with its symbol as used in analysis and tables, alongside summary statistics.

4.3. Analytical Procedures

4.3.1. Log Transformations

Several variables in this study, including Google search volumes and media mention counts, exhibit exponential growth over time. To address this, we applied natural log (ln) transformations, which helped reduce positive skew, compress extreme values, and stabilize variance [104]. This is particularly important given our limited sample size, where outliers could otherwise exert an undue influence on correlation estimates. Log transformation is a standard practice for count-based variables in time-series and cross-sectional analysis, as it enables a clearer interpretation of proportional differences and aligns more closely with the assumptions of normality and linearity in statistical models.

4.3.2. Correlation and Assumption Checks

We first computed Pearson correlation coefficients among all variables of interest (

Table 3. Correlations.

Variables	Protect	Polar	DIGI_AUT	Protest	DEG_NYT	DEG_FB	DEG_NEWS	DEG_GOO	GEG_FAC	Pages	Entropy
PROTECTIONS	--
POLARIZATION	0.946 **	--
DIGITAL_AUTH	0.930 **	0.848 **	--
PROTESTS	0.952 **	0.882 **	0.901 **	--
DEGLOBAL_NYT	0.814 **	0.772 **	0.861 **	0.802 **	--
DEGLOBAL_FACEBOOK	0.970 **	0.880 **	0.913 **	0.942 **	0.832 **	--
DEGLOBAL_NEWS	0.804 **	0.769 **	0.848 **	0.788 **	0.998 **	0.818 **	--
DEGLOBAL_GOOGLE	0.937 **	0.871 **	0.935 **	0.926 **	0.910 **	0.929 **	0.897 **	--
DEGLOBAL_FACTOR	0.768 **	0.793 **	0.729 **	0.770 **	0.839 **	0.772 **	0.859 **	0.794 **	--
GOOGLE_PAGES	0.908 **	0.763 **	0.833 **	0.896 **	0.623 *	0.932 **	0.593 *	0.850 **	0.581 *	--
ENTROPY	0.916 **	0.834 **	0.893 **	0.832 **	0.809 **	0.861 **	0.803 **	0.896 **	0.704 **	0.778 **	--

Significance: p < 0.01 (1-tailed) marked by **; p < 0.05 marked by *. Note: “--” indicates the variable’s correlation with itself, which is 1.0 by definition or left blank. The cell for PROTECT–PROTECTIONS is not shown, but these two are highly correlated at >0.95 as expected and are indicated qualitatively as “(high)” since Protections is cumulative of Protect.

reports the full correlation matrix). Pearson’s r assumes a linear relationship and is sensitive to outliers and non-normality. As noted, we mitigated these issues through log transformations and by inspecting scatter plots. We found that relationships were monotonic and roughly linear, with many variables trending upward over time. We also considered Spearman rank correlations, which are robust against any remaining non-normality, yielding similar results (all high correlations remained high in rank form). Given the consistency and our interest in linear effect size, we report Pearson’s r. We acknowledge that N = 14 is low for stable correlation estimates; common trends might inflate an observed high r, while a moderate r could be unstable. We interpret correlations above ~0.7 as “strong” in this context and between 0.5 and 0.7 as “moderate”. However, due to low power, statistical significance (p-value) is used as a rough guide rather than a definitive test. All hypothesis tests for correlation use one-tailed p-values (since the directions are predicted) at α levels of 0.05 and 0.01; however, with so few data points, p-values should be interpreted cautiously.

Before finalizing the correlation results, we checked for any obvious violations, such as a pair of variables that might be non-independent. Arguably, our four “deglobalization mention” variables are not independent, so we combined them via PCA to reduce dimensionality and avoid redundancy. Similarly, “Inceptions” and “Protections in Force” are related (Protections is roughly the cumulative sum of Inceptions with some decay if measures expire). Indeed, Inceptions and Protections correlated highly (we observed r > 0.95 between them), so including both in a regression would be problematic. We largely analyze them separately or use one as primary and qualitatively mention the effects of the other. Multicollinearity is present among many predictors (all the discourse variables also strongly intercorrelate, as shown in Table 3), so we refrain from extensive multivariate modeling, such as multiple regression; the data do not support it. Instead, our analytic focus is on bivariate correlations and patterns.

4.3.3. Principal Component Analysis (PCA)

As mentioned under variables, we used PCA to create the deglobalization discourse factor. The procedure involved inputting the standardized (z-score) series of four indicators (Google hits, Nexis news count, NYT count, and Facebook count for “deglobalization” per year) and extracting the components. The first principal component had an eigenvalue greater than three and explained ~95% of the variance, reflecting that all four series exhibited a similar exponential increase pattern. Loadings were 0.94 (Google), 0.96 (news), 0.90 (NYT), 0.95 (Facebook)—all very high and positive. We retained only this first component (it effectively averages the standardized inputs) and interpret it as an overall index of deglobalization discourse intensity. This approach reduces measurement noise (e.g., if one source undercounts in a given year but others capture it, the PCA smooths out the discrepancy) and prevents overweighting of any single platform. We note that PCA assumes a linear combination and that the underlying concept (discourse intensity) is one-dimensional, which seems reasonable here given the high correlations among sources.

4.3.4. Partial Correlations

After examining raw correlations, we computed partial correlations controlling for the Google Pages variable (total web content growth). This was to ensure that observed relationships between trade protectionism and communication variables are not solely due to general time trends or the expansion of the internet. The partial correlation for variables X and Y controlling Z is essentially the correlation of the residuals from regressing X on Z and Y on Z, respectively. We report partial correlation coefficients in

Table 4. Partial correlations controlling for Google Pages.

Variables	Prot.	DEG_FAC	Polar	DIG_AUT	Protest
PROTECTIONISM	--
DEGLOBALIZATION FACTOR	0.730 **	--
POLARIZATION	0.901 **	0.493	--
DIGITAL_AUTH	0.689 **	0.61 *	0.519 *	--
PROTESTS	0.697 **	0.420	0.512 *	0.579 *	--
ENTROPY	0.782 **	0.703 **	0.594 *	0.669 **	0.467

** Correlation is significant at the 0.01 level (1-tailed). * Correlation is significant at the 0.05 level (1-tailed).

. Assumptions for partial correlation are similar to Pearson’s (linearity, etc.), and given the small N, we again interpret these with caution. In practice, when including the control, it did not change the direction of any correlations; it generally reduced their magnitude slightly (as expected if some trend is shared with Z). We ensured the control itself was not causing multicollinearity issues; indeed, Google Pages is highly correlated with year and many variables, but that is precisely why we control for it—to partially remove that common variance.

4.3.5. Robustness and Exploratory Checks

Given the limited data points, we conducted a few robustness checks to see if the results hold under slight variations:

• Year omissions: We omitted potential outlier years and recalculated correlations. Specifically, we excluded 2009 (the first year after the crisis, which may have had unstable data) and, separately, excluded 2020 (the year of COVID-19 onset, which saw unusual disruptions). In each case, we compared the correlation patterns with the full sample.
• Inceptions vs. Protections: We examined whether using Inceptions (new measures) vs. Protections in Force (ongoing measures) yields different strengths of association with outcomes. This addresses whether short-term spikes or cumulative pressure are more important.
• Alternate keywords: For the digital authoritarianism variable, we explored different combinations of terms (e.g., using only “internet shutdown” or only “censorship”) to determine if a particular term was driving the results. We found the trend to be robust across different terms; all showed upward movement and correlated with the rise in protectionism.
• Data transformations: We tested whether not logging the variables (or logging the trade measures) changed conclusions. Without logs, the correlations remained positive but were even more dominated by time trends (nearly all variables increased over time), yielding extremely high correlations (0.9+) across the board. That was not informative, so the logged results are preferred. Logging the trade measures also yielded similar qualitative findings (results were slightly strengthened for polarization when using logged protectionism; for consistency, we present the unlogged trade measure results in the main tables).
• Autocorrelation and time-lag exploration: Although 14 points are too few for serious time-series modeling, we performed a rudimentary check for autocorrelation. Year-to-year changes were not our primary focus, but all series exhibit strong autocorrelation, largely due to their trend. We considered first-differencing to see if year-over-year protectionism changes correlate with outcomes. Those correlations were weaker and often not significant, which is expected if effects are cumulative or if the signal-to-noise ratio in annual deltas is low. Thus, our main analysis focuses on the level values. Still, we acknowledge this limitation; more sophisticated time-series models (e.g., ARIMA or cross-correlation) would require longer series or more granular data (e.g., quarterly) to be effective.

Our interdisciplinary methodology integrates computational text analytics (for entropy and mentions) with quantitative social science analysis. We have taken steps to ensure clarity of variable definitions, appropriateness of statistical techniques (with normalizing transforms and checks), and caution in interpretation. All analyses were conducted using Python 3.11 and R 4.3.2; the code and data can be made available upon request for verification.

5. Results

5.1. Descriptive Statistics

Table 2 summarizes each variable’s range and central tendency. All variables cover the 14 years (2009–2022 or 2023, depending on data availability). The values for discourse-related variables (polarization, digital auth, protests, etc.) are provided in their logged forms as used in the analysis.

Trends: Broadly, by the 2010s, all discourse variables increased substantially (e.g., polarization mentions ranged from ~10^9 to ~10^14 hits in log terms, and protests ranged from ~10^11 to ~10^18). Deglobalization was barely mentioned in 2009 (log ~0.7 in Facebook and 2.4 in NYT, meaning practically a handful of mentions) but reached thousands of mentions by the late 2010s (log ~7). Language entropy roughly doubled from 1.12 to 2.29 bits, indicating a diversification of languages on the sampled Facebook posts. Protections in Force had a mean of ~12.8 k measures, with a large standard deviation, reflecting the sharp rise over time. Below, we describe patterns in each key area:

• Trade Protectionism: Inceptions ranged from about 1600 in 2009 to ~25,000 at their peak in the late 2010s (with an average of ~8000 per year if we calculated the mean of yearly introductions). There was a notable jump around 2016, paralleling events like the Brexit referendum and the U.S. presidential election, and then some fluctuation through 2023. Protections in Force rose consistently yearly, reflecting that most new measures remained in effect and accumulated. By the end of the period, there were over 25,000 active protectionist measures globally, up from just ~1600 in 2009, illustrating a significant shift toward trade restrictiveness.
• Deglobalization Discourse: Mentions of “deglobalization” were negligible in 2009–2010 (often fewer than 100 hits in mainstream media databases). Post-2015, the term’s usage increased exponentially, peaking around 2019–2020. This timing coincided with intense coverage of U.S.–China trade tensions and discussions of retreating globalization in policy circles. For instance, 2019 saw a surge in think-tank reports and news stories asking if globalization was in reverse. Our PCA-based factor captures this spike; notably, the 2019–2020 period has the highest factor scores (around 2–3 SDs above the mean).
• Language Entropy: In Figure 2, we track the percentage of English in Facebook “communication” posts, and in Figure 3, we track the entropy. English accounted for ~80% of such posts in 2009, but dropped to around 60% by 2020, indicating an increase in posts in other languages. Correspondingly, Shannon entropy rose from ~1.10 bits to ~2.20 bits over the period. An entropy of 1.1 bits suggests one language (English) heavily dominates with a few others making minor contributions; 2.2 bits suggests a more even mix (though still not uniform by any means; English was still the largest share, but smaller relative to before). This empirical trend supports the idea that global English dominance in online discourse slightly receded in the late 2010s, allowing more multilingual content, which is consistent with H2’s expectation under deglobalization.
• Political Polarization, Protests, and Digital Authoritarianism: All three showed marked increases in textual mentions starting in the mid-2010s, with prominent surges around 2016–2017 and again around 2019–2020. Key real-world referents include the 2016 U.S. election and Brexit (which amplified talk of polarization and also triggered protests in some places); the late 2010s rise in populist governments (some imposing digital controls and hence more talk of digital authoritarianism); and the 2019–2020 period which included global protest waves (e.g., climate strikes, Hong Kong protests, various protests in Latin America), plus the onset of COVID-19 (prompting both protests and emergency powers that sometimes curbed digital freedoms). According to our data, polarization discourse roughly doubled (in log terms) from the early 2010s to the late 2010s; protest mentions showed a similar jump. Digital authoritarianism-related terms were scarcely mentioned pre-2014 but were common in policy reports and news by 2020 (for example, the term “digital authoritarianism” gained currency around 2018, and many reports of internet shutdowns in 2019–2021 added to the count).

These descriptive observations set the stage: protectionism and our outcome variables generally trended upward post-2009. The next section examines the correlation between these trends using correlation analysis.

5.2. Correlations

Table 3 presents the zero-order Pearson correlations among the primary study variables, including trade protectionism measures, deglobalization discourse, language entropy, political polarization, protests, and digital authoritarianism. As shown, the correlations between trade protectionism and the communication-related outcomes are uniformly positive and, in most cases, strong to very strong.

For instance, trade protectionism (measured by Protections in Force) correlates highly with mentions of deglobalization (r ≈ 0.85–0.90), language entropy (r ≈ 0.80–0.90), protest discourse (r ≈ 0.90–0.95), and digital authoritarianism (r ≈ 0.90–0.93). The association with political polarization is somewhat more moderate but still positive and significant (r ≈ 0.65–0.70). These patterns are consistent with the expectations outlined in hypotheses H1 through H5.

In addition to strong correlations between protectionism and individual outcomes, the outcome variables (e.g., polarization, protests, and digital authoritarianism) also exhibit high intercorrelations. This suggests that these phenomena may have co-evolved during the study period, potentially reflecting broader systemic economic and social change shifts.

However, it is important to interpret these initial correlations with caution. Given the nature of the data—covering a period of significant global expansion in internet content and digital discourse—there is a possibility that the observed associations partly reflect shared time trends rather than substantive linkages. Many of the variables studied (including trade protectionism and digital discourse measures) exhibit overall upward trajectories from 2009 to 2023. Consequently, some of the strong correlations may be inflated by secular growth in global digital content rather than reflecting true underlying relationships.

To strengthen the validity of these associations and rule out confounding factors due to secular trends in internet content growth, we examine partial correlations controlling for overall internet expansion.

5.3. Partial Correlations Controlling for Internet Growth

To address the possibility that secular trends in global internet content growth inflated the observed zero-order correlations, we computed partial correlations controlling for the annual expansion in total Google-indexed pages (GOOGLE_PAGES). Table 4 reports these partial correlations between trade protectionism (Protections in Force) and the five main communication outcome variables.

As shown in Table 4, the associations between trade protectionism and each outcome variable—deglobalization discourse (partial r = 0.730), language entropy (partial r = 0.782), political polarization (partial r = 0.901), protests (partial r = 0.697), and digital authoritarianism (partial r = 0.689)—remain positive and statistically significant (p < 0.01 for most associations). Although the magnitudes of the partial correlations are somewhat lower than the corresponding zero-order correlations reported in Table 3, the direction and significance of the relationships are preserved, reinforcing the strength of the findings.

The partial correlation between trade protectionism and political polarization becomes even stronger (r = 0.901) after controlling for general internet growth, suggesting a particularly robust connection between protectionist economic environments and rising ideological polarization. While slight attenuation is observed for variables such as protests and digital authoritarianism, the relationships remain substantively meaningful and statistically reliable.

Overall, the partial correlation analysis confirms that the relationships observed are not merely artifacts of parallel growth in digital communication but reflect genuine linkages between rising protectionism and broader communicative and social transformations. The next section briefly summarizes additional robustness checks conducted to assess these results further.

5.4. Robustness Checks

We conducted several supplementary robustness checks to assess the stability of the observed relationships further. First, we recalculated correlations omitting potentially atypical years, such as 2009 (immediately following the global financial crisis) and 2020 (the onset of the COVID-19 pandemic). The exclusion of these years did not materially alter the pattern of associations: trade protectionism remained positively correlated with deglobalization discourse, language entropy, political polarization, protest activity, and digital authoritarianism.

Second, we compared results using both new protectionist measures introduced each year (“Inceptions”) and the cumulative stock of active protectionist measures (“Protections in Force”). While both indicators were positively associated with the outcome variables, the cumulative measure generally yielded stronger and more consistent correlations, suggesting that the entrenchment of protectionist policies over time has a more durable influence on communication and social dynamics than short-term fluctuations.

Third, we tested the sensitivity of the digital authoritarianism measure by varying the keyword combinations (e.g., using only “internet shutdowns” versus a broader set including “censorship” and “social media bans”). The main results remained stable across these variations, supporting the robustness of the findings.

Finally, we examined the effects of using logged versus unlogged forms of trade protectionism variables. Although logging variables compressed the distribution and slightly moderated some correlations, the overall pattern of positive associations persisted, with logged versions providing a more conservative estimate.

Together, these robustness checks confirm that the relationships observed between trade protectionism and the key communication and social variables are not sensitive to modest variations in modeling choices, sample years, or operational definitions. Having established that the primary associations are robust across multiple analytic approaches and not merely artifacts of temporal trends, we now summarize the degree of support for each research hypothesis (H1–H5).

5.5. Summary of Hypothesis Tests (H1–H5)

Having presented the results of both zero-order correlations and partial correlations controlling for overall internet content growth, and after conducting several robustness checks, we now summarize the degree of support for each of the five hypotheses (H1–H5). Our interpretation reflects both the simple associations and the more conservative controlled analyses.

Deglobalization Discourse (H1)

The correlation and partial correlation analyses (Table 3 and Table 4) show that Protections in Force correlates strongly with the deglobalization discourse factor (r ≈ 0.85–0.90, p < 0.01). This strong association remains even after controlling for global internet expansion, with only slight attenuation. Thus, periods of expanded trade restrictions coincided with intensified public discourse on “deglobalization”. While causality cannot be asserted, the alignment suggests that the concept of deglobalization gained traction in tandem with real protectionist policy growth, lending strong support to H1.

This pattern remains robust across both controlled and robustness checks.

Language Entropy (H2)

The analyses demonstrate that Protections in Force is positively associated with language entropy (r ≈ 0.80–0.90, p < 0.01). This relationship remains strong even after adjusting for general internet content growth (partial r ≈ 0.78). Thus, higher cumulative protectionism is associated with greater multilingualism in online communication environments. This finding supports H2 and aligns with the theoretical expectation that economic deglobalization would erode the dominance of English in favor of greater linguistic diversity.

This result holds consistently after controlling for confounding factors and under robustness checks.

Political Polarization (H3)

The correlation and partial correlation analyses (Table 3 and Table 4) show a positive relationship between trade protectionism and political polarization (r ≈ 0.65–0.70, p < 0.05 to p < 0.01). Although this association is moderately weaker than for other outcomes, it remains significant after controlling for content growth. Thus, H3 is moderately supported: the data suggest that rising trade protectionism is associated with greater political polarization, although other social and media forces likely also contribute to polarization trends.

This relationship remains evident after robustness tests but is more modest in strength compared to H1, H2, H4, and H5.

Protests (H4)

Trade protectionism is highly correlated with increased protest discourse, with r ≈ 0.90–0.95 (p < 0.01) correlations across both zero-order and partial correlation analyses. Even after controlling for overall internet content growth, the association between protectionism and protest mentions remains strong (partial r ≈ 0.70). These findings robustly support H4, consistent with theoretical expectations that economic grievances linked to protectionism can fuel public unrest.

This association remains strong and stable across robustness checks.

Digital Authoritarianism (H5)

Finally, digital authoritarianism discourse correlates highly with trade protectionism (r ≈ 0.90–0.93, p < 0.01). The partial correlation analysis also shows a strong and significant association (partial r ≈ 0.69), confirming that as protectionist measures accumulate, references to digital repression also rise. This provides robust support for H5, suggesting that governments may respond to the tensions of economic deglobalization by tightening digital controls.

This pattern persists even after accounting for secular internet growth and holds across alternative operationalizations.

These findings prove that rising trade protectionism coincides with increases in deglobalization discourse, language diversity, political polarization, protest activity, and digital authoritarian practices. Although the study’s correlational nature precludes causal claims, the robustness of the associations across partial correlation analyses and supplementary checks strengthens confidence in the observed relationships.

6. Discussion

This study demonstrates that discourse about deglobalization—encompassing nationalism, protectionism, and protest—co-evolves with structural indicators, including trade openness, FDI, and global mobility. The results suggest that as global integration recedes, communicative patterns respond accordingly: protest and polarization increase, entropy falls, and governments become more likely to adopt digital authoritarian measures. While these associations are exploratory and correlational, they reveal consistent co-movement across multiple political and communicative domains, suggesting a broader systemic pattern exists.

Importantly, these results resonate with the theoretical frameworks outlined in Section 2.3 and Section 2.4, including World Systems Theory and Optimal Information Theory. These frameworks help explain why shrinking global ties produce measurable impacts on internal discourse, communication structures, and collective responses to uncertainty. Optimal Information Theory supports the idea that reductions in external information flow trigger compensatory internal dynamics—such as signal amplification, network densification, or polarization—that we observe here.

Recent work by Acemoglu and Robinson [105] highlights the crucial role that discourse plays not only in reflecting societal conditions but also in actively shaping institutional outcomes and policy directions. Their findings suggest that changes in public narratives and information environments can influence political alignment, institutional trust, and the implementation of economic reforms. This aligns closely with the present study’s focus on how deglobalization is not merely a structural or economic process, but also a communicative one, manifesting through shifts in discourse that may precede or amplify changes in protest behavior, polarization, and governance responses. By integrating these insights with theories from political communication and information processing, our analysis extends Acemoglu and Robinson’s argument into global network disintegration and its domestic reverberations in the information age.

At the same time, it is essential to recognize variations across different political regimes, regional media environments, and institutional capacities. The global-level findings presented here reflect average relationships across diverse national contexts; however, the mechanisms may differ depending on regime type, levels of media freedom, or degree of dependency on global trade. In authoritarian regimes, digital repression may occur earlier and more intensively, while in democratic settings, protest responses may be more visible in discourse before being met with institutional pushbacks.

Overall, the findings suggest that deglobalization is not only an economic transformation but a shift in communicative and institutional dynamics that deserves attention across disciplines. Future research can build on this work by identifying how these dynamics vary within and across countries and examining how discourse responds to and shapes real-world political outcomes.

6.1. Linking Empirical Patterns to Theory

The empirical patterns observed can be interpreted through the theoretical lenses outlined earlier:

• World Systems Theory: We found that as global interconnectedness weakened (due to rising tariffs, etc.), internal complexities increased: local languages gained ground (resulting in higher entropy), political factions hardened (leading to polarization), and governments exercised greater control over online spaces (manifesting as digital authoritarianism). This resonates with Wallerstein’s idea that breaking down a global core–periphery structure leads to intensifying internal reconfigurations. For instance, the dominance of the core-country language (English) diminished slightly, consistent with a diffusion of power. The strong correlations support a WST-informed view that deglobalization has systemic effects domestically: as the “core” influence recedes, nations turn inward, which in our data meant more inward-facing behaviors and challenges.
• Optimal Information Theory (OIT): Danowski’s extension of OIT suggested an inverse relationship between external ties and internal network density. Our results show that higher protectionism (fewer external economic ties) is associated with more intense internal dynamics, whether new language patterns or conflictual discourse requiring crackdowns. The strong correlation of protectionism with language entropy, protests, and digital authoritarianism [62,63,64] aligns with the notion that when external variety is reduced, internal signals amplify to compensate. In essence, a country facing less external economic input may generate more internal complexity (not all positive, as indicated by polarization and unrest). This provides indirect empirical evidence in support of OIT at a societal level.
• Social Strain and Resource Mobilization: The high correlation between protectionism (Protections in Force) and protest references reinforces the idea that economic disruptions or shifts can mobilize discontented groups. This is consistent with resource mobilization theory’s [76] claim that grievances, paired with available resources (such as communication networks and organizations), result in social movement activity. Likewise, social strain theory predicts that sustained pressure (cumulative protectionism) yields cumulative strain, culminating in protests. Our data cannot prove the mechanism, but the temporal alignment is suggestive. It is plausible that tariffs and trade shocks contributed to conditions (e.g., rising prices and unemployment in certain sectors) that sparked protests, or at least that both stem from a broader context of discontent.
• Political/Cultural Backlash: We observed a rise in language entropy alongside the emergence of protectionism, which aligns with the notion that local identities reassert themselves as globalization recedes [14]. Cultural Backlash Theory posits that segments of society react against global or liberal norms by emphasizing traditional or regional identities. The use of local languages could be one such expression. Additionally, the correlation between protectionism, polarization, and digital authoritarianism can be seen as a double-edged backlash: one portion of society pushing nationalist policies (a backlash against globalism), which then triggers another backlash in the form of protests or opposition, prompting an authoritarian response from the state. It is a chain of reactions consistent with polarization, where two camps pull apart and conflict ensues.

In sum, the empirical findings do not neatly “prove” any single theory but are congruent with multiple theoretical expectations. World Systems provides a macro-structural explanation, OIT gives a systemic information-flow explanation, and social movement theories explain the unrest aspect. The theories depict a feedback loop: economic policy shifts (such as trade) change social conditions, influencing communication and unrest, which may provoke a state response (including authoritarian measures). Our correlations are consistent with such a feedback loop, though we caution that we cannot confirm the sequence without time-lag analysis.

6.2. Political and Social Implications

Our results carry several implications for contemporary sociopolitical dynamics:

• Amplified Polarization: The correlation between protectionism and polarization discourse is evident in real-world examples. For instance, the U.S.’s “America First” trade policies and the UK’s Brexit (essentially, deglobalization moves) went hand in hand with highly polarized political climates. Our data suggest this is not coincidental; economic nationalism often comes packaged with narratives that split societies (globalist vs. patriot, etc.). This implication worries governance: countries could face more gridlock and less social cohesion if pursuing protectionist policies tends to deepen internal divides. Policymakers should be mindful that trade policy is not just a matter of technocratic economics; it has significant social ramifications.
• Rise in Protest Activity: History shows that sudden economic policy shifts (like sharp tariff increases) can trigger unrest, e.g., farmers protesting export bans and consumers protesting price hikes. Our data from the 2010s support this: as protectionism rose, so did protests in many parts of the world. Whether directly causal or due to underlying discontent, the association suggests that governments embracing protectionism may need to prepare for managing public dissent. For example, several countries that increased tariffs on fuel or food saw immediate street protests due to cost-of-living issues. Our findings align with this pattern. Thus, even policies intended to protect may have the paradoxical effect of inciting protest if they impose burdens on population segments.
• Digital Authoritarianism: One of the most striking findings is the high correlation between trade protectionism and references to digital control. This suggests a coupling of economic nationalism with information nationalism. A possible interpretation is that regimes that turn inward economically tend to tighten control over information, finding justification to maintain a cohesive narrative. For example, a government might censor online criticism of its protectionist policies or generally suppress dissent during economically tough times, which are often exacerbated by trade restrictions. The normative implication is significant: deglobalization may threaten digital rights and freedoms, as it could be accompanied by censorship and surveillance. Societies should be vigilant to ensure that “shielding the economy” is not used as a pretext to shield the public from information.

6.3. Relevance for Policymakers

From a policy perspective, our findings suggest several points worth noting:

• Policymakers should recognize the knock-on effects of protectionist economic decisions. Implementing a tariff to protect the domestic industry may not occur in a vacuum; it may alter domestic social dynamics and intensify political tensions. For instance, if protectionism leads to higher prices, this can become a political issue, polarizing public opinion and leading to protests. A government focusing on “economic sovereignty” might inadvertently fuel internal discord.
• A government that builds physical or regulatory trade barriers might also be more likely to erect “digital barriers”. Our analysis suggests a parallel: censorship or surveillance often accompanies shifts in nationalist policy. Therefore, advocates for free expression and internet freedom should pay attention to shifts in economic policy as potential early warning signs. Conversely, trade policymakers should consult with human rights advisors when advocating protectionist agendas, as such measures may have implications for civil liberty.
• Integrated Monitoring: This could be beneficial to monitor a dashboard of indicators: trade measures alongside social media sentiment, protest incidence, and internet freedom metrics. Such holistic monitoring might help anticipate crises. For example, suppose a country is ramping up trade barriers, and one sees simultaneous spikes in polarization on Twitter and protests being organized on Facebook. In that case, it may signal instability brewing that could escalate unless addressed.
• Communication Strategy: If a state chooses protectionist measures for legitimate reasons (say, to safeguard a critical sector), it should proactively communicate the rationale and address potential misconceptions to mitigate polarization. Transparency and inclusive dialogue can help prevent an “us vs. them” narrative from taking over. Without clear communication, people may fill the void with speculation or propaganda, worsening polarization.

In summary, policy architects need a broad lens. Economic nationalism might achieve certain goals at the expense of social cohesion and digital freedoms. Balancing these trade-offs is crucial.

6.4. Limitations and Future Directions

Despite the intriguing findings, this study has significant limitations stemming from both data and methodology:

• Small Sample of Annual Observations: We have only 14 time points. This raises the risk of spurious or inflated correlations; even random trends could appear significant with so few data points. While using large textual corpora per point lends some substantive validity, statistically, the sample is underpowered. Any conclusions are tentative. Future research should extend the time frame (as more years pass or possibly look historically further back to see if data can be compiled) or increase granularity (e.g., quarterly data and country-level panels) to provide more robust tests.
• Reliance on Mentions as Proxies: Our measures for polarization, protest, and other phenomena are based on counts of mentions in media and online content, rather than direct measures of the phenomena themselves. This is an important caveat: increased talk of “protests” does not always mean more actual protests occurred; it could reflect media sensationalism or other factors. Similarly, a society can be polarized without explicitly using the term “polarization” in articles. We chose these proxies due to global data availability. However, they imperfectly represent reality. Triangulating with more direct indicators would strengthen the case. For example, future studies could incorporate actual protest event data (counts of protests by year from databases such as ACLED), survey-based polarization metrics, or observed measures of language use (such as the percentage of web content in English versus other languages). Using multiple measures can also help alleviate biases (e.g., media censorship may underreport protests, so combining this with social media counts helps, which we partially did).
• Causality and Endogeneity: Our design is correlational, and we explicitly acknowledge this. We cannot disentangle whether trade protectionism leads to social tensions or whether a broader underlying factor, such as a wave of populism, drives both protectionism and the social outcomes. Indeed, it is plausible that a rise in nationalist sentiment around 2015 caused both anti-globalization policies and more polarized, protest-prone societies. The relationships are likely bidirectional and complex. To address causation, future research could employ time-series techniques (e.g., Granger causality tests or vector autoregressive models) to determine if changes in one variable precede changes in another more clearly. Alternatively, examining country-specific cases or natural experiments (e.g., a country that suddenly raises tariffs due to a political event and observes whether its protest frequency changes compared to similar countries) could provide insight. Cross-lagged panel analyses with country-level data may reveal whether past protectionism predicts future polarization better than vice versa.
• Interdisciplinary Data Integration Issues: Combining data from economics, media studies, and computational linguistics presents significant challenges. Each comes with different biases and error structures. For instance, Google’s hit counts are notoriously imprecise and can vary by orders of magnitude based on unknown indexing quirks. Nexis Uni’s coverage of newspapers might be biased toward English-speaking outlets or certain types of outlets. Although large, Facebook data only cover public posts and may not be representative of the broader discourse (mostly in English or a subset of engaged users). Additionally, while novel, our language entropy measure depends on accurate language detection and the representativeness of “posts about communication” as a sample. Our interdisciplinary approach may thus suffer from comparability issues; we treated all data streams with equal weight, but perhaps they should not be. For example, an uptick in Google hits might not be as meaningful as the same percentage uptick in a curated news database. Future work could improve this by calibrating measures (e.g., validating Google counts against known values or focusing on more consistent data sources).
• Variable Definition Clarity: While we strived to define each variable, some ambiguity remains. For instance, what constitutes a “protectionist measure” in GTA could range from a tariff to a subsidy, and not all measures have an equal economic impact. Similarly, “digital authoritarianism” encompasses various actions (censorship and surveillance) that may have different social consequences. Future studies may break these down further, perhaps as a separate analysis for “internet shutdowns” versus “social media censorship” to determine if one correlates more closely with unrest than the other. Ensuring consistent variable symbols and understanding is crucial. In our tables, we used shorthand like ‘PROTECT’ and ‘PROTECTIONS’, which could confuse readers if unclear. We have clarified these terms in the text (e.g., PROTECT = new measures and PROTECTIONS = cumulative). However, further clarity or standardization (such as using Inceptions vs. Protections in Force labels) would be helpful, especially as this work spans disciplines where terminology can differ.
• Cross-Cultural Variation: Our analysis treated the world somewhat monolithically. These relationships likely vary by country or region. For example, a country with strong institutions might implement tariffs without experiencing significant polarization, whereas in a more volatile political system, even minor protectionist measures could spark substantial conflicts. Our global aggregate could be dominated by the patterns of a few large countries (e.g., the U.S., China, and India), where content volume is substantial. A future direction is to disaggregate and examine country-level data, if possible. Some of our metrics can be compiled by country, as Global Trade Alert reports measures per country. Additionally, news and social media can be filtered by country or language. A panel study with countries as units could reveal whether the correlations hold broadly or are driven by specific cases. It could also test for moderation: for example, perhaps the link between protectionism and protests is stronger in democracies, where people can protest, than in autocracies, where protests are suppressed.

In conclusion, our study provides a macro-level, exploratory perspective that should be interpreted cautiously. We highlight robust cross-domain linkages but also stress their tentative nature due to data limitations. Armed with more data and refined methods, future research should further investigate these connections to establish causal narratives or rule out spuriousness. Our work opens a conversation across disciplines—between economists, communication scholars, and political scientists—about how to comprehensively assess the societal shifts accompanying deglobalization.

7. Conclusions

7.1. Main Contributions

This analysis expands the literature on deglobalization by empirically connecting trade protectionism to various communication and societal variables. Many prior studies focus on the economic ramifications of protectionism; we show that limiting cross-border trade may parallel or even accelerate changes in language usage, intensify domestic political rifts, trigger or amplify protest movements, and embolden digital repression by governments. These results are consistent with multiple theoretical approaches, ranging from systems-oriented frameworks (Optimal Information Theory and World Systems Theory), which predict internal adjustments to external change, to social movement perspectives (resource mobilization and social strain) that view economic stress as translating into collective action.

Our interdisciplinary approach demonstrates the value of integrating diverse data sources. We introduced a novel measure of cultural change—language entropy in social media—to the deglobalization debate. We empirically corroborated often theorized links, such as those between economic nationalism and polarization. Perhaps counterintuitively, we find that even a phenomenon as technical as trade policy is deeply intertwined with information and communication dynamics: what languages people speak online, how citizens align politically, what they protest about, and how governments manage information.

Crucially, our contribution lies not in establishing causality but in mapping a multifaceted correlation space from 2009 to 2023. The observed patterns form a basis for future work investigating the underlying mechanisms. For example, now that we know these variables move together, researchers can investigate specific case studies (e.g., how did the trade war impact Chinese social media discourse? Did countries that avoided deglobalization avoid polarization?).

By highlighting these broad correlations, we emphasize that deglobalization is not merely an economic reversal; significant societal shifts accompany it. This has implications for understanding the current era of global change as a complex system where economics, politics, culture, and technology are interlinked.

7.2. Policy Recommendations

While our study is global and macro in nature, a few policy recommendations emerge:

• Consideration of Social Externalities: Policymakers deliberating tariffs or quotas should weigh broader ramifications. Economic protection may achieve certain goals, such as protecting jobs in a sector, but can also stoke domestic political tensions and inadvertently prompt authorities to curtail freedoms if unrest escalates. A cost–benefit analysis of a trade policy should include social costs, such as increased polarization or protests, which have economic impacts of their own (e.g., instability can deter investment).
• Multilateral Engagement: Our findings suggest that engaging with the international community may mitigate some negative domestic effects. If absolute economic nationalism tends to polarize the population, participating in selective cooperation forms could ease the “us vs. them” narrative. For instance, leaders can pair protectionist moves with diplomatic efforts in other areas to demonstrate a balanced approach. Discussing digital rights and openness in trade negotiations could help forestall a slide into digital authoritarianism. International bodies, such as the WTO or G20, could explicitly address the interplay between trade policy and social cohesion, encouraging members to adopt measures that minimize societal disruption.
• Fostering of Inclusive Communication Environments: Civil society and the media play a crucial role. In times of deglobalization, maintaining open channels of discourse is key. Efforts to promote media literacy and resist disinformation can help mitigate worsening polarization. Encouraging multilingual content and dialogue might turn the language entropy finding positive. Instead of seeing the decline of a lingua franca as a fracturing, it can be embraced as a form of cultural pluralism. Educational and cultural institutions might emphasize national languages in a way that unites rather than divides (e.g., bilingual education that values both English and local languages). At the same time, they should guard against extreme nationalist narratives. Social media platforms should be vigilant about how their algorithms might amplify polarization in a politically charged environment, building on research into the filter bubble phenomenon.

In short, a holistic approach to policy is recommended. If a nation chooses an economically deglobalizing path, then it should proactively implement measures to keep its society inclusive and its government accountable, thereby avoiding the darker correlated outcomes we have identified.

7.3. Closing Remarks

In sum, deglobalization is not purely an economic phenomenon but a societal one. By examining the interplay between trade protectionism and communication variables, this study underscores the broad societal shifts that can accompany a turn away from global markets. Despite relatively few observations, the consistent correlations suggest that states and societies must know how economic nationalism can alter the informational and discursive environment. As the contours of globalization continue to evolve—be it through further unraveling or partial resurgence—monitoring these socio-communicative dimensions will be crucial for scholars, policymakers, and citizens alike.

We emphasize caution in interpretation: correlation does not imply causation. Countries may continue to find ways to reap some benefits of deglobalization, such as local empowerment, without suffering all the downsides, like repression. Achieving that balance requires awareness of these interconnections. We hope this research stimulates further inquiry into how the global and local systems interact, and how policy choices in one domain reverberate across the entire social system.