Two decades of changes in social cohesion in Latin America (2004-2023)

Authors

Affiliations

Juan Carlos Castillo

Universidad de Chile

Centro de estudios del conflicto y cohesión social (COES)

University of Bremen

Gabriel Cortés Paredes

Centro de estudios del conflicto y cohesión social (COES)

Tomás Urzúa

Centro de estudios del conflicto y cohesión social (COES)

Julio Iturra-Sanhueza

University of Bremen

Abstract

In a regional context marked by political crises, persistent inequalities, and episodes of social unrest, understanding the evolution of social cohesion is essential for assessing democratic stability and institutional legitimacy. Although there are numerous studies on the causes and consequences of mistrust or polarization in Latin America, there is still a gap in the systematic and longitudinal analysis of social cohesion as an integral phenomenon. This paper seeks to fill that gap by developing a set of indicators that allow for the analysis of the evolution of the different dimensions of social cohesion with temporal and regional comparability.

This article seeks to fill these gaps by proposing and validating a measurement model that allows for a comparative, longitudinal, and multilevel analysis of social cohesion in Latin America. Specifically, we seek to advance: (i) a clear and validated operationalization that integrates key dimensions from the existing literature and available data for the region; (ii) the estimation of regional and national trajectories over the last two decades; and (iii) the identification of factors associated with these changes through the application of hybrid multilevel regression models. This is expected to provide robust evidence on changes in the region over the last two decades, contributing to the academic and political discussion on the challenges and opportunities of social cohesion in Latin America.

Keywords

social cohesion, multilevel analysis, longitudinal analysis

Introduction

Social cohesion has emerged as a critical dimension of societal well-being and democratic stability, particularly in regions experiencing rapid social and political transformation. In Latin America, recent years have been marked by episodes of political instability, persistent inequalities and low economic growth, as well as cycles of social conflict (Salazar-Xirinachs, 2023; United Nations Development Programme, 2023). Social tensions therefore appear to have increased in the region, reflecting a lack of trust in democratic institutions and widespread discontent with corruption and inequality. In this context, social cohesion has climbed the public and academic agendas, with recent diagnoses from both international organizations and national governments warning about its strains and the challenges it poses for democratic governance and inclusive development (Castillo et al., 2022; Ministerio de Desarrollo Social y Familia, 2020; Salazar-Xirinachs, 2023; United Nations Development Programme, 2023).

Despite its widespread and everyday use in public discussion, defining social cohesion theoretically and operationally remains a challenge. The literature ranges from studies focused on one or several specific dimensions of social cohesion (Ariely, 2013; Castillo et al., 2022; Castillo et al., 2023) to efforts to synthesize the phenomenon into comprehensive indices (Delhey et al., 2018; Delhey & Dragolov, 2016; Dragolov et al., 2013; Janmaat, 2010; Langer et al., 2016). Such conceptual and methodological heterogeneity makes it difficult to compare countries and to detect transformations over time. Moreover, most of these definitions and their correspondent operationalizations have been tested mainly in European or high-income countries (Ariely, 2013; Delhey & Dragolov, 2016), with only partial references to Latin America (Janmaat, 2010). This limitation persists even though evidence suggests that national and regional differences in cultural, historical, and institutional contexts shape the cohesion of societies and the factors that determine it (Delhey & Dragolov, 2016; Janmaat, 2010). Thus, despite the widespread perception that social cohesion is under strain in Latin American societies,there is a lack of empirical evidence of differences between countries, trends over time, and the factors that explain these changes.

This article has a double aim. On the one hand, and based on previous comprehensive approaches to social cohesion, it seeks to propose and validate a measurement model that enables comparative, longitudinal, and multilevel analysis of social cohesion in Latin America. On the second hand, based on this model, we aim to advance on the estimation of regional and national trajectories over the past two decades, as well as to identufy factors associated with these changes through longitudinal multilevel regression models. This is expected to provide robust evidence on the changes in the region over the last two decades, contributing to the academic and policy discussion on the challenges and opportunities of social cohesion in Latin America.

What is social cohesion?

Social cohesion is a multifaceted construct that plays a critical role in community interactions, impacting various social and health outcomes. The conceptualization of social cohesion encompasses various dimensions, such as connectedness, shared values, and participation in collective activities. Although its theoretical roots date back to the origins of social theory — and it has widespread use in public discourse — defining social cohesion remains a difficult task (Chan et al., 2006). Situated at the intersection of theoretical and descriptive approaches that emphasize cooperation and integration, as well as normative considerations about how societies stay united in the face of crises, tensions, and transformations, social cohesion has been defined in multiple ways, reflecting diverse perspectives, scopes, and indicators (Castillo et al., 2021; Chan et al., 2006).

Social cohesion has been associated to different outcomes. Scholars as Turchi et al. (2023) articulate that social cohesion becomes paramount in emergency contexts, where interpersonal connections can provide essential support and resilience. This assertion is supported by Stafford et al., who reveal that diminished social cohesion is correlated with health detriments, including increased mortality rates and adverse mental health indicators (Stafford et al., 2003). Lê et al. (2013) et al. further corroborate this perspective, exploring how cohesive communities can mitigate mental health issues in the aftermath of traumatic events. Moreover, context plays a significant role in the assessment and impacts of social cohesion. Nazmi et al. (2010) illustrate how neighborhood characteristics influence health outcomes, suggesting that communities with stronger social interactions and solidarity can buffer against health stresses (Nazmi et al., 2010). In contrast, other research has shown that lower levels of perceived neighborhood cohesion are associated with adverse health conditions (Neergheen et al., 2019).

Most definitions of social cohesion share some common elements (Castillo et al., 2021):

• It is an attribute of the collective, not of individuals.
• It is a multidimensional construct
• It is a quality of social relations that enables the achievement of shared goals.
• Given that subjects have a common objective, they are capable of cooperating to achieve these objectives.

Therefore, a robust definition of social cohesion should be able to take the previous elements into consideration, and at the same time it should allow for a comprehensive operationalization and measurement. Taking this into consideration, in this article we will begin with the definition proposed by Chan, To and Chan (2006, p. 90), for whom social cohesion “is a state of affairs concerning both the vertical and the horizontal interactions among members of society as characterized by a set of attitudes and norms that includes trust, a sense of belonging and the willingness to participate and help, as well as their behavioural manifestations”. From this definition, two main dimensions emerge. First, social cohesion has a horizontal dimension. By this we refer to the daily interactions among members of society. It contains subjective aspects (such as sense of belonging and interpersonal trust), as well as objective aspects, such as social networks and material conditions that facilitate or hinder these interactions. Second, social cohesion has a vertical dimension, which refers to the relations between individuals and social and political institutions. This dimension includes aspects such as trust in institutions, the legitimacy of the political system, and civic participation.

Measuring Social Cohesion

The measurement of social cohesion is equally critical and can vary considerably depending on the metrics employed. Bottoni emphasizes this by asserting that social cohesion is inherently multidimensional, incorporating structural components such as trust and social networks (Bottoni, 2016). Mair et al. and Cooper et al. both utilized factor analysis in their work to develop robust scales that effectively capture different facets of social cohesion, indicating the complex nature of social bonds within communities (Mair et al., 2009; Cooper et al., 2014). These scales assess variables like mutual responsibility and shared values, which are essential for understanding neighborhood dynamics.

A notable contribution to the framework of social cohesion measurement is provided by Delhey et al. (2023), whose work emphasizes the necessity of incorporating a broad spectrum of attributes beyond mere trust and conflict perceptions. In their review, Delhey et al. (2023) critique many existing measures for focusing narrowly on isolated components of social cohesion, which may lead to oversimplified interpretations. By highlighting social cohesion as a community asset that fosters solidarity and reduces conflict, Delhey et al. articulate its role in enhancing collective welfare and health outcomes. Moreover, their advocacy for multidimensional measurement approaches reflects an understanding that the impacts of social cohesion can vary significantly across different socio-cultural contexts. This adaptation is crucial for ensuring that social cohesion measures remain relevant and effective, addressing the criticisms of traditional models that fail to encompass the complexity of cohesive interactions.

Studying social cohesion

Social Cohesion is usually understood as a gradual phenomenon, meaning that societies can display higher or lower levels of cohesion (Delhey et al., 2023; Schiefer & van der Noll, 2017). This can be interpreted both longitudinally, that is, countries can increase or decrease their cohesion levels over time; as well as comparatively, that is, some countries can show higher levels of cohesion than others.

Despite the assumption that social cohesion levels change over time, the difficulty of conducting repeated measurements on population attitudes and behaviors hinders the longitudinal analysis of social cohesion. An important exception to this trend is the Mapping Social Cohesion Survey, conducted by the Scanlon Foundation Research Institute since 2007, which enables the construction of the Scanlon-Monash Index (SMI) of Social Cohesion. Thus, the SMI allows observing synthetically the evolution of Social Cohesion in Australia, recording a 22% decline since 2007 and significant decreases in 4 of its 5 dimensions (O’Donnell et al., 2024). Another notable example of time comparison is the Social Cohesion Radar, whose index can be reconstructed for four time periods between 1989 and 2012 (Delhey et al., 2023). However, its scope is limited to European and Western societies.

The literature on social cohesion appears to be more focused on establishing which countries show higher or lower levels of cohesion (Delhey et al., 2018; Delhey & Dragolov, 2016), proving the existence of cohesion regimes (Delhey & Dragolov, 2016; Janmaat, 2010), and determining the factors related to social cohesion (Delhey et al., 2018; Delhey & Dragolov, 2016; Green et al., 2011; Janmaat, 2010). These associated factors can be grouped into economic factors, political factors, and demographic factors.

Economic prosperity, often measured through GDP per capita, has been associated with higher levels of social cohesion. This association has been found across diverse cultural contexts (Delhey et al., 2023; Janmaat, 2010), although it appears to attenuate in some high-income Asian societies (Delhey et al., 2018). These findings align with universalist or modernist perspectives on social cohesion, which view cohesion as a phenomenon closely related to different stages of socioeconomic development (Janmaat, 2010). From these findings, we derive the first hypothesis of this article:

H1: Countries with higher economic development will exhibit greater levels of social cohesion

Conversely, economic inequality has generally been associated with lower levels of social cohesion (Delhey et al., 2018, 2023; Janmaat, 2010). Again, however, this association is attenuated in Asian societies, where moderate levels of inequality are associated with higher levels of cohesion (Delhey et al., 2018; Janmaat, 2010).

To date, there is no evidence regarding how inequality levels in Latin America relate to social cohesion. On one hand, despite being considered one of the world’s most unequal regions, Latin Americans display exceptionally high expectations of social mobility (Somma & Valenzuela, 2015), possibly due to increased educational coverage in many regional countries or the predominance of meritocratic justifications for educational and economic inequality (Castillo et al., 2024; Somma & Valenzuela, 2015). Nevertheless, demands for greater equality, redistribution, and fair treatment (Araujo et al., 2022; Cabib et al., 2025), which have translated into mass protests in several regional countries, suggest that economic inequality is a relevant factor for social cohesion in Latin America. Therefore, we propose the following hypotheses:

H2(a): Increases in a country’s economic inequality will be associated with lower levels of social cohesion

H2(b): Countries with higher levels of inequality will exhibit lower levels of social cohesion

The state’s capacity to respond effectively to population challenges, particularly for vulnerable groups, has been identified as one of the primary sources of social cohesion in diverse societies (ANDREWS & JILKE, 2015; Kustov & Pardelli, 2024; Njozela et al., 2016). In this vein, effective governance and robust institutions are typically associated with higher levels of social cohesion in contexts as diverse as Asia, Latin America, and Europe (Delhey et al., 2018; Green et al., 2011). In contrast, the relationship between political systems and democratic regimes with social cohesion appears to be more context-dependent. While in wealthy Western societies liberal democracy tends to reinforce social cohesion levels, in some Asian countries authoritarian regimes have been linked to equally high levels of cohesion (Delhey et al., 2018).

In Latin America, states are typically characterized by lower levels of efficiency and higher corruption indices, accompanied by citizenries with low levels of trust in political institutions and lower participation compared to the developed world. However, this institutional distrust coexists with a strong sense of national identification, even more pronounced than in the developed world (Somma & Valenzuela, 2015). For Somma (2015), this apparent paradox relates to the propensity of Latin American societies toward personalist and populist leadership. In this context, consistent with what Guillermo O’Donnell described as delegative democracies (Toppi, 2018), broad sectors of the citizenry appear willing to trade liberal democratic principles in favor of authoritarian leadership when they perceive that democratic institutions fail to resolve their daily problems. This suggests that governance effectiveness—rather than political regime type—is the institutional factor that carries greater weight on social cohesion levels in Latin American societies. From this discussion, we propose the following hypotheses:

H4(a): Increases in a country’s governance levels will be associated with higher levels of social cohesion

H4(b): Countries with higher levels of governance will exhibit higher levels of social cohesion

Cultural and demographic variables, including ethnic, religious, and national identities, have demonstrated complex effects on social cohesion levels, varying substantially by context. This contextual character is particularly evident in the impact of ethnic diversity on social cohesion. On one hand, several studies have identified negative effects of ethnic diversity on indicators such as sense of belonging or the strength of neighborhood networks (Ariely, 2013; Gijsberts et al., 2011), an association that is particularly pronounced in contexts of high racial segregation, such as the United States (Meer & Tolsma, 2014). Nevertheless, other research has shown that institutional factors and the implementation of inclusive policies can moderate, or even reverse, this relationship (Delhey et al., 2018, 2023; Kustov & Pardelli, 2024; Meer & Tolsma, 2014; Reeskens & Wright, 2012).

The management of ethnic diversity in Europe and North America contrasts significantly with the Latin American reality, characterized by greater ethnic and linguistic heterogeneity. Despite episodes of conflict arising from relationships between states and indigenous peoples, Latin American countries have demonstrated strong national identities (Somma & Valenzuela, 2015). However, recent changes in migration patterns, which have favored intraregional migration in both relative and absolute terms (Stefoni, 2018), could constitute new sources of tension for social cohesion in Latin American societies. Although Latin Americans tend to display levels of ethnic-religious and national tolerance higher than those observed in several European countries (Somma & Valenzuela, 2015), cases such as Chile—in response to massive Venezuelan migration—evidence signs of deterioration in relations between migrants and non-migrants in recent years (Castillo et al., 2023). In light of this, we propose the following hypotheses:

H5(a): Increases in the proportion of migrant population in a country will be associated with decreases in social cohesion levels

H5(b): Countries with higher levels of ethnic diversity will exhibit lower levels of social cohesion

H5(c): The impact of migration on social cohesion will be smaller in countries with better governance levels

Methodology

Data

The main source of data for this study is the AmericasBarometer of the Latin American Public Opinion Project, also known as the LAPOP Survey. The survey aims to collect data on public opinion about democracy and governance in the Americas. The survey design is probabilistic and representative of the adult population in each country (LAPOP LAb, 2023).

The survey has been conducted regularly since 2004. To date, nine waves have been carried out, covering between 11 and 23 countries, with a total of over 400,000 interviews in two decades. The questionnaire is administered through face-to-face surveys, with the exception of Canada and the United States.

As a criterion, this study included only those countries in the region that had data available for the main indicators of the study at least five points in time. As summarized in Table 1, this study includes a total of 238,257 individuals nested in 174 country waves in 25 countries in the Americas.

Table 1: Data availability by waves and countries

Country	2004	2006	2008	2010	2012	2014	2016	2018	2023	Total
Argentina	0	0	1231	1235	689	1249	2816	2884	1469	11573
Belize	0	0	1096	1341	715	1380	0	0	1411	5943
Bolivia	2802	2563	2590	2603	2605	2848	2900	3060	0	21971
Brazil	0	0	1203	2112	681	1390	2906	2782	1407	12481
Chile	0	1412	1358	1723	699	1268	2900	2978	1525	13863
Colombia	1304	1268	1318	1272	624	1353	2886	1530	1400	12955
Costa Rica	1386	1422	1350	1373	674	1425	2758	2806	1447	14641
Dominican Republic	0	1312	1265	1292	656	1400	1288	1360	2974	11547
Ecuador	2644	2657	2768	2589	628	1315	1388	2934	1414	18337
El Salvador	1407	1573	1468	1500	620	1448	1450	1353	1457	12276
Guatemala	1237	1133	1157	1217	643	1309	1355	1362	1425	10838
Guyana	0	1151	2069	1267	654	1315	0	0	0	6456
Haiti	0	1363	1351	1596	845	1318	1681	0	0	8154
Honduras	1220	1366	1275	1437	717	1452	1349	1328	1385	11529
Jamaica	0	1219	1272	1297	706	1202	1180	1151	1209	9236
Mexico	1366	1351	1402	1385	697	1320	1374	1387	1545	11827
Nicaragua	1098	1491	1278	1373	1573	1402	1383	1376	0	10974
Panama	1528	1374	1405	1435	706	1414	2868	2924	1454	15108
Paraguay	0	0	1065	1300	676	1240	1170	1346	1310	8107
Peru	0	1340	1385	1382	670	1337	2478	1449	1481	11522
Trinidad & Tobago	0	0	0	0	0	0	0	0	1444	1444
Uruguay	0	1068	1359	1321	671	1342	2720	2866	1383	12730
Venezuela	0	1386	1330	1340	686	1377	2794	0	0	8913
Total	15992	26449	31995	33390	17835	31104	41644	36876	27140	262425

For contextual data on countries, various data sources were used, including:

1. Open data from the World Bank. This contains various indicators on social and economic development for most countries in the world. The data portal is accessible at: https://datos.bancomundial.org/.

2. The World Bank’s Worldwide Governance Indicators. This is a survey of experts that collects data on various governance indicators, covering multiple countries with information updated between 1996 and 2003. The data is available at: https://www.worldbank.org/en/publication/worldwide-governance-indicators.

3. The V-Dem Dataset. It collects a multidimensional set of data that seeks to measure the quality of democracy around the world. The database is accessible through the R package vdemdata (Maerz et al., 2025).

Variables

Dependent variables

A Social Cohesion Index was constructed, comprising two dimensions which, in turn, are summary indices constructed from LAPOP indicators. The selection of indicators, sub-dimensions, and dimensions is based on previous work at the aggregate level by the Social Cohesion Observatory, accessible here: https://ocscoes.github.io/medicion-cohesion-LA/.

The Horizontal Cohesion Index consists of two sub-dimensions: Urban Safety and Interpersonal Trust. Urban Safety includes indicators of objective safety and subjective safety. Interpersonal Trust, meanwhile, is a single indicator of how trustworthy people are in general[¹].

The Vertical Cohesion Index consists of two dimensions: Trust in Institutions and Attitudes toward Democracy. Trust in Institutions includes indicators related to citizens’ trust in Congress, the judiciary, and political parties. Attitudes toward Democracy consists of two indicators on support for the democratic system and satisfaction with the functioning of democracy in one’s country[²].

The indicators were standardized so that all sub-dimensions and dimensions of the indices have a range from 0 to 10, with 0 indicating low levels of social cohesion and 10 indicating high levels of cohesion.

Independent variables

Economic, institutional, and cultural factors were included as independent variables. Recognizing the hierarchical structure of the data, predictors were considered at the individual level, at the wave-country level, and at the country level.

Individual variables

The main individual predictor used in this study is the educational level of individuals. The multiple LAPOP codes for this indicator were unified to create a variable with three categories, distinguishing between individuals with primary, secondary, and tertiary education.

In addition, gender, age, and political position were added as control variables.

Contextual variables

Economic prosperity was measured using the logarithm of GDP per capita at purchasing power parity (PPP) values. Economic inequality was measured using the Gini index. Both indicators were extracted from the World Bank database. In addition, the percentage of individuals with tertiary education in a country was included as a proxy for educational opportunities.

In terms of institutional factors, the Electoral Democracy Index, or polyarchy, was used to measure the democratic quality of countries. On the other hand, a governance index (\(\alpha\) = 0.96) calculated from the World Bank’s Worldwide Governance Indicators was included (Kaufmann & Kraay, 2024).

Cultural diversity will be measured as the percentage of the migrant population relative to the total population of the country, using data from the World Bank. Given that the series is available every five years, an annual series was constructed by imputing the intermediate years using logistic interpolation.

Method

Confirmatory factor analysis

A confirmatory factor analysis was performed to test the model constructed by the Social Cohesion Observatory (2025) and the theoretical proposal by Chan et al. (2006). As can be seen in Figure 1, social cohesion is understood as a latent construct consisting of two latent dimensions: vertical cohesion and horizontal cohesion.

Figure 1: Conceptual Framework for Social Cohesion

Multilevel analysis

Given the hierarchical structure of the data, hybrid multilevel regression models were estimated. This technique allows individual-level data to be used to decompose country-level effects into their components between countries (between effects) and within a country over time (within effects) (Fairbrother, 2014; Schmidt-Catran & Fairbrother, 2016). The models were estimated using the R package lme4 (Bates et al., 2015).

The proposed model could be formally expressed as:

\[ y_{jti} = \beta_{0}(t) + \beta_{1}X_{jti} + \gamma_{we}(Z_{jt}-\bar{Z}_{j}) + \gamma_{be}\bar{Z}_{j} + v_j + u_{jt} + e_{jti} \] The model integrates three levels with individuals \(i\) nested in waves-countries \(t\) nested in countries \(j\). \(X_{jti}\) represents individual-level variables, while \(Z_{jt}\) are contextual variables at the wave-country level. Given that \(Z_{jt}\) contains variance at both level 2 and level 3, it was broken down into the average of the variable across all its waves (\(\bar{Z}_{j}\)) and the intra-country deviation in a given wave (\(Z_{jt}-\bar{Z}_{j}\)). Thus, \(\gamma_{we}\) represents the within effect, that is, the effect of change in a country over time, while \(\gamma_{be}\) represents the between effect, that is, the structural differences between countries. In addition, \(\beta_{0}(t)\) controls for changes over time not explained by the model. Finally, \(v_j\), \(u_{jt}\) and \(e_{jti}\) represent the errors at the country, wave-country, and individual levels.

Results

Confirmatory Factor Analysis

Figure 2: Confirmatory Factor Analysis for Social Cohesion Measurement Model

Figure 2 reports the confirmatory factor analysis for the proposed two-factor measurement model. At first, overall fit appears strong (RMSEA = 0.03; CFI=0.0996). However, these indices should be interpreted with caution, given that including only two indicators per factor results in the model being just overidentified (df=1), which limits the diagnostic value of global fit measures.

Standardized factor loading are moderate. For Horizontal Cohesion, loadings range from 0.44 to 0.49, and for Vertical Cohesion from 0.56 to 0.58. The latent correlation between both factors is also moderate (r = 0.51), implying related but empirically distinguishable dimensions. Following Dickes and Valentova (2013), this pattern supports interpretation of social cohesion as a multidimensional construct. However, the statement that a single composite indicator is “not possible” (2013, p. 836) might be discussed. Empirically, the moderate latent correlation observed here suggest a shared component that could justify, on pragmatic grounds, a composite summary for descriptive purposes. Nevertheless, such composite cannot be treated as evidence of unidimensionality nor a higher-order general factor. Accordingly, subsequent analyses will primarily treat Vertical and Horizontal dimensions as separate construct, with any exceptions explicitly noted.

Descriptives

Figure 3: Evolution of vertical and horizontal cohesion (2004-2023)

Figure 3 shows the averages for each year for the different types of cohesion. First, we observed that horizontal cohesion maintained an upward trend between until 2012, reaching 6.54 (out of 10) that year. After that, horizontal cohesion faced a sustained decline until 2018, with a regional mean of 5.93, before rebounding to somewhat similar levels as 2004 (6.17).

Meanwhile, vertical cohesion show a similar trends, with a persistente trends overt time until 2010, when it suffered a sharp decline. Between 2010 and 2016, the regional mean went from 5.24 to 4.44. After that, the index showed a slight increase, reaching 4.63 in 2023.

Across the entire period under study, horizontal cohesion remains consistently higher than vertical cohesion, with an average gap of 1.38 points. This persistent difference may reflect the structural and enduring institutional weaknesses present in many Latin American states, which might tend to reinforce the role of interpersonal relationships, family ties, and informal networks as key resources for facing social life (Araujo & Martuccelli, 2014; Brinks et al., 2020).

At first glance, these trends appear closely aligned with the broader political, economic, and social context of Latin America over the period analyzed. The upward trajectories observed until 2010–2012 coincide with the commodities boom, during which rising prices of commodities such as oil, copper, and agricultural products—driven largely by demand from emerging economies—supported sustained economic growth in the region. This favorable context was leveraged by progressive and populist governments to implement redistributive policies, contributing to notable reductions in poverty and income inequality across several countries (Sánchez-Ancochea, 2021).

Following the slowdown of the Chinese economy around 2012 and the subsequent end of the commodities boom, many Latin American countries entered a period of economic, political and social struggles, which might be reflected in the decline of both dimensions of cohesion. Meanwhile, the slight rebound observed in 2023 may be interpreted as a process of social and political recomposition in the aftermath of the COVID-19 pandemic, during which the expansion of state support and the reinforcement of solidarity mechanisms may have partially mitigated previous erosions of social cohesion.

Relationship between and within country level factors with social cohesion

Figure 4 presents bivariate associations between both dimensions of social cohesion and macrostructural variables such as economic inequality, GDP per capita (PPP), governance, the V-Dem democracy index, and the proportion of migrant population. The panels allow us to visually distinguish variation between countries and association within countries, represented by the gray lines. The red line represents the linear fit including the United States and Canada. Since these two countries tend to appear clearly as outliers, the fit excluding them is also reported (in black). In general, the exclusion of these cases does not alter the direction of the associations, but it does reduce their strength, shifting them toward moderate or low ranges.

Figure 4: Country-level association of macro-level factors and social cohesion. Red line represents the association including Canada and United States.

The correlation between economic inequality and horizontal cohesion is -0.2 (excluding Canada and the United States), with associations within countries ranging from -0.77 in Argentina to 0.64 in Brazil. The association with GDP per capita (PPP) is slightly positive (0.14), with countries such as Venezuela (r=0.88), El Salvador (r=0.67), and Brazil (r=0.42) showing correlations above the average, and countries such as Colombia (r=-0.9), Mexico (r=0.79), and Ecuador (r=-0.73) showing a negative association. The average correlation between the V-Dem Democracy Index and Horizontal Cohesion is 0.14, with Venezuela showing a strong positive association (r=0.73) and El Salvador a strong negative one (r=-0.85). The association between governance and horizontal cohesion is the most pronounced, with a coefficient of 0.43. However, it is possible to find countries that show the opposite trend, such as Colombia (r=-0.73), Paraguay (r=-0.63), and Ecuador (r=-0.62). Finally, the proportion of migrants in the country shows a positive association with horizontal cohesion (r=0.24), but with strong negative trends in countries such as Ecuador (-0.87), Colombia (-0.84), and Mexico (-0.78).

The associations between vertical cohesion and macro indicators follow the same general trends as in the horizontal dimension, although differences in magnitude are observed. Thus, the association with economic inequality is stronger in the vertical dimension (-0.36), although countries such as Chile (r=0.77), Brazil (r=0.74), and Colombia (r=0.73) show an association in the opposite direction. The average effect of GDP per capita (PPP) is practically flat (r=-0.01), but there are also countries with a strongly negative trend, such as Chile (-0.82) and Colombia (-0.81). Something similar occurs with the association between the V-Dem Index and Vertical Cohesion, where an average correlation of 0.04 does not exclude the presence of strongly positive associations in Venezuela (r=0.92) or strongly negative ones in countries such as Ecuador (-0.90). The association between governance and vertical cohesion is less pronounced than in the horizontal dimension (r=0.18), with a range from -0.87 in Colombia to 0.91 in Venezuela. Finally, the effect of the percentage of migrant population is greater in the vertical dimension than in the horizontal dimension (r=0.41), with strongly negative effects observed in countries such as Chile (-0.78) and Peru (-0.77).

Differences and changes in country level factors and social cohesion

Table 2: Multilevel regression for individual level factors and social cohesion

	Horizontal		Vertical
	Model 1	Model 2	Model 3	Model 4
(Intercept)	6.205***	6.240***	5.254***	5.029***
	(0.115)	(0.114)	(0.125)	(0.123)
Income decile	0.027***	0.026***	0.004*	0.002
	(0.002)	(0.002)	(0.002)	(0.002)
Education (ref.=Secondary)
Primary	0.202***	0.196***	0.132***	0.122***
	(0.013)	(0.013)	(0.010)	(0.010)
Tertiary	0.099***	0.102***	0.003	0.002
	(0.013)	(0.013)	(0.010)	(0.010)
Political position (ref.=Undefined)
Left		-0.162***		0.073***
		(0.017)		(0.013)
Center		-0.052**		0.193***
		(0.017)		(0.013)
Right		0.041*		0.568***
		(0.017)		(0.013)
Time	-0.051***	-0.048***	-0.100***	-0.098***
	(0.011)	(0.011)	(0.015)	(0.015)
Male (ref.=Female)	0.202***	0.204***	0.044***	0.033***
	(0.010)	(0.010)	(0.008)	(0.008)
Age	0.098***	0.095***	-0.007	-0.012*
	(0.006)	(0.006)	(0.005)	(0.005)
Age2	0.048***	0.046***	0.094***	0.090***
	(0.004)	(0.004)	(0.003)	(0.003)
Controls	Yes	Yes	Yes	Yes
BIC	958966.852	958777.616	849552.485	846499.344
Num. obs.	213130	213130	213130	213130
Num. Country-wave	161	161	161	161
Num. Country	22	22	22	22
Var: Country-wave (Intercept)	0.095	0.093	0.195	0.187
Var: Country (Intercept)	0.199	0.194	0.170	0.167
Var: Residual	5.016	5.010	2.999	2.956
***p < 0.001; **p < 0.01; *p < 0.05

The models in Table XX present the estimates of the multilevel models examining the association between individual characteristics and social cohesion. For horizontal social cohesion, Model 1 indicates that a higher position in the relative income distribution is associated with greater social cohesion (\(\beta = 0.027\), \(p < .001\)). This contrasts with differences between educational groups, where, compared to the group with secondary education, individuals with primary education (\(\beta = 0.202\), \(p < .001\)) and those with tertiary education (\(\beta = 0.099\), \(p < .001\)) exhibit higher horizontal social cohesion. Model 2 incorporates political position into the estimation, showing that, compared to the ideologically undefined group, both centrists (\(\beta = -0.052\), \(p < .01\)) and leftists (\(\beta = -0.162\), \(p < .001\)) report lower levels of horizontal cohesion. In contrast, right-leaning individuals display higher levels of social cohesion (\(\beta = 0.041\), \(p < .05\)).

Similar to the results for horizontal cohesion, Model 3 shows that individuals with higher incomes exhibit higher vertical cohesion, though with a smaller coefficient and a more moderate level of significance (\(\beta = 0.004\), \(p < .05\)). Differences between educational groups are more attenuated compared to the horizontal dimension. Relative to the group with secondary education, those with primary education report higher levels of vertical cohesion (\(\beta = 0.132\), \(p < .001\)), whereas those with tertiary education show no significant differences (\(\beta = 0.003\), \(p > .05\)). When political position is incorporated into Model 4, differences from Model 3 emerge. Income loses strength and statistical significance (\(\beta = 0.002\), \(p > .05\)), while the relationship with educational level remains robust. Regarding political groups, individuals who declare a political position generally exhibit higher levels of vertical cohesion. Compared to those who are undecided, people on the right show the highest levels of vertical cohesion (\(\beta = -0.052\), \(p < .01\)), followed by centrists (\(\beta = -0.052\), \(p < .01\)) and leftists (\(\beta = -0.052\), \(p < .01\)).

For the time variable in Models 2 and 4, results indicate a decline in both horizontal (\(\beta = -0.048\), \(p < .001\)) and vertical (\(\beta = -0.098\), \(p < .001\)) cohesion during the analyzed period. The decrease in the vertical dimension is nearly double that observed in the horizontal dimension. Additionally, a quadratic term was included to assess whether the decline accelerated in recent periods; however, no significant differences were found when incorporating the quadratic term of time (see supplementary material).

In line with our hypotheses, Table XX presents the estimates of the hybrid multilevel models that show the between- and within-country differences in social cohesion.

Horizontal Cohesion

The results display the associations between (BE) and within (WE) for inequality (Gini) and economic prosperity (GDP). Regarding inequality, across all specifications (Models 1 to 3), differences between countries in levels of economic inequality are not associated with social cohesion. However, an increase in inequality within countries is associated with lower social cohesion. Similarly, the results show that neither between-country differences nor increases in economic prosperity appear to affect horizontal cohesion. Regarding political variables, we observe that both levels of electoral democracy (Democracy) and governance affect social cohesion in different ways. First, differences between countries in levels of democracy are associated with lower levels of horizontal cohesion. However, this relationship is not observed within countries. Second, differences between countries in governance levels indicate that governance is associated with higher levels of horizontal cohesion. Nevertheless, changes in governance within countries do not appear to be associated with horizontal cohesion. Finally, the results in Model 3 show that differences between countries in the percentage of the migrant population are not associated with levels of horizontal cohesion. However, an increase in the migrant population within countries is negatively associated with horizontal cohesion.

Vertical Cohesion

Similar to the results for horizontal cohesion, we observe distinct patterns for vertical cohesion. Differences between countries in levels of economic inequality indicate that countries with higher inequality exhibit lower vertical cohesion. Additionally, rising economic inequality within countries is negatively related to levels of vertical cohesion. Furthermore, both the between-country and within-country associations for economic prosperity are weak, showing a positive association for the within-country effect only when controlling for political characteristics and migration.

For political variables, we observe noteworthy results. First, differences between countries in levels of democracy do not show a robust association with vertical cohesion; this effect loses statistical significance when controlling for the percentage of the migrant population in the country. However, changes within countries reduce vertical cohesion. In other words, countries that increase their levels of electoral democracy also experience declines in vertical cohesion. Second, the results for governance are robust. Both differences between countries and changes within countries show a positive and significant association with social cohesion.

Finally, differences between countries and changes within countries in the percentage of the migrant population do not show an association with levels of vertical cohesion.

Table 3: Multilevel regression for country level factors and social cohesion

	Horizontal			Vertical
	Model 1	Model 2	Model 3	Model 4	Model 5	Model 6
Time	-0.086**	-0.094**	-0.105***	-0.180***	-0.208***	-0.214***
	(0.031)	(0.031)	(0.030)	(0.040)	(0.038)	(0.038)
Gini (BE)	-1.811	-1.791	-1.767	-3.757*	-3.691*	-3.718*
	(2.320)	(1.959)	(2.054)	(1.901)	(1.637)	(1.697)
Gini (WE)	-3.401*	-3.538*	-3.503*	-6.648**	-5.916**	-5.876**
	(1.581)	(1.600)	(1.556)	(2.164)	(2.029)	(2.015)
GDP (BE)	0.198	0.161	0.120	0.258	0.223	0.207
	(0.176)	(0.189)	(0.229)	(0.149)	(0.165)	(0.194)
GDP (WE)	0.171	0.197	0.546	0.401	0.468	0.730*
	(0.277)	(0.276)	(0.294)	(0.368)	(0.342)	(0.372)
Democracy (BE)		-3.079*	-2.972*		-2.111*	-2.058
		(1.206)	(1.263)		(1.015)	(1.050)
Democracy (WE)		-0.830	-0.631		-3.792***	-3.632***
		(0.599)	(0.586)		(0.762)	(0.762)
Governace (BE)		0.995**	0.973**		0.756**	0.744**
		(0.319)	(0.334)		(0.267)	(0.277)
Governace (WE)		-0.085	-0.210		0.743*	0.640*
		(0.251)	(0.248)		(0.320)	(0.322)
Migration (BE)			0.019			0.007
			(0.050)			(0.041)
Migration (WE)			-0.111**			-0.090
			(0.038)			(0.050)
AIC	958631.141	958627.080	958631.660	846342.480	846318.283	846327.710
BIC	958815.995	958853.013	958878.132	846527.334	846544.215	846574.182
Log Likelihood	-479297.571	-479291.540	-479291.830	-423153.240	-423137.141	-423139.855
Num. obs.	213130	213130	213130	213130	213130	213130
Num. groups: country_wave	161	161	161	161	161	161
Num. groups: pais	22	22	22	22	22	22
Var: country_wave (Intercept)	0.091	0.090	0.085	0.178	0.151	0.149
Var: pais (Intercept)	0.202	0.136	0.148	0.121	0.084	0.090
Var: Residual	5.010	5.010	5.010	2.956	2.956	2.956
***p < 0.001; **p < 0.01; *p < 0.05

Interactions

fig-coefplot

Figure ?@fig-coefplot shows the coefficients for the between- and within-country effects for the country-level characteristics and the time trend (wave). When examining the between-country effects (BE × wave), the results indicate no significant differences for any of the socioeconomic, political, or migration variables.

In contrast, the within-country effects (WE × wave) reveal nuances in their relationship with social cohesion. For economic inequality, an increase in within-country inequality is associated with a decline in social cohesion in both the horizontal and vertical dimensions as time progresses.

Additionally, the effect of economic prosperity within countries becomes increasingly positive for horizontal cohesion over time. However, this pattern is not observed for vertical cohesion.

Regarding political variables, changes in levels of democracy do not show differences over time for either dimension of social cohesion. In contrast, the results for governance indicate that, as time passes, increases in governance become more positively associated with social cohesion.

Finally, for migration, the results show no significant changes in the within-country effect or its interaction with time for either dimension of social cohesion.

Table 6…

Table 4: Multilevel regression for macro factors and general social cohesion

Table 5

References

ANDREWS, R., & JILKE, S. (2015). Welfare States and Social Cohesion in Europe: Does Social Service Quality Matter? Journal of Social Policy, 45(1), 119–140. https://doi.org/10.1017/s0047279415000513

Araujo, K., & Martuccelli, D. (2014). Beyond institutional individualism: Agentic individualism and the individuation process in Chilean society. Current Sociology, 62(1), 24–40. https://doi.org/10.1177/0011392113512496

Araujo, K., Orchard, M., Rasse, A., & Stecher, A. (2022). Primer Informe de Resultados Encuesta Nacional de Autoridad NUMAAP 2021. NUMAAP.

Ariely, G. (2013). Does Diversity Erode Social Cohesion? Conceptual and Methodological Issues. Political Studies, 62(3), 573–595. https://doi.org/10.1111/1467-9248.12068

Bates, D., Mächler, M., Bolker, B., & Walker, S. (2015). Fitting Linear Mixed-Effects Models Using Lme4. Journal of Statistical Software, 67(1). https://doi.org/10.18637/jss.v067.i01

Brinks, D. M., Levitsky, S., & Murillo, M. V. (Eds.). (2020). The Politics of Institutional Weakness in Latin America (1st ed.). Cambridge University Press. https://doi.org/10.1017/9781108776608

Cabib, I., Miranda, D., Ormeño, J. P., & Moyano, D. (2025). Nuestras Trayectorias. Dinámicas intra-individuales e inter-individuales en Chile (2016-2023). https://doi.org/10.13140/RG.2.2.23888.62720

Castillo, J. C., Bonhomme, M., Miranda, D., & Iturra, J. (2023). Social cohesion and attitudinal changes toward migration: A longitudinal perspective amid the COVID-19 pandemic. Frontiers in Sociology, 7, 1009567. https://doi.org/10.3389/fsoc.2022.1009567

Castillo, J. C., Espinoza, V., & Barozet, E. (2022). Cohesión social en Chile en tiempos de cambio: indicadores, perfiles y factores asociados.

Castillo, J. C., Salgado, M., Carrasco, K., & Laffert, A. (2024). The Socialization of Meritocracy and Market Justice Preferences at School. Societies, 14(11), 214. https://doi.org/10.3390/soc14110214

Castillo, J.-C., Olivos, F., & Iturra, J. (2021). Conceptos y medición de cohesión social en proyectos internacionales (Documentos de {{Trabajo COES}} 47; pp. 1–37). COES.

Chan, J., To, H.-P., & Chan, E. (2006). Reconsidering social cohesion: Developing a definition and analytical framework for empirical research. Social Indicators Research. https://doi.org/10.1007/s11205-005-2118-1

Delhey, J., Boehnke, K., Dragolov, G., Ignácz, Z. S., Larsen, M., Lorenz, J., & Koch, M. (2018). Social Cohesion and Its Correlates: A Comparison of Western and Asian Societies. Comparative Sociology, 17(3-4), 426–455. https://doi.org/10.1163/15691330-12341468

Delhey, J., & Dragolov, G. (2016). Happier Together. Social Cohesion and Subjective Well-Being in Europe: HAPPIER TOGETHER-COHESION AND SWB. International Journal of Psychology, 51(3), 163–176. https://doi.org/10.1002/ijop.12149

Delhey, J., Dragolov, G., & Boehnke, K. (2023). Social Cohesion in International Comparison: A Review of Key Measures and Findings. KZfSS Kölner Zeitschrift Für Soziologie Und Sozialpsychologie, 75(S1), 95–120. https://doi.org/10.1007/s11577-023-00891-6

Dickes, P., & Valentova, M. (2013). Construction, Validation and Application of the Measurement of Social Cohesion in 47 European Countries and Regions. Social Indicators Research, 113(3), 827–846. https://doi.org/10.1007/s11205-012-0116-7

Dragolov, G., Ignácz, Z., Lorenz, J., Delhey, J., & Boehnke, K. (2013). Social Cohesion Radar Measuring Common Ground: An International Comparison of Social Cohesion Methods Report.

Fairbrother, M. (2014). Two Multilevel Modeling Techniques for Analyzing Comparative Longitudinal Survey Datasets. Political Science Research and Methods, 2(1), 119–140. https://doi.org/10.1017/psrm.2013.24

Gijsberts, M., van der Meer, T., & Dagevos, J. (2011). “Hunkering Down” in Multi-Ethnic Neighbourhoods? The Effects of Ethnic Diversity on Dimensions of Social Cohesion. European Sociological Review, 28(4), 527–537. https://doi.org/10.1093/esr/jcr022

Green, A., Janmaat, G., & Cheng, H. (2011). Social Cohesion: Converging and Diverging Trends. National Institute Economic Review, 215, R6–R22. https://doi.org/10.1177/0027950111401140

Janmaat, J. G. (2010). Social Cohesion as a Real-life Phenomenon: Assessing the Explanatory Power of the Universalist and Particularist Perspectives. Social Indicators Research, 100(1), 61–83. https://doi.org/10.1007/s11205-010-9604-9

Kaufmann, D., & Kraay, A. (2024). The Worldwide Governance Indicators (Policy {{Research Working Paper}} WPS10952). World Bank.

Kustov, A., & Pardelli, G. (2024). Beyond Diversity: The Role of State Capacity in Fostering Social Cohesion in Brazil. World Development, 180, 106625. https://doi.org/10.1016/j.worlddev.2024.106625

Langer, A., Stewart, F., Smedts, K., & Demarest, L. (2016). Conceptualising and Measuring Social Cohesion in Africa: Towards a Perceptions-Based Index. Social Indicators Research, 131(1), 321–343. https://doi.org/10.1007/s11205-016-1250-4

LAPOP LAb. (2023). AmericasBarometer.

Lê, F., Tracy, M., Norris, F. H., & Galea, S. (2013). Displacement, county social cohesion, and depression after a large-scale traumatic event. Social Psychiatry and Psychiatric Epidemiology. https://doi.org/10.1007/s00127-013-0698-7

Maerz, S. F., Esgell, A. B., Hellemeier, S., Illchenko, N., & Fox, L. (2025). Vdemdata: An R package to load, explore and work with the most recent V-Dem (Varieties of Democracy) dataset.

Meer, T. van der, & Tolsma, J. (2014). Ethnic Diversity and Its Effects on Social Cohesion. Annual Review of Sociology, 40(1), 459–478. https://doi.org/10.1146/annurev-soc-071913-043309

Ministerio de Desarrollo Social y Familia. (2020). Informe Final Consejo Asesor para la Cohesión Social. Diagnóstico para una aproximación a la cohesión social en Chile y recomendaciones para fortalecer el aporte de la política social.

Nazmi, A., Roux, A. D., Ranjit, N., Seeman, T. E., & Jenny, N. S. (2010). Cross-sectional and longitudinal associations of neighborhood characteristics with inflammatory markers: Findings from the multi-ethnic study of atherosclerosis. Health & Place. https://doi.org/10.1016/j.healthplace.2010.07.001

Njozela, L., Shaw, I., & Burns, J. (2016). Towards measuring social cohesion in South Africa.

O’Donnell, J., Guan, Q., & Prentice, T. (2024). Mapping Social Cohesion 2024 (Scanlon Foundation Research Institute).

Observatorio de Cohesión Social. (2025). Medición Cohesión Social América Latina. COES.

Reeskens, T., & Wright, M. (2012). Nationalism and the Cohesive Society. Comparative Political Studies, 46(2), 153–181. https://doi.org/10.1177/0010414012453033

Salazar-Xirinachs, J. M. (2023). Repensar, reimaginar, transformar: los “qué” y los “cómo” para avanzar hacia un modelo de desarrollo más productivo, inclusivo y sostenible. Revista de la CEPAL, 2023(141), 11–43. https://doi.org/10.18356/16820908-2023-141-2

Sánchez-Ancochea, D. (2021). The surprising reduction of inequality during a commodity boom: What do we learn from Latin America? Journal of Economic Policy Reform, 24(2), 95–118. https://doi.org/10.1080/17487870.2019.1628757

Schiefer, D., & van der Noll, J. (2017). The Essentials of Social Cohesion: A Literature Review. Social Indicators Research, 132(2), 579–603. https://doi.org/10.1007/s11205-016-1314-5

Schmidt-Catran, A. W., & Fairbrother, M. (2016). The Random Effects in Multilevel Models: Getting Them Wrong and Getting Them Right. European Sociological Review, 32(1), 23–38. https://doi.org/10.1093/esr/jcv090

Somma, N. M., & Valenzuela, E. (2015). Las Paradojas de La Cohesión Social En América Latina. Revista Del CLAD Reforma y Democracia, 61, 43–74.

Stafford, M., Bartley, M., Sacker, A., Marmot, M., Wilkinson, R., Boreham, R., & Thomas, R. S. (2003). Measuring the social environment: Social cohesion and material deprivation in english and scottish neighbourhoods. Environment and Planning a Economy and Space. https://doi.org/10.1068/a35257

Stefoni, C. (2018). Panorama de la migración internacional en América del Sur.

Toppi, H. P. (2018). Guillermo O’Donnell y su aporte al desarrollo de la democracia en América Latina desde la tercera ola de democratización. REVISTA IUS, 12(42). https://doi.org/10.35487/rius.v12i42.2018.407

Turchi, G. P., Bassi, D., Cavarzan, M., Camellini, T., Moro, C., & Orrù, L. (2023). Intervening on global emergencies: The value of human interactions for people’s health. Behavioral Sciences. https://doi.org/10.3390/bs13090735

United Nations Development Programme. (2023). Trapped: High Inequality and Low Growth in Latin America and the Caribbean: Regional Human Development Report 2021. United Nations. https://doi.org/10.18356/9789210057844