The Influence of Loneliness on Perceived Connectedness and Trust Beliefs – Longitudinal Evidence from the Netherlands

Loneliness, what it is and what it does

Similar to hunger or thirst, loneliness is a universal human experience and, consequentially, the topic of countless philosophical and cultural writings (Svendsen, 2017). Despite its cultural prominence, the notion of loneliness suffers from multiple misconceptions and gets confused with related concepts in public as well as in academia. Therefore, it is important to discuss in more detail what loneliness is, how it relates to objective social relationships and activities, and what is known about its consequences.

As reliable relationships were a central resource for survival throughout history, humans have developed a strong desire to form and maintain reliable social relationships and belong to a larger social group (Baumeister & Leary, 1995; Gere & Macdonald, 2010). Loneliness stems from a perceived discrepancy between one’s desired and actual relationships and can be understood as the emotional expression of a perceived insufficiency in personal social relationships, either in respect of quantitative or qualitative aspects (Franklin & Tranter, 2021; Perlman & Peplau, 1981). The fields of psychology and neuroscience established that loneliness developed as a painful emotional warning signal that motivates humans to re-affiliate with others to maintain a protective relationship network (Cacioppo & Cacioppo, 2018; Qualter et al., 2015), and many scientists concerned with human motivation consider affiliation as one of the most important drivers of human action (Kovač, 2016).

This close conceptual relationship between loneliness and social relationships led to the common misconception that loneliness is a synonym for aloneness. Although conceptually intermingled at times, loneliness and being alone are distinct empirical phenomena and loneliness is more dependent on the quality, rather than quantity, of social relationships and social activities (de Jong Gierveld et al., 2018). For instance, research repeatedly found statistically significant, but modest correlations between social isolation, social activity and loneliness (Coyle & Dugan, 2012; Russell et al., 2012). These findings manifest themselves in an everyday observation: people can feel lonely despite being in company but feel happy with very few contacts. In contrast, social connectedness and social embeddedness are often used interchangeably and refer to the objective social network of individuals (Reilly, 2017). Simply put, aloneness and connectedness are physical states of being, but loneliness is a perception and as such a state of mind.

Due to its motivating function to re-affiliate, loneliness is a useful emotion in the short term. However, being unresolved, chronic loneliness has severe negative consequences. Keeping in mind that loneliness is a warning signal for seemingly insufficient or eroding social relationships, it is of little surprise that chronically lonely individuals tend to experience various negative emotions during social encounters such as anxiousness, insecurity, and risk aversion coupled with a self-preservation mindset (Qualter et al., 2015; Spithoven et al., 2017).

Given that social ties provide security, loneliness is associated with increased stress which results in an elevated threat sensitivity, anxiety and risk avoidance (Cacioppo & Cacioppo, 2018). Mediated through these stress reactions, loneliness is known to cause reduced sleep quality and unhealthy coping behaviours, which results in severe consequences for health and psychological well-being. These outcomes include dementia, depression, cardiovascular diseases, and reduced life expectancy (Hawkley & Cacioppo, 2010; Lim et al., 2020). This led scientists to promote loneliness first and foremost as an issue for public health (Holt-Lunstad et al., 2017).

However, as elaborated in the former section, there is a lack of research concerned with the effect of loneliness on other issues of societal interest. In light of survey results suggesting staggering rates of loneliness in the US and Europe, this might be a crucial shortcoming. For instance, US American surveys found that roughly 35% of adults aged 45 or older report to feel lonely, a trend that likely worsened during the Covid-19 pandemic (National Academies of Sciences Engineering and Medicine, 2020). Likewise, survey data from the adult British population indicate that roughly 14% often or always feels lonely (Jo Cox Commission on Loneliness, 2017). This is mirrored by data gathered in Germany in 2013 and 2017 which indicate that about 10% of the adult population considered themselves frequently lonely (Eyerund & Orth, 2019). Using data from the European Social Survey, social scientists found that loneliness is a substantial issue in almost all European countries with up to 34% of individuals reporting feeling lonely (Yang & Victor, 2011). Global Meta-Analysis and several European studies show the global scale of the issue (Ernst et al., 2022). These findings suggest that a potential influence of loneliness on trust beliefs and connectedness might be of considerable scale.

Loneliness, connectedness and trust beliefs

As reviewed in the preceding section, loneliness is a response to insufficient or eroding social relationships. This is accompanied by feelings of anxiousness, insecurity, and risk aversion (Spithoven et al., 2017). Spithoven and colleagues also conclude that loneliness alters social perception and behaviour as soon as it becomes a chronic state. While situational loneliness motivates individuals to seek out social interactions, repeated failure to overcome their loneliness causes additional fear of negative social interactions and, ultimately, the paradoxical behaviour of social withdrawal (Qualter et al., 2015).

This dynamic likely radiates from the sense of being part of the immediate social community on the perception of belonging in society in general (Pretty et al., 1994). As elaborated before, humans have a strong need for social belonging (Baumeister & Leary, 1995). This does not only refer to individual contacts but also the person’s perception of whether they fit into the broader social setting (for instance, the village or city) and to be an integral part of society (Franklin & Tranter, 2021). Hence, feeling lonely logically relates to increased uncertainty about the own social standing, the reliability of others, and to what degree one has a generalized perceived connection to others. Qualitative studies of loneliness in marginalized groups suggest that a sense of disconnectedness from the greater society and other people is part of experiencing loneliness (Bower et al., 2018; Rokach, 2014). Likewise, studies investigating war veterans highlight the link between loneliness and disconnectedness from society (Stein & Tuval-Mashiach, 2015). However, it is an open question whether these insights can be replicated with survey data of the general population.

Likewise, the prolonged experience of feeling lonely likely fosters distrust against other people. As discussed, lonely individuals tend to grow anxious and insecure in social settings and interpret social encounters more negatively (Spithoven et al., 2017). This likely corresponds with a higher probability to experience negative social situations and, in the long run, becoming more distrusting. Unsurprisingly, observational studies found that loneliness correlated with interpersonal distrust in multiple countries and with various operationalizations (Nyqvist et al., 2016; Qualter et al., 2009; Rotenberg, 1994). However, loneliness and trust beliefs are associated with time-invariant characteristics such as personality traits and correlations potentially due to self-selection (Buecker et al., 2020). Studies that experimentally manipulate situational loneliness support the idea that loneliness fosters anxiety (Cacioppo et al., 2006), but given that experimental manipulation of chronic loneliness is not feasible and cross-sectional data do not observe changes within individuals, it is unclear whether within-person changes in loneliness over longer timespans affects social distrust.

Finally, if loneliness causes distrust and disconnectedness, it potentially exerts an influence on the citizens’ trust towards political actors and institutions as well. If chronically lonely individuals evaluate other people more negatively and become more distrustful in general, the idea that this might lead to generalized distrust against politicians and political organizations as well is not far-fetched. While several studies could not find a link between social and political trust (Uslaner, 2017), more recent studies found an association (Newton & Zmerli, 2011). Likewise, longitudinal evidence from Denmark suggests a bidirectional relationship between social and political trust (Sønderskov & Dinesen, 2016).

All this supports the idea that loneliness potentially influences social as well as political trust. However, lonely individuals show these attitudinal changes out of latent insecurity and fear of further social exclusions. Political actors and institutions are more abstract concepts for most individuals. Whether the social distrust radiates to a general distrust against political actors and their spheres of action (i.e., political institutions) is an open empirical question. There are a couple of multinational cross-sectional studies that find the association between loneliness and political trust using the European Social Survey (Schobin, 2018; Yang, 2019). However, the theoretical relationship between loneliness and political distrust is less clear compared to the other two discussed considered outcomes and suffers from the same issue that empirical findings are mostly based on cross-sectional data.

Taken together, despite a strong argument why loneliness likely causes insecurity and anxiety that, in turn, cause distrust and a reduced sense of connectedness, the empirical evidence is fairly limited. Therefore, this study tests the hypotheses that loneliness is associated with reduced connectedness, and social as well as political trust.

H1

Increases in loneliness are associated with a decrease in perceived connectedness.

H2

Increases in loneliness are associated with a decrease in social trust beliefs.

H3

Increases in loneliness are associated with a decrease in political trust beliefs.

Data

The data utilized in this study are administered by the CentERdata Institute for Data Collection and Research, funded by the Dutch governmental organization for scientific research. The “Longitudinal Internet Studies for the Social Sciences”, hereafter LISS, is the central panel data project of the “Measurement and Experimentation in the Social Sciences” (MESS) project and is openly available for scientists. The survey is a true probability sample of the registered population in the Netherlands. Although the LISS is mainly organized as an internet survey, the institute is gathering data on people without internet connection. Participants without computer or internet access are provided with devices so that they can participate. Therefore, the survey collection mode does reduce the potential bias of the sample. In the context of this study, this is particularly important because older age groups and low-income households are less likely to have internet access and show higher probabilities to suffer from loneliness (Hawkley et al., 2020; van Deursen & van Dijk, 2019).

Importantly, loneliness as well as perceived societal connectedness and trust beliefs are fairly stable over time (Marien, 2011; Mund et al., 2020). Therefore, long-running panel models are needed to track within-person variations in the central variables of interest. Fielded the first time in 2008, the panel gathers data annually with the most recent wave fielded in October 2020. Furthermore, the LISS continuously included a 6-item version of the de Jong Gierveld loneliness scale since the first observation period. As loneliness is a complex latent construct that is best captured with specific validated scales (Marangoni & Ickes, 1989), the long-running panel is uniquely suitable for the purpose of this study.

This study is based on 12 out of 13 waves gathered between 2008 and 2020. The fifth wave (2012) was not included in the analysis as the variables measuring loneliness varied in their operationalization in this specific wave. The panel itself is unbalanced, with permanent and temporary drop outs (average participation duration is 4.17 years). After excluding underage participants from the analysis (n = 803) and list-wise deletion of missing cases, the final analysis sample contains 41,508 observations from 9954 individuals.

Measures

Control variables

As elaborated in the theory section, many studies attempt to explain trust and perceived connectedness with measures of objective social inclusion. To ensure that the observed relationship between loneliness and the outcomes are not mere symptoms of limited social contacts, I control for membership in civic organizations as well as interaction frequency with family and friends, both measured on a seven-point scale. Other potential confounders are life events such as divorce (van Tilburg et al., 2015). To account for major life events that trigger loneliness, I include dummies measuring widowhood, divorce, separation from a partner, being married, and whether a person lives with their partner in the same household.

Furthermore, the analysis accounts for several socio-demographic variables. First, the study controls for sex (male vs. female) and level of education (“primary school”, “intermediate secondary education/junior high school”, “higher secondary education”, “intermediate vocational education/junior college”, and “university”). Furthermore, to account for age effects, the models control for the age categories “18–24 years”, “25–34 years”, “35–44 years”, “45–54 years”, “55–64 years”, and “65 years and older”. Also, as loneliness is correlated with poor health, I include a self-reported measure of general health, ranging from one to five, with high values indicating good health. Likewise, financial hardship can increase political and social distrust and is a predictor of social exclusion. To account for this, I include a variable measuring subjective satisfaction with the personal financial situation. The subjective measure was preferred over self-reported income because questions about self-reported income tend to have very high proportions of missing values and are biased due to social desirability (Hariri & Lassen, 2017; Korinek et al., 2006). Moreover, financial satisfaction captures financial hardship better, as it accounts for household size, changing living standards, and other social forces such as social comparisons.

Finally, longitudinal data might show period effects, due to specific events in time. This is often tackled by including arbitrary time dummies, but this can cause biases and introduce issues in respect of effect interpretation (Kropko & Kubinec, 2020). Following recommendations from the literature (Brüderl & Ludwig, 2014), the analysis does not rely on arbitrary time dummies but includes specific, theory-driven period events. It’s worth noting, however, that the results do not change under the model specification with time dummies instead of event dummies.

First, the Netherlands elected their parliament three times during this period. As explained in the description of the dataset, this wave from the 2012 election had to be dropped due to inconsistent measurement of loneliness. Therefore, the analysis considers only two election periods. As election periods are often accompanied by negative media reports and aggressive campaigns, scholars suspect that this affects trust beliefs (Bovens & Wille, 2008). To account for this, the analysis contains a dummy variable indicating whether an observation was gathered during an election year.

Furthermore, the European Union experienced the so-called refugee crisis as well as the European sovereign debt crisis during the survey period. Both events are in a way ongoing and it is impossible to clearly define an official start and end to them. However, the Netherlands, similar to most other European countries, experienced record numbers of migrants and asylum seekers during 2015/16 (Tolsma et al., 2021). This is mirrored by the frequency of related media articles published in print media, which were highest in 2015 and 2016 (Czymara & Klingeren, 2021). Hence, a dummy indicating the years 2015–2016 is included. In respect of the fiscal crisis, the sovereign fiscal crisis is typically considered to have started in 2010 and ended with the introduction of the central banks’ Outright Monetary Transactions program that effectively ended immediate concerns about speculations against the currency (Ehrmann & Fratzscher, 2017). Table 1 provides an overview of the descriptive statistics of all variables.

OPEN IN VIEWER

Table 1.

Descriptive statistics (pooled).

n = 9954; N = 41508	Mean	Std	Min	Max
Social trust	6.14	2.13	0	10
Political trust	5.44	1.77	0	10
Sense of connectedness	6.02	2.52	0	10
Loneliness	2.84	2.40	0	12
Female	0.51	—	0	1
Age categories
18–24	0.07	—	0	1
25–34	0.11	—	0	1
35–44	0.14	—	0	1
45–54	0.19	—	0	1
55–64	0.23	—	0	1
65+	0.27	—	0	1
Education
Primary	0.07	—	0	1
Secondary	0.23	—	0	1
Higher secondary	0.11	—	0	1
Vocational	0.23	—	0	1
Higher vocational	0.25	—	0	1
University	0.10	—	0	1
Contact friends	3.60	1.46	1	7
Contact family	4.44	1.41	1	7
Financial satisfaction	6.89	1.73	0	10
Health	3.12	0.76	1	5
Living with partner
Yes	0.73	—	0	1
Civic status
Never been married (ref)	—	—	0	1
Being widowed	0.06	—	0	1
Being divorced	0.10	—	0	1
Being separated	0.01	—	0	1
Being married	0.59	—	0	1
Group membership	0.07	—	0	1
Time periods
Election year	0.17	—	0	1
Migration crisis	0.18	—	0	1
Fiscal crisis	0.18	—	0	1

Statistical analysis plan

For the analysis, I estimate between-person effects (BE) as well as fixed, within-person variability, effects (FE) panel models (Bell & Jones, 2015; Brüderl & Ludwig, 2014). To investigate the hypotheses, the analyses will mainly focus on FE models, which control for unobserved heterogeneity by design. However, as an additional analytical step BE models are calculated as well. This is done for several reasons: First, comparisons between individuals are better suited to describe differences between sociodemographic groups, while FE models are preferable for causal inference (Brüderl & Ludwig, 2014). As loneliness and trust beliefs are fairly robust over time, BE are especially useful to describe more permanent differences between the most and least lonely parts of the population. Also, given that most previous research is based on cross-sectional data, that is between-person comparisons, the BE models can be seen as replications of these results.

Still, BE models suffer from issues of self-selection and confounding from unobserved individual characteristics (Andreß et al., 2013). As FE models only use within-person variation to estimate coefficients, they are particularly useful to account for these shortcomings. Therefore, to show the dynamic part of the relationship and to ensure that the relationship is not biased by time-invariant confounders, FE models are calculated. Please note that time-invariant variables such as biological sex are omitted in the FE models by design. The FE models are calculated with cluster robust standard errors.

After establishing the relationship using the models described above, the analysis concludes by investigating the possibility of a reverse causal relationship between loneliness and the outcomes. In contrast to experimental designs, observational data struggle inherently with issues related to reverse causality (Leszczensky & Wolbring, 2019). It is plausible that not only loneliness affects trust beliefs and sense of connectedness, but that the individuals’ loneliness is itself influenced by the outcomes. That is, a person growing less trusting is likely becoming lonelier (Rapolienė & Aartsen, 2021); likewise, a person that feels disconnected might become lonelier as well. In order to establish that variation in loneliness actually causes a change in the outcome variables, this potential reciprocal dependency needs to be accounted for. Traditionally, this issue is approached with random intercept cross-lagged panel models. However, these models came under criticism for their strict model assumptions and the, consequentially, often biased estimates (Allison et al., 2017; Mund & Nestler, 2019). Recent advances in statistical inference allow to control for the issue of reverse causality by combining panel fixed effect estimation with cross-lagged panel models (Allison et al., 2017; Leszczensky & Wolbring, 2019). To take this into account, I used the Stata command xtdpdml to compute linear dynamic panel-data models using full information maximum likelihood and robust standard errors (Williams et al., 2018). This approach combines the advantages of fixed-effect estimation (accounts for unobserved heterogeneity) with cross-lagged panel models (solves the issue of reverse dependency between x and y). With that, the estimates control for the effect of the outcome variable on the dependent variable, that is loneliness.

The findings

As elaborated in the preceding section, I estimate between-effects (BE) as well as fixed-effects (FE) panel models to investigate the relationship between loneliness and the outcomes. Figure 2 summarizes the effects of interest out of six regression models. The full regression table can be found in Appendix Tables A2 and A3.

OPEN IN VIEWER

Figure 2.

Effect of loneliness on social trust, political trust, and sense of connectedness.

The first three coefficients report the between-effect of loneliness on the sense of connectedness, social trust, and political trust. On average, every additional point of loneliness is associated with a lower perceived connectedness of roughly 0.2 points (p ≤ 0.001). Over the full range of the loneliness scale, this accumulates to a difference of roughly 2.39 points on the 10-point connectedness scale between the most and the least lonely individuals. Likewise, loneliness is negatively associated with social trust (b = −0.153; p ≤ 0.001) as well as political trust (b = −0.059; p ≤ 0.001). Considering the scale of the variables, this accumulates to a difference between the most and the least lonely of 1.824 and 0.708, respectively. This confirms the previously found relationships: lonely individuals are substantially more likely to feel disconnected and to be socially and politically distrusting compared to less lonely individuals.

However, these effects are still confounded by self-selection and time-invariant unobserved confounders. This is especially important in the context of loneliness, as personality characteristics such as extraversion, agreeableness, conscientiousness, neuroticism, and openness are known predictors for loneliness (Buecker et al., 2020). To account for this, the last three coefficients reported in Figure 2 are based on panel fixed effects which are not biased by time-invariant unobserved confounders.

The within-person effects shrink considerably in size compared to the between effect coefficients, but remain statistically significant in two out of three models. With respect to the effect size, every increase in loneliness is, on average, associated with a decrease in perceived connectedness by roughly 0.059 (p ≤ 0.001), accumulating to a difference of −0.708 points on the 10-point connectedness scale between the most and the least lonely. With respect to trust beliefs, every additional point in loneliness is associated with a reduction in social trust of 0.034 (p ≤ 0.001). The effect of loneliness on political trust, however, is no longer statistically significant (p = 0.393).

The results so far suggest that rising loneliness causes reduced social trust and a diminished sense of connectedness. In contrast, the results provide just limited support for the effect of loneliness on political trust. As discussed in the section “Statistical Analysis Plan”, the following part of the analysis is concerned with the possible bias caused by a dependence of loneliness from the outcome variables themselves.

As summarized in Table 2, within-person variation of loneliness at t and at t-1 predicts social trust beliefs statistically significant. With respect to perceived connectedness, loneliness at t has a significant negative effect, while the coefficient of loneliness at t-1 shows the expected direction but is not significant. Substantially, this indicates that the effect of loneliness on the outcomes is robust even under consideration of a reverse influence from the outcomes on loneliness. The effect of loneliness on social trust seems to be particularly strong and lasting, given that loneliness exerts a statistically significant influence on social trust up to 2 years later.

OPEN IN VIEWER

Table 2.

Cross-lagged fixed effect regression using full maximum likelihood.

	Connected	Connected	Social trust	Social trust
Explanatory variable
Loneliness	−0.056***	−0.063***	−0.033***	−0.037***
Loneliness t-1	−0.005	−0.006	−0.027**	−0.249**
Loneliness t-2		−0.008		−0.017*
Auto-regressive effect on outcome
Connectedness	0.089***	0.079***	—	—
Social trust	—	—	0.159***	0.156***
BIC	344,904	327,762	323,023	307,473

To conclude, the analysis supports Hypotheses 1 and 2, namely that loneliness erodes perceived connectedness and social trust. In contrast, the analysis finds only limited support for Hypothesis 3. While it is still a possibility that the effect of loneliness on political trust is too small to be detected on a within-person level, the significant BE and insignificant FE models suggest that the relationship between loneliness and political trust potentially stems from self-selection or unobserved confounders.

Robustness tests – alternative model specifications

To get a better understanding of the robustness of the results, I calculated several alternative model specifications. First, the reported results are based on adults only. Including the excluded underage participants does not change the results.

Secondly, the last wave was fielded in October 2020, a time the COVID-19 pandemic likely had a strong exogenous effect on loneliness as well as perceived social solidarity and political trust. However, removing this wave from the analysis does not change the results either.

Third, while panel fixed effects are powerful tools for inference statistics, they can be biased if the exogeneity assumption is not met. Intuitively speaking, if unobserved traits such as maturation or learning effects do influence the effect of loneliness on the outcome of interest, these estimates can be biased (Wooldridge, 2011). The potential issue that the effect of loneliness on beliefs and perceived connectedness might change based on the duration/chronicity of loneliness stands to reason. Fixed effects with individual slopes relax the exogeneity assumption and serve as a remedy of this problem (Brüderl & Ludwig, 2014). Using the STATA program XTFEIS (Ludwig, 2019), repeating the analysis using panel fixed effects with individual slopes leads to the same conclusions. All robustness tests are reported in the online appendix.

Discussion

The central goal of this study is to investigate whether perceived loneliness erodes people’s generalized sense of connectedness and political and social trust beliefs. With that, the study sheds light on the influence of the subjective perception of the personal social relationships on three indicators of perceived social cohesion. In accordance with Hypotheses 1 and 2, the results show that rising loneliness is indeed negatively associated with social trust and perceived cohesion, a relationship that was neglected previously. In respect of political distrust, the results are mixed as the relationship can be found in between-person comparisons, but not in fixed-effects models. Substantially, this means that lonely individuals are, on average, less trusting toward politicians and political institutions, compared to not lonely individuals. However, changes in the loneliness levels did not correlate with changes in political distrust, either because the effect is too small to be detected or because it is not present at all. It is noteworthy that out of the three outcome variables considered in this study, political trust has the lowest variation over time across the observed time period (SD = 0.86). Hence, the relationship between loneliness and political trust may be simply too small to be detected. Still, this finding indicates that the negative influence of loneliness on trust is not generalizable to political actors and institutions.

The results have important implications for existing as well as upcoming empirical research concerned with social capital, social cohesion, and political attitudes formation. The analysis supports the idea that loneliness exerts an influence on political attitudes and behaviour, potentially partially mediated through the three investigated outcomes. For instance, studies have found that loneliness is associated with a reduced sense of duty to vote and lower self-reported voter turnout (Langenkamp, 2021). Assuming lonely individuals feel less connected to other people and therefore not as an incremental part of society, this potentially explains a reduced perceived moral obligation to participate in the democratic process. Likewise, findings that loneliness is associated with a higher probability to sympathize with conservative beliefs, hold more xenophobic attitudes, and an increased probability to vote for populist parties, might directly relate to the enhanced distrust and anxiety of lonely individuals (Floyd, 2017; Langenkamp & Bienstman, 2022).

Furthermore, the sense of generalized connectedness reflects on a more general perceived unity and interdependence with others (Cojuharenco et al., 2016). A lack of feeling connected with friends while feeling lonely is trivial, the question whether lonely individuals feel disconnected from wider society is less so. Importantly, a strong sense of social connectedness is associated with pro-social behaviour and contributing to public goods such as activity in pro-environmental organizations (Cojuharenco et al., 2016) and higher salience of social values and cooperation (Triandis, 2018; Utz, 2004). Hence, loneliness potentially threats these societal resources as well. Upcoming research might investigate whether a potential negative influence of loneliness on these outcomes is mediated through social trust and the sense to connectedness. Likewise, upcoming studies should investigate which moderating factors might lead to variability in the association between loneliness and political attitudes. For instance, the anxiety and distrust triggered by loneliness might radiate more strongly onto attitudes about politicians for individuals with a high political interest as political representatives and topics are more salient for this population. Same might be true for individuals with a high political efficacy.

In the greater theoretical picture, the study speaks to the debate of how (and which) emotions and social relationship characteristics influence social behaviour and cognition. Research established that objective social interaction, social organizations, and network characteristics foster trust beliefs and cohesion. The current paper adds to this literature and suggests that loneliness plays an independent role in predicting indicators of social capital/social cohesion and is a useful additional measurement complementing these more established constructs.

To deepen our understanding of the underlying mechanisms, it will be useful for upcoming studies to take a more nuanced take on loneliness and investigate how different subdimensions of loneliness such as emotional, social, and cultural/collective loneliness relates to the considered outcomes. This will allow identifying which specific unfulfilled social needs cause what kind of social and political attitudes. This will provide not only a deeper theoretical understanding of the mechanisms but also a fruitful basis for specifically designed political interventions.

For instance, publicly funded social places such as parks, car-free city centres, or community centres can reduce loneliness significantly and increase well-being (Astell-Burt et al., 2022; Bell et al., 2014). It is unclear, however, to what extent these interventions reduce different sub-dimensions of loneliness. On a theoretical level, socio-structural interventions such as community centres providing social support are more likely to reduce social loneliness, that is the perceived lack of an adequate social support network. In contrast, emotional loneliness refers to the absence of satisfactory intimate attachment figures less unlikely to be met this way. Exploring which social needs are associated with political outcomes, such as non-voting or radical right-wing support, will allow designing interventions against loneliness with democracy promotion in mind.

However, the results and conclusions of this study have to be interpreted in light of some limitations. First, while the within-person analysis ensures that the effects is not confounded by time-invariant variables, the threat of time-variant confounders remains. To reduce this risk, the analysis took major life events such as changes in family life, unemployment, and changes in financial satisfaction into account. Still, confounding remains an inherent issue of any observational analysis. Secondly, loneliness is a stigmatized and painful emotional state. Therefore, self-reported loneliness might suffer from under-reporting, a potential bias that is difficult to assess. Third, the study relies on single-item measures of trust and connectedness. Similar to the measurement loneliness, these latent constructs should ideally be captured with multi-item scales to improve their reliability. Finally, the survey did not contain information about participants’ race, gender identity, or sexual orientation. Given that effect heterogeneity is possible for these and other demographic groups, upcoming studies should replicate the findings for these demographic strata.

In response to these potential shortcomings, I like to add that research does not happen in a vacuum and results need to be interpreted in the context of the literature. For instance, eye-tracking studies indicate that lonely children have enhanced hypervigilance for social threats and studies using a prisoners dilemma paradigm found that lonely individuals grow less trusting faster if they get crossed in the game (Qualter et al., 2013; Rotenberg, 1994). This supports the conclusion of this study that lonely individuals become distrusting based on experiences faster compared to non-lonely individuals. Likewise, attempts to manipulate state loneliness experimentally suggest that loneliness causes anxiety, anger and fear of negative evaluation (Cacioppo et al., 2006). Other methodological approaches come to similar results Qualitative research of marginalised groups suggests that feeling disconnected from society is part of the loneliness experience (Rokach, 2014; Stein & Tuval-Mashiach, 2015). These and other studies support the idea that loneliness fosters distrust and alienation which make spurious correlation unlikely.

In respect of the threat of social desirability bias, one has to keep in mind that the used loneliness scale measures loneliness indirectly and was developed with that issue in mind. Furthermore, the scale is one of the most broadly validated scales in the field (de Jong Gierveld & van Tilburg, 2010).

Taken together, while all the mentioned limitations of this study have to be taken seriously, this study in conjunction with the empirical context allows us to conclude with some confidence that loneliness causes a reduction in social connectedness and shrinking social trust beliefs. With that, loneliness likely has serious implications for political attitude formation and behaviour.