Friends make us happy: evidence from three European panel studies

Many studies suggest that having friends and being embedded into social networks is important for individuals’ life satisfaction (e.g. Bartolini & Sarracino, 2014; Rodríguez-Pose & Berlepsch, 2014) and subjective health (e.g. Arezzo & Guidici, 2017; Giordano & Lindstrom, 2010; Oshio, 2016; Rodgers et al., 2019; Xue et al., 2020). A network of friends is an important component of social capital. According to a widely used definition, social capital refers to “the aggregate of the actual or potential resources which are linked to possession of a durable network of more or less institutionalized relationships of mutual acquaintance or recognition” (Bourdieu, 1986: 248; for similar definitions see also Coleman, 1988, 1990, and Putnam, 1995, 2000). Social capital is not only important for individuals’ well-being, but also for societies as a whole, since it supports solidarity and social cohesion. Social capital can also contribute to economic prosperity and the functioning of democracy (e.g. Franzen & Botzen, 2011; Jungbauer-Gans, 2002; Putnam, 2000). In his famous publication “Bowling alone”, Robert D. Putnam (1995) diagnosed a general decline in social capital in the United States of America since the 1960s. His results were much discussed and partly refuted by more recent studies (e.g., Clark, 2015; Jungbauer-Gans, 2002; Paxton, 1999). However, modern societies face the threat of a decline in the density of social relations. This tendency is due to various developments. First, most modern societies face a decline in the size and stability of family relations, e.g. declining fertility rates and increasing divorce rates. Smaller familial networks and the stress of family dissolution often also reduce other network relations. Second, growing prosperity leads to more and more single-person households, since individuals can increasingly afford to live by themselves. Third, the digitalization of society provides the possibility of working more flexibly from a home-based office. Digitalization can also increase the amount of time spent on social media and may prevent individuals, particularly the younger generations, from spending their leisure time with real social contacts (e.g. Franzen, 2000). All of these developments have the potential to decrease personal face-to-face contacts and hence may reduce individuals’ social networks. Last, but not least, many countries implemented social distancing policies in response to the COVID-19 pandemic, which may also have resulted in a decrease in social capital. The pandemic and the policies reducing social contacts ended in 2021, but they might nevertheless have left a scar in individuals’ social networks.

The purpose of this study is twofold. First, we provide an update on the development of social capital in Germany, Switzerland, and the United Kingdom. These countries have launched longitudinal studies that track some aspects of individuals’ social networks, such as the number of close friends, for more than two decades. We are particularly interested in whether there is a trend towards smaller networks visible in the data, particularly in the aftermath of the COVID-19 pandemic. Second, we investigate whether social networks contribute to individuals’ well-being. Most studies on the effect of social capital on subjective well-being suggest that a network of close friends increases life satisfaction (e.g., Lucchini et al., 2015) and makes people healthier (Rodgers et al., 2019). There are several reasons why friends should increase subjective well-being. First, the search for social approval is a fundamental need of humans. This notion dates back to Adam Smith (1759) and is mirrored by many scholars, for instance by Abraham Maslow (1942). Friends are one important way to attain this warm glow of social contact and social approval. Positive social contacts cause physiological reactions like the production of oxytocin, and endorphins. These hormones increase individuals’ subjective well-being (Colonnello et al., 2017). Second, friends, and particularly close friends, provide social support, both emotionally and practically (Coleman, 1988; Pancheva & Vásquez, 2022). This includes many relevant realms like education, economic success and physical and mental health and thereby increases subjective well-being through many different mechanisms. However, there are only a few studies exploiting individual-level national panel data and utilizing causal modeling approaches. More precisely, there are only two studies on the effect of network resources on life satisfaction using general population panel data and applying fixed effects regression models (Lucchini et al., 2015 for Switzerland, Pancheva & Vásquez, 2022 for the United States of America).

Lucchini et al. (2015) use yearly data of the Swiss Household Panel (SHP) from 2002 to 2009 and apply fixed effects (FE) panel regression models to account for selection on time-constant individual unobservables. The authors follow a y-centered, so-called “kitchen sink” approach including more than 30 parameters to “control” for many factors that may be associated with happiness. Though widely used, this procedure may induce an overcontrol bias, because the models do not only include potentially confounding factors, but also colliders (e.g., Morgan & Winship, 2014). Lucchini et al. (2015) find that a doubling in the number of close friends increases life satisfaction (measured on an 11-point-scale from 0 “not at all satisfied” to 10 “completely satisfied”) by 0.05 scale points for men and 0.08 scale points for women.

Pancheva and Vásquez (2022) use three waves (1995/1996, 2004/2005, and 2012/2013) from the “Midlife in the U.S.” (MIDUS) panel study and apply FE models. The authors follow an x-centered approach and cover many dimensions of social capital. In general, Pancheva and Vásquez (2022) find that some indicators of social capital are associated with life satisfaction, while the frequency of contact with friends is not related to it. However, their study does not analyze the effect of the number of close friends. Moreover, it is limited regarding the number of surveyed years (three), which may be an explanation for the absence of a substantial effect of the frequency of contact with friends in their analysis.

Furthermore, both studies (Lucchini et al., 2015; Pancheva & Vásquez, 2022) do not adequately account for potential bias introduced by selection on growth or by feedback processes. In this study, we address these shortcomings. We demonstrate that the average number of close friends has been relatively stable in the three European countries Switzerland (CH), Germany (DE), and the United Kingdom (UK) over the past two decades. Furthermore, using national individual-level panel data (German Socio-economic Panel (GSOEP), Swiss Household Panel (SHP), and United Kingdom Household Longitudinal Study (UKHLS)), we show that the number of friends has a small but robust positive effect on happiness. To better identify causal effects, we apply panel regression models that account for the potential influence of unobserved individual time-constant heterogeneity (FE models) and time-varying heterogeneity (fixed effects with individual slopes (FEIS) models) as well as feedback processes (first-difference (FD) models).

The remainder of this paper proceeds as follows: Section two describes the data used and the methods applied. Section 3 presents the results of the data analyses. Finally, Sect. 4 concludes and discusses the results.

Figure 2 demonstrates that the average number of close friends in Germany, Switzerland, and the United Kingdom has been relatively stable during the past two decades lying at around five for Switzerland and the UK, and between four and five for Germany. Hence, we do not find evidence that social capital as measured by the number of close friends declined in these three European countries.

Applying the explained FE, FEIS, and FD models we consistently observe a small but robust positive effect of the number of close friends on happiness (see Fig. 3 and Table 3). The effect is highest for the GSOEP followed by the UKHLS, and lowest for the SHP. A doubling in the number of close friends in Germany yields an increase in life satisfaction of 0.11 standard deviations (sd). This is a very small effect. In the United Kingdom a doubling in the number of friends increases subjective well-being by 0.08 sd and in Switzerland this effect is only half the size (0.04 sd for each doubling of friends).

Bild vergrößern

These effects remain stable in size and statistical significance when applying FEIS or FD instead of FE models. The only exception is the UKHLS, where the effect in the FEIS model is less than half the size of the effect in the FE model (0.03 n.s. vs 0.08***). Hence, the results show that selection on growth and feedback do not substantially affect the relation between the number of friends and life satisfaction. Only in the UK there seems to be an influence of selection on growth. Hence, individuals in the UK with a steeper happiness trend may select into having more friends as compared to subjects with a less steep happiness trend.

To investigate the robustness of these results, we analysed the potential influence of an omitted variable bias from selection on observables following the procedure by Young and Holsteen (2017), which also includes models with no confounders, and from both selection on observables and unobservables following the procedure suggested by Frank (2000). Neither test shows that the inference presented in Fig. 3 is substantially affected by selection on observables or unobservables. Furthermore, we recalculated all models excluding the variable health, because it might be a mediator biasing the estimate of the total effect of friends on life satisfaction. However, the estimation results do not differ substantially. Hence, health is not a mediator in our samples. In order to mitigate potential bias induced by influential cases, we reran all models limiting the value range of the number of close friends between 1 and 20. Limiting the distribution does not affect the results in any substantial way. Furthermore, we also reanalysed all models using dummy variables for each number of friends instead of taking the natural logarithm. The resulting pattern confirms our assumption of a positive but marginally decreasing effect of the number of friends on life satisfaction. In addition, one might also be concerned with the question whether the effect of friendships is asymmetrical, meaning that gaining friends might affect life satisfaction differently from losing friends. To investigate this question, we used asymmetric fixed effects regressions (Allison, 2019). The results suggest that the effect is symmetric. The absolute effect size of gaining friends is the same as the absolute effect size of losing friends. We also analyzed the question whether there are gender differences in the effect of friends on happiness. For this so we calculated all models separately for men and women. We did not find any substantial differences.

Our study reaffirms that social capital as measured by the number of close friends has not declined in Germany, Switzerland or the United Kingdom during the first two decades of the twenty-first century. Even the developments related to the COVID-19 pandemic between 2020 and 2022 with its deep repercussions on daily life and the widespread polarization of opinions regarding governmental regulations did not have an effect on the average number of friends (at least in the short run).

Furthermore, we provide robust evidence using FE, FEIS and FD models that social capital, measured by the number of close friends, increases life satisfaction. However, the effect is very small. For illustration, the immediate effect of a doubling in the number of friends in Germany (0.12) is about half the size of life events such as first partnership, and cohabitation, separation from partner, first marriage, and childbirth, about a fourth of the effect of unemployment, and less than a tenth of the effect of the death of the partner or of a child (GSOEP data; Krämer et al., 2025).

As the FD model results suggest, feedback does not seem to bias the estimate of the effect of friends on happiness. This does not eliminate the possibility of reverse causality due to other processes. The FD model only ensures the unbiased estimation of the effect of the friends on happiness due to feedback. Yet, selection on growth may be at work in the UKHLS population.

Nonetheless, there may be a methodological explanation for the smallness of the friends-happiness-effect. The number of friends that subjects report may partly be just “paper friends”, i.e., relatively close individuals, but with only sparse interaction. For example, it could well be that subjects count their former best classmates as friends, even if graduation is 20 years ago and if they meet them only once a year. Hence, if the number of friends also captures paper friends, the effect of friendship on happiness that we observe may be underestimated due to measurement error. Therefore, a better indicator to embrace the importance of friends for subjective well-being could be to also ask for the number of additional social contacts or weak ties or the “time spent with friends”. Albeit, data availability on this measure is very limited, which leaves this question open for future research. Another specious alternative for the number of friends may be the social (emotional and practical) support provided by friends. Yet, social support might be consequence of the number of friends (or the time spent with them) and hence might work as an explanation of the effect of the number of friends on life satisfaction. And again, data availability on social support of friends is limited.

In sum, our study extends and reaffirms previous work on the positive effect of social capital on subjective well-being. This may be useful for social and public health policies tailored to the enhancement of social capital to promote subjective well-being. Thus, the general well-being of a population would benefit from policies that support opportunities to meet other people and to form friendships. As already outlined by Robert Putnam (2000), the creation of social bonds is facilitated for example by associations such as social clubs, sports clubs, or cultural events. Communities could support such organizations by providing the infrastructure to foster social activities for example by community centers, and to encourage citizens to participate in social events