Elsevier

Social Networks

Volume 44, January 2016, Pages 219-225
Social Networks

Extroversion and neuroticism affect the right side of the distribution of network size

https://doi.org/10.1016/j.socnet.2015.10.004Get rights and content

Highlights

  • I use quantile regression to describe the entire distribution of network size.

  • Extroversion affects the right side of the inner and outer layer size distribution.

  • High neuroticism cuts off the right tail of the inner layer size distribution but not the outer layer size distribution.

Abstract

Previous studies explored various predictors of network size by examining their effects on mean network size. However, such predictors may not affect the entire distribution of network size uniformly. In the present study, I theoretically predict that extroversion and neuroticism affect the right side of the network size distribution to a greater extent than the left side and test these predictions using quantile regression analysis. The results showed that the effects of extroversion on the size of the inner and outer layers of personal network were significantly greater when the prediction target was the third quartile than when it was the first quartile. The results also showed that although the effects of neuroticism on the inner layer size were significant, that on the outer layer size were not. When the target variable was inner layer size, the relationship between neuroticism and inner layer size was stronger when the prediction target was the third quartile than when it was the first quartile.

Introduction

There are considerable individual differences in the size of personal networks (egocentric networks), which refer to social networks consisting of alters that the ego can directly contact (Bernard et al., 1990, Pollet et al., 2011a). Previous studies have revealed that these individual differences are partially explained by demographic variables and socio-economic status (Fischer, 1982, McPherson et al., 2006).

Extroversion is another candidate predictor of network size. Extroversion is one of the Big Five traits (Costa and McCrae, 1992). Extroverts tend to be socially active and enjoy it. As a result, they are expected to have a larger personal network. Although sample characteristics and network size measures vary across studies, previous research has found a positive relationship between extroversion and network size (Asendorpf and Wilpers, 1998, Bolger and Eckenrode, 1991, Pollet et al., 2011a, Roberts et al., 2008, Swickert et al., 2002).

The objective of the present study is to re-examine the effect of extroversion on network size using quantile regression. This method is more appropriate for analysis of network size than the methods used in previous studies, because it allows researchers to treat effects of predictors on different sides of a target variable's distribution differently. In the present study, I hypothesize that extroversion and neuroticism have different effects on the right and left sides of the network size distribution.

In previous studies, researchers have focused their attention mostly on mean network size (or mean network layer size) and examined the effect of the determinants on network size by predicting the mean value of network size or calculating correlation coefficients (for example, for the analysis of the effects of personality traits on the mean size of the inner layer to the outer one, see Roberts et al. (2008) and Pollet et al. (2011a)). I will oversimplify these methods and call them OLS-like (ordinary-least-square-like) methods. Though OLS-like methods are popular, there are two problems with addressing network size and the effects of predictors on network size using such methods.

The first problem is the skewness of the network size distribution. Regardless of definitions and measurements, the network size distribution is, most heavily, right-skewed (Bernard et al., 1991, Dunbar and Spoors, 1995, Fischer, 1982, Freeman and Thompson, 1989, Hill and Dunbar, 2003, Killworth et al., 1990, Roberts et al., 2009, Wang and Wellman, 2010). Researchers have tried to solve this problem by log-transforming the network size distribution for the sake of normality or merely excluding the right tail as outliers. However, neither strategy is always appropriate. Although log-transformation changes the distribution form and function, the theoretical meaning of the change is hardly discussed. There are no theoretical bases to exclude the respondents belonging to the right tail in most studies. Far from that, these respondents are potential research targets as Barabási's network hubs (Barabási, 2002). In social network research, we cannot merely exclude outliers from analyses.

The second and more critical problem is that the assumption that effects of predictors on network size are uniform over the entire distribution has no theoretical basis. When we apply OLS-like methods to network size analysis, we assume that only the mean value changes as explanatory variables change, not the shape of the distribution, that is, the change in the network size distribution is a purely parallel move against the change of explanatory variables. We can comprehend the characteristics of the target variable and the effect of explanatory variables with their means and SDs only if the normality and homoscedasticity assumptions hold true. However, if these assumptions do not hold true, the information regarding the value and the change of means is insufficient to describe the distribution of the target variable and the effect of predictors on it. As a result, researchers who employ OLS-like strategies possibly overlook important phenomena. Although often ignored, while these assumptions are very powerful, they are, in fact, merely assumptions for convenience. Indeed, when we address the relationships of network size and explanatory variables, these assumptions are unrealistic. How can we assume that lonely people and network hubs would respond to the change of an explanatory variable like extroversion in exactly the same manner?

When we take heteroscedasticity into account, we can build and test theoretical predictions that treat people on different sides of the distribution differently. In the present study, I demonstrate that the effect of extroversion and neuroticism on network size is not uniform across the distribution. Rather, we should predict different effects of extroversion and neuroticism for the right and left sides of the distribution.

Section snippets

The effect of extroversion and neuroticism is stronger on the right side of the network size distribution

As previous studies have shown, the central position of the network size distribution would move toward the right as extroversion increases. By definition, extroverts are socially active, likely to have more contact with others, and a result, tend to have a larger personal network than do introverts. Positive relationships of extroversion and the central position of network size would be found even without assuming homoscedasticity.

However, I predict that while the positive relationship is

Hypotheses

Based on the discussion above, I predicted that extroversion positively affects the sizes of the inner and outer layers of personal networks and that this effect is stronger on the right sides of the distributions than on the left sides. Similarly, I predicted that neuroticism has a negative impact on inner and outer layer size and that this effect is stronger on the right sides of the distributions than on the left sides. To facilitate comparison with previous studies that addressed only mean

Respondents

Respondents who had registered with the Japanese survey company “Borders Inc.” were invited via email to participate in a survey of interpersonal relationships. I planned to include at least 80 respondents each in five age groups (respondents in their 20s, 30s, 40s, 50s, and 60s or older). I also planned to include 40 male and 40 female respondents in each age group. In total, 478 respondents completed the questionnaire. After excluding respondents with missing values or who did not follow

Results

When all respondents are included into the computation, the mean and standard deviation of the inner layer size are 15.76 and 17.44 and those of the outer layer size are 172.45 and 211.14, respectively, as shown in Table 1. The minimum values of the inner and outer layer sizes are both zero. The maximum values are 114.00 and 2357.48, respectively. The 25th, 50th, and 75th percentiles (denoted as p25, p50, and p75, respectively) of inner layer size are 5, 10, and 20, and those of the outer layer

Discussion

The results support Hypotheses 1a and 1b. As extroversion increases, inner and outer layer sizes increase as well. Furthermore, as extroversion increases, both layer size distributions become right-skewed. The results also support Hypothesis 1b, Hypothesis 2b. Directly opposite in direction to the effects of extroversion, neuroticism decreases the central position of the inner layer size distribution and reduces the right-skewness of the distribution. The results also support Hypothesis 2b,

Acknowledgement

The discussion with Kazunori Inamasu, Keiko Katagiri, Eri Shigemasu, and Ikuko Sugawara helped to elaborate the idea of the present study.

References (40)

  • A.-L. Barabási

    Linked: The New Science of Networks

    (2002)
  • H.R. Bernard et al.

    Estimating the size of an average personal network and of an event subpopulation: some empirical results

    Soc. Sci. Res.

    (1991)
  • N. Bolger et al.

    Social relationships, personality, and anxiety during a major stressful event

    J. Pers. Soc. Psychol.

    (1991)
  • D. Byrne et al.

    Attraction as a linear function of proportion of positive reinforcements

    J. Pers. Soc. Psychol.

    (1965)
  • O. Curry et al.

    Do birds of a feather flock together? The relationship between similarity and altruism in social networks

    Hum. Nat.

    (2013)
  • R.I.M. Dunbar et al.

    Social networks, support cliques, and kinship

    Hum. Nat.

    (1995)
  • L. Festinger et al.

    Social Pressures in Informal Groups

    (1950)
  • C.S. Fischer

    To Dwell Among Friends: Personal Networks in Town and City

    (1982)
  • L.C. Freeman et al.

    Estimating Acquaintanceship Volume

  • T. Furukawa et al.

    Social support and adjustment to a novel social environment

    Int. J. Soc. Psychiatry

    (1998)
  • Cited by (13)

    • Uncovering the effects of learning capacity and social interaction on the experienced utility of low-carbon lifestyle guiding policies

      2021, Energy Policy
      Citation Excerpt :

      Structural characteristics refers to the social network topology structure in which individuals interact with others in the context of low-carbon lifestyle guiding policies (Yeh and Wu, 2014). On one hand, the density of the individual social network (defined as network density in this study) will affect structural characteristics (Ishiguro, 2016); on the other, the frequency of interaction between individuals and other social members (defined as interaction intensity in this study) will also affect structural characteristic (Cheng et al., 2019b). The former provides possible or objective conditions for social interaction, while the latter reflects the active choice of individual social characteristic.

    • Personalized networks? How the Big Five personality traits influence the structure of egocentric networks

      2019, Social Science Research
      Citation Excerpt :

      For example, Klein et al. (2004) detected that neuroticism influences centrality scores in friendship networks. In similar ways, Ishiguro (2016) examined the effect of extraversion and neuroticism on network size. Zhu et al. (2013) investigated all five personality traits and their impact on subjective wellbeing, treating social networks as intermediary variables.

    • Changes in core network size in Japan: Comparisons between the 1990s and 2010s

      2018, Social Networks
      Citation Excerpt :

      When used as covariates of regression models, the following variables were means-centred: years of residency, age, commuting time, working time, and annual household income. Network size is a count variable, and in many cases, its distribution is right-skewed (Bernard et al., 1991; Dunbar and Spoors, 1995; Fischer, 1982; Hill and Dunbar, 2003; Ishiguro, 2016; Killworth et al., 1990; Roberts et al., 2009; Wang and Wellman, 2010). Therefore, I used negative binomial models to examine relationships of various network sizes and covariates.

    • THE ROLE OF FREELANCERS’ EGO NETWORKS IN INCREASING THEMSELVES ACCUMULATED CAPITAL

      2022, Monitoring Obshchestvennogo Mneniya: Ekonomicheskie i Sotsial'nye Peremeny
    View all citing articles on Scopus
    View full text