New ICTs and social interaction: Modelling communication frequency and communication mode choice

Abstract

New information and communication technologies (ICTs) have provided new ways of communicating and maintaining social networks. However, relatively little is known on the effect of ICT-use on social interaction. Therefore, this article aims to explore the factors influencing individuals’ communication frequency and choice of communication mode. The analyses are based on social interaction diary data gathered in 2008 in the Eindhoven region in the Netherlands among 747 respondents. Using these data two models are estimated analysing the number of social interactions in two days and the choice of a communication mode used for the social interaction. Many significant effects of personal and household characteristics were found. In addition, the results for communication mode choice show the importance of including characteristics of the contacted person(s) to explain communication mode choice. The findings allow us to reconstruct the generation of social activities and the relationship between face-to-face and ICT-mediated communication.

Keywords

communication mode choice interaction mixed logit negative binomial social network

Introduction

Over the last decades, new information and communication technologies (ICTs) such as the internet and mobile phones have been developed. These new technologies have changed people’s communication patterns and the way they can maintain social networks that are becoming more geographically spread. Although the effect of new ICTs on interpersonal communication has received a substantial amount of attention in recent literature, a number of issues remain unresolved.

Communication between people by ICT does not take place independently of face-to-face contacts (Dijst, 2009). The relationship between face-to-face and ICT-mediated activities has long been discussed. Usually four possible relationships between face-to-face and ICT-mediated contact are distinguished: substitution, complementarity, neutrality and modification (Graham and Marvin, 1996; Mokhtarian, 1990; Salomon, 1986). The substitution effect, meaning that increasing ICT-use decreases face-to-face contact (and travel), has attracted most attention, especially with regard to work or shopping activities (Mokhtarian, 1990). However, whereas the effect of ICT-use on working or shopping has received much attention, relatively little is known on the effect of ICT-use on social activities (Mokhtarian and Meenakshisundaram, 1999; Mokhtarian et al., 2006; Senbil and Kitamura, 2003).

In addition, there is little evidence of studies at the level of social networks. Some recent exceptions are Carrasco and Miller (2006), Frei and Axhausen (2009), Hlebec et al. (2006), Larsen et al. (2006), Lonkila and Gladarev (2008) and Tillema et al. (2010). These studies indicate the importance of explicitly incorporating characteristics of those with whom people communicate.

Moreover, not much is known about the factors that influence the use of different communication modes. Many studies into ICT use have focused on internet use rather than on electronic communication, while cell phone conversations, short message service (SMS) and email play an important role in social network maintenance (Tillema et al., 2010). In addition, there might be substantial differences between countries and cultures in these relationships.

Recognizing that more knowledge is needed about social interaction behaviour, the aim of this article is to explore the factors influencing communication frequency and communication mode choice in the Dutch context. For the purpose of this study social interaction diary data were collected in the Eindhoven region in the Netherlands in 2008. This article describes the results of two models: a negative binomial regression model to analyse the number of social interactions in two days and a mixed logit model to analyse the choice of a communication mode used for the social interaction. The ability to predict these variables is important, as it allows us to reconstruct the generation of social activities and the relationship between face-to-face social activities and the use of different ICT-mediated communication modes in maintaining social contacts. Moreover, social interactions are important for people as they provide access to a variety of resources, such as instrumental and emotional support (social capital). This study is therefore also relevant with regard to broader concerns in society, such as social exclusion, social capital and the quality of life of people.

The remainder of the article is organized as follows. The next section discusses the existing literature on factors influencing social interaction frequency and communication mode choice for social interaction. The third section describes the data collection and analysis methods. In the fourth section the analysis results of the two estimated models are presented: a negative binomial regression model to analyse the number of social interactions in two days and a mixed logit model to analyse the choice of a communication mode used for the social interaction. The final section contains the conclusions and discusses the implications of the findings.

Theoretical framework

The maintenance of social networks is realized through social interaction. This maintenance depends on the frequency of social contact and the communication mode that is used. So far, not much is known about the factors that affect the frequency of contact and the choice of various communication technology tools (Lo and Lie, 2008), especially for social interaction. This section discusses the existing literature on factors influencing social interaction frequency and communication mode choice for social interaction.

Factors influencing social interaction frequency

The needs to communicate or perform social activities differ between individuals. Personal characteristics, such as gender, age and work status have been found to affect contact frequency between social network members (e.g. Carrasco and Miller, 2006, 2009; Farber and Páez, 2009; Frei and Axhausen, 2009; Tillema et al., 2010). For example, Carrasco and Miller (2006) and Lu and Pas (1999) found that females undertake fewer social activities than males. Farber and Páez (2009) found that the elderly are less likely to perform social activities while the young are more likely. Carrasco and Miller (2006) found a smaller propensity to perform social activities for people who are employed, as people with more free time (less work or study) have more possibilities for more frequent social activities.

In addition, household composition is likely to affect the number of social interactions people have. The presence of a partner and children (understood here as up to and including 18 years of age) in the household may lessen the need or even be a constraint for out-of-home social activities (Carrasco and Miller, 2006).

Car ownership offers opportunities for (face-to-face) social activities and may have a positive effect on the number of social interactions people have. In a survey of the quality of life of the elderly, Banister and Bowling (2004) found that people with access to a vehicle (and people with access to good local transport) were likely to undertake more social activities. However, Farber and Páez (2009) found automobile-reliant people to participate in significantly fewer social activities.

Social interaction frequencies are also likely to be influenced by social network characteristics. Carrasco and Miller (2006) and Silvis et al. (2006) suggest that people with a large social network are likely to have more social interactions than people with a small social network. However, Boase et al. (2006) found a negative relationship between the size of the social network and the total number of face-to-face and mediated interactions. These findings seem to indicate that larger social networks involve lower frequency of social contacts. Also according to Dijst (2009) people seem to be capable of maintaining large social networks by reducing contact frequency.

Finally, the day of the week may also affect social-interaction variables. For instance, Mokhtarian and Meenakshisundaram (1999) found face-to-face and mediated social interaction frequencies to be lower at the weekend.

Factors influencing communication mode choice

ICTs (the internet and mobile phones) offer a number of services, such as voice call, email, text messaging (SMS) and instant messaging (IM). These communication modes differ in nature. They can be fixed in place, such as the landline telephone, or flexible (mobile). Contacts can be synchronous (phone calls, IM), or asynchronous (email, SMS). It is important to differentiate between these communication modes, instead of regarding ICT-mediated communication as one homogenous category.

The existing literature suggests that the use of these different modes can be explained by a number of factors. For example, with respect to the use of email, young, highly educated males were the forerunners and they still have a higher ICT-mediated contact frequency (e.g. Frei and Axhausen, 2009).

According to Dijst (2009), the choice between synchronous and asynchronous modes can be explained by costs. Synchronous contacts often require more money, time and effort than asynchronous contacts. For instance, trying to establish synchronous contact involves the risk that the contact cannot be reached, making asynchronous forms of contact more efficient. According to Dijst (2009) the cost argument can explain the higher use of SMS and email by younger people. The opportunities people have for travelling can also explain their use of communication means (Dijst, 2009; Larsen et al., 2006).

It is also to be expected that social interaction behaviour depends on the time available. People who work more hours per week will have less time for (face-to-face) social interactions, which may result in fewer face-to-face social interactions relative to ICT-mediated interactions. During the weekend, people will have more time for face-to-face social activities. For instance, Kemperman et al. (2006) found people to have more face-to-face social activities on Saturdays and Sundays.

In addition, social interaction is influenced by the alter(s) with whom individuals interact (Carrasco and Miller, 2006, 2009). Relational role and emotional closeness has been found to affect people’s communication with different modes (e.g. Lonkila and Gladarev, 2008; Tillema et al., 2010). Several studies found that all communication means are used more for very close ties than for less close ties (Boase et al., 2006; Tillema et al., 2010). Rivière and Licoppe (2005), comparing text messaging in Japan and France, found that in Japan, text messages are sent to all contacts, independent of relational distance. However, telephone calls, which are relatively expensive, are mainly used to contact people who are emotionally very close. They found that in France SMS messages are mainly sent to the closest ties and not to acquaintances or professional contacts (Rivière and Licoppe, 2005). Frei and Axhausen (2009) found that workmates are contacted less often face-to-face and by telephone, whereas relatives are contacted more often by telephone. Chen et al. (2002) have shown that friends are more likely than family members to use the internet to communicate with one another. Baym et al. (2007) found respondents to have lower proportions of face-to-face interaction with family members than with friends, romantic partners and acquaintances. They also found people to have higher proportions of telephone interaction with family members than with friends, romantic partners and acquaintances. Lo and Lie (2008) suggest that the level of trust towards the communication partner will affect the choice of communication tool employed. When the level of trust is lower, communication tools that provide more clues and information (synchronous modes) will be chosen.

Finally, the geographical distance between people is an important factor in explaining social interaction behaviour with different communication modes. As social networks are becoming more geographically spread, ICTs are becoming more important because they provide opportunities to maintain contacts over longer distances (Axhausen, 2002; McPherson et al., 2006; Urry, 2003). Boase et al. (2006) found that face-to-face contacts diminish with geographical distance. Telephone calls (mainline and mobile phone) showed no relationship with geographical distance, while the frequency of email use increased with geographical distance. Larsen et al. (2006) found the frequency of face-to-face and telephone contact to decrease with geographical distance, and the frequency of email communication to increase, because of the relatively low costs. Frei and Axhausen (2009) found that face-to-face contact frequency decreases fastest with distance. They found distance to have a negative effect on telephone and SMS contact frequency as well. For email no effect was found. Tillema et al. (2010) also found that face-to-face and electronic communication frequencies decline with increasing physical distance.

The available literature shows that some research attempts have been made to study the effects of new ICTs on social interaction behaviour. However, the knowledge about the factors that influence social interaction frequency and communication mode choice is rather limited and needs further investigation. Moreover, there might be substantial differences between countries and cultures in these relationships. The purpose of this study is therefore to explore the factors that influence social interaction frequency and communication mode choice in the Dutch context. The results will help to better understand the maintenance of social networks in a changing society.

Data and methods

Data collection

For this study we use data that were collected between January and June 2008 in a number of neighbourhoods in and around Eindhoven. A data collection instrument was designed, consisting of a paper and pencil social interaction diary, in which the respondents were asked to record all their social interactions during two days. Compared with other data collection methods, such as observations and interviews, diaries are less time consuming for researchers and more familiar and unobtrusive to respondents, as diaries enable people to self-record their contacts with other people (Duck, 1991; Fu, 2007; Reis and Wheeler, 1991). To prevent bias due to respondents omitting certain interactions in the diary, they were asked to record their social interactions as soon as possible after they occurred. They also received an interaction worksheet which they could use during the day to remember their social interactions, in case it was impossible to take the diary along.

Social interactions were defined as all forms of social contact, for instance visiting, performing a joint activity, having a conversation (face-to-face as well as over the phone or online), sending or receiving an email, an SMS, a letter or a fax. Interactions at work or school about work or school matters were not recorded; nor were interactions at home with only household members. Interactions that had a more business-like nature than a social nature, like interactions with unknown shop assistants when paying for merchandise, were not recorded either.

In the social interaction diary, detailed information was gathered about the interactions, e.g. when, where and with whom the interaction took place, and which communication mode was used.

To recruit respondents, a personal approach was employed. We went by people’s homes to ask them if they were willing to participate in this study. The visits took place at varying times of day, also in the evening, to prevent under-representation of working people. The personal approach was employed to increase respondents’ participation. However, it may have caused some bias in the sample of people who were not home (and therefore possibly more likely to be socializing). The completed diaries were collected approximately a week later. Out of 3699 people who answered the door, 1648 (45%) accepted a diary. Out of these 1648, 747 useful diaries were returned. This results in an overall response rate of 20%, which is a good response rate for this type of survey.

In addition to the social interactions, a number of personal characteristics were collected. Table 1 shows the descriptive statistics of the variables that are relevant for this study. The sample is not completely representative of the Dutch population. Females are somewhat over-represented in the sample, as are older people and higher educated people. However, weighting was not applied, as the aim of the study is to analyse the effects of the explanatory variables on social interaction frequency and communication mode choice (and not to derive results in terms of predictions of dependent variables that are representative of the Dutch population).

Table 1.

Respondent characteristics (N=747 respondents).

		N	%
Gender	Male	293	39
Age	Young (<30)	96	13
	Old (>50)	281	38
Partner	Living with partner	582	78
Children	Children under 18	333	42
Education	Primary	173	23
	BSc or higher	442	46
Network size	Small (0–11)	216	29
	Large (>32)	135	18
Involvement in clubs	1 or more clubs	526	71
Car	1 or more cars	670	90
		Mean	St. dev.
Work hours	Per week	19.85	18.72

In total, the 747 respondents reported 8074 social interactions on 1494 diary-days (2 days per respondent). This is an average of 5.4 social interactions per person per day. It is precarious to compare the recorded number of social interactions with findings of other studies, as most other studies using a communication diary used a wider definition of social interaction or communication and therefore find a higher number of interactions (Fu, 2007; Silvis et al., 2006; Thulin and Vilhelmson, 2004). Our findings are, however, in line with Baym et al. (2004) and Zumkeller (2002).

The characteristics of the sample of social interactions are shown in Table 2. As can be seen, 52% of the interactions were face-to-face. This means that face-to-face communication is the dominant mode of interaction for our sample. Baym et al. (2004) also found face-to-face to be the most used communication mode; 64% of college students’ interpersonal interactions were face-to-face. However, Chen et al. (2002) and Thulin and Vilhelmson (2004) found that the telephone was the dominant communication mode.

Table 2.

Social interaction characteristics (N=8074 social interactions).

		N	%
Communication mode	Face-to-face	4177	52
	Mobile phone	836	10
	Landline phone	1506	19
	SMS	449	6
	Email	858	11
	Instant messaging (IM)	162	2
	Other	60	1
Day of the week	Saturday	951	12
	Sunday	636	8
Interaction alter	Group (2 or more)	2121	26
	1 relative	1950	24
Duration of relationship	Known at least 15 years	2379	30
Strength of tie	Very strong tie	2273	28
	Somewhat strong tie	1938	24
		Mean	St.dev.
Distance	(km)	81.99	670.46

Methods

Using the data from the social interaction diaries, the number of social interactions and the communication mode choice are analysed. We use a negative binomial regression model to analyse the number of social interactions in two days, as a function of a number of personal and household characteristics. To analyse the choice of a communication mode used for the social interaction, we use a mixed logit model.

Negative binomial regression model

As the number of social interactions individual i had in two days is a count variable, Poisson regression or negative binomial regression can be used for the first model. Negative binomial regression can be considered a generalization of Poisson regression that allows the variance to differ from the mean. It assumes that the mean λ_i of y_i is not only determined by x_i but also a heterogeneity component ε_i unrelated to x_i. The formulation can be expressed as:

P (Y = y_{i}) = e^{- λ_{i}} λ_{i}^{y_{i}} / y_{i}!, y = 0, 1, \dots,

where ln λ_i = β’x_i + ε_i.

In this model exp(ε_i) has a gamma distribution with mean 1.0 and variance α. The formulation of the negative binomial distribution which can be used to model count data with overdispersion is then derived as:

P (Y = y_{i}) = Γ (θ + y_{i}) / [Γ (θ) y_{i}!] u_{i}^{θ} {(1 - u_{i})}^{y i}

where Γ is the gamma distribution

u_{i} = θ / (θ + λ_{i})

θ = 1 / α

α is a dispersion parameter, such that

Var [y_{i}] = E [y_{i}] {1 + α E [y_{i}]}

Mixed logit model

In the second model the communication mode choice for social interaction is analysed. As communication mode is a categorical variable, a multinomial logit model could be used. Using a mixed multinomial logit model some limitations of the standard multinomial logit model are overcome. In the standard multinomial logit model, the alternative specific constants and coefficient parameters (β’s), are fixed, which means they are the same for everyone. Moreover, standard logit does not take into account any unobserved factors that persist over time for a given decision maker (with repeated choices). In mixed logit models, both random taste variation and correlation in unobserved factors over time, can be accounted for. For the mixed logit model that takes into account heterogeneity and repeated measures, the utility for individual i for alternative j on choice occasion t would be:

U_{i j t} = β_{i}^{,} x_{i j t} + ε_{i j t}

where ε_ijt is an unobserved random term that is distributed iid extreme value, independent of vectors β_i and x_ijt. Each random parameter in vector β_i is defined as the average preference in the population, b, and an individual deviation, η_i, which represents the individual’s preference relative to the average preference for, in this case, a particular communication mode. The utility is:

U_{i j t} = b^{,} x_{i j t} + η_{i}^{,} x_{i j t} + ε_{i j t}

If β_i were known, the probability that individual i chooses alternative j at choice occasion t would be standard logit:

L_{i j} (β i) = e^{β_{i} x_{i j t}} / \sum j^{,} e^{β_{i} x_{i j^{,} t}}

However, since β_i is random and not known, the (unconditional) choice probability is the integral of this logit formula over the density of β_i. Assuming that the preferences vary in the population with density f(β|θ), where θ are the parameters of this distribution, the actual probability is (Greene, 2002):

P_{i j} (θ) = \int L_{i j} (β) f (β | θ) d β

Results

In this section the results of the two models are discussed. We used Nlogit (Greene, 2002) to estimate the models.

The number of social interactions

In order to examine the effect of the explanatory variables on the number of social interactions a negative binomial model was estimated. The estimation results of model 1 are shown in Table 3. The frequency of face-to-face social interaction is found to be influenced by a number of personal and household characteristics. Whereas others (Carrasco and Miller, 2006; Lu and Pas, 1999) found that females undertake fewer social activities than males, we did not find a significant gender effect. This indicates that, when controlling for other variables, men and women have equal numbers of social interactions.

Table 3.

Estimation results for number of social interactions in two days.

	β
Constant	2.505 **
Male	−0.001
Age <30	0.117 **
Age >50	−0.117 **
Partner	−0.150 **
Children	0.150 **
Low education	−0.144 **
High education	0.145 **
Work hours per week	−0.003 *
Car	0.000
Small social network	−0.263 **
Large social network	0.176 **
Club member	0.106 *
Weekend days	−0.192 **
Dispersion parameter	0.168 **
Number of observations	747
Log likelihood function	−2264.952
Restricted log likelihood	−2551.352
Chi squared	572.801
McFadden Pseudo R-squared	.112

Coefficient is significant at the 0.05 level. **Coefficient is significant at the 0.01 level.

With regard to age we find fewer social interactions for the oldest age group, and more for the youngest group. This is in line with findings of other studies (e.g. Carrasco and Miller, 2006; Farber and Páez, 2009).

Living with a partner has a negative effect on the number of social interactions. This finding is in line with Tillema et al. (2010). It is a plausible finding as people living with a partner tend to perform more social activities at home with their partner and those interactions occurring inside the home with only household members were not recorded. However, the presence of children in the household has a significant positive effect on the number of social interactions. This indicates that the presence of children in the household is a stimulus rather than a constraint for social interactions outside the household. A possible explanation for this is that when having children, people expand their social networks or intensify contacts with other families with children.

The results show that education level significantly affects the number of social interactions per day. When controlling for the other characteristics, the average number of social interactions is lower for people with primary education and higher for people who are higher educated. This might be explained by the fact that higher educated people meet more people and expand their social networks through study and work.

The number of work hours per week is found to have a negative effect on the number of social interactions. This is a plausible finding as more time spent on working implies less time for social activities. Whereas others (Banister and Bowling, 2004; Farber and Páez, 2009) found car ownership to affect the number of face-to-face social activities, we did not find car ownership to have a significant effect on the total number of face-to-face and ICT-mediated social interactions. This difference may be related to the complementarity of the different communication modes.

Social network size is found to have an effect on the number of social interactions per day. The average number of interactions per day is found to be lower for people with a small social network. People with a larger social network on average have more social interactions. This is a plausible finding which is consistent with Carrasco and Miller, (2006) and Silvis et al. (2006), but in contrast with Boase et al. (2006).

The parameter for club membership shows that involvement in one or more clubs is positively related to the number of social interactions. This is a plausible finding as clubs generate activities for their members and they offer opportunities for new social contacts.

Finally, the results show a negative coefficient for the number of social interactions in the weekend. Whereas most people have more free time during the weekend, and probably spend more time on social interaction, the total number of social interactions is lower compared to weekdays. Mokhtarian and Meenakshisundaram (1999) also found the number of personal meetings as well as mediated communications to be lower in the weekend compared to weekdays. Studying only face-to-face social visits, Kemperman et al. (2006) found larger numbers during the weekend. However, many social interactions such as mediated contacts and short talks were not included in their study, which might explain this difference.

Communication mode choice

The second model is used to analyse the communication mode choice for social interaction. A mixed logit model is estimated because each respondent has several choice situations and we expect preference heterogeneity between respondents. After deleting cases with missing values, 6237 useful cases among 743 individuals were entered in the analysis.

The choice between face-to-face, landline phone, mobile phone, SMS, email and IM is considered. The first category, face-to-face, serves as the reference category in the model. Thus, the coefficients estimated are interpreted relative to choosing face-to-face social interaction. As access to ICTs is very high in the sample (96% for both internet and mobile phone), in the model all individuals are assumed to have access to all communication modes.

Random parameters (standard deviations of unobserved heterogeneity) were estimated for the alternative specific constants. Simulated maximum likelihood estimation, using Halton draws, was used to estimate the parameters of the choice model. The number of Halton draws was set to 100.

The constants in Table 4 are all negative. This indicates that, if all explanatory variables are evaluated at zero, people on average are more likely to choose face-to-face contact relative to the other modes. As expected, IM has the largest negative coefficient.

Table 4.

Estimation results for communication mode choice.

	Landline phone	Mobile phone	SMS	Email	IM
Constant	−1.700**	−2.872**	−2.730**	−2.953**	−5.185**
Male	−0.106	0.310*	−0.426*	0.088	0.756*
Age <30	−0.881**	0.718**	0.980**	−0.179	1.979**
Age >50	0.278*	−0.743**	−1.744**	0.278	−1.001
Partner	0.035	0.016	0.132	0.328*	−0.659
Children	0.372**	−0.244	−0.266	0.334*	−0.141
Low education	0.094	−0.257	0.376	−0.324	0.474
High education	0.035	0.163	0.129	0.702**	−0.791*
Work hours	−0.014**	0.005	−0.002	−0.001	−0.003
Car	−0.341*	0.219	−0.276	−0.245	0.020
Small network	0.144	0.027	0.105	0.198	−0.213
Large network	−0.180	0.014	−0.154	−0.027	−0.111
Club member	0.207*	−0.009	−0.033	0.150	0.180
Saturday	−0.366**	−0.082	0.081	−0.777**	0.100
Sunday	−0.167	−0.467*	0.195	−0.182	0.095
Group	−2.438**	−2.637**	−2.363**	−0.453**	−1.055**
1 Relative	0.590**	0.862**	−0.196	0.225	−0.025
Very strong tie	0.695**	1.448**	1.891**	−0.317*	1.088**
Somewhat strong	0.442**	0.660**	1.146**	−0.184	0.632
Known 15 years	0.436**	−0.051	−0.054	−0.214	−0.293
Ln distance	0.318**	0.157**	0.172**	0.428**	0.499**
St. deviations	0.538**	0.603**	0.733**	0.790**	1.520**
Number of observations		6237
Number of groups		743
Log likelihood function		−7085.658
Restricted log likelihood		−11175.20
R-squared		0.366
R-squared adjusted		0.364

Coefficient is significant at the 0.05 level. **Coefficient is significant at the 0.01 level.

Regarding gender, when controlling for the other characteristics, we find males to have a larger likelihood of choosing IM and mobile phone for social interaction and a smaller likelihood of choosing SMS. This finding is in contrast with Tillema et al. (2010) who found male respondents to communicate less frequently via electronic modes. It is, however, in line with the notion that the forerunners in adopting ICTs are still more likely to use the (newest) ICT-mediated modes.

With respect to age we find that the youngest group is more likely to choose mobile phone, SMS and IM and less likely to choose the landline phone. According to Oksman and Turtiainen (2004) the popularity of the mobile phone among younger people may be explained by the privacy it offers from parental control, compared to the landline phone. The oldest age group has a smaller likelihood of choosing mobile phone and SMS and a larger likelihood of choosing the landline phone. This indicates that younger people are faster in adopting new ICT-based modes and the older generation tends to hold on to communication modes they are familiar with.

Living with a partner is found to have a positive effect on choosing email. The explanation for this is unclear. People with children under 18 have a larger likelihood of choosing landline telephone and email. This may be explained by the fact that people with children spend more time at home, where a landline phone (and PC) is available.

Higher educated people have a larger likelihood of choosing email and a smaller likelihood of choosing IM. People who work or study more are less likely to choose the landline phone, which is a plausible finding as they spend more time out of home. Car ownership is found to have a negative effect on choosing the landline phone relative to face-to-face contact. This might be explained by the fact that car ownership offers more opportunities for spending time out of home and travelling for face-to-face social activities.

With respect to social network size no significant effects are found. Whereas one might expect that people with large networks substitute (time-consuming) face-to-face contacts with faster ICT communications, our results suggest that communication mode choice is independent of social network size.

Club membership is found to have a positive effect on landline phone interactions. This is a surprising finding as one might expect clubs to mostly generate face-to-face social interactions between members.

With respect to day of the week the results indicate that on Saturdays people are less likely to have landline phone and email contact. On Sundays we find people to be less likely to have mobile phone interactions relative to face-to-face interactions. The reason for this is probably that during the weekend people have more time for longer face-to-face social interactions.

The negative coefficients for group indicate that social interactions with more than one person (a group) are most likely to be face-to-face interactions. The highest negative coefficients are for telephone interactions. This finding was expected, as telephone is not very suitable for interacting with more than one other person.

For interacting with relatives, positive effects are found for landline and mobile telephone. Frei and Axhausen (2009) also found that relatives are contacted more often by telephone.

People who have known each other long (15 years or more) are more likely to contact each other by landline telephone. We find that if the tie between ego and alter is very strong, they are less likely to interact by email, and more likely to interact using telephone, SMS and IM, relative to face-to-face. This finding is surprising, as one would expect face-to-face contact to be most intimate and therefore more used in closer circles. This finding is also in contrast with Tillema et al. (2010) who found that asynchronous modes, such as email and other modes, become more important at the expense of face-to-face as relational distance increases. For somewhat strong ties positive effects are found for telephone and SMS.

The results for geographical distance show positive coefficients for all ICT-mediated modes. This indicates that, if distance increases, people are more likely to choose ICT-mediated modes over face-to-face contacts. Email and IM have the largest coefficients, indicating that these modes are most likely to be chosen if distance increases. This finding is in line with other studies (e.g. Boase et al., 2006; Larsen et al., 2006). As social networks are becoming more geographically spread, this finding suggests that ICTs will become more important for the maintenance of social networks.

There is evidence of heterogeneity in the communication mode choice. The unobserved preference heterogeneity terms for all five modes are highly significant. This indicates substantial variation across individuals in the overall preference for a communication mode. An R-square of .355 indicates a high goodness of fit of the model.

Conclusions and discussion

In this article two models were estimated that provide insight in face-to-face and ICT-mediated social interaction behaviour. The models were based on data collected in the Netherlands, using a social interaction diary.

First, a negative binomial model was estimated to analyse for each respondent the number of social interactions. The results suggest that younger, higher educated people on average have more social interactions. Work was found to have a negative effect. The presence of a partner was found to reduce the number of social interactions, whereas the presence of children in the household was found to have a positive effect. Social network size and club membership were found to have a positive effect on the number of social interactions.

The second model, a mixed logit model, was used to analyse the communication mode choice for each social interaction. Personal characteristics were found to significantly affect communication mode choice. The results indicate that younger people are faster in adopting new ICT-based modes, whereas older people tend to hold on to familiar modes. People who spend more time at home where a landline phone is available (people who work less, have children and no car), tend to choose the landline phone. Gender, education, social network characteristics and day of the week were also found to affect communication mode choice.

Communication mode choice was also found to be affected by characteristics of the contacted person. Social interactions with a group are most likely to be face-to-face interactions and least likely by telephone. For interacting with relatives, positive effects are found for landline and mobile telephone and email. People who have known each other long (15 years or more), are more likely to contact each other by landline telephone.

With regard to emotional distance we found that if the tie between ego and alter is very strong, they are more likely to interact using telephone, SMS and IM, relative to face-to-face, which is in contrast with other findings (Tillema et al., 2010). The findings with regard to geographical distance were in line with other findings, suggesting that if geographical distance increases, ICT-based modes (especially email and IM) become more important at the expense of face-to-face.

The models that were estimated in this article provide interesting results on social interaction behaviour and the way ICTs can offer opportunities to maintain social networks. Our findings show the importance of differentiating between different ICT-mediated communication modes, instead of treating ICT communication as a single category.

New ICTs have increased the possibilities for maintaining and establishing new contacts over longer distances. Moreover, ICT-mediated communication modes have been found to become more important at the expense of face-to-face contacts as geographical distance increases. As social networks are becoming more geographically spread, the ICTs will become more important for the maintenance of social networks, especially as access to ICTs is expected to increase around the world in the coming years. In spite of the increasing overall levels of access to ICTs, social equity and social exclusion are current policy issues that require continued research into social interaction behaviour.

In addition, as face-to-face contacts require travel, the revealed communication patterns between social network members (frequency and type of communication media used) are relevant to an understanding of social travel demand. New ICTs have increased possibilities to maintain and to build new social ties over long distances. Although we found the frequency of face-to-face communication (and trips) to decrease with geographical distance, our data show that (occasional) physical meetings tend to take place with most contacts. As the access to and the use of ICTs is still increasing, long distance contacts can be maintained better, which may imply that long distance trips will increase in the near future.

Although this study provides interesting results on social interaction behaviour, there are factors that may affect communication mode choice which were not included in our empirical data. For instance, information content, quality, intimacy or urgency were not taken into account in this study, whereas these are likely to affect the choice of communication mode (Tillema et al., 2010).

Finally, the literature suggests that communication frequencies with different modes might differ substantially between different cultures. Therefore similar analyses from other socio-cultural and spatial contexts would be desirable to further investigate face-to-face and ICT-mediated social interaction behaviour.

Footnotes

Funding

This research received no specific grant from any funding agency in the public, commercial or not-for-profit sectors.

Pauline E.W. van den Berg is a PhD student at the Urban Planning Group at the Eindhoven University of Technology. Her research interests concern human behaviour and its social and cultural context, urban and transport planning, activity-travel behaviour and data collection techniques.

Theo A. Arentze is an Associate Professor in the Urban Planning group at the Eindhoven University of Technology. His main fields of expertise and current research interests are decision support systems, behavioural modelling, knowledge discovery and learning with applications in urban planning and marketing.

Harry J.P. Timmermans is chaired Professor of the Urban Planning Group at the Eindhoven University of Technology. His main research interests concern the study of human choice processes, mathematical modelling of urban systems and spatial interaction and choice patterns and the development of decision support and expert systems for application in urban planning.

References

Axhausen

(2002) A dynamic understanding of travel demand: A sketch. Arbeitsberichte Verkehr und Raumplanung 119. Zürich: Institut für Verkehrsplanung und Transportsysteme, ETH Zürich.

Banister

Bowling

(2004) Quality of life for the elderly: The transport dimension. Transport Policy 11: 105–115.

Baym

Zhang

Lin

(2004) Social interaction across media: Interpersonal communication on the Internet, face-to-face, and telephone. New Media & Society 6(1): 41–60.

Baym

Zhang

Kunkel

. (2007) Relational quality and media use in interpersonal relationships. New Media & Society 9(5): 735–752.

Boase

Horrigan

Wellman

. (2006) The Strength of Internet Ties. Washington, DC: Pew Internet and American Life Project. Available at: http://www.pewinternet.org/~/media/Files/Reports/2006/PIP_Internet_ties.pdf.pdf (accessed 31 January 2012).

Carrasco

Miller

(2006) Exploring the propensity to perform social activities: A social networks approach. Transportation 33(5): 463–480.

Carrasco

Miller

(2009) The social dimension in action: A multilevel, personal networks model of social activity frequency between individuals. Transportation Research Part A-Policy and Practice 43(9): 90–104.

Chen

Boase

Wellman

(2002) The global villagers: Comparing internet users and uses around the world. In: Wellman

Haythornthwaite

(eds) The Internet in Everyday Life. Malden, MA: Blackwell, 74–113.

Dijst

(2009) ICT and social networks: Towards a situational perspective on the interaction between corporeal and connected presence. In: Kitamura

Yoshii

Yamamoto

(eds) The Expanding Sphere of Travel Behaviour Research. New York: Emerald Publishing, 45–76.

10.

Duck

(1991) Diaries and logs. In: Montgomery

Duck

(eds) Studying Interpersonal Interaction. New York: Guilford, 141–161.

11.

Farber

Páez

(2009) My car, my friends, and me: An exploratory analysis of automobility and social activities. Journal of Transportation Geography 17(3): 216–225.

12.

Frei

Axhausen

(2009) Modelling the frequency of contacts in a shrunken world. Submitted to Social Networks. Available at: http://www.ivt.ethz.ch/oev/ped2012/vpl/publications/reports/ab532.pdf (accessed 1 February 2012).

13.

Y-C

(2007) Contact diaries: Building archives of actual and comprehensive personal networks. Field Methods 19(2): 194–217.

14.

Graham

Marvin

(1996) Telecommunications and the City: Electronic Spaces, Urban Places. London: Routledge.

15.

Greene

(2002) Nlogit Version 3.0 Reference Guide. Plainview, NY: Econometric Software Inc.

16.

Hlebec

Lozar Manfreda

Vehovar

(2006) The social support networks of Internet users. New Media & Society 8(1): 9–32.

17.

Kemperman

Arentze

Timmermans

(2006) Social commitments and activity-travel scheduling decisions. In: Proceedings of the 85th Annual Meeting of the Transportation Research Board, Washington DC.

18.

Larsen

Urry

Axhausen

(2006) Social Networks and Future Mobilities: Report to the Horizons Programme of the Department for Transport. Lancaster/Zurich: Department of Sociology, University of Lancaster/IVT, ETH Zürich.

19.

Lie

(2008) Selection of communication technologies – a perspective based on information richness theory and trust. Technovation 28: 146–153.

20.

Lonkila

Gladarev

(2008) Social networks and cellphone use in Russia: Local consequences of global communication technology. New Media & Society 10(2): 273–293.

21.

Pas

(1999) Socio-demographics, activity participation and travel behavior. Transportation Research Part A – Policy and Practice 33: 1–18.

22.

McPherson

Smith-Lovin

Brashears

(2006) Social isolation in America: Changes in core discussion networks over two decades. American Sociological Review 71(3): 353–375.

23.

Mokhtarian

(1990) A typology of relationships between telecommunications and transportation. Transportation Research Part A – Policy and Practice 24(3): 231–242.

24.

Mokhtarian

Meenakshisundaram

(1999) Beyond tele-substitution: Disaggregate longitudinal structural equations modeling of communication impacts. Transportation Research Part C – Emerging Technologies 7: 33–52.

25.

Mokhtarian

Salomon

Handy

(2006) The impacts of ICT on leisure activities and travel: A conceptual exploration. Transportation 33: 263–289.

26.

Oksman

Turtiainen

(2004) Mobile communication as a social stage: Meanings of mobile communication in everyday life among teenagers in Finland. New Media & Society 6(3): 319–339.

27.

Reis

Wheeler

(1991) Studying social interaction with the Rochester interaction record. In: Zanna

(ed.) Advances in Experimental Social Psychology. Vol. 24. San Diego, CA: Academic Press, 270–318.

28.

Rivière

C-A

Licoppe

(2005) From voice to text: Continuity and change in the use of mobile phones in France and Japan. In: Harper

Palen

Taylor

(eds) The Inside Text: Social and Cultural Design Perspectives on SMS. Dordrecht: Springer, 103–126.

29.

Salomon

(1986) Telecommunications and travel relationships: A review. Transportation Research Part A – Policy and Practice 20: 223–238.

30.

Senbil

Kitamura

(2003) Simultaneous relationships between telecommunications and activities. Paper presented at the tenth international conference on travel behaviour research, Lucerne, 10–15 August 2003.

31.

Silvis

Niemeier

D’Souza

(2006) Social networks and travel behavior: Report from an integrated travel diary. In: Proceedings of the 11th International Conference on Travel Behaviour Research, Kyoto, Japan, 16–20 August 2006.

32.

Thulin

Vilhelmson

(2004) Virtual mobility of urban youth: ICT-based communication in Sweden. Tijdschrift voor Economische en Sociale Geografie 96(5): 477–487.

33.

Tillema

Dijst

Schwanen

(2010) Face-to-face and electronic communications in maintaining social networks: The influence of geographical and relational distance and of information content. New Media & Society 12(6): 965–983.

34.

Urry

(2003) Social networks, travel and talk. British Journal of Sociology 54(2): 155–175.

35.

Van den Berg

Arentze

Timmermans

(2008) Social networks, ICT use and activity-travel patterns: Data collection and first analyses. Paper presented at the 9th international conference on design & decision support systems in architecture and urban planning, The Netherlands, 7–10 July 2008.

36.

Zumkeller

(2002) Transport and telecommunication: First comprehensive surveys and simulation approaches. In: Mahmassani

(ed.) In Perpetual Motion: Travel Behavior Research Opportunities and Application Challenges. Oxford: Pergamon, 183–207.