«NETWORKING IN EVERYDAY LIFE by Bernard J. Hogan A thesis submitted in conformity with the requirements for the degree of Doctor of Philosophy ...»
This quote is from the survey, although the interview used virtually the same wording.
The wording of this question is drawn almost exactly from Boase et al.’s (2006) study of media use and social networks in America. This is not a coincidence as both Boase and Wellman were members of the Connected Lives design team as well as designers of the earlier study.
As can be seen from the wording, one can superﬁcially pass over the details of what ‘very close’ means and still have a reasonably good sense of what it encapsulates. It is also worth noting that while the survey includes an indicator of the frequency of inter
those that one frequently contacts, we carefully chose ‘regularly keep in touch with’.
This latter phrase does not necessarily mean the most contact, but it does imply that ego shares a mutual interest in maintaining contact. It also frames the kind of contact as being oriented towards mutual affection. One ‘regularly reports’ to their boss, but one ‘regularly keeps in touch’ with alters. This is in keeping with Marsden and Campbell’s criticism of contact frequency as a measure of tie strength (1984). They point out that numerous individuals are in frequent contact (such as neighbours or coworkers) because of shared structural location. This is not to say that one should exclude these individuals, but it implies that frequent contact is not a sufﬁcient condition for inclusion in one’s personal network.
A socioemotional deﬁnition of network membership is especially important for this particular analysis. This work is focused on the styles and strategies employed by ego in everyday life. There are a number of settings where ego’s pattern of networking may be especially constrained (Webster, Freeman, and Aufdemberg, 2001).
By focusing on the voluntary and emotionally close ties rather than structurally imposed ties (such as one’s coworkers) I am able to look at the relationships for whom ego has the most latitude in organizing daily affairs. This should lead to a more appreciable and realistic appraisal of the differences in everyday communication and social engagement.
4.7.3 Survey measures The survey did not ask individuals to explicitly name all of the very close and somewhat close ties in their network. Such a strategy would likely be abandoned by many as overly tedious and possibly invasive. Instead, we employed a ‘summation method’.
This method, originally developed by McCarty et al. (McCarty, Killworth, Bernard, Johnsen, and Shelley, 2000), was designed to partition the network into manageable
asks the respondent to give the number of people in each of eight possible roles, for both somewhat close and very close ties. Thus, instead of asking for one or two estimated ﬁgures, the respondent can break this onerous task into 16 manageable smaller
counts. The eight roles are:
1. Members of your immediate family who do not live with you (such as parents, siblings, children)
2. Other relatives
4. People you currently work with, or go to school with
5. People you only know online
6. People from organizations (such as church, sports leagues, business associa
7. Friends not included above
8. Other people not included above These roles are similar to the ones used in Boase et al. (2006). However, they only included ﬁve roles. Roles ﬁve through eight in the Connected Lives survey were all combined in Boase et al.’s study into “Other people who are not co-workers or neighbours, who you are very close to”. For this reason and the fact that Boase et al.’s study was a telephone survey rather than a self-administered questionnaire, comparability is possible but limited.
4.7.4 Survey results Results from the survey show that individuals have a diversity of social network structures. Most individuals include a few family members as well as several friends, but otherwise their networks vary considerably. Figure 4.1 displays the distribution of
show signiﬁcant positive skew. While the mean number of very close alters is 13.4, six individuals have at least 51 alters (outliers). One individual reports having 155 very close alters. I believe that the reliability of the count of network size decreases with the number of alters included. As respondents include more individuals, they tend to ‘round up’ the numbers. For example, the individual who has 155 very close alters has 50 very close alters online as well as 50 very close alters from organizations and 20 friends. Nevertheless, I believe that the numbers represent relative network sizes faithfully, even if the speciﬁc counts are overinﬂated. I elaborate on this claim below after reviewing the network sizes from the interview.
Apart from a few outliers there are still novel descriptive insights to be gleaned from the composition of the networks. Figure 4.1 shows the share of the individuals in the network by role plotted by network size. Each line represents the percentage of the sample who have a certain number of individuals in their network by role. The line with square marks shows family members. Intuitively, it shows that most individuals at least one family member in their network, but also that it is very uncommon to have many family members (as opposed to a large extended family). This is indicated by the sharp decline between 3 and 5 immediate family members. The networks also show that very few people have individuals known only online. Only ten percent of the networks report knowing anyone online. However, as can be seen by the ﬂat slope, if people are to include alters only known online it is common to know several of them. Another ﬁnding can be interpreted from examining all of the lines at the 6-8 member mark. This shows that people have a diversity of different roles in their network. Forty percent have at least eight friends, three neighbours, four co-workers and/or seven family members. Since the slopes for all of the lines descend very close to each other, it seems that no role is particularly prominent in most networks, nor is
The summation method used in the survey is an efﬁcient method for capturing alters.
Prior research suggests that in most cases it reasonably approximates the number of network members (McCarty et al., 1997, 2000). Nevertheless, better measures exist although they are more costly and complex to implement. The most common and wellvetted measure for ascertaining network size is a ‘name generator’ (Burt, 1984). This method, best done in an interview, asks the respondent to elicit the speciﬁc names of individuals with whom they have a speciﬁc relationship. The Connected Lives study uses a modiﬁed name generator based on socioemotional closeness. The initial name generating stage remains the same, but we use a novel technique for capturing the edges between alters as well as capturing additional data about the alters. This novel technique involves producing a real-time visual display of a person’s social network.
Name generators have a long history in social network analysis. The earliest name generators date back to the mid-1960s. Laumann’s Detroit Area study is often recognized as the ﬁrst study to include a name-generator (Laumann, 1973; Marin and Hampton, 2007). It was quickly followed by Wellman’s ﬁrst East York study and Fisher’s Detroit and Northern California studies (Wellman, 1979; Fischer, 1982). These studies were persuasive enough in their depiction of personal relationships that a name generator was included in the 1984 American General Social Survey (hereafter the American GSS; Burt, 1984; Marsden, 1987).4 There are two strategies for deciding who to include in a name generator (Hogan
et al., 2007):
1. Free recall with deﬁned scope conditions. This is the approach used in the interview. The scope condition is that individuals have to be close to ego. Another scope condition is “people with whom you discuss important matters”. This was the scope condition used in the American GSS.
2. A range of questions designed to elicit a diversity of supportive alters. This is the approach used by Fischer. The respondent would be asked to name someone “who could loan you $500”, or “who could babysit your children”. This approach was also used in the fecund Social Survey of the Networks of the Dutch (Van Der Gaag and Snijders, 2005).
Once the names have been elicited, there are two additional common stages in the process. The ﬁrst is edge-generating and the second is name interpreting. The edge generation stage is widely acknowledged to be a slow and onerous task (McCarty and Govindaramanujam, 2005). It is traditionally done using a triangular matrix. The names of the alters appear in the rows and the columns of the matrix. For each cell in To give the reader a sense of the potential impact of name generators, the 2004 American GSS included a name generator identical to the one used in 1984. Researchers comparing the two noticed a clear and distinct decline in the number of core network members (McPherson et al., 2006). This led to a mainstream media frenzy about social isolation in America in the summer of 2006, with stories in many major American newspapers (e.g., Piccalo, July 23, 2006; Vedantam, June 23, 2006).
CHAPTER 4. RESEARCH METHODS AND MEASURES 77the matrix the interviewer asks, “Does A know B?” or “Are A and B close?”. The question is usually designed to be symmetric, meaning that if A knows B, B also knows A.
This way one only has to ask half as many questions. The matrix method worked ﬁne in the General Social Survey since each respondent could only list up to six alters. The formula for determining the number of questions is n(n-1)/2. Thus the GSS needed to ask a maximum of 15 additional questions. The Connected Lives interview was meant to capture a much wider scope of alters. Given the distribution of somewhat close and very close alters, as well as past work on network size by McCarty (2005), we opted to use a maximum of 66 alters. If a respondent named this many alters (as four members of the interview sample did), it would lead to 2145 unique questions of the form “Does A know B”. This was considered unduly burdensome. For this reason we opted to use a more visual approach to edge-generation.
Rather than writing down names on a single sheet of paper, we designed a name generator template that allowed individuals to write down names in an orderly fashion, but then rearrange the names afterwards. This template (as seen in Figure 4.2), includes a center layer on which we lay 66 small “Post-Its R ” (hereafter referred to as name tags). Thirty-three on one side are colour coded and marked as ‘very close’.
Thirty-three on the other side are in a different colour and marked as ‘somewhat close’.
To keep the name tags in place, we used heavy cardboard plates with windows cut out. This allows the person to see most of the name tag, while keeping the loose edge of the name tag snugly in place. The cardboard plates were attached with binder clips so that once the respondent had ﬁnished writing down names the interviewer could remove the plates allowing the participant to easily pick-up and rearrange the name tags.
We asked respondents to lay out the name tags on a large 17” by 22” inch sheet (which is the same as 2-by-2 letter-sized sheets). On these sheets, we had printed
sets of post-its and a divider.
Each tag contains the following information:
Name, number denoting role and a ‘rank number’ that denotes order that the alters were recalled.
that the respondent could lay the name tags on the rings without overlapping between rings.
Individuals were given only two instructions for laying out the network. The ﬁrst was to place those who were closest to themselves closest to the centre and the second was to place people who knew each other closer together. These instructions made it easier to draw edges between people once the network was arranged on the page.
Respondents were given a chance to tweak the arrangement until they were satisﬁed.
While we believe we had independently stumbled upon the idea of using concentric rings of closeness, a subsequent literature review of social network techniques revealed at least three prior instances of this kind of work. One was done in the laboratory (Freeman, 2000) and two in the ﬁeld (Antonucci, 1986; Spencer and Pahl, 2006).
Of the two who used this work in an interview setting, neither had included links
within these networks. We consider this independent use a testament to the underlying logic of using this particular method.
One of the advantages of performing this task is that laying out the alters on the concentric rings allowed individuals to reassess the closeness of a tie. Individuals commonly deliberated on whether a tie was somewhat or very close. In this case, they were able to place a tie on the middle ring between the closest and the least close individuals. This also gives us a secondary measure of closeness which I suggest is more accurate than a strict dichotomy. The consequences of this are discussed in the interview results section below.
Once the name tags were arranged on the paper, the interviewer asked respondents to draw edges between the individuals in four steps:
1. Draw a circle around alters who were all very close with each other.
2. Draw edges between pairs of individuals who were very close with each other.
3. Draw a circle around alters who were all somewhat close with each other. This might include pairs of individuals who were considered very close, or even a subgroup of people who were somewhat close.
4. Draw edges between people who were somewhat close to each other.
These steps are also shown in 4.3 which is a stylized version of the actual network and layout of one of the participants (#232).
The colour of the pen used for the somewhat and very close edges corresponded to the colour of the somewhat and very close name tags. This sort of ‘mapping’ (Norman,