In this longitudinal study of retiring public sector employees in Finland, we observed a loss of, on average, one tie in personal social network during the retirement transition period but no corresponding change after retirement. The decrease was most marked in less close ties, those in the outer circle. No corresponding change was noticed in closer ties, those in the middle and inner circles. Our results are thus consistent with the previous studies suggesting the less close relationships, for example work-related network ties (van Tilburg 1992, 2003) to decrease over time and closer relationships to be more enduring (Shaw et al. 2007; Suanet et al. 2013; Wrzus et al. 2013). Our findings are also in accordance with the social convoy theory (Antonucci and Akiyama 1987) which assumes that a decrease in the number of social network ties is more likely seen in more peripheral ties than in the closest ties. In addition, the fact that this change was apparent during the retirement transition period but not during the post-retirement period suggests that a decrease in social network ties is due to retirement, thus supporting the assumptions of the social convoy model that the relations are affected by changes in a person’s circumstances.
The decrease in the number of ties in the middle and outer circles was similar during the retirement transition period in all subgroups stratified by sex, occupational status, marital status and health, suggesting that our finding is robust. The number of ties in the inner circle practically did not change during the retirement transition period or the post-retirement period in the total study population. However, during the retirement transition period there was a decrease in these closest ties among those in upper grade non-manual occupations but not among those in lower grade non-manual or manual occupations. Those in upper grade non-manual occupations may be more work-oriented than those with lower occupational status. It may be that they have larger share of coworkers in their closest ties, and thus retirement decreases the number of these ties among them.
A decrease in the closest ties was also noticed among married or cohabiting subjects but not among those who were not married. This may be related to the loss of a spouse/partner due to death or divorce, both events a person living alone cannot encounter (Wrzus et al. 2013). Also, it might be that married or cohabiting subjects are more likely to focus on their families after retirement and thus lose even the closest relationships with former coworkers, while single individuals may have a stronger incentive to stay in contact with their former coworkers or replace these ties with other close ties.
The mean number of the closest ties decreased marginally among those who had no mental disorders, while among those with mental disorders it rather appeared to increase. We have no conclusive explanation to this difference. However, since information about mental disorders was assessed before retirement (at wave − 1), it may have changed during the follow-up potentially affecting the results.
Our findings on decreasing social ties differ from the results of Sabbath et al. (2015) who found in their longitudinal study on French utility workers an overall increase in the size of close social network with family and friends across the retirement. The differences in these findings may be partly related to the operationalization of social network ties. In the study of Sabbath et al. (2015) the size of close family and friend social network was assessed, while we used categorization based on the social convoy model including three concentric circles representing different levels of closeness, potentially including also other types of close relationships. In addition, there was a marked age and gender difference between the studies. In the study of Sabbath et al. (2015) the mean retirement age was around 55 years and 81% of the study sample comprised of men, while in our study the mean retirement age was over 63 years and most participants were women (84%). Therefore, the difference in results may at least partly reflect potential age and gender differences in social network changes across the retirement transition.
It is of note that the time frame used in our study was relatively short, from around 0.5 years before retirement till around 1.5 years after retirement. As retirement may be regarded as a process during which the contacts with former coworkers may be lost first and then compensated by other ties later, the timing of the measurements around retirement may affect the noticeable changes (Moen 2000). This may also explain the different results between our study and the study of Sabbath et al (2015). In their study the assessment points were 13 years apart from each other, and therefore, the timing of retirement relative to pre- and post-retirement measurements has differed considerably from our study. It is thus possible that participants of these studies have been at different stages of retirement process. This idea was supported by the fact that those participants in the study of Sabbath et al. (2015) who had been retired longer were more likely to increase the number of network ties compared to those who had retired more recently.
To our knowledge, this study is one of the first attempts to prospectively examine the changes in individuals’ social networks across the retirement. The main strength of the study is therefore study design allowing an assessment of annual changes in ties of different levels of closeness using repeat data on social network ties before and after retirement. It is important to assess ties with different levels of closeness separately since as it has been shown, less close ties may be particularly beneficial for health and well-being of an individual, even irrespective of the number of the closest ties (Sandstrom and Dunn 2014; Kauppi et al. 2018; Pan and Chee 2019). This may at least partly relate to that compared to the closest ties, less close ties are more often voluntary, less time-consuming and intimate, and are thus likely to avoid some negative social influences. Therefore, the changes in ties with different level of closeness may affect differently on individual’s well-being. Another strength was the homogeneity of our sample which did not include early retirement due to work disability. Thus, there were no major differences in health-related selection around the retirement transition between the participants.
However, some limitations are noteworthy. First, we had no information about the members or quality of relations in personal social network. Yet, as it has been shown previously, the closest ties most often include family members, while less close ties are likely to include for example less close family members and friends (Antonucci et al. 2004). Also, the size of social network, especially the number of the closest ties, can be regarded as an indicator of potentially available support even if we had no information about the quality of relations or actual support provided through them. Second, we could not assess the degree of potential turnover within the social networks. For example, the number of ties in the innermost circles appeared to decrease only slightly across the retirement transition. However, we did not know whether the ties in different circles were the same in each waves of retirement process. It might be that some of the closest ties are being lost and new close ties being achieved as replacements. Also, it might be that some ties which earlier were less close are converted into closer ties during and after retirement. It has been shown that social network turnover may be notable over longer time period, the rate of which is largely affected by some major life events as well as by some background characteristics, such as health and socioeconomic status (Wellman et al. 1997; Cornwell 2015; Badawy et al. 2019). However, since the time frame of the present study was relatively short, it is quite unlikely that the degree of social turnover would be high. Yet, since it has been shown that social network turnover, social disadvantage and poor health may be interlinked (Cornwell 2015), future studies should assess the potential turnover in social networks during the retirement transition and to examine whether the changes in social network characteristics associate, e.g., with an individual’s adaptation to a new life situation. In addition, due to relatively short time period it was not possible to accurately examine potential effects of aging on social network change. However, our finding that the change was apparent during the retirement transition period, and plateaued after that suggest a decrease to be rather due to retirement than to aging.
Our study sample comprised mostly of women which might be a limitation since women are often regarded more gregarious, to have larger and more diverse networks than men and are more likely to participate in the study, which might bias the results. However, this gender distribution represents well public sector employees in Finland (Statistics Finland 2016). Also, our results did not show any differences in changes between the genders over time.
Finally, although our final analytical sample was highly similar to the eligible population in terms of age, gender and occupation, and findings may thus be generalizable to public sector employees in Finland, there might be other, e.g., personality factors, such as loyalty and conscientiousness, which differ between participants and those who did not respond. Future studies should examine whether similar changes in social ties occur in other countries with different work cultures and pension systems.