Download as pdf or txt
Download as pdf or txt
You are on page 1of 51

INVEST

#NewWelfareState

INVEST Working Papers 70/2023

Social Media Contact with Kin


and Happiness in Two
Generations

Kristiina Tammisalo
Mirkka Danielsbacka
Antti O. Tanskanen
Bruno Arpino

30.03.2023

ISSN 2737-0534
The Inequalities, Interventions, and New Welfare State (INVEST) ecosystem brings together the
efforts of over 200 researchers and advances cross-disciplinary research. It forms an international
collaboration platform that connects researchers, practitioners, public organisations, third sector
organisations, companies, students, families and other citizens that work with us in our mission to
increase the wellbeing of the Finnish society during childhood, youth and early adulthood. The
INVEST ecosystem can be harnessed to solve welfare state and public sector problems and provide
greater insights to governments as they form policy. It is involved in large-scale policy reforms.
INVEST is funded by the Academy of Finland Flagship Programme and the host institutions and
affiliates multiple projects funded by various external research funders. The INVEST working paper
series has subseries for the long-term projects funded by the Strategic Research Council (SRC).
Social media contact with kin and happiness in two generations

Tammisalo, K.1*, Danielsbacka, M. 1, Tanskanen, A. O. 1 & Arpino, B. 2

* Corresponding author

kristiina.a.tammisalo@utu.fi

1
Department of Social Research, University of Turku. Assistentinkatu 7, 20014 Turku, FInland
2
Bocconi University. Via Roberto Sarfatti, 25, 20100 Milano MI, Italia

Competing interests

The authors declare none.


Abstract

The prior literature suggests that older adults, in particular, may benefit from social media use in

terms of increased well-being because social media facilitate family communication. Although,

social media’s potential for increasing well-being among older adults has been recognized, it has

not been rigorously explored. In this study, we test whether social media contact (SMC) with family

members serves as a source of happiness for older (aged 68–74) versus younger and middle-aged

(aged 19–56 years) adults. Using population-based data from Finland, we examine to what extent

self-rated happiness is dependent on whether or not the respondents sustain SMC with given family

members. In the analyses we use 2,807 social media users. The family members examined are

daughters, sons, sisters, and brothers for both generations, and in addition, grandchildren for the

older generation, and mothers and fathers for the younger generation. Propensity score matching

was used in order to improve the credibility of the estimated associations. Conflicting with the

socioemotional selectivity theory, it was found that SMC with family members did not increase

happiness in the generation of older adults. However, in the generation of younger adults, SMC

with some of the tested family members was associated with increased happiness. This study adds

to the knowledge about life-stage specific factors that contribute to well-being in the digital era.

Keywords: Happiness, Family, Social media, Strong ties

Introduction

Social media, such as Facebook, Instagram, Twitter, and Snapchat, increase the overall potential for

communication by overcoming barriers of time and place. They also allow users to communicate

more frequently, in more spontaneous ways, to more than one recipient at once, and in ways that

complement other forms of communication (Danielsbacka, Tammisalo and Tanskanen 2022;

Dienlin, Masur and Trepte 2017). Social media may also gratify a range of idiosyncratic needs in
that they provide a wider range of communication affordances than traditional methods of

communication (e.g., (Taipale 2019). These positive aspects of social media have prompted some

scholars to consider social media’s potential to enhance the well-being that individuals derive from

their personal relationships (Chan 2015; Chan 2018; Dienlin, Masur and Trepte 2017; Ishii 2017;

Newman, Stoner and Spector 2019; Vriens and van Ingen 2018).

The positive potential of social media as an enhancer of well-being has been a prominent focus in

studies considering older adults (Nef et al. 2013; Newman, Stoner and Spector 2019). The findings

have been promising; for example, social media has been found to be associated with the

preservation of personal social capital in older age (Erickson 2011; Simons et al. 2021) and various

measures of well-being such as lower levels of social isolation and loneliness (Chang, Choi, Y. H.:

Bazarova, N. N. and Löckenhoff 2015; Chen and Schulz 2016; Hutto et al. 2015).

However, not all studies have found positive effects which emphasizes the need for improved

understanding of the mechanisms by which social media affects well-being (Newman, Stoner and

Spector 2019). A spate of studies have attempted to unmask potential mechanisms by differentiating

between types of social media use. For example, using social media for communication and content

creation is associated with increased well-being through increased social capital and feelings of

connectedness; conversely, passively viewing content is associated with decreased well-being

through negative emotions such a envy (Seabrook, Kern and Rickard 2016; Verduyn et al. 2017). In

addition, differences in the social media environment, such as proportion of strangers in one’s social

media network, have been found to moderate the effect of social media on well-being (Seabrook,

Kern and Rickard 2016).


A mechanism that has not been rigorously explored thus far is the use of social media for family

communication. We propose that a potential mechanism explaining the association between social

media use and well-being among older adults is that social media is used to interact with family

members.

Literature review

As individuals age, relationships with family tend to gain relative importance (Wrzus et al. 2013).

Of the typically non-resident family members, sibling, parent-adult child, and grandparent-

grandchild relationships are the closest and more enduring across life transitions than non-family

relationships (Roberts and Dunbar 2011; Tanskanen and Danielsbacka 2019; Buchanan and

Rotkirch 2021). This age-related “family shift” has been found to apply to social media networks

too. Older adults use social media to connect with family members more than younger adults do

(Brandtzaeg, Heim and Kaare 2010; Hutto et al. 2015; Tammisalo et al. 2022; Zickuhr and Madden

2012). Further attesting to the significance of family in older age, older adults’ most frequent self-

appointed reason for adopting social media is to connect with their children and grandchildren (Nef

et al. 2013; Newman, Stoner and Spector 2019; Zickuhr and Madden 2012). Despite the abundant

evidence, social media’s effects on well-being have rarely been explicitly attributed to older adults’

preference for communicating with family members. In other words, it has not been considered a

major factor in the association between social media use and well-being that social media enable

older adults to be more involved in the lives of their close family members.

Partially addressing this issue, two studies from Hong Kong have considered the role of age and

social network composition in the association between multimodal communication (including social

media) and well-being (Chan 2015; Chan 2018). Notably, network composition in these studies

means the proportion of strong ties, without differentiating between family and non-family. These
studies suggest that social media’s benefits to older adults are due to age-related changes in the

proportion of strong ties in their networks. In the findings, older age groups (from 35 and 55 years

to over 70 years) experienced positive emotions from using technology to connect with their strong

ties (including both close friends and family members), while the younger age groups did not.

Younger adults, on the other hand, experienced a decrease in well-being from using media to

connect with a larger network of weak-ties (i.e., non-intimate friends and acquaintances). These

findings suggest that age-related changes in network composition may play an important role in the

association between social media use and well-being.

The theoretical basis of the aforementioned findings is provided by the socioemotional selectivity

theory. According to this theory, older adults are more strongly motivated to uphold a smaller

number of social ties and they derive emotional gratification from such social interaction

(Carstensen, Isaacowitz and Charles 1999; Löckenhoff and Carstensen 2004). This prediction

derives from the theory’s assumption that with advancing age there is a motivational shift from

informational goals towards emotionally gratifying goals prompted by the realization of finite time

towards the end of life. As a result of this process, older individuals more often than younger

individuals are expected to invest in a smaller number of emotionally gratifying relationships –

most often family (Carstensen, Isaacowitz and Charles 1999; Löckenhoff and Carstensen 2004;

Wrzus et al. 2013). In line with this theory, studies show that smaller strong-tie networks and

family communications are indeed more typical for older adults communication (Wrzus et al. 2013)

and this applies also to social media networks (Brandtzaeg, Heim and Kaare 2010; Chang, Choi, Y.

H.: Bazarova, N. N. and Löckenhoff 2015). However, while these studies show that older adults

selectively communicate with a smaller number of close ties and do so also in the social media

environment, it is still an open question whether older adults are able to increase their well-being by

doing so. Furthermore, family relationships have thus far only been examined together with friends
(Chan 2015; Chan 2018; Chang, Choi, Y. H.: Bazarova, N. N. and Löckenhoff 2015), and different

relationship types in the family, such as parent-child, grandparent-grandchild, or siblings, have not

been considered separately.

To our knowledge, the present investigation is the first to consider age-variant associations between

SMC with family members and happiness. While social media and psychological well-being have

been studied extensively, different age groups have been mostly studied in isolation. To address this

limitation, we use a data set which comprises a large age range to investigate whether SMC with

family members is associated with happiness in different ways depending on age. Our study

incorporates a generation of older adults aged from 68 to 74 years and the generation of their adult

children ranging in age from 19 to 56 years.

Research questions

Based on the socioemotional selectivity theory and the indicative evidence from earlier studies, we

pose the following research questions:

- Do older adults use social media for family communication more than younger adults?

- Does SMC with family members increase happiness among younger and older adults?

Data

We use the third wave of the Generational Transmissions in Finland (Gentrans) survey gathered in

2018 and 2019. The Gentrans surveys gather information about lining conditions, social networks,

and well-being of Finnish adults. The data include detailed information about networks,

communication with kin as well as a large range of sociodemographic background information. For
the purpose of this study, we focus on questions related to social media use to connect with family

members and happiness as well as relevant background variables as controls.

The data comprise two family generations: The older generation represents Finnish baby boomers

born in 1945–1950 (n = 2,663, response rate 66.4 %) and the younger generation is the baby

boomers’ adult children born in 1962–1999 (n = 1,945, response rate 55.6 %). During the time of

data collection, the respondents were aged 68–74 years in the older generation and 19–56 years in

the younger generation. Because of the large age range of the younger generation, we further

divided them into 19–39-year-olds (younger adults) and 40–56-year-olds (middle-aged adults) for

further analyses in addition to treating them as one age group in the main analyses. In order to

investigate the association between SMC with family members and happiness, we exclude from the

main analyses those who are not social media users; the final sample sizes were 1,265 for the older

generation and 1,542 for the younger generation (475 younger adults; 1,045 middle-aged adults).

The sub-analyses testing SMC with each family member type (daughter, son, mother, etc.) are run

only using those respondents who have the family member in question (sample sizes for each sub-

analysis are provided in the online supplementary material).

Our dependent variable is happiness measured as self-reported happiness on a 11-point scale

ranging from 0 (“Very unhappy”) to 10 (“Very happy”). The explanatory variable is SMC with a

family member. The respondents indicated with whom (from a list of kin, affinal kin, and non-kin)

they interact via social media. To guide their responses, the following example of social media were

listed: Facebook, Instagram, Twitter, and Snapchat. From the responses, we use the following

indicators of SMC with family:

1) SMC with a daughter (yes/no)

2) SMC with a son (yes/no)


3) SMC with mother (yes/no, only for younger generation)

4) SMC with father (yes/no, only for younger generation)

5) SMC with a grandchild (yes/no, only for older generation)

6) SMC with a sisters (yes/no)

7) SMC with a brothers (yes/no)

In order to detect whether and how SMC with family members is associated with happiness, we

control for a range of sociodemographic variables. The sociodemographic variables used as controls

are: Age calculated according to the year of birth; Sex collected as a binary variable

(male/female); Education collected as a nine-level ordinal variable ranging from incomplete

elementary school to doctorate degree and subsequently recoded into a four-level ordinal variable

from 1 = completed or incomplete elementary school to 4 = higher education; Self-rated economic

situation measured on a four-level ordinal scale ranging from 1 = high income to 4 = low income,

subsequently inverted for analysis so that 4 indicates the highest income level; Marital

status collected as a categorical variable with six response options and subsequently recoded as the

following categories: 1 = unpartnered, 2 = cohabiting/married, 3 = divorced, and 4 = widowed;

Health measured as a dichotomous variable 1 = absence of long-term medical issues or 0 =

presence thereof.

The control variables also included two variables that capture characteristics of respondents’ social

networks: Number of children and Number of close friends. They were included because having

close friends and relatives may affect both the independent variable (SMC) and the dependent

variable (happiness). Outlier values from 21 and up were set at 20.


In addition, age of daughters and sons in the younger age group and age of grandchildren in the

older age group were added as controls in the corresponding sub-analyses (i.e., SMC with

daughter/son/grandchild) because some children/grandchildren of the respondents may be less

likely to use social media due to young age. Ages were approximated using the available

information. In the case of children, the mean age of the four eldest children was used; in the case

of grandchildren the mean age of the youngest children of the four eldest children was used.

Ethical permission for the Gentrans surveys has been granted by the Ethical Board of Statistics

Finland in 2006. The respondents have given their informed consent to the use of their data for

research, and the data have been anonymized. All users of the data have committed to follow the

Statistics Finland ethical rules by accepting The Pledge of Secrecy of Holder of Permission to Use

Data.

Methods

Three estimates of the associations between SMC with a family member and happiness are

presented for each family member type: linear regression without controls, linear regression with

controls, and propensity score matching (PSM) approach (excluding the analyses in which sample

sizes did not allow for PSM). Note that the same control variables (presented in detail previously)

are used in the regression and PSM approach. First, the regressions show whether there is an

association between SMC with a family member and happiness, and whether the association is a

matter of selection (i.e., whether those who are happier due to factors accounted for by the control

variables are also more in SMC with family members). After the regressions, the PSM approach is

used to improve the credibility of causal inference. PSM is a quasi-experimental method which is

based on retrospectively forming comparable groups (“treatment” and “control”) from existing data.
The “treatment group” comprises those who have SMC with a given family member and the

“control group” comprises those who do not have SMC with that given family member.

In order to form comparable groups, a propensity score is calculated for each individual using the

control variables. The score is an estimate of the probability for an individual to have SMC with a

given family member given their observed control variables. Observed variables that influence the

probability of adopting social media and happiness, i.e., the possible confounders, are included as

controls and, therefore, contribute to the propensity score. After calculating the propensity scores

for each individual, individuals with equivalent or similar scores from the treatment (SMC with

given family member) and control groups (no SMC with given family member) are matched.

Unmatched individuals are dropped out of the sample.

The PSM method allows us to estimate the counter-factual happiness of respondents had they

belonged to the other group by matching respondents from the treatment and control groups with

equivalent or similar values in the propensity scores, and so of the control variables. The key

advantage of PSM over regression is that PSM accounts for control variables in a semi-parametric

way; thus, its results are less sensitive to violations of regression model assumptions. In addition,

PSM allows checking whether the matching across the controls was successful in a given

application (the so-called balancing of control variables).

For the propensity score matching, all age variables (excluding the ages of the older generation’s

respondents whose age range is relatively narrow) were recoded into 5-year-intervals to guarantee a

finer matching of the age distribution. Results from PSM in the main text are presented in terms of

Average Treatment effects on the Treated (ATT), which measure how mean happiness of those who

have SMC with a specific family member would change if they did not have SMC with that family
member. The balances of the matching across control variables for each analysis are presented in

the online supplementary material.

Results

Descriptive findings

Among the older generation, 47.5 per cent are social media users, while among the younger

generation, 79.3 per cent use social media Tables 1 and 2 present descriptive statistics of the social

media users in the older and younger generations, respectively.

Table 1. Descriptive statistics. Social media users in the older generation (68–74 years) (n and

%/mean).

n %/mean SD

Age (mean) 1265 70.4 1.67

Gender (%)
Women 802 63.4 -
Men 463 36.6 -

Education (%)
Elementary school 333 27.1 -
Upper secondary/Vocational 604 49.1 -
College/Lower university degree 115 9.3 -
Higher academic degree 178 14.5 -

Economic situation1(mean) 1265 1.81 0.80

Marital status (%)


Married/Cohabiting 888 71.1 -
Unmarried 68 5.4 -
Divorced 160 12.8 -
Widowed 133 10.6 -

Health (long-term medical issue) (%)


Yes 767 60.6 -
No 470 37.2 -

Number of children (mean) 1265 2.2 1.59

Number of grandchildren (mean) 1265 3.6 5.10

Number of friends (mean) 1265 5.5 3.95


1
1 = “low income,” 4 = “high income”
Table 2. Descriptive statistics. Social media users in the younger generation (19–56 years) (n and

%/mean).

n %/mean SD

Age (mean) 1542 42.0 5.85

Gender (%)
Women 1071 85.5 -
Men 471 68.0 -

Education (%)
Elementary school 27 1.8 -
Upper secondary/Vocational 541 35.3 -
College/Lower university degree 449 29.3 -
Higher academic degree 516 33.7 -

Economic situation1(mean) 1542 2.00 0.75

Marital status (%)


Married/Cohabiting 1168 75.7 -
Unmarried 231 15.0 -
Divorced 137 8.9 -
Widowed 6 0.4 -

Health (long-term medical issue) (%)


Yes 609 39.8 -
No 921 59.7 -

Number of children (mean) 1542 1.8 1.49

Number of friends (mean) 1542 4.3 3.27


1
1 = “low income,” 4 = “high income”

The prevalence of SMC with family members varies across generations (Figures 1 and 2). In the

older generation, SMC with children is high: 87.1 per cent of those who have at least one daughter

have SMC with a daughter and 78.2 per cent of those with at least one son have SMC with a son.

The corresponding percentages among the younger generation are 32.9 per cent (SMC with

daughter) and 27.1 per cent (SMC with son). In addition, in the younger generation 35.9 and 20.5

per cent have SMC with their mother and father, respectively. In the younger generation, friends

and coworkers are the most typical SMCs. All in all, older adults’ social media use can be described

as more family-oriented.
100%

90%

80%

70%

60%

50%

40%

30%

20%

10%

0%
Daughter Son (697) Grandchild Sister (920) Brother Friend Coworker Neighbor Someone
(714) (895) (915) (1265) (1265) (1265) else (1265)

Figure 1. Percentage of social media users in the older generation (aged 68 to 74) who sustain SMC

with kin and non-kin. In the case of relatives, (n) indicates the number of respondents with the

given relative.
100%

90%

80%

70%

60%

50%

40%

30%

20%

10%

0%
Daughter Son (1286) Mother Father Sister Brother Friend Coworker Neighbor Someone
(1264) (1453) (1302) (1432) (1437) (1542) (1542) (1542) else (1542)

Figure 2. Percentage of social media users in the younger generations (aged 19–56) who sustain

SMC with kin and non-kin. In the case of relatives, (n) indicates the number of respondents with the

given relative.

Linear regressions and PSM approach

Using linear regression and PSM we test whether SMC with family members is associated with

happiness separately for the two generations. We test this separately for contact with each kin type:

daughter, son, sister, and brother for both generations, and additionally, grandchild for the older

generation and mother and father for the younger generation. The analyses are performed on

respondents who reported using social media and each analysis is performed only on respondents

who have the family member in question. Because of the large age range of the younger generation,

we further divided them into 19–39-year-olds (younger adults) and 40–56-year-olds (middle-aged

adults) to check for differences between age groups. In addition, the results of the analyses are
presented both separately for men and women as well as together. The analyses are presented for

the older generation in Table 3, for the whole younger generation in Table 4, and for the younger

generation divided in to younger and middle-aged adults in Table 5 and 6, respectively. For the

PSM, the balance across the control variables for the matched and unmatched samples are provided

in the online supplementary material along with the specific matching algorithm used in each

analysis. In addition, sample sizes for each of the analyses are also provided in the online

supplementary material.

Older generation

The results of the analyses do not display the expected pattern. That is, older adults’ SMC with

family members does not contribute to happiness (Table 3). The only exception is men’s SMC with

their daughters, however, this association is detected only using the linear regression with controls

and with statistical significance at the 10 per cent level (β=.516, p=.081). The PSM method does not

confirm the statistical significance of this association. In addition, women’s SMC with a brother

shows a negative association in the linear regressions with controls, although also statistically

significant only at the 10 per cent level (β=-.282, p=.06). The PSM method does not confirm the

statistical significance of this association either.


Table 3 Estimated coefficients for SMC with family members and self-rated happiness (1=very unhappy, 10=very happy) in the older generation

(68–74 years).

Regression without controls Regression with controls PSM1


Coef. 95% CI Coef. 95% CI Coef. 95% CI2
Daughter .343* -.009 .695 .133 -.235 .500 .062 -.348 .472
Women .151 -.321 .622 .002 -.473 .476 .1323 -.476 .741
Men .599** .063 1.135 .516* -.064 1.097 .3113 -.660 1.281
Son .048 -.247 .343 -.060 -.369 .248 -.129 -.419 .160
Women -.002 -.376 .371 -.119 -.513 .274 -.097 -.484 .289
Men .137 -.346 .621 .140 -.367 .646 .253 -.320 .827
Grandchild .145 -.079 .369 .161 -.094 .416 .117 -.178 .413
Women .051 -.249 .351 .214 -.133 .552 .194 -.231 .618
Men .338* -.006 .682 .095 -.279 .468 .016 -.389 .421
Sister .252** .020 .484 .060 -.182 .302 .006 -.251 .262
Women .175 -.117 .468 .043 -.257 .344 .050 -.240 .339
Men .345* -.048 .738 .144 -.267 .556 .193 -.284 .669
Brother -.033 -.259 .194 -.160 -.396 .076 -.163 -.422 .097
Women -.112 -.397 .172 -.282* -.577 .012 -.276 -.595 .043
Men .110 -.267 .487 .019 -.381 .419 .056 -.389 .500
1
Coefficients are the average treatment effects on the treated (ATT)
2
Bootstrapped confidence intervals
3
PSM sample mean bias > 5 %

Significance levels: * p < .1, ** p < .05, *** p < .01.


Table 4 Estimated coefficients for SMC with family members and self-rated happiness (1=very unhappy, 10=very happy) in the younger

generation (19–56 years)

Regression without controls Regression with controls PSM1


Coef. 95% CI Coef. 95% CI Coef. 95% CI2
Daughter .020 -.202 .242 .278** .020 .535 .371** .008 .733
Women .028 -.230 .286 .310** .001 .619 .429* -.010 .867
Men -.005 .442 .432 .098 -.388 .583 .174 -.602 .950
Son -.015 -.238 .207 .204 -.053 .460 .202 -.129 .533
Women -.063 -.316 .188 .237 -.060 .534 .328 -.147 .802
Men .114 -.342 .569 .072 -.439 .584 -.1663 -.812 .480
Mother .270*** .086 .454 .182** .012 .351 .174** .004 .345
Women .150 -.058 .358 .132 -.063 .332 .110 -.090 .310
Men .580*** .216 .944 .205 -.124 .533 .289* -.042 .621
Father .281** .050 .512 .070 -.144 .283 .041 -.146 .229
Women .078 -.176 .333 -.032 -.277 .213 -.049 -.290 .192
Men .739*** .276 1.203 .129 -.281 .539 .155 -.200 .511
Sister .358*** .088 .627 .154 -.094 .403 .104 -.170 .377
Women .094 -.223 .411 .015 -.288 .319 .010 -.319 .339
Men .682*** .175 1.189 .340 -.094 .775 .398 -.136 .931
Brother .107 -.128 .343 .077 -.140 .295 .050 -.212 .312
Women .007 -.254 .268 .019 -.232 .269 .046 -.233 .326
Men .435* -.063 .933 .130 -.306 .567 .082 -.372 .536
1
Coefficients are the average treatment effects on the treated (ATT)
2
Bootstrapped confidence intervals
3
PSM sample mean bias > 5 %

Significance levels: * p < .1, ** p < .05, *** p < .01.


Younger generation

Conversely, in the younger generation there were several age and gender specific significant

associations (p<.05) between SMC with family member and happiness. Table 4 shows the results

for the entire younger generation including the results from the linear regressions and the PSM

approach. The PSM detected that, SMC with a daughter (ATT=.371, p=.045) and mother

(ATT=.174, p=.045) increase happiness. In addition, when examining the results by gender, positive

associations that were significant at higher levels (p=.05–.100) suggest that the aforementioned

positive associations were driven by women’s SMC with their daughters (ATT=.429, p=.055) and

men’s SMC with their mothers (ATT=.289, p=.087). The regressions with controls detected similar

results.

An additional and relatively consistent result among men across generations is that SMC with many

of the tested family members is positively associated with happiness before controlling for

covariates (Tables 4). This means that happier men are more likely to have SMC with their family

members.

Table 5 and 6 show the results for the younger generation divided into younger (19–39-year-olds)

and middle-aged adults (40–56-year-olds). These analyses are based only on the linear regression

models, because the sample sizes do not allow implementing the PSM method in these groups.

Based on the additional analyses with the divided younger generation’s sample, the results reported

above for the younger generation appear to be mostly driven by the middle-aged adults. However,

the limited samples especially in some of the analysis for the youngest adults do not allow further

interpretations (online supplementary material for the sample sizes in each analysis).
Table 5 Estimated coefficients for SMC with family members and self-rated happiness (1=very

unhappy, 10=very happy) in middle-aged adults (40–54 years)

Regression without controls Regression with controls


Coef. 95% CI Coef. 95% CI
Daughter .169 -.100 .439 .331** .034 .628
Women .269* -.049 .587 .442** .081 .803
Men -.097 -.609 .414 -.021 -.567 .525
Son .129 -.136 .395 .202 -.085 .490
Women .115 -.188 .419 .289* -.048 .626
Men .162 -.368 .692 -.037 -.593 .518
Mother .280** .054 .506 .209** .002 .416
Women .138 -.122 .397 .116 -.127 .359
Men .657*** .226 1.089 .346* -.039 .732
Father .268* -.024 .561 .050 -.217 .317
Women .063 -.263 .389 -.063 -.376 .250
Men .711** .139 1.282 .155 -.329 .640
Sister .357** -.012 .701 .175 -.143 .494
Women .298 -.113 .710 -.183 -.217 .583
Men .328 -.303 .959 .331 -.185 .847
Brother .168 -.131 .467 .116 -.059 .391
Women .200 -.136 .535 .223 -.095 .541
Men .190 -.440 .820 -.006 -.540 .528
Significance levels: * p < .1, ** p < .05, *** p < .01

Table 6 Estimated coefficients for SMC with family members and self-rated happiness (1=very

unhappy, 10=very happy) in young adults (19–39 years)

Regression without controls Regression with controls


Coef. 95% CI Coef. 95% CI
Daughter -.262 -.727 .202 .134 -.415 .682
Women -.485* -1.011 .041 -.157 -.803 .488
Men .407 -.565 1.379 1.356** .198 2.515
Son -.164 -.672 .345 .322 -.318 .962
Women -.379 -.931 .173 .063 -.646 .772
Men .444 -.723 1.612 1.296 -.318 2.910
Mother .247 -.074 .567 .117 -.180 .415
Women .168 -.187 .523 .205 -.135 .545
Men .375 -.317 1.068 -.109 -.771 .552
Father .302 -.072 .675 .107 -.244 .459
Women .102 -.303 .508 .032 -.357 .421
Men .807** .011 1.602 .129 -.664 .923
Sister .366* -.067 .799 .145 -.261 .551
Women -.244 -.739 .250 -.137 -.613 .345
Men 1.408*** .558 2.258 .674 -.222 1.570
Brother -.007 -.386 .373 -.105 -.468 .259
Women -.380* -.788 .028 -.373* -.780 .034
Men .911** .086 1.736 .543 -.329 1.415
Significance levels: * p < .1, ** p < .05, *** p < .01
Discussion

In this study, we investigated whether older adults more than younger adults use social media for

family communication and whether this influences happiness. First, we asked whether older adults

use social media for family communication more than younger adults. In line with the earlier

literature (Brandtzaeg, Heim and Kaare 2010; Hutto et al. 2015; Tammisalo et al. 2022; Wrzus et

al. 2013; Zickuhr and Madden 2012) we found that older adults’ social media use is more family-

oriented than that of younger adults.

Second, we asked whether SMC with family members increases happiness particularly among older

adults. We did not find evidence for this in the case of older adults’ social media communication

with family members. This finding contradicts earlier evidence, which suggests that digital

communication with close ones increases the well-being of older adults (Brandtzaeg, Heim and

Kaare 2010; Chan 2015; Chan 2018; Chang, Choi, Y. H.: Bazarova, N. N. and Löckenhoff 2015;

Chen and Schulz 2016; Erickson 2011; Hutto et al. 2015; Nef et al. 2013; Newman, Stoner and

Spector 2019). While our study does find that older adults have proportionally more family

members in their social media contacts, our study does not find that SMC with family members

would be associated with subjective ratings of happiness. Hence, our results support the

socioemotional selectivity theory only partly (Carstensen, Isaacowitz and Charles 1999; Löckenhoff

and Carstensen 2004).

However, it was detected that younger adults did display associations between SMC with family

members and happiness in the case of some of the tested family relationships. In earlier studies, the

age at which the theorized selective investment in family communication occurs has varied with

most studies focusing only on old age. However, some studies have also reported the so-called

family shift and the well-being benefits to occur in middle adulthood (Chan 2015; Chan 2018;
Wrzus et al. 2013). In this vein, it has been suggested that middle-aged individuals who need to

sustain contact with both dependent children and ageing parents benefit from social media because

such platforms are highly efficient and provide low-effort means for staying updated in family

members’ affairs (Robinson et al. 2015). As for older adults, while contact with family may be an

important source of well-being for them, social media may not be the most emotionally gratifying

method of communication as social media lack many qualities of rich and intimate communication,

such as voice tones and synchronized effort (Daft and Lengel 1986; Goodman-Deane et al. 2016;

Tammisalo and Rotkirch 2022). Notably, instant messaging too, despite being text-based, has been

found to have positive effects on satisfaction in life among older adults in a number of countries

including Finland (Rosenberg and Taipale 2022). This may be because instant messages have the

ability to create a synchronous and personal sense of “connected presence” (Cui 2016). This quality

is less salient in social media.

Importantly, it is possible that while older adults gain emotional benefit from family communication

more generally they do not gain additional benefits from SMC with them. Since all the individuals

included in our analyses were social media users and social media use is correlated with other forms

of contact with family, such as meeting face-to-face and phone calls (Danielsbacka, Tammisalo and

Tanskanen 2022; Shen et al. 2017), it is possible that our sample represents those who have above-

average levels of contact in their family relationships. Furthermore, it is also important to

acknowledge that our sample may mask trends that apply to sub-groups in specific circumstances;

for example, there can be a well-being benefit when lack of other communication options are

compensated with social media (Arpino et al. 2022).

Closer examination of the younger generation shows that the associations between SMC with

family and happiness vary by age and gender. For example, the positive association between SMC
with daughters and happiness appear to be driven by middle-aged women, whereas the association

between SMC with mothers and happiness was driven by middle-aged men. These results support

the notion, that SMC with family benefits mostly middle-aged adults (Robinson et al. 2015).

In interpreting the results, the following limitations and strengths should be recognized. First,

despite controlling for a range of possible covariates that have been previously identified as factors

determining social media use and are likely also effect happiness (Tammisalo et al. 2022), the effect

of unobserved confounders cannot be entirely ruled out when using non-experimental data. In other

words, those with SMC with family may differ in some unobserved systematic way from those who

do not have SMC with family and this differentiating factor may also be associated with happiness

(e.g. sociability or extraversion). The PSM method, however, improves the credibility of the

estimated associations and is therefore a strength of the study. In fact, it allows focusing on

comparisons between groups only for comparable individuals, while regression models can be

biased due to extrapolations. A second limitation pertains to the definition of social media. Social

media are rapidly transforming technologies that defy precise definitions. Absent more detailed

information about the respondents’ social media use, our measure leans on the respondents’

interpretation of the question and the brief description provided (Facebook, Instagram, Twitter, and

Snapchat were the provided examples). Finally, the outcome measure of happiness is based on only

one question, therefore, capturing only one aspect of overall well-being. Other possible positive

repercussion of SMC that were not considered here could be, for example, digital inclusion, access

to information, and possibilities to participate in society which all contribute to a more universal

concept of well-being. Although our study did not find SMC with family to increase happiness

among older adults, both sustaining contact with family as well as the use of digital media for

various activities can both be considered positive contributions to active ageing.


Statement of ethical approval

Ethical permission for the Gentrans surveys has been granted by the Ethical Board of Statistics

Finland in 2006.

Statement of funding

The study is part of the NetResilience consortium funded by the Strategic Research Council at the

Academy of Finland (Grant number 345183) and INVEST flagship funded by the Academy of

Finland (Grant number 320162). Additional funding was received from Academy of Finland (Grant

numbers 338869 and 331400 and 325857 and 317808). The financial sponsors have not influenced

the design, execution, analysis or interpretation of data, or writing of the study.

Declaration of contribution of authors

Concept: KT, MD & AOT

Literature review, text: KT

Data analyses: KT, BA

Data administration: AOT, MD

Text editing: MD, AOT, BA

Competing interests

The authors declare none.


References

Arpino, B., Meli, E., Pasqualini, M., Tomassini, C. and Cisotto, E. 2022. Determinants of

grandparent–grandchild digital contact in Italy. Genus, 78, 1, 1–22.

Brandtzaeg, P. B., Heim, J. and Kaare, B. H. 2010. Bridging and bonding in social network sites –

investigating family-based capital. International Journal of Web Based Communities, 6, 3,

231.

Buchanan, A. and Rotkirch, A. (eds) 2021. Brothers and Sisters: Sibling Relationships Across the

Life Course. Palgrave Macmillan, Cham.

Carstensen, L. L., Isaacowitz, D. M. and Charles, S. T. 1999. Taking time seriously: A theory of

socioemotional selectivity. American Psychologist, 54, 3, 165–81.

Chan, M. 2018. Digital communications and psychological well-being across the life span:

Examining the intervening roles of social capital and civic engagement. Telematics and

Informatics, 35, 6, 1744–54.

Chan, M. 2015. Multimodal connectedness and quality of life: Examining the influences of

technology adoption and interpersonal communication on well-being across the life span.

Journal of computer-mediated communication: JCMC, 20, 1, 3–18.

Chang, P. F., Choi, Y. H.: Bazarova, N. N. and Löckenhoff, C. E. 2015. Age differences in online

social networking: Extending socioemotional selectivity theory to social network sites.

Journal of broadcasting & electronic media, 59, 2, 221–39.

Chen, Y.-R. R. and Schulz, P. J. 2016. The effect of information communication technology

interventions on reducing social isolation in the elderly: A systematic review. Journal of

medical Internet research, 18, 1, e18.


Cui, D. 2016. Beyond “connected presence”: Multimedia mobile instant messaging in close

relationship management. Mobile Media & Communication, 4, 1, 19–36.

Daft, R. L. and Lengel, R. 1986. Organizational information requirements, media richness and

structural design. Management science, 32, 5, 554–71.

Danielsbacka, M., Tammisalo, K. and Tanskanen, A. O. 2022. Digital and traditional

communication with kin: displacement or reinforcement? Journal of Family Studies.

Dienlin, T., Masur, P. K. and Trepte, S. 2017. Reinforcement or displacement? The reciprocity of

FtF, IM, and SNS communication and their effects on loneliness and life satisfaction.

Journal of Computer-Mediated Communication.

Erickson, L. B. 2011. Social media, social capital, and seniors: The impact of Facebook on bonding

and bridging social capital of individuals over 65. In Proceedings of the Seventeenth

Americas Conference on Information Systems.

Goodman-Deane, J., Mieczakowski, A., Johnson, D., Goldhaber, T. and Clarkson, P. J. 2016. The

impact of communication technologies on life and relationship satisfaction. Computers in

human behavior, 57, 219–29.

Hutto, C. J., Bell, C., Farmer, S., Fausset, C., Harley, L., Nguyen, J. and Fain, B. 2015. Social

media gerontology: Understanding social media usage among older adults. Web intelligence,

13, 1, 69–87.

Ishii, K. 2017. Online communication with strong ties and subjective well-being in Japan.

Computers in human behavior, 66, 129–37.


Löckenhoff, C. E. and Carstensen, L. L. 2004. Socioemotional selectivity theory, aging, and health:

the increasingly delicate balance between regulating emotions and making tough choices.

Journal of personality, 72, 6, 1395–1424.

Nef, T., Ganea, R. L., Müri, R. M. and Mosimann, U. P. 2013. Social networking sites and older

users: A systematic review. International psychogeriatrics, 25, 7, 1041–53.

Newman, L., Stoner, C. and Spector, A. 2019. Social networking sites and the experience of older

adult users: a systematic review. Ageing & Society, 41, 2, 377–402.

Roberts, S. G. B. and Dunbar, R. I. M. 2011. The costs of family and friends: an 18-month

longitudinal study of relationship maintenance and decay. Evolution and Human Behavior,

32, 3, 186–97.

Robinson, L., Cotten, S. R., Ono, H., Quan-Haase, A., Mesch, G., Chen, W., Schulz, J., Hale, T. M.

and Stern, M. J. 2015. Digital inequalities and why they matter. Information,

Communication and Society, 18, 5, 569–82.

Rosenberg, D. and Taipale, S. 2022. Social and satisfied? Social uses of mobile phone and

subjective wellbeing in later life. Human Technology, 18, 1, 45–65.

Seabrook, E. M., Kern, M. L. and Rickard, N. S. 2016. Social Networking Sites, Depression, and

Anxiety: A Systematic Review. JMIR Mental Health, 3, 4, e50.

Shen, C., Wang, M. P., Chu, J. T., Wan, A., Viswanath, K., Chan, S. S. C. and Lam, T. H. 2017.

Sharing family life information through video calls and other information and

communication technologies and the association with family well-being: Population-based

survey. JMIR mental health, 4, 4, e57.


Simons, M., Reijnders, J., Peeters, S., Janssens, M., Lataster, J. and Jacobs, N. 2021. Social network

sites as a means to support personal social capital and well-being in older age: An

association study. Computers in Human Behavior Reports, 3, 100067.

Taipale, S. 2019. Intergenerational Connections in Digital Families. Springer, Cham.

Tammisalo, K., Danielsbacka, M., Andersson, E. and Tanskanen, A. O. 2022. Predictors of Social

Media Use in Two Family Generations. Frontiers in Sociology, 6.

Tammisalo, K. and Rotkirch, A. 2022. Effects of information and communication technology on the

quality of family relationships: A systematic review. Journal of social and personal

relationships, 39, 9, 2724–65.

Tanskanen, A. O. and Danielsbacka, M. 2019. Intergenerational Family Relations: An Evolutionary

Social Science Approach. Routledge.

Verduyn, P., Ybarra, O., Résibois, M., Jonides, J. and Kross, E. 2017. Do social network sites

enhance or undermine subjective well-being? A critical review. Social issues and policy

review, 11, 1, 274–302.

Vriens, E. and van Ingen, E. 2018. Does the rise of the Internet bring erosion of strong ties?

Analyses of social media use and changes in core discussion networks. New media &

society, 20, 7, 2432–49.

Wrzus, C., Hänel, M., Wagner, J. and Neyer, F. J. 2013. Social network changes and life events

across the life span: A meta-analysis. Psychological bulletin, 139, 53–80.

Zickuhr, K. and Madden, M. 2012. Older adults and Internet use. Pew Research Center.
Appendix 1

Table 1 Sample sizes in each of the sub-analyses in the older generation (68–74 years)
All Women Men
Daughter
Total sample 855 529 326
Treatment 742 468 274
Control 113 61 52
Son
Total sample 846 542 304
Treatment 673 432 241
Control 173 110 63
Grandchild
Total sample 1,013 657 356
Treatment 650 462 188
Control 363 195 168
Sister
Total sample 949 604 345
Treatment 542 376 166
Control 407 228 179
Brother
Total sample 978 632 346
Treatment 397 246 151
Control 581 386 195

Table 2 Sample sizes in each of the sub-analyses in the younger generation (19–56 years)
All Women Men
Daughter
Total sample 828 599 229
Treatment 419 304 115
Control 409 295 114
Son
Total sample 841 598 243
Treatment 352 249 103
Control 489 349 140
Mother
Total sample 1,456 1,010 446
Treatment 524 353 171
Control 932 657 275
Father
Total sample 1,306 897 409
Treatment 271 181 90
Control 1035 716 319
Sister
Total sample 928 652 276
Treatment 727 530 197
Control 201 122 79
Brother
Total sample 909 645 264
Treatment 570 395 175
Control 339 250 89

Table 3 Sample sizes in each of the sub-analyses in the younger generation’s middle-aged adults
(40–56 years)
All Women Men
Daughter
Total sample 619 442 177
Treatment 368 268 100
Control 251 174 77
Son
Total sample 624 438 186
Treatment 317 223 94
Control 307 215 92
Mother
Total sample 1004 685 319
Treatment 345 220 125
Control 659 465 194
Father
Total sample 877 591 286
Treatment 180 117 63
Control 697 474 223
Sister
Total sample 608 418 190
Treatment 480 342 138
Control 128 76 52
Brother
Total sample 610 426 184
Treatment 386 258 128
Control 224 168 56

Table 4 Sample sizes in each of the sub-analyses in the younger generation’s younger adults (19–39
years)
All Women Men
Daughter
Total sample 209 157 52
Treatment 51 36 15
Control 158 121 37
Son
Total sample 217 160 57
Treatment 35 26 9
Control 182 134 48
Mother
Total sample 452 325 127
Treatment 179 133 46
Control 273 192 81
Father
Total sample 429 306 123
Treatment 91 64 27
Control 338 242 96
Sister
Total sample 320 234 86
Treatment 247 188 59
Control 73 46 27
Brother
Total sample 299 219 80
Treatment 184 137 47
Control 115 82 33
Appendix 2

Table 1 Balance of independent variables before and after matching. Older generation (68–74 years) SMC with daughter

Mean/All1 Mean/Women2 Mean/Men1


Variable Treated Control %bias Treated Control %bias Treated Control %bias
age Unmatched 70.336 70.389 -3.100 70.281 70.474 -11.100 70.429 70.263 9.9
Matched 70.318 70.343 -1.400 70.278 70.371 -5.400 70.399 70.398 .1

gender U 0.628 .6 5.7 - - - - - -


M 0.635 0.636 -0.200 - - - - - -

edu_dum1 U 0.245 0.263 -4.200 0.259 0.333 -16.200 0.221 0.158 16.000
M 0.244 0.232 2.6 0.260 0.250 2.3 0.220 0.197 5.8

edu_dum2 U 0.505 0.474 6.3 0.496 0.439 11.500 0.521 0.526 -1.100
M 0.510 0.526 -3.300 0.510 0.524 -2.800 0.552 0.611 -11.900

edu_dum3 U 0.102 0.084 6.2 0.109 0.123 -4.400 0.092 0.026 27.900
M 0.100 0.095 1.7 0.106 0.098 2.4 0.063 0.062 .2

edu_dum4 U 0.147 0.179 -8.500 0.136 0.105 9.30 0.167 0.289 -29.300
M 0.146 0.146 0 0.124 0.128 -1.200 0.166 0.130 8.70

med_issue U 0.619 0.663 -9.300 0.630 0.614 3.2 .6 0.737 -29.200


M 0.619 0.617 .4 0.631 0.575 11.500 0.601 0.576 5.2

income U 1.833 1.674 19.000 1.748 1.509 30.600 1.975 1.921 6.1
M 1.819 1.744 9 1.737 1.656 10.400 1.951 2.021 -8.000

marital_dum1 U 0.011 0.032 -14.400 0.012 0.053 -22.700 0.008 0 12.900


M 0.010 0.007 1.8 0.003 0.002 .1 0 0 0

marital_dum2 U 0.766 0.674 20.600 0.714 0.544 35.500 0.854 0.868 -4.100
M 0.764 0.761 .8 0.719 0.721 -0.300 0.857 0.878 -6.300

marital_dum3 U 0.135 0.158 -6.500 0.165 0.228 -15.700 0.083 0.053 12.100
M 0.139 0.168 -8.300 0.173 0.204 -7.900 0.085 0.080 2.1

marital_dum4 U 0.088 0.137 -15.300 0.109 0.175 -19.100 0.054 0.079 -9.900
M 0.088 0.064 7.3 0.106 0.073 9.4 0.058 0.042 6.6
no. of children U 2.693 2.453 17.300 2.603 2.421 12.900 2.846 2.5 25.400
M 2.631 2.567 4.6 2.559 2.540 1.4 2.731 2.664 4.9

no. of friends U 5.595 4.632 25.100 5.116 4.754 11.900 6.404 4.447 40.600
M 5.355 5.084 7.1 4.951 5.113 -5.300 6.206 5.417 16.400

income squared U - - - 3.699 2.842 27.000 - - -


M - - - 3.650 3.276 11.800 - - -

med_issue x friends U - - - - - - 1.779 1.790 -0.600


M - - - - - - 1.659 1.504 8.800

no. of friends squared U - - - - - - 65.696 40.868 25.800


M - - - - - - 61.964 49.216 13.300

Mean bias U 11.5 16.5 16.7


M 3.5 5.2 6.5
1
Kernel matching and conditioning on the common support
2
Radius matching (caliper=0.01) with replacement and conditioning on the common support

Table 2 Balance of independent variables before and after matching. Older generation (68–74 years) SMC with son

Mean/All1 Mean/Women1 Mean/Men1


Variable Treated Control %bias Treated Control %bias Treated Control %bias
age Unmatched 70.285 70.579 -17.800 70.245 70.587 -20.900 70.355 70.563 -12.300
Matched 70.304 70.369 -3.900 70.253 70.387 -8.100 70.356 70.440 -5.000

gender U 0.640 0.657 -3.600 - - - - - -


M 0.642 0.645 -0.600 - - - - - -

edu_dum1 U 0.229 .3 -16.100 0.242 0.359 -25.500 0.206 0.188 4.5


M 0.228 0.214 3.2 0.245 0.249 -0.800 0.204 0.158 11.600

edu_dum2 U 0.508 0.457 10.300 0.513 0.402 22.300 .5 0.562 -12.500


M 0.515 0.523 -1.800 0.518 0.497 4.3 0.534 0.586 -10.300

edu_dum3 U 0.109 0.079 10.600 0.116 0.087 9.5 0.098 0.062 13.100
M 0.107 0.101 2.2 0.109 0.111 -0.700 0.073 0.070 1.1

edu_dum4 U 0.153 0.164 -3.000 0.129 0.152 -6.700 0.196 0.188 2.2
M 0.150 0.162 -3.200 0.128 0.143 -4.400 0.188 0.186 .6

med_issue U 0.625 0.579 9.4 0.608 0.543 13.000 0.654 0.646 1.7
M 0.618 0.623 -1.000 0.599 0.609 -2.000 0.628 0.642 -2.900

income U 1.843 1.757 10.100 1.750 1.598 18.700 2.009 2.062 -6.000
M 1.836 1.846 -1.100 1.725 1.723 .2 2.026 1.984 4.7

marital_dum1 U 0.008 0.014 -5.500 0.013 0.022 -6.500 - - .


M 0.009 0.008 .2 0.014 0.011 2.1 - - .

marital_dum2 U 0.774 0.757 4.1 0.721 0.663 12.500 0.869 0.938 -23.200
M 0.772 0.800 6.600 0.717 0.753 -7.900 0.906 0.889 5.8

marital_dum3 U 0.120 .15 -8.900 0.147 0.207 -15.500 0.070 0.042 12.300
M 0.123 0.112 3.2 0.153 0.144 2.3 0.063 0.076 -5.700

marital_dum4 U 0.098 0.079 6.7 0.118 0.109 3.1 0.061 0.021 20.200
M 0.097 0.080 5.9 0.117 0.092 7.9 0.031 0.035 -2.100

no. of children U 2.783 2.350 32.000 2.655 2.293 28.200 3.009 2.458 37.700
M 2.627 2.533 7 2.501 2.433 5.4 2.675 2.605 4.8

no. of friends U 5.562 5.050 13.600 5.234 4.793 14.900 6.145 5.542 12.300
M 5.446 5.340 2.8 5.161 5.057 3.5 5.811 5.550 5.3

Mean bias U 10.8 15.2 13.2


M 3 3.8 5
1
Kernel matching and conditioning on the common support

Table 3 Balance of independent variables before and after matching. Older generation (68–74 years) SMC with grandchild

Mean/All1 Mean/Women2 Mean/Men1


Variable Treated Control %bias Treated Control %bias Treated Control %bias
age Unmatched 70.358 70.436 -4.800 70.287 70.380 -5.700 70.535 70.500 2.1
Matched 70.364 70.440 -4.600 70.317 70.423 -6.500 70.513 70.500 .8

grchild_age_dum1 U 0.113 0.328 -53.800 0.098 0.321 -56.800 0.148 0.336 -44.700
M 0.112 0.121 -2.100 0.098 0.103 -1.300 0.151 0.156 -1.100

grchild_age_dum2 U 0.327 0.336 -2.000 0.305 0.401 -20.300 0.381 0.262 25.500
M 0.331 0.346 -3.200 0.311 0.304 1.6 0.388 0.374 3.1

grchild_age_dum3 U 0.338 0.201 31.200 0.333 0.153 42.800 0.348 0.254 20.600
M 0.336 0.312 5.6 0.335 0.357 -5.200 0.355 0.354 .4

grchild_age_dum4 U 0.161 0.097 19.200 0.186 0.058 39.700 0.097 0.139 -13.200
M 0.157 0.148 2.6 0.177 0.162 4.6 0.099 0.108 -2.800

grchild_age_dum5 U 0.054 0.035 9.1 0.065 0.058 2.6 0.026 0.008 13.600
M 0.054 0.062 -3.900 0.066 0.062 1.8 0.007 0.009 -2.100

grchild_age_dum6 U 0.009 0.004 6.6 0.013 0.007 5.6 0 0 .


M 0.009 0.011 -2.300 0.013 0.012 .7 0 0 .

gender U 0.714 0.529 38.800 - - - - - -


M 0.712 0.680 6.7 - - - - - -

edu_dum1 U 0.262 0.216 10.700 0.282 0.263 4.2 0.213 0.164 12.500
M 0.265 0.294 -6.700 0.288 0.335 -10.600 0.211 0.236 -6.600

edu_dum2 U 0.496 0.502 -1.100 0.499 0.438 12.200 0.490 0.574 -16.700
M 0.503 0.468 6.9 0.501 0.481 4 0.487 0.481 1.2

edu_dum3 U 0.114 0.070 15.600 0.124 0.073 17.200 0.090 0.066 9.2
M 0.105 0.110 -2.000 0.113 0.108 1.7 0.092 0.059 12.500

edu_dum4 U 0.127 0.212 -22.800 0.096 0.226 -36.000 0.206 0.197 2.4
M 0.127 0.127 0.000 0.098 0.076 6.1 0.211 0.225 -3.500

med_issue U 0.625 0.595 6.3 0.618 0.555 12.800 0.645 0.639 1.2
M 0.626 0.615 2.3 0.615 0.588 5.5 0.638 0.646 -1.500

income U 1.801 1.873 -8.800 1.695 1.759 -8.000 2.064 2.0 7.9
M 1.802 1.816 -1.700 1.699 1.718 -2.400 2.072 2.088 -2.000

marital_dum1 U 0.011 0.004 8.4 0.015 0.007 7.7 0 0 .


M 0.007 0.008 -1.100 0.011 0.005 5.7 0 0 .

marital_dum2 U 0.766 0.764 .3 0.711 0.686 5.3 0.903 0.852 15.500


M 0.768 0.780 -2.800 0.718 0.690 6.0 0.901 0.910 -2.500
marital_dum3 U 0.131 0.139 -2.300 0.158 0.182 -6.600 0.065 0.090 -9.600
M 0.133 0.118 4.4 0.156 0.189 -8.800 0.066 0.052 5

marital_dum4 U 0.092 0.093 -0.100 0.116 0.124 -2.400 0.032 0.057 -12.100
M 0.092 0.094 -0.800 0.116 0.117 -0.200 0.033 0.038 -2.500

no. of children U 2.679 2.552 8.6 2.584 2.372 15.300 2.916 2.754 10.300
M 2.669 2.802 -9.000 2.557 2.547 .7 2.882 2.839 2.7

no. of friends U 5.483 5.467 .4 5.238 4.927 9.9 6.097 6.074 .5


M 5.492 5.607 -2.900 5.224 5.619 -12.600 5.921 6.306 -7.900

age x no of friends U - - - 368.300 346.940 9.6 - - -


M - - - 367.540 396.340 -12.900 - - -

Mean bias U 12.5 16 12.8


M 3.6 4.9 3.4
1
Kernel matching and conditioning on the common support
2
Radius matching (caliper=0.01) with replacement and conditioning on the common support

Table 4 Balance of independent variables before and after matching. Older generation (68–74 years) SMC with sister

Mean/All1 Mean/Women1 Mean/Men2


Variable Treated Control %bias Treated Control %bias Treated Control %bias
age Unmatched 70.225 70.388 -9.800 70.149 70.371 -13.300 70.396 70.411 -0.900
Matched 70.240 70.266 -1.600 70.147 70.189 -2.500 70.426 70.400 1.6

gender U 0.692 0.580 23.200 - - - - - -


M 0.689 0.684 1.2 - - - - - -

edu_dum1 U 0.246 0.264 -4.200 0.251 0.292 -9.300 0.236 0.226 2.4
M 0.247 0.253 -1.400 0.252 0.270 -3.900 0.240 0.200 9.6

edu_dum2 U 0.497 0.506 -1.800 0.523 0.470 10.600 0.438 0.555 -23.500
M 0.509 0.498 2.2 0.527 0.508 3.9 0.481 0.478 .6

edu_dum3 U 0.105 0.086 6.4 0.093 0.109 -5.300 0.132 0.055 26.700
M 0.101 0.102 -0.100 0.093 0.093 -0.100 0.093 0.138 -15.400

edu_dum4 U 0.152 0.144 2.4 0.133 0.129 1.3 0.194 0.164 7.8
M 0.143 0.148 -1.300 0.128 0.130 -0.600 0.186 0.185 .3
med_issue U 0.585 0.629 -9.200 0.582 0.619 -7.500 0.590 0.644 -11.000
M 0.584 0.590 -1.400 0.581 0.587 -1.100 0.589 0.588 .2

income U 1.833 1.793 5 1.762 1.673 11.600 1.993 1.959 4.2


M 1.828 1.811 2.2 1.744 1.741 .4 1.961 2.091 -16.000

marital_dum1 U 0.051 0.069 -7.400 0.062 0.079 -6.700 0.028 0.055 -13.600
M 0.053 0.052 .2 0.064 0.067 -1.300 0.031 0.029 1

marital_dum2 U 0.730 0.690 8.9 0.669 0.624 9.4 0.868 0.781 23.000
M 0.725 0.737 -2.700 0.665 0.669 -0.900 0.868 0.874 -1.600

marital_dum3 U 0.107 0.144 -11.100 0.139 0.173 -9.300 0.035 0.103 -27.000
M 0.110 0.105 1.6 0.144 0.144 -0.100 0.031 0.039 -3.100

marital_dum4 U 0.111 0.098 4.5 0.130 0.124 1.9 0.069 0.062 3.1
M 0.112 0.106 2.2 0.128 0.120 2.3 0.070 0.058 4.7

no. of children U 2.501 2.135 23.200 2.415 1.990 28.300 2.694 2.336 20.900
M 2.319 2.298 1.3 2.230 2.214 1.1 2.419 2.390 1.6

no. of friends U 5.527 5.330 5 5.056 5.173 -3.700 6.583 5.548 21.000
M 5.412 5.388 .6 5.026 4.933 2.9 6.326 5.757 11.600

age x no. of children U - - - - - - 189.240 164.240 20.900


M - - - - - - 169.970 168.090 1.6

Mean bias U 8.70 9.1 14.7


M 1.4 1.6 4.9
1
Kernel matching and conditioning on the common support
2
Radius matching (caliper=0.01) with replacement and conditioning on the common support

Table 5 Balance of independent variables before and after matching. Older generation (68–74 years) SMC with brother

Mean/All2 Mean/Women1 Mean/Men1


Variable Treated Control %bias Treated Control %bias Treated Control %bias
age Unmatched 70.233 70.325 -5.600 70.066 70.351 -17.200 70.508 70.272 14.400
Matched 70.228 70.184 2.7 70.072 70.123 -3.100 70.496 70.465 1.9

gender U 0.622 0.675 -11.000 - - - - - -


M 0.628 0.639 -2.500 - - - - - -

edu_dum1 U 0.263 0.253 2.2 0.265 0.271 -1.300 0.258 0.215 10.000
M 0.255 0.255 0 0.263 0.261 .4 0.256 0.229 6.2

edu_dum2 U 0.496 0.506 -2.100 0.507 0.512 -1.000 0.477 0.494 -3.400
M 0.501 0.510 -1.700 0.507 0.505 .5 0.488 0.505 -3.300

edu_dum3 U 0.100 0.093 2.6 0.095 0.098 -0.900 0.109 0.082 9.20
M 0.102 0.094 2.7 0.096 0.103 -2.600 0.104 0.116 -3.900

edu_dum4 U 0.142 0.148 -1.900 0.133 0.119 4.2 0.156 0.209 -13.600
M 0.141 0.141 .1 0.134 0.131 1 0.152 0.150 .5

med_issue U 0.608 0.644 -7.500 0.583 0.631 -9.900 0.648 0.671 -4.700
M 0.607 0.603 .8 0.589 0.593 -1.000 0.64 0.645 -1.000

income U 1.796 1.790 .8 1.711 1.701 1.2 1.938 1.975 -4.600


M 1.799 1.798 .2 1.708 1.706 .3 1.936 1.954 -2.200

marital_dum1 U 0.032 0.068 -16.300 0.043 0.073 -13.100 0.016 0.057 -22.200
M 0.033 0.035 -1.100 0.043 0.031 5.1 0.016 0.019 -1.500

marital_dum2 U 0.749 0.683 14.700 0.668 0.643 5.2 0.883 0.766 31.000
M 0.748 0.749 -0.300 0.670 0.687 -3.500 0.88 0.866 3.6

marital_dum3 U 0.112 0.152 -11.900 0.161 0.168 -1.800 0.031 0.120 -34.000
M 0.114 0.110 1.2 0.163 0.157 1.5 0.032 0.035 -1.300

marital_dum4 U 0.106 0.097 3.1 0.128 0.116 3.7 0.070 0.057 5.5
M 0.105 0.105 0 0.124 0.125 -0.200 0.072 0.079 -3.000

no. of children U 2.457 2.158 19.000 2.303 2.116 12.300 2.711 2.247 28.300
M 2.372 2.461 -5.600 2.258 2.288 -1.900 2.552 2.570 -1.100

no. of friends U 6.029 5.105 23.700 5.578 4.918 20.400 6.773 5.494 26.400
M 5.802 5.628 4.4 5.479 5.268 6.5 6.456 6.514 -1.200

Mean bias U 8.70 7.1 15.9


M 1.7 2.1 2.4
1
Kernel matching and conditioning on the common support
2
Radius matching (caliper=0.01) with replacement and conditioning on the common support

Table 6 Balance of independent variables before and after matching. Younger generation (19–56 years) SMC with daughter

Mean/All1 Mean/Women1 Mean/Men1


Variable Treated Control %bias Treated Control %bias Treated Control %bias
age_dum1 Unmatched 0 0 . 0 0 . 0 0 .
Matched 0 0 . 0 0 . 0 0 .

age_dum2 U 0 0 . 0 0 . 0 0 .
M 0 0 . 0 0 . 0 0 .

age_dum3 U 0.017 0.124 -42.400 0.020 0.133 -43.500 0.011 .1 -39.700


M 0.019 0.015 1.7 0.022 0.018 1.5 0.016 0.018 -0.700

age_dum4 U 0.112 0.263 -39.200 0.111 0.281 -43.600 0.116 0.218 -27.600
M 0.112 0.121 -2.300 0.108 0.102 1.4 0.115 0.162 -12.800

age_dum5 U 0.314 0.348 -7.200 0.298 0.353 -11.700 0.358 0.336 4.5
M 0.349 0.373 -5.000 0.323 0.362 -8.200 0.361 0.333 5.9

age5_dum6 U 0.375 0.224 33.200 0.389 0.201 41.900 0.337 0.282 11.900
M 0.391 0.375 3.5 0.405 0.359 10.300 0.377 0.350 5.9

age5_dum7 U 0.182 0.041 45.700 0.183 0.032 49.900 0.179 0.064 35.700
M 0.128 0.116 4.1 0.142 0.159 -5.600 0.131 0.138 -2.000

daughter_age_dum1 U 0.006 0.253 -79.000 0.004 0.255 -80.600 0.011 0.245 -74.800
M 0.006 0.006 .1 0.004 0.004 .1 0.016 0.015 .6

daughter_age_dum2 U 0.101 0.371 -67.100 0.095 0.370 -68.800 0.116 0.373 -62.400
M 0.112 0.120 -1.900 0.103 0.108 -1.200 0.180 0.204 -5.700

daughter_age_dum3 U 0.349 0.193 35.500 0.321 0.176 34.000 0.421 0.236 39.900
M 0.381 0.386 -1.000 0.341 0.344 -0.800 0.508 0.460 10.400

daughter_age_dum4 U 0.277 0.139 34.300 0.286 0.151 33.000 0.253 0.109 37.800
M 0.292 0.300 -2.100 0.302 0.300 .4 0.230 0.229 .2

daughter_age_dum5 U 0.173 0.028 49.400 0.187 0.032 50.800 0.137 0.018 45.300
M 0.122 0.121 .2 0.155 0.178 -7.500 0.016 0.043 -10.300
daughter_age_dum6 U 0.095 0.015 35.300 0.107 0.014 39.500 0.063 0.018 22.800
M 0.087 0.067 8.5 0.095 0.066 12.500 0.049 0.049 0

gender U 0.726 0.716 2.2 - - - - - -


M 0.724 0.728 -0.800 - - - - -

edu_dum1 U 0.017 0.018 -0.600 0.016 0.011 4.4 0.021 0.036 -9.100
M 0.019 0.034 -11.200 0.017 0.024 -5.600 0.033 0.026 3.9

edu_dum2 U 0.427 0.265 34.300 0.444 0.230 46.400 0.379 0.355 5


M 0.410 0.419 -1.900 0.435 0.452 -3.700 0.361 0.391 -6.400

edu_dum3 U 0.245 0.345 -22.100 0.230 0.367 -30.200 0.284 0.291 -1.500
M 0.247 0.232 3.2 0.233 0.228 1.1 0.262 0.275 -2.700

edu_dum4 U 0.311 0.371 -12.600 0.310 0.392 -17.300 0.316 0.318 -0.500
M 0.324 0.315 1.9 0.315 0.296 3.8 0.344 0.307 7.9

med_issue U 0.435 0.353 16.800 0.472 0.381 18.400 0.337 0.282 11.900
M 0.439 0.415 4.9 0.470 0.431 7.8 0.328 0.321 1.4

income U 2.046 2.051 -0.700 1.972 1.971 .1 2.242 2.255 -1.700


M 2.042 2.044 -0.400 1.961 1.925 5.1 2.147 2.138 1.3

marital_dum1 U 0.043 0.031 6.5 0.052 0.043 4 0.021 0 20.600


M 0.045 0.039 2.9 0.052 0.042 4.4 0 0 0

marital_dum2 U 0.818 0.902 -24.300 0.790 0.885 -26.000 0.895 0.945 -18.700
M 0.827 0.821 1.8 0.797 0.774 6.3 0.918 0.941 -8.400

marital_dum3 U 0.138 0.067 23.600 0.159 0.072 27.400 0.084 0.055 11.600
M 0.128 0.140 -3.800 0.151 0.183 -10.300 0.082 0.059 9

marital_dum4 U 0 0 . 0 0 . 0 0 .
M 0 0 . 0 0 . 0 0 .

no. of children U 2.994 2.291 53.600 2.972 2.295 52.800 3.053 2.282 55.600
M 2.724 2.775 -3.900 2.759 2.759 0 2.508 2.448 4.4

no. of friends U 3.951 4.165 -7.400 3.956 4.050 -3.700 3.937 4.455 -14.100
M 3.962 3.955 .2 3.918 3.799 4.7 4.213 4.009 5.6

income x no of friends U - - - 8.024 7.986 .6 - - -


M - - - 7.879 7.394 7.8 - - -

income squared U - - - 4.393 4.417 -0.800 - - -


M - - - 4.358 4.243 3.7 - - -

Mean bias U 29.3 30.4 25.1


M 2.9 4.7 4.8
1
Kernel matching and conditioning on the common support

Table 7 Balance of independent variables before and after matching. Younger generation (19–56 years) SMC with son

Mean/All1 Mean/Women1 Mean/Men1


Variable Treated Control %bias Treated Control %bias Treated Control %bias
age_dum1 Unmatched 0 0 . 0 0 . 0 0 -
Matched 0 0 . 0 0 . 0 0 -

age_dum2 U 0 0 . 0 0 . 0 0 -
M 0 0 . 0 0 . 0 0 -

age_dum3 U 0.010 0.103 -40.800 0.014 0.103 -38.300 0 0.102 -47.500


M 0.010 0.008 1.2 0.015 0.015 -0.100 0 0.002 -1.000

age_dum4 U 0.099 0.271 -45.600 0.100 0.284 -47.800 0.094 0.241 -39.900
M 0.101 0.077 6.3 0.091 0.069 5.8 0.113 0.087 6.9

age_dum5 U 0.320 0.353 -6.900 0.287 0.344 -12.300 .4 0.372 5.7


M 0.326 0.363 -7.700 0.303 0.352 -10.600 0.423 0.462 -8.100

age5_dum6 U 0.381 0.235 32.000 0.397 0.230 36.600 0.341 0.248 20.400
M 0.378 0.383 -1.100 0.404 0.405 -0.100 0.394 0.363 6.9

age5_dum7 U 0.190 0.038 49.100 0.201 0.039 51.200 0.165 0.036 43.400
M 0.184 0.169 4.7 0.187 0.159 8.80 0.070 0.086 -5.200

son_age_dum1 U 0 0.256 -83.000 0 0.230 -77.100 0 0.321 -96.900


M 0 0 0 0 0 0 0 0.007 -2.000

son_age_dum2 U 0.116 0.342 -55.900 0.105 0.338 -58.400 0.141 0.350 -49.900
M 0.118 0.110 1.9 0.111 0.077 8.5 0.169 0.149 4.7

son_age_dum3 U 0.296 0.233 14.300 0.278 0.242 8.20 0.341 0.212 29.100
M 0.302 0.326 -5.500 0.283 0.347 -14.500 0.394 0.454 -13.400

son_age_dum4 U 0.293 0.103 49.100 0.287 0.115 43.900 0.306 0.073 61.900
M 0.288 0.303 -3.900 0.283 0.289 -1.500 0.268 0.220 12.800

son_age_dum5 U 0.177 0.043 43.900 0.182 0.045 44.000 0.165 0.036 43.400
M 0.174 0.176 -0.700 0.177 0.141 11.400 0.141 0.142 -0.300

son_age_dum6 U 0.119 0.024 37.700 0.148 0.030 42.200 0.047 0.007 24.500
M 0.118 0.084 13.300 0.146 0.146 .1 0.028 0.029 -0.500

gender U 0.711 0.707 .8 - - - - - -


M 0.712 0.728 -3.500 - - - - - -

edu_dum1 U 0.010 0.017 -5.900 0.005 0.012 -8.000 0.024 0.029 -3.500
M 0.010 0.010 .2 0.005 0.007 -2.500 0.028 0.032 -2.600

edu_dum2 U 0.452 0.263 40.300 0.459 0.242 46.800 0.435 0.314 25.200
M 0.455 0.456 -0.300 0.449 0.461 -2.500 0.437 0.395 8.5

edu_dum3 U 0.248 0.340 -20.100 0.234 0.353 -26.300 0.282 0.307 -5.300
M 0.25 0.244 1.3 0.242 0.264 -4.800 0.254 0.257 -0.800

edu_dum4 U 0.289 0.380 -19.400 0.301 0.393 -19.200 0.259 0.350 -19.900
M 0.285 0.289 -0.900 0.303 0.267 7.5 0.282 0.315 -7.300

med_issue U 0.432 0.355 15.800 0.469 0.384 17.300 0.341 0.285 12.100
M 0.431 0.451 -4.100 0.475 0.485 -2.000 0.338 0.354 -3.400

income U 2 2.051 -7.200 1.890 2.012 -17.200 2.271 2.146 18.200


M 1.996 2.051 -7.600 1.884 1.904 -2.900 2.253 2.269 -2.300

marital_dum1 U 0.027 0.028 -0.300 0.029 0.039 -5.800 0.024 0 21.800


M 0.028 0.039 -6.900 0.030 0.036 -2.900 0.028 0 26.100

marital_dum2 U 0.820 0.891 -20.300 0.809 0.873 -17.700 0.847 0.934 -28.100
M 0.823 0.811 3.3 0.803 0.811 -2.200 0.831 0.804 8.800
marital_dum3 U 0.146 0.077 22.100 0.153 0.082 22.300 0 0 -
M 0.142 0.140 .7 0.157 0.140 5.3 0 0 -

marital_dum4 U 0.007 0.004 3.4 0.010 0.006 4 0 0 -


M 0.007 0.009 -3.300 0.010 0.014 -4.300 0 0 -

no. of children U 3.068 2.259 60.400 3.077 2.266 62.100 3.047 2.241 56.500
M 2.951 2.894 4.3 2.843 2.861 -1.300 2.859 2.743 8.1

no. of friends U 4.163 4.019 4.8 4.019 3.933 3.1 4.518 4.226 8.300
M 4.049 4.284 -7.900 3.884 3.827 2.1 4.535 4.515 .6

no. of children x no. of - - - 12.517 8.695 39.200 14.176 9.321 31.200


friends U
M - - - 10.773 10.725 .5 14.014 13.571 2.8

income squared U - - - - - - 5.565 5.124 14.200


M - - - - - - 5.493 5.551 -1.900

no. of friends x incomeU - - - - - - 6.812 4.869 56.400


M - - - - - - 6.451 6.293 4.6

Mean bias U 28.3 31.2 31.8


M 3.8 4.3 5.8
1
Kernel matching and conditioning on the common support

Table 8 Balance of independent variables before and after matching. Younger generation (19–56 years) SMC with mother

Mean/All1 Mean/Women2 Mean/Men1


Variable Treated Control %bias Treated Control %bias Treated Control %bias
age_dum1 Unmatched 0.004 0.001 5.6 0.006 0.002 7 0 0 .
Matched 0.004 0.005 -1.800 0.003 0 4.8 0 0 .

age_dum2 U 0.021 0.013 6.2 0.029 0.016 9 0.006 0.008 -2.000


M 0.021 0.020 1.3 0.029 0.026 2.4 0.006 0.007 -1.200

age_dum3 U 0.103 0.092 3.8 0.113 0.089 7.9 0.083 0.098 -5.300
M 0.104 0.106 -0.800 0.111 0.112 -0.400 0.087 0.086 .2

age_dum4 U 0.214 0.188 6.5 0.232 0.189 10.400 0.179 0.186 -1.800
M 0.215 0.216 -0.400 0.234 0.235 -0.400 0.174 0.176 -0.700
age_dum5 U 0.318 0.323 -1.200 0.299 0.330 -6.800 0.357 0.307 10.700
M 0.318 0.316 .5 0.298 0.297 .2 0.366 0.361 1.1

age5_dum6 U 0.265 0.280 -3.400 0.232 0.283 -11.800 0.333 0.273 13.200
M 0.266 0.257 1.9 0.234 0.231 .8 0.323 0.313 2.1

age5_dum7 U 0.074 0.102 -9.800 0.090 0.091 -0.300 0.042 0.129 -31.500
M 0.072 0.080 -2.600 0.091 0.098 -2.700 0.043 0.055 -4.300

gender U 0.673 0.708 -7.600 - - - - - -


M 0.672 0.679 -1.600 - - - - - -

edu_dum1 U 0.008 0.021 -11.100 0.009 0.011 -2.300 0.006 0.045 -25.100
M 0.006 0.008 -1.600 0.009 0.009 -0.600 0.006 0.006 .1

edu_dum2 U 0.322 0.358 -7.600 0.319 0.335 -3.400 0.327 0.413 -17.700
M 0.322 0.328 -1.200 0.319 0.321 -0.500 0.335 0.339 -0.800

edu_dum3 U 0.279 0.299 -4.500 0.290 0.315 -5.400 0.256 0.261 -1.200
M 0.279 0.274 1.1 0.289 0.277 2.8 0.261 0.243 4

edu_dum4 U 0.392 0.322 14.500 0.383 0.340 9 0.411 0.280 27.600


M 0.393 0.390 .6 0.383 0.393 -2.000 0.398 0.412 -3.000

med_issue U 0.390 0.399 -1.800 0.443 0.421 4.5 0.280 0.345 -14.000
M 0.389 0.393 -0.800 0.442 0.438 .7 0.286 0.282 .7

income U 2.053 1.989 8.5 1.936 1.944 -1.000 2.292 2.099 25.200
M 2.053 2.039 1.9 1.936 1.934 .3 2.286 2.252 4.3

marital_dum1 U 0.129 0.146 -5.100 0.139 0.131 2.2 0.107 0.182 -21.300
M 0.127 0.128 -0.200 0.135 0.133 .5 0.112 0.115 -0.900

marital_dum2 U 0.776 0.759 4.1 0.754 0.765 -2.700 0.821 0.742 19.200
M 0.777 0.774 .7 0.757 0.757 .1 0.826 0.810 3.8

marital_dum3 U 0.092 0.092 -0.100 0.101 0.099 1 0.071 0.076 -1.700


M 0.092 0.094 -0.900 0.102 0.104 -0.500 0.062 0.075 -4.900

marital_dum4 U 0.004 0.003 1 0.006 0.005 1.5 0 0 .


M 0.004 0.004 .2 0.006 0.006 -0.700 0 0 .

no. of children U 1.918 1.802 7.7 1.881 1.901 -1.300 1.994 1.561 28.300
M 1.922 1.887 2.3 1.895 1.905 -0.700 1.870 1.866 .2

no. of friends U 4.717 4.098 19.000 4.661 4.024 21.400 4.833 4.277 14.500
M 4.725 4.675 1.5 4.564 4.583 -0.600 4.522 4.626 -2.700

Mean bias U 6.5 5.7 15.3


M 1.2 1.1 2.1
1
Kernel matching and conditioning on the common support
2
Radius matching (caliper=0.01) with replacement and conditioning on the common support

Table 9 Balance of independent variables before and after matching. Younger generation (19–56 years) SMC with father

Mean/All2 Mean/Women2 Mean/Men1


Variable Treated Control %bias Treated Control %bias Treated Control %bias
age_dum1 Unmatched 0 0.004 -8.900 0 0.006 -10.700 0 0 .
Matched 0 0 0 0 0 0 0 0 .

age_dum2 U 0.030 0.017 8.5 0.039 0.022 10.200 0.011 0.006 5


M 0.026 0.028 -1.000 0.034 0.027 4 0.011 0.011 .5

age_dum3 U 0.104 0.102 .5 0.117 0.103 4.4 0.078 0.100 -7.900


M 0.106 0.107 -0.300 0.119 0.124 -1.600 0.080 0.090 -3.700

age_dum4 U 0.204 0.205 -0.100 0.201 0.210 -2.100 0.211 0.194 4.2
M 0.208 0.209 -0.300 0.205 0.220 -3.900 0.216 0.201 3.7

age_dum5 U 0.372 0.321 10.600 0.369 0.312 12.000 0.378 0.343 7.2
M 0.377 0.376 .3 0.375 0.376 -0.200 0.386 0.389 -0.600

age5_dum6 U 0.219 0.281 -14.200 0.196 0.282 -20.300 0.267 0.278 -2.600
M 0.211 0.214 -0.700 0.188 0.173 3.5 0.25 0.250 .1

age5_dum7 U 0.071 0.070 .4 0.078 0.066 4.7 0.056 0.078 -8.800


M 0.072 0.067 1.9 0.080 0.080 0 0.057 0.059 -1.000

gender U 0.665 0.693 -5.800 - - - - - -


M 0.660 0.672 -2.500 - - - - - -
edu_dum1 U 0.004 0.016 -12.400 0 0.009 -13.200 0.011 0.032 -14.600
M 0.004 0.002 1.9 0 0 0 0.011 0.011 .4

edu_dum2 U 0.264 0.355 -19.800 0.268 0.332 -13.900 0.256 0.408 -32.700
M 0.260 0.254 1.4 0.261 0.260 .3 0.25 0.276 -5.500

edu_dum3 U 0.312 0.287 5.6 0.290 0.302 -2.500 0.356 0.252 22.500
M 0.317 0.318 -0.200 0.295 0.303 -1.700 0.352 0.347 1.2

edu_dum4 U 0.420 0.342 16.000 0.441 0.358 17.100 0.378 0.307 14.800
M 0.419 0.426 -1.500 0.443 0.437 1.3 0.386 0.367 4.1

med_issue U 0.353 0.390 -7.600 0.402 0.412 -2.100 0.256 0.340 -18.400
M 0.355 0.357 -0.500 0.403 0.392 2.4 0.25 0.276 -5.600

income U 2.171 1.988 24.600 2.062 1.940 17.100 2.389 2.097 37.500
M 2.166 2.161 .7 2.057 2.065 -1.100 2.398 2.375 2.9

marital_dum1 U 0.108 0.155 -14.000 0.134 0.144 -2.800 0.056 0.181 -39.600
M 0.109 0.108 .3 0.136 0.132 1.3 0.057 0.068 -3.400

marital_dum2 U 0.803 0.755 11.500 0.765 0.759 1.6 0.878 0.748 33.700
M 0.808 0.806 .3 0.773 0.778 -1.300 0.875 0.862 3.3

marital_dum3 U 0.082 0.088 -2.100 0.089 0.095 -1.900 0.067 0.071 -1.800
M 0.083 0.085 -0.900 0.091 0.090 .4 0.068 0.070 -0.700

marital_dum4 U 0.007 0.002 7.9 0.011 0.003 9.9 0 0 .


M 0 0 0 0 0 0 0 0 .

no. of children U 1.963 1.775 12.000 1.877 1.861 1.1 2.133 1.583 34.200
M 1.921 1.930 -0.600 1.881 1.867 .9 2.034 2.060 -1.600

no. of friends U 4.725 4.300 12.800 4.492 4.244 8.5 5.189 4.424 19.100
M 4.706 4.606 3 4.460 4.431 1 4.852 4.989 -3.400

Mean bias U 9.8 8.2 17.9


M 0.9 1.3 2.5
1
Kernel matching and conditioning on the common support
2
Radius matching (caliper=0.01) with replacement and conditioning on the common support
Table 10 Balance of independent variables before and after matching. Younger generation (19–56 years) SMC with sister

Mean/All1 Mean/Women3 Mean/Men1


Variable Treated Control %bias Treated Control %bias Treated Control %bias
age_dum1 Unmatched 0.006 0 10.600 0.008 0 12.500 0 0 .
Matched 0 0 0 0 0 0 0 0 .

age_dum2 U 0.024 0.021 2.3 0.031 0.017 9.1 0.005 0.026 -16.900
M 0.025 0.020 3.3 0.027 0.024 1.6 0.005 0.001 3.4

age_dum3 U 0.111 0.103 2.7 0.114 0.102 4 0.104 0.105 -0.400


M 0.113 0.113 .1 0.110 0.126 -4.900 0.109 0.105 1.3

age_dum4 U 0.202 0.237 -8.600 0.207 0.263 -13.100 0.188 0.197 -2.500
M 0.208 0.217 -2.300 0.215 0.219 -1.000 0.191 0.171 5.1

age_dum5 U 0.326 0.304 4.7 0.311 0.288 5.1 0.365 0.329 7.5
M 0.328 0.323 1 0.313 0.311 .3 0.372 0.382 -2.300

age5_dum6 U 0.244 0.263 -4.300 0.244 0.246 -0.500 0.245 0.289 -10.100
M 0.244 0.245 -0.300 0.249 0.239 2.5 0.251 0.271 -4.400

age5_dum7 U 0.087 0.072 5.6 0.085 0.085 .1 0.094 0.053 15.800


M 0.083 0.082 .3 0.086 0.081 1.6 0.071 0.069 .7

gender U 0.729 0.608 25.900 - - - - - -


M 0.726 0.714 2.4 - - - - - -

edu_dum1 U 0.017 0.026 -6.100 0.015 0.008 6.4 0.021 0.053 -16.900
M 0.017 0.017 .3 0.014 0.012 1.9 0.022 0.020 .7

edu_dum2 U 0.336 0.392 -11.700 0.331 0.390 -12.300 0.349 0.395 -9.400
M 0.338 0.335 .6 0.335 0.331 1 0.339 0.351 -2.600

edu_dum3 U 0.296 0.263 7.4 0.308 0.254 11.900 0.266 0.276 -2.400
M 0.295 0.284 2.3 0.299 0.288 2.4 0.273 0.263 2.3

edu_dum4 U 0.351 0.320 6.7 0.346 0.347 -0.300 0.365 0.276 18.900
M 0.350 0.363 -2.900 0.352 0.369 -3.700 0.366 0.365 .2

med_issue U 0.392 0.397 -1.000 0.420 0.424 -0.800 0.318 0.355 -7.900
M 0.386 0.387 -0.300 0.417 0.422 -0.900 0.317 0.326 -1.900

income U 1.977 1.979 -0.300 1.913 1.881 4.2 2.151 2.132 2.5
M 1.975 1.993 -2.300 1.922 1.909 1.8 2.180 2.150 3.8

marital_dum1 U 0.128 0.170 -11.700 0.133 0.119 4.5 0.115 0.25 -35.400
M 0.126 0.113 3.6 0.123 0.121 .4 0.120 0.132 -3.200

marital_dum2 U 0.793 0.716 17.700 0.781 0.737 10.300 0.823 0.684 32.400
M 0.794 0.789 1.2 0.793 0.772 5.1 0.814 0.795 4.6

marital_dum3 U 0.079 0.113 -11.700 0.085 0.144 -18.500 0.062 0.066 -1.300
M 0.080 0.098 -6.200 0.084 0.107 -7.200 0.066 0.073 -3.100

marital_dum4 U 0 0 . 0 0 . 0 0 .
M 0 0 . 0 0 . 0 0 .

no. of children U 1.972 1.727 15.700 1.971 1.898 4.6 1.974 1.460 33.600
M 1.933 1.912 1.4 1.982 1.980 .1 1.831 1.774 3.7

no. of friends U 4.402 3.938 14.200 4.422 3.780 22.700 4.349 4.184 4.2
M 4.180 4.188 -0.300 4.043 4.314 -9.600 4.224 4.207 .4

Mean bias U 8.9 7.8 12.8


M 1.6 2.6 2.6
1
Kernel matching and conditioning on the common support
3
5-nearest neighbor matching caliper (=0.01) with replacement and conditioning on the common support

Table 11 Balance of independent variables before and after matching. Younger generation (19–56 years) SMC with brother

Mean/All2 Mean/Women1 Mean/Men1


Variable Treated Control %bias Treated Control %bias Treated Control %bias
age_dum1 Unmatched 0.004 0.003 .9 0.005 0.004 1.5 0 0 .
Matched 0.004 0.003 .6 0.005 0.005 -0.500 0 0 .

age_dum2 U 0.018 0.021 -2.500 0.026 0.021 3.3 0 0 .


M 0.018 0.020 -1.400 0.026 0.025 .7 0 0 .

age_dum3 U 0.118 0.107 3.5 0.121 0.100 6.9 0.111 0.131 -6.100
M 0.118 0.123 -1.500 0.121 0.117 1.2 0.121 0.120 .2
age_dum4 U 0.184 0.217 -8.200 0.196 0.212 -3.900 0.158 0.238 -20.100
M 0.185 0.177 2 0.196 0.202 -1.600 0.172 0.171 .1

age_dum5 U 0.326 0.272 11.700 0.312 0.278 7.4 0.357 0.262 20.500
M 0.324 0.322 .6 0.312 0.312 0.000 0.338 0.312 5.6

age5_dum6 U 0.254 0.291 -8.200 0.242 0.290 -10.900 0.281 0.298 -3.700
M 0.254 0.248 1.4 0.242 0.224 4.2 0.280 0.316 -7.800

age5_dum7 U 0.097 0.089 2.7 0.098 0.095 .8 0.094 0.071 8


M 0.097 0.108 -3.800 0.098 0.114 -5.600 0.089 0.080 3.1

gender U 0.694 0.737 -9.500 - - - - - -


M 0.695 0.695 0.000 - - - - - -

edu_dum1 U 0.018 0.009 7.5 0.010 0.004 7.3 0.035 0.024 6.6
M 0.018 0.011 5.9 0.010 0.001 10.600 0.038 0.040 -0.800

edu_dum2 U 0.349 0.376 -5.700 0.330 0.365 -7.400 0.392 0.393 -0.200
M 0.349 0.354 -0.800 0.330 0.352 -4.700 0.389 0.377 2.3

edu_dum3 U 0.308 0.245 14.100 0.330 0.241 19.800 0.257 0.262 -1.000
M 0.306 0.306 0 0.330 0.313 3.8 0.248 0.248 0

edu_dum4 U 0.326 0.370 -9.300 0.330 0.390 -12.500 0.316 0.321 -1.200
M 0.326 0.329 -0.600 0.330 0.334 -0.800 0.325 0.335 -2.200

med_issue U 0.392 0.446 -11.100 0.425 0.465 -7.900 0.316 0.381 -13.600
M 0.392 0.379 2.6 0.425 0.415 2.1 0.331 0.298 7.1

income U 1.957 1.985 -3.700 1.874 1.975 -14.200 2.146 2.024 14.700
M 1.957 1.943 1.8 1.874 1.863 1.5 2.127 2.121 .7

marital_dum1 U 0.152 0.156 -1.100 0.160 0.145 4 0.135 0.179 -12.100


M 0.152 0.151 .4 0.160 0.156 1 0.146 0.130 4.5

marital_dum2 U 0.758 0.752 1.4 0.742 0.747 -1.100 0.795 0.774 5.2
M 0.758 0.763 -1.100 0.742 0.752 -2.100 0.803 0.818 -3.700

marital_dum3 U 0.086 0.086 .1 0.093 0.100 -2.300 0.070 0.048 9.6


M 0.086 0.083 1 0.093 0.089 1.3 0.051 0.052 -0.500
marital_dum4 U 0.004 0.006 -3.700 0.005 0.008 -3.800 0 0 .
M 0.004 0.003 1.1 0.005 0.003 2.4 0 0 .

no. of children U 1.966 1.817 9.6 1.928 1.925 .2 2.053 1.548 32.200
M 1.959 1.935 1.5 1.928 1.985 -3.700 1.777 1.745 2

no. of friends U 4.458 4.284 5.2 4.407 4.029 13.300 4.573 5.024 -10.300
M 4.464 4.386 2.4 4.407 4.263 5 4.605 4.260 7.9

Mean bias U 6 6.8 10.3


M 1.5 2.8 3
1
Kernel matching and conditioning on the common support
2
Radius matching (caliper=0.01) with replacement and conditioning on the common support

You might also like