The Impact of an Adult Child’s
Emigration on the Mental Health
of Older Parents
Alan Barrett and Irene Mosca
31 January 2014
Structure of the talk
• Motivation
– why are we interested in the question of whether a
child’s emigration might impact upon the mental
health of parents?
– where is this paper placed in the literature?
• The data, the method and the variables – TILDA, a
fixed effects approach and the mental health
measures
• The sample and descriptive stats
• Results
• Conclusions
Motivation (1)
• Why is the research question of interest?
• From a national (Irish) perspective
– Emigration is often characterised as a
consequence of recession that affects younger
people; but maybe there is an effect on older
people too
• From an international perspective
– A growing literature on the impact of migration on
the family members left behind (Antman, 2013)
Motivation (2)
• What are the impacts of migration on the
family members left behind?
• On children’s education
• On children’s health
• On spouse’s labour supply
• On parents’ health
• This is where we come in
Data, method and variables (1)
• TILDA Wave 1 collected between 2009 and 2011
• Extensive information collected on 8,500 people
aged 50 and over; response rate 62%
• The data covers economic, social and health
circumstances
• Intensive efforts to keep people engaged between
Wave 1 and Wave 2: birthday cards, newsletters,
certificate of participation
• TILDA Wave 2 collected in 2012
• Response rate was 90% (including an end-of-life
interview and proxy interview)
Data, method and variables (2)
• Critically for our purposes, we have measures of
mental health for the respondents and we know
where their children were living in Waves 1 and 2
• We also know a lot about other changes between
Waves 1 and 2 such as bereavement, retirement,
onset of illness
• Hence, we can explore whether mental health
changed in response to a child’s emigration
controlling for other changes over this period
Data, method and variables (3)
• By focusing on changes, we difference away
time-invariant unobservables which may be
correlated with both the child’s emigration
and parental mental health
• We follow Lindeboom et al (2003) and
Wooldridge and estimate a fixed effects model
by differencing the basic equation and
applying OLS
Data, method and variables (4)
• Measuring mental health
• Depression
– CESD is 20-item Center for Epidemiological Studies
Depression Scale. It measures the degree to which
respondents have experienced a wide variety of
depressive symptoms within the past week. Each of the 20
items is measured on a 4 point scale leading to a min score
of 0 and a max score of 60.
– Self-rated emotional/mental health on a 1 (excellent)-5
(poor) scale
• Loneliness
– UCLA Loneliness Scale. Cross-sectional score ranges
between 0 (not lonely) to 10 (extremely lonely).
The sample
• We select people who are parents of children
aged 16 and over at Wave 1; we exclude
parents with children who are younger than
16
• We only look at parents all of whose children
were living in Ireland at Wave 1
• This gives a sample of 2,912 parents
• Of this group, 361 had seen a child emigrate
between Waves 1 and 2
Descriptive Stats – Variables in
regressions
Men & Women
together
No
1+
children
children
emigrating emigrating
Outcome variables
Change in CES-D score, mean
Change in self-reported mental health
score, mean
Change in UCLA loneliness score, mean
-0.518
0.151
0.135
0.029
0.166
0.226
Demographic changes:
Widowhood
Decrease in number of close relatives/friends
Health changes:
New ADL
New IADL
Cardiovascular disorder
Chronic illness
1-point deterioration in self-reported health
2-point deterioration in self-reported health
Economic changes:
Retired
Unemployed
Change in weekly individual gross income, mean
Changes in children’s conditions:
1+ children unemployed
1+ children widowed/separated/divorced/single
1.5%
43.8%
0.3%
47.3%
4.2%
5.6%
21.9%
26.6%
19.8%
4.8%
1.4%**
1.1%***
21.7%
22.7%
18.2%
4.0%
4.6%
1.1%
1.458
6.3%
2.8%**
-15.755
11.6%
5.1%
10.1%
5.1%
Variables not in regressions
Men & Women together
Age, mean
No children
emigrating
66.3
1+ children
emigrating
60.5***
Education:
Low
Medium
41.5%
43.8%
22.7%***
52.1%***
High
14.7%
25.2%***
Depression score at wave 1
6.07
4.68***
Loneliness score at wave 1
1.60
1.47
SR mental/emotional health at wave 1
2.22
2.17
SR physical health at wave 1
2.67
2.47**
Descriptives for the children
Age, mean
High education
at w1
Non-emigrating
children
35.9
Emigrating children
30.8%
43.8%***
28.6***
Results
ΔCESD – full sample
Widowhood
Decrease in number of close relatives/friends
New ADL
New IADL
Cardiovascular disorder
Chronic illness
1-point deterioration in self-reported health
2-point deterioration in self-reported health
Retired
Unemployed
Change in income (000s)
1+ children emigrated
1+ children unemployed
1+ children widowed/separated/divorced/single
Constant
N
Men & women
Coeff.
t stat.
5.156***
(3.88)
0.731**
(2.42)
1.828**
(2.00)
0.544
(0.64)
0.945***
(2.68)
0.546
(1.58)
0.944***
(2.87)
2.024***
(2.97)
1.072***
(2.65)
0.634
(0.49)
-0.0431
(-0.46)
0.809*
(1.78)
0.305
(0.69)
0.965
(1.22)
-1.799***
(-7.67)
2912
ΔCESD – men and women
Women only
Men only
Coeff.
t stat.
Coeff.
t stat.
Widowhood
Decrease in number of close relatives/friends
5.136***
1.165***
(2.89)
(2.88)
5.001***
0.128
(3.20)
(0.33)
New ADL
2.378**
(2.01)
0.643
(0.53)
New IADL
Cardiovascular disorder
0.473
1.024**
(0.45)
(2.01)
0.880
0.838**
(0.65)
(2.05)
Chronic illness
0.808*
(1.75)
0.0946
(0.19)
1.760***
1.363
(3.80)
(1.51)
-0.0973
3.080***
(-0.23)
(2.81)
Retired
1.179*
(1.80)
0.821*
(1.71)
Unemployed
Change in income (000s)
-2.660
-0.0563
(-1.03)
(-0.41)
1.980
-0.0225
(1.50)
(-0.20)
1+ children emigrated
1.229**
(2.06)
0.292
(0.54)
0.418
(0.71)
0.0508
(0.09)
1.011
-2.380***
(0.96)
(-7.05)
0.725
-1.019***
(0.86)
(-3.27)
1-point deterioration in self-reported health
2-point deterioration in self-reported health
1+ children unemployed
1+ children widowed/separated/divorced/single
Constant
N
1707
1205
Δ self-rated mental health
Women only
Men only
Coeff.
t stat.
Coeff.
t stat.
0.365
(1.46)
0.338
(1.16)
-0.00512
(-0.10)
0.0345
(0.55)
New ADL
0.0809
(0.51)
0.0711
(0.42)
New IADL
0.0717
(0.51)
0.0177
(0.10)
Cardiovascular disorder
0.00740
(0.11)
0.0122
(0.17)
Chronic illness
0.113**
(2.05)
0.0443
(0.54)
1-point deterioration in self-reported health
0.419***
(5.98)
0.449***
(6.44)
2-point deterioration in self-reported health
0.871***
(6.54)
1.020***
(5.69)
Retired
0.0717
(0.64)
0.261**
(2.45)
Unemployed
0.203
(0.81)
0.224
(0.99)
-0.00082
(-0.03)
-0.0251
(-1.31)
0.166**
(2.20)
-0.0738
(-0.93)
1+ children unemployed
0.117
(1.44)
-0.0930
(-0.88)
1+ children widowed/separated/divorced/single
0.188*
(1.67)
-0.0592
(-0.41)
Constant
-0.0677
(-1.44)
-0.0436
(-0.87)
Widowhood
Decrease in number of close relatives/friends
Change in income (000s)
1+ children emigrated
N
1707
1205
Δ loneliness score
Widowhood
Decrease in number of close relatives/friends
Change in positive exchanges score
Change in negative exchanges score
New ADL
New IADL
Cardiovascular disorder
Chronic illness
1-point deterioration in self-reported health
2-point deterioration in self-reported health
Retired
Unemployed
Change in income (000s)
1+ children emigrated
1+ children unemployed
1+ children widowed/separated/divorced/single
Constant
N
Women only
Men only
Coeff.
t stat.
Coeff.
t stat.
0.495
(0.65)
1.712*
(1.94)
0.167
(1.47)
0.0468
(0.36)
-0.0940*** (-5.56) -0.0504***
(-3.21)
0.0595*** (4.10) 0.0517***
(3.69)
-0.480
(-0.88)
0.574
(0.94)
-0.136
(-0.40)
-0.200
(-0.33)
-0.213
(-1.40) 0.00918
(0.06)
0.244*
(1.76)
-0.243
(-1.28)
0.252*
(1.70)
-0.0500
(-0.30)
0.446*
(1.67)
-0.277
(-1.18)
-0.0707 (-0.33) -0.0696
(-0.39)
0.0838
(0.13)
0.506
(1.34)
-0.164** (-2.29)
0.0135
(0.41)
0.432***
(2.78)
-0.0320
(-0.14)
-0.203
(-1.01)
0.362
(1.49)
-0.00888 (-0.04)
0.127
(0.40)
-0.0945 (-0.89)
0.0338
(0.33)
983
749
A potential problem
• (Apparently), people with poor mental health
experience faster declines in mental health
• If this is the case, our fixed effects approach
could still be producing misleading results
• To deal with this, we (1) restrict the sample to
those with low CESD scores in Wave 1 (<16)
and (2) we interact the “child emigrate”
variable with a retrospective indicator of
mental health problems
ΔCESD – CESD <16 at W1
Women only
Men only
Coeff.
t stat.
Coeff.
t stat.
4.450***
(3.01)
4.300***
(3.17)
Decrease in number of close relatives/friends
0.419
(1.21)
0.162
(0.48)
New ADL
1.927
(1.40)
1.142
(0.96)
New IADL
1.478
(1.56)
1.330
(1.04)
Cardiovascular disorder
1.055**
(2.23)
0.590
(1.57)
Chronic illness
1.405***
(3.53)
0.216
(0.43)
1-point deterioration in self-reported health
0.866*
(1.96)
-0.0946
(-0.28)
2-point deterioration in self-reported health
1.201
(1.63)
2.602**
(2.45)
Retired
0.526
(0.83)
0.148
(0.31)
Unemployed
-0.149
(-0.07)
1.267
(0.94)
Change in income (000s)
-0.144
(-1.14)
-0.0742
(-0.90)
1+ children emigrated
0.917*
(1.66)
-0.0847
(-0.17)
1+ children unemployed
0.120
(0.20)
0.296
(0.60)
1+ children widowed/separated/divorced/single
1.410
(1.49)
0.332
(0.40)
-0.972***
(-3.40)
-0.291
(-1.06)
Widowhood
Constant
N
1511
1137
Δ self-rated mental health – good or better at W1
Widowhood
Decrease in number of close relatives/friends
New ADL
New IADL
Cardiovascular disorder
Chronic illness
1-point deterioration in self-reported health
2-point deterioration in self-reported health
Retired
Unemployed
Change in income (000s)
1+ children emigrated
1+ children unemployed
1+ children
widowed/separated/divorced/single
Constant
N
Women only
Coeff.
t stat.
0.358
(1.38)
-0.00675 (-0.13)
0.129
(0.75)
0.251*
(1.75)
0.0326
(0.48)
0.0853
(1.49)
0.381***
(5.49)
0.776***
(5.55)
0.0274
(0.24)
0.246
(1.11)
-0.00045 (-0.02)
0.181**
(2.46)
0.149*
(1.78)
0.209*
(1.88)
0.0527
1528
(1.12)
Men only
Coeff.
t stat.
0.297
(0.81)
0.0468
(0.75)
0.0365
(0.20)
0.0824
(0.39)
0.0445
(0.60)
0.0323
(0.38)
0.427***
(6.07)
1.064***
(5.76)
0.185*
(1.73)
0.156
(0.68)
-0.0241
(-1.29)
-0.128
(-1.58)
-0.0335
(-0.31)
-0.0803
(-0.58)
0.0300
1120
(0.60)
ΔCESD; with interactions between history of mental health problems and
child emigrate
Mothers only
Coeff.
t stat.
5.028***
(2.82)
1.132***
(2.83)
2.286*
(1.94)
0.727
(0.71)
1.010**
(1.99)
0.756*
(1.65)
1.752***
(3.77)
1.356
(1.50)
1.206*
(1.82)
-2.709
(-1.04)
-0.0470
(-0.34)
Widowhood
Loss of close relatives/friends
Loss in functional capacity (new ADL)
Loss in functional capacity (new IADL)
Cardiovascular disorder
Chronic illness
1-point deterioration in self-rated health
2-point deterioration in self-rated health
Retirement
Unemployment
Change in income (000s)
Ref: No child’s emigration * no history of depression
No child’s emigration * history of depression
-1.995
Child’s emigration * no history of depression
0.870
Child’s emigration * history of depression
7.108***
Child’s unemployment
0.430
Child’s marital breakdown/widowhood
0.886
Constant
-2.229***
N
1706
(-1.64)
(1.44)
(2.97)
(0.73)
(0.85)
(-6.70)
Fathers only
Coeff.
t stat.
4.892***
(3.15)
0.124
(0.32)
0.712
(0.58)
0.909
(0.67)
0.858**
(2.09)
0.0768
(0.15)
-0.120
(-0.29)
3.012***
(2.74)
0.771
(1.60)
2.013
(1.52)
-0.0348
(-0.32)
-1.741
0.308
-1.182
0.109
0.690
-0.934***
1205
(-1.20)
(0.60)
(-0.33)
(0.18)
(0.82)
(-3.06)
• Does this suggest that there may be reverse
causality?
• Were children whose parents had suffered a
mental health problem more likely to
emigrate?
• We test this by running a probit regression
where “child emigrate” is now the dependent
variable.
Probit with child emigrate as dependent variable
Age
Medium education
High education
Employed
Other
Another town/city
Rural area
Income 2nd quintile
Income 3rd quintile
Income 4th quintile
Income 5th quintile
Married/cohabiting
Number of children
Return migrant
Past diagnosis of depression
CES-D score
Good self-rated health
Fair/poor self-rated health
Past diagnosis of cancer
Past diagnosis of heart attack
Constant
N
Mothers only
Coeff.
-0.0457***
0.164
0.285*
-0.175
-0.150
-0.179
-0.185
0.0349
-0.0703
0.149
0.0847
0.0471
0.144***
0.160
-0.106
-0.0145**
0.0272
-0.103
-0.0406
-0.417
1.290***
1,589
t stat.
(-6.47)
(1.31)
(1.94)
(-1.13)
(-1.10)
(-1.35)
(-1.58)
(0.26)
(-0.41)
(1.00)
(0.49)
(0.41)
(5.41)
(1.30)
(-0.45)
(-1.98)
(0.27)
(-0.64)
(-0.22)
(-0.92)
(2.66)
Fathers only
Coeff.
t stat.
***
-0.0519
(-5.52)
0.161
(1.22)
0.407***
(2.73)
-0.0969
(-0.61)
0.0242
(0.12)
-0.0101
(-0.07)
0.0221
(0.16)
-0.182
(-0.80)
-0.267
(-1.23)
-0.0902
(-0.49)
-0.163
(-0.94)
0.273
(1.63)
***
0.187
(5.57)
0.346***
(2.61)
0.306
(1.31)
-0.0119
(-1.09)
0.0175
(0.13)
-0.0704
(-0.36)
0.195
(0.73)
-0.918**
(-2.23)
1.271**
(2.09)
1,134
2013
2012
2011
2010
2009
2008
2007
2006
2005
2004
2003
2002
2001
2000
1999
1998
1997
1996
1995
1994
1993
1992
1991
1990
1989
1988
1987
And can we take the push to emigrate
as being an exogenous shock?
Emigrants in 1,000s
100
90
80
70
60
50
40
30
20
10
0
Does it matter that the emigrants’
parents are younger?
CESD regression for those aged 65 and under – only significant coefficients shown here
Widowhood
Decrease in number of close relatives/friends
Cardiovascular disorder
Chronic illness
1-point deterioration in self-reported health
2-point deterioration in self-reported health
Retired
1+ children emigrated
1+ children widowed/separated/divorced/single
Constant
N
Women only
Coeff.
t stat.
7.455*** (3.22)
1.479** (2.56)
1.455** (1.97)
1.409** (2.17)
1.712*** (2.59)
0.894
(0.58)
1.549
(1.55)
1.594** (2.30)
3.200** (2.37)
-2.788*** (-5.44)
1038
Men only
Coeff.
t stat.
6.266***
(11.54)
-0.282
(-0.53)
1.318**
(2.33)
-0.103
(-0.14)
0.300
(0.50)
4.377***
(2.73)
0.981*
(1.65)
0.410
(0.62)
0.453
(0.36)
-0.979**
(-2.22)
652
Does it matter if the emigrating child is
a son or daughter?
CESD regression – only showing coefficients for emigrant child dummy variables –
Different point estimates but not statistically significant
Women only
Child emigrating is male
Child emigrating is female
N
Men only
Coeff.
t stat.
Coeff.
t stat.
1.162
(1.33)
0.161
(0.21)
1.940**
(2.21)
0.534
(0.59)
1733
1207
Does it matter if the emigrating child
was living with the parents at W1?
CESD regression – no statistically significant difference
Women only
Men only
Coeff.
t stat.
Coeff.
t stat.
Child emigrating was co-resident at w1
1.036
(1.01)
0.571
(0.76)
Child emigrating was NOT co-resident at
w1
N
1.355*
(1.91) 0.0217
(0.03)
1753
1226
Other things we looked at...
• Age of emigrating child
• Does the emigrating child have children (ie.
grandchildren of our participants)
• Results as expected but no statistically
significant differences between estimated
coefficients
Conclusion
• There appears to be reasonably robust evidence
that the emigration of an adult child affects of the
mental health of mothers.
• But effect strongest among mothers with a
previous history of depression
• This is important in itself but also in terms of how
mental health impacts upon physical health
• There might also be impacts on potential
emigrants and another dimension to Mincer’s
(1978) “tied stayers”