Intergenera(onal  Living  Arrangements  in  South  Korea:  Health  and  Economic  Implica(ons  

Young  Kyung  Do  

Program  in  Health  Services  and  Systems  Research  

Duke-­‐NUS  Graduate  Medical  School  Singapore  

Asia  Health  Policy  Program  Seminar  Walter  H.  Shorenstein  Asia-­‐Pacific  Research  Center  Stanford  University  •  November  12,  2012  

Acknowledgements: I thank Chetna Malhotra, Kelvin Foo and Chandima Arambepola.

Three  studies  on  elderly  living  arrangements  

1.  The  effect  of  coresidence  with  an  adult  child  on  depressive  symptoms    among  older  widowed  women  in  South  Korea:  An  instrumental  variables  es=ma=on  (with  Malhotra  C,  published  in  Journal  of  Gerontology  B:  Social  Sciences,  2012)  

2.  Expecta=ons  about  bequests  and  informal  care  (by  coresiden=al  status)  (with  Foo  K,  under  review)  

3.  Adult  children’s  coresidence  with  parents  and  labor  force  par(cipa(on:  Genera=onal  differences  (under  development)  

To  put  it  another  way…    

Elderly  parents  

Adult  children  

Coresidence with adult child(ren)

Depressive sx.

Providing parental care

Expectation of receiving bequest

Expectation of receiving care from child

Expectation of giving bequest

Coresidence with parent(s)

Labor force participation

Three  studies  in  one  presenta(on:  ra(onale?  

  Same  theme:  Elderly  living  arrangements  

  Same  data:  Korea  Longitudinal  Study  of  Aging    Same  methodological  issue:  Living  arrangement  (&  informal  

care)  is  a  choice  

  Same  approach  to  this  methodological  issue:  Instrumental  variable  es=ma=on  

Popula(on  aging  in  South  Korea  1960

2010

2050

Source: Korea National Statistical Office (2006)

1960

2010

2050

Popula(on  aging:  doomed,  disastrous  future?  

•  Silver  tsunami  

•  Agequake  

Individual  aging:  rosy  and  flowery?  

•  Successful  aging;  Produc(ve  aging;  Ac(ve  aging;  Healthy  aging  

Missing  links?  

•  Micro-­‐macro  integra=on  

•  Holis=c  life  experience  of  elderly  and  ramifica=ons  

Trends  in  eldelry  living  arrangements  in  S.  Korea  

1990 1995 2000 2005

Elderly (≥ 65) (1,000) 2,162 2,624 3,347 3,372

Living arrangements

(%)

Total 100 100 100 100 1-person 8.9 13.3 16.2 32.0

1-generation 16.9 23.3 28.7 34.6 2-generation 23.4 23.0 23.9 23.8 3-generation 48.0 38.0 30.0 9.0

Living with non-family 1.2 0.8 0.4 0.4

Source: Census data, National Statistical Office

The  effect  of  coresidence  with    an  adult  child  on  depressive  symptoms    among  older  widowed  women  in  South  Korea:  An  instrumental  variables  es(ma(on  

Do YK, Malhotra C. The effect of coresidence with an adult child on depressive symptoms among older widowed women in South Korea: an instrumental variables estimation. Journal of Gerontology B: Psychological Sciences and Social Sciences. 2012 May;67(3):384–91. (A previous version was published as Stanford Asia Health Policy Program working paper #20.)

High  suicide  rates  in  elderly  Koreans  

Source: Korea Joongang Daily (Sept 10,2012) “Suicide a growing trend in elderly”

Unusual:  High  suicide  rates  driven  by  elderly  in  S.  Korea  

Source: Society at a glance 2009: OECD Social Indicators - ISBN 978-92-64-04938-3 - © OECD 2009

Trends  in  suicide  rates  Per  100  000  persons  

0

5

10

15

20

25

30

1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009

Austria Korea Japan OECD average

Source: OECD Factbook 2011: Economic, Environmental and Social Statistics - ISBN 978-92-64-11150-9 - © OECD 2011

Change  in  suicide  rates  (original  graph  by  OECD)  Percentage,  1995-­‐2009  or  latest  available  period    

-60

-50

-40

-30

-20

-10

0

10

20

30

40

50

60 153.6

Source: OECD Factbook 2011: Economic, Environmental and Social Statistics - ISBN 978-92-64-11150-9 - © OECD 2011

Public health crisis

Background:  Widowed  women  living  alone  in  an  aging  Korea      Rapid  aging  in  South  Korea  

  Elderly  (>  65)  from  9%  to  38%  (2050)    Life  expectancy  at  65  years:  21  years  (women);  16.6  years  (men)  

  Stresses  to  older  widowed  women  in  “Anomie”    Declining  intergenera=on  coresidence  &  family  support  

(“old”  norm)  vs.  s=ll  limited  personal  (re=rement)  savings  and  social  support  (“new”  norm):  vulnerable  to  stresses  

  High  rates  of  depression  and  increasing  rates  of  suicide  in  elderly  

Background:  Coresidence  with  adult  child  may  be  protec(ve    Women’s  expecta=ons  sa=sfied  (social  norm  of  filial  piety)  

  Emo=onal/instrumental/economic  support  from  adult  child  coresiding  

  Cultural  differences  (vs.  Western)  

  A  testable  hypothesis:  “Widowed  women  coresiding  with  adult  child  have  less  depressive  symptoms  than  widowed  women  living  alone”  

Methodological  issue:  Es(ma(ng  the  causal  effect  of  coresidence  

  Randomized  controlled  trial?:  Unthinkable  

  Comparison  of  group  means  of  depressive  symptoms  scores  

  Mul=ple  regression  controlling  for  observable  confounders  

  S=ll,  selec=on  on  unobservables  (endogeneity)  is  an  issue  

AHPP  Working  paper  #20,  Do  and  Malhotra  (2010)  

Comment  on  our  working  paper  (2010)    in  Johar  and  Maruyama  (2011)  “Does  Coresidence  Improve  an  Elderly  Parent’s  Health?”  

Causal  associa(on  between  coresidence  and  depressive  symptoms?  

Coresidence  with  an  

adult  child  

Depressive  symptoms  

Instrumental  variable  (IV)  

 mimics  the  random  assignment  of  coresidence  

•  Number of sons •  Whether eldest child is daughter : Statistical tests suggested these are good IVs

Data  and  Variables  

•  Data:  Korea  Longitudinal  Study  of  Ageing  (KLoSA),  2006/8  •  Sample:  Widowed  women  ≥65  (N=2,446)  with  at  least  one  

living  child  

•  Outcome:  10-­‐item  CES-­‐D  (0  to  30,  mean  10.0)  

•  Main  predictor:  Coresidence  with  adult  child  (mean:  57.9%)  •  Covariates:  demographic,  socioeconomic,  health  variables  

•  Instrumental  variables  used  –  Number  of  sons  (con=nuous)  

– Whether  eldest  child  is  a  daughter  (binary)  

Results  

•  OLS  es=mate:  −0.700  (95%  CI:  −1.196,  −0.204)  

– Close  to  group  mean  difference,  −0.097  

•  IV-­‐2SLS  es=mate:  −7.749  (95%  CI:  −14.092,  −1.407)  

•  Sta=s=cal  test  suggests  that  coresidence  is  endogenous  in  our  model  of  CES-­‐D  score.  –  OLS  underes=mates  the  protec=ve  effect  of  coresidence  on  depressive  symptoms.  

Discussion  •  Accoun=ng  for  the  endogeneity  of  coresidence  using  IV  method  

revealed  a  larger  protec=ve  effect  than  found  in  naïve  OLS.  

•  Decreasing  rates  of  coresidence  with  children  may  pose  a  public  health  concern  among  widowed  women  in  South  Korea.  

•  Limita=ons  –  Did  not  measure  the  type  and  quality  of  actual  or  perceived  support  –  Possible  differences  may  exist  by  coresidence  type  

–  Limited  to  older  community-­‐dwelling  widows  

Expecta(ons  about  bequests  and  informal  care  

(with Kelvin Foo, under review)

An  inconvenient  truth?  

•  Adult  children  provide  informal  care  to  their  parents  (at  least  in  part)  because  they  have  received,  are  receiving,  and/or  expect  to  receive  money  from  their  parents.  

•  Parents  who  can  give  money  to  their  adult  children  are  more  likely  to  receive  informal  care  from  their  adult  children  (when  the  need  for  informal  care  arises).  

Hypothesis  

•  H1:  Providing  informal  care  to  his  or  her  parent  increases  an  adult  child’s  expected  probability  of  receiving  an  inheritance  in  the  future.  

•  H2:  An  elderly  parent's  expected  probability  of  leaving  an  inheritance  increases  his  or  her  an=cipa=on  that  any  adult  children  will  be  willing  and  able  to  provide  informal  care  in  the  future.  

•  Secondary  hypothesis:  In  both  hypotheses,  the  es=mated  magnitude  of  effect  for  the  main  explanatory  variable  is  smaller  for  parents  and  children  who  are  coresiding  than  those  who  are  not.  

Literature  on  intergenera(onal  transfers  and  informal  care  •  Agreement  on  which  mo=ve  (in  transfers)  dominates  –  

exchange  or  altruism  –  is  unclear  (Sloan  et  al.  1997;  Perozek  1998).  

•  Exchange  mo(ve:    

Parents  and  children  provide  =me-­‐related  services  or  money  to  each  other  because  of  expecta=ons  that  the  other  party,  governed  by  the  moral  norm  of  reciprocity,  would  give  back  accordingly  (Gouldner  1960).    

•  Altruis(c  mo(ve:    

Parents  and  children  provide  =me-­‐related  services  or  money  to  each  other  simply  out  of  selfless  inten=ons  (Becker  1974).  

Exchange  Mo(ve  

•  Cox  (1987):  Children  with  higher  incomes  receive  greater  financial  transfers  from  their  parents  than  do  lower  income  children  –  contradicts  the  altruis3c  mo3ve    

•  Bernheim  and  colleagues  (1985):  a  significant  posi3ve  rela3onship  between  the  bequeathable  wealth  of  parents  in  poor  health  and  the  asen=on  received  from  their  children.    

•  Henresa  et  al.  (1997):  a  posi=ve  and  sta=s=cally  significant  rela=onship  between  parents’  past  financial  transfers  and  children’s  current  caregiving  propensi=es.    

•  Cox  and  Rank  (1992):  Children’s  =me  help  to  parents  posi=vely  related  to  their  propensity  to  receive  financial  transfers  from  their  parents,  whereas  financial  transfers  are  intended  as  payments  for  =me  help.  

Altruism    •  Parents  will  give  transfers  to  their  children  regardless  of  whether  their  

children  provide  financial  support  to  them  or  take  care  of  them  during  old  age;  rela=vely  more  transfers  are  given  to  children  with  greater  financial  needs  (Becker  1974;  Cox  1987).    

•  McGarry  and  Schoeni  (1995):  an  inverse  rela=onship  between  income  and  transfers,  sugges=ng  altruis=c  mo=va=ons.    

•  Parents’  expecta=ons  about  giving  a  bequest  do  not  significantly  influence  their  expecta=ons  about  entering  a  nursing  home  (Lindrooth  et  al.  2000)  –  suggests  that  expecta=ons  about  giving  a  bequest  have  no  effect  on  their  expecta=ons  of  receiving  informal  care  from  their  children  in  the  future.  

•  Sloan  and  colleagues  (2002):  less  affluent  parents  received  more  money  from  children.    

•  Compared  with  parents  of  a  lower  economic  standing,  parents  of  a  higher  economic  status  did  not  have  a  higher  likelihood  of  receiving  help  from  their  children  (Sun  2002).  

Exchange  vs  (?)  Altruism  

•  Is  it  possible  that  these  two  coexist  as  equally  important  reasons  for  intergenera=onal  transfers?    

(Silverstein  et  al.  1995;  Caputo  2002;  Light  and  McGarry  2004;  Koh  and  MacDonald  2006;  Norton  and  Van  Houtven  2006).  

Models  Exp_INHERIT  =  f  (IC,  number  of  siblings,  age,  marital  status,  educa=on  level,  amount  of  assets,  home  ownership,  parent's  home  ownership,  residen=al  area,  year  effect):  Adult  children’s  expecta=on    

Exp_IC  =  f  (Exp_BEQUEST,  number  of  children,  age,  marital  status,  educa=on  level,  amount  of  assets,  home  ownership,  residen=al  area,  year  effect):  Parent’s  expecta=on  

→  Separate  analysis  by  gender  &  coresiden=al  status  (revealed  preference)  

Causal  associa(on  between  informal  care  and  bequest  expecta(on?  

Informal  care  to  parents  

Expecta(on  of  bequest  

Instrumental  variable  (IV)  

 mimics  the  random  assignment  of  caregiving  

•  Father’s ADL limitation (1 vs. 0) •  Mother’s ADL limitation (1 vs. 0) : Statistical tests suggested these are good IVs

Adult  children’s  IC  (giving)  →  ↑Bequest  (receiving)  expecta(on?  

Not coresiding with parent(s) Coresiding with parent(s) Male Female Male Female

OLS IV OLS IV OLS IV OLS IV Any parental care (IC=1 vs. 0) 5.24 5.80 5.24 28.17* −0.58 −4.58 2.98 5.86

Mean of predicted values at IC=0 19.7% 19.7% 15.0% 15.0% 18.4% 18.8% 12.4% 12.0%

Models control for number of siblings, age, marital status, education level, amount of assets, home ownership, parent's home ownership, residential area, year effect. * p < 0.05.

Parent’s    Bequest  (giving)  expecta(on  →  IC  (receiving)  expecta(on  (%)?  

27.8%   32.5%  38.2%  

47.6%  

9.3%   1.8%  

9.6%  

3.3%  

0%  

10%  

20%  

30%  

40%  

50%  

60%  Male,  

non-­‐

cores

iding  

Male,  

cores

iding  

Fema

le,  

non-­‐

cores

iding  

Fema

le,  

cores

iding  

Incremental  probability  when  E(bequest)=1  Predicted  probability  when  E(bequest)=0  

*

*

Models control for number of children, age, marital status, education level, amount of assets, home ownership, residential area, year effect. * p < 0.05.

Discussion  

•  Some  evidence  on  the  exchange  mo=ve  for  parental  care  

•  Who  is  less  likely  to  receive  care  from  children?  (Poor  elderly!)  

–  Inequality  in  late  life  

–  Demand  for  public  long-­‐term  care  

•  Coresidence  masers  

–  Effect  modifica=on  (revealed  preference  for  future  informal  care)  

–  ↓  Intergenera=onal  coresidence  →  ↑  role  of  the  exchange  mo=ve  in  

parental  caregiving  

Adult  children's  coresidence  with  parents  and  labor  force  par(cipa(on:  Genera(onal  differences  

(under development)

Literature  (from  Asian  countries)  •  Japan    

–  Coresidence  with  parents  or  in-­‐laws:    •  Significant  posi=ve  impact  on  married  women’s  LFP  (Sasaki,  2002)  

•  Coresidence  with  husband’s  parents  increased  LFP  of  married  women  by  24%  and  by  19%  when  co-­‐residing  with  own  parents  (Oishi  and  Oshio,  2006)  

–  Over  a  quarter  of  older  adults  co-­‐reside  with  unmarried  children    

•  Ques3on  of  whether  the  unmarried  children  are  ‘parasites’  or  the  parents  are  dependents  

•  Indonesia:    •  Unmarried  children  more  likely  to  cohabit  with  parents  

who  are  healthy,  have  more  wealth,  and  have  non-­‐market  value  =me  (Johar  and  Maruyama,  2011)  

•  South  Korea  –  Coresidence  with  parents-­‐in-­‐law  has  no  effect  on  labor  force  par=cipa=on  of  married  women  (Chun  and  colleagues,  2009)  

Sample  construc(on  

Respondent

45

Respondent’s adult children

1

63 105 64 45

M: 2,910 F: 2,726

M: 5,776 F: 5,370

25 45 44 64

Exclude adult children of respondents with any activity limitations

response

Causal  associa(on  between  coresidence  and  labor  force  par(cipa(on?  

Coresidence  with  parent(s)  

Labor  force  par(cipa(on  

(LFP)  

Instrumental  variable  (IV)  

 mimics  the  random  assignment  of  coresidence  

•  Number of brothers •  Number of sisters •  Being eldest son in family (for males only) : Statistical tests suggested these are good IVs

Probit  and  bivariate  probit  (BVP)  regression  of  LFP  as  func(on  of    coresidence  with  parent(s),  by  gender  

Male Female Probit IV-BVP Probit IV-BVP

Younger adult children (25−44) −0.210** −0.767** 0.094 −0.738**

Midlife adult children (45−64) −0.063 0.140 0.648** 0.827

Models control for adult child's characteristics (age, education, marital status, home ownership, number of own children) and parents’ characteristics (marital status, education, home ownership, total assets, residential region). ** p < 0.01.

•  Younger: For both males and females, coresidence reduces likelihood of labor force participation, even after controlling for endogeneity of coresidence. •  Midlife: For females, coresidence increases likelihood of labor force participation. No statistical evidence on endogeneity.

Discussion  

•  Genera=onal  difference:  cannot  differen=ate  between  age  and  cohort  effects  (or  both)  

•  Younger  genera=on’s  coresidence  with  parents  may  have  nega=ve  economic  consequences.  

•  Will  younger  genera=on’s  coresidence  con=nue?  (less  likely:  what  if  they  move  out  later?)  

•  Limited  savings  for  parents’  late  life  

Grand  summary  

•  Intergenera=onal  living  arrangements  in  South  Korea,  coupled  with  rapid  popula=on  aging,  have  health  and  economic  implica=ons  for  both  genera=ons  (and  society  as  a  whole).  

•  Op=mal  public  policies  should  consider  elderly  living  arrangements.  Ex:  –  At-­‐risk  groups  (widowed  women  living  alone)  

–  Public  LTCI  →  [↓  Coresidence  →]  ↑↑  Demand  for  public  LTC  (esp.  among  the  poor  elderly)  

www.duke-­‐nus.com.sg  

Contact  informa(on:  Young  Kyung  Do,  MD,  MPH,  PhD  Assistant  Professor  Duke-­‐NUS  Graduate  Medical  School  Singapore    8  College  Road,  Singapore  169857    Tel:  (65)  6516  2790  |  Fax:  (65)  6534  8632    Email:  young.do@duke-­‐nus.edu.sg