Educational Sciences: Theory & Practice - 13(1) • Winter • 50-53
©
2013 Educational Consultancy and Research Center
www.edam.com.tr/estp
Examination of the Film “My Father and My Son”
according to the Basic Concepts of Multigenerational
Family Therapy
a
b
Tülin ACAR
Nilüfer VOLTAN-ACAR
Hacettepe University
Hacettepe University
Abstract
The aim of this study was to evaluate the basic concepts of Multigenerational Family Therapy and
to evaluate the scenes of the film ‘’My Father and My Son’’ according to these concepts. For these
purposes firstly basic concepts of Multigenerational Family Therapy such as differentiation of self,
triangles/triangulation, nuclear family emotional system, family projection process, emotional cut
off, multigenerational transmission process, sibling position, societal regression and genogram
technique of this therapy were explained. Then, in order to clarify family relations in the film, the
family genogram of the family was drawn. Finally, sixteen scenes of the film which depict the basic
concepts of Multigenerational Family Therapy were examined. Results of this examination showed
that family members of the three generation usually had low levels of differentiation. Furthermore, some basic concepts of Multigenerational Family Therapy such as triangles, nuclear family
emotional system, emotional cut off, family projection process, differentiation of self, multigenerational transmission process were evident in all three generations family. In short, the film My
Father and My Son can be a rich source in Multigenerational Family Therapy training.
Key Words
Multigenerational Family Therapy, “My Dad and My Son”, Film, Family Therapy.
There are many approaches of family therapy each
having different views on family and thus different
approaches for helping families with their difficulties. One of these approaches is the Multigenera-
a
Tülin ACAR, M.C. is currently research assistant
at the Department of Educational Sciences,
Program of Psychological Counseling and Guidance. Her research interests include bullying,
counseling process, counselor education, career
counseling and psychodrama. Correspondence: Hacettepe University, Faculty of Education,
Department of Educational Sciences, Program
of Psychological Counseling and Guidance, Beytepe, Ankara/Turkey. E-mail: tulina@hacettepe.
edu.tr; tulinac@gmail.com. Phone: +90 312 297
85 50/128 Fax: +90 312 299 20 27.
b
Nilüfer VOLTAN-ACAR, M.C.; M.S.W.; PhD., Hacettepe University, Faculty of Education, Department
of Educational Sciences, Program of Psychological Counseling and Guidance, Beytepe, Ankara/
Turkey.
tional Family Therapy. Murray Bowen’s Multigenerational Family Therapy is one of the systems therapies that views the family as a whole. According to
this approach when there is a low level of anxiety
in the families the emotional system of the family
does not develop symptomatic issues. On the other
hand when there is a high level of anxiety some
problems may arise in the family (Gladding, 2006).
Multigenerational Family Therapy differs from the
other approaches by its emphasis on the emotional system and history of the family (Corey, 2008;
Sharf, 2008). The approach insists that interactional
patterns among family members are transmitted
from one generation to the other. Family is viewed
as a dynamic unit. It contents that changes in one
member can affect the system and changes in the
system can affect individual members of the family.
TV shows and other media of visual media sit have
influences on the functioning of today’s families.
The film “My Father and My Son” revolves around
events, traumas and sufferings resulting from the
ACAR, VOLTAN-ACAR / Examination of the Film “My Father and My Son” according to the Basic Concepts of...
September 12th Queue in Turkey (Wikipedia, 2010,
2011). Üstüner Wambach (2009) categorizes this
film as a “trauma film”. This film is depicted as a film
which mentions the generation conflict (film.com.
tr, 2010). Also, the film critiques accentuates that
the film is successful in artistic perspective (Vardar,
2008). This film received many awards (film.com.tr,
2011). Films may be utilized for training of mental
health professionals (Paddock, Terranova, & Giles,
2001; Shepard & Brew, 2005; Toman & Rak, 2000;
Villalba & Redmond, 2008).
Basic Concepts of Multigenerational Family
Therapy
Differentiation of the self, triangles/triangulation,
nuclear family emotional system, family projection
process, emotional cut-off, multigenerational transmission process, sibling position, societal regression and genogram are basic concepts of the Multigenerational Family Therapy (Bitter, 2009; Fenell
& Weinfold, 2003; Gehart, 2010; Gladding, 2006;
Goldenberg & Goldenberg, 2008; Nazlı, 2003).
Main Characters of the Film “My Father and My
Son”
Deniz (the main character- 5-6 year old boy), Hüseyin
Bey (Mr. Hüseyin; grandfather/father), Nuran Hanım
(Mrs. Nuran; grandmother/mother), Sadık (father/
son), Salim (uncle/brother), Hanife (aunt), Gülbeyaz
(grand aunt/aunt) and Özkan (Sadık’s friend).
Examination of Some Scenes of “My Father and
My Son”
In this study, 16 scenes from the film, My Father
and My Son, were sampled and examined. The below abstract presents several of these scenes with
respect to basic concepts of Multigenerational
Family Therapy.
Train Ride to the Village: The little boy, Deniz,
fantasizes as he often does. He imagines himself as a
highly strong and invincible person. In Sadık’s reactions and expressions one can see the influences of
his past political experiences on the nuclear family
emotional process. Furthermore, Sadık’s political
views and acts appear to be central to the tension
and conflict between him and his father, Hüseyin
Bey. The conflict has significant negative impact on
Deniz. According to Multigenerational Family Therapy, in order to understand the family, family should
be evaluated in a holistic- systemic way. Therefore, it
can be said that the burn out and disgust that Sadık
experiences can be the result of his preferences.
These effects the present relationship and interaction between Sadık and his son Deniz. On the other
hand, Sadık is aware that he has a terminal illness
and keeps distant to his son. In other words, looking
at it from the concept of nuclear family emotional
system, he deals with the anxiety within the family
unit by distancing himself.
***
Deniz and Sadık while coming home by truck…
In this scene Sadık cannot explain to his son that
he is going to die and he tries to keep his son away
from himself. Sadık’s pattern in distancing is quite
similar to that of his father who also avoids expressing his feeling to his son. In the multigenerational
transmission process strategies of coping methods
are transmitted through generations. Here in reality Deniz does the same as his grandfather and father. Instead of talking about the hurt he prefers to
sweep it under the rug. It can be said that there is
similarity of expression of feelings between members of the respective generations.
***
First Encounter of Deniz with Hüseyin (grandfather)
First encounter of Deniz with the third generation,
Deniz visualizes (imagines) that his grandfather is
a hero.
Figure 1
Family Genogram of the film “My Father and My Son”
In the proceeding parts, Deniz depicts his father as
a witch, bad man, and pirate. This may be valued as
51
EDUCATIONAL SCIENCES: THEORY & PRACTICE
Deniz perceives his grandfather as strong, mighty.
This viewpoint, is similar to Sadık’s feelings toward
Hüseyin, the father. Hüseyin Bey is a strong character in the eyes of his wife, his son and his grandson.
On the other hand, Hüseyin, the father is viewed
as an authoritarian figure by Sadık. Deniz, grandson thinks of Hüseyin, his grandfather, as strong
too. Therefore, being parallel to multigenerational
transmission process, it can be said that the emotional process of the previous generations is transmitted to the present family emotional process. In
this scene Nuran, the wife, mother and grandmother, in order to reduce the anxiety between Hüseyin,
husband and Sadık, the son intervenes in the event.
Thus, the family experiences the tension by Sadık’s
coming back home years later. The relationship between Nuran, the mother and Hüseyin, the father
deteriorates. This position can be given as an example of forming triangles. However, Sadık, the son,
who re-joins the family causes a negative effect.
On the other hand, Nuran, the mother, in order to
maintain the balance, tries to reduce anxiety and
tension. Hence, the new triangle is formed.
***
First dinner at home…
In this scene, in spite of the Sadık’s efforts Hüseyin,
the father, experiences the emotional cut off. The
tension between Hüseyin, the father and Sadık, the
son emerges thoroughly. At this point Hüseyin, the
father, forms a triangle by calling Nuran, the mother to a dialogue. Nuran, by suggesting her son to be
calm, tries to preserve the balance of the triangle.
***
While coming back from the field, Salim, the older
son, and Sadık, the younger son encountering with
the father..
Sadık’s relationship with his brother, Salim is also
remarkable. Salim is a family member who is
slightly mentally retarded. Salim’s level of differentiation of self is lower than Sadık. Their relationship is that of two loving brothers.
***
Sadık with Hüseyin while in the hospital…
In this scene, Hüseyin cannot express his feelings
and love toward Sadık. In the previous scenes,
Sadık complains about his not being expressive toward his son. Deniz, the grandson, cannot reflect
his feelings, either. Thus, the difficulty of reflecting
feelings passes through the nuclear family emotional system and multigenerational transmission
processes are evident.
***
52
While coming back home after the funeral… Hüseyin stops the car and gets off…
Hüseyin feels guilty because of his son’s death.
Hüseyin did not permit his son for self-differentiation Hüseyin cannot distinguish his feelings form
his thoughts. Sadık’s death causes a great shock in
the family and no one can be able to cope with this
shock. In order to cope with this intensive anxiety
Sadık’s friend Özkan tries to help the family but
cannot succeed.
Conclusion and Recommendations
When the film My Father and My Son is reviewed
in terms of Multigenerational Family Therapy,
Deniz’s family on his father’s side has members
who have a lower level of differentiation. In the
third generation family, the concepts of triangles
nucleus family emotional system, emotional cut
off, family projection system, differentiation of self,
multigenerational transmission process are observed. The results of this study show that anxiety is
evident in three generations of the family. Furthermore, in each generation members of the family do
not deal with this anxiety in functional manners.
The film illustrates that the family patterns have a
multi-generational quality to them. Therefore, the
film can be used in family therapy training- particularly with respect to Multigenerational Family
Therapy. Also, within the “cinema therapy” films
can be used in therapy or counseling as creative
tools. Future studies can examine the film My Father and My Son with respect to other theories of
family therapy. Likewise, various other films can
be used in exploring concepts of Multigenerational
Family Therapy.
References/Kaynakça
Bitter, J. R. (2009). Theory and practice of family therapy and
counseling. Belmont: Brooks/Cole Cengage Learning.
Corey, G. (2008). Psikolojik danışma, psikoterapi kuram ve
uygulamaları. Ankara: Mentis Yayıncılık. (Orijinali 2005’te
yayımlanmıştır).
Fenell, D. L., & Weinhold, B. K. (2003). Counseling families an
introduction to marriage and family therapy (3rd ed.). Denver:
Love Publishing Company.
Film.com.tr. (2010). Kuşak çatışması. http://film.com.tr/search.cfm?etiket=kusak_catismasi adresinden 12.12.2010 tarihinde edinilmiştir.
Film.com.tr. (2011). TYB de “Babam ve Oğlum”u Seçti. http://
film.com.tr/haber/index.cfm?hid=2496 adresinden 05.01.
2011 tarihinde edinilmiştir.
ACAR, VOLTAN-ACAR / Examination of the Film “My Father and My Son” according to the Basic Concepts of...
Gehart, D. (2010). Mastering compentencies in family therapy a
practical approach to theories and clinical case documentation.
Belmont: Brooks/Cole Cengage Learning.
Gladding, S. T. (2006). Family therapy (4th ed.). Ohio: Pearson
Merrill Prentice Hall.
Goldenberg, H., & Goldenberg, I. (2008). Family therapy an
overview (7th. ed.). Belmont: Thomson Higher Education.
Nazlı, S. (2003). Aile danışmanlığı (3. bs.). Ankara: Anı Yayıncılık.
Paddock, J. R., Terranova, S., & Giles, L. (2001). Sasb goes hollywood: teaching personality theories through movies. Teaching
of Psychology, 28 (2), 117-121.
Sharf, R. S. (2008). Theories of psychotherapy and counseling
consepts and cases (4th ed.). Belmont: Thomson Brooks/Cole.
Shepard, D. S., & Brew, L. (2005). Teaching theories of couples
counseling: The use of popular movies. The family journal, 13
(4). Retrieved June 01, 2012 from http://tfj.sagepub.com/content/13/4/406.full.pdf+html.
Toman, S. M., & Rak, J. F. (2000). The use of cinema in the counselor education curriculum: strategies and outcomes. Counselor Education & Supervision, 40 (2). Retrieved June 18, 2012
from http://ha6uq4xy8k.scholar.serialssolutions.com/?sid=go
ogle&auinit=SM&aulast=Toman&atitle=The+use+of+cinema
+in+the+counselor+education+curriculum:+Strategies+and+
outcomes&id=doi:10.1002/j.15566978.2000.tb01242.x&title=
Counselor+education+and+supervision&volume=40&issue=
2&date=2000&spage=105&issn=0011-0035.
Üstüner Wambach, Ö. (2009). Trauma cinema: A critical view
on beynelmilel and babam ve oğlum. Yayımlanmamış yüksek
lisans tezi, Bilkent Üniversitesi, Ankara.
Vardar, A. (2008). Babam ve oğlum. http://www.sinemafilm.
net/babam_ve_o%C4%9Flum.htm adresinden 04. 02. 2011
tarihinde edinilmiştir.
Villalba, J. A., & Redmond, R. E, ( 2008). Crash: Using a popular film as an experiential learning activity in a multicultural
counseling. Counselor Education & Supervision, 47, 264-276.
Wikipedia.org. (2010). Babam ve oğlum. http://tr.wikipedia.
org/wiki/Babam_ve_O%C4%9Flum adresinden 13. 01. 2010
tarihinde edinilmiştir.
Wikipedia.org. (2011). Babam ve oğlum. http://tr.wikipedia.
org/wiki/Babam_ve_O%C4%9Flum#Ald.C4.B1.C4.9F.
C4.B1_.C3.96d.C3.BCller
adresinden 05.01.2011tarihinde
edinilmiştir.
53
Copyright of Educational Sciences: Theory & Practice is the property of EDAM- Education Consultancy
Limited and its content may not be copied or emailed to multiple sites or posted to a listserv without the
copyright holder's express written permission. However, users may print, download, or email articles for
individual use.
635850
research-article2016
JOS0010.1177/1440783316635850Journal of SociologyEasthope et al.
Article
Changing perceptions of
family: A study of
multigenerational
households in Australia
Journal of Sociology
2017, Vol. 53(1) 182–200
© The Author(s) 2016
Reprints and permissions:
sagepub.co.uk/journalsPermissions.nav
https://doi.org/10.1177/1440783316635850
DOI: 10.1177/1440783316635850
journals.sagepub.com/home/jos
Hazel Easthope
University of New South Wales
Edgar Liu
University of New South Wales
Ian Burnley
University of New South Wales
Bruce Judd
University of New South Wales
Abstract
Many people around the world live in households with multiple generations of related adults
(multigenerational households). While more prominent in certain cultures, multigenerational living
is also an important part of the lives of millions in societies where this arrangement has not been
seen as ‘the norm’. Australia is one such case, where one in five people live in a multigenerational
household. This article presents findings of a research project on multigenerational households
in Australia, including a survey of 392 people, 21 diaries and 21 follow-up interviews to explore
how multigenerational household members understand their own experiences of living together.
It focuses particularly on whether they feel multigenerational living is a socially accepted
living arrangement. The article concludes with a discussion about how these experiences and
understandings of multigenerational family members may reflect changing social norms regarding
the form and role of families in Australian society.
Keywords
Australia, dependence, family, multigenerational, social norms
Corresponding author:
Hazel Easthope, Faculty of Built Environment, UNSW Australia, Sydney, New South Wales 2052, Australia.
Email: hazel.easthope@unsw.edu.au
Easthope et al.
183
This article draws on findings of a three-year research project on multigenerational
households in Australia. It focuses on the reflections of the research participants about
their experiences of living in a multigenerational household and whether this household
form has become more accepted over time. This question is important because the social
acceptability of multigenerational living affects the everyday lives and wellbeing of people who live in multigenerational households. This is no insignificant number of people;
one in five people in Australia currently live in a multigenerational household made up
of two or more generations of related adults. If these people are stigmatised, then this
will impact upon their health and wellbeing and their relationships with other family
members.
This question is also important because of its implications for the uptake of multigenerational living arrangements in the future and the resulting impact on the ability of families to provide financial, practical and social supports to household members. Australian
government policies, like those in many other western countries, increasingly rely on the
family to provide support in place of the state.
The majority of the academic literature on multigenerational living to date has focused
on the implications of this living arrangement for financial and other support provided
within the family. Much of this interest has centred on the delayed home-leaving of
young adults, particularly the reasons for this change (e.g. Flatau et al., 2007) and also
the financial implications for older generations in these households (e.g. Cobb-Clark,
2008). International research on multigenerational households in western countries has
also largely focused on delayed home-leaving (e.g. Iacovou, 2010 in Europe) and financial implications (e.g. CMF, 2014 in the UK).
Interestingly, while interest in multigenerational living has increased in Australian
academic and policy circles, this does not reflect a rapid increase in multigenerational
living. In fact, similar proportions of the population have lived in multigenerational
households since at least the mid 1980s. A likely reason why multigenerational households have received little attention to date is that the Australian Bureau of Statistics
(ABS) does not report on multigenerational households as a distinct household form,
making analysis more difficult. In this article, we report on customised cross-tabulations
purchased from the ABS to overcome this limitation.
Given that such a large proportion of the population lives in multigenerational households, and these households have been relatively neglected in academic and policy literature despite being a common part of Australian society for at least 30 years, it is timely
to consider how people who live in multigenerational households understand their experiences of living together.
The article begins with a summary of recent sociological research on the family.
It then presents a summary of the study approach and methods followed by detailed
information about multigenerational households in Australia. The remainder of the
article considers whether the research participants felt that multigenerational living
is an accepted household form in Australian society and the implications of these
findings. While this research was undertaken in Australia, the findings will be of
interest in other countries where multigenerational living is also not viewed as traditional, yet is nevertheless receiving increased recognition as an important household form.
184
Journal of Sociology 53(1)
Sociology of the family
For the past half-century, there have been debates within sociology about ‘the family’.
Over that period, there has been a move away from seeing the family as an institution
(the family) towards a focus on the practices of families (what people do in their family
life) and the importance of emotion in these practices (e.g. Morgan, 2011). There has also
been a shift in focus from examining the importance and functioning of ‘nuclear’ and
‘extended’ families towards an increasing recognition of family diversity and complexity
(e.g. Farrell et al., 2012; Widmer, 2010).
Gilding (2010: 760) explains that within sociological research on the family, there has
been a move to recognising ‘reflexivity over and above convention’. Conservative family sociologists have despaired at the decline of the family as a social institution, while
critical accounts have welcomed ‘the decline of patriarchal and heteronormative institutions’ (Gilding, 2010: 760); yet both sides have recognised reflexivity (in the form of
deinstitutionalisation, individualisation, and new forms of connectedness and embeddedness) over conventional understandings of the family.
However, some family sociologists have recently called for recognition of the importance of convention as well as reflexivity in understanding families. These academics
(e.g. Gilding 2010 and Liefbroer and Billari, 2010) recognise the continued influence of
social norms and constraints on the timing, sequencing and frequency of life transitions
(such as the institutions of marriage, monogamy and childbirth) and resulting implications for household formation and living arrangements without over-prioritising individualisation and personal choice. As Gilding (2010: 774) explains:
the family is best understood as an institutional regime; that is, an assemblage of institutions,
configured in different ways in different places at different times. This framework […] provides a
vehicle to understand the dynamics of family relationships and practices, giving due weight to both
their reflexive reconfiguration on the one hand, and their institutional embeddedness on the other.
In many ways, this debate parallels the longstanding discussion in sociology around the
relative importance of structure and agency, where the reflexive approach to the family
can be seen to prioritise agency, while conventional approaches prioritise structure. The
approach we have taken in this project is to recognise the importance of both agency and
structure in understanding family life.
Method
This article reports on selected findings from a larger project examining multigenerational households in the two Australian cities of Sydney and Brisbane. The aim of the
research was to determine the principal drivers of the prevalence of multigenerational
households in these cities and how they affect the day-to-day lives of families. The
research included:
1.
Statistical analysis of custom cross-tabulations over the period 1986–2011 from
the ABS with socio-demographic information about multigenerational households in Sydney and Brisbane.
Easthope et al.
2.
3.
4.
185
A detailed online survey of 392 people1 living in multigenerational households
in Sydney and Brisbane undertaken between August 2012 and July 2013. The
survey asked about the respondents’ history of living together (how long they
have lived together; the timing of the arrangement; whether the offspring had
previously left home but since then returned), their reasons for living together,
satisfaction with their current arrangement, their reasons for (dis)satisfaction,
and their preferred living arrangements for the long term. A series of demographic questions were also asked, including the age, birthplace, ancestry, education, employment and occupational status of household members, household
size and household composition. The survey was made available in English,
Arabic, Simplified Chinese and Spanish, and was promoted in 11 local and
migrant newspapers as well as in the 50Something magazine for seniors, the
Gumtree online classifieds paper, and online by the Tenants Union New South
Wales, the University of Queensland and the University of New South Wales.
People who self-identified as members of multigenerational households opted
in to complete the survey. This meant that multigenerational households that
did not meet our definition as stipulated for the census analysis also completed
the survey, including those with members of their household living in ancillary
dwellings such as granny flats, as well as households where no members of the
youngest generation were over 18 years old. All returned surveys were in
English, though the responding households represented a diverse range of cultural backgrounds (see Table 1).
Diaries completed by survey respondents who agreed to continue with the
research, detailing their thoughts about, and experiences of living in, a multigenerational household over a four-week period. A total of 21 diaries were received
from 15 households during August and September 2013.
In-depth interviews with survey respondents who agreed to continue with the
research, including most of those who completed the diaries. A total of 21 interviews were conducted with members of 18 households between October 2013
and February 2014. Four of the interviews included multiple participants.
This article draws largely from the results of the diaries and interviews. However, the
survey results are referenced where these provide supporting information for the issues
raised in the diaries and interviews.
An important limitation of this article is that perceptions of social norms regarding the
family and the acceptability of multigenerational living were drawn only from the experiences of people living in multigenerational households. Nonetheless, the experiences
and understandings of multigenerational household members about how they feel others
see their living arrangements provide a rich and informed view of social change in this
context.
Multigenerational households in Australia
In 2011, over 4 million people in Australia were living in a multigenerational household. That is, one in five Australians lived in a dwelling where two or more
186
Journal of Sociology 53(1)
Table 1. Cultural diversity of survey respondents and other members of their household.
Top 5 primary ancestries
Australian
East Asian
Western European
South-East Asian
South Asian
Other
n*
Top 5 brithplaces
Australia
China
UK
Vietnam
Hong Kong
Other
n*
Top 5 languages spoken at home
English
Cantonese
Mandarin
Other Chinese languages
Vietnamese
Other
n*
Sydney
Brisbane
Total
35%
17%
11%
11%
8%
18%
998
56%
4%
13%
2%
4%
21%
469
42%
13%
12%
8%
6%
19%
1467
56%
7%
4%
4%
4%
26%
852
73%
2%
6%
0%
1%
19%
435
62%
5%
4%
3%
3%
23%
1287
65%
9%
5%
4%
4%
14%
992
87%
2%
1%
1%
0%
9%
471
72%
6%
3%
3%
3%
12%
1463
Note: Respondents were asked their, and other household members’ (for up to 10 members), primary
ancestry, birthplace and languages spoken at home. % based on total number of responses.
generations of related adults (where the oldest of the youngest generation is aged 18
or over) were living together. This definition follows the many common academic
descriptions of multigenerational household forms; including young adults delaying
their first home-leaving, or an elderly parent joining their adult child’s family household for reasons of care. We acknowledge some multigenerational household forms
will be omitted from these figures, such as three-generational households where no
household member of the youngest generation (i.e. grandchildren) has turned 18.
Additionally, residents of secondary dwellings such as granny flats are counted as
separate households, and as such any households where some members live in a
granny flat are also excluded from the census analysis. As the construction of secondary dwellings in Australian cities was not permissible until relatively recently, there
are likely to be few households excluded from our analysis for this reason. This definition thus accounts for the majority of multigenerational households, although we
acknowledge that it still likely provides an undercount of all multigenerational
households.
Easthope et al.
187
The proportion of the total population living in multigenerational households has
changed little since 1986. While the number of people living in multigenerational
households increased by almost one million (30%) between 1986 and 2011, the population as a whole increased by 34%. However, the growth of multigenerational households has not been uniform across the country and in some areas their growth has
outstripped population growth. For example, the number of multigenerational households in Sydney increased by 44% between 1986 and 2011, outstripping the city’s
household growth of 40%.
Large proportions of multigenerational households are middle-aged couples (45–54
years old) living with their young adult children (18–24 years old) (see Figure 1) .
However, this does not simply reflect young people staying at home for longer and studying while ‘living off’ their parents. While a third of all 18–29-year-olds live in a multigenerational household, only 30% are dependent students2 and 70% are non-dependents.
There are also three-generation multigenerational households (where a member of the
youngest generation is aged over 18), and households where the youngest generation is
in their 30s or 40s.
Some people living in multigenerational households have moved to Australia from
countries and regions where multigenerational living is common. For example, 35% of
Australians born in North Africa and the Middle East and 28% of Australians born in
South East Asia lived in multigenerational households in 2011 (see Figure 2). This may
provide a partial explanation as to why the growth in multigenerational households has
been greater in Sydney, which has the largest overseas-born population in Australia, with
many migrants from South East Asia (ABS, 2014). However, three-quarters of all multigenerational household residents in Australia were born in Oceania (mainly Australia but
also New Zealand and other Pacific island countries; see Figure 3).
While changes in the acceptance of multigenerational living are likely to be influenced by the changing cultural makeup of Australia’s cities, available evidence indicates
that this is not the sole reason for this change. In their analysis of the Household, Income
and Labour Dynamics in Australia survey dataset, Flatau et al. (2007) found that even
after controlling for education, family background and ethnicity there has been a gradual
increase in the age of offspring when they first leave the parental home in Australia.
Indeed, our census analysis shows significant absolute increases in the 45-54 (+307,201),
55-64 (+201,193), 25-34 (+146,806) and 20-24 (+143,488) age groups between 1986
and 2011; the next most sizeable absolute increase was for the 75+ age group (+79,434).
Because of the more sizeable increases in the older age groups, however, slight proportional decreases were noticed for the 18-19 and 20-24 age groups (Figure 4). These
numbers indicate that the dominate family arrangement of multigenerational households
in 2011 being middle-aged parents and adult children in their early 20s to early 30s,
unlike in 1986 when younger age groups in each generation were more readily represented. This shift is particularly important among young adults aged in their late 20s and
early 30s at the time at which Flatau et al. (2007) conducted their analysis (the mid
2000s, and therefore prior to the global financial crisis, which some writers argue further
delayed some young adults’ first home-leaving). They suggest that broader changes in
norms and values have influenced these changed household outcomes.
188
Journal of Sociology 53(1)
Figure 1. Household breakdown: Multigenerational households and other family households,
Australia 2011.
Is multigenerational living accepted in Australia?
The analysis of the census data indicated that multigenerational living is common. This
raises the question of how people who live in multigenerational households understand
Easthope et al.
189
Figure 2. Percentage of the population born in each region who live in a multigenerational
household in Australia, 2011.
Figure 3. Percentage of total Australian population who live in a multigenerational household
born in each region, 2011.
190
Journal of Sociology 53(1)
Figure 4. Change in percentage of persons living in multigenerational and total households by
age group, Australia, 1986–2011.
their own living arrangement, and what this might tell us about changing social expectations regarding the form and role of families.
To address this question, the article focuses on the research participants’ reflections of multigenerational households as a socially accepted living arrangement,
and how this was perceived to have changed over time. Interviewees were asked
Easthope et al.
191
whether they believed that multigenerational living was accepted. Some participants also wrote about this issue in their diaries. Their reflections on this topic were
mixed.
The case of yes
Some of the research participants said they believed multigenerational households were
accepted in contemporary Australian society. Some parents who had adult children living
with them said they thought this was influenced by their more relaxed attitude to parenting compared with that of older generations:
I think that because my generation are probably a lot more easy-going and a lot more modern …
it’s a lot different than when I was the age of my daughter.… I just think we’re a lot more easygoing and open to the relationships and the lives that our kids are leading.… It’s not everybody
but I think that in general it is, it’s just a much more accepting society that we live in. (QLD422I,
40s, F, young adult/middle-aged parents, young adult never left, Anglo-Australian)3
Parents of my generation, my husband’s generation are maybe a little bit more relaxed when it
comes to their children. Like, I know that my parents.… They were very, very strict, and parents
of that generation were. (NSW222I, late 40s, F, young adult/middle-aged parents, young adult
never left, second-generation western European-Australian)
These arguments are supported by Campbell and Gilmore’s (2007: 141) questionnaire on
parenting practices with 560 Australian parents, which found evidence of ‘a socio-cultural shift from more authoritarian to more democratic child rearing practices’. Similar
findings have also been reported in other western countries (e.g. Trifan et al., 2014 in
Sweden).
Others explained that they thought multigenerational living was more widely accepted
because it is now more common, especially as a result of the financial difficulties the
younger generation faces in affording independent living (Cobb-Clark and Gørgens,
2012; Kahn et al., 2013), particularly if they are still studying:
Well incomes aren’t proportionate to the price to buy a house. You can’t do it. You cannot live
on an income and pay a mortgage, it doesn’t balance out and I have every expectation that my
girls will never leave home because it’s too expensive and they won’t be able to afford to leave
home. I’m alright with that. I think it’s more common, a lot more common especially at uni.
They’re mostly young. They’re all living at home, no one cares, no one’s oh my god I can’t
believe you’re still living at home. It’s just expected, normal. (QLD425I, 30s, F, threegeneration, older ‘child’ and grandchildren boomeranged,4 Anglo-Australian)
I think it’s not feasible for most families to move their children out straight away especially if
they’re not going into the workforce full time. If they’re doing any sort of study whether it’s an
apprenticeship or like tertiary education it’s way too expensive. (QLD432I, 20s, F, young adult/
middle-aged parent, young adult boomeranged, Anglo-Australian)
Certainly, among the survey participants, financial considerations, including the financial barriers faced by young people moving out of the family home, played a large role
192
Journal of Sociology 53(1)
Figure 5. What are the reasons that prompted multiple generations of your family to live
together? (n = 392, multiple responses allowed).
Figure 6. Please indicate how strongly you agree/disagree with the following … (n = 392).
in their reasons for living in a multigenerational household. Asked in an open-ended
question, Figure 5 presents the back-coded reasons survey respondents gave for living in
a multigenerational household.
While financial considerations were the most common reason given for living in a
multigenerational household (43%), few noted this as the sole reason (16%). The majority who said financial considerations were a driver also said their decisions to live
together were influenced by other reasons (63%), notably continuing or returning to
education, and providing care and support to other household members.
These findings are also supported by responses to another survey question that asked
people to indicate to what extent they agreed with a series of statements, presented in
Figure 6. This finding is in contrast to existing studies on multigenerational households in
other countries (Gee et al., 2003; Izuhara, 2005), which highlighted cultural backgrounds
rather than financial considerations were as a more significant influence in people’s decisions regarding the timing of first home-leaving and living in a multigenerational arrangement. While 39% of our survey respondents agreed that it was traditional in their cultural
background to live this way (Figure 6), only 6% said it was for cultural reasons that they
did so (Figure 5).
Easthope et al.
193
The case of no
While some research participants believed that multigenerational living has become
more common (or at least is spoken about more), others said it was not necessarily more
accepted. For example, this interviewee who is purchasing her home with her single
mother explained:
It’s either more common or it’s [that] you talk about it a bit more. Like you would know more
people because they will admit to living at home, but I don’t think it’s any more accepted. Like
when I tell people that I still live at home, they will look down at me, like, ‘Are you not smart
enough to move out?’ ‘You can’t afford it or you’re just living off your parents’ kind of thing.
So that’s really hard to take. (QLD441I, 20s, F, young adult/middle-aged single parent, young
adult never left, Anglo-New Zealander)
This issue was elaborated in her diary:
I think the majority of society, in Brisbane/Australia, has a set perception of adults who live
with their parents/remain in the family home, and that this hasn’t become more acceptable,
even if it seems to be more widely practised. I often get comments like ‘Aren’t you lucky’ and
‘Wish I still lived at home and didn’t have to do anything’ which I find pretty hurtful and
offensive … (QLD441-D, 20s, F, young adult/middle-aged single parent, young adult never
left, Anglo-New Zealander)
Another interviewee of an older generation made a similar observation:
I think it’s become possibly more common, but I don’t think it’s more accepted … I know a lot
of it’s got to do with culture and stuff like that, but I think in Australian society and culture – I
don’t think it’s really accepted. I think it’s definitely become more common because of the
financial pressures.… It seems to be more common because of the lack of jobs, and the high
cost in rent … (QLD430I, 30s, F, three-generation, older parent moved in, Anglo-Australian)
The case of yes, on condition that …
Other research participants clarified their response by explaining that they considered
multigenerational living to be more accepted only in some circumstances. For households
where the older generation had moved in with the younger generation, some participants
noted that outsiders would often consider their multigenerational arrangement strange.
I grew up in Vanuatu and people live as a family, your extended family.… So it wasn’t such a
big shock or a big change for me… it was just what people do … it’s a cultural thing to do.…
People or my friends or people I work with ask me do your in-laws live with you. I say yeah
they do and there’s a shocked look on their face like ‘Really, how do you cope with that?’ I say
‘Well we just get on with it.’ (QLD428I, 30s, M, three-generation, Anglo-Australian in-laws
moved in, Pacific Islander)
This was especially the case where there was not an obvious need for the older generation to live with them, such as illness or widowhood.
194
Journal of Sociology 53(1)
I think if there was an obvious medical reason like yeah mum has to live with us because she’s
got diabetes or something that would be just really obvious. But I think … it’s just not within
their sphere of recognition or whatever … you can’t just want to, you’ve got to have a clear cut
reason for it and then it’s acceptable. (QLD428I, 30s, F, three-generation, older parents moved
in, Anglo-Australian)
What I have seen as being more acceptable are when one parent often times dies or is incapacitated,
the other parent will move in with the couple to survive, because they can’t do it on their own.
(QLD423I, 60s, F, three-generation, moved in with adult ‘child’, Anglo-American)
In the case of young adults living with their parents, discussions about acceptability
of these living arrangements often focused on the life stages in which it was acceptable
for adults to remain financially dependent on their parents.
So when you’re still studying, it’s okay to live at home still. But once you’re earning – once
you’re working full time and earning your wage, I think that’s when you start looking at, OK, I
need to move, I need to purchase property, I need to actually get on with my life now. (QLD446I,
20s, F, young adult/middle-aged single parent, young adult never left, Anglo-Australian)
However, it is not the case that all adult children living with their parents are financially
dependent. In some cases, young adults were responsible for financially contributing to
the upkeep or even purchase of their home. For example, the young participant quoted
earlier who was purchasing her home with her mother explained the assumption that
younger people who live in multigenerational households are dependent on their parents
can lead to frustration.
I find the biggest drawbacks or negative aspect of living in a multigenerational household are
actually external to the household/family. Specifically, other people’s perceptions.… I find that
when I tell people – of all different ages or backgrounds – that I ‘live at home’ or with my mum,
they immediately presume that what I mean is that I live off my parent’s charity and everything is
done for me, that I’m some kind of adult-sized child. It’s super irritating, considering that my
family home has evolved into more of a share house/flat style of living arrangement with shared/
divided household duties, dinners, and purchases. What is especially annoying is that it’s rude and
inappropriate to explain that actually I own more of the house than mum, have put more money
into up-grading and maintaining it – like with kitchen renovations and a new hot water cylinder
being paid for by me – and I have paid all the rates so far. (QLD441D, 20s, F, young adult/middleaged single parent, young adult never left, Anglo-New Zealander, original emphasis)
Reflecting these findings in the US context, Niederhaus and Graham (2013: 240) note
that both older family members living with their children, and young adults living with
their parents in multigenerational households ‘suffer from the cultural stigma of “dependence”’, but in reality they may more likely be interdependent. It is also important to
recognise that there are different types of support that they can receive from family
members, including emotional support and general company.
My daughter who is studying is financially dependent upon parents to provide food and lodging,
car and pocket money. We pay all bills. I am dependent upon her to help with chores, shopping
Easthope et al.
195
Figure 7. Who usually provides the following types of support in your household? (n = 392).
and for social stimulus at home and technology trouble shooting. We have great discussions on
all issues and many laughs and a few tears. (QLD 422D, 40s, F, young adult/middle-aged
parents, young adult never left, Anglo-Australian)
I am happy for her or both of them to stay here as long as they want to. I have plenty of room
here and she is a big help both physically – helping to pick up the kids and helping with cooking
and cleaning up – and good company and emotional support. (QLD 426D, 50s, F, young adult/
middle-aged single parent, young adult never left, Anglo-Australian)
However, results from the survey suggest that in many cases the oldest generation was
more likely to provide support than the younger generations (see Figure 7). The contrast
is especially stark where direct and indirect financial supports are concerned, highlighting
the stronger financial position the oldest generation may be in within their households.
Shifting social norms about family
From the above descriptions, it seems as though there has been a shift in normalised
ideas surrounding the form and function of family as this relates to multigenerational
households. However, this does not seem to be a case of universal acceptance. The
degree to which multigenerational households are accepted or normalised relies in part
on the reasons for multigenerational living and the form and nature of the household.
The discourses around acceptability are tied up with notions of dependence and individualism. In cases where one or more household member is dependent on others – for example
when a younger family member is studying, or when an older family member is ill – then
multigenerational living is considered acceptable. However, in cases where dependence is
not essential – for example when a younger person has a job or when an older person is
physically capable of independent living – such a living arrangement is seen to be less
accepted. It appears to be especially unacceptable if a family member can financially afford
196
Journal of Sociology 53(1)
independent living but chooses to enjoy a multigenerational family life instead, as is the case
for the young woman who is sharing a mortgage with her mother. Her continuing to live
with her parent is pathologised as a ‘failure’, and this ‘failure’ is outwardly perceived by
others as financial dependence on her mother. In part this appears to result from normalised
ideas about dependence, in particular the assumption that an adult ‘child’ will be dependent
on their middle-aged parents and an elderly parent will be dependent on their adult children
within a family context, when in fact this is not necessarily the case.
These findings suggest that in the case of multigenerational living the de-institutionalisation thesis among family sociologists may be overstated. It seems that only some forms
of family flexibility – those arrangements entered into in order to support a dependent family member – are fully legitimised in the case of multigenerational households in Australia.
Others – such as a desire to embrace feelings of connectedness between generations for
their own sake – are less well understood. Put another way, structural reasons for multigenerational living (provision of care, availability of employment, affordability of housing
etc.) are legitimised, while agentic reasons (such as a desire for close personal relationships) are seen as less legitimate. For the remainder of this section, we discuss these structural reasons in more depth, in order to better understand why this shift in normalised ideas
around the form and function of the family has occurred in those cases where a member of
the family is dependent on others. If we are to understand the family as an ‘institutional
regime’ (Gilding, 2010) then we can start to examine how ideas of the family can change
in response to social, technological and economic change (Eichler, 1988).
In the interviews and diaries, people spoke in particular about the constraints of housing affordability in Sydney and Brisbane, and the influence of public policy decisions
about areas of family significance – notably higher education, child care and aged care.
Housing affordability
Many research participants noted that housing was much less affordable for young people to rent or buy than it was for older generations when they were at the same life stage.
These observations are partly supported by recent Australian research (Burke et al.,
2014) that found a decline in home ownership for younger households over the 30 years
from 1981 to 2011. However, this study also found that much of this decline occurred
during the period 1981–91 and home purchase rates have actually increased for younger
households since 1991. At the same time, the ability to achieve outright ownership at an
early age has declined, likely related to the availability of longer-term mortgaging
arrangements. Burke et al. (2014: 1) do note, however, that since 2001 ‘certain types of
younger households’, such as single-income and low to moderate-income households,
have been less able to buy a home.
Concurrently, private rental has also presented itself to be a less attractive housing
option for young adults. While the private rental market was traditionally seen as a transitional tenure for younger households, as Stone et al. (2013: 6) argue, changes in migration and international education policies, contraction of the social housing sector as well
as socio-cultural changes such as greater female workforce participation have significantly increased the demand for private rental properties. This increased demand was
accompanied by significant increases in rent, making it a less affordable option for young
adults (Stone et al., 2013: 20).
Easthope et al.
197
Many of our research participants said that concern over the affordability of housing
was a major reason why multigenerational living was common, and for some, why they
thought it was more acceptable than in the past.
People can’t afford to move out, so unless you have double income, and you really want to have
a mortgage and set up a house and a life in the same house in the suburbs somewhere, more and
more people I guess, that I meet now, are either living at home or in some arrangement where
they pay less rent or are negotiating a mortgage or something, because it’s too hard to afford to
live out of home. (QLD444I, 30s, F, young adult/older parents, young adult boomeranged,
Anglo-Australian)
I don’t think there’s stigma. I think certainly maybe 10–20 years ago, there was a bit of that
stigma but these days, I think it’s pretty accepted that the housing market is tough to get into …
and everyone that I speak to that rents wishes that they could buy but their rent is just too high
for them to save anything. (QLD446I, 20s, F, young adult/older single parent, young adult
never left, Anglo-Australian)
Public policy decisions
The reduction in government support in Australia in almost all areas of life was also
raised by some participants as a reason for multigenerational living.
I think it’s really different from when I was young. Because when I was young, I actually left
home when I was 18 or something like that. But then we had a student allowance.… I didn’t
have to work when I was studying. So it was really a different kind of situation. Now they’ve
got HECS5 debt and … they’re trying to work and they’re trying to do all these things. It’s
really a very different situation. So I think it’s financially, things are just not the same. Because
of that, society has to change as well to accommodate that. (QLD426I, 50s, F, young adult/
middle-aged single parent, young adult never left, Anglo-Australian)
One participant said that they thought this change away from government support and
towards a greater reliance on the family to provide support would pose a challenge for
many Australians.
I guess most people don’t have the skills, because it’s not really a part of Australian culture to
[live in] a multigenerational household.… So people don’t have the skills or the knowledge of
how to deal with it. So we grow up thinking that we can rely on aged care homes and the public
health system … but we can’t trust them now. Now we see that you can’t rely on these services.
(QLD444I, 30s, F, young adult/older parents, young adult boomeranged, Anglo-Australian)
Indeed, government policies that have reduced welfare for individuals and required a
greater reliance on the family for support seem to be having an impact on the way in
which these multigenerational household members think about family, and the way
in which they think their living arrangements are seen by others. Notable examples in
Australia include the deinstitutionalisation of care and a reduction in government support
for people with a disability (Wiesel and Bigby, 2015), the promotion of ageing in place
as a strategy to address aged care (Olsberg and Winters, 2005) and means testing of
198
Journal of Sociology 53(1)
government support for students based on their parents’ income until the age of 21
(Australian Government, 2015).
Conclusion
It appears that the social acceptance of multigenerational living is not complete, and that
acceptance of this living arrangement is relative to the reasons for, and circumstances of,
multigenerational living. The concept of ‘dependence’ plays an important role in determining acceptability. In those cases where it is considered appropriate for family members to be dependent on others – for example, where they require care and support due to
illness or disability, or financial support to enable them to study or save for a home
deposit – multigenerational living is considered more acceptable. Situations such as
these can be expected to arise more frequently as governments withdraw services for the
young, the elderly and the unemployed who will increasingly need to rely on their families. However, multigenerational households who are living together because they want
to, because they enjoy each other’s company, or because it makes financial sense to pool
resources as a family, do not appear to be so well understood or accepted.
This suggests that social expectations around the form and role of families are changing
relatively rapidly in response to the increased need for families to provide support as governments withdraw from this function. However, changing social norms regarding individualism and the role of the family in social relationships (beyond support for dependants)
are not changing at an equal pace and multigenerational families who live together not
because they have to, but because they want to, still face some discrimination.
In the language of the family studies debates, what we are witnessing is an increased
recognition in society of the impact of structural factors (notably housing affordability, increased educational participation into adulthood and policies encouraging
home-based care of the elderly and differently abled by the family rather than the
state) on the form and function of the family, and multigenerational family households in particular. However, at the same time, we are seeing less recognition of the
important role played by the agency of individuals who choose to live in multigenerational households for reasons other than those considered ‘acceptable’ as a result of
structural changes.
In most cases, a combination of choice and necessity (or at least practicality) drives
people’s decisions to live in a multigenerational household. However, based on the interviews and diaries of participants in this research, we can conclude that multigenerational
living appears to have become more accepted in those situations where it is seen to result
more from necessity than choice. This suggests that there is still a long way to go in
Australia before multigenerational living, in all of its forms, is normalised.
Acknowledgements
Our sincere thanks to those people who gave their time to share their experiences of multigenerational living with us through participating in the survey, interview and diary exercises. An
earlier version of this article was presented at The Australian Sociological Association
Conference in 2014, and we would like to thank colleagues in the audience for their thoughtful
comments.
Easthope et al.
199
Funding
The research reported in this paper was supported under the Australian Research Council’s
Discovery Projects funding scheme [project number DP120100956].
Notes
1.
2.
3.
4.
5.
These 392 people lived in 382 households. In 10 households, 2 members of the household
completed the survey.
The ABS defines a dependent student as a natural, adopted, step or foster offspring of a couple
or lone parent, aged 15–24 years, usually resident in the same household, who is currently
attending ‘a secondary or tertiary education institution as a full-time student and for who there
is no identified partner or child of his/her own’ (ABS, 2011).
Descriptions of the participants are given in brackets to provide context. In order of
appearance, the description states the participants’ state of residence (New South Wales or
Queensland); if the quote was from their diary (D) or interview (I); age group; gender; household form; circumstance which led to multigenerational living, and cultural background.
‘Boomeranging’ is a term for young adults leaving the parental home and then returning.
Higher Education Contribution Scheme.
References
ABS (2011) ‘Census Dictionary, 2011’, URL (consulted February 2016): http://www.abs.gov.au/
ausstats/abs@.nsf/mf/2901.0.
ABS (2014) ‘Where Do Migrants Live?’ 4102.0 Australian Social Trends, URL (consulted
February 2016): http://www.abs.gov.au/ausstats/abs@.nsf/Lookup/4102.0main+featu
res102014
Australian Government, Department of Human Services (2015) ‘Eligibility for Youth Allowance’,
URL (consulted 10 February 2016): http://www.humanservices.gov.au/customer/enablers/
centrelink/youth-allowance/eligibility-for-youth-allowance.
Burke, T., W. Stone and L. Ralston (2014) Generational Change in Home Purchase
Opportunity in Australia, Final report 232. Melbourne: Australian Housing and Urban
Research Institute.
Campbell, J. and L. Gilmore (2007) ‘Intergenerational Continuities and Discontinuities in
Parenting Styles’, Australian Journal of Psychology 59(3): 140–50.
CMF (2014) Family Generations and Financial Pressures 2014. Edinburgh: Centre for the
Modern Family.
Cobb-Clark, D. A. (2008) ‘Leaving Home: What Economics Has to Say About the Living
Arrangements of Young Australians’, Australian Economic Review 41(2): 160–76.
Cobb-Clark, D. A. and T. Gørgens (2012) ‘Parents’ Economic Support of Young-adult Children:
Do Socioeconomic Circumstances Matter?’, Melbourne Institute Working Paper 4/12.
Melbourne: Melbourne Institute of Applied Economics and Social Research.
Eichler, M. (1988) Nonsexist Research Methods: A Practical Guide. Boston: Allen and Unwin.
Farrell, B., A. VandeVusse and A. Ocobock (2012) ‘Family Change and the State of Family
Sociology’, Current Sociology 60(3): 283–301.
Flatau, P., I. James, R. Watson, G. Wood and P. H. Hendershott (2007) ‘Leaving the Parental
Home in Australia over the Generations: Evidence from the Household, Income and
Labour Dynamics in Australia (Hilda) Survey’, Journal of Population Research 24(1):
51–71.
Gee, E.M., B.A. Mitchell and A.V. Wister (2003) ‘Home Leaving Trajectories in Canada: Exploring
Cultural and Gendered Dimensions’, Canadian Studies in Population 30(2): 245–70.
200
Journal of Sociology 53(1)
Gilding, M. (2010) ‘Reflexivity over and above Convention: The New Orthodoxy in the Sociology of
Personal Life, Formerly Sociology of the Family’, British Journal of Sociology 61(4): 757–77.
Iacovou, M. (2010) ‘Leaving Home: Independence, Togetherness and Income’, Advances in Life
Course Research 15(4): 147–60.
Izuhara, M. (2005) ‘Residential Property, Cultural Practices and the “Generational Contract”
in England and Japan’, International Journal of Urban and Regional Research 29(2):
327–40.
Kahn, J.R., F. Goldscheider and J. García-Manglano (2013) ‘Growing Parental Economic Power
in Parent–Adult Child Households: Coresidence and Financial Dependency in the United
States, 1960–2010’, Demography 50(4): 1449–75.
Liefbroer, A. and F. Billari (2010) ‘Bringing Norms Back In: A Theoretical and Empirical
Discussion of Their Importance for Understanding Demographic Behaviour’, Population,
Space and Place 16(4): 287–305.
Morgan, D. (2011) Rethinking Family Practices. Hampshire: Palgrave Macmillan.
Niederhaus, S.G. and J.L. Graham (2013) All in the Family: A Practical Guide to Successful
Multigenerational Living. Lanham, MD: Taylor Trade Publishing.
Olsberg, D. and M. Winters (2005) Ageing in Place: Intergenerational and Intrafamilial Housing
Transfers and Shifts in Later Life. Final report No. 88. Melbourne: Australian Housing and
Urban Research Institute.
Stone, W., T. Burke, K. Hulse and L. Ralston (2013) Long-term Private Rental in a Changing
Australian Private Rental Sector. Final report 209. Melbourne: Australian Housing and
Urban Research Institute.
Trifan, T.A., H. Stattin and L. Tilton-Weaver (2014) ‘Have Authoritarian Parenting Practices and
Roles Changed in the Last 50 Years?’, Journal of Marriage and Family 76(4): 744–61.
Widmer, E.D. (2010) Family Configurations: A Structural Approach to Family Diversity.
Farnham: Ashgate.
Wiesel, I. and C. Bugby (2015) ‘Movement on Shifting Sands: Deinstitutionalisation and People
with Intellectual Disability in Australia, 1974–2014’, Urban Policy and Research 33(2):
178–94.
Biographical notes
Hazel Easthope is an Australian Research Council Future Fellow and Senior Research Fellow at
UNSW Australia’s City Futures Research Centre. She has background in sociology, anthropology and
human geography, and has a strong research track record in urban studies. She has a particular research
interest in residential decision-making and the intersections between mobility, identity and home.
Edgar Liu is a Research Fellow at UNSW Australia’s City Futures Research Centre. He has background in cultural and human geography, with research interests spanning a number of subdisciplines within the social sciences: family forms, public policy changes, social impacts of urban
renewal, and access to housing.
Ian Burnley is an Emeritus Professor at UNSW Australia’s City Futures Research Centre. He is a
human geographer and demographer whose research activities have focused on international
migration to Australia. He was elected as a Fellow to the Academy of the Social Sciences in
Australia in 2010.
Bruce Judd is the Director of the Australian School of Architecture and Design at UNSW Australia.
With a background in architecture, his research interests include housing design and human behaviour; medium-density housing; public housing estate renewal; urban renewal; and ageing and the
built environment.
Published April 13, 2017
The impact of multi-generational genotype imputation
strategies on imputation accuracy and subsequent genomic predictions1
M. M Judge,* D. C. Purfield,* R. D. Sleator,† and D. P. Berry*2
*Teagasc, Animal & Grassland Research and Innovation Center, Moorepark, Ireland;
and †Department of Biological Sciences, Cork Institute of Technology, Bishopstown, Ireland.
ABSTRACT: The objective of the present study was
to quantify, using simulations, the impact of successive
generations of genotype imputation on genomic predictions. The impact of using a small reference population
of true genotypes versus a larger reference population
of imputed genotypes on the accuracy of genomic predictions was also investigated. After construction of
a founder population, high-density (HD) genotypes
(n = 43,500 single nucleotide polymorphisms, SNP)
were simulated across 25 generations (n = 46,800 per
generation); a low-density genotype panel (n = 3,000
SNP) was developed from these HD genotypes, which
was then used to impute genotypes using 7 alternative
imputation strategies. Both low (0.03) and moderately
(0.35) heritable phenotypes were simulated. Direct
genomic values (DGV) were estimated using imputed
genotypes from the investigated scenarios and the accuracy of predicting the simulated true breeding values
(TBV) were expressed relative to the accuracy when
the true genotypes were used. Mean allele concordance
rate and the rate of change in mean allele concordance
per generation differed between the imputation strategies investigated. Imputation was most accurate when
the true HD genotypes of sires and 50% of the dams
of the generation being imputed were included in the
reference population; the average allele concordance
rate for this scenario across generations was 0.9707.
The strongest correlation between the TBV and DGV
of the last generation was when the reference population included sequentially imputed HD genotypes of
all previous generations, plus the true HD genotypes
of all sires of the previous generations (0.987 as efficient as when the true genotypes were used in the
reference population). With a moderate heritability,
the correlation between the TBV and the DGV using a
small reference population of accurate genotypes were,
on average, 0.07 units stronger compared to DGV
generated using a larger population of imputed genotypes. When the heritability was low, the accuracy of
genomic predictions benefited from a larger reference
population, even if SNP were imputed. The impact
on the accuracy of genomic predictions from the
accumulation of imputation errors across generations
indicates the need to routinely generate HD genotypes
on influential animals to reduce the accumulation of
imputation errors over generations.
Key words: genomic selection, heritability, imputation accuracy, single nucleotide polymorphism
© 2017 American Society of Animal Science. All rights reserved.
INTRODUCTION
The ability to predict individual animal breeding values using genome-wide information was il1Funding from the Irish Department of Agriculture, Food and
the Marine FIRM research grant GENOTRACE, the FP7 project
SEQSEL and Science Foundation Ireland grant 14/1A/2576 are
greatly appreciated.
2Corresponding author: donagh.berry@teagasc.ie
Received November 19, 2016.
Accepted January 25, 2017.
J. Anim. Sci. 2017.95:1489–1501
doi:10.2527/jas2016.1212
lustrated by Meuwissen et al. (2001). What has since
become known as genomic selection has become
the norm in animal (Hayes et al., 2009; Spelman
et al., 2013) and plant (Heffner et al., 2009) breeding programs. The accuracy of genomic prediction
is a function of, among other things, the size of the
reference population used to estimate single nucleotide polymorphism (SNP) effects (Daetwyler et al.,
2008). VanRaden et al. (2009) and Liu et al. (2011)
documented an increase in the accuracy of genomic
predictions, when the number of genotyped sires
present in the reference population increased.
1489
1490
Judge et al.
Deployment of genomic predictions on-farm requires candidate animals to be genotyped and reference animals to be genotyped and phenotyped; both of
which incur a cost for producers. Low-density genotype panels imputed to higher density are used globally across species as a strategy to reduce the cost of
implementing genomic selection programs (Berry et
al., 2014; Mulder et al., 2012). However, the potential
accumulation of imputation errors across generations
remains unknown. Previous studies in cattle (Berry
and Kearney, 2011; Weigel et al., 2010) and other species (Cleveland and Hickey, 2013) have documented
a reduction in the accuracy of genomic predictions
when imputed genotypes were used in the prediction
process. Furthermore, the impact on genomic predictions of using less accurate (i.e., imputed) genotypes,
taking cognizance that they are most likely closely related to the candidate population, is also unknown.
The objective of the present study was to quantify the impact of multiple generations of successive
imputation on the accuracy of imputed genotypes as
well as the consequential implications on genomic
predictions. Also evaluated was the impact of using a
small reference population of very accurate genotypes
versus a larger reference population of less accurate
(imputed) genotypes on genomic predictions.
MATERIALS AND METHODS
Simulation Parameters
Twenty-nine autosomes, each 1 Morgan in length,
were simulated using QMsim (Sargolzaei and Schenkel,
2009), a program which simulates large-scale genomic
information in livestock populations. A total of 1,500
evenly distributed genotyped bi-allelic markers were
simulated per chromosome, to represent after editing, the density of the Illumina Bovine50 Beadchip
(Matukumalli et al., 2009). A historical population
of 9,000 females and 1,000 males were simulated for
2,000 generations. The last generation of the historical population was assumed to represent the founder
population (i.e., generation zero) which was selected
for a further 25 generations; the founder population
consisted of 100 males and 9,000 females selected at
random from the last generation of the historical population. In total, 100 quantitative trait nucleotides (QTN)
were assumed to be randomly distributed across each
chromosome; QTN effects were sampled from a γ distribution with a shape factor of 0.4. A mutation rate per
generation for both markers and QTN of 2.5 × 10−5 was
assumed. QTN allele frequency in the first generation
of the historical population was assumed to be equal,
and the number of QTN alleles in the first generation
of the historical population was assumed to be random.
Sire and dam replacement rate was 40% and 20%, respectively, and culling and selection were both based on
estimated breeding values using pedigree-based BLUP.
Mating between males and females was assumed to be
random. Sex ratio was 52% males.
The heritability of the simulated trait was 0.35.
The phenotypic variance was assumed to equal one.
The simulated QTN were assumed to explain all the
genetic variance and phenotypes were available on all
animals. The simulation was replicated 5 times. On
completion of each replicate, the phenotypes were adjusted to also represent a heritability of 0.03; this was
done by multiplying the residual by 5 and adding this
figure to the true breeding value (TBV) to generate a
new phenotype with a lower heritability value.
Development of a Low-density
Genotype Panel for Imputation
Genotype edits were undertaken based on the SNP
data from generation 15 animals; SNP with a minor allele frequency (MAF) of < 0.02 were discarded leaving, on average, 41,856 SNP per replicate.
A low-density genotype panel (3,000 SNP) was
developed using a combination of SNP MAF and linkage disequilibrium (LD) in the generation 15 animals.
Linkage disequilibrium between SNP was generated
for each replicate separately using PLINK version
1.90 (Purcell et al., 2007). The Block method of SNP
selection, as described by Judge et al. (2016), was used
to select the 3,000 SNP from the simulated SNP. Each
chromosome was first divided into either 103 or 104
blocks with equal numbers of SNP per blocks. Single
nucleotide polymorphisms were ranked, within block,
on an index comprised of a combination of MAF plus
the average LD of that SNP with all other candidate
SNP in that block. The highest ranking SNP (i.e., high
MAF and strong average LD) was chosen within
each block. For the blocks at the start and end of each
chromosome, the second most informative SNP was
also selected as described by Judge et al. (2016). The
partial correlation of each candidate SNP in the peripheral block with all other candidate SNP, following
adjustment for the correlation with the SNP already
chosen, was calculated and the highest-ranking SNP
on an index of MAF plus the average partial correlations between that SNP and all other remaining SNP
in that block (standardized to have equal variances)
was selected (Judge et al., 2016). Imputation to high
density (i.e., the original simulated genotype panel
density) was undertaken across the whole genome simultaneously using FImpute (Sargolzaei et al., 2014).
This was done separately for each of the 5 replicates.
Multi-generational imputation strategies
Imputation Scenarios
Several alternative scenarios for generating a genomic prediction reference population of animals with
(imputed) high-density genotypes were evaluated and
are summarized in Appendix 1.
Imputation scenario 1. The true high-density
genotypes of all sires of generation 15 animals were
used to impute generation 15. For generations 16 to
20, the true high-density genotypes of all sires of the
generation being imputed, plus the true high-density
genotypes of the sires in all previous generations were
used to impute each generation individually. For generation 21, the true high-density genotypes of the sires
of that generation were used in the reference population and for generations 22 to 25 the same strategy applied in generations 16 to 20 was undertaken. No true
high-density genotypes of the dams were considered
in this scenario.
Imputation scenario 2. For generation 15, the
true high-density genotypes of all sires and a random
50% of the dams were used to impute that generation.
For generations 16 to 20, the true high-density genotypes of all sires and a random 50% of the dams of the
generation being imputed, plus the true high-density
genotypes of the sires and 50% of the dams of all previous generations were used to impute each generation
individually. For generation 21, the true high-density
genotypes of all sires and a random 50% of the dams
of that generation were used to impute that generation
and for generations 22 to 25, the same strategy undertaken for generations 16 to 20 was used.
Imputation scenario 3. The true high-density
genotypes of all sires and a random 50% of the dams
of just generation 15 were used as the reference population to impute generations 15 to 20 individually. For
generations 21 to 25, the true high-density genotypes
of all sires and a random 50% of the dams of just generation 21 were used as the reference population to
impute each generation individually.
Imputation scenario 4. For generation 15, the
true high-density genotypes of all sires and a random
50% of the dams of just generation 15 were used in
the reference population as undertaken for scenarios 2
and 3. For generations 16 to 20, the true high-density
genotypes of all sires and a random 50% of the dams
of just generation 15, plus the low-density genotypes
(i.e., 3,000 SNP) of just the immediately previous generation were used as a reference population. For generation 21 the true high-density genotypes of all sires
and a random 50% of the dams of just generation 21
were used as the reference population. For generations
22 to 25, the true high-density genotypes of all sires
and a random 50% of the dams of just generation 21,
plus the low-density genotypes of only the immedi-
1491
ately previous generation were used to impute each
subsequent generation individually.
Imputation scenario 5. For generation 15, the true
high-density genotypes of all sires and a random 50%
of the dams of just generation 15 were used as the reference population as undertaken in scenarios 2, 3, and
4. The true high-density genotypes of all sires and a
random 50% of the dams of just generation 15, plus the
low-density genotypes (i.e., 3,000 SNP) of all previous
generations (as opposed to just the immediately previous generation in imputation scenario 4), were used to
impute generations 16 to 20 individually. For generation 21, the true high-density genotypes of all sires and
a random 50% of the dams of just generation 21 were
used as the reference population. For generations 22
to 25, the true high-density genotypes of all sires and
a random 50% of the dams of just generation 21, plus
the low-density genotypes of all previous generations
were used to impute each generation individually.
Imputation scenario 6. The true high-density
genotypes of all sires and a random 50% of the dams
of just generation 15 were used to impute generation
15, as was the case for scenarios 2 to 5. For each subsequent generation, the sequentially imputed high-density genotypes of all previous generations were used to
impute all generations individually. For generations 21
to 25, the true high-density genotypes of all sires and
a random 50% of the dams of just generation 21 were
used to impute generation 21. For each subsequent
generation, the sequentially imputed high-density
genotypes of all previous generations from generation
21 were used to impute all generations individually.
Imputation scenario 7. The true high-density
genotypes of all sires and a random 50% of the dams
of just generation 15 were used to impute generation 15, as was the case for scenarios 2 to 6. For each
subsequent generation the sequentially imputed highdensity genotypes of all previous generations, plus the
true high-density genotypes of all sires of the current
and all previous generations were used to impute all
generations individually. For generation 21, the true
high-density genotypes of all sires and a random 50%
of the dams of just generation 21 were used to impute
generation 21. For generations 22 to 25, the sequentially imputed high-density genotypes of all previous
generations plus the true high-density genotypes of
the sires of the generation being imputed were used to
impute all generations individually.
The accuracy of imputation achieved per generation was determined using the allele concordance rate;
in all instances, accuracy was calculated by also including the true genotypes of animals from the lowerdensity panels to mimic a real-life scenario.
1492
Judge et al.
Table 1. Simulation characteristics for each of the 5 replicates
Simulation characteristics
SNP number
Mean minor allele frequency1
Mean linkage disequilibrium between adjacent SNP1
Heritability trait 1
Heritability trait 2
1+Figures
Replicate 1
41,744
0.29
0.21
0.33
0.04
Replicate 2
41,896
0.29
0.20
0.33
0.04
Replicate 3
41,736
0.29
0.21
0.30
0.03
Replicate 4
42,017
0.30
0.19
0.32
0.03
Replicate 5
41,888
0.29
0.20
0.32
0.03
taken from generation 15 data.
Genomic Predictions
Genetic and residual variance components for the
simulated phenotypes of generations 22, 23, and 24 (n =
140,400) were estimated using ASReml (Gilmour et
al., 2009). Traditional genetic evaluations were undertaken using MiX99 (MiX99 Development Team, 2015)
and included all animals from generations 10 to 24 (n =
655,200); variance components used were those estimated
from ASReml. The effective record contribution (ERC)
was calculated for each genotyped animal in generations
22, 23, and 24 using the reliability estimates from the genetic evaluation. The EBV of animals from generations
22, 23, and 24 were subsequently deregressed. Genomic
predictions were undertaken for each scenario separately
using SNP-BLUP (MiX99 Development Team, 2015)
which is a random regression SNP marker model where
the marker effects are estimated with BLUP and the DGV
can subsequently be calculated by summing the effects of
the alleles across all markers. All the genetic variance was
assumed to have been explained by the SNP. The fitted
genetic and residual variances were those estimated using ASReml (Gilmour et al., 2009). The dependent variable of the deregressed EBV was weighted by the ERC.
Genomic predictions for all animals in generation 25 (n =
46,800 animals) were subsequently estimated as the sum
of the product of the estimated SNP effects for each scenario times the genotype of the individual.
The correlation between the TBV of each animal
generated in the simulation and the predicted direct genomic values (DGV) from the SNP-BLUP was used as
a measure of accuracy of prediction. A gold-standard genomic prediction for each of the 5 simulation replicates
was generated for the animals in generation 25. This was
calculated using the true high-density genotypes of animals in generations 22, 23, and 24 to estimate the SNP
effects which were subsequently applied to the true highdensity genotypes of generation 25 animals. The accuracy of prediction of the different scenarios evaluated were
expressed relative this gold-standard prediction.
Finally, the effect of using a small reference population of accurate (i.e., true) genotypes versus a larger reference population of less accurate (i.e., imputed) genotypes
on the accuracy of genomic predictions was quantified.
Genomic predictions were estimated for all animals in
generation 25 (n = 46,800) based on SNP effects generated using either of two reference population scenarios:
1) The genomic prediction reference population representing a small population of highly accurate genotypes consisted of the true genotypes of all sires
(n = 180) and either 0%, 10% (n = 13,208), 20% (n =
24,367), 30% (n = 33,765), 40% (n = 41,632), or 50%
(n = 48,124) of the true genotypes of the dams of generations 22 to 24. The SNP effects generated were applied to the true genotypes of generation 25 animals.
2) The genomic prediction reference population representing a large population of less accurate genotypes
consisted of the imputed genotypes of all animals of
generations 22 to 24 (n = 140,400); this was done
individually for each of the 7 reference imputation
scenarios. The SNP effects generated were applied to
both the true genotypes and the imputed genotypes
for the scenario under investigation of generation 25
animals separately.
Genomic predictions were also undertaken for
each of the scenarios described above using a trait
with a heritability of approximately 0.03. This was undertaken separately for all replicates.
RESULTS
Results shown are the average of all 5 replicates,
unless otherwise stated. Table 1 presents an overview of
the simulation characteristics for each replicate including the number of SNP after edits, the average MAF
per SNP in generation 15, and the mean pairwise linkage disequilibrium (LD) of adjacent SNP in generation
15. Inbreeding levels averaged at 0.0528 in generation
15 across all replicates and increased by, on average,
0.0065 per generation (R2 = 0.9958). Excluding animals with no progeny, the median number of progeny
for males was 491 while the modal number of progeny
for females was 6. The effective population size on average was, 78 across all replicates.
Imputation Accuracy
The construction of the reference population for
the imputation of generation 15 was identical across
Multi-generational imputation strategies
1493
Figure 1. Mean animal allele concordance rates for each generation (15 to 25) in each reference population scenario; scenario 11 (solid black line;
square markers) scenario 21 (solid black line), scenario 31 (solid gray line), scenario 41 (solid gray line; square markers), scenario 51 (broken gray line),
scenario 61 (double black line), and scenario 71 (broken black line). Error bars represent the standard error associated with the mean across all 5 replicates.
1A description of each scenario is in Appendix 1.
scenarios 2 to 7 and, as such, the accuracy of imputation was identical for these scenarios (the same
was true for generation 21). Allele concordance rate
achieved per generation differed substantially among
the different scenarios investigated; mean concordance
rate per generation within scenario was, however, relatively similar for each of the 5 replicates (Fig. 1).
When the reference population for imputation included just the true genotypes of the sires of the generation being imputed, plus the true genotypes of the
sires of all previous generations (i.e., scenario 1), the
allele concordance rate was the least accurate of all scenarios for generations 15, 16, and 17 (Fig. 1). Mean animal allele concordance rate improved for this scenario
with each generation; allele concordance rate increased
by 5% (4.3% units) from generation 15 to 20, and 3%
(2.6% units) from generation 21 to 25. When the true
genotypes of all sires and 50% of the dams of the generation being imputed (and all previous generations)
were also included in the reference population (i.e., scenario 2), the mean animal allele concordance rate was,
on average, 7% (i.e., 6% units) more accurate across
all generations compared to when the genotypes of just
all the sires alone were included in the reference population. For scenario 2, as generation number increased,
allele concordance rate also increased but just for 2 gen-
erations after when the true genotypes of the animals
started to be collated for the first time (i.e., generation
15 and 20) and plateaued thereafter; regardless of this,
mean animal allele concordance rate for this scenario
was the most accurate of all constructed reference populations evaluated in the present study.
For scenarios 3 to 7, the mean animal allele concordance rate declined as generation number increased
(Fig. 1). Scenarios 3 to 7 all shared the same reference population for the imputation of generation 15
and generation 21 but differed thereafter; the density
(low or high) and quality (true or imputed) of the genotypes differed among scenario. The rate at which the
imputation accuracy declined per generation varied by
reference population scenario evaluated. The greatest
decline in accuracy with increasing generation was
when the true genotypes of all the sires and 50% of
the dams of generation 15 only were used as a reference population to impute generations 15 to 20 (i.e.,
scenario 3); allele concordance rate declined in total
by 11.8% from generations 15 to 20. The addition of
the low-density genotypes (i.e., 3,000 SNP) of the immediately previous generation to the generation being
imputed, along with the true genotypes of all the sires
and 50% of the dams of generation 15 (i.e., scenario 4),
reduced the rate of deterioration in allele concordance
1494
Judge et al.
Figure 2. Mean allele concordance rate per SNP for different minor allele frequency bins in each reference population scenario1; MAF ≤ 0.05 diagonal black lines; > 0.05, ≤ 0.10 dark gray bars; > 0.10, ≤ 0.20 white bars with black dots; > 0.20, ≤ 0.30 vertical black lines; > 0.30, ≤ 0.40 light gray bars; >
0.40 white bar. Error bars represent the standard deviation associated with the mean across all 5 replicates. 1A description of each scenario is in Appendix 1.
rate over the 5 generations (i.e., generation 15 to 20)
from 11.8% to 10.6%. Furthermore, when the lowdensity genotypes of all the previous generations (i.e.,
scenario 5), as opposed to just the immediately previous generation (i.e., scenario 4), were included in the
reference population, the rate at which the allele concordance deteriorated was just 9.1% from generations
15 to 20. When the reference population for imputation included the sequentially imputed high-density
genotypes of all previous generations (i.e., scenario
6) the decline of imputation accuracy was substantially reduced to 2% over the 5-generation period (i.e.,
generations 15 to 20); it was further reduced to 0.7%
when the true high-density genotypes of the sires of
the generation being imputed were also included in
the reference population (scenario 7).
For scenarios 3 to 7, the reintroduction of true
genotypes to the reference population in generation 21
resulted in an abrupt improvement in imputation accuracy; mean animal allele concordance rates reached
similar values to those observed in generation 15 for
the respective scenario. From generations 21 to 25,
reference population scenarios 3 to 7 followed the
same trend and rate of deterioration in imputation accuracy with an increase in generation number, as that
observed in generations 15 to 20.
The mean allele concordance per SNP of generation
25 across all reference population scenarios for 6 different MAF bins is shown in Fig. 2; allele concordance
rate decreased almost linearly with an increase in SNP
MAF (with the exception of SNP with a MAF > 0.4).
Genomic Predictions
The extent of the variability in the correlation between the TBV and the predicted DGV of generation
25, observed across the 5 replicates, was dependent
on reference population scenario (Fig. 3). The mean
correlation across all 5 replicates between the TBV of
generation 25 and the estimated DGV for each reference population scenario are also shown in Fig. 3.
The scenario of imputation yielding the most accurate
DGV prediction (i.e., 0.987 as efficient as when the
true genotypes were used in the reference and validation population) was when the reference population for imputation included the sequentially imputed
high-density genotypes of all previous generations,
along with the true high-density genotypes of all sires
(i.e., scenario 7; Fig. 3). The next most accurate genomic prediction in generation 25 was achieved when
just the sequentially imputed genotypes of all previous
generations were included in the reference population
Multi-generational imputation strategies
1495
Figure 3. The correlation between the true breeding value (TBV) of generation 25 and the direct genomic value (DGV; relative to the gold-standard
prediction) using imputed genotypes from each of the 7 reference population scenarios1. Generations 22, 23, and 24 were used in the generation of DGVs.
Error bars represent the standard deviation associated with the mean across all 5 replicates. 1A description of each scenario is in Appendix 1.
as depicted by scenario 6 (i.e., the true high-density
genotypes of the sires were not included in the reference population as was the case for scenario 7). The
weakest correlation between the TBV and DGV in
generation 25 animals was when the reference population was generated using the imputation reference
population constituted of only the true genotypes of all
sires and 50% of dams of generation 21 (scenario 3).
Effect on Reference Population Size and Accuracy
of Genotypes on Genomic Predictions
Unless otherwise stated, the DGV were generated
by applying the SNP effects of each reference population scenario to the respective imputed genotypes of
generation 25. When the heritability of the simulated
trait was 0.35, the correlation between TBV and DGV
generated when the reference population included the
true high-density genotypes of all sires and at least 10%
of the dams of generations 22 to 24 (i.e., the small but
very accurate reference population) was stronger than
when the gold-standard reference population (i.e., the
true genotypes of all animals in generations 22 to 24)
was used (0.5692 versus 0.5521; Fig. 4). There was
therefore no obvious benefit of having > 10% of genotyped dams in the reference population when the heri-
tability was 0.35. However, when the heritability of the
trait in question was just 0.03, the correlation between
the TBV of generation 25 and the DGV generated using the gold-standard reference population was stronger
than the correlations generated using any of the small
(very accurate) reference populations (Fig. 5).
Regardless of the heritability of the trait, as the
percentage of dams with true high-density genotypes
included in the genomic prediction training population
increased, the correlation between the TBV and the
DGV of generation 25 animals strengthened (Fig. 4
and Fig. 5); this effect, however, plateaued more rapidly when the heritability was 0.03. When the heritability of the trait was 0.35, DGV generated from any of
the 7 large populations of imputed genotypes in generations 22 to 24 were more strongly correlated (on
average, 0.0748 stronger across all scenarios) with the
TBV of generation 25, compared with DGV generated
from the true genotypes of just the sires of generations
22 to 24 (Fig. 4). However, when the true genotypes of
10% of the dams of generations 22 to 24, plus the true
genotypes of the sires of generations 22 to 24 were included in the reference population, the correlation between the TBV and the DGV was stronger than any of
correlations generated from the large reference populations of imputed genotypes (0.0560 stronger, on
1496
Judge et al.
Figure 4. The correlation between the true breeding value (TBV) and the direct genomic value (DGV) of generation 25 when genomic predictions were
undertaken using the gold-standard reference population (gray dotted line), a small reference population of highly accurate (true) genotypes (double black line;
square markers), or a large population of less accurate (imputed) genotypes when imputation was undertaken using reference scenario 11 (solid black line), scenario
21 (solid black line; square marker), scenario 31 ( solid gray line), scenario 41 (solid gray line; square markers), scenario 51 (gray line; circle markers), scenario
61 (broken black line), and scenario 71 (broken black line; square markers). The heritability of the trait was 0.35 1A description of each scenario is in Appendix 1.
average, regardless of reference population scenario).
When the heritability of the trait was 0.03, however,
the correlation between the TBV and the DGV generated using a large reference population of imputed
genotypes (when the true high density genotypes of
the sires and 50% of the dams [scenario 2], or the sequentially imputed genotypes [with or without the true
high-density genotypes of the sires; scenarios 6 and 7]
were included in the reference population for imputation) was on average, 0.03 units stronger than any
of the correlations using a small reference population
consisting of the true genotypes of the sires and up to
50% of the dams of generations 22 to 25 (Fig. 5).
Finally, when the SNP effects generated using
each of the 7 large reference population scenarios
were applied to the true genotypes of generation 25
(as opposed to the imputed genotypes) the strength of
the correlation of the TBV with the DGV varied by
scenario. When using the true genotypes of generation
25, the correlation between TBV and DGV was stronger for scenarios 1 to 5 (0.055 on average across scenarios), but was weaker for scenarios 6 and 7 (0.021
on average across replicates).
DISCUSSION
Imputation of genotypes from low to high density
is now accepted practice in both genome-wide association studies (Meredith et al., 2013; Purfield et al., 2015;
Santana et al., 2014) and genomic evaluations (Berry
and Kearney, 2011; Mulder et al., 2012; Weigel et al.,
2010). Most studies to date have simply established the
impact of imputation on genomic predictions (Berry and
Kearney, 2011; Mulder et al., 2012; Weigel et al., 2010).
However, the impact of imputed genotypes originating
from previous generations with phenotypes on the accuracy of genomic predictions remains unclear. Therefore,
the motivation of the present study was to determine,
using simulations, the impact of multiple generations
of successive imputation on the accuracy of genotype
imputation and the consequential impact on the accuracy
of genomic predictions. Any potential compounding effect of imputation errors across generations is particularly relevant given the shortening of generation intervals with the implementation of whole-genome enabled
selection (Schaeffer, 2006). Moreover, given that the
greatest resource constraint in the development of reference populations for estimating SNP effects is likely to
Multi-generational imputation strategies
1497
Figure 5. The correlation between the true breeding value (TBV) and the direct genomic value (DGV) of generation 25 when genomic predictions were undertaken using the gold-standard reference population (gray dotted line), a small reference population of highly accurate (true) genotypes (double black line; square
markers) or a large population of less accurate (imputed) genotypes when imputation was undertaken using reference scenario 11 (solid black line), scenario 21
(solid black line; square marker), scenario 31 (solid gray line), scenario 41 (solid gray line; square markers), scenario 51 (gray line; circle markers), scenario 61
(broken black line), and scenario 71 (broken black line; square markers). The heritability of the trait was 0.03. 1A description of each scenario is in Appendix 1.
be financial, the impact of using a small reference population of accurate genotypes (i.e., higher cost) versus a
larger reference population of less accurate (imputed)
genotypes (i.e., lower cost) on genomic predictions
was also determined. Although genotype panels are
increasing in density (especially in the cattle industry)
the necessity to use genome-wide imputation in current
routine genomic evaluations is reducing. However, it is
still not financially feasible to sequence all animals, and
consequently the results from the present study are still
pertinent for imputation to higher-density genotypes.
The practical implications of the results from the
present study suggest that:
1. Imputation errors accumulate across generations
but the realized rate of error is dependent on the
composition of the imputation reference population and the imputation strategy employed.
2. The accumulation of imputation errors across generations is reduced when A) the true high-density
genotypes of the sires and 50% of the dams of the
generation being imputed are included in the reference population, and B) the direct progeny of animals genotyped using the high-density genotype
are first imputed to higher density prior to inclusion
in the reference population for imputation of subsequent generations (i.e., scenario 6 and scenario 7).
3. Including the low-density genotypes of previous generations does aid imputation but this strategy is not a
good substitute for sequentially imputing generations.
4. For lower heritability traits (i.e., 0.03), genomic predictions are more accurate when a large population
of animals with imputed genotypes are used in the
reference population to generate SNP effects compared to a smaller population of animals with true
genotypes. However, for traits with higher heritabilities (i.e., 0.35) genomic predictions are more accurate when the reference population includes a smaller population with accurate (i.e., true) genotypes.
As with any simulation, the results reported herein
are only reflective of (a) the population parameters (e.g.,
linkage disequilibrium patterns among SNP, effective
population size, selection, and mating strategies) simulated, (b) the imputation software and algorithms used, and
(c) the genomic prediction algorithm used. The autosomal genome simulated in the present study was modeled
on a within-breed cattle population using the commonly
used Illumina Bovin...
Purchase answer to see full
attachment