The limitations of understanding multi-partner fertility in Australia.

Gray, Edith ; Evans, Ann

INTRODUCTION

Over the past few decades there have been enormous changes in family formation, structures and processes. In Australia, and many other western-industrialised countries, there has been a large increase in the proportion of people who cohabit outside of marriage, divorce is commonplace, and exnuptial births have increased rapidly over the last twenty years. Childbearing certainly no longer occurs just within marriage; in fact 33 per cent of births occurring in 2006 were exnuptial. (1) Such exnuptial births may take place in cohabiting relationships, to single persons, or after relationship dissolution or death of a partner. With the increase in non-marital fertility, partnership dissolution, and repartnering, comes an increased risk or chance of having children with multiple partners, a concept known as multipartner fertility.

Despite the significant implications that multipartner fertility has for both the parents and the children involved, relatively little is known about this phenomenon. Most of what is known about the issue comes from the United States, but even in the U.S. it has been difficult to establish the level and pattern of multipartner fertility because research is predominately based on surveys of low-income or disadvantaged populations that are not representative of the wider national population. (2)

This paper examines the data issues involved in measuring multipartner fertility, then, using the Household, Income and Labour Dynamics in Australia (HILDA) survey, (3) we explore patterns of multipartner fertility in Australia focusing particularly on generational changes over time. We conclude with a summary of the difficulties of studying multipartner fertility in Australia and discuss future directions in data collection.

WHAT IS KNOWN ABOUT MULTIPARTNER FERTILITY?

Multipartner fertility can be defined as having children with more than one partner. It is important at the outset to distinguish the measurement of multipartner fertility from the measurement of step families or blended families. The measurement of step families typically occurs at the household level, although it could be at the level of the individual, for example, 'Are you a step-mother/father?' In the U.S. the measurement of step families tends to be limited to narrowly defined versions of step families, specifically those that include only legally (re)married families after divorce. (4) This is not the case in Australia, and a recent paper demonstrates the ability to measure the type of family in which children live, regardless of the parent's marital status. (5) Using census data it is now possible to determine whether the household consists of an 'intact', 'step' or 'blended' family, that is, a family with children from past relationships and children from the current union.

But measuring step or blended families is not the same as the measurement of multipartner fertility. The measurement of step families within households can not tell us: whether there are other children of the parents who live outside the home, or, if the parents do have children who live elsewhere, whether these children are from the current relationship or another relationship. In the case of parents with children from past relationships, there is no way to determine whether the children all come from just one past relationship or a number of past relationships.

Unlike the more established field of research on step families, the study of multipartner fertility is still in its infancy. Research in the U.S. has shed some light on the prevalence, correlates and consequences of multipartner fertility. However, as most research is based on non-representative populations there are limitations in terms of the generalisability to the wider population in the U.S., let alone other countries such as Australia.

The research that comes from the U.S. shows that there are varying estimates of the prevalence of multipartner fertility depending on the sources of data used. Estimates produced from surveys of disadvantaged or low-income populations tend to show quite high levels of multipartner fertility. For example using the Fragile Families and Child Well-being Survey, Mincy found that over a third of the mothers and fathers in the sample exhibited multipartner fertility. (6) But importantly this survey is limited to new unmarried parents in 20 U.S. cities, and therefore it is highly unrepresentative of the wider population. (7) Other research based on administrative data from welfare recipients in Wisconsin, also found that around 30 per cent of women were confirmed to have had children with at least two partners. (8)

One recent study which did analyse a representative sample of American men from the National Survey of Family Growth (NSFG), found a much lower level of multipartner fertility; around eight per cent of men aged 15 to 44 had children with more than one partner. (9) The authors also examined sub groups of the population, and found that older age groups, being of black non-Hispanic background and having a low income were associated with higher levels of multipartner fertility. For example among low-income black men aged 35 to 44, over one third reported having children with at least two women, and 16 per cent reported multipartner fertility with three or more women. In the U.S. ethnicity appears to be one of the strongest correlates of multipartner fertility with black non-Hispanic parents being much more likely to have children with multiple partners. Other important correlates include a young age at first birth for mothers and the marital status of the parents. In the Fragile Families study, unmarried parents were also significantly more likely to have experienced multipartner fertility compared to the control group who were married. (10)

IMPLICATIONS OF MULTIPARTNER FERTILITY

There are significant implications of multipartner fertility for both children and adults. Children whose parents have children in a number of relationships are at greater risk of experiencing a lower quality and quantity of parental investment. This investment includes less time and financial resources from parents. Further, there are indications that multipartner fertility has negative consequence for all children involved, whether they are children bom in an earlier relationship or new children born to a current relationship. Investment in new biological children appears to be undermined if a father has previous children with another partner, illustrating the responsibilities the father has for children from past relationships. (11)

However, the situation may be exacerbated for children of non-residential parents, where the non-residential parent has 'new' biological children. In terms of the previous children, some U.S. research finds evidence that fathers swap families and reduce child support payments to nonresidential children when they set up a different family with new biological children. (12) Other research has indicated that visits to non-residential children may be hampered on a practical level, as having children living in several different places increases the transportation cost and time burden on parents. (13)

In addition to complicating the investment that parents make in terms of time and money with children from previous and current relationships, having children from multiple partners has implications for both children and parents who have to negotiate the complexities of family instability. (14) Children from previous relationships are faced with significant adjustment issues when a parent's partner joins the household and also when new children are born to that relationship.

The implications for parents are also substantial. For parents who do not live with their children, there is evidence that they will not experience many of the benefits of parenthood, such as emotional benefits and social connections. (15) There are also different implications for parents based on whether they are a coresidential parent or non-coresidential parent. Coresidential parents with children from different relationships have the complexities associated with living in a blended family. Non-coresidential parents have the complexities associated with having children living in different households. In terms of relationship formation, Stewart et al. have found that having children from a previous relationship has the effect of reducing the chance of marriage. (16)

In sum, there are additional complexities associated with multipartner childbearing, both for children and parents, yet we know little about the level of this phenomenon in Australia. This is mainly due to data availability. The following section outlines issues associated with data adequacy and measurement of multipartner fertility in Australia.

DATA ISSUES

In Australia even basic information about the prevalence of multipartner fertility is not available, and this is primarily because of the lack of adequate data. We could reasonably expect that information on the timing of previous births, and the relationship context of the previous births, might be obtained in birth registration data, but there is a great deal of inconsistency between states and territories in what is collected about births in other relationships. (17) Similarly in the census, very limited information is collected on children and the relationship context in which they were born. In the latest 2006 census there was a question on family blending (18) but, as previously discussed, this information does not provide information for studying multipartner fertility as information on children outside the household is not collected. This leads to the situation where information about childbearing must be collected from large sample surveys.

For this study we use data from the HILDA survey, a nationally representative longitudinal study. The first wave of HILDA was collected in 2001, surveying over 13,000 individuals aged 15 and over. Respondents are surveyed annually and this study uses the latest available wave, collected in 2006. While HILDA collects information at each wave which can be used to examine transitions over time prospectively, in this paper we use the retrospective information available on childbearing and relationship histories. We focus on the retrospective information so as to be able to compare the situations of different generations and examine changes in multipartner fertility over time.

While HILDA contains detailed information on prior fertility and marriage histories, it contains very little information on cohabitation histories of respondents. The information collected on cohabitation includes the timing of a respondent's first cohabitation, the total number of cohabitations that the respondent has had, and whether marriages are preceded by cohabitation. Unfortunately this means that it is not always possible to tell whether or not a child born outside of marriage was born within a cohabiting relationship or not, making it difficult to create an accurate picture of multipartner fertility. (19) We note previous research that indicates an important distinction between marital and non-marital fertility, with fathers being more likely to have multipartner fertility across a series of non-marital partnerships than marital partnerships. (20) So the distinction between marriage and non-marriage is important.

To be able to accurately identify all incidences of multipartner fertility, the ideal data would contain the full relationship histories of respondents, including the start and end date of all cohabitations and marriages as well as the timing of each childbirth. This would enable us to accurately match each child to the relationship context in which it was born, and thereby identify respondents who have had children born to multiple partners. However, there is enough information on marital histories and birth histories in the HILDA data to provide instructive information on the relationship context of childbearing.

DATA AND ASSUMPTIONS

The sixth wave of HILDA has a total sample size was 12,905. To compare multipartner fertility across the generations, we restricted the sample to individuals aged 38 years and over in order to exclude those who had not yet finished childbearing. Using a 38 year cut-off is appropriate for women as fertility past this age is negligible. However for men it is likely that, particularly for the youngest cohort, individuals may still have more children in the future. Of the 7,617 respondents who were aged 38 and over, we further restricted the analysis to the 5,672 individuals with two or more children, that is, those people who could have had their children in different relationships. As shown in Table 1, of respondents aged 38 and over, 71 per cent of men and 77 per cent of women had two children or more.

Table 1: Children ever born by sex and birth cohort of respondent (a)

 1905-1929 1930-1944 1945-1959

Children ever born Freq. per cent Freq. per cent Freq. per cent

Male respondents

0 27 9 82 10 240 16
1 31 10 64 8 169 11
2 81 27 271 32 544 36
3 74 24 239 28 367 24
4 53 17 110 13 127 8
5+ 39 13 78 9 76 5
Total 305 100 844 100 1,523 100

Female respondents

0 35 8 60 7 166 11
1 59 14 66 8 176 11
2 119 29 243 29 623 40
3 82 20 233 28 384 25
4 73 18 130 16 135 9
5+ 46 11 106 13 62 4
Total 414 100 838 100 1,546 100

Total respondents with 2+ children

Male respondents 247 81 698 83 1,114 73
Female respondents 320 77 712 85 1,204 78

 1960-1968 Total

Children ever born Freq. per cent Freq. per cent

Male respondents

0 314 29 662 18
1 143 13 406 11
2 348 32 1,244 33
3 183 17 863 23
4 54 5 344 9
5+ 35 3 228 6
Total 1,077 100 3,747 100

Female respondents

0 150 14 411 11
1 165 15 466 12
2 418 39 1,403 36
3 212 20 911 24
4 91 8 429 11
5+ 36 3 250 6
Total 1,072 100 3,870 100

Total respondents with 2 children

Male respondents 620 58 2,679 71
Female respondents 757 71 2,993 77

Source: HILDA Survey, 2006

Note: (a) respondents aged 38+
Total N = 7617
Total N with children = 5672

In order to examine generational changes over time, four birth cohorts are compared: those born in 1905 to 1929, 1930 to 1944, 1945 to 1959 and 1960 to 1968. We weight the data using a person population weight that is rescaled to sum to the number of responding persons.

The procedure for identifying instances of multipartner fertility in the sample involves classifying each child of a respondent into the relationship context in which it was born. Due to the limited data on cohabitation, children could only be classified as having been born to: (1) a marriage; (2) a cohabitation that resulted in a marriage (pre-marital cohabitation); or, (3) outside of marriage. Based on this information three groups of people were identified:

1. The first category contains those who had not experienced multipartner fertility as they had had all their children within one marriage. Children are counted as having been born to a particular marriage if they were born after the start of the marriage, and before the end of the marriage if the marriage had ended. Any period of cohabitation which can be identified as pre-marital is counted as part of the duration of the marriage concerned. Any child who was born in the same year as the marriage is also counted as having been born in that marriage.

2. The second category identifies the people with multipartner fertility. These are the respondents who had their children across more than one marriage, as well as those who had their children outside of marriage and in marriage. As it was not possible to identify whether the partner involved in the exnuptial and nuptial births was the same person, there is a possibility that this category is overestimating multipartner fertility to individuals who had had all their children with the same partner, but with some children born before and some after the marriage. However since the period of pre-marital cohabitation is included as belonging to the marriage duration, this would only occur if the couple had children but were not cohabiting, and later married and had one or more additional children. (21)

3. The third category contains individuals who had had all their children outside of any marriage. It is likely that a high percentage of these childbirths occurred in cohabiting relationships, to the same partner. However with no additional information on cohabitation we are unable to make assumptions as to whether individuals in this category have in fact experienced multipartner fertility or not. If there is substantial multipartner fertility across non-marital relationships as found in the U.S. (22) the analysis will under represent multipartner fertility.

RESULTS

The dominant trend evident in the data is one of a stable majority having all their children within the same marriage, but that an increasing number in the younger cohorts have experienced multipartner fertility (see Table 2). This section reports on our analysis of men and women with two or more children only. Overall 85 per cent of children born to both men and women were born to the same marriage, but this figure did vary across cohorts. For example over 90 per cent of women born in the oldest cohort (1905 to 1929) had their children in the same marriage compared to 76 per cent of women in the 1960 to 1968 birth cohort. With regard to multipartner fertility the first two cohorts, those born in 1905 to 1929 and 1930 to 1944, exhibit fairly stable behaviour with around eight or nine per cent of the children being born to different relationships, with a similar pattern for men and women. (23)

Table 2: Relationship context of births for respondents with two or
more children, by cohort

Relationship 1905-1929 1930-1944 1945-1959
context of births Freq. % Freq. % Freq. %

Male respondents

All within one 223 90 630 90 939 84
marriage

All outside 2 1 9 1 31 3
marriage

Across more than 23 9 59 8 144 13
one relationship

Total 248 100 698 100 1,114 100

Female respondents

All within one 293 92 643 90 1,042 87
marriage

All outside 0 0 9 1 41 3
marriage

Across more than 27 8 60 8 121 10
one relationship

Total 320 100 712 100 1,204 100

Relationship context of births 1960-1968 Total
 Freq. % Freq. %

Male respondents

All within one marriage 489 79 2,281 85

All outside marriage 53 9 95 4

Across more than one relationship 77 12 303 11

Total 619 100 2,679 100

Female respondents

All within one marriage 572 76 2,550 85

All outside marriage 61 8 111 4

Across more than one relationship 124 16 332 11

Total 757 100 2,993 100

Source: HILDA survey, 2006

The percentage of births not born to the same marital relationship appears to have increased over time, particularly in the youngest cohort of females and in the two youngest cohorts of males. For men a steady decline in the percentage of births born in the same marriage is evident, falling from 90 per cent in the oldest cohorts to 84 per cent for those born from 1945 to 1959 to 79 per cent of those born in 1960 to 1968. This appears to be mostly due to the increase in multipartner fertility for the 1945 to 1959 cohort, and then to an increased percentage of men who have all their children outside of marriage for the youngest cohort. The results for the youngest cohort of males should be interpreted with some caution however as they may not have completed childbearing yet.

For females the patterns are similar, but the real change for women seems to have occurred only for the latest generation (1960 to 1968) where the percentage of women having children in at least two different relationships increased to 16 per cent, and where a substantial percentage also had all their children outside of marriage.

DISCUSSION

Multipartner fertility is a growing phenomenon in Australia and other western-industrialised countries. Yet information about the extent of multipartner fertility is severely inadequate. The major barrier to understanding the level of multipartner fertility in Australia is data limitations. Relationship and birth histories are not routinely collected in national statistics or vital registration. Survey data which collects these histories are the closest source available, but as demonstrated, even these have limitations.

The most likely way social researchers will be able to investigate multipartner fertility is through surveys which collect information prospectively over time. For people who are currently in the childbearing stage of the life course, future research will be able to use HILDA. Such prospective information on all childbirths and relationship transitions will make it possible to create a more accurate picture of the prevalence of multipartner fertility, and to extend the research by exploring the correlates and consequences of multipartner fertility in Australia.

However, the data needed to compare the experience of past generations are necessarily retrospective. Measuring multipartner fertility across generations is only possible if details on all relationships individuals have experienced throughout their lifetime, including cohabitations and marriages, are collected, as well as information on when each child was born. If these data were available each child could be accurately matched to the relationship within which it was born.

The current paper provides some understanding of the patterns of multipartner childbearing in Australia over time. The lack of information on cohabitation was an important limitation in this study and it has implications for the interpretation of results. For the most recent generation around eight per cent of people who have had two or more births have had them outside of a marriage. Unfortunately we do not know whether these births occurred in a cohabiting relationship, although given that cohabitation is fairly well accepted in Australia this is likely. Further, we can not tell whether these births outside of marriage are to the same partner or not. If multipartner fertility often occurs to a series of non-marital relationships as is found in the U.S., then at least some of the eight per cent would contribute to multipartner fertility in Australia. But we just cannot tell.

These results do however demonstrate that a substantial percentage of people in the youngest generation are having children in more than one marital relationship. An estimated 12 per cent of men, and 16 per cent of women, who were born between 1960 and 1968 and who have two or more children, have had children in more than one marriage.

The structure of families and family processes is changing dramatically over time, with implications for children, parents, service providers, and collectors of social statistics.

Acknowledgements

The authors gratefully acknowledge the Australian Research Council (DP0772544) for its support of this research. This paper uses unit record data from the Household, Income and Labour Dynamics in Australia (HILDA) Survey. The HILDA Survey was initiated and is funded by the Australian Government through the Department of Families, Housing, Community Services and Indigenous Affairs (FaHCSIA) and is managed by the Melbourne Institute of Applied Economic and Social Research (MIAESR). The findings and views reported in this paper, however, are those of the authors and should not be attributed to either FaHCSIA or the MIAESR. We would also like to thank Anna Reimondos for research assistance.

References

(1) Australian Bureau of Statistics (ABS), Births, Australia 2006, Catalogue no. 3301.0, ABS, Canberra, 2007

(2) For example, R.B. Mincy, 'Who should marry whom?: multiple partner fertility among new parents', Princeton University, Woodrow Wilson School of Public and International Affairs, Center for Research on Child Wellbeing Working Paper 2002-03-FF 2002 <http://crew.princeton.edu/publications/publications.asp> accessed 25 July 2008; M.J. Carlson and F.F. Furstenberg, 'The prevalence and correlates of multipartnered fertility among urban U.S. parents', Journal of Marriage and Family, vol. 68, no. 3,2006, pp. 718-732; D.R. Meyer, M. Cancian and S.T. Cook, 'Multiple-partner fertility: Incidence and implications for child support policy', Social Science Review, vol. 79, no. 4, 2005, pp. 577-601; See K. B. Guzzo and F. F. Furstenberg, 'Multipartnered fertility among American men', Demography, vol. 44, no. 3, 2007, pp. 583-601 for research based on a representative sample.

(3) Melbourne Institute of Applied Economic and Social Research (MIAESR), HILDA Survey Annual Report 2006, 2007, University of Melbourne, Melbourne

(4) M.J. Carlson and F.F. Furstenberg, The consequences of multi-partner fertility for parental involvement and relationships, 2007, Princeton University, woodrow Wilson School of Public and International Affairs, Center for Research on Child Wellbeing Working Paper 2006-28-FF <http://crew.princeton.edu/publications/publications.asp> accessed 23 July 2008

(5) J. Forster-Jones, 'Family diversification in Australia--the increasing share of blended and step families', People and Place, vol. 15, no. 4, 2007, pp. 9-19

(6) Mincy, 2002, op. cit.

(7) The Fragile Families and Child Well-being Survey also includes information on a comparison group of 1,186 children born to married parents. Carlson and Furstenberg, 2006, op. cit., found that when unmarried and married parents are compared the prevalence of multipartnered fertility declines to around one quarter of women.

(8) Meyer, et al., 2005, op. cit.

(9) Guzzo and Furstenberg, 2007, op. cit.

(10) ibid.

(11) ibid.

(12) W.D. Manning and P.J. Smock, '"Swapping" families: Serial parenting and economic support for children', Journal of Marriage and the Family, vol. 62, no. 1, 2000, pp. 111-122

(13) Carlson and Furstenberg, 2006, op. cit.

(14) P. Fomby, and A. Cherlin, 'Family instability and child well-being', American Sociological Review, vol. 72, no. 2, 2007, pp. 181-204; O. Cynthia and S. McLanahan, 'Partnership instability and child well-being', Journal of Marriage and Family, vol. 69, no. 4, 2007, pp. 1065-1083.

(15) Guzzo and Furstenberg, 2007, op. cit.

(16) S.D. Stewart, W.D. Manning and P.J. Smock, 'Union formation among men in the U.S.: does having prior children matter?', Journal of Marriage and Family, vol. 65, no. 1, 2003, pp. 90-104; D.M Upchurch, L.A. Lillard and C.W.A Panis, 'The impact of nonmarital childbcaring on subsequent marital formation and dissolution', in L.L. Wu and B. Wolfe (Eds), Out of Wedlock: Causes and Consequences of Nonmarital Fertility, Russell Sage Foundation, New York, 2001, pp. 344-380

(17) P. Corr and R. Kippen, 'The case for parity and birth-order statistics', Australian and New Zealand Journal of Statistics, 2006, vol. 48, no. 2, pp. 171-200

(18) Forster-Jones, 2007, op cit.

(19) However for the older respondents in particular, even if information had been gathered for all cohabitations, estimation of the start and end dates of past cohabitations may have been seriously affected by recall error. See S.R Hayford and S.P Morgan, 'The quality of retrospective data on cohabitation', Demography, 2008, vol. 45, no. 1, pp. 129-149.

(20) J. Manlove, C. Logan, E. Ikramullah and E. Holcombe, 'Factors associated with multiple-partner fertility among fathers', Journal of Marriage and Family, 2008, vol. 70, no. 2, pp. 536-548.

(21) It is also possible that multi-partnered fertility was over-estimated due to the way in which pre-marital cohabitation information was collected. Rather than asking for the specific year that the cohabitation preceding a marriage began, respondents were asked about how long they had lived together before marriage in terms of the number of years or months. The start date of the pre-marital cohabitation was then calculated as the start date of the marriage minus the number of years or months of cohabitation. If respondents gave a rough estimation of the length of cohabitation it is possible that the calculated date may not have been truly accurate and children born around the time of the pre-marital cohabitation may have been classified as not having been born to that marriage.

(22) Manlove, et al., 2008, op. cit.

(23) We note that recall error may be an issue for the oldest cohorts, although major events, like marriages and births are more easily remembered than other life events. See L.L. Bumpass and R.K. Raley, 'Measuring divorce and separation', in S.L Hofferth and L.M Casper (Eds), Handbook of Measurement Issues in Family Research. Lawrence Erlbaum Associated, New Jersey, 2007, pp. 125-144.

Edith Gray and Ann Evans

In Australia family forms have changed dramatically over the last 30 years. One of these changes is the increase in the number of people who have children in more than one relationship, a concept known as multipartner fertility. This article assesses what is known about multipartner fertility in Australia, and highlights the difficulties in measuring this phenomenon.