You have reached the Federal
Depository Library Program Electronic Collection (FDLP/EC) Archive. This
archive assures ongoing access to publications in the FDLP/EC. Access to publications
in the archive is available through GPO's suite of cataloging and locator
services, including the Catalog of U.S. Government Publications (CGP).
Archived copies of publications are retrieved in the event that publications
are no longer available from the original publishing agency site.
Title:
Unbinding the ties: edit
effects of marital status on same gender couples
SuDoc Number: C 3.223/27:34
Item Number:
0154-B-55 (online)
CGP system
Number: 001151946
2021-06-08
Publication content below:
Unbinding
the Ties: Edit Effects of Marital Status on Same Gender Couples
Working Paper Number POP-WP034 Jason M. Fields and Charles L. Clark Fertility and Family Statistics Branch Population Division U.S. Census Bureau Washington, DC ABSTRACT
DISCLAIMER
This paper reports
results of research and analysis undertaken by Census Bureau staff. It has
undergone a more limited review than official Census Bureau publications.
This report is released to inform interested parties of research and to
encourage discussion. CONTENTS
TABLES
Table
1. Number of households by household type and Census 2000 Dress Rehearsal
site. Table
2. Number of households by household characteristics: Sacramento,
California. (6k) Table
3. Number of households by household characteristics: Columbia, South
Carolina. (7k) Table
4. Number of households by household characteristics, Sample
characteristics only: Sacramento, California. (5k) Table
5. Number of households by household characteristics, Sample
characteristics only: Columbia, South Carolina. (5k) Table
6. Odds ratios (a) for being married: Census 2000 Dress Rehearsal. (4k) Introduction
The origins of this
edit process come from two directions. First, the "legal"
definition of marriage evolved in this century from the common usage and
normative understanding of the term marriage and became defined as a
"[l]egal union of one man and one woman, as
husband and wife." The census edit process recognizes this commonly
understood definition of marriage, resulting in a defined universe of married
couples that excludes same gender marriages. Congress has
recently addressed recognition of same gender marriages. Legislation recently
passed, H.R.3396 of the 104th Congress, the Federal Defense of Marriage Act,
is summarized as follows: Defense of Marriage Act - Amends the Federal judicial code to provide
that no State, territory, or possession of the United States or Indian tribe
shall be required to give effect to any marriage between persons of the same
sex under the laws of any other such jurisdiction or to any right or claim
arising from such relationship. An implication of
this legislation for the edit process of Census data products is that,
although a person may hypothetically be legally married to someone of the same
gender in one state, the edit process would need to be informed by the
specific marriage laws of the state where the couples reside at the time of
Census enumeration. This legislation further defines marriage and spouse as
follows: `In determining the meaning of any Act of Congress, or of any ruling,
regulation, or interpretation of the various administrative bureaus and
agencies of the United States, the word `marriage' means only a legal union
between one man and one woman as husband and wife, and the word `spouse'
refers only to a person of the opposite sex who is a husband or a wife.' While this is not a
binding definition for state legislation and decisions, it would present some
inconsistencies for Census Bureau data processing should states decide to
recognize same gender marriages. Second, same gender
unions create inconsistencies in the presentation of data where
characteristics in a table stub assume that the spouse of the householder is
the opposite gender of the householder. An example of this type of tabulation
would be presentation of a table of male householders showing the children
ever born to the spouse of the householder. The implication is that all of
the spouses are females, and therefore at risk of childbirth. Additionally,
balancing the number of husbands and wives is a priority for presenting
tables and making a consistent presentation of data, where users are able to
match universes across tables. These are both
institutional reasons for maintaining the status quo in data collection and
processing. Any suggested change to the current collection methodology and
editing would have to be demonstrated and defended through research and then
carefully planned and implemented, taking into account
the potential policy and processing implications. This work addresses
one of the research issues. This paper examines the heterogeneity of the
populations self-identifying as same gender married couples, compared to
those identifying as same gender unmarried partners. If these populations
prove distinct, then consideration would need to be taken when analyzing the
results of the Census 2000 enumeration, because researchers will not be able
to distinguish between these two differently self-identified households once
the editing procedure has occurred. Both published tabulations and micro-data
files released to the public will only contain the final edited results. The data for this
paper include responses to the Census 2000 Dress
Rehearsal conducted in April 1998 from sites in Sacramento, California and
South Carolina. We evaluate the characteristics of the same-gender couples'
households that would be expected to differentiate married from unmarried.
The prevailing discussion of the determinants and characteristics of marriage
and cohabitation are primarily from a heterosexual perspective. We attempt to
integrate this perspective with lesbian and gay perspectives on the
institution of marriage. From this, we address three hypotheses
differentiating the characteristics of same gender couples identifying as
married from those identifying as unmarried partners:
|
<hr size=1 width="100%" align=center> |
||||
Self-identified
Household type |
CENSUS 2000 DRESS
REHEARSAL SITE |
|||
Sacramento, CA |
Columbia, SC |
|||
ALL Households |
"Long
Form" |
ALL Households |
"Long
Form" |
|
|
||||
TOTAL HOUSEHOLDS |
128,498 |
17,607 |
227,053 |
33,246 |
|
|
|
|
|
SPOUSE |
|
|
|
|
Total Married |
49,942 |
6,832 |
109,571 |
16,128 |
Opposite gender couple |
49,582 |
6,770 |
108,722 |
15,963 |
Same gender couple |
360 |
62 |
849 |
165 |
Male/male |
165 |
27 |
353 |
61 |
Female/female |
195 |
35 |
496 |
104 |
|
|
|
|
|
UNMARRIED PARTNER |
|
|
|
|
Total Unmarried Partner |
7,184 |
771 |
7,516 |
829 |
Opposite gender couple |
6,365 |
658 |
7,190 |
785 |
Same gender couple |
819 |
113 |
326 |
44 |
Male/male |
400 |
64 |
161 |
19 |
Female/female |
419 |
49 |
165 |
25 |
|
|
|
|
|
NEITHER |
|
|
|
|
Total Neither |
71,372 |
10,004 |
109,966 |
16,289 |
Male householder ONLY |
26,829 |
3,753 |
35,684 |
5,233 |
Female householder ONLY |
44,543 |
6,251 |
74,282 |
11,056 |
|
This is a
descriptive analysis, and as such, an examination of simple univariate and
bivariate distributions comprise the bulk of this work. First, we compare
unedited distributions for "married" and unmarried same gender couples,
and compare these with their combined distribution as would be the case once
the Census 2000 edits are performed. These distributions are shown separately
for male/male and female/female couples. We attempt to further identify
relationships that persist in the presence of other covariates. To this end
we summarize the descriptive findings with multivariate logistic regressions
that measure the strength of the associations between the covariates and same
gender couples who identify as married rather than unmarried partners.
An initial
examination of the data in Tables 1, 2, and 3 shows some basic patterns. The
relative proportions of same gender married and unmarried partner households
are reversed for the two sites, with unmarried partners more prevalent than
same gender married couples in California, but the reverse in South Carolina
(Table 1). This pattern repeatedly emerges both in the total and sample data
sets. The similarity in the numbers is coincidental. One possible reason for
the reversal is the provision for persons in California to have domestic
partnerships, and the fact that the population in California may be more
familiar or comfortable with the concept of an "unmarried partner."
However, an argument could be made that California has a more active
homosexual community and there may be more interest and awareness in the
same-gender marriage debate. This is conjecture and something that will not
be revealed by this data. Nonetheless we continue to examine patterns of the
other covariates by marital status, and by site. If there were significant
differences in the meaning of "marriage" by site, it is our hope
that site specific tables will illuminate these differences.
Tables 2 and 3
present distributions of basic "short-form" characteristics; age,
race, presence and age of children, and number of unrelated adults in the
household for same gender couples by gender and type of relationship in
California and South Carolina respectively. These results show age
distributions that reveal striking differences. For both sites, the
proportion of respondents (person 1) over age 50 in married couple households
is much greater than in unmarried partner
households, indicating a relatively older population for married households.
This is true whether one examines male/male or female/female households. The
mean age difference between person 1 and their spouse/partner is less than 3
years although there are sizable standard deviations, indicating some spread
in individual age differences. A greater proportion of married male couples
are in the same age group than are male unmarried partner couples.
The distributions
of same gender couple households by the race of person 1 show differences for
both the site at which the data was collected and for married versus
unmarried partner households. In California (Table 2) less than 50% of the
married couple households have a householder identifying only white as their
race. Compared to about 80% of the householders in unmarried partner
households. This pattern persists in South Carolina, although there is only
about a 10-15 percentage point difference between married and unmarried
couple households. Overall, South Carolina has less racial diversity than California
with approximately 95% of householders either White or Black regardless of
couple status; so this finding was not unexpected.
For both sites, in both married-couple or unmarried partner households, the
majority of partners identify the same racial group as person 1. There are
only marginally greater proportions of interracial couple households in
California (about 15 to 20 percent) than in South Carolina (5 to 10 percent).
Household
composition proved to hold some interesting differences between married and
unmarried partner households. Children were coresident
in a much greater proportion of married couple households than in unmarried
partner households. Between 50 and 60% of married couple households in
California and between 45 and 50 % of married couple households in South
Carolina have children in the household. This contrasts sharply with the
unmarried partner households where only 8 to 20% of households in California,
and between 18 and 22% in South Carolina, had coresident
children. We also examined the number of coresident
unrelated adults, and found very few households (generally less than 5
percent), either married couple or unmarried partner, that had any coresident unrelated adults, other than the respondent
and their spouse/partner. This finding was similar by gender of the couple
and survey site.
The two columns of
"edited" data in tables 2 and 3 indicate the estimated number of
same gender unmarried couples which would result after all married couple
households were edited into unmarried partner households. For the California
site, about 30 percent of unmarried partner households would be comprised of
households originally designating themselves as married-couple households.
For South Carolina, this percentage would be 70 percent. A combination of
these two demographically different household structures into a single edited
household type could affect the analyses of same gender households,
especially if different areas produce an aggregate total from different
proportions of household types.
Table
2. Number of households by household characteristics: Sacramento,
California. (6k)
Table
3. Number of households by household characteristics: Columbia,
South Carolina (7k)
Long form
distributions for California and South Carolina are presented in tables 4 and
5. These tables present data from the sample "long" form for same
gender couples by gender and type of relationship. For this set of tables we include: the respondents' reported marital
status, the educational attainment of the respondent and the similarity of
this level for the spouse/partner of person 1, the employment status of
person 1 and of their spouse/partner in the last week, and an indicator of
their residential stability. The frequency distributions in tables 4 and 5
are based on very small numbers of households. The percentages should be
considered cautiously. However, it should be noted that when aggregated and
analyzed in a logistic regression the relationships cursorily observed in
these tables persist. Discussion of the regression results follows the
discussion of these tables.
The reported
marital status of person 1 is the first of the variables added from the
sample forms. This was included to evaluate the responses entered for
relationship to household reference person (person 1). The distribution of
the marital status item for both men and women, in both data collection
sites, and for married couples and unmarried partners, prove to be consistent
and validate the responses to the relationship item. The distributions do not
reflect perfect agreement, but the overwhelming trend is in a direction that
is consistent and distinguishes these two groups of households. In
California, 74 percent of partners were reported to be in
agreement with the respondent, that they were married spouse present.
In South Carolina, between 80 and 85 percent of partners were reported to be in agreement with the respondent, that they are married
spouse present in married couple households. The percentage of respondents
and partners in unmarried partner households both reported in agreement that
they were "currently not married" ranges between 84 and 97 percent
of the households for the two sites and gender combinations.
There are greater
proportions of respondents in unmarried partner households that have
experience in college, and there are correspondingly higher proportions of
respondents in married couple households that have high school degrees or
less. Current employment status follows a trend consistent with educational
status between these two types of couples. For both male and female couple
households, a greater proportion of unmarried partner households have both
the respondent and spouse/partner currently employed than in married couple
households. Residential stability also indicates some differences by type of
couple. Higher proportions of married couple households have both the
respondent and their spouse reporting that they lived in that house five
years ago. These patterns are consistent with the age pattern of the
householders shown in tables 2 and 3 which indicate a higher proportion of
householders age 50 years and over among the married couples than unmarried
partner households. Younger couples have probably benefitted from more recent
trends in increasing educational levels, and are now more mobile, and likely
to be employed than are people who are older.
Table
4. Number of households by household characteristics, Sample
characteristics only: Sacramento, California. (5k)
Table
5. Number of households by household characteristics, Sample
characteristics only: Columbia, South Carolina. (5k)
Table 6 presents
the results of simple multivariate regressions on the probability that a same
gender household is also a married couple household by various
characteristics. It is important to point out that these are probabilities of
associations, not cause. The data collection sites are combined in these
models, and a dummy variable is added to indicate the collection site.
Although we believe that there are interactions present in the underlying
relationship, for the purpose of this paper we do not attempt to model them.
In model A we include the variables available from the short form (100%)
only. The relationship between having children in the household and being
married is quite strong. Same gender households with children are 7 times
more likely to be married couple households. Households in South Carolina are
also more likely to identify as married couple households, 5 times more than
those households from Sacramento. Controlling for the relative age of the
partner, householders who are 40 years or older are more likely to identify
as married couple households (3 times more likely), and partners who are in
the same five year age group are more likely to also
be married compared to those whose partners are in the adjacent or more
distant age groups.
In model B, we add
variables from the long form, and correspondingly reduce the sample to long
form respondents. The relationships between presence of children, age of the
respondent, and data collection site and the likelihood of being married
persist in the presence of additional covariates from the sample form. In
model B, the variable indicating the presence of other unrelated adults is
dropped because it failed to attain significance in the 100 percent models.
In addition to these relationships, respondents with any college experience
are less likely to be in married couple households 74% less likely than
respondents with a high school degree. Respondents in stable households,
those where both partners lived at that residence five years ago, are more
likely to also be married. Households in which both the respondent and their
partner worked last week, are less likely to identify themselves as married
couples, 53% less than households where only one of the partners is working.
Models C and D show
results for the variables included in model A, but stratified by site. The
results in these models only differ from the combined model in two places.
First, for the Sacramento site, couples in which the respondent and partner
identify the same race category are more likely to also be married. This
effect would not be expected to be present at the Columbia, South Carolina
site because of the very high overall level of racial homogamy in both
married and unmarried partner households. At the Columbia site, female
couples are 38% more likely to also be married compared to male couples.
Appendix A shows
results for the same four models, only exchanging marriage for the presence
of children as a dependent outcome. These models present a consistent picture
with that shown for marriage. Some of the interactions present in the data
begin to be apparent when these results are examined. Because the presence of
children was such a strong correlate for being in a married couple household,
this appendix is included as reference.
Table
6. Odds ratios (a) for being married: Census 2000 Dress Rehearsal
(4k)
Our second
hypothesis, that married couple households would be more likely to also have coresident children, was supported. Across gender and
site, and controlling for other characteristics, couples with children in the
household are much more likely to also be married. Remember that we are using
the term "married" in a slightly different context than normal.
These are not state sanctioned legal marriages if they are accurately
reported as same gender unions. This is more likely the adoption of a social
meaning of marriage that is being used in household environments with
children. There is no way to validate this speculation from these data.
However, it is certainly clear that children are more likely to be found in
same gender households reported as "married".
Regarding our third
hypothesis, the results are not obvious. Due to data availability issues we
were unable to include income, household ownership, occupation, and other
measures of individual and household wealth and financial stability from the
unedited data file. Our first proxy for financial stability is current work
status. While dual earner couples are likely have
more disposable income, and in our data set are less likely to be married, it
is also true that, on average, financial stability increases with age, and so
we found older respondents more likely to be married. Pending additional
information about financial stability, and further consideration of
interactions, additional conclusions about trends in marital status for same
gender couples by financial stability would be premature.
It is important to
warn the reader that these data cannot be extrapolated to the nation as a
whole. However, these results clearly illustrate that there is a potential
for significant increases in the same gender unmarried partner counts. This
would seriously impact analytic results, by inflating the number of unmarried
partner couples with couples who have clearly different demographic
characteristics. Same gender unmarried partners represented about one sixth
of one percent of all households in the 1990 Census. This is of course a
national average and will vary significantly by region, state, and especially
by smaller geography. In the dress rehearsal results, same gender couples
represent 0.5 percent of households at the Columbia site, and 0.9 percent of
households at the Sacramento site. Altogether, this combined total of
same-gender married couple and unmarried partner households makes up a very
small percentage of all households. However, after the edit reassignments are
made, same gender households will make up about one sixth of all unmarried
partner households. On average, adding same gender couples who indicate that
they are "married" to same gender unmarried partner counts
significantly increases the latter estimate for both sites combined.
It is our
conclusion that these two groups of self identified
couples are distinct, and consideration of alternative edit decisions and
tabulation methods should be explored. Further examination of these
populations once the covariates are fully edited, and examination of data
from the full Census 2000 enumeration in April, 2000 will further illuminate
the degree of heterogeneity caused by our edit process in this population.
Boswell, John.
1995. Same-Sex Unions in Premodern Europe. New York, NY: Vintage
Books.
Eskridge, Jr.,
William N. 1996. From Sexual Liberty to Civilized Commitment - The Case For Same-Sex Marriage. New York, NY: The Free Press.
Folbre, Nancy. 1994. Who Pays for the Kids? Gender and the Structures of
Constraint. London: Routledge.
Risman, Barbara J. 1998. Gender Vertigo: American Families in Transition,
Yale University Press.
Stiers, Gretchen A. 1999. From This Day Forward - Commitment, Marriage, and
Family in Lesbian and Gay Relationships. New York, NY: St. Martin's
Press.
Strasser, Mark.
1997. Legally Wed - Same-Sex Marriage and the Constitution. Ithaca,
NY: Cornell University Press.
Sullivan, Maureen.
1998. "Alma Mater: Family 'Outings' and the Making of the Modern Other
Mother (Mom)." Unpublished.