30
Number of Siblings in Childhood and the Likelihood of Divorce in Adulthood Mary Lopez Economics of the Family Fall 2014 Abstract Despite a decline in fertility across many developed countries, relatively little is known about the consequences of being raised with fewer children. Some studies indicate that fewer siblings could have a positive effect, as children with few siblings tend to have better educational

Number of Siblings in Childhood and the Likelihood of Divorce in Adulthood

Embed Size (px)

Citation preview

Page 1: Number of Siblings in Childhood and the Likelihood of Divorce in Adulthood

Number of Siblings in Childhood and the Likelihood of Divorce in Adulthood

Mary Lopez

Economics of the Family

Fall 2014

AbstractDespite a decline in fertility across many developed countries, relatively little is known about the consequences of being raised with fewer children. Some studies indicate that fewer siblings could have a positive effect, as children with few siblings tend to have better educational outcomes. However, there has also been evidence that siblings could increase social skills and lead to better long term relationships as an adult. With data from the General Social Surveys 1972-2012, I attempt to quantify the effects of siblings on the likelihood of relationship formation and dissolution. I find that there is no effect of siblings on the likelihood of marriage or divorce, net of covariates.

Page 2: Number of Siblings in Childhood and the Likelihood of Divorce in Adulthood

The question of whether there is a child quantity – quality tradeoff within a family

is one of the most studied topics in family economics. Most of these studies focus on the

dependent variables of educational attainment or wages but few question how the number

of siblings a person has affects that individual’s ability to develop long-term meaningful

relationships. Specifically, this paper will examine how the number of siblings affects the

likelihood of marriage and divorce. This topic has been underexplored despite being

particularly interesting from a policy standpoint. Especially as the fertility rate decreases

in many developed countries and families become smaller, whether by requirement as in

China, or by natural progression, it will be important to understand the social

consequences that may arise from this change in family structure.

There are multiple channels by which family size might affect marriage and

divorce. For example, the resource dilution perspective states that parents have a finite

amount of resources such as time and money to give to their children, so the more

children there are, the less each child gets. As such, children with more siblings should

have more negative outcomes than those with few or no siblings. Proponents of this view

often point to studies showing that children with few or no siblings perform better in

school and on cognitive ability tests than children with many siblings (Blake 1981).

Using Blake’s work as a foundation, Downey analyzed data from the National Education

Longitudinal Study and found a consistent negative relationship between sibship size and

parental resources such as money saved for college, having a computer in the house, and

time spent speaking about school-related matters (Downey 1995).

However, there are several important questions raised to the resource dilution

theory. Much of the research on resource dilution stems from cross-sectional data that

may not fully control for the differences between parents who have many children versus

2

Page 3: Number of Siblings in Childhood and the Likelihood of Divorce in Adulthood

those who have few. This could affect the ability to interpret the effect of siblings on

educational outcomes as causal. Additionally, an important limitation to the dilution

model is its nearly exclusive focus on educational or wage outcomes. Siblings matter in

many other ways and the focus on education effects may have obscured the positive

aspects of sibling interaction, which are explored next.

An alternative to the dilution model is that siblings serve as resources themselves,

separate from those provided by parents. Sibling relationships are particularly unique,

given their long duration, shared familial environment, heritage, and experiences, and

similar expanded network of relationships (White 2001). Growing up with siblings could

provide individuals with an opportunity to develop conflict resolution skills, which could

facilitate the maintenance of future relationships in adulthood. In addition, siblings may

indirectly affect development by influencing how parents rear their children. After

success with previous children, parental confidence may increase along with the use of

effective parenting techniques.

Both the resource dilution model and the siblings as resources model are plausible

and further previous research on both models is presented in the next section, in the hopes

of creating a clearer picture of the expected effect of siblings on marriage and divorce

patterns. In this paper I attempt to contribute to the research on this topic by quantifying

the effects of the number of siblings on the likelihood of marriage and divorce, while also

considering the possible effect of the sex mix of the family.

Previous Literature

In much of previous economic research, such as Arleen Leibowitz’ article “Home

Investment in Children,” there has been a widely observed negative association between

family size and academic achievement (Leibowitz 1974). Despite previous beliefs

3

Page 4: Number of Siblings in Childhood and the Likelihood of Divorce in Adulthood

however, recent research has found that there is not significant evidence of a child

quantity – quality tradeoff within families. Agrist, Levy, and Schlosser use instrumental

variables to account for the likelihood of omitted variables in previous estimates of the

effect of family size and finds that the IV estimates are considerably smaller than the

corresponding ordinary least squares (OLS) estimates (Angrist 2010).

While there may be less of a child quantity – quality tradeoff than previously

thought, family size still seems to have some effect on educational achievement and

wages. Black, Devereax, and Salvanes argue that much of the negative relationship

between family size and economic outcomes seems to be driven by birth order. A higher

birth order is negatively associated with wages and average years of education, especially

for women. Consequently, as families get smaller, average outcomes will improve, but

only because there are fewer high birth order kids (Black 2005). Behrman comes to a

similar conclusion as Black, finding that birth-order differences occur despite parental

preferences or prices by birth order favoring later borns, apparently because of stronger

endowment effects that favor firstborns (Behrman 1986).

Clearly, family size and birth order have been shown to have an effect on

economic and academic factors, but little research has been done on the effect of siblings

on divorce. Doug Downey and Donna Bobbitt-Zeher explore this relationship in their

forthcoming paper, “Number of Siblings During Childhood and the Likelihood of

Divorce in Adulthood,” and find that more siblings means less chance of divorce as an

adult. Downey finds that each additional sibling a person has, up to about seven, reduces

the likelihood of divorce by 3% (Downey 2014). Even more intriguingly, it doesn’t

appear to be the difference between being an only child and having any siblings at all that

was significant, but rather how family dynamics change with the incremental effect of

4

Page 5: Number of Siblings in Childhood and the Likelihood of Divorce in Adulthood

each additional sibling. One would expect that having any siblings at all would give an

individual the experience with personal relationships need to have a successful marriage,

but it appears that each additional sibling increases the ability of an individual to deal

with a marital relationship as an adult. Downey uses data from the General Social Survey,

which involves interviews with approximately 57,000 adults across the United States at

28 points between 1972 and 2012. To account for the argument that smaller families are

more likely to have a single parent or some other issue that might adversely affect

children in their future marriages, Downey controls for education, socioeconomic status,

family structure, race, age at marriage, whether the respondents had children, gender role

attitudes and religious affiliation of both the respondents and their parents. Even with

controlling for all of these factors, the relationship between siblings and later divorce

remained relatively unchanged.

One possible explanation Downey offers for this relationship between siblings

and later divorce is that having siblings allows for better social and interpersonal skills,

that later translate into helpful skills in a marriage. This idea is supported by an earlier

work by Downey, in which he found that kindergarteners negotiate peer relationships

better when they have at least one sibling (Downey 2004). Another possible explanation

is that sibling relationships are intimate, with both positive and negative emotions, and

could match the dynamics that many individuals have in marriage relationships.

Downey’s findings are very intriguing but further research is needed on the

relationship between siblings and later divorce, especially since this link has been

explored by so few. While Downey explores the effect of sibship size in general, it would

5

Page 6: Number of Siblings in Childhood and the Likelihood of Divorce in Adulthood

be interesting to see if the effect of siblings on later divorce varies by gender or if the sex

mix of the family has any effect.

MethodsSample

Data from the General Social Surveys (GSS) from 1972-2012 will be used in this study.

The GSS represents data from approximately 57,000 adults collected at 28 points over

four decades and the data are particularly well suited for this study because of the level of

detail on measures of marital and family composition outcomes. The key independent

variable is number of siblings, which is measured in a continuous fashion and includes

full, half, step, and adopted siblings. The primary dependent variable of interest is a

binary variable created for ever having been divorced. There are also a number of

variables that need to be controlled for, which are discussed below, and are included in

the vector term Xi. The basic OLS equation is as follows:

Ever having been divorcedi = β0 + β1number of siblingsi + XiB2 + εi

Independent variables

The fact that the kinds of parents who have many children are potentially different

than those that have few makes it difficult to isolate sibling effects. As a result, I attempt

to control for potential differences between family sizes that might also affect marriage

and divorce. Background factors are considered first. Respondent’s education is captured

with two dummy variables, high school or less and completed college, with high school

or less serving as the base case. Similarly, I measure years of mother’s education, which

serves as a proxy for socioeconomic status of family of origin, with corresponding

dummy variables mother high school or less and mother completed college. Again

mother high school or less serves as the base case. Race is gauged with three

6

Page 7: Number of Siblings in Childhood and the Likelihood of Divorce in Adulthood

dichotomous dummy variables based on self reporting of racial group: white, black, or

other, with other serving as the base case. Sex is captured with a dummy variable for

males, with females as the base case. The number of brothers the respondent reported

having is captured by the continuous variable, number of brothers. Respondent’s age is

captured with a continuous variable measured in years. Given the potential for non-linear

effects, models include a term for age squared as well. Finally, I include a continuous

variable for survey year to control for the time at which the survey was administered.

I also consider variables for economic status, family formation, and geography. In

all models I include family income, a logged, continuous variable that measures total

family income in constant 2000 dollars. I also include a dummy variable for home

ownership, where 1 indicates that the respondent owns their own home and 0 indicates

that they do not. Based on the respondent’s location at the time of survey, I created

dummy variables for geographic region, coded as North, East, West, and South, with

South serving as the base case.

Finally, since large families may foster more traditional worldviews that may

affect marriage and divorce patterns, I include measures for gender role attitudes and

religiosity. I created an index to measure gender role attitudes, based on the level of

agreement with the following statements: 1) “It is much better for everyone involved if

the man is the achiever outside the home and the woman takes care of the home and

family,” and 2) “A preschool child is likely to suffer if his or her mother works.” The

responses to each statement were captured on a scale of 1 (strong agreement) to 4 (strong

disagreement) and then the responses were added to create an index with possible values

of 2 to 8. To capture religion, I created two measures. Religious affiliation is a dummy

variable coded 1 for reporting any religious affiliation and 0 for reporting none. I also

7

Page 8: Number of Siblings in Childhood and the Likelihood of Divorce in Adulthood

measure religious attendance based on answers to the question, “How often do you

attend religious services?” Dummy variables are created for attending religious services

less than monthly, less than weekly, and weekly, with weekly serving as the base case.

It is important to note that several of these variables, particularly respondent’s

education, family income, home ownership, and gender role attitudes, could be potential

mechanisms through which the effect of the number of siblings is working. As such, the

results will be discussed differently for models that control for these factors and those

that do not.

Dependent variables

I consider two dependent variables. First, I constructed a binary variable for ever

having been married. The reference category is never been married. Second, for

respondents who have ever been married, I created a binary variable for ever having been

divorced. Never having been divorced serves as the reference category. Together, these

two outcomes allow for a view into the effect of the number of siblings on relationship

formation and dissolution.

Table 1 shows the descriptive statistics from the sample, and has been broken

down into those who have ever been divorced and those who have never been divorced,

to see if there are any obvious differences between the two groups that might indicate a

larger story. Those who have been divorced are generally older than those who have not,

with a mean of about 50 compared to 43, which is not too surprising considering it takes

time to get married and subsequently divorced. Both categories have grown up with

multiple siblings (mean = 3.8 for divorced and 3.7 for never divorced) though there is

significant variation in sibship size (standard deviation = 3.2 and 3.0, respectively).

Considering current fertility trends in the U.S., this seems like an unusually high number

8

Page 9: Number of Siblings in Childhood and the Likelihood of Divorce in Adulthood

of siblings but that could possibly be explained by the fact that it includes step and

adopted siblings as well. The mean number of brothers is low for both categories, though

there is some variation in that as well which could account for the low mean. Those who

have been divorced have more children on average, though there is also significant

variation in number of children as well. Surprisingly, on average there does not seem to

be a difference in religious affiliation or gender role attitudes between those who have

been divorced and those who have not. There are only 5,504 responses for those who

have ever been divorced, which seems low considering there are nearly 40,000 for never

having been divorced but that is likely due to the fact that the second category includes

those that have never been married in the first place.

Analytic Strategy

I test the effect of sibling-ship size on the likelihood of ever marrying using the

full sample, considering first the bivariate and then the multivariate models. I then

predict the likelihood of ever having been divorced, specifically testing for the effect of

the number of siblings. Again, I test first the bivariate and then the multivariate models.

In the multivariate models, I focus on controlling for variables that have the potential to

affect the relationship between siblings and the marriage or divorce outcome. Finally, I

examine the effect of the sex mix of the family by testing the effect of the proportion of

brothers on the likelihood of divorce. For the regressions regarding sex mix, I report total

results, as well as separate results for both men and women.

9

Page 10: Number of Siblings in Childhood and the Likelihood of Divorce in Adulthood

Results

Do siblings affect the likelihood of ever marrying? I predict the likelihood of marrying in

Table 2. Results indicate a significant but small positive of the number of siblings on the

likelihood of marriage, with each additional sibling increasing the likelihood of marriage.

Specifically, with no controls in the model as represented in Model 1, each additional

sibling is associated with a one percent increase in the odds of ever marrying. However,

that effect decreases with the addition of background, economic, and geographic controls

in Model 2, and decreases again to half a percent with the addition of religious and

gender role controls in Model 3, though the effect remains significant.

Does the number of siblings one has affect the likelihood of divorcing? Table 3 shows

results related to the effect of siblings on the likelihood of divorce. There is essentially no

effect of the number of siblings on the likelihood of divorce in any of the models,

whether bivariate as in Model 1 or with the controls as in Models 2 and 3.

Does the number of brothers one has affect the likelihood of divorcing? Table 4 shows

the results related to the proportion of brothers on the likelihood of divorce. In a one

sibling family, having a brother is associated with a 5.3% increase in the likelihood of

divorce, an effect that is statistically significant. As the number of siblings increases, this

effect decreases and is no longer statistically significant. All regressions include controls

for background, economic, and geographic variables, as well as religious affiliation and

attendance, and gender role attitudes.

Table 5 similarly shows the results related to the effect of the proportion of

brothers on the likelihood of divorce, considering only male respondents. The effect has a

10

Page 11: Number of Siblings in Childhood and the Likelihood of Divorce in Adulthood

similar magnitude as the total results in a one sibling family, but is not statistically

significant. The effect is only statistically significant in three and five sibling families,

where an additional brother is associated with about a 3% increase in the likelihood of

divorce.

Table 6 shows the results when considering only female respondents. The effect

of an additional brother is again a similar magnitude as the total results in a one-sibling

family, about 5%, but again is not statistically significant. In fact, the effect is not

statistically significant in any family size.

Discussion

As the fertility rate decreases in many developing countries, more and more

children are being raised without siblings but social scientists know little about the

consequences of this change. I attempt to advance the literature in this area by examining

the relationship between the number of siblings and the likelihood of marriage and

divorce. My results indicate that siblings do not have a significant effect on the likelihood

of marriage or divorce. Even if the effect were working through one of the potential

mechanisms mentioned earlier, such as respondent’s education or income, the effect

should have been visible, statistically significant, and even potentially overestimated in

Model 1, considering that it includes no controls.

However, in a one-sibling family, having a brother is associated with a 5.3%

increase in the likelihood of divorce, which is rather surprising. Perhaps only interacting

with a brother does not lead to the same social skills as interacting with sisters does, and

leads to fewer conflict resolution skills and thus less successful long term relationships.

This interpretation is consistent with the other regressions within Table 4 as well,

11

Page 12: Number of Siblings in Childhood and the Likelihood of Divorce in Adulthood

considering the effect of an additional brother decreases and becomes statistically

insignificant as the number of siblings increases. However, the results from Table 5 and 6

do not support this interpretation, as the effect is statistically insignificant, even in a one-

sibling family. Women’s likelihood of divorce especially does not seem to be affected by

the number of brothers, as the result is not statistically significant in any family size.

Men’s likelihood of divorce is still affected, though only in sibling sizes of three and five.

Further research into this phenomenon is needed to illuminate the inconsistencies within

these results.

The result that siblings have no effect on relationship formation and dissolution is

different from what Downey and Bobbitt-Zeher found in their 2014 paper, where they

found that each additional sibling decreases an individual’s chance of divorce by 3%.

This is likely due to the fact that Downey and Bobbitt-Zeher handle missing data with

multiple imputations. They replace missing values on independent variables with

plausible estimates developed from an imputation model constructed from all variables in

their regression models (Downey 2014). Had I pursued a similar strategy, I also may have

found a larger effect of siblings on relationship formation and dissolution.

Another reason for the perceived lack of effect could be that the resource dilution

effect and the siblings as resources effect are offsetting each other, making it seem that

siblings have no effect on long term relationship formation and dissolution. Future

research should attempt to separate those effects and quantify their effects. It would also

be interesting to find out if spacing has any effect, as one might expect that respondents

might benefit most from having a widely spaced older sibling who is relatively mature.

The GSS also represents data only from the United States, where the decline in fertility

has been less steep than in European countries so it would be interesting to expand this

12

Page 13: Number of Siblings in Childhood and the Likelihood of Divorce in Adulthood

question to countries like Italy and France where fertility rates have dropped well below

replacement.

13

Page 14: Number of Siblings in Childhood and the Likelihood of Divorce in Adulthood

Table 1Descriptive Statistics, General Social Survey 1972-2012

Year of survey 1992 11.5 1991 11.7Age 49.7 14.1 43.2 16.9Number of siblings 3.8 3.2 3.7 3.0Number of children 2.4 1.7 1.7 1.7Family income 50686 39783 41673 36547Number of brothers 0.09 0.51 0.08 0.49White 0.86 0.33 0.81 0.38Black 0.10 0.30 0.13 0.33Other 0.02 0.16 0.04 0.21Male 0.43 0.49 0.44 0.49Homeowner 0.37 0.48 0.27 0.44North 0.24 0.43 0.26 0.44East 0.13 0.33 0.20 0.40West 0.21 0.41 0.19 0.39High school or less 0.56 0.49 0.49 0.49Completed college 0.18 0.39 0.25 0.43Mother high school or less 0.83 0.37 0.77 0.41Mother completed college 0.07 0.25 0.10 0.30Attends religious services less than monthly

0.54 0.49 0.49 0.49

Attends religious services less than weekly

0.15 0.36 0.16 0.37

Attends religious services weekly 0.30 0.45 0.34 0.47Religiously affiliated 0.90 0.28 .89 0.31Composite of gender role questions 2.4 2.7 2.3 2.8

N=5504 N=39251

Notes: Sample taken from General Social Survey from 1972-2012. The sample includes 57,000 adults but is reduced to 44,755 after dropping missing observations. Data is reported separately by the binary variable of ever having been divorced, regardless of current marital status. Religiously affiliated is a binary variable that accounts for identifying with any religion at all, regardless of what that religion is. The composite of gender role questions is an index of the responses to questions regarding gender roles, which can be found in the description of data. The higher the gender index, the more progressive the respondent is regarding traditional gender roles.

Ever Been Divorced

Standard DeviationMean

Never Been Divorced

MeanStandard Deviation

14

Page 15: Number of Siblings in Childhood and the Likelihood of Divorce in Adulthood

Table 2OLS Estimates of Probability of Ever Marrying

Model 1 Model 2 Model 3Number of siblings 0.011 0.006 0.005

(0.0006)** (0.0005)** (0.0005)**Age 0.053 0.052

(0.0005)** (0.0005)**Age squared -0.000 -0.000

(0.00)** (0.00)**Male -0.061 -0.055

(0.003)** (0.003)**White 0.054 0.050

(0.007)** (0.007)**Black -0.077 -0.083

(0.008)** (0.008)**Completed college -0.041 -0.038

(0.004)** (0.004)**Mother completed college

-0.070(0.005)**

-0.064(0.005)**

Income 0.000 0.000(0.00)** (0.00)**

North -0.015 -0.012(0.004)** (0.004)**

East -0.054 -0.049(0.004)** (0.004)**

West -0.038 -0.027(0.004)** (0.004)**

Religious 0.080(0.005)**

Attends religious services less than monthly

-0.029(0.003)**

Attends religious services less than weekly

-0.005(0.004)

Gender role attitude composite

-0.007(0.0005)**

_cons 0.745 -0.570 -0.595(0.003)** (0.013)** (0.014)**

R2 0.01 0.34 0.35N 44,755 44,755 44,755

Notes: Sample is described in notes of Table 1. Dependent variable is a binary indicator for ever marrying. Column 1 reports OLS estimates from model described in Equation 1. Column 2 reports OLS estimates from Model 2, which adds controls for background, economic, and geographic variables. Column 3 reports OLS Table 3

15

Page 16: Number of Siblings in Childhood and the Likelihood of Divorce in Adulthood

OLS Estimates of Probability of Ever Divorcing

Model 1 Model 2 Model 3Number of siblings 0.001 -0.001 -0.001

(0.0005)* (0.0005)* (0.0005)Age 0.017 0.017

(0.0005)** (0.0005)**Age squared -0.000 -0.000

(0.00)** (0.00)**Male -0.003 -0.008

(0.003) (0.003)*White 0.048 0.048

(0.007)** (0.007)**Black 0.025 0.029

(0.008)** (0.008)**Completed college -0.066 -0.064

(0.003)** (0.003)**Mother completed college -0.008 -0.009

(0.005) (0.005)Income 0.000 0.000

(0.00)** (0.00)**North -0.036 -0.037

(0.004)** (0.004)**East -0.073 -0.077

(0.004)** (0.004)**West -0.010 -0.015

(0.004)* (0.004)**Religious 0.012

(0.005)*Attends religious services less than monthly

0.055(0.003)**

Attends religious services less than weekly

0.029(0.004)**

Gender role attitude composite

0.002(0.005)**

_cons 0.127 -0.305 -0.359(0.002)** (0.013)** (0.014)**

R2 0.00 0.05 0.06N 44,755 44,755 44,755Notes: Sample is described in notes of Table 1. Dependent variable is a binary indicator for ever divorcing. Column 1 reports OLS estimates from model described in Equation 4. Column 2 reports OLS estimates from Model 5, which adds controls for background, economic, and geographic variables. Column 3 reports OLS estimates from Model 6, which additionally adds controls for gender role attitudes, religious affiliation and attendance. Gender role attitude variable explained in notes of Table 1, as well as in description of data. All regressions weighted using the composite weight variable provided by the GSS, which deals with experimental

randomization and black oversamples. Non-robust standard errors are presented in parentheses. *p<0.05; ** p<0.01. 16

Page 17: Number of Siblings in Childhood and the Likelihood of Divorce in Adulthood

Table 4OLS Estimates of Probability of Ever Divorcing Based on Proportion of Brothers

1 Sibling 2 Siblings 3 Siblings 4 Siblings 5 Siblings 6 or more Siblings

Number of brothers

0.053(0.02)**

0.019(0.012)

0.009(0.009)

0.003(0.009)

0.012(0.009)

-0.006(0.004)

_cons -0.435 -0.350 -0.365 -0.389 -0.252 -0.301(0.034)** (0.032)** (0.036)** (0.042)** (0.049)** (0.032)**

R2 0.09 0.06 0.07 0.07 0.05 0.04N 7,909 8,701 7,231 5,178 3,600 9,791

Notes: Sample is described in notes of Table 1. Dependent variable is a binary indicator for ever divorcing. All regressions control for background, economic, and geographic variables, as well as religious affiliation and attendance, and gender role attitudes. Gender role attitude variable explained in notes of Table 1, as well as in description of data. Non robust standard errors are presented in parentheses. *p<0.05; ** p<0.01.

Table 5

OLS Estimates of Male Probability of Ever Divorcing Based on Proportion of Brothers

1 Sibling 2 Siblings 3 Siblings 4 Siblings 5 Siblings 6 or more Siblings

Number of brothers

0.058(0.031)

0.018(0.018)

0.033(0.014)*

-0.008(0.012)

0.030(0.013)*

-0.002(0.006)

_cons -0.479 -0.427 -0.409 -0.497 -0.370 -0.355(0.049)** (0.045)** (0.054)** (0.062)** (0.070)** (0.048)**

R2 0.09 0.08 0.07 0.08 0.06 0.04N 3,658 3,937 3,184 2,215 1,551 4,056

Notes: Sample is described in notes of Table 1. Dependent variable is a binary indicator for ever divorcing. All regressions control for background, economic, and geographic variables, as well as religious affiliation and attendance, and gender role attitudes. Gender role attitude variable explained in notes of Table 1, as well as in description of data. Non-robust standard errors are presented in parentheses. *p<0.05; ** p<0.01.

17

Page 18: Number of Siblings in Childhood and the Likelihood of Divorce in Adulthood

Table 6

OLS Estimates of Female Probability of Ever Divorcing Based on Proportion of Brothers

1 Sibling 2 Siblings 3 Siblings 4 Siblings 5 Siblings 6 or more Siblings

Number of brothers

0.052(0.027)

0.021(0.016)

-0.013(0.012)

0.014(0.013)

-0.004(0.012)

-0.008(0.005)

_cons -0.389 -0.285 -0.328 -0.314 -0.146 -0.271(0.048)** (0.046)** (0.049)** (0.056)** (0.068)* (0.043)**

R2 0.08 0.05 0.07 0.07 0.05 0.05N 4,251 4,764 4,047 2,963 2,049 5,735

Notes: Sample is described in notes of Table 1. Dependent variable is a binary indicator for ever divorcing. All regressions control for background, economic, and geographic variables, as well as religious affiliation and attendance, and gender role attitudes. Gender role attitude variable explained in notes of Table 1, as well as in description of data. Non-robust standard errors are presented in parentheses. *p<0.05; ** p<0.01.

18

Page 19: Number of Siblings in Childhood and the Likelihood of Divorce in Adulthood

Works Cited

Angrist, Joshua, Victor Lavy, and Analia Schlosser. "Multiple Experiments for the

Causal Link between the Quantity and Quality of Children." Journal of Labor

Economics 28.4 (2010): 773-824. JSTOR. Web. 5 Oct. 2014.

Behrman, Jere R., and Paul Taubman. "Birth Order, Schooling, and Earnings."Journal

of Labor Economics 4.S3 (1986): 121-45. JSTOR. Web. 5 Oct. 2014.

Black, Sandra E., Paul J. Devereux, and Kjell G. Salvanes. "The More the Merrier? The

Effect of Family Size and Birth Order on Children's Education*." Quarterly

Journal of Economics 120.2 (2005): 669-700. Web. 5 Oct. 2014.

Blake, J. “The Only Child in America: Prejudice Versus Performance.” Population and

Development Review 7(1): 43-54. (1981).

Downey, D. B. “When Bigger Is Not Better: Family Size, Parental Resources,

and Children’s Educational Performance.” American Sociological Review

60(5) (1995): 746-761.

Downey, Douglas B., and Donna Bobbitt-Zeher. “Number of Siblings During

Childhood and the Likelihood of Divorce in Adulthood.” Manuscript accepted

at Journal of Family Issues. Forthcoming as of 6 Oct. 2014.

Downey, Douglas B., and Dennis J. Condron. "Playing Well with Others in

Kindergarten: The Benefit of Siblings at Home." Journal of Marriage and

Family66.2 333-50. (2004). JSTOR. Web. 5 Oct. 2014.

Leibowitz, Arleen. “Home investment in children.” Journal of Political

Economy 82:S111–S131. (1974).

White, L. "Sibling Relationships Over the Life Course: A Panel Analysis.

Journal of Marriage and Family 63(2): 555-568. (2001).

19