Submission for PAA
In response to changes in family formation patterns that marked the past several decades, researchers have been intent on understanding the causes and consequences of these transitions, particularly divorce. One of the major outcomes researchers have focused on is health. Lots of research shows that people who get married and stay married are healthier and happier than those who never marry or get divorced (cite*). Some have suggested that marriage provides benefits in the form of institutionalization, social roles, social support, and commitment (Musick & Bumpass, 2006). Others have focused on the issue of selection (cite*).
Yet this research is relatively narrow in scope for two reasons: time frame and health outcomes. Most studies on health in relation to marriage and divorce use national datasets such as NSFH which collect data across predetermined periods of time. This data set is optimal for capturing information about individuals both before and after marriage and divorce and issues of selection (e.g., Lamb, Lee, & Demaris, 2003). However, there is no control for how close in time the survey is able to measure the effects of marital dissolution on health. Interviews occur five* years apart and the process of dissolving the marriage could occur at any point. There is evidence that recency of marital dissolution is associated with poorer outcomes (Willitts, Benzeval, & Stansfeld, 2004) suggesting that a more proximal and controlled measure of well being following divorce is warranted. The period immediately following divorce is more challenging for families as changes occur in household structure, financial resources, and family relationships.
The second limitation of current divorce literature is the use of restricted health outcomes. The majority of research tends to focus on two dimensions of health: mental and physical health. Yet even these dimensions of health are often narrowly defined. Typically, mental health is defined by individuals' symptoms of depression (cite*). Physical health is less often examined but can include stress and immune responses from health researchers (Kiecolt-Glaser, Blane, Glaser, & Malarkey, 2003; Kiecolt-Glaser, Fisher, Ogrocki, Stout, Speicher, & Glaser, 1987) and ** from family demographers (cite*). A broader understanding of the effects of marital dissolution should be understood through a wider definition of health and more proximal measurement of well being.
A more proximal measurement of well being would also allow the effects of additional factors that are often considered part and parcel of divorce and may impact individuals reaction to and recovery from divorce. Collecting health data immediately following a formal attempt to dissolve the marriage (filing for divorce) would capture their reactions to marital transitions. However, even interviewing individuals following filing for divorce generally follows a separation period, which could last for any period of time (variation between separation and divorce and who does what and why - Sweet 1973; Sweet and Bumpass 1987; Thornton 1977). The negative effects on health are strongest when separation and divorce is more recent (Booth & Amato, 1991; Kincaid & Caldwell, 1995; Kitson & Holmes, 1992). It is believed that a less recent separation and divorce provides individuals with more time to recover. African Americans tend to have longer periods between separation and divorce than Whites which often leads researchers to conclude that African Americans are less negatively impacted than Whites (Kitson & Holmes, 1992; Maneker & Rankin, 1997; Sweet & Bumpass, 1987). However, controlling for length of separation accounts for racial differences (Barrett, 2003). Less work has focused on how the length of the dissolving marriage impacts health. Limited evidence suggests that individuals with poorer well being tend to have longer marriages (Berman & Turk, 1981).
There is evidence that the emotional hurt following marital dissolution depends on who initiates the divorce (Kalmijn & Poortman, 2006; Goode, 1993). Sweeny and Hortiwz (2001) describe the "magnifying or softening" effects of initiating divorce. For the partners who initiate divorce, the negative effects of divorce are not as strong. Alternatively, those who do not initiate and perceive themselves to have little control over the divorce feel more distress. This distress may manifest itself in feelings of despair, guilt, and loss for the dissolving relationship. Individuals may also feel more distress and have poorer health as a result when the marriage and separation period is marked by high levels of conflict. Following separation, conflict often increases during negotiation of major family transitions (Hetherington, Stanley-Hagan, & Anderson, 1989).
This paper addresses a gap in the current divorce literature by addressing a wider array of health indicators and interviewing families immediately following filing for divorce. We examine women's physical, mental, and sexual health approximately three months after filing for divorce. We are also examining the interaction of marital and family characteristics. In particular, we expect that a longer separation period would give women longer to recover after the initial shock of marital dissolution. However, we are unclear what effect length of marriage will have on health indicators. Women with longer separation periods will report better physical, mental, and sexual health. Conversly, the negative residual effect of high marital conflict and the ex-spouse initiating the divorce will be linked with poorer health. It is unclear what effect family size will have on health. More children require more resources but could also provide support to women. Finally, women who report a more negative emotional reaction to the divorce will have poorer health outcomes.The data for this study come from a National Institute of Child Health and Human Development (NICHD) funded research project called Texas Families Project. This is an ongoing longitudinal study of the family environment following divorce. Participants were recruited through divorce court records from a metropolitan area in the Southern United States. Trained staff collected publicly available divorce court records and recorded contact information for eligible families. Families were eligible for participation if the household included at least one elementary school aged child (age 4 through 11 years) who resided in the mother's household at least 50% of each week. After contacting the families, an additional eligibility requirement of English fluency was assessed.
After both the mother and target child agreed to participate, a series of semi-annual visits occurred for two years. Women and their children were first interviewed within 120 days after filing for divorce. This paper uses self report data from the mother at the baseline interview. 319 eligible women and the target child participated in the baseline interview. The average age of the women was 37 years. Sixty-four percent of women were white, 9% were African American, and 27% were Hispanic. Most women had two children. Over 75% of women had been married only once making this their first divorce. Six percent did not finish high school, 13% had a high school diploma or GED, 44% attended some college, and 37% graduated from college. At the baseline visit, 81% of women had paid employment outside of the home. None of the families had finalized their divorces at the baseline interview.
Physical Health. Four different indicators were tapped to measure physical health. Self reported health was a single item question that ranged from very poor to excellent on a five point scale. A count of health concerns was computed from a list of 13 ailments including items such as allergies, diabetes, back pain, anxiety, and cholesterol. BMI was computed from women's reports of their height and weight. A high BMI is indicated by scores above 25. The final indicator measured how much women perceive their health as interfering in their lives. Women rated how much their health interfered with their job, parenting, and relationship with others. An average score of health interference across domains is based on a five point scale where high scores indicate greater perceptions of interference.
Mental Health. Two dimensions of women's mental health is assessed through depression, anxiety, and substance use. Women reported their symptoms of depression using the CES-D (Radloff, 1977). This 19-item self-report questionnaire focuses on feelings and symptoms of depression. The author has reported good internal consistency (alpha .85 to .90) and test-retest reliabilities in the .5 range for up to 12 months. Anxiety was reported with a single item where women reported if a doctor or nurse had ever told them they were anxious (fix*). Women's substance use was limited to alcohol use and was indicated by reports of being drunk at least once in the past month. Reports of drug use in the study were restricted to marijuana use. Nearly no women reported drug use and it was not included in the substance use measure.
Sexual Health. Sexual health is assessed through two questions dealing with sexual activity during the past month. Women indicate if they have had sex with someone the first time they met. Women also report if they have had unprotected sex.
Based on the literature review, the effect of several variables on women's health were the focus of this study. These variables included length of separation, length of the dissolving marriage, and the number of children. Initiator status was assessed through women's reports of who suggested the divorce. Possible responses included "you, your ex-spouse, or both of you." Women reported on conflict in the form of psychological and physical attacks between the participant and the ex-spouse. The conflict could have occurred during the marriage, after physical separation, or up until the time of interview. A composite score was computed from variables measuring psychological aggression and violence. Finally, women's emotional reaction to the divorce was assessed with 15 questions including items such as "I feel as if I've been dumped," "I feel as if this is all a horrible mistake," and "I feel I will never get over the divorce." Responses were assessed on a five point scale and an average score was computed from the 15 items. Additional variables were included as controls. Age, race/ethnic category, education, and employment status served as controls.
We first examined the descriptive statistics of the women's physical, mental, and sexual health to examine trends in women's health immediately following filing for divorce. The second analytic step include running regression models for the health outcomes. The health outcomes included both continuous and categorical outcomes so both OLS and Logistic Regression was used. OLS Regression was used for scores on the CES-D, self reported health, BMI, count of health concerns, and health interference. Logistic regression was used for substance use and both indicators of sexual health. Two models were run for each outcome variable. The first model includes only the control variables and the second model adds the additional independent variables including length of marriage and separation, number of children, initiator status, conflict, and emotional reaction to the divorce.
For the majority of the women, this divorce represents the dissolution of their first marriage. The length of the average marriage was 9.66 years (SD = 5.38) and the median was 9 years, with a range from 1 to 26 years. The average length of separation was 1.2 years (SD = 1.54) though the median length of separation was seven months. There is a great deal of variation in separation length for the women in this sample. Nearly 50% of sample have been separated for 6 months or less while over 10% of the sample have been separated for more than 3 years. Approximately 25% of the women indicated that their spouse had suggested the divorce while 75% felt they had some say in suggesting the divorce. The average score for conflict in the marriage was 10.44 (SD = 7.25). On average, women scored 3.84 on the emotional reaction to divorce indicating a somewhat negative emotional reaction to the divorce.
Most women reported somewhat good physical health. On average, women reported good to excellent health (M = 4.22, SD = -.73). Women reported few health concerns (M = 1.85) and perceive their health as not interfering with their lives (M = 1.46, SD = 0.75). However, the average BMI was 26.07, falling into the overweight category. Women also reported high levels of depression symptoms (M = 15.53, SD = 9.69) and substance use. 34% of the women reported being drunk at least once during the past month and 30.5% of women reported being diagnosed as anxious. In the past month, 16% of the women reported having unprotected sex. Only 3% of the women reported having sex with someone the first time the met.
African American women were more likely to report higher BMI but better self reported health than White women. Hispanic women also reported high BMI than White women. Women who attended or graduated from college reported better self reported health, lower BMI, and fewer health concerns than women who did not finish high school. High school graduates also reported fewer health concerns than women who did not finish high school. Employment outside the home seemed to serve as a protective factor for self reported health and health concerns. These differences remained significant in Model 2 when characteristics of the dissolving marriage were added. Model 2 included variables for family size, length of the marriage and separation, initiator status, conflict, and emotional reaction to the divorce. Women reported significantly more health concerns when the ex-spouse suggested the divorce. All indicators of women's physical health were negatively impacted when there were higher levels of conflict in the dissolving marriage. Poorer scores on self reported health, BMI, and health interference were associated with negative emotional reactions to the divorce.
There were no differences in levels of depression after controlling for age, ethnicity, education, or employment. However depression did differ by characteristics of the dissolving marriage in Model 2. Higher levels of conflict and more negative emotional reactions to the divorce were linked to higher scores on the CES-D.
Hispanic women were less likely to be diagnosed as anxious than White women. Women who attend college and those who graduated from college had lower odds of reporting anxiety than women who did not graduate from high school. The results changed little in Model 2. However, women who reported higher levels of conflict had marginally higher odds of reporting anxiety.
Results from the logistic regression for substance use did yield differences across control variables. Older women African American women were less likely to report substance use though women with some college attendance were more likely to report substance use. In Model 2, having more children reduced the odds of substance use though higher levels of conflict increased the odds of substance use.
Two indicators were used to measure sexual health: having sex with someone the first time they met and unprotected sex during the last month. None of the results for unprotected sex fell within the 95% Confidence Interval though some results were marginally significant and are worth reviewing. Hispanic women were more likely to have unprotected sex than White women. Women employed outside the home were less likely to have unprotected sex. Women who were leaving longer marriages were more likely to have unprotected sex as were women who reported more negative emotional reactions to the divorce.
Few results were significant for the second indicator of sexual health though older women and those who graduate from college were less likely to have sex with someone the first time they met them. In Model 2, the effect of having higher education attainment became insignificant. Though marginally significant, women leaving longer marriages were less likely to have sex with someone the first time they met them. A history of longer marriages reduces a woman's chances of having sex with someone the first time they meet but increases the chances they will have unprotected sex.
Discussion
race differences consistent with previous work - future work should build on Barrett's interaction work for race and sep/marriage length