Determinants of Divorce Rates in the United States: An Econometric Study of 1987-2017

Abstract: The looming idea that nearly half of all marriages end in divorce has been haunting the United States for decades. However, research proves that the divorce rate per 1,000 persons in the U.S. has been continually falling since the late 1970s. This paper seeks to observe the determinants of divorce and underline the trends of a 30-year span, from 1987 to 2017. This paper explores the relationships between key economic and religious variables and divorce rates.

Angelica Urrego

• INTRODUCTION

The persistence of high divorce rates in the United States has been a topic of discussion for many decades. The current divorce rate, at 2.9%, is highly referenced statistic, given that the United States became accustomed to experiencing aggressively high divorce rates from 1980 to 2000. When analyzing the trends of divorce rates from 1987 to 2017, there are numerous distinguishable factors that follow divorce trends or seem to have an indirect relationship. This study seeks to analyze the determinants of divorce rates in the United States by looking at select variables. Four independent variables have been selected based on societal characteristics that relate to divorce in America. A regression analysis will be used to help explain the divorce trends from 1987 onward. Furthermore, this study will analyze the individual correlation between divorce and specific variables by exploring the reasoning behind correlations or lack thereof.

The organization of this paper follows 4 large sections. Section II provides an outline of previous scholarly work on this topic; the review of the literature creates a basis for comparison on the results obtained from the regression analysis. Section III is an empirical analysis containing descriptive statistics and the outlining methodology behind this paper. Section IV presents the results of the regression analysis and explores the logic behind the multiple regression and individual correlations. This section goes in depth in explaining the results of relationships through a comparison of the previous literature on the topic. Section V provides a conclusion on the regression analysis and how this relates to current divorce trends in the United States. An outlook on the future of divorce and social behavior is briefly mentioned as an encouragement for future research on this subject.

• REVIEW OF LITERATURE

The phenomena of divorce is widely studied within the US, as it falls into both cultural and economic concerns. The relationship between poverty and divorce rates is a growing area of study in poverty studies mostly because of the prevalence of single mothers below the poverty line. The study by Espenshade, “The Economic Consequences of Divorce”, published in the Journal of Marriage and Family explores the historical significance of divorce in current poverty statistics. Espenshade explores the likelihood that historically high divorce rates are a leading factor in the growing number of single mothers that live below the poverty line. This is often referred to as the feminization of poverty. When married couples divorce, they are more likely to face economic hardship because of loss of economies of scale, they can no longer enjoy low per capita costs by dividing costs among each other (Espenshade 620).

Another key variable that is being tested on its effect on divorce rates is the unemployment rate. The relationship between divorce and unemployment is often mentioned to be an interesting one, for it seems to be that case that couples prefer to stay together during times of economic hardship to better their chances of stability. When the unemployment rate lowers, and the economy creates more opportunities for advancement; couples who were previously together for the stability choose to opt for divorce. A study published in the Journal of Social Science Research by Amato and Beattie, named “Does the unemployment rate affect the divorce rate? An analysis of the state data 1960-2005”, finds results that mirror previous conclusions by other researchers. According to Amato and Beattie, “when the sample is divided into time periods, unemployment is negatively and significantly associated with divorce after 1980. These findings provide the strongest support for a “cost of divorce” perspective and suggest that a high rate of unemployment decreases the rate of divorce.”

Following the topic of the job market in relation to divorce, this study takes an interest in observing the relationship between divorce rates and women labor force participation throughout the years. Both statistics have displayed notable trends over recent decades, allowing researchers to establish and quantify the relationship. A study by Bremmer and Kesselring for the Atlantic Economic Journal on “Divorce and Female Labor Force Participation” resulted in highlighting the overall gender trends that are being witnessed in the United States over time. Along with increasing divorce rates, women’s LFPR increased and there was a notable reduction in the birth rates for married women. Though these trends hold truth in general gender studies, this study was performed in 2004, before the time period that divorce rates began decreasing at a faster rate. It will be interesting to further analyze the decreasing divorce rates in a LFPR that is plateauing for both men and women.

The final characteristic that this study analyzes in relation to divorce rates is religiosity. It is a part of our common knowledge to view divorce as frowned upon within religious communities, and further research on the topic displays that attitudes towards divorce varies among religious affiliation. The study “Religion as a Determinant of Marital Stability” by researchers Lehrer and Chriswick explores the role of religious composition within families as a determinant of marital stability. The results of this study suggest that spouses of Protestant and Catholic affiliations have stronger marital stability because of the factor of religion. This trend is noted in various other studies, causing “Catholic” and “Protestant” to be key variables in this econometric analysis of divorce rates. This study will also aim to observe how the strength of the relationship between divorce and religious affiliation varies between protestant affiliation and catholic affiliation.

• EMPIRICAL ANALYSIS

The following equation is produced to observe the determinants of divorce rates from 1987 to 2017 using regression analysis from STATA: DIV= (WLFPR, POV, UNEMP, PRT, CTH) + error term

Building on the findings of previous literature, this study formulates the hypothesized effects of each independent variable that is being studied. The following table displays an explanation on the meaning of each independent variable in the regression analysis and its expected correlation with the dependent variable of Divorce rates.

Table 1: Description of Independent Variables and their Expected Correlation the Divorce Rates

Observational Limits

It is important to mention that given that divorce is a behavioral variable, we must acknowledge the exclusion of personal feelings of love or lack thereof in this analysis. There is a plethora of behavioral variables that are tied to divorce, such as how much couples argue, their techniques for conflict resolution, and their priorities. It is often noted that most couples do not get divorced because of the event of situational problems, rather, their marriage depends on how they communicate to solve the problems that are presented. This study cannot expand its scope to analyze these variables, it instead uses five quantifiable variables that are commonly cited in divorce research.

Description of Statistics

Table 2: Description of Statistics: Means, Standard Deviations, Maximums, and Minimums for each

Table 2 displays the average of all observations over the 1987-2017 time period that is being examined. This allows for a clearer view of the rand and skew of all the variables being observed. Through an observation of the max and mix of variables, one can see the expected trends over the years in comparison to divorce. The variable of poverty is expected to be positively related to divorce, though the characteristics of the trend lines are especially notable because there seems to be a time lag in the relationship between these two variables. Religiosity trends display a positive relationship with divorce rates, given their correlation in years on max and mins. The relationship between divorce and percentage of women in the labor force continues to be an interesting one given that divorce rates broke the parallel when they began decreasing in the late 1980s.

• REGRESSION RESULTS

Table 3:

Number of Obs. = 31 R-Squared: 0.9580

This model predicts that as women’s labor force participation rate increases by one percentage point, the divorce rate is expected to decrease by 0.3115 percentage points, all else equal. This variable is statistically significant and displays an evident relationship. This result disproves my hypothesized expected effect of a positive relationship between DIV and WLFPR (Table 1). As stated previously, the characteristics of this relationship is an interesting area of study given that divorce rates seem to continue to decrease and the labor force participation rate for both men and women is reaching a point of plateau. Figure 1 shows that since 1992, divorce rates in the United States have been steadily decreasing while the labor force participation rate slowly continued to increase at decreasing rates. Previous research on this topic places more of an emphasis on highlighting the evident relationship of labor force participation and divorce in the 1970s and 1980s. “Women, Work, and Divorce”, a 1985 study by William Sander for the American Economic Association is a prime example of the type of academic research that has been previously done on this topic. This study affirms that “the [high] divorce rate is significantly and substantially affected by the earning ability of women in market work” (519). However, as observable through the regression equation, this relationship has evolved considerably in the recent years. Figure 2 displays the recent appearance of the negative relationship between DIV and WLFPR.

Figure 2

The regression model also predicts that when the poverty rate increases by one percentage point, divorce rates are expected to increase by 0.0546 percentage points, all else equal. This effect is small, and the variable of poverty displays significance at a 10% level, however, this is expected given that there are many omitted variables that are excluded from this relationship. Nonetheless, the relationship between poverty and divorce falls in line with the arguments of poverty economics, as poor couples are unfortunately more likely to have marital instability.

The variable of unemployment is not significant at a 1%, 5%, or 10% level and therefore cannot be interpreted.

Regarding religiosity, the model predicts that with a one percentage point increase in the number of Protestants in the United States, the divorce rate is expected to increase 0.0466 percentage points, all else equal. As expected, this is not a large effect since strong protestant religiosity serves as a disincentive for divorce. The effect is alike regarding Catholic affiliation; with a one percentage point increase in the number of Catholics in the United States, the divorce rate is expected to increase 0.0589 percentage points, all else equal.

Tables 4 and 5 display the strength is individual relationships through single regressions. It can be clearly observed through the high R-squared of the single regression analysis that there seems to be stronger relationship with the percent of protestant individuals in the United States and the divorce trends from 1987 through 2017. This may be interpreted through the simple fact that the American population is predominantly Protestant, however, the percentage of religious affiliation in the United States has been slowly decreasing overall due to the shift away from traditional ideals. This decrease mirrors the decrease of divorce rates over the recent decades.

• CONCLUSION

It can be argued that the decrease divorce in the United States has gone without its deserved attention, as the public sentiment seems to be frozen in the times of extraordinarily high divorce rates of the 70s and early 80s. It is essential to reiterate the existence of low divorce rates in our current state of development as a characteristic of our progression towards social stability. This paper sought to test the effect that five independent variables had on divorce rates and found that four variables have significant relations to divorce from 1987 to 2017. The R-squared of the multiple regression analysis conveys that 95.8% of the variance in divorce rates from 1987 to 2017 can be explained by the 5 independent variables in the study. Though this high R-squared may be influenced by conditions that will be discussed in the Limitations section, the relationship continues to be evidently true.

Limitations

The strong relationship displayed in the OLS model, and as visible through the R-squared, may be exaggerated by the fact time-series data was being used and most variables follow a similar trend over the 30-year timeline. Though this is beneficial to my analysis, there are many variables influencing these relationships throughout the years that are out of the scope of my study. For this reason, I am only able to acknowledge the results of the relationship between poverty and divorce but cannot tie these variables together with certainty given that there are omitted variables. The same follows with the unemployment rate; though previous literature finds a relationship, this paper is limited to the observation of the span of 30 years, impeding a careful analysis of the interesting relationship between unemployment and divorce.

Outlook

The results of this regression test lay out a base for understanding who is getting divorced and allows for further research on the why’s of things. The common statement among news sources is that millennials are not only less likely to get divorced; they are delaying marriage overall. Census figures released in 2018 show that the median age at first marriage is now 30 for men and 28 for women. It seems to be the case that millennials have been conditioned by the marital instability of their previous generations to the point where they are skeptical to accept commitments. Moreover, these social changes come with a variety of issues. An article by Time magazine summarizes the trends, “people who are getting married are increasingly staying married. But that group is an even smaller and more privileged group of individuals. Marriage is becoming one of the institutions from which the poor, less-educated and disadvantaged are excluded” (Luscombe). Ongoing trends require further academic research on the effects of marriage and divorce on social behavior, and how these behaviors differ among people of different socioeconomic status. Though the divorce rate has significantly decreased from 1987 to 2017, and is expected to stay low, this may not be explicitly a good sign.

WORKS CITED

Amato, P. R., & Beattie, B. (2011, January 4). Does the unemployment rate affect the divorce rate? An analysis of state data 1960–2005. Retrieved from https://www.sciencedirect.com/science/article/pii/S0049089X10002917.

Bremmer, D., & Kesselring, R. (n.d.). Divorce and female labor force participation: Evidence from time-series data and cointegration. Retrieved from https://link.springer.com/article/10.1007/BF02299436.

Espenshade, T. (1979). The Economic Consequences of Divorce. Journal of Marriage and Family, 41(3), 615-625. doi:10.2307/351631

Lehrer, E. L., & Chiswick, C. U. (n.d.). Religion as a determinant of marital stability. Retrieved from https://link.springer.com/article/10.2307/2061647.

Luscombe, B. (2018, November 26). The Divorce Rate Is Dropping. But That May Not Be Good News. Retrieved from https://time.com/5434949/divorce-rate-children-marriage-benefits/.

Sander, W. (1985). Women, Work, and Divorce. The American Economic Review, 75(3), 519-523. Retrieved from www.jstor.org/stable/1814817