As the New York Times has reported more than once, the oft-quoted statistic that 50% of marriages in this country end in divorce hasn’t actually been true since the early 1980s. In fact, the Times reports that for people married since 2000, the divorce rate is just a little over 10% and on track to keep falling. If this is true, it’s a phenomenon that begs explanation, and the Times offers one: in the late ’70s and early ’80s, people got married younger than they do now. Waiting until later in life means making a better choice of partner, which is less likely to result in divorce.
Investigating the causes and effects of human behavior is complex, to put it very mildly. We at the Digital Humanities Working Group aren’t sociologists, but we do want to learn how to access the kind of data that articles like this are based on, and how to test whether a variable like ‘divorce’ correlates to a variable like ‘age-at-marriage.’ As a learning exercise, Jennifer Cook, David Hadbawnik, and I set out to do just that.
First, we ran our idea past a few colleagues, and we were eventually directed to the blog of a sociologist who has approached the same question. By following roughly the steps of Dr. Cohen’s investigation with slightly different data, we hope to:
- Learn to download and analyze large databases of census data
- Discuss the principles of statistical inference and experiment design
- Examine the reliability of the statistics published by the NY Times
- Challenge or support the analysis of Dr. Cohen (depending on outcomes)
It is our hope that projects like this will make us better analysts of data, and therefore better humanists. A future blog post will narrate in detail the steps we have taken so far, publish the R code we wrote for this project, and discuss the limitations of census data which we discovered. For now, I’d like to share a simple graph we created from our data, which says something interesting about the relationship between age at the time of marriage and the duration of marriage:
Our overall data set includes men living in New York who are currently either married or divorced. (The blog article by Dr. Cohen linked above examined married and divorced women throughout the country.) The graph above represents only the divorcés. For these men, the younger their age at the time of marriage, the greater the range of years their marriage might last. Because this data doesn’t include ongoing marriages, it doesn’t speak directly to the divorce rate’s dependency on age at marriage. What we can say is that the older a man is when he’s married, the less time he will spend in a marriage that will eventually end. What accounts for this pattern? How does it affect the NY Times narrative about the wisdom of marrying a bit older? The discussion is ongoing, and we have more questions to ask of our data set this semester. Consider joining us on this or another project, as we learn to design and execute meaningful experiments at the intersection of humanities and data science.
Authors: James Gawley, David Hadbawnik, Jennifer Cook.