We will not even know whether marriage and divorce rates are rising or falling. For all the talk of evidence-based policy, the result will be that important debates on issues including family law, welfare reform, same-sex marriage and the rise of nontraditional families will proceed in a statistical void.
Much of what I, an economist who has studied family issues, and my colleagues in this field have learned about recent trends in marriage and divorce has come from questions in the American Community Survey. It asks people whether they have given birth, married, divorced or been widowed in the past year. Their answers allow demographers to track marriage and divorce rates by age, gender, race and education.
These data have revealed many important social trends, including the rise of sharply different marriage and divorce patterns between rich and poor, and the increase in divorce among older Americans, even as it has fallen for younger people. And they have provided the only statistical window into the adoption of same-sex marriage.
The Census Bureau is proposing to eliminate these questions. It would follow a series of steps taken over recent decades that have collectively devastated our ability to track family change. This isn’t being done as a strategic policy choice but rather is the result of a series of isolated decisions made across several decades by statisticians scattered across various government agencies who have failed to understand the cumulative effect of their actions.
In principle, tracking marriage and divorce shouldn’t be too hard. Every wedding, like every divorce, requires a trip to City Hall or the county courthouse to file the relevant paperwork. The resulting paper trail should be enough to allow analysts to map the contours of our changing family life over time. Indeed, until the mid-1990s, the federal government collated data from all those marriage and divorce certificates into a coherent set of marriage and divorce statistics that detailed the changing nature of marriage.
But in 1996, the National Center for Health Statistics stopped collecting these detailed data. If you subsequently got married or divorced, the forms you filled out still exist, but only as unexamined documents in a filing cabinet at your county courthouse.
Today, states report only the total number of marriages and divorces each year — providing no detail on who is marrying, which marriages persist or whether children are involved. And because the government devotes so few resources to collecting these data, several states don’t even bother counting how many divorces they grant. As a result, estimates of the divorce rate for the United States do not include data on a large share of the country; they are missing all of the divorces in California, Georgia, Hawaii, Indiana, Louisiana and Minnesota.
The rationale the health statisticians offered for no longer collecting the more detailed data was that much of this information could be gleaned from a special survey taken every five years as a supplement to the Current Population Survey. But a different set of government statisticians killed that supplement in the late 1990s.
All was not lost, as demographers could still rely on surveys asking people about their marital histories as part of a separate poll, the Survey of Income and Program Participation. These are useful data as far as they go — and indeed, we relied heavily on them in a recent analysis of national divorce trends here at The Upshot. However, these marital histories are taken every five years, they’re only as reliable as people’s memories, and the results are released only years later. As such, the most recent year we have data for is 2008.
Moreover, the sample size is so small that it is impossible to track trends by state. For instance, the latest iteration includes only 23 New Yorkers who got divorced in the most recent three years. This is particularly problematic because most family policies are implemented at the state level, and so the survey can’t be used to track, say, the implications of New York having adopted no-fault divorce laws in 2010.
It gets worse. The Bush administration decided to kill that remaining family survey. Even though it later reversed itself in the face of widespread criticism, it did so with a much reduced budget, which has necessitated a host of changes in how the survey is conducted. It remains unclear just how comparable these new data will be with earlier survey rounds.
It is this emerging statistical void that makes the debate about whether the Census Bureau should continue to collect marriage and divorce information so critical. It’s also an issue ripe for confusion, because dozens of government surveys ask people about their current marital status. But current marital status is not the relevant statistic for most policy debates. For instance, it would be a mistake to infer from Zsa Zsa Gabor’s current marital status (still married) that her children enjoyed a stable family life (he’s husband No. 9). It is far more relevant to track the flow of new marriages and divorces each year, and this is the unique contribution of the questions that are to be cut from the American Community Survey. If the cuts proceed, then the United States will be the only developed country lacking annual estimates of the rates of new marriage and divorce for each age group.
There’s a bigger issue here, too. The federal government has dozens of statistical bureaus spread across countless government agencies. The result is fragmented expertise, and incentives to make decisions that reflect narrow departmental interests rather than a broader sense of the public interest. Many other countries have consolidated the various statistical groups into a coherent national statistical agency.
When I asked Jim Treat, the Census Bureau division chief in charge of the American Community Survey, whether his proposal meant that it would be impossible to measure the divorce rate in 2016, he responded: “I don’t know the answer to that question.” I found this troubling, because I know that Mr. Treat’s proposal will eliminate our only measure of the national divorce rate.
When I asked Mr. Treat what led him to his decision, he described a processthat was focused not on whether these questions should be asked, but whether they should be asked on the particular survey he manages. His is a survey focused on generating statistics for small areas or small groups, and divorce remains sufficiently rare that estimates at the level of, say, a county, remain quite unreliable. Unfortunately, the fact that this survey yields the only estimates of marriage and divorce in many states appears not to have been considered significant.
In the end, the decision to shorten the survey reflects political calculation – an effort to mollify Tea Party Republicans who tried to eliminate the American Community Survey altogether, arguing that it is an unconstitutional breach of privacy. A briefer questionnaire may yield less political opposition. The Census Bureau targeted the questions about marriage and divorce not because people object to answering the questions posed (it turns out that they don’t), but instead because they judged the resulting data to be of little benefit, since no legislative formulas are linked directly to them.
The leading academic association of demographers has argued for the Census Bureau to reverse course, as have many individual analysts. Steven Ruggles, the incoming president of the Population Association of America,argues that cutting these data “would severely damage our ability to understand ongoing changes in American society and to implement effective policy responses.” His sentiments echo earlier analysis by the Census Bureau that “no other data sources exist that can provide the level of detail necessary to plan for and evaluate the effects of federal policies and programs related to marriage.”
The silver lining to all this is that this is the first time that I have seen people on both the conservative and progressive sides of the family policy wars agree on something: the value of continuing to collect useful data about family life. After all, each realizes that without actual data to rely on, the politically charged opinions of the other side will become more important forces shaping policy.