Marriage among among ages 18-35

Hi everyone,

Something puzzling (to me) has been coming up as I have been working with marriage data in the ACS, and I was hoping you could help me find out what is going on. I apologize if I’m doing some wrong or missing something obvious.

I am studying the effect of a policy on marriage rates, and that has required me to understand marriage trends among Hispanics, ages 18-35. My figure in this post have the weighted average of “Married, Spouse Present” on the y-axis. Below is a figure that includes only individuals in households (1970 definition), to rule out that this is caused by the group quarters changes in 2006. I would post the other figures I describe later in the post, but as a new user I can only post one.

The decline in marriage rates beginning in 2005, until perhaps 2011, are steep enough to raise the eyebrows of my PhD advisor, so I need to figure out what is going on. Between 2005 and 2006 there is a huge drop-off. But even after that, the drop-off is quite steep until the after 2010. I tried to replicate this using the CPS ASEC (below). The drop-off is again very steep, though differently shaped. Also note that in the ACS, the marriage rate settles down well-below .3 after 2010, and in the CPS it settles above .3. A big difference!

I’ve looked at this in non-Hispanic whites and blacks as well, and while the levels by race are different, the same general pattern remains. The main difference is that the gap between the ACS and the CPS is more prominent in Hispanics.

I have also looked at this by age within my sample and by education, and have not found an obvious explanation from this either.

Does anyone have any insight on what is going on here?

Thanks so much for your help.


I encourage you to take a look at the comparability tabs and data quality tabs for the variables you are using, both in IPUMS USA and IPUMS CPS. Comparability tabs note the change to the universe or question phrasing that might affect the comparability of the measure over time. Data quality flags are useful for understanding how values were allocated for individual cases. After taking a look myself, here are a couple of things that stuck out to me:

  • In MARST, “Married, spouse absent” is defined differently between CPS and USA, as described in the CPS MARST comparability tab.
  • The questionnaire wording for HISPAN (in USA data) changed in 2008 from “Hispanic” to “Hispanic Origin”, which may have captured more native-born hispanics who had not been included before. There was also a note included on the survey questionnaire at this time to distinguish that Hispanic origins are not races, which could have impacted responses (see page 8 of this Census Bureau research note).

Though you have already accounted for the major differences between USA and CPS ASEC data by controlling for the institutional difference in Group Quarters, you may want to read more on the sample design differences between the two datasets here.

I hope this helps you start to track down the differences. Please feel free to follow up with additional questions.