What does it mean to have a null for median year built for occupied housing, but have values for owner- and/or renter-occupied? My coworker found these two interesting cases in the 2010-2014 data for table B25037.
Any insights would be helpful.
I recreated your table call and looked in the "Notes" section (when you click "Customize Table", notes is in the upper left hand corner in the bar across the top). I found two notes that refer to the negative/dash symbol in the estimate cells:
Since this is a median estimate, it looks like the median year is very close to the lowest year in the distribution ("falls in the lowest interval"). When I looked at the estimates for the 5-year ACS for 2013 and 2015 (on either side of this), the years are very similar. In the Hawaii census tract, the median year in all three cells hovers around 1974. In the Texas census tract, the median year hovers around 1990.
To get around this, I would use a different year, if possible. If that is not possible, you could average the estimates in 2013 and 2015. Both 2013 and 2015 share 4/5 of the same data as 2014 so averaging them will give you something close to what 2014 would be.