If I had had the same problem, I would not have spent time on SPSS and use another software for reformatting date like Excel, linux shell etc. I do not have the solution for SPSS, of course.

Dear Rasmus,

I think the easiest way to transform your data from "19881031-3456" to "198810031" is to navigate to register card 'variable view' (lower area of display) in SPSS and then go to variable property 'width' (3rd column upper area of display). Here you change the 'width' form '13' to '8' and SPSS then will only keep the first eight digits (also make sure that your variable is a 'numeric' one). This should work. Hope I was of help to you.

Best regards

Florian

Thanks Florian! After I followed your suggested steps I had 19881031 as a nominal and could easily use Data--->Define variable properties and choose Date and format yy/mm/dd.

When I try to do the same thing with 150330 and format yy/mm/dd though it doesn't work. 150330 is interpreted as a scale in variable properties, and when I change to Date it returns a date in Oct in the 16th century (apparently the date when we switched to the gregorian calendar) for all my dates.

Mehmet: well, yeah, but I want a permanent solution. In Sweden we work constantly with personal numbers and dates in those formats, so this wont be the last time I will encounter this problem. Needs to be fixed once and for all.

I finally got it! Both variables have to be in String, then changed the width, then interpreted as a Date by Data-->Define variable properties and yy/mm/dd. The problem was 150330 had 7 columns, thus including an extra space, that couldn't be converted into a date so SPSS returned 14 Oct 1532-something.

Thank you so much for posting this Rasmus. It worked for my problem, too.