I just looked at the women's file for the Ethiopia 2011 survey (ETIR61).  You referred to v119 and v125, but it seems that the relevant variables are v120 (household has radio) and v121 (household has television).  The labels are as follows:

v120:
           0 no
           1 yes
           7 not a dejure resident
v121:
           0 no
           1 yes
           7 not a dejure resident

Category 7 is "not a de jure resident".  Note the "not" in the label.  The identification of those cases means that all other cases ARE de jure.  The variable in the IR file that gets at this status is v140.  I suspect that this category is included in order to match with a MICS variable, but there may  be some other reason.

The tabulation of these two  variables (unweighted) is given below:

. tab v120 v121,m

                      |          household has: television
household has: radio |        no        yes  not a dej          9 |     Total
----------------------+--------------------------------------------+----------
                   no |     7,776        714          0          1 |     8,491 
                  yes |     4,637      2,873          0          1 |     7,511 
not a dejure resident |         0          0        505          0 |       505 
                    9 |         0          0          0          8 |         8 
----------------------+--------------------------------------------+----------
                Total |    12,413      3,587        505         10 |    16,515 

A few cases are coded 9.  I would code them "." in a composite variable, just  so  they  would be omitted from tabulations or regression.

If you want to construct a composite variable that includes radio and tv, I would not combine "radio or tv"  into a single category, because the great majority of those cases have a radio and NOT a tv.  I would distinguish between the category with 4637 cases and the category with 714 cases.  If you do not know the steps to construct the composite variable, you can post another message to the forum.