Following is a response from DHS staff member, Tom Pullum:
It appears that the Ghana 2022 survey had a somewhat different set of education categories than the 2014 survey. In the 2014 survey the classification for v106 was as follows:
. label list V106
V106:
0 no education
1 primary
2 secondary
3 higher
The number of years in each level is v107. The cumulative number of years of schooling is v133, calculated as follows:
v133=v107 if v106==1
v133=v107+6 if v106==2
v133=v107+12 if v106==3
This is the standard classification and coding used in most DHS surveys.
However, an alternative was used in the Ghana 2022 survey, with a survey-specific variable s105, having these categories:
. label list S105
S105:
1 primary
2 middle
3 jss/jhs
4 sss/shs
5 secondary
6 higher
I don't know what jss, jhs, sss, and shs refer to. Perhaps another user can clarify.
The construction of v133 was changed to the following:
v133=s105 if s105=1
v133=s105+6 if s105=2
v133=s105+6 if s105=3
v133=s105+9 if s105=4
v133=s105+10 if s105=5
v133=s105+12 if s105=6
The correspondence between s105 and v106 in the IR file for the 2022 survey, unweighted, is the following:
. tab s105 v106
highest |
educationa | highest educational level
l level | primary secondary higher | Total
-----------+---------------------------------+----------
primary | 960 0 0 | 960
middle | 0 79 0 | 79
jss/jhs | 0 1,732 0 | 1,732
sss/shs | 0 876 0 | 876
secondary | 0 45 0 | 45
higher | 0 0 356 | 356
-----------+---------------------------------+----------
Total | 960 2,732 356 | 4,048
The new categories were classified as secondary. I believe that what you are observing is due mainly to this change in the classification and in the effective replacement of v106 with s105 in the 2022 survey. This may have disrupted the trend based on the earlier classification. If you are studying the trend, you may want to re-calculate the STATcompiler estimates to get better comparability over time. Hope this helps.