| Month of birth recording in IR [message #17817] |
Wed, 12 June 2019 07:38  |
Alexis
Messages: 3
Registered: June 2019
Location: United Kingdom
|
Member |
|
|
Hello,
I'm working on multiple DHS datasets and I wanted to use the information on the month of birth of women in IR datasets. However, it is common that the distribution of months of birth is not evenly distributed and that more women are born at the beginning of the year than at the end of the year. This is still the case when I use the v014 variable to only keep observations with complete age information.
As an example please see below some surveys with a large bias but there are many more surveys that seems to show this pattern. This is using the month of birth of women; months of birth of children seem to be more evenly distributed. I understand that there might be no birth certificate but I'm surprised that it is always biased towards the beginning of the year. Do interviewers record January if they are not sure of the month of birth?
Burkina Faso 2014 (BFIR70FL):
. tab v009 if v014==1 [aw=v005]
respondent'
s month of
birth Freq. Percent Cum.
1 3,052.2068 47.88 47.88
2 327.201778 5.13 53.01
3 309.809501 4.86 57.87
4 314.853944 4.94 62.81
5 262.622618 4.12 66.93
6 267.081004 4.19 71.12
7 271.175631 4.25 75.37
8 256.894687 4.03 79.40
9 227.786846 3.57 82.97
10 318.383412 4.99 87.97
11 212.438137 3.33 91.30
12 554.545677 8.70 100.00
Total 6,375 100.00
Afghanistan 2015 (AFIR70FL)
. tab v009 if v014==1 [aw=v005]
respondent' |
s month of |
birth | Freq. Percent Cum.
------------+-----------------------------------
1 | 1,614.1386 9.09 9.09
2 | 2,544.3889 14.33 23.42
3 | 2,428.1078 13.67 37.09
4 | 1,774.7469 9.99 47.08
5 | 1,863.0132 10.49 57.57
6 |1,601.14519 9.02 66.59
7 | 1,208.168 6.80 73.39
8 | 1,007.6137 5.67 79.07
9 | 1,256.037 7.07 86.14
10 | 1,034.7895 5.83 91.97
11 | 808.320749 4.55 96.52
12 |618.5304505 3.48 100.00
------------+-----------------------------------
Total | 17,759 100.00
Cote d'Ivoire 2011-12 (CIIR62FL)
. tab v009 if v014==1 [aw=v005]
respondent' |
s month of |
birth | Freq. Percent Cum.
------------+-----------------------------------
1 | 2,546.0023 29.45 29.45
2 | 715.87848 8.28 37.74
3 | 656.813988 7.60 45.33
4 | 542.819071 6.28 51.61
5 | 572.299899 6.62 58.23
6 | 532.295654 6.16 64.39
7 | 509.972343 5.90 70.29
8 | 508.296997 5.88 76.17
9 | 411.078436 4.76 80.93
10 | 400.15638 4.63 85.56
11 | 398.032321 4.60 90.16
12 | 850.354166 9.84 100.00
------------+-----------------------------------
Total | 8,644 100.00
|
|
|
|
|
|
|
|
| Re: Month of birth recording in IR [message #25126 is a reply to message #25021] |
Thu, 01 September 2022 16:31  |
Janet-DHS
Messages: 888
Registered: April 2022
|
Senior Member |
|
|
Following is a response from DHS Research & Data Analysis Director, Tom Pullum:
The reason for this pattern was described in https://www.dhsprogram.com/pubs/pdf/MR26/MR26.pdf, beginning on page 7. It is basically due to a tendency by interviewers to calculate birth year as the year of the survey minus the completed years of age, and then to look for a month and day that are compatible with that birth year and age. This forces births into the first part of the year. The effect is most serious for interviews that are earlier in the calendar year.
|
|
|
|
The DHS Program User Forum
Members
Search
Help
Register
Login
Home
