Hi Bruno! I wonder how you determined this? I did the following:

set more off
use e:\DHS\DHS_data\IR_files\RWIR41FL.dta, clear

* youngest child has bidx_k==1
* oldest child has bord_k=1

rename b*_0* b*_*

gen first_cmc=.
gen last_cmc=.

local li=1
while `li'<=20 {

replace first_cmc=b3_`li' if bord_`li'==1
replace last_cmc=b3_`li' if bidx_`li'==1

local li=`li'+1
}

gen interval_survey_to_last=v008-last_cmc
gen interval_survey_to_first=v008-first_cmc

tab v201, summarize(interval_survey_to_last) means
tab v201, summarize(interval_survey_to_first) means


use e:\DHS\DHS_data\MR_files\RWMR41FL.dta, clear

gen interval=mv008-sm213c

* Question: is the following closer to the first or second distribution above??
tab mv201, summarize(interval) means

* Answer: much closer to the second. Thus sm213* refer to the first birth, not the last

So--I agree with you that the label should be changed and I will inform Data Processing. I wonder, however, whether the label is what they wanted and the calculation was done wrong. Good catch, as we say--Tom

The men's questionnaire refers to the "premier enfant" (first child). The only problem is with the label.