Number of children in household variable from individual level dataset

[Sophie Gebers]

Dear all
I am using the PALMS dataset (Post-Apartheid Labour Market Survey) - it is a stacked crossed sectional dataset of the labour market dynamic survey in South Africa (and its predecessors) across the years 1994 to 2017. It is at the individual level.
I am wanting to create a variable for the number of children in the household under the ages of 7 (i.e. 'Young children present'). I gather you do this using the 'egen' command. There is a variable for the household and individual identifier.

The code that i have tried to run is:

egen Y_CHILD = count(personnr) if age <= 7 , by (uqnr)
* Counts number of young children in every uqnr (household identifier)
egen YChild = max (YCHILD), by (uqnr)
*Assigns total number of children to each individual within the same household
tab YChild
recode YChild (0/2=1 "1. 0 - 2 children") (3/4=2 "2. 3 - 4 children") (5/6=3 "3. 5 - 6 children") (7/max=4 "4. 7 or more children"), gen(YChild_cat)
label var YChild_cat "Categorical variable: Number of Young Children present in Household"
*Assess years
tab YChild_cat if year == 2007
tab YChild_cat if year == 2012
tab YChild_cat if year == 2017

However, this does not produce the desired result. I believe i am missing something fundamental about the egen command.
Could anyone help?

[Nick Cox]

this does not produce the desired result

That's on all fours with "doesn't work". What is wrong specifically?

Here is some technique that may help:

egen num_child = total(age <= 7) , by (uqnr)
egen tag = tag(uqnr)
tab num_child if tag

The first command gives you the number of children under 8 in every observation in every household. The second and third insist that each household is tabulated just once; otherwise you are tabulating observations, not households. I guess that's the main problem here.

See also

help egen

for more on tag(). Recoding and restricting to years is extra: I am guessing that is not where your problem lies.

By the way, watch out that

egen num_others = total(age > 7) , by (uqnr)

catches missing values on age too; if you didn't want that, then

egen num_others = total(inrange(age, 8, .)) , by (uqnr)

will not include those missings.

[Sophie Gebers]

Thank you for your response.

My confusion lies in the fact that I do not know how to follow one household using the 'tag' command, and then to specify the number of children within a single household.

I have now attempted to use it, however i am unsure if it is still correct.
The code:

egen hh_one=tag(uqnr)

* 'tag' - creates a dummy variable: where = 1 for the first observation for each household
* and = 0 to subsequent observations

egen num_ychild = total(age <= 7) , by (uqnr)
tab num_ychild if hh_one==1

This produces the result:

num_ychild | Freq. Percent Cum.
------------+-----------------------------------
0 | 339,982 56.12 56.12
1 | 76,915 12.70 68.81
2 | 57,762 9.53 78.34
3 | 34,183 5.64 83.99
4 | 42,876 7.08 91.06
5 | 9,409 1.55 92.62
6 | 11,541 1.90 94.52
7 | 5,636 0.93 95.45
8 | 13,499 2.23 97.68
9 | 3,197 0.53 98.21
10 | 2,099 0.35 98.55
11 | 1,749 0.29 98.84
12 | 3,353 0.55 99.40
13 | 759 0.13 99.52
14 | 599 0.10 99.62
15 | 518 0.09 99.71
16 | 799 0.13 99.84
17 | 203 0.03 99.87
18 | 200 0.03 99.90
19 | 122 0.02 99.92
20 | 180 0.03 99.95
21 | 68 0.01 99.97
22 | 43 0.01 99.97
23 | 38 0.01 99.98
24 | 56 0.01 99.99
25 | 23 0.00 99.99
26 | 8 0.00 99.99
27 | 5 0.00 99.99
28 | 13 0.00 100.00
29 | 1 0.00 100.00
30 | 4 0.00 100.00
31 | 6 0.00 100.00
32 | 6 0.00 100.00
33 | 1 0.00 100.00
34 | 3 0.00 100.00
35 | 1 0.00 100.00
36 | 1 0.00 100.00
37 | 1 0.00 100.00
40 | 1 0.00 100.00
42 | 1 0.00 100.00
------------+-----------------------------------
Total | 605,861 100.00

.
How do you know restrict it to one variable for the number of children within a single household - i.e. i would like to get rid of having to use the 'if' statement in the command.

Regards

[Nick Cox]

Sorry, but I don't think I understand your new question.

But at its simplest, the number of children in a household is the same for all people in a household. If you want to tabulate households not people, then either you need to use an if restriction to see each household just once, or you need a reduced dataset with one observation per household.