Keeping only households with one wife and one husband

[Enrique Alameda]

// Generating indicators for husband, wife, son, and daughter
gen is_wife = (sex == 3 & (relationtohead == 1 | relationtohead == 2))
gen is_husband = (sex == 1 & (relationtohead == 1 | relationtohead == 2))
gen is_daughter = (sex == 3 & (relationtohead == 3 | relationtohead == 4))
gen is_son = (sex == 1 & (relationtohead == 3 | relationtohead == 4))

// Checking the number of husbands and wives in each household and wave
bysort household_id wave: egen wives_per_wave = total(is_wife)
bysort household_id wave: egen husbands_per_wave = total(is_husband)

// Determining if there is at most one husband and one wife in each household
bys household_id: egen max_husbands = max(husbands_per_wave)
bys household_id: egen max_wives = max(wives_per_wave)
bys household_id: gen unique_husband = (max_husbands == 1)
bys household_id: gen unique_wife = (max_wives == 1)

// Keeping households with at most one husband and one wife
bys household_id: keep if unique_husband == 1 & unique_wife == 1

However, although when browsing the data it appears that there is only when husband and one wife (across various waves) for each household, when I do:

distinct pidlink if is_wife == 1 and distinct pidlink if is_husband == 1, I still get more wives than husbands (regardless of waves given the distinct command. Thus, implying, that I have kept households where more than one wife exists).

I provide a data example below:

clear
input str10(household_id person_id) float(wave is_husband is_wife)
"0010651" "001060004" 5 0 1
"0010651" "001065102" 5 1 0
"0010651" "001065103" 5 0 0
"0010651" "001065104" 5 0 0
"0010851" "001080003" 5 1 0
"0010851" "001080012" 5 0 1
"0010851" "001085103" 5 0 0
"0010851" "001085104" 5 0 0
"0010851" "001085105" 5 0 0
"0010851" "001085106" 5 0 0
"0010851" "001085107" 5 0 0
"0012200" "001220001" 1 1 0
"0012200" "001220001" 2 1 0
"0012200" "001220001" 3 1 0
"0012200" "001220001" 4 1 0
"0012200" "001220001" 5 1 0
"0012200" "001220002" 1 0 1
"0012200" "001220002" 2 0 1
"0012200" "001220002" 3 0 1
"0012200" "001220002" 4 0 1
"0012200" "001220002" 5 0 1
"0012200" "001220003" 1 0 0
"0012200" "001220003" 2 0 0
"0012200" "001220003" 3 0 0
"0012241" "001220003" 4 1 0
"0012241" "001220003" 5 1 0
"0012200" "001220004" 1 0 0
"0012200" "001220004" 2 0 0
"0012200" "001220004" 3 0 0
"0012200" "001220004" 4 0 0
"0012200" "001220004" 5 0 0
"0012200" "001220005" 1 0 0
"0012200" "001220005" 2 0 0
"0012200" "001220005" 3 0 0
"0012200" "001220005" 4 0 0
"0012200" "001220005" 5 0 0
"0012200" "001220006" 1 0 0
"0012200" "001220006" 2 0 0
"0012200" "001220006" 3 0 0
"0012200" "001220006" 4 0 0
"0012200" "001220006" 5 0 0
"0012200" "001220007" 1 0 0
"0012200" "001220007" 2 0 0
"0012200" "001220007" 3 0 0
"0012200" "001220007" 4 0 0
"0012200" "001220007" 5 0 0
"0012200" "001220008" 1 0 0
"0012200" "001220008" 2 0 0
"0012200" "001220008" 3 0 0
"0012200" "001220009" 1 0 0
"0012200" "001220009" 2 0 0
"0012200" "001220009" 3 0 0
"0012200" "001220009" 4 0 0
"0012200" "001220010" 1 0 0
"0012200" "001220010" 2 0 0
"0012200" "001220010" 3 0 0
"0012200" "001220010" 4 0 0
"0012200" "001220010" 5 0 0
"0012241" "001220011" 4 0 1
"0012241" "001220011" 5 0 1
"0012241" "001220013" 4 0 0
"0012241" "001220013" 5 0 0
"0012200" "001220014" 4 0 0
"0012200" "001220014" 5 0 0
"0012200" "001220015" 4 0 0
"0012200" "001220015" 5 0 0
"0012241" "001224104" 4 0 0
"0012241" "001224104" 5 0 0
"0012241" "001224105" 4 0 0
"0012241" "001224105" 5 0 0
"0012241" "001224106" 5 0 0
"0012400" "001240001" 1 1 0
"0012400" "001240001" 2 1 0
"0012400" "001240001" 5 1 0
"0012400" "001240002" 1 0 1
"0012400" "001240002" 2 0 1
"0012400" "001240002" 5 0 1
"0012400" "001240003" 1 0 0
"0012400" "001240003" 2 0 0
"0012400" "001240003" 5 0 0
"0012400" "001240004" 1 0 0
"0012400" "001240004" 2 0 0
"0012400" "001240004" 5 0 0
"0012400" "001240005" 1 0 0
"0012400" "001240005" 2 0 0
"0012451" "001240005" 5 0 1
"0012400" "001240006" 1 0 0
"0012400" "001240006" 2 0 0
"0012400" "001240006" 5 0 0
"0012400" "001240007" 1 0 0
"0012400" "001240007" 2 0 0
"0012400" "001240007" 5 0 0
"0012400" "001240008" 1 0 0
"0012400" "001240008" 2 0 0
"0012400" "001240008" 5 0 0
"0012400" "001240009" 1 0 0
"0012400" "001240009" 2 0 0
"0012452" "001240009" 5 0 1
"0012400" "001240010" 1 0 0
"0012400" "001240010" 2 0 0

Thank you in advance,

Enrique

[Devra Golbe]

Hi Enrique,

Thank you for supplying some sample data.

list if inlist(household, "0012451", "0012452")

produces the following, indicating that your sample data contain two household-wave pairs where there is a wife but no husband.

     +--------------------------------------------------+
     | househ~d   person_id   wave   is_hus~d   is_wife |
     |--------------------------------------------------|
 86. |  0012451   001240005      5          0         1 |
 98. |  0012452   001240009      5          0         1 |
     +--------------------------------------------------+

When I run the following part of your code those two observations are dropped. Note that they satisfy the condition in the comment -- "at most one husband and one wife"-- though not the condition "exactly one husband and one wife" your code appears to check for.

// Checking the number of husbands and wives in each household and wave
bysort household_id wave: egen wives_per_wave = total(is_wife)
bysort household_id wave: egen husbands_per_wave = total(is_husband)

// Determining if there is at most one husband and one wife in each household
bys household_id: egen max_husbands = max(husbands_per_wave)
bys household_id: egen max_wives = max(wives_per_wave)
bys household_id: gen unique_husband = (max_husbands == 1)
bys household_id: gen unique_wife = (max_wives == 1)

bys household_id: keep if unique_husband == 1 & unique_wife == 1

pidlink is not defined, so I can't comment on that part of your code.

[George Ford]

drop if state=="Utah"

[Enrique Alameda]

Hi Enrique,

Thank you for supplying some sample data.

list if inlist(household, "0012451", "0012452")

produces the following, indicating that your sample data contain two household-wave pairs where there is a wife but no husband.

+--------------------------------------------------+
| househ~d person_id wave is_hus~d is_wife |
|--------------------------------------------------|
86. | 0012451 001240005 5 0 1 |
98. | 0012452 001240009 5 0 1 |
+--------------------------------------------------+

When I run the following part of your code those two observations are dropped. Note that they satisfy the condition in the comment -- "at most one husband and one wife"-- though not the condition "exactly one husband and one wife" your code appears to check for.

// Checking the number of husbands and wives in each household and wave
bysort household_id wave: egen wives_per_wave = total(is_wife)
bysort household_id wave: egen husbands_per_wave = total(is_husband)

// Determining if there is at most one husband and one wife in each household
bys household_id: egen max_husbands = max(husbands_per_wave)
bys household_id: egen max_wives = max(wives_per_wave)
bys household_id: gen unique_husband = (max_husbands == 1)
bys household_id: gen unique_wife = (max_wives == 1)

bys household_id: keep if unique_husband == 1 & unique_wife == 1

pidlink is not defined, so I can't comment on that part of your code.

Hi Devra,

Thank you so much for your help. Pidlink is actually person_id in the data example. Do you know how I could change the code to exactly one husband and one wife?

Thanks a lot,

Enrique

[George Ford]

bys household wave: egen husbandwife = total(is_husband + is_wife)

looking for a 2

[Enrique Alameda]

bys household wave: egen husbandwife = total(is_husband + is_wife)

looking for a 2

Hi George,

Thanks a lot, this is a smart and elegant solution.

However, some wives appear in certain waves where husbands do not (although both are present in the household). This causes the "distinct pidlink if is_husband == 1" to yield slightly different distinct values to "distinct pidlink if is_wife == 1" (i.e., there are still slightly more wives than husbands) upon keeping instances where husbandwife == 2. Do you know how I can fix this?

Thanks a lot,

Enrique.

[Devra Golbe]

Enrique,

I think your code gives exactly one husband and one wife. Indeed, when I run George's code after the code I quoted in my post husbandwife =2 for all observations. And when I run

distinct person_id if is_wife == 1
distinct person_id if is_husb == 1

I get identical results from each command.

Perhaps you need to supply a different sample of observations in order to illustrate the issues you are having.

[Enrique Alameda]

Enrique,

I think your code gives exactly one husband and one wife. Indeed, when I run George's code after the code I quoted in my post husbandwife =2 for all observations. And when I run

distinct person_id if is_wife == 1
distinct person_id if is_husb == 1

I get identical results from each command.

Perhaps you need to supply a different sample of observations in order to illustrate the issues you are having.

Hi Devra and George,

Here is a data example prior to running mine or George's code. I have changed notation to make it easier to understand:

clear
input str10(pidlink hhid) float(wave husband wife)
"272010001" "2720100" 1 1 0
"272010001" "2720100" 2 1 0
"272010001" "2720100" 3 1 0
"272010001" "2720100" 4 1 0
"272010001" "2720100" 5 1 0
"272010002" "2720100" 1 0 1
"272010002" "2720100" 2 0 1
"272010002" "2720100" 3 0 1
"272010002" "2720100" 4 0 1
"272010002" "2720151" 5 0 0
"272010003" "2720100" 1 0 0
"272010003" "2720100" 2 0 0
"272010003" "2720100" 3 0 0
"272010003" "2720100" 4 0 0
"272010003" "2720100" 5 0 0
"272010004" "2720100" 2 0 0
"272010004" "2720100" 3 0 0
"272010004" "2720100" 4 0 0
"272010004" "2720100" 5 0 0
"272010005" "2720100" 5 0 0
"272010006" "2720100" 5 0 1
"272020001" "2720200" 1 1 0
"272020001" "2720211" 2 1 0
"272020001" "2720211" 3 1 0
"272020001" "2720211" 4 1 0
"272020001" "2720211" 5 1 0
"272020002" "2720200" 1 0 1
"272020002" "2720211" 2 0 1
"272020002" "2720211" 3 0 1
"272020002" "2720211" 4 0 1
"272020002" "2720211" 5 0 1
"272020003" "2720200" 1 0 0
"272020003" "2720211" 2 0 0
"272020003" "2720211" 3 0 0
"272020003" "2720211" 4 0 0
"272020003" "2720211" 5 0 0
"272020004" "2720200" 2 1 0
"272020004" "2720200" 3 1 0
"272020004" "2720200" 4 1 0
"272020004" "2720200" 5 1 0
"272020005" "2720200" 2 0 1
"272020005" "2720200" 3 0 1
"272020005" "2720200" 4 0 1
"272020005" "2720200" 5 0 0
"272021104" "2720211" 2 0 0
"272021104" "2720211" 3 0 0
"272021104" "2720211" 4 0 0
"272021104" "2720211" 5 0 0
"272030001" "2720300" 1 1 0
"272030001" "2720300" 2 1 0
"272030001" "2720300" 3 1 0
"272030001" "2720300" 4 1 0
"272030001" "2720300" 5 1 0
"272030002" "2720300" 1 0 1
"272030002" "2720300" 2 0 1
"272030002" "2720300" 3 0 1
"272030002" "2720300" 4 0 1
"272030002" "2720300" 5 0 1
"272030003" "2720300" 1 0 0
"272030003" "2720300" 2 0 0
"272030003" "2720300" 3 0 0
"272030003" "2720341" 4 0 1
"272030004" "2720300" 2 0 0
"272030004" "2720300" 3 0 0
"272030004" "2720300" 4 0 0
"272030004" "2720300" 5 0 0
"272030005" "2720300" 3 0 0
"272030005" "2720300" 4 0 0
"272030005" "2720300" 5 0 0
"272034101" "2720341" 4 1 0
"272034101" "2720341" 5 0 0
"272034103" "2720341" 4 0 0
"272034104" "2720341" 5 1 0
"272040001" "2720400" 1 1 0
"272040001" "2720400" 2 1 0
"272040001" "2720400" 3 1 0
"272040001" "2720400" 4 1 0
"272040001" "2720400" 5 1 0
"272040002" "2720400" 1 0 1
"272040002" "2720400" 2 0 1
"272040002" "2720400" 3 0 1
"272040002" "2720400" 4 0 1
"272040002" "2720400" 5 0 1
"272040003" "2720400" 1 0 0
"272040003" "2720400" 2 0 0
"272040003" "2720400" 3 0 0
"272040003" "2720400" 4 0 0
"272040003" "2720451" 5 1 0
"272040004" "2720400" 1 0 0
"272040004" "2720400" 2 0 0
"272040004" "2720400" 3 0 0
"272040004" "2720400" 4 0 0
"272040004" "2720400" 5 0 0
"272040005" "2720400" 4 0 0
"272040005" "2720400" 5 0 0
"272050001" "2720500" 1 1 0
"272050001" "2720500" 2 1 0
"272050001" "2720500" 3 1 0
"272050001" "2720500" 4 1 0
"272050001" "2720500" 5 1 0
end


Upon running the code, if I do:

keep if husbandwife == 2
(41,234 observations deleted)

And then the distinct:

. distinct pidlink if wife == 1

| Observations
| total distinct
---------+----------------------
pidlink | 39619 15910

. distinct pidlink if husband == 1

| Observations
| total distinct
---------+----------------------
pidlink | 40025 16008



As you can see there is a different number of husband and wives. I provide a data example following this step below:

clear
input str10(pidlink hhid) float(wave husband wife)
"001060005" "0010600" 1 0 0
"001060006" "0010600" 1 0 0
"001060001" "0010600" 1 1 0
"001060002" "0010600" 1 0 1
"001060004" "0010600" 1 0 0
"001060003" "0010600" 1 0 0
"001060004" "0010600" 2 0 0
"001060002" "0010600" 2 0 1
"001060006" "0010600" 2 0 0
"001060005" "0010600" 2 0 0
"001060003" "0010600" 2 0 0
"001060001" "0010600" 2 1 0
"001065104" "0010651" 5 0 0
"001060004" "0010651" 5 0 1
"001065102" "0010651" 5 1 0
"001065103" "0010651" 5 0 0
"001080008" "0010800" 1 0 0
"001080006" "0010800" 1 0 0
"001080003" "0010800" 1 0 0
"001080005" "0010800" 1 0 0
"001080007" "0010800" 1 0 0
"001080009" "0010800" 1 0 0
"001080002" "0010800" 1 0 1
"001080010" "0010800" 1 0 0
"001080004" "0010800" 1 0 0
"001080001" "0010800" 1 1 0
"001080009" "0010800" 2 0 0
"001080008" "0010800" 2 0 0
"001080010" "0010800" 2 0 0
"001080001" "0010800" 2 1 0
"001080004" "0010800" 2 0 0
"001080006" "0010800" 2 0 0
"001080005" "0010800" 2 0 0
"001080003" "0010800" 2 0 0
"001080007" "0010800" 2 0 0
"001080002" "0010800" 2 0 1
"001085107" "0010851" 5 0 0
"001085104" "0010851" 5 0 0
"001085105" "0010851" 5 0 0
"001080012" "0010851" 5 0 1
"001080003" "0010851" 5 1 0
"001085106" "0010851" 5 0 0
"001085103" "0010851" 5 0 0
"001220009" "0012200" 1 0 0
"001220008" "0012200" 1 0 0
"001220002" "0012200" 1 0 1
"001220001" "0012200" 1 1 0
"001220003" "0012200" 1 0 0
"001220004" "0012200" 1 0 0
"001220006" "0012200" 1 0 0
"001220005" "0012200" 1 0 0
"001220010" "0012200" 1 0 0
"001220007" "0012200" 1 0 0
"001220010" "0012200" 2 0 0
"001220001" "0012200" 2 1 0
"001220004" "0012200" 2 0 0
"001220002" "0012200" 2 0 1
"001220008" "0012200" 2 0 0
"001220006" "0012200" 2 0 0
"001220009" "0012200" 2 0 0
"001220007" "0012200" 2 0 0
"001220005" "0012200" 2 0 0
"001220003" "0012200" 2 0 0
"001220009" "0012200" 3 0 0
"001220002" "0012200" 3 0 1
"001220003" "0012200" 3 0 0
"001220007" "0012200" 3 0 0
"001220001" "0012200" 3 1 0
"001220005" "0012200" 3 0 0
"001220008" "0012200" 3 0 0
"001220010" "0012200" 3 0 0
"001220006" "0012200" 3 0 0
"001220004" "0012200" 3 0 0
"001220004" "0012200" 4 0 0
"001220007" "0012200" 4 0 0
"001220006" "0012200" 4 0 0
"001220001" "0012200" 4 1 0
"001220005" "0012200" 4 0 0
"001220009" "0012200" 4 0 0
"001220002" "0012200" 4 0 1
"001220014" "0012200" 4 0 0
"001220015" "0012200" 4 0 0
"001220010" "0012200" 4 0 0
"001220004" "0012200" 5 0 0
"001220006" "0012200" 5 0 0
"001220007" "0012200" 5 0 0
"001220014" "0012200" 5 0 0
"001220002" "0012200" 5 0 1
"001220005" "0012200" 5 0 0
"001220015" "0012200" 5 0 0
"001220001" "0012200" 5 1 0
"001220010" "0012200" 5 0 0
"001220003" "0012241" 4 1 0
"001224104" "0012241" 4 0 0
"001224105" "0012241" 4 0 0
"001220011" "0012241" 4 0 1
"001220013" "0012241" 4 0 0
"001220003" "0012241" 5 1 0
"001220013" "0012241" 5 0 0
"001224105" "0012241" 5 0 0
end
[/CODE]


I am grateful for any help, as I cannot comprehend my mistake.

Thank you in advance,

Enrique

[George Ford]

I get the same with this.

bys hhid wave: egen husbandwife = total(husband + wife)

. keep if husbandwife==2
(3 observations deleted)

. 
. distinct pidlink if wife == 1


total   distinct

pidlink         12          6


. distinct pidlink if husband ==    1


total   distinct

pidlink         12          6


. 
. distinct hhid if wife == 1


total   distinct

hhid         12          6


. distinct hhid if husband == 1


total   distinct

hhid         12          6

[George Ford]

If the problem is missing data, then any solution is going to be a bit of a kluge.

One option is to assume that if it's 2 2 1 2 2, then it should be a 2 in between due to missing values. I'd have more confidence if the household member count followed the same pattern.

bys hhid wave: egen hhsize = count(hhid)

[Enrique Alameda]

If the problem is missing data, then any solution is going to be a bit of a kluge.

One option is to assume that if it's 2 2 1 2 2, then it should be a 2 in between due to missing values. I'd have more confidence if the household member count followed the same pattern.

bys hhid wave: egen hhsize = count(hhid)

Hi George,

Thanks a lot for all your help.

Here is the output from your code. Note that the dataset also contains sons and daughters of these couples (i.e., to identify child-wife pairs):

bys hhid wave: egen hhsize = count(hhid)

.
end of do-file

. ta hhsize

hhsize | Freq. Percent Cum.
------------+-----------------------------------
2 | 10,314 5.96 5.96
3 | 34,371 19.88 25.84
4 | 48,292 27.93 53.77
5 | 36,365 21.03 74.80
6 | 21,594 12.49 87.28
7 | 11,137 6.44 93.72
8 | 6,056 3.50 97.23
9 | 2,943 1.70 98.93
10 | 1,160 0.67 99.60
11 | 495 0.29 99.88
12 | 132 0.08 99.96
13 | 52 0.03 99.99
15 | 15 0.01 100.00
------------+-----------------------------------
Total | 172,926 100.00

[George Ford]

This might get you started. I've dropped the middle observation of wife for hhid==0012200 to create some interest.

I think the kids condition may be redundant in markreplace.

The math to address a change in kids should be easy enough to implement, but this might tell you how big the problem is.

clear
input str10(pidlink hhid) float(wave husband wife)
"001060005" "0010600" 1 0 0
"001060006" "0010600" 1 0 0
"001060001" "0010600" 1 1 0
"001060002" "0010600" 1 0 1
"001060004" "0010600" 1 0 0
"001060003" "0010600" 1 0 0
"001060004" "0010600" 2 0 0
"001060002" "0010600" 2 0 1
"001060006" "0010600" 2 0 0
"001060005" "0010600" 2 0 0
"001060003" "0010600" 2 0 0
"001060001" "0010600" 2 1 0
"001065104" "0010651" 5 0 0
"001060004" "0010651" 5 0 1
"001065102" "0010651" 5 1 0
"001065103" "0010651" 5 0 0
"001080008" "0010800" 1 0 0
"001080006" "0010800" 1 0 0
"001080003" "0010800" 1 0 0
"001080005" "0010800" 1 0 0
"001080007" "0010800" 1 0 0
"001080009" "0010800" 1 0 0
"001080002" "0010800" 1 0 1
"001080010" "0010800" 1 0 0
"001080004" "0010800" 1 0 0
"001080001" "0010800" 1 1 0
"001080009" "0010800" 2 0 0
"001080008" "0010800" 2 0 0
"001080010" "0010800" 2 0 0
"001080001" "0010800" 2 1 0
"001080004" "0010800" 2 0 0
"001080006" "0010800" 2 0 0
"001080005" "0010800" 2 0 0
"001080003" "0010800" 2 0 0
"001080007" "0010800" 2 0 0
"001080002" "0010800" 2 0 1
"001085107" "0010851" 5 0 0
"001085104" "0010851" 5 0 0
"001085105" "0010851" 5 0 0
"001080012" "0010851" 5 0 1
"001080003" "0010851" 5 1 0
"001085106" "0010851" 5 0 0
"001085103" "0010851" 5 0 0
"001220009" "0012200" 1 0 0
"001220008" "0012200" 1 0 0
"001220002" "0012200" 1 0 1
"001220001" "0012200" 1 1 0
"001220003" "0012200" 1 0 0
"001220004" "0012200" 1 0 0
"001220006" "0012200" 1 0 0
"001220005" "0012200" 1 0 0
"001220010" "0012200" 1 0 0
"001220007" "0012200" 1 0 0
"001220010" "0012200" 2 0 0
"001220001" "0012200" 2 1 0
"001220004" "0012200" 2 0 0
"001220002" "0012200" 2 0 1
"001220008" "0012200" 2 0 0
"001220006" "0012200" 2 0 0
"001220009" "0012200" 2 0 0
"001220007" "0012200" 2 0 0
"001220005" "0012200" 2 0 0
"001220003" "0012200" 2 0 0
"001220009" "0012200" 3 0 0
/*"001220002" "0012200" 3 0 1*/
"001220003" "0012200" 3 0 0
"001220007" "0012200" 3 0 0
"001220001" "0012200" 3 1 0
"001220005" "0012200" 3 0 0
"001220008" "0012200" 3 0 0
"001220010" "0012200" 3 0 0
"001220006" "0012200" 3 0 0
"001220004" "0012200" 3 0 0
"001220004" "0012200" 4 0 0
"001220007" "0012200" 4 0 0
"001220006" "0012200" 4 0 0
"001220001" "0012200" 4 1 0
"001220005" "0012200" 4 0 0
"001220009" "0012200" 4 0 0
"001220002" "0012200" 4 0 1
"001220014" "0012200" 4 0 0
"001220015" "0012200" 4 0 0
"001220010" "0012200" 4 0 0
"001220004" "0012200" 5 0 0
"001220006" "0012200" 5 0 0
"001220007" "0012200" 5 0 0
"001220014" "0012200" 5 0 0
"001220002" "0012200" 5 0 1
"001220005" "0012200" 5 0 0
"001220015" "0012200" 5 0 0
"001220001" "0012200" 5 1 0
"001220010" "0012200" 5 0 0
"001220003" "0012241" 4 1 0
"001224104" "0012241" 4 0 0
"001224105" "0012241" 4 0 0
"001220011" "0012241" 4 0 1
"001220013" "0012241" 4 0 0
"001220003" "0012241" 5 1 0
"001220013" "0012241" 5 0 0
"001224105" "0012241" 5 0 0
end

bys hhid wave: egen husbandwife = total(husband + wife)
bys hhid wave: egen hhsize = count(hhid)
g kids = hhsize - husbandwife

bys hhid: egen max_husbandwife = max(husbandwife)
keep if max_husbandwife==2 // delete cases not desired

bys hhid: egen max_hhsize = max(hhsize)
bys hhid: egen max_kids = max(kids)
bys hhid: egen min_wave = min(wave)
bys hhid: egen max_wave = max(wave)

g markreplace = (husbandwife - max_husbandwife)==-1 & (hhsize - max_hhsize)==-1 & (kids==max_kids) & (wave != min_wave) & (wave != !max_wave)
* wave included as condition since you don't have data to either side to confirm.
g markcheck = (husbandwife - max_husbandwife)==-1 & (hhsize - max_hhsize)==-1 & (kids!=max_kids)
replace husbandwife = 2 if markreplace

[Devra Golbe]

Enrique,

The data sample you provide cannot be the data sample on which you ran your code (> 40,000 observations deleted.) Try to extract a small set of data which produces the phenomenon that concerns you: more distinct wives than husbands.

I suspect also that you will get better answers if you describe your data more fully. I guess that you are using some kind of household sample which is resurveyed multiple times (waves.) I don't know anything about this sample-- even how a household is defined. I would not be surprised if the number of children changes over time in the same HH ID. But suppose a head of household changes spouses. Does the HH ID change or stay the same? If the spouse who is not designated head of HH gets a new spouse does he drop out of the survey or get a new HH ID? I'm not certain, given the possible changes in HH composition, whether the numbers of distinct husbands & wives should be the same.