Mahapick with foreach loop

[Edgar Goldemer]

Dear Statalists,

I want to find several (5) company matches based on financial data, always for a given year and a given industry. The industry is expressed as a code ranging from 100 to 9997 in one-unit steps. When there are less than 5 matching observations for a company, I would like to run mahapick again, but then searching for companies in the next industry (industry code + 1). If there are still less than 5 matches, the search shall be repeated with industry code + 2, and so on...

Moreover, the "treated companies" are listed twice in my dataset, both as treated and not treated, because for other treated companies they might serve as a good match. Therefore, running mahapick leads to the result that many of the first matches are companies with themselves.

If someone has a good idea, please help me out!

Below you see an dataex example and the coding so far. I´m working with MP 15.0.

clear
input float id byte pair_id int year byte(treat industry) int assets float(bm_ratio match_id_1 match_id_2 match_id_3 match_id_4 match_id_5 match_id_6)
 1  1 2015 1 10 100       .1 1 9 5 7  3 .
 1  2 2016 1 10 110      .11 1 9 5 3  7 .
 1  3 2017 1 10 120      .12 1 9 3 5  7 .
 1  4 2018 1 10 130      .13 1 9 3 7  5 .
 1  5 2019 1 10 140      .14 1 9 3 5  7 .
 2  6 2015 1 20 200       .2 2 4 8 6 10 .
 2  7 2016 1 20 210      .21 2 4 6 8 10 .
 2  8 2017 1 20 220      .22 2 4 8 6 10 .
 2  9 2018 1 20 230      .23 2 6 8 4 10 .
 2 10 2019 1 20 240      .24 2 6 4 8 10 .
 1  . 2015 0 10 100       .1 . . . .  . .
 1  . 2016 0 10 110      .11 . . . .  . .
 1  . 2017 0 10 120      .12 . . . .  . .
 1  . 2018 0 10 130      .13 . . . .  . .
 1  . 2019 0 10 140      .14 . . . .  . .
 2  . 2015 0 20 200       .2 . . . .  . .
 2  . 2016 0 20 210      .21 . . . .  . .
 2  . 2017 0 20 220      .22 . . . .  . .
 2  . 2018 0 20 230      .23 . . . .  . .
 2  . 2019 0 20 240      .24 . . . .  . .
 3  . 2015 0 10 105      .98 . . . .  . .
 3  . 2016 0 10 115     .101 . . . .  . .
 3  . 2017 0 10 125      .12 . . . .  . .
 3  . 2018 0 10 135      .15 . . . .  . .
 3  . 2019 0 10 145      .12 . . . .  . .
 4  . 2015 0 20 205     .202 . . . .  . .
 4  . 2016 0 20 215     .203 . . . .  . .
 4  . 2017 0 20 225      .23 . . . .  . .
 4  . 2018 0 20 235      .21 . . . .  . .
 4  . 2019 0 20 245      .23 . . . .  . .
 5  . 2015 0 10 130  .111111 . . . .  . .
 5  . 2016 0 10 120 .1099999 . . . .  . .
 5  . 2017 0 10 140  .132222 . . . .  . .
 5  . 2018 0 10 180  .102222 . . . .  . .
 5  . 2019 0 10 150      .12 . . . .  . .
 6  . 2015 0 20 290      .23 . . . .  . .
 6  . 2016 0 20 250      .23 . . . .  . .
 6  . 2017 0 20 340      .24 . . . .  . .
 6  . 2018 0 20 240      .25 . . . .  . .
 6  . 2019 0 20 250      .24 . . . .  . .
 7  . 2015 0 10  90      .13 . . . .  . .
 7  . 2016 0 10 105      .14 . . . .  . .
 7  . 2017 0 10  80      .14 . . . .  . .
 7  . 2018 0 10  90      .15 . . . .  . .
 7  . 2019 0 10 110      .15 . . . .  . .
 8  . 2015 0 20 190      .24 . . . .  . .
 8  . 2016 0 20 170      .24 . . . .  . .
 8  . 2017 0 20 200  .245555 . . . .  . .
 8  . 2018 0 20 205      .23 . . . .  . .
 8  . 2019 0 20 206  .243333 . . . .  . .
 9  . 2015 0 10  98      .09 . . . .  . .
 9  . 2016 0 10 101     .093 . . . .  . .
 9  . 2017 0 10 100     .096 . . . .  . .
 9  . 2018 0 10 104     .098 . . . .  . .
 9  . 2019 0 10 105     .104 . . . .  . .
10  . 2015 0 20 197      .08 . . . .  . .
10  . 2016 0 20 199     .089 . . . .  . .
10  . 2017 0 20 201     .092 . . . .  . .
10  . 2018 0 20 200     .095 . . . .  . .
10  . 2019 0 20 208     .099 . . . .  . .
11  . 2015 0 11 101      .11 . . . .  . .
11  . 2016 0 11 102      .12 . . . .  . .
11  . 2017 0 11 103   .13444 . . . .  . .
11  . 2018 0 11 102  .135666 . . . .  . .
11  . 2019 0 11 104     .147 . . . .  . .
12  . 2015 0 21 206      .23 . . . .  . .
12  . 2016 0 21 218      .23 . . . .  . .
12  . 2017 0 21 198     .235 . . . .  . .
12  . 2018 0 21 212     .237 . . . .  . .
12  . 2019 0 21 220     .239 . . . .  . .
13  . 2015 0 12 130    .1134 . . . .  . .
13  . 2016 0 12 128    .1139 . . . .  . .
13  . 2017 0 12 134    .1142 . . . .  . .
13  . 2018 0 12 138    .1134 . . . .  . .
13  . 2019 0 12 140    .1156 . . . .  . .
14  . 2015 0 22 230     .212 . . . .  . .
14  . 2016 0 22 228     .219 . . . .  . .
14  . 2017 0 22 220     .221 . . . .  . .
14  . 2018 0 22 231     .229 . . . .  . .
14  . 2019 0 22 235      .24 . . . .  . .
end


gen match_id_1 = .
gen match_id_2 = .
gen match_id_3 = .
gen match_id_4 = .
gen match_id_5 = .
gen match_id_6 = .
recast float id
mahapick assets bm_ratio, idvar(id) pickids(match_id_1 match_id_2 match_id_3 match_id_4 match_id_5 match_id_6) treated(treat) matchon(year industry)

[Edgar Goldemer]

Hey Statalists,

i made some progress on my issue, but it´s still not solved. I added the variable "match_industry" that increases by one every time mahapick hasn´t found enough matches and the variable "start_missing" that displays the digit of the first match_id thatl is still missing.
Below in the last line of the foreach loop, how can I ensure, that Stata for each observation starts with mahapick from the match_id that was the first missing in the last mahapick match?

clear
input byte(id pair_id) int year byte(treat industry) int assets float bm_ratio
 1  1 2015 1 10 100       .1
 1  2 2016 1 10 110      .11
 1  3 2017 1 10 120      .12
 1  4 2018 1 10 130      .13
 1  5 2019 1 10 140      .14
 2  6 2015 1 20 200       .2
 2  7 2016 1 20 210      .21
 2  8 2017 1 20 220      .22
 2  9 2018 1 20 230      .23
 2 10 2019 1 20 240      .24
 1  . 2015 0 10 100       .1
 1  . 2016 0 10 110      .11
 1  . 2017 0 10 120      .12
 1  . 2018 0 10 130      .13
 1  . 2019 0 10 140      .14
 2  . 2015 0 20 200       .2
 2  . 2016 0 20 210      .21
 2  . 2017 0 20 220      .22
 2  . 2018 0 20 230      .23
 2  . 2019 0 20 240      .24
 3  . 2015 0 10 105      .98
 3  . 2016 0 10 115     .101
 3  . 2017 0 10 125      .12
 3  . 2018 0 10 135      .15
 3  . 2019 0 10 145      .12
 4  . 2015 0 20 205     .202
 4  . 2016 0 20 215     .203
 4  . 2017 0 20 225      .23
 4  . 2018 0 20 235      .21
 4  . 2019 0 20 245      .23
 5  . 2015 0 10 130  .111111
 5  . 2016 0 10 120 .1099999
 5  . 2017 0 10 140  .132222
 5  . 2018 0 10 180  .102222
 5  . 2019 0 10 150      .12
 6  . 2015 0 20 290      .23
 6  . 2016 0 20 250      .23
 6  . 2017 0 20 340      .24
 6  . 2018 0 20 240      .25
 6  . 2019 0 20 250      .24
 7  . 2015 0 10  90      .13
 7  . 2016 0 10 105      .14
 7  . 2017 0 10  80      .14
 7  . 2018 0 10  90      .15
 7  . 2019 0 10 110      .15
 8  . 2015 0 20 190      .24
 8  . 2016 0 20 170      .24
 8  . 2017 0 20 200  .245555
 8  . 2018 0 20 205      .23
 8  . 2019 0 20 206  .243333
 9  . 2015 0 10  98      .09
 9  . 2016 0 10 101     .093
 9  . 2017 0 10 100     .096
 9  . 2018 0 10 104     .098
 9  . 2019 0 10 105     .104
10  . 2015 0 20 197      .08
10  . 2016 0 20 199     .089
10  . 2017 0 20 201     .092
10  . 2018 0 20 200     .095
10  . 2019 0 20 208     .099
11  . 2015 0 11 101      .11
11  . 2016 0 11 102      .12
11  . 2017 0 11 103   .13444
11  . 2018 0 11 102  .135666
11  . 2019 0 11 104     .147
12  . 2015 0 21 206      .23
12  . 2016 0 21 218      .23
12  . 2017 0 21 198     .235
12  . 2018 0 21 212     .237
12  . 2019 0 21 220     .239
13  . 2015 0 12 130    .1134
13  . 2016 0 12 128    .1139
13  . 2017 0 12 134    .1142
13  . 2018 0 12 138    .1134
13  . 2019 0 12 140    .1156
14  . 2015 0 22 230     .212
14  . 2016 0 22 228     .219
14  . 2017 0 22 220     .221
14  . 2018 0 22 231     .229
14  . 2019 0 22 235      .24
end



gen match_id_1 = .
gen match_id_2 = .
gen match_id_3 = .
gen match_id_4 = .
gen match_id_5 = .
gen match_id_6 = .
gen match_industry = industry
gen start_missing = 0
recast float id
mahapick assets bm_ratio, idvar(id) pickids(match_id_1 match_id_2 match_id_3 match_id_4 match_id_5 match_id_6) treated(treat) matchon(year industry)
foreach x in start_missing {
replace start_missing = cond(match_id_1==.,1,cond(match_id_2==.,2,cond(match_id_3==.,3,cond(match_id_4==.,4,cond(match_id_5==.,5,cond(match_id_6==.,6,0)))))) if treat == 1
replace match_industry = match_industry+1 if `x' >0 & treat == 1
mahapick assets bm_ratio, idvar(id) pickids(match_id_`x' UNTIL match_id_6) treated(treat) matchon(year match_industry)
}

[Phil Bromiley]

If your industry is an SIC code, then your method is flawed. Industries break at the 1000's so 1999 is not at all like 2000. What folks would normally do is move to a higher level (e.g., 3 digit sic code) level. I'm not sure how other industry codes work.