Different Confidence Intervals with ID Fixed Effects.

[Jose Ignacio Torrealba]

Hello!

Im posting this because I need help understanding this figure:

(Im using STATA /SE 18.0)

* Example generated by -dataex-. For more info, type help dataex
clear
input float(mmm_lag12 mmm_lag11 mmm_lag10 mmm_lead12 mmm_lead11 mmm_lead10 migrante post) byte cotiza_pension double idrph float TT
0 0 0 0 0 0 0 0 . 1535675 -12
0 0 0 0 0 0 0 0 . 191661279 -12
0 0 0 0 0 0 0 0 . 216373956 -12
0 0 0 0 0 0 0 0 . 1101216208 -12
0 0 0 0 0 0 0 0 1 294155744 -12
0 0 0 0 0 0 0 0 1 1272437932 -12
0 0 0 0 0 0 0 0 1 327913644 -12
0 0 0 0 0 0 0 0 . 851797487 -12
0 0 0 0 0 0 0 0 . 653799035 -12
0 0 0 0 0 0 0 0 1 1366324161 -12
0 0 0 0 0 0 0 0 1 536703141 -12
0 0 0 0 0 0 0 0 . 331497713 -12
0 0 0 0 0 0 0 0 . 1006721492 -12
0 0 0 0 0 0 0 0 1 828891842 -12
0 0 0 0 0 0 0 0 . 401513138 -12
0 0 0 0 0 0 0 0 1 641707372 -12
0 0 0 0 0 0 0 0 1 1265778039 -12
0 0 0 0 0 0 0 0 1 386102030 -12
0 0 0 0 0 0 0 0 0 939707001 -12
0 0 0 0 0 0 0 0 1 995646170 -12
1 0 0 0 0 0 1 0 . 1184902819 -12
0 0 0 0 0 0 0 0 1 1337717970 -12
0 0 0 0 0 0 0 0 . 181896984 -12
1 0 0 0 0 0 1 0 . 745607211 -12
0 0 0 0 0 0 0 0 . 1211121027 -12
0 0 0 0 0 0 0 0 . 378703313 -12
0 0 0 0 0 0 0 0 1 442076797 -12
0 0 0 0 0 0 0 0 1 616110332 -12
0 0 0 0 0 0 0 0 . 1300390826 -12
0 0 0 0 0 0 0 0 1 1277958431 -12
0 0 0 0 0 0 0 0 1 682110931 -12
0 0 0 0 0 0 0 0 . 38774903 -12
0 0 0 0 0 0 0 0 . 292233846 -12
0 0 0 0 0 0 0 0 . 1144728268 -12
1 0 0 0 0 0 1 0 . 915522803 -12
0 0 0 0 0 0 0 0 0 1226553104 -12
0 0 0 0 0 0 0 0 1 1368387141 -12
0 0 0 0 0 0 0 0 . 1169104253 -12
0 0 0 0 0 0 0 0 1 142489124 -12
0 0 0 0 0 0 0 0 . 863020048 -12
0 0 0 0 0 0 0 0 1 775285234 -12
0 0 0 0 0 0 0 0 1 292467595 -12
0 0 0 0 0 0 0 0 1 291249966 -12
0 0 0 0 0 0 0 0 0 761800780 -12
0 0 0 0 0 0 0 0 . 1246822159 -12
0 0 0 0 0 0 0 0 . 1185440791 -12
0 0 0 0 0 0 0 0 . 419265334 -12
0 0 0 0 0 0 0 0 . 252458391 -12
0 0 0 0 0 0 0 0 . 265554605 -12
0 0 0 0 0 0 0 0 . 435686873 -12
0 0 0 0 0 0 0 0 . 877092537 -12
0 0 0 0 0 0 0 0 . 862236040 -12
0 0 0 0 0 0 0 0 1 425819289 -12
0 0 0 0 0 0 0 0 . 821155020 -12
0 0 0 0 0 0 0 0 0 1190318874 -12
0 0 0 0 0 0 0 0 . 582105161 -12
1 0 0 0 0 0 1 0 0 310385353 -12
0 0 0 0 0 0 0 0 . 769567464 -12
0 0 0 0 0 0 0 0 0 163268899 -12
0 0 0 0 0 0 0 0 1 1141060852 -12
0 0 0 0 0 0 0 0 . 1082744188 -12
0 0 0 0 0 0 0 0 . 1440324312 -12
0 0 0 0 0 0 0 0 1 584139900 -12
0 0 0 0 0 0 0 0 1 973942267 -12
0 0 0 0 0 0 0 0 . 1131004703 -12
0 0 0 0 0 0 0 0 . 1449672384 -12
0 0 0 0 0 0 0 0 1 433439596 -12
0 0 0 0 0 0 0 0 1 1271370829 -12
0 0 0 0 0 0 0 0 1 415779682 -12
0 0 0 0 0 0 0 0 . 465952258 -12
0 0 0 0 0 0 0 0 . 806267061 -12
0 0 0 0 0 0 0 0 . 259353393 -12
0 0 0 0 0 0 0 0 . 601893986 -12
0 0 0 0 0 0 0 0 . 583509185 -12
0 0 0 0 0 0 0 0 . 309350474 -12
0 0 0 0 0 0 0 0 . 927274273 -12
0 0 0 0 0 0 0 0 1 1420772545 -12
0 0 0 0 0 0 0 0 . 501346490 -12
0 0 0 0 0 0 0 0 . 1311890685 -12
0 0 0 0 0 0 0 0 . 1060799671 -12
0 0 0 0 0 0 0 0 . 1020012690 -12
0 0 0 0 0 0 0 0 1 901744789 -12
0 0 0 0 0 0 0 0 . 346203046 -12
0 0 0 0 0 0 0 0 . 1120106474 -12
1 0 0 0 0 0 1 0 . 29401317 -12
0 0 0 0 0 0 0 0 . 12952474 -12
1 0 0 0 0 0 1 0 0 288469124 -12
0 0 0 0 0 0 0 0 . 1252650566 -12
0 0 0 0 0 0 0 0 . 1442339449 -12
0 0 0 0 0 0 0 0 0 399176808 -12
0 0 0 0 0 0 0 0 . 741347426 -12
0 0 0 0 0 0 0 0 1 1423505443 -12
0 0 0 0 0 0 0 0 . 555861467 -12
0 0 0 0 0 0 0 0 . 1141650802 -12
0 0 0 0 0 0 0 0 1 345798408 -12
0 0 0 0 0 0 0 0 1 913189941 -12
0 0 0 0 0 0 0 0 . 37700077 -12
0 0 0 0 0 0 0 0 1 836137025 -12
0 0 0 0 0 0 0 0 . 121543511 -12
0 0 0 0 0 0 0 0 . 688041026 -12
end
label values cotiza_pension snna_val
label def snna_val 1 "Sí", modify
[/CODE]
------------------ copy up to and including the previous line ------------------

Listed 100 out of 280743 observations
Use the count() option to list more



I use same data for this 2 graphs. The only thing that is changing is the specification.
1. Fixed Effect ID (idrph)
2. Whithout FE

Any ideas of why is happening the change in CI size? the patron of "2 positive, 1 negative"?
For the size of CI i try using e(sample) to check the specification with the long CI v/s shorts, but there is no change in obs. Maybe missing values? Or other exercise to check that?

I would appreciate any idea with this

[Jose Ignacio Torrealba]

Here is the CODE:

//Estimation FE: mes encuesta & ID


use "/Users/macbookair/Library/CloudStorage/OneDrive-UniversidadCatólicadeChile/Trabajo/Investigación/Migración/Base de Datos Completa (raw)/ene_2021-2023_ES_mensual.dta", clear

global out_emp trabajo cotiza_pension
local demograph sexo edad nivel
rename migrante_post migrante_post_mmm


foreach var in $out_emp {
qui reghdfe `var' mmm_lead0 mmm_lag2 mmm_lag3 mmm_lag4 mmm_lag5 mmm_lag6 mmm_lag7 mmm_lag8 mmm_lag9 mmm_lag10 mmm_lag11 mmm_lag12 mmm_lead1 mmm_lead2 mmm_lead3 mmm_lead4 mmm_lead5 mmm_lead6 mmm_lead7 mmm_lead8 mmm_lead9 mmm_lead10 mmm_lead11 mmm_lead12 migrante if TT>-13 & TT < 13, absorb(`demograph' mes_encuesta comuna idrph) cl(comuna)

}

//Estimation NO FE ID

use "/Users/macbookair/Library/CloudStorage/OneDrive-UniversidadCatólicadeChile/Trabajo/Investigación/Migración/Base de Datos Completa (raw)/ene_2021-2023_ES_mensual.dta", clear



global out_emp trabajo cotiza_pension
local demograph sexo edad nivel
rename migrante_post migrante_post_mmm


foreach var in $out_emp {
qui reghdfe `var' mmm_lead0 mmm_lag2 mmm_lag3 mmm_lag4 mmm_lag5 mmm_lag6 mmm_lag7 mmm_lag8 mmm_lag9 mmm_lag10 mmm_lag11 mmm_lag12 mmm_lead1 mmm_lead2 mmm_lead3 mmm_lead4 mmm_lead5 mmm_lead6 mmm_lead7 mmm_lead8 mmm_lead9 mmm_lead10 mmm_lead11 mmm_lead12 migrante if TT>-13 & TT < 13, absorb(`demograph' mes_encuesta comuna) cl(comuna)

}

[Andrew Musau]

Why would you expect the standard errors (and coefficients) to be the same or even similar? You are estimating different models. Better to choose the correct model and stick with that.

[Jose Ignacio Torrealba]

Why would you expect the standard errors (and coefficients) to be the same or even similar? You are estimating different models. Better to choose the correct model and stick with that.

Andrew Musau

Thank u Andrew. I just want to check what im doign is right, or maybe is something wrong in the construction of the Data. For example, is there any way to see how many obs are being estimated for each lag? how many missing values ? (For example: fig 1: i would expect to find a difference in obs. between -10 and -9, but i dont know how to check that)

[Andrew Musau]

I do not follow what you mean by

difference in obs. between -10 and -9

As this question does not appear related to graphing, I suggest that you start a new thread and provide more details.