Extremely large coefficient

[Filmon Kidane]

Hi,
I am in the process of investigating (for a thesis) on how average income can be / is impacted by the constant inflow of immigrants into a country. I have average income data on all provinces as well as shift-share of immigrants (for both sex) across 2013-2022. While the average income is in hundreds of thousands, my shift-share is in decimal points. I performed panel data regression on average income including other independent variables beside shift-share, but the coefficient of the shift-share I have gotten is astronomically large which I cannot make sense of it. I am looking for guidance on how I can approach this issue, and any tips would be handy, thanks!

[George Ford]

Standard practice is to use ln(average income) as the dependent variable. That will fix your problem and make the coefficient have a sensible interpretation. (%change in income given a percentage point change in share. You can use margins to get elasticity).

The problem with using levels is that you get a "mean" effect on the coefficient, at the mean may be larger than average income for some members of the sample.

[Filmon Kidane]

Thank you for the response, George. The problem is that I am facing the same issue even when I transform my variable average income into logarithm form. Tried a couple of things to see if i have misspecification issue but so far no luck.

[George Ford]

You have fixed effects for country?

[Filmon Kidane]

The data I have is for Denmark.

[Filmon Kidane]

Here is a glimpse of the data I have with regards to average income and shift-share for male
Province Year Average income SS_M
"PCC" 2013 337639 .00076
"PCC" 2014 344384 .0011
"PCC" 2015 353485 .00162
"PCC" 2016 362413 -.00104
"PCC" 2017 372443 -.001
"PCC" 2018 385552 -.00036
"PCC" 2019 392441 .00006
"PCC" 2020 415306 -.00098
"PCC" 2021 423726 .00012
"PCC" 2022 438224 .00356
"PCA" 2013 391499 .000686230552207768
"PCA" 2014 397910 .001013596704871724
"PCA" 2015 408161 .0015457631147393956
"PCA" 2016 413425 -.0010331644745475625
"PCA" 2017 419996 -.0010390775996814703
"PCA" 2018 429152 -.0003867465675674716
"PCA" 2019 433814 .00006390682285565869
"PCA" 2020 455120 -.0010798979620496095
"PCA" 2021 463258 .00013768990664465847
"PCA" 2022 473808 .004128844331519971
"PNZ" 2013 429648 .0003094505530339657
"PNZ" 2014 435857 .000466584729375131
"PNZ" 2015 445827 .0007599083968506439
"PNZ" 2016 448280 -.0005238051296756172
"PNZ" 2017 451584 -.000519118799545795
"PNZ" 2018 462121 -.00019165480601086228
"PNZ" 2019 463925 .000031089450912855137
"PNZ" 2020 487227 -.0005175072259926876
"PNZ" 2021 494522 .00006335767622219998
"PNZ" 2022 502521 .0018757523505842607
"PB" 2013 260309 .00012358869722133508
"PB" 2014 265995 .00020264423475218433
"PB" 2015 276367 .0003961055478484748
"PB" 2016 276603 -.00028886598610145014
"PB" 2017 282598 -.00029618322769119293
"PB" 2018 293672 -.00011067409332641549
"PB" 2019 286506 .000018052423367683287
"PB" 2020 297557 -.000292737531698579
"PB" 2021 298592 .000036401804651967235
"PB" 2022 309923 .001089058614636829
"PEZ" 2013 380571 .0003232229693875865
"PEZ" 2014 385215 .0004906784158649971
"PEZ" 2015 394276 .0007917585145957408
"PEZ" 2016 400507 -.0005360043342380299
"PEZ" 2017 407111 -.0005236054293812254
"PEZ" 2018 418391 -.0001922016592759227
"PEZ" 2019 420406 .00003154600650332961
"PEZ" 2020 446642 -.0005307084784381331
"PEZ" 2021 447593 .00006566477516925226
"PEZ" 2022 456970 .001993057346005828
"PWSZ" 2013 308935 .00025
"PWSZ" 2014 315513 .00038
"PWSZ" 2015 323197 .00064
"PWSZ" 2016 327101 -.00045
"PWSZ" 2017 332944 -.00044
"PWSZ" 2018 341831 -.00016
"PWSZ" 2019 343200 .00003
"PWSZ" 2020 363087 -.00043
"PWSZ" 2021 365495 .00005
"PWSZ" 2022 375601 .00158
"PF" 2013 298486 .0003319050026183776
"PF" 2014 305412 .0005001948324480789
"PF" 2015 311842 .0007985081569473459
"PF" 2016 315506 -.0005416040798654688
"PF" 2017 320698 -.0005278528938937743
"PF" 2018 331096 -.0001929149880995439
"PF" 2019 332542 .00003110190786916103
"PF" 2020 350593 -.000510943167329122
"PF" 2021 354370 .00006256227287905896
"PF" 2022 366195 .0018336153644756272
"PSJ" 2013 317624 .0003032423437646849
"PSJ" 2014 325445 .00045871571071148923
"PSJ" 2015 333137 .0007455511650518901
"PSJ" 2016 339167 -.0005129785836156105
"PSJ" 2017 345501 -.0005048792341583758
"PSJ" 2018 354252 -.00018642195294759853
"PSJ" 2019 355988 .00003056991663826257
"PSJ" 2020 374018 -.0005155872242686359
"PSJ" 2021 382676 .00006462658919457169
"PSJ" 2022 393944 .001947324091091339
"PEJ" 2013 318089 .0003513502705178821
"PEJ" 2014 325765 .0005208560433512126
"PEJ" 2015 333562 .0008220077975867973
"PEJ" 2016 338511 -.0005549730283112333
"PEJ" 2017 346611 -.0005487273964922709
"PEJ" 2018 356381 -.00019959740587580042
"PEJ" 2019 358598 .00003237971139437413
"PEJ" 2020 378460 -.0005333326806562341
"PEJ" 2021 385640 .00006567528642814265
"PEJ" 2022 400627 .00195741840957212
"PWJ" 2013 303624 .00023743869668571368
"PWJ" 2014 311619 .00036552986927307617
"PWJ" 2015 320044 .0006125858675777523
"PWJ" 2016 325736 -.00043365079504739814
"PWJ" 2017 330832 -.0004358822681560128
"PWJ" 2018 340009 -.00016083665049098188
"PWJ" 2019 340835 .000026137490525394816
"PWJ" 2020 358475 -.0004390260180741426
"PWJ" 2021 366523 .00005485026827635915
"PWJ" 2022 374489 .0016747602453279266

[George Ford]

g lincome = ln(income)
reghdfe lincome ssm , absorb(province year)

coefficent is 1.27.

[Filmon Kidane]

Thank you very much, George. This last bit was very helpful.

[George Ford]

If you drop the provence fixed effect, you get the big coefficient.

[Filmon Kidane]

I intend to keep province; it is looking good now. I appreciate the help.