Hello,

I am currently writing my undergrad dissertation, titled "The Macroeconomic Determinants of Mental Health". I originally used GDPpc as one of my explanatory variables, however my supervisor advised me to use lnGDPpc instead. But when I make this change, the significance of my results change dramatically.

My previous code was:
Code:
 xtreg DepressionPrevM GDPpc Gini Unemp Inflation FLaborforce GovExpE HealthExp i.Year, fe cluster(CountryNum)
estimates store OMAllCountriesDY
xtreg DepressionPrevM GDPpc Gini Unemp Inflation FLaborforce GovExpE HealthExp i.Year if EconomicRegion2019=="H", fe cluster(CountryNum)
estimates store OMHIncomeDY
estimates table OMAllCountriesDY OMHIncomeDY, star(0.1 0.05 0.01) stats(N r2)
This gave the results:
----------------------------------------------
Variable | OMAllCountr~Y OMHIncomeDY
-------------+--------------------------------
GDPpc | -1.623e-08 9.117e-09
Gini | -.00002133 -.00002211
Unemp | .00006135** .00008061**
Inflation | -4.896e-06 7.371e-06
FLaborforce | -.0000146 -.00001161
GovExpE | -.00010854 .00001162
HealthExp | .00003076 .00003139
|
Year |
2001 | -.00003004 -6.992e-06
2002 | -.00011852 -.00025828
2003 | -.00016475 -.00048657*
2004 | -.00011298 -.00070361**
2005 | -.00017278 -.00083285**
2006 | -.00013591 -.00082165*
2007 | -.0001865 -.00091721**
2008 | -.00024751 -.0010785**
2009 | -.00045626** -.00130067***
2010 | -.00056189** -.00145002***
2011 | -.00055431** -.00146602***
2012 | -.0005653** -.00155055***
2013 | -.00060111** -.00162533***
2014 | -.00056984** -.0016286***
2015 | -.0005531* -.00158836***
2016 | -.00047691 -.00144049**
2017 | -.00046323 -.00149202**
2018 | -.00006073 .00048034
|
_cons | .0226525*** .02112168***
-------------+--------------------------------
N | 1604 606
r2 | .12887502 .27891163
----------------------------------------------
legend: * p<.1; ** p<.05; *** p<.01


Sorry if the formatting is not clear, but as you can see, the significance on the unemployment variables are significant at the 90% and 95% significant levels respectively. However, when I use lnGDPpc instead:

Code:
 xtreg DepressionPrevM lnGDPpc Gini Unemp Inflation FLaborforce GovExpE HealthExp i.Year, fe cluster(CountryNum)
estimates store MAllCountriesDY
xtreg DepressionPrevM lnGDPpc Gini Unemp Inflation FLaborforce GovExpE HealthExp i.Year if EconomicRegion2019=="H", fe cluster(CountryNum)
estimates store MHIncomeDY
estimates table MAllCountriesDY MHIncomeDY, star(0.1 0.05 0.01) stats(N r2)
This gave the results:

----------------------------------------------
Variable | MAllCountri~Y MHIncomeDY
-------------+--------------------------------
lnGDPpc | -.00016724 -.0005759
Gini | -.00003128 -.00001951
Unemp | .00005978* .00006034
Inflation | -5.716e-06 2.233e-06
FLaborforce | -.00002548 4.515e-07
GovExpE | -.00011046 .00002899
HealthExp | .00002526 9.274e-06
|
Year |
2001 | -.00001808 -.00002353
2002 | -.0001139 -.00021238
2003 | -.00015365 -.0003143
2004 | -.00010997 -.00041622
2005 | -.00015597 -.00049005
2006 | -.00011203 -.00043508
2007 | -.00016932 -.00040614
2008 | -.00022669 -.00047712
2009 | -.00042016 -.00077883
2010 | -.0005279 -.00088548
2011 | -.00052688 -.00080701
2012 | -.0005281 -.00092485
2013 | -.0005665 -.00098783
2014 | -.00053793 -.00099949
2015 | -.00049237 -.00108376
2016 | -.00041956 -.00093399
2017 | -.00041368 -.00094756
2018 | .00001754 .00122708
|
_cons | .02489984*** .02621443***
-------------+--------------------------------
N | 1604 606
r2 | .12285483 .28114665
----------------------------------------------
legend: * p<.1; ** p<.05; *** p<.01

Now the unemployment variable is significant at the 90% level in the first instance, and insignificant in the second instance. Can anyone tell me why this is happening?

For reference, this is how I generated my lnGDPpc variable:
Code:
gen lnGDPpc=ln(GDPpc)