Adding time dummies or adding country dummies do not affect the results at all! panel data regression

[svetlana zvyagintseva]

I am trying to find out if conflict (dependent var) is related to i)food import prices and ii)food import price volatility (independent vars)

I have two main questions:

1. I have panel data for 39 countries and 20 years for each of them. I control for i) time fixed effects., ii)country specific fixed effects, iii) linear time trend step by step AND the results are exactly the same, can not explain this
2. When i add country specific time effects, the coefficient of the variable of main interest (food price) goes insignificant. Is this a reason to rule out the model? how could I explain this is not?



adding time dummies + country fixed effects (by fe):

xtset id year

xtreg  number final_foodprice cv2_food i.year, fe robust

------------------------------------------------------------------------------

Adding trend on top of country and time fixed effects, no change in results

xtreg number  final_foodprice cv2_food trend i.year, fe

----------------------- the coefficients of the two regressions are the same, only the significance levels of time dummies change, how to explain this?---------------------------


2. I generate a variable which is 1 for each country-year pair.

gen csp1=_n-0 if id=1
gen csp1=_n if id=1
gen csp1=_n if id==1
replace csp1=0 if id>1
gen csp2=_n-2.22 if id==2
replace csp2=0 if id<2 & id>2
drop csp2
gen csp2=_n-44 if id==2
replace csp2=0 if id<2 & id>2
drop csp2
gen csp2=_n-22 if id==2
replace csp2=0 if id>2
replace csp2=0 if id<2
gen csp3=_n-44 if id==3
replace csp3=_n-66 if id==4
drop csp1 csp2 csp3
gen cs1=_n if id==1
replace cs1=0 if id>1
gen cs2=_n-22 if id==2
replace cs2=0 if id>2 & id<2
replace cs2=0 if id<2 & id>2
replace cs2=0 if id>2
replace cs2=0 if id<2
gen cs3=_n-88 if id==3
drop cs3
gen cs3=_n-66 if id==3
drop cs3
gen cs3=_n-33 if id==3
drop cs3
gen cs3=_n if id==3
drop cs3
gen cs3=_n-44 if id==3
replace cs3=0 if id>3
replace cs3=0 if id<3
replace cs4=_n if id=4
gen cs4=_n if id==4
drop cs4
gen cs4=_n-66 if id==4
replace cs4=0 if id>4
replace cs4=0 if id<4
gen cs5=_n if id=5
gen cs5=_n if id==5
drop cs5
gen cs5=_n-88 if id==5
replace cs5=0 if id>5
replace cs5=0 if id<5
gen cs6=_n-110 if id==6
replace cs6=0 if id<6
replace cs6=0 if id>2
drop cs6
gen cs6=_n-88 if id==6
replace cs6=0 if id>6
replace cs6=0 if id<6
drop cs6
gen cs6=_n-110 if id==6
replace cs6=0 if id>6
replace cs6=0 if id<6
gen cs7=_n-132 if id==7
replace cs7=0 if id<7
replace cs7=0 if id<7
replace cs7=0 if id>7
gen cs8=_n-154 if id==8
replace cs8=0 if id<8
replace cs8=0 if id>8
gen cs9=_n-176 if id==9
replace cs9=0 if id<9
replace cs9=0 if id>9
gen cs10=_n-198 if id==10
replace cs10=0 if id>10
replace cs10=0 if id<10
gen cs11=_n-220 if id==10
drop cs11
gen cs11=_n if cs==11
gen cs11=_n if id==11
drop cs11
gen cs11=_n-220 if id==11
replace cs11=0 if id<11
replace cs11=0 if id>11
gen cs12=_n-242 if id==12
replace cs12=0 if id>12
replace cs12=0 if id<12
gen cs13=_n-264 if id==13
replace cs13=0 if id>13
replace cs13=0 if id<13
gen cs14=_n-286 if id==14
replace cs14=0 if id>14
replace cs14=0 if id<14
gen cs15=_n-308 if id==15
replace cs15=0 if id>15
replace cs15=0 if id<15
gen cs16=_n-330 if id==16
replace cs16=0 if id>16
replace cs16=0 if id<16
gen cs17=_n-352 if id==17
replace cs17=0 if id>17
replace cs17=0 if id<17
gen cs18=_n-374 if id==18
replaec cs18==0 if id<18
replace cs18=0 if id>18
replace cs18=0 if id<18
gen cs19=_n-396 if id==19
replace cs19=0 if id>19
replace cs19=0 if id<19
gen cs20=_n-428 if id==20
drop cs20
gen cs20=_n-418 if id==20
replace cs20=0 if id>20
replace cs20=0 if id<20
gen cs21=_n-440 if id==21
replace cs21=0 if id<21
replace cs21=0 if id>21
gen cs22=_n-462 if id==22
replace cs22=0 if id<22
replaec cs22=0 if id>22
replace cs22=0 if id>22
gen cs23=_n-484 if id==23
replace cs23=0 if id>23
replaec cs23=0 if cs<23
replace cs23=0 if cs<23
replace cs23=0 if id<23
gen cs24=_n-506 if id==24
replace cs24=0 if id>24
replace cs24=0 if id<24
gen cs25=_n-528 if id==25
replace cs25=0 if id>25
replace cs25=0 if id<25
gen cs26=_n-550 if id==26
replace cs26=0 if id>26
replace cs26=0 if id<26
gen cs27=_n-572 if id==27
replace cs27=0 if id<27
replace cs27=0 if id>27
gen cs28=_n-594 if id==28
replace cs28=0 if id>28
replace cs28=0 if id<28
gen cs29=_n-616 if id==29
replace cs29=0 if id>29
replace cs29=0 if id<29
gen cs30=_n-638 if id==30
replace cs30=0 if id<30
replace cs30=0 if id>30
gen cs31=_n-660 if id==31
replace cs31=0 if id>31
replace cs31=0 if id<31
gen cs32=_n-682 if id==32
replace cs32=0 if id<32
replace cs32=0 if id>32
gen cs33=_n-704 if id==33
replace cs33=0 if id>33
replaec cs33=0 if id<33
replace cs33=0 if id<33
gen cs34=_n-726 if id==34
replace cs34=0 if id>34
replace cs34=0 if id<34
gen cs35=_n-748 if id==35
replace cs35=0 if id>35
replaec cs35=0 if id<35
replace cs35=0 if id<35
gen cs36=_n-770 if id==36
replace cs36=0 if id>36
replace cs36=0 if id<36
gen cs37=_n-792 if id==37
replace cs37=0 if id<37
replace cs37=0 if id>37
gen cs38=_n-814 if id==37
replace cs38=0 if id>37
replace cs38=0 if id<37
drop cs38
gen cs38=_n if id==38
drop cs38
gen cs38=_n-814 if id==38
replace cs38=0 if id>38
replace cs38=0 if id<38
gen cs39=_n-836 if id==39
replace cs39=0 if id>39
replace cs39=0 if id<39
gen cs40=_n-858 if id==40
replace cs40=0 if id>40
replaec cs40=0 if id<40
replace cs40=0 if id<40

xtreg number final_foodprice  cv2_food i.year i.id  cs1-cs40, fe

MANY THANKS FOR YOUR ADVICE

[Phil Bromiley]

With year dummies, the time trend is irrelevant. The year dummies completely take care of any time trend effect. You will also see that you cannot reject the hypothesis that all the year effects have zero parameters. Some would say this means you can drop the year dummies (although some might disagree). The same goes for your cs variables.

You can't have country-year effects if your observation is the country-year. You'd have as many dummies as you have observations. I think your code doesn't do what I interpreted your text to say.

If you're going to do all these, take a look at factor variables notation. It will do all these interactions for you automatically and make them all work with margins.

By the way, thank you for including the code. It might be easier if next time you cut it down so we didn't have quite as much to read. I am not going to try to interpret your very long coding example. I strongly suspect you have logical error in your code. When you set something equal to _n that _n will be the observation number in the entire data set. [With by variables, sometimes it will be the number within the by panel, but not without the by.] I don't see why you'd want to do that.