Dear All,
I have a panel data of 2016 country-pairs for the period between 1990-2010 (5-year intervals). Dependent variable (RTA) is a binary and indicates whether dyads have a regional trade agreements or not and I have following explanatories as independent variables (GDP sum, GDP difference, Human Development Index diference, distance, remoteness and common colonizer dummy).
First, I run random effect probit model.
Then, in order to adress the issue of unobservable heterogeneity, I estimate Chamberlein probit random effects in which the averages of time variant variables are adding as controls.
Comparing the results show that except dhdi (Human development index difference) sign of the variables are the same.However, this variable is one of the main explanatory that I want to include and I expect negative sign like in the first regression. On the other hand why Chamberlein probit gave positive sign? Which one is reliable? Is there anything else that I can do?
Thank you in advance.
Elif
I have a panel data of 2016 country-pairs for the period between 1990-2010 (5-year intervals). Dependent variable (RTA) is a binary and indicates whether dyads have a regional trade agreements or not and I have following explanatories as independent variables (GDP sum, GDP difference, Human Development Index diference, distance, remoteness and common colonizer dummy).
First, I run random effect probit model.
Code:
iis id
. tis t
. xtset id t
panel variable: id (strongly balanced)
time variable: t, 1990 to 2010, but with gaps
delta: 1 unit
. xtprobit rta sgdp dgdp dhdi lndist remote i.commoncolonizer, re
Random-effects probit regression Number of obs = 9050
Group variable: id Number of groups = 2016
Random effects u_i ~ Gaussian Obs per group: min = 2
avg = 4.5
max = 5
Wald chi2(6) = 363.47
Log likelihood = -1115.682 Prob > chi2 = 0.0000
------------------------------------------------------------------------------
rta | Coef. Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
sgdp | .8437854 .0556821 15.15 0.000 .7346505 .9529204
dgdp | -.9613412 .1090365 -8.82 0.000 -1.175049 -.7476336
dhdi | -1.972117 1.092439 -1.81 0.071 -4.113259 .1690245
lndist | -3.615695 .2682057 -13.48 0.000 -4.141368 -3.090021
remote | .408067 .0518054 7.88 0.000 .3065303 .5096037
1.commonco~r | 1.584577 .7582801 2.09 0.037 .0983755 3.070779
_cons | -15.47044 3.007589 -5.14 0.000 -21.3652 -9.575671
-------------+----------------------------------------------------------------
/lnsig2u | 2.686554 .1181208 2.455041 2.918067
-------------+----------------------------------------------------------------
sigma_u | 3.831579 .2262947 3.412758 4.301799
rho | .9362285 .0070524 .9209293 .9487323
------------------------------------------------------------------------------
Code:
egen sgdpbar=mean( sgdp), by(id)
. egen dgdpbar=mean( dgdp), by(id)
. egen dhdibar=mean( dhdi), by(id)
. xtprobit rta sgdp dgdp dhdi sgdpbar dgdpbar dhdibar lndist remote i.commoncolonizer, re
Random-effects probit regression Number of obs = 9050
Group variable: id Number of groups = 2016
Random effects u_i ~ Gaussian Obs per group: min = 2
avg = 4.5
max = 5
Wald chi2(9) = 326.03
Log likelihood = -1070.8547 Prob > chi2 = 0.0000
------------------------------------------------------------------------------
rta | Coef. Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
sgdp | 1.083604 .0770198 14.07 0.000 .9326484 1.23456
dgdp | -1.308282 .2440045 -5.36 0.000 -1.786522 -.8300415
dhdi | 14.29419 2.598298 5.50 0.000 9.20162 19.38676
sgdpbar | -.4882945 .0742988 -6.57 0.000 -.6339175 -.3426715
dgdpbar | .7428298 .2560823 2.90 0.004 .2409177 1.244742
dhdibar | -21.8249 2.956264 -7.38 0.000 -27.61907 -16.03073
lndist | -3.335184 .2451413 -13.61 0.000 -3.815652 -2.854716
remote | .2710329 .0455683 5.95 0.000 .1817207 .3603452
1.commonco~r | .7676127 .7005804 1.10 0.273 -.6054998 2.140725
_cons | -4.201492 2.8389 -1.48 0.139 -9.765633 1.362649
-------------+----------------------------------------------------------------
/lnsig2u | 2.36992 .1290648 2.116957 2.622882
-------------+----------------------------------------------------------------
sigma_u | 3.270555 .2110567 2.881983 3.711518
rho | .9145046 .0100911 .8925404 .9323198
------------------------------------------------------------------------------
Likelihood-ratio test of rho=0: chibar2(01) = 1194.90 Prob >= chibar2 = 0.000
Thank you in advance.
Elif