Panel threshold regression

[Ermiyas Gebrie]

Dear all,
I am trying to do panel threshold regression developed by Hansen(1999). Our dependent variable is economic growth while our independent variables are labour force, gross capital formation, inflation rate, total land area and trade openness. And our threshold variable and regime dependent variable is external debt to GDP ratio and external debt. When I run the threshold regression I got the following result. Based on the threshold effect test the model does not have a threshold value because the RSS is less than any of the critical values. According to this result, the model is linear, not non-linear. Having said this I have the following questions
1. is our model specification correct especially the threshold variable and the regime dependent variable?
2. According to Hansen(1999) threshold effect test if the model does not have threshold B1 = B2 but, in our case, the threshold effect test showed me no threshold effect but the beta value is different. In this case how to interpret the two betas and how to interpret the effect of external debt on economic growth in our model.

xthreg g l a k opp i , rx( DebtGDP ) qx( D ) thnum(1) grid(400)trim(0.01) bs(300)
Estimating  the  threshold  parameters:   1st ......  Done
Boostrap for single threshold
.................................................. +   50
.................................................. +  100
.................................................. +  150
.................................................. +  200
.................................................. +  250
.................................................. +  300

Threshold estimator (level = 95):

model     Threshold         Lower         Upper

Th-1       22.3286       22.2560       22.3389


Threshold effect test (bootstrap = 300):

Threshold        RSS        MSE      Fstat    Prob   Crit10    Crit5    Crit1

Single    32.2469     0.0448      74.74  0.0767  67.1511  83.4262  107.4726


Fixed-effects (within) regression               Number of obs      =       738
Group variable: country                         Number of groups   =        41

R-sq:  within  = 0.8471                         Obs per group: min =        18
between = 0.2962                                        avg =      18.0
overall = 0.2378                                        max =        18

F(7,690)           =    546.21
corr(u_i, Xb)  = -0.9914                        Prob > F           =    0.0000


g       Coef.   Std. Err.      t    P>t     [95% Conf. Interval]

l    1.931645   .0806895    23.94   0.000     1.773219    2.090071
a   -5.830976   2.001644    -2.91   0.004     -9.76102   -1.900932
k    .2711044   .0178865    15.16   0.000     .2359859    .3062229
opp   -.1428224   .0297305    -4.80   0.000    -.2011954   -.0844493
i    1.91e-07   .0003268     0.00   1.000    -.0006415    .0006419

_cat#c.DebtGDP
0    -.0004341   .0001691    -2.57   0.010    -.0007662   -.0001021
1     -.003356   .0003512    -9.56   0.000    -.0040455   -.0026665

_cons    59.16148   24.48589     2.42   0.016     11.08568    107.2373

sigma_u   9.8258185
sigma_e   .21673881
rho   .99951368   (fraction of variance due to u_i)

F test that all u_i=0: F(40, 690) = 60.55                    Prob > F = 0.0000 

If you could provide me with some better suggestions on how to approach such a problem

Thank you
Ermiyas;

[Phil Bromiley]

You didn't get a quick answer. You'll increase your chances of a useful answer by following the FAQ on asking questions - provide Stata code in code delimiters, readable Stata output and sample data using dataex.

With a user-written procedure, a response often depends on someone active on the list using the procedure. You may need to contact the author.