Dear Statalist

I'm having difficulties trying estimate the following model working with panel data and a binary dependent variable

Code:
 xtlogit Gov i.v2x_regime_lag cgdppc_lag max_rdiscl_lag NHIxl_lag cinc_lag Total_Oil_Income_PC_lag peace_year_lag decay_function_lag Americas Europe MENA Asia Africa if e2==1, vce(cluster country_name_numeric)
Gov = binary
Regions (Americas, Europe, MENA, Asia, Africa) = binary
v2x_regime_lag = categorical
Other variables = continuous

+6000 observations

The result from the xtlogit is as follow:

Code:
 Fitting comparison model:
Iteration 0:   log pseudolikelihood = -665.12183  
Iteration 1:   log pseudolikelihood = -638.32856  
Iteration 2:   log pseudolikelihood = -630.81981  
Iteration 3:   log pseudolikelihood = -630.49594  
Iteration 4:   log pseudolikelihood = -630.48068  
Iteration 5:   log pseudolikelihood = -630.47707  
Iteration 6:   log pseudolikelihood = -630.47634  
Iteration 7:   log pseudolikelihood = -630.47622  
Iteration 8:   log pseudolikelihood = -630.47619  
Iteration 9:   log pseudolikelihood = -630.47618 

Fitting full model:

tau =  0.0     log pseudolikelihood = -630.47618
tau =  0.1     log pseudolikelihood =  -630.8826

Iteration 0:   log pseudolikelihood =  -630.8826  
Iteration 1:   log pseudolikelihood = -630.49654  
Iteration 2:   log pseudolikelihood = -630.45437  
Iteration 3:   log pseudolikelihood = -630.44726  
Iteration 4:   log pseudolikelihood = -630.44694  
Iteration 5:   log pseudolikelihood = -630.44694  (not concave)
Iteration 6:   log pseudolikelihood = -630.44694  (not concave)
Iteration 7:   log pseudolikelihood = -630.44694  (not concave)
Iteration 8:   log pseudolikelihood = -630.44694  (not concave)
Iteration 9:   log pseudolikelihood = -630.44694  (not concave)
.
.
Iteration 257:   log pseudolikelihood = -630.44694  (not concave)
We have tried to insert the difficult option, however, that does not work. Furthermore we have inspected for any possible measurement errors in the regional dummies because without the model works. If we drop one regional dummy from the regression the model converges. Any suggestions?

//Marco Liedecke