Sinem Hacioglu Hoke and George Kapetanios, "Common Correlated Effect 
Cross-Sectional Dependence Corrections for Non-Linear Conditional Mean 
Panel Models", Journal of Applied Econometrics, Vol. 36, No. 1, 2021, 
pp. 125-150.

DATA

* First Empirical Application

The dataset is downloaded from
https://www.aeaweb.org/articles?id=10.1257/aer.p20151092

To get a monthly balanced sample, we selected the banks that have
reasonably long periods of data, which leads to 96 banks over
1996:3-2011:2. The final dataset is for Tier 1 and asset beta values
and both are in the file Tier1AssetBeta.csv .

* Second Empirical Application

Annual data for 27 OECD countries for 1951-2006. Countries are:
Australia, Austria, Belgium, Canada, Chile, Denmark, Finland, France,
Greece, Iceland, Ireland, Israel, Italy, Japan, Luxembourg, Mexico,
Netherlands, New Zealand, Norway, Portugal, Spain, Republic of Korea,
Sweden, Switzerland, Turkey, United Kingdom, United States. 

Data downloaded from World Penn tables, version 7.1. All the
variables, investment ratio, savings ratio and openness are in
the file SavingInv.csv


CODE

The replication files can be used to replicate our simulation results
and can be adapted to use the data provided for the empirical
exercises. Parallel computing in a server for the simulations is
recommended. An additional ReadMe file is provided with the simulation
files.

The main files to run are

1) nIcce_allsim to generate similar results to the results in Tables
1-4 in the paper.

2) nIcce_simplesim to generate the similar results to the additional
exercise in the Appendix, Tables A2-A7.

First add path to the functions from the respective directory.

For 1, nIcce_allsim:

ExpA = 1 or 2;  running experiment A or B, if 1 we run full rank
condition, if 2, we run rank deficient version

exp = 1 or 2: running experiment 1 or 2, exp = 1 heterogeneous beta,
exp = 2 homogeneous beta

For 2, nIcce_simplesim:

choose the number of factors by changing nf.
choose the number of explanatory variables to include, x, by changing nx.
nx and nf together make the conditions for full rank or rank deficient
cases.

Exp=1 or 2 % running experiment 1 or 2, exp = 1 heterogeneous beta,
exp = 2 homogeneous beta

Despite all the effort that has been made to ensure that the codes are
error-free, mistakes are still possible. Please get in touch if you
spot a problem in the code:

sinemhacioglu [AT] gmail.com