Antoine A. Djogbenou, "Comovements in the Real Activity of Developed and 
Emerging Economies: A Test of Global versus Specific International 
Factors", Journal of Applied Econometrics, Vol. 35, No. 3, 2020, pp. 
344-370.

All files are ASCII files in DOS format and are zipped in ad-files.zip.
Unix/Linux users should use "unzip -a".


A. Description of the datasets

I used four datasets, each in two files.

1. The first dataset is quarterly and was downloaded from the Global
Economic Monitor DataBank. It covers the period from the third quarter
of 1996 to the last quarter of 2018. The data were constructed as
described in the empirical application. The file dev1.csv contains 22
series of real gross domestic product and 15 series of industrial
production for developed countries. The file eme1.csv includes 30
series of gross domestic product and 21 series of industrial
production for emerging countries.

2. The second dataset is yearly. It includes real GDP, real private
consumption, and real fixed asset investment as in Kose, Otrok, and
Prasad (2012). I reconstructed this dataset based on the data used by
Hirata, Kose, and Otrok (2013) by focusing on the developed economies
and the emerging economies as presented in the paper. In dev2.csv and
eme2.csv, there are 69 and 72 real economic activity variables,
respectively. To apply our test to the same data as in Kose, Otrok, we
simply restrict the time periods to 1961−2008. However, the results
using the full sample from Hideaki Hirata and Ayhan M. Kose can be
obtained using the overall time periods from 1961 to 2015 by removing
the time periods restriction in the code.

3. The third dataset is the same as in Walti (2012). It contains real
GDP data measured at an annual frequency for the period from 1980 to
2008. As in Walti (2012), dev3.csv contains 26 real GDP series for
developed economies, and eme3.csv includes 31 real GDP series for
emerging economies. To address the data irregularities, we use an
expectation-maximization algorithm. The main program below includes
comments on how to remove the time periods or the variables with
missing values before performing the test.

4. The last dataset is from Mumtaz, Simonelli, and Surico (2011). It
contains the change in percentage of yearly real GDP from 1951 to
2006. The data are publicly available and can be downloaded from
https://econpapers.repec.org/software/redccodes/09-235.htm. To focus
on developed and emerging economies, I use the same classification of
countries as above. The files dev4.csv and eme4.csv contain 21 series
on real GDP growth for developed economies and 15 series on real GDP
growth for emerging economies, respectively.


B. Matlab Codes

The main program for replication is the Matlab file main.m. This code
produces

a) the computed test statistic (test_stat1), the critical value
(critical_value1), the P-value (p_value1) and the decision from the
test (reject1, 1 if we reject the null and 0 if not) if the data in 1
are used.

b) the computed test statistic (test_stat2), the critical value
(critical_value2), the P-value (p_value2) and the decision from the
test (reject2, 1 if we reject the null and 0 if not) if the data in 2
are used.

c) the computed test statistic (test_stat3), the critical value
(critical_value3), the P-value (p_value3) and the decision from the
test (reject1, 1 if we reject the null and 0 if not) if the data in 3
are used.

d) the computed test statistic (test_stat4), the critical value
(critical_value4), the P-value (p_value4) and the decision from the
test (reject4, 1 if we reject the null and 0 if not) if the data in 4
are used.


The Matlab file main.m uses the following functions:

1. lm_test.m : It provides the computed test statistic, the critical
   value, the P-value and the decision.

2. standard.m : It standardizes the variables.

3. uptriangle.m : It is used to obtain the Vech of matrices.

4. hpfilter.m : It uses the Hodrick-Prescott filter to extract the
   trend of a time series.

5. EM.m : This program uses an expectation-maximization algorithm to
   address data irregularities by iteratively fitting the missing values.

6. nbplog.m : It gives the number of factors using Bai and Ng (2002).
   This function and its sub-functions are publicly available. They can
   be downloaded from
   https://drive.google.com/file/d/0B-aG4lrQrBYsTGNvYTNjWld2X1U/view


Antoine A. Djogbenou
daa [AT] yorku.ca