Error in Bilateral Data Set  updated May 21, 2010

Eileen Brooks has found a small error in a bilateral trade data set of mine; I thank her for pointing this out to me. In particular, the World Bank country income classifications have been slightly miscoded in a bilateral trade data set.

The mistake was that Gabon’s IFS country code (646) was incorrectly replaced by Guinea-Bissau’s country code (654) in the program generating the data.  This means that Gabon is missing an income classification (it is in fact a middle-income country), while some of Guinea-Bissau’s observations are classified as middle-income (it is in fact a low-income country).

The consequences of this error are small in practice. This error affects two rows in table 4 of my paper “Do We Really Know that the WTO Increases Trade?”  With the error, the relevant rows read:

 

Both in GATT/WTO

One in GATT/WTO

GSP

 

 

 

Middle Income

-.05

(.06)

-.03

(.05)

.93

(.04)

Low Income

-.38

(.08)

-.37

(.08)

1.11

(.05)

With corrected data, the rows should read:

 

Both in GATT/WTO

One in GATT/WTO

GSP

 

 

 

Middle Income

-.05

(.06)

-.04

(.05)

.92

(.04)

Low Income

-.38

(.08)

-.36

(.08)

1.11

(.05)

 

The error also affects a single row of different tables in my paper “Which International Institutions Promote International Trade?”  With the error, the relevant rows of tables 2 and A5 respectively are:

 

                                      -----------------------------------------------------Coefficients----------------------------------------------

 

Both GATT

/WTO

One GATT

/WTO

Both

IMF

One

 IMF

Both OECD

One OECD

Regional

FTA

 

 

 

 

 

 

 

No Low-Income Countries

.15

(.07)

.10

(.06)

-.43

(.11)

-.26

(.11)

.37

(.08)

.36

(.05)

1.15

(.12)

 

 

Both GATT

/WTO

One GATT

/WTO

Both

IMF

One

 IMF

Both OECD

One OECD

Regional

FTA

 

 

 

 

 

 

 

No Low-Income Countries

.17

(.02)

.02

(.02)

-.42

(.04)

-.18

(.04)

.84

(.04)

.34

(.02)

.67

(.04)

With corrected data, the rows should read:

                                      -----------------------------------------------------Coefficients----------------------------------------------

 

Both GATT

/WTO

One GATT

/WTO

Both

IMF

One

 IMF

Both OECD

One OECD

Regional

FTA

 

 

 

 

 

 

 

No Low-Income Countries

.14

(.07)

.09

(.06)

-.41

(.11)

-.24

(.11)

.36

(.08)

.35

(.05)

1.16

(.12)

 

 

Both GATT

/WTO

One GATT

/WTO

Both

IMF

One

 IMF

Both OECD

One OECD

Regional

FTA

 

 

 

 

 

 

 

No Low-Income Countries

.16

(.02)

.01

(.02)

-.40

(.04)

-.17

(.04)

.84

(.04)

.34

(.02)

.66

(.04)

The data sets on the web have not been corrected, but the results in the papers have been.

****

On Dec 15, 2003 Chris Pearce brought to my attention a potential issue associated with rebasing of GDP series between 1997 and 1998.  This seems to make no difference to the final results (almost surely because of the year dummies), and has thus not been pursued further.  Still, users are warned!  I thank Chris for bringing this to my attention.

*****

On May 2, 2005 Johnannes Voget alerted me to the fact that the Euro-zone countries are coded as custrict=0 for 1999 even though EMU technically began on Jan 1, 1999.  I like to date EMU from the introduction of the physical currency in 2002, since that’s when prices switched over to the Euro.  Still, be warned!  Johannes also alerted me to that fact that Luxembourg enters the data set in 1997 but is incorrectly coded as not being a member of the EU.  Apologies for this mistake; please be aware.  I thank Johannes for his messages.

*****

On July 27, 2005 Ariell Reshef alerted me to the fact that Guyana and Suriname are coded as carib=1 (they are not in the Caribbean), while Jamaica and St. Lucia are coded carib=0 (they are).  Thanks to him, and apologies.

*****

On October 21, 2005 Russell Hillberry alerted me to the fact that EFTA is not included in the RTAs (as noted in the paper, so this is a clarification rather than an error).  Thanks to him.

*****

On June 30, 2008 Trang Tran alerted me to the facts that a) Rwanda was coded as a non-landlocked country is landlocked; and b) Syria was coded as landlocked but it has a 193km coastline (according to the CIA world factbook).  Thanks!

 *****

On May 21, 2010 Yukihiro Kumeno alerted me to the fact that "join2" is often mis-coded as 1948 when cty1==112 (the UK).  As always, my thanks and apologies.

 *****

To my knowledge, no other observations or variables are in error.

I apologize for these mistakes, and again ask future users of my data sets to bring any problems to my attention.