Table B1. Estimation Results for Knitted Apparel: Heterogeneous Effects across Export Countries
(I) (II)
GSP x Year2015 0.825*** 0.646***
(Cambodia as a base country) [0.043] [0.022]
* 1 for Laos -1.111*** -0.562***
[0.152] [0.054]
* 1 for Myanmar 0.207*** 0.153***
[0.066] [0.034]
Method OLS Frac
Country FE YES
HS FE YES
Regime FE YES
Year FE YES
Country-HS-Regime FE YES
Country-HS-Year FE YES
Number of obs 1,892 2,054
R-squared 0.8016
Log pseudolikelihood -738
Notes: ***, **, and * represent significance at the 1%, 5%, and 10% statistical levels, respectively.
Parentheses contain the standard error clustered at the country-HS-regime level. The dependent variable is the share of imports of product p from country i under preference regime r in year t out of total imports of product p from country i in year t. Sample countries include Cambodia, Laos, and Myanmar, products do HS nine-digit codes in the HS 61, and years do from 2013 to 2016. The preference regimes include AJCEP and GSP-LDC.
30
Table B2. Correlation between the Trade Values of Knitted Apparel and Woven Apparel
ln Knitted Trade Export Export Import ln Woven Trade 1.015*** 1.273*** 0.994***
[0.10] [0.338] [0.054]
Constant -0.498*** -6.094 0.022
[0.159] [6.920] [0.137]
Sample All LDC3 All
Number of obs 840 12 841
R-squared 0.9279 0.5863 0.9529
Notes: We estimate by OLS. *** represents significance at the 1% statistical level. Parentheses contain the heteroscedasticity-consistent standard error. The dependent variable in column “Export” (“Import”) is a log of total exports (total imports) of knitted apparel in country i in year t. Similarly, the independent variable in column “Export” (“Import”) is a log of total exports (total imports) of woven apparel in country i in year t. Sample countries include 212 countries and sample years do from 2012 to 2015. When
“Sample” indicates “LDC3,” we restrict exporting countries to only Cambodia, Laos, and Myanmar.
Table B3. Impacts on the Share and Value of Preferential Imports
Share Value Value
HS61 x Year2015 x GSP 0.568*** 2.026*** 2.315***
[0.026] [0.167] [0.204]
HS61 x Year2015 x AJCEP -0.199*** -0.329*** -0.390*
[0.023] [0.107] [0.233]
ln Imports 0.006 0.209*** 0.866***
[0.021] [0.045] [0.227]
Plus one in a dep. var. YES NO
Number of obs 3,468 10,434 1,968
R-squared 0.8436 0.8121 0.8370
Notes: *** represents significance at the 1% statistical level. Parentheses contain the standard error clustered at the country-HS level. The dependent variable is the share of preferential imports under each regime in column “Share.” In column “Value,” it is a log of the preferential imports. In “YES” (“NO”) for
“Plus one in a dep. var.,” we add (do not add) the value one to the preference imports before taking a log.
We estimate by OLS and include country-product-regime and country-regime-year fixed effects in both columns. Sample countries include Cambodia, Laos, and Myanmar, products do HS nine-digit codes in HS 61 and HS62, and years do from 2013 to 2016. The preference regimes include AJCEP and GSP-LDC.
31
Table B4. Separate Estimation Results for GSP-LDC and AJCEP
(I) (II) (III) (IV)
HS61 x Year2015 0.559*** 0.291*** -0.192*** -0.101***
[0.026] [0.022] [0.023] [0.028]
ln Imports 0.063** 0.059* -0.052* -0.036
[0.029] [0.034] [0.031] [0.039]
Method OLS Frac OLS Frac
Country FE YES YES
HS FE YES YES
Year FE YES YES
Country-HS FE YES YES
Country-Year FE YES YES
Number of obs 1,692 1,815 1,692 1,815
R-squared 0.8293 0.6317
Log pseudolikelihood -412 -382
GSP-LDC AJCEP
Notes: ***, **, and * represent significance at the 1%, 5%, and 10% statistical levels, respectively.
Parentheses contain the standard error clustered at the country-HS level. The dependent variable is the share of imports of product p from country i under preference regime r in year t out of total imports of product p from country i in year t. Sample countries include Cambodia, Laos, and Myanmar, products do HS nine-digit codes in HS 61 and HS62, and years do from 2013 to 2016. The preference regimes include only AJCEP in column “AJCEP” and only GSP-LDC in column “GSP-LDC.” In this table, we exclude woven apparel products of which the RoOs for GSP-LDC are the two-stage processing rule.
Table B5. Estimation Results: Impacts on Trade Values by Tariff Regimes
GSP-LDC AJCEP MFN
HS61 x Year2015 2.318*** -0.388* -0.786***
[0.204] [0.232] [0.176]
ln Imports 0.796*** 1.112** 0.247
[0.248] [0.522] [0.315]
Number of obs 1,321 647 940
R-squared 0.8601 0.7145 0.7818
Notes: ***, **, and * represent significance at the 1%, 5%, and 10% statistical levels, respectively.
Parentheses contain the standard error clustered at the country-HS level. The dependent variable is a log of imports of product p from country i in year t. We estimate by OLS. Sample countries include Cambodia, Laos, and Myanmar, products do HS nine-digit codes in HS 61 and HS62, and years do from 2013 to 2016.
In all specifications, we include country-HS and country-year fixed effects.
32
Table B6. Estimation Results: Impacts on Trade Values by Tariff Regimes
(I) (II) (III) (IV) (V) (VI)
HS61 x Year2015 1.960*** 1.767*** -0.292** -0.139 -0.611*** -0.885***
[0.177] [0.481] [0.114] [0.187] [0.121] [0.205]
ln Imports 0.331*** 1.384*** 0.107*** 1.558*** 0.035 0.825**
[0.088] [0.237] [0.040] [0.448] [0.048] [0.364]
Method OLS PPML OLS PPML OLS PPML
Country FE YES YES YES
HS FE YES YES YES
Year FE YES YES YES
Country-HS FE YES YES YES
Country-Year FE YES YES YES
Number of obs 4,908 3,309 4,908 2,052 4,908 2,724
R-squared 0.817 0.6686 0.7791 0.6408 0.8157 0.7169
Log pseudolikelihood -1.5.E+08 -2.3.E+07 -9.3.E+06
GSP-LDC AJCEP MFN
Notes: ***, **, and * represent significance at the 1%, 5%, and 10% statistical levels, respectively.
Parentheses contain the standard error clustered at the country-HS level. The dependent variable is a log of one-plus imports of product p from country i in year t in columns (I), (III), and (V), i.e., the case of OLS estimation. It is simply imports of product p from country i in year t in columns (II), (IV), and (VI), i.e., PPML estimation. Sample countries include Cambodia, Laos, and Myanmar, products do HS nine-digit codes in HS 61 and HS62, and years do from 2013 to 2016. In this table, we exclude woven apparel products of which the RoOs for GSP-LDC are the two-stage processing rule.
33
Table B7. Estimation Results: Impacts on Trade Values
GSP-LDC AJCEP MFN GSP-LDC AJCEP MFN
HS61 x Year2015 2.335*** -0.408* -0.778*** 2.318*** -0.388* -0.786***
[0.204] [0.239] [0.182] [0.204] [0.232] [0.176]
ln Imports 0.787*** 1.122** 0.207 0.796*** 1.112** 0.247 [0.249] [0.524] [0.315] [0.248] [0.522] [0.315]
Number of obs 1,298 639 913 1,321 647 940
R-squared 0.8576 0.7096 0.7799 0.8601 0.7145 0.7818
Model (I) Model (II)
Notes: ***, **, and * represent significance at the 1%, 5%, and 10% statistical levels, respectively.
Parentheses contain the standard error clustered at the country-HS level. The dependent variable is a log of imports of product p from country i in year t. We estimate by OLS. In all specifications, we include country-HS and country-year fixed effects. Sample countries include Cambodia, Laos, and Myanmar, products do HS nine-digit codes in HS 61 and HS62, and years do from 2013 to 2016. In Model (I), we exclude woven apparel products of which the RoOs for GSP-LDC are the two-stage processing rule. In Model (II), we do not include country-product-year pairs in which any imports, regardless of tariff regimes not being observed.
34
Table B8. Estimation Results: Impacts on Trade Values
(I) (II) (III) (IV) (V) (VI)
HS61 x Year2015 5.907*** 1.670*** -1.346*** -0.169 -2.375*** -0.914***
[0.281] [0.487] [0.318] [0.179] [0.308] [0.189]
ln Imports 0.962*** 1.355*** 0.236 1.346*** 0.517 0.810**
[0.370] [0.247] [0.447] [0.438] [0.444] [0.382]
Method OLS PPML OLS PPML OLS PPML
Country FE YES YES YES
HS FE YES YES YES
Year FE YES YES YES
Country-HS FE YES YES YES
Country-Year FE YES YES YES
Number of obs 1,734 1,817 1,734 1,344 1,734 1,641
R-squared 0.8204 0.6546 0.749 0.6445 0.762 0.7145
Log pseudolikelihood -1.4.E+08 -2.0.E+07 -8.3.E+06
GSP-LDC AJCEP MFN
Notes: ***, **, and * represent significance at the 1%, 5%, and 10% statistical levels, respectively.
Parentheses contain the standard error clustered at the country-HS level. The dependent variable is a log of one-plus preferential imports of product p from country i in year t, in columns (I) and (III), i.e., the case of OLS estimation. It is simply preferential imports of product p from country i in year t in columns (II) and (IV), i.e., PPML estimation. Sample countries include Cambodia, Laos, and Myanmar, products do HS nine-digit codes in HS 61 and HS62, and years do from 2013 to 2016. Compared with Table 6, this table does not include country-product-year pairs in which any imports, regardless of tariff regimes not being observed.