Post-colonial trade between Russia and former Soviet republics: back to big brother?

Table 4
Summary statistics
Definitions of variables are in Table 
9
. Country i is the exporter, and j is the importer throughout the 
tables
Variable
Observations
Mean
SD
Min
Max
ln(Trade flow)
19,522
1.91
3.15
− 6.91
13.85
ln(Trade cost)
19,432
0.73
0.41
0.00
3.84
ln(GDP)
23,273
5.03
2.24
− 0.56
10.24
ln(Distance)
23,273
7.89
0.87
4.38
9.48
ln(1+Tariff)
23,273
0.03
0.05
0.00
0.79
Contiguous
23,273
0.08
0.27
0.00
1.00
RTA Between i and j
23,273
0.53
0.50
0.00
1.00
Colony
23,273
0.03
0.18
0.00
1.00
Common Language
23,273
0.04
0.19
0.00
1.00
i and j are Same Country
23,273
0.03
0.16
0.00
1.00
i and j Both Landlocked
23,273
0.43
0.49
0.00
1.00
Country Pair Group Dummies
RUS
j
_CIS
+
i
23,273
0.01
0.09
0.00
1.00
RUS
i
_CIS
+
j
23,273
0.01
0.09
0.00
1.00
SIB
j
_SIB
i
23,273
0.07
0.26
0.00
1.00
proRUS_RUS
23,273
0.005
0.076
0.00
1.00
23
The baseline estimations in Table 
12
are with Poisson pseudo-maximum likelihood (PPML) on the 
lines of Silva and Tenreyro (
2006
, 
2011
) and two-stage least squares. All other methods confirm the GLS 
results. The one exception is PPML, which provides some unexpected signs for tariffs and the colony 
variable but on average higher coefficient estimates. We think this is either due to inclusion of zero trade 
flows, or due to its nature of being a semilog type of method, while all the others are log-log.
24
Note that, since trade flows are in value terms, the coefficient on (1 + Tariff ) is an estimate of 1 + 𝜎.
Economic Change and Restructuring (2021) 54:877–918
899

---

1 3
Table 5
Regressions with trade flows
Robust standard errors in parentheses ∗∗∗ p < 0.001, ∗∗ p < 0.01, ∗ p < 0.05 . All regressions include 
year dummies. Columns 2–5 have exporter, importer and year dummies, while column 1 allows monadic 
group dummies. Variable definitions are in Table 9. i denotes exporter and j denotes importer
Variables
(1)
(2)
(3)
(4)
ln(Trade flow)
ln(Trade flow)
ln(Trade flow)
ln(Trade flow)
Year/country dummies
Yes
Yes
Yes
Yes
ln
(GDP
i
)
0.14***
0.14***
0.14***
0.14***
(0.037)
(0.037)
(0.038)
(0.038)
ln
(GDP
j
)
0.78***
0.78***
0.78***
0.78***
(0.046)
(0.046)
(0.046)
(0.046)
ln(Distance)
− 1.12***
− 1.03***
− 1.02***
− 1.02***
(0.081)
(0.075)
(0.076)
(0.076)
Contiguous
0.41**
0.41**
0.41**
0.42**
(0.143)
(0.145)
(0.146)
(0.146)
i and j Both Landlocked
− 0.76***
− 0.40*
− 0.35*
− 0.35*
(0.191)
(0.172)
(0.172)
(0.172)
ln(1+Tariff)
− 3.61***
− 3.58***
− 3.58***
− 3.58***
(0.517)
(0.515)
(0.515)
(0.515)
RTA Between i and j
0.41***
0.37***
0.37***
0.37***
(0.047)
(0.048)
(0.048)
(0.048)
Colony
0.66**
0.14
0.53*
0.51*
(0.216)
(0.208)
(0.263)
(0.255)
Common Language
0.52*
0.24
0.25
0.24
(0.211)
(0.181)
(0.180)
(0.180)
i and j are Same Country
4.11***
3.92***
3.92***
3.93***
(0.295)
(0.281)
(0.283)
(0.282)
Country pair group dummies
CIS
+
ij
2.18***
(0.179)
RUS
j
_CIS
+
i
1.25**
1.12*
(0.452)
(0.475)
RUS
i
_CIS
+
j
1.11**
0.98*
(0.376)
(0.406)
SIB
j
_SIB
i
2.26***
2.27***
(0.186)
(0.186)
proRUS_RUS
0.45
(0.355)
Constant
5.32***
2.96***
2.80***
2.77***
(0.843)
(0.802)
(0.811)
(0.812)
Observations
19,522
19,522
19,522
19,522
Number of I_ij
1352
1352
1352
1352
Economic Change and Restructuring (2021) 54:877–918
900

---

1 3
Finally, we investigate a dynamic form of the model, following the logic of HRM 
(2010), in the sense that we investigate the degree of decline over time of post-
colonial, and specifically post-Soviet trade flows. This is done by incorporation of a 
lagged dependent variable, and interactions thereof with post-Soviet dummies.
5 Results and discussion
5.1 Static estimation results
5.2 First baseline table: trade flows
‘
Appendix
’ Table 
10
 carries out estimation omitting country fixed effect dum-
mies. While this is not generally recommended in gravity analysis (as discussed 
in AVW (
2003
)), this allows us to include time-invariant monadic characteristics 
(such as landlockedness) and group dummies for the CIS+ countries as importers 
and exporters. Hence, it confirms the impression that, even after correcting for land-
locked location, the former Soviet countries are less open to trade than the average 
of the rest of the World, with the partial exception that they still trade relatively 
more with each other than with non-Soviet countries.
Table 
5
 shows our initial baseline estimation in terms of log trade flows, based 
upon (
11
). This shows the initial version of the static model, estimated by gener-
alized least squares (GLS), and the effects of incorporation of various post-Soviet 
dyadic dummies.
The four columns of Table 
5
show variants of the country fixed effects model. 
Note that most of the main estimated coefficients are stable and significant through-
out. Country-specific, time-invariant variables cannot be included due to collinearity 
with the fixed effect dummies, and we have to be careful not to carry out regressions 
on combinations of country groupings which are collinear with either exporter or 
importer dummies.
25
Trade flows have very stable and significant positive estimated elasticities of 0.14 
with respect to exporter GDP and 0.78 with respect to importer GDP. The exporter 
GDP elasticity is perhaps lower than in much of the literature, indicating that larger 
countries do not necessarily produce a greater number of product varieties (perhaps