Trade Potential and Trade Integration of the Russian Far East: A Regional Perspective

Over the past decade, the Russian government has embarked on an ambitious program of economic development in the Russian Far East (RFE), envisioning the transformation of the region into a hub for trade with the Asia Pacific. This paper explores the extent of RFE’s trade integration with both key partners around the world and the rest of Russia. In particular, we calculate the region’s trade potential on the basis of mean predicted values from a gravity model using three samples that offer different perspectives. Actual trade flows are then evaluated relative to the potential and the resulting index is analyzed for various years and countries. The results suggest that RFE exports to Northeast Asia have intensified over the period 2008-17, allowing the region to surpass its potential, although there is room to grow with respect to China. Moreover, the deepening integration with the Asia Pacific has been achieved at the expense of trade links with the rest of Russia. Lastly, imports from Japan and Korea are shown to be far below potential, although these two countries could be useful in promoting the economic development of RFE.


INTRODUCTION
In the three decades since the end of the Cold War, trade has been celebrated as an engine of economic growth and countries around the world have been encouraged to liberalize their trade relations, open their markets, and negotiate free trade agreements. Empirical evidence based on various country samples, time periods, and model specifications has lent support to this argument by showing that trade openness can facilitate the diffusion of knowledge and technology, improve efficiency, raise productivity, and promote growth (for surveys of the literature, see, for instance : Edwards, 1993;Rodriguez, Rodrik, 1999;Singh, 2010). Trade liberalization can also be employed at the subnational level to foster regional economic development. Such place-based policies could either target specific regions that are lagging behind or they could be part of an unbalanced growth strategy that favors more advanced regions in the hope of spillover effects. An example of the latter are China's Special Economic Zones (SEZ) that have generated local growth by attracting foreign investment, setting up export-oriented industries, and deepening their integration with world markets (Wang, 2013). Other emerging economies, such as India, have also established regional SEZs hoping to replicate China's success (Alkon, 2018).
This paper focuses on exploring the extent of trade integration of the Russian Far East (RFE) with the world and the rest of Russia. RFE is a federal district of Russia bordering China and the Asia-Pacific region 1 . It has a large territory (36% of Russia's total) and is rich in natural resources but low living standards and persistent outmigration have decimated the already sparse population and labor force. The central government declared the economic development of RFE a national priority and established a special federal ministry in 2012 tasked with promoting the economic and social advancement of the district. But the implicit motives behind this strategy were linked to the need of expanding Russia's trade with Northeast Asia via RFE following the slump in European demand caused by the global financial crisis (Minakir, 2017). Besides megaprojects like the "Power of Siberia" gas pipeline, a number of regional cooperation agreements were signed between RFE and China aimed at improving cross-border infrastructure and attracting Chinese investment in the region (Izotov, 2014). Recently, the federal government, hoping to boost RFE's trade openness and industrial capacity, introduced several place-based programs inspired by SEZ. Vladivostok, the district's largest city, was declared a free port, while Territories of Accelerated Socioeconomic Development (TOSER) offer various incentives, ПЭ № 4 2018

METHODOLOGY
The gravity model, which is widely used in the trade literature, postulates that trade flows between two countries are a function of three groups of factors which account for the traits of the exporter and importer, as well as for components that either facilitate or impair trade relations between them. Accordingly, a stochastic representation of the model can be defined as follows: 3 investigation and discusses various estimation strategies. Section 3 describes the data a descriptive statistics. Section 4 reports the results of the regression and the estimated indi potential. Section 5 summarizes the findings and draws conclusions.

METHODOLOGY
The gravity model, which is widely used in the trade literature, postulates that trade flo two countries are a function of three groups of factors which account for the traits of the e importer, as well as for components that either facilitate or impair trade relations between them. a stochastic representation of the model can be defined as follows: (1) where stands for the exports of country i to country j, which is proportional to the aggrega country i ( ), aggregate expenditure of country j ( ), and bilateral trade costs ( ). deno term, which has an expected value of 1 and is assumed to be independent of the regressors. Th , where 3 investigation and discusses various estimation strategies. Section 3 describes the descriptive statistics. Section 4 reports the results of the regression and the estima potential. Section 5 summarizes the findings and draws conclusions.

METHODOLOGY
The gravity model, which is widely used in the trade literature, postulates that two countries are a function of three groups of factors which account for the traits importer, as well as for components that either facilitate or impair trade relations betwee a stochastic representation of the model can be defined as follows: (1) where stands for the exports of country i to country j, which is proportional to the country i ( ), aggregate expenditure of country j ( ), and bilateral trade costs ( ).
term, which has an expected value of 1 and is assumed to be independent of the regre stands for the exports of country i to country j, which is proportional to the aggregate output of country i ( 3 investigation and discusses various estimation strategies. Sec descriptive statistics. Section 4 reports the results of the regre potential. Section 5 summarizes the findings and draws conclusion

METHODOLOGY
The gravity model, which is widely used in the trade lite two countries are a function of three groups of factors which importer, as well as for components that either facilitate or impair a stochastic representation of the model can be defined as follows: where stands for the exports of country i to country j, which country i ( ), aggregate expenditure of country j ( ), and bilat term, which has an expected value of 1 and is assumed to be ind ), aggregate expenditure of country j ( 3 investigation and discusses various estimation strategies. Sectio descriptive statistics. Section 4 reports the results of the regressi potential. Section 5 summarizes the findings and draws conclusions.

METHODOLOGY
The gravity model, which is widely used in the trade literatu two countries are a function of three groups of factors which acc importer, as well as for components that either facilitate or impair tra a stochastic representation of the model can be defined as follows: (1) where stands for the exports of country i to country j, which is country i ( ), aggregate expenditure of country j ( ), and bilatera term, which has an expected value of 1 and is assumed to be indep ), and bilateral trade costs ( 3 ion strategies. Section 3 describes the data and presents results of the regression and the estimated indices of trade nd draws conclusions. sed in the trade literature, postulates that trade flows between of factors which account for the traits of the exporter and facilitate or impair trade relations between them. Accordingly, defined as follows: (1) to country j, which is proportional to the aggregate output of try j ( ), and bilateral trade costs ( ).
denotes the error s assumed to be independent of the regressors. The aggregate ). (1) to country j, which is proportional to the aggregate output of try j ( ), and bilateral trade costs ( ). denotes the error s assumed to be independent of the regressors. The aggregate denotes the error term, which has an expected value of 1 and is assumed to be independent of the regressors. The aggregate output of the exporting country serves as a proxy for the size of its economy, accounting for total potential supply of goods, and is thus expected to have a positive effect on exports. The aggregate expenditure of the importing country reflects total potential demand for goods, which is also predicted to have a positive impact. The third variable, trade costs, is a broad category that can include geographic as well as policy-based factors. The anticipated sign of the variable depends on which factors are included in the model. Geographical distance between the trading partners, the presence of high mountains or deserts on the trade routes, and landlockedness would obviously impair trade relations, while geographic proximity and access to waterways would have the opposite effect. Moreover, low tariffs and non-tariff barriers, free trade agreements, cultural and political ties, and shared history and language can facilitate trade between countries.
The usual approach in the literature is to linearize Eq. (1) by taking natural logs. The resulting equation is given by: output of the exporting country serves as a proxy for the size of its economy, accounting for t supply of goods, and is thus expected to have a positive effect on exports. The aggregate expen importing country reflects total potential demand for goods, which is also predicted to have a po The third variable, trade costs, is a broad category that can include geographic as well as factors. The anticipated sign of the variable depends on which factors are included in Geographical distance between the trading partners, the presence of high mountains or deserts routes, and landlockedness would obviously impair trade relations, while geographic proximity waterways would have the opposite effect. Moreover, low tariffs and non-tariff barrier agreements, cultural and political ties, and shared history and language can facilitate trade betw The usual approach in the literature is to linearize Eq. (1) by taking natural logs. T equation is given by: Trade costs are generally broken down into various components that are mea quantitatively (e. g., distance) or via dummy variables (e. g., the presence of a free trade agre lack thereof). The estimation strategy typically employs OLS. However, Anderson and Van Wi show that OLS estimates are biased, if the model does not account for trade hurdles between th countries and their other trading partners not involved in the given bilateral exchange. Th multilateral resistance terms can be estimated via fixed effects in a panel regression (Rose and V 2001), and indeed fixed-effect models have become standard in the gravity literature.
We adopt the specification in Eq. (2) and adapt it to the objectives of the paper as follow . ( Trade costs are generally broken down into various components that are measured either quantitatively (e. g., distance) or via dummy variables (e. g., the presence of a free trade agreement, or the lack thereof). The estimation strategy typically employs OLS. However, Anderson and Van Wincoop (2003) show that OLS estimates are biased, if the model does not account for trade hurdles between the two trading countries and their other trading partners not involved in the given bilateral exchange. These so-called multilateral resistance terms can be estimated via fixed effects in a panel regression (Rose and Van Wincoop, 2001), and indeed fixed-effect models have become standard in the gravity literature.

ПЭ № 4 2018
We adopt the specification in Eq. (2) and adapt it to the objectives of the paper as follows: 4 show that OLS estimates are biased, if the model does not account for trade hurdles between countries and their other trading partners not involved in the given bilateral exchange. T multilateral resistance terms can be estimated via fixed effects in a panel regression (Rose and 2001), and indeed fixed-effect models have become standard in the gravity literature.
We adopt the specification in Eq. (2) and adapt it to the objectives of the paper as follo The dependent variable is the log of exports (in real terms), while the dependent variab aggregate output of the exporting and importing country proxied by their corresponding re costs are broken down into bilateral distance (� �� ), a dummy variable for contiguity (���� �� ) variable for home bias (���� �� ). 2 The coefficients for economic size and contiguity are positive, while the one for distance is predicted to have a negative effect on trade. Home shown to result in intranational trade flows being significantly higher than international trade Wei, 1996). Furthermore, Eq. (3) takes into account factors that vary across countries but not exporter and importer fixed effects (� � and � � , respectively). Similarly, factors that vary acro across countries are controlled for by including time-fixed effects (� � ). 2 The home bias variable takes the value of one, if the trade flows occur within a country, and zero, if border trade. This is an important variable when dealing with intranational trade between RFE and Rus 4 y variables (e. g., the presence of a free trade agreement, or the lly employs OLS. However, Anderson and Van Wincoop (2003) odel does not account for trade hurdles between the two trading not involved in the given bilateral exchange. These so-called d via fixed effects in a panel regression (Rose and Van Wincoop, become standard in the gravity literature. and adapt it to the objectives of the paper as follows: exports (in real terms), while the dependent variables include the orting country proxied by their corresponding real GDP. Trade e (� �� ), a dummy variable for contiguity (���� �� ), and a dummy efficients for economic size and contiguity are expected to be dicted to have a negative effect on trade. Home bias has been being significantly higher than international trade (Evans, 2001; account factors that vary across countries but not across time via � � , respectively). Similarly, factors that vary across time but not ing time-fixed effects (� � ).
, if the trade flows occur within a country, and zero, if it denotes crossdealing with intranational trade between RFE and Russia.

4
show that OLS estimates are biased, if the model does not account for trade countries and their other trading partners not involved in the given bila multilateral resistance terms can be estimated via fixed effects in a panel reg

2001), and indeed fixed-effect models have become standard in the gravity l
We adopt the specification in Eq.
(2) and adapt it to the objectives of The dependent variable is the log of exports (in real terms), while the dependent variables include the aggregate output of the exporting and importing country proxied by their corresponding real GDP. Trade costs are broken down into bilateral distance ( 4 The estimation strategy typically employs OLS. However, Anderson and Van Wincoop (2003) estimates are biased, if the model does not account for trade hurdles between the two trading their other trading partners not involved in the given bilateral exchange. These so-called sistance terms can be estimated via fixed effects in a panel regression (Rose and Van Wincoop, eed fixed-effect models have become standard in the gravity literature. opt the specification in Eq. (2) and adapt it to the objectives of the paper as follows: pendent variable is the log of exports (in real terms), while the dependent variables include the put of the exporting and importing country proxied by their corresponding real GDP. Trade en down into bilateral distance (� �� ), a dummy variable for contiguity (���� �� ), and a dummy ome bias (���� �� ). 2 The coefficients for economic size and contiguity are expected to be e the one for distance is predicted to have a negative effect on trade. Home bias has been lt in intranational trade flows being significantly higher than international trade (Evans, 2001;urthermore, Eq. (3) takes into account factors that vary across countries but not across time via mporter fixed effects (� � and � � , respectively). Similarly, factors that vary across time but not es are controlled for by including time-fixed effects (� � ). s variable takes the value of one, if the trade flows occur within a country, and zero, if it denotes crossis is an important variable when dealing with intranational trade between RFE and Russia.
), a dummy variable for contiguity ( 4 ). The estimation strategy typically employs OLS. However, Anderson and Van Wincoop (2003) LS estimates are biased, if the model does not account for trade hurdles between the two trading d their other trading partners not involved in the given bilateral exchange. These so-called resistance terms can be estimated via fixed effects in a panel regression (Rose and Van Wincoop, ndeed fixed-effect models have become standard in the gravity literature.
adopt the specification in Eq.
(2) and adapt it to the objectives of the paper as follows: ), and a dummy variable for home bias ( 4 show that OLS estimates are biased, if the model does not account for trade hurdles b countries and their other trading partners not involved in the given bilateral exch multilateral resistance terms can be estimated via fixed effects in a panel regression (R 2001), and indeed fixed-effect models have become standard in the gravity literature.
We adopt the specification in Eq. (2) and adapt it to the objectives of the paper The dependent variable is the log of exports (in real terms), while the depende aggregate output of the exporting and importing country proxied by their correspo costs are broken down into bilateral distance (� �� ), a dummy variable for contiguity ( variable for home bias (���� �� ). 2 The coefficients for economic size and contigu positive, while the one for distance is predicted to have a negative effect on trade shown to result in intranational trade flows being significantly higher than internatio Wei, 1996). Furthermore, Eq. (3) takes into account factors that vary across countries exporter and importer fixed effects (� � and � � , respectively). Similarly, factors that v across countries are controlled for by including time-fixed effects (� � ). 2 The home bias variable takes the value of one, if the trade flows occur within a country, and border trade. This is an important variable when dealing with intranational trade between RFE ) 1 . The coefficients for economic size and contiguity are expected to be positive, while the one for distance is predicted to have a negative effect on trade. Home bias has been shown to result in intranational trade flows being significantly higher than international trade (Evans, 2001;Wei, 1996). Furthermore, Eq. (3) takes into account factors that vary across countries but not across time via exporter and importer fixed effects ( 4 e biased, if the model does not account for trade hurdles between the two trading rading partners not involved in the given bilateral exchange. These so-called can be estimated via fixed effects in a panel regression (Rose and Van Wincoop, ct models have become standard in the gravity literature. cation in Eq. (2) and adapt it to the objectives of the paper as follows: (2) and adapt it to the objectives of the paper as follows: , respectively). Similarly, factors that vary across time but not across countries are controlled for by including time-fixed effects ( 4 show that OLS estimates are biased, if the model does not account for trade hurdles between the two tra countries and their other trading partners not involved in the given bilateral exchange. These so-ca multilateral resistance terms can be estimated via fixed effects in a panel regression (Rose and Van Winc 2001), and indeed fixed-effect models have become standard in the gravity literature.
We adopt the specification in Eq. (2) and adapt it to the objectives of the paper as follows: The dependent variable is the log of exports (in real terms), while the dependent variables include aggregate output of the exporting and importing country proxied by their corresponding real GDP. T costs are broken down into bilateral distance (� �� ), a dummy variable for contiguity (���� �� ), and a dum variable for home bias (���� �� ). 2 The coefficients for economic size and contiguity are expected t positive, while the one for distance is predicted to have a negative effect on trade. Home bias has shown to result in intranational trade flows being significantly higher than international trade (Evans, 2 Wei, 1996). Furthermore, Eq. (3) takes into account factors that vary across countries but not across time exporter and importer fixed effects (� � and � � , respectively). Similarly, factors that vary across time bu across countries are controlled for by including time-fixed effects (� � ). 2 The home bias variable takes the value of one, if the trade flows occur within a country, and zero, if it denotes c border trade. This is an important variable when dealing with intranational trade between RFE and Russia.
). Santos Silva, Tenreyro (2006) argue that OLS estimates of log-linearized models like the one in Eq. (3) produce biased estimates in the presence of heteroscedasticity. Although fixed effects control for heteroscedasticity, the log-linearization could still generate misleading results. Santos Silva, Tenreyro (2006) suggest instead estimating the gravity equation in levels by employing Pseudo Poisson Maximum Likelihood (PPML). This technique has since become a standard approach in the gravity literature to deal with the problem of heteroscedasticity 2 . Accordingly, we test the robustness of our results by using both the traditional OLS specification with fixed effects and PPML to estimate Eq. (3).
In the second step of the analysis, we calculate the fitted value of exports for a given country on the basis of the estimated coefficients from Eq. (3). The resulting number is the mean predicted value from the sample, which can be interpreted as the export potential of the country. More importantly, it can serve as a benchmark to evaluate actual exports relative to their potential. For this purpose, we follow De Benedictis, Vicarelli (2005) and define an index given by: ПЭ № 4 2018 5 basis of the estimated coefficients from Eq. (3). The resulting number is the mean predicted sample, which can be interpreted as the export potential of the country. More importantly, it benchmark to evaluate actual exports relative to their potential. For this purpose, we follow Vicarelli (2005) and define an index given by: is the fitted value of exports from country i to country j generated by Eq.
indicates that actual exports match the mean predicted value and therefore the export pote reached. Values less than one signal that the level of exports is below potential, while val imply that exports outperform the mean levels predicted by the sample.

DATA
where 5 basis of the estimated coefficients from Eq. (3). The resulting number is the mean pred sample, which can be interpreted as the export potential of the country. More importan benchmark to evaluate actual exports relative to their potential. For this purpose, we fo Vicarelli (2005) and define an index given by: where is the fitted value of exports from country i to country j generated by Eq.
indicates that actual exports match the mean predicted value and therefore the expor reached. Values less than one signal that the level of exports is below potential, whi imply that exports outperform the mean levels predicted by the sample. is the fitted value of exports from country i to country j generated by Eq. (3) 1 . A value of one indicates that actual exports match the mean predicted value and therefore the export potential has been reached. Values less than one signal that the level of exports is below potential, while values above one imply that exports outperform the mean levels predicted by the sample.

DATA
Data on Russia's bilateral trade flows over the period 2008-2017 are collected from the IMF's Direction of Trade Statistics (DOTS) database. Two different sets of data are available due to separate reporting by the exporting and the importing nations. In the literature, the two sets are usually averaged, but we opt instead for the data reported by Russia to ensure consistency with the regional trade series for RFE. RFE's bilateral trade flows with foreign countries are obtained from Russia's Federal Customs Service, while its trade with the rest of Russia is measured using data from Russia's Federal State Statistical Service (FSSS). Twenty-nine trading partners included in the sample account for more than 85% of Russia's exports and imports over the sample period 2 . RFE is added to the sample as the thirtieth trading partner. Therefore, Russia is now defined as an entity consisting of all of its regions bar the Russian Far East. All variables are adjusted accordingly and for the rest of the analysis we will refer to this new entity simply as Russia.
Bilateral trade flows are modeled as exports in line with Eq.
(3) and are initially measured in current US dollars. We convert them in real terms (constant 2010 US dollars) by employing the index of export prices as deflator, which is reported by the Economist Intelligence Unit 3 . In the absence of a regional export price index, RFE's exports are deflated by its local Producer Price Index (PPI RFE's trade is only a tiny fraction of Russia's exports (6%) and imports (3.7%). Moreover, the two entities differ markedly in the geographic structure of their trade relations. Russia trades predominantly with Europe (40-50%), while RFE's foreign economic relations are focused intensely on East Asia (70-80%). Geographic proximity certainly explains in part these patterns. However, only an empirical investigation can determine whether these levels of trade match RFE's potential.

RESULTS
The sample is divided into three groups. The Russia sample includes Russia's bilateral trade with foreign countries and RFE, while the RFE sample consists of the region's bilateral trade interactions with Russia and the world. The total sample combines the trade flows of Russia and RFE with the world and among themselves. First, we estimate the gravity model for each of the three samples, followed by calculations of RFE's trade potential with respect to various countries.

Gravity estimates
The results of the gravity model estimation are presented in table 2. For each sample we estimate two specifications ((1) OLS with fixed effects and (2) PPML) to test the robustness of the estimates. The goodness-of-fit is very high, while coefficients have mostly the expected signs and attain statistical significance across models and samples. The size of the exporting and importing economies has a large and significant positive effect on trade as anticipated, with the exception of the RFE sample where the exporting country's GDP does not attain statistical significance at conventional levels.  Note: Model (1) refers to OLS with fixed effects. Model (2) is estimated using PPML. All specifications include exporter, importer, and time fixed effects. Robust standard errors are in parentheses. * p < 0,10; ** p < 0,05; *** p < 0,01.

ПЭ № 4 2018
Distance represents a serious hurdle to trade as illustrated by the high magnitude, statistical significance, and negative sign of the corresponding coefficient. By contrast, contiguity has a positive and significant effect on trade except for the PPML estimates in the RFE sample. The substantial size of the home bias coefficient suggests that intranational trade between Russia and RFE is significantly higher than with other countries.
The estimates are generally consistent across model specifications and samples. However, OLS estimates exhibit larger magnitudes than PPML, especially in the Russia sample and the total sample. Given the bias of OLS estimates in the presence of heteroscedasticity and the advantages of PPML as described in the methodology section, we choose to employ the latter estimates in the calculations of the trade potential.

Estimates of trade potential
We focus first on RFE's trade potential with respect to foreign countries using the coefficients from the RFE sample and the total sample. The annual indices of export potential are shown in table 3. The numbers in the left panel evaluate RFE's actual exports to a given country relative to the mean predicted value of exports to all of RFE's major trading partners. In that sense, the average across countries and years of 1,03 suggests that RFE is exporting slightly above its estimated potential. However, this conceals considerable variation in sample.

ПЭ № 4 2018
In the early years, the export potential is below unity meaning that RFE was underperforming. The levels increase gradually, though not linearly, reaching values of around 1,5 in the period 2015-2017. In other words, RFE has managed to attain and exceed its potential in a matter of a decade.  In recent years RFE has exceeded its export potential with EU member states, despite the fact that these countries are less relevant as trading partners. No single country within the EU seems to be driving the results. The general picture is of rather low indices for most years interspersed with high performances in a few given years without a clear pattern. Belgium is an exception in that RFE's export potential begins at almost zero in 2008 and increases steadily to 1,8 in 2017. With regards to CIS, RFE's exports are far below their potential until 2013 when a dramatic reversal occurs with actual exports exceeding projected ones by a factor of between two and three. This is propelled initially by exports to Ukraine but Belarus and Kazakhstan are responsible for the high levels in the last two years ПЭ № 4 2018 of the sample. Lastly, RFE underperforms with respect to most other countries, although India and Brazil exceed unity in four of the ten years in the sample.
The right panel of table 3 displays the results for the total sample. RFE's export potential is now assessed against a benchmark derived from RFE's and Russia's trade with the world. In other words, Russia's trade patterns are now reflected in the mean predicted value, broadening the scope of comparison to include interactions between large economies. The most interesting changes in the estimates occur with respect to RFE's Northeast Asian neighbors. RFE's actual exports to China are now on average only 68% of the predicted mean value, whereby the trade potential is reached only in 2016. By contrast, exports to Japan and Korea surpass the potential in eight of the ten years of the sample. This result suggests that RFE as a region of Russia is not using its full potential for exports to China but its performance with respect to Japan and Korea is outstripping expectations. The high indices for Europe are determined almost entirely by excessive exports to Belgium, while nearly all other EU countries record levels far below potential 1 . Similarly, the rest of the world reaches the dismal level of 0,56 only thanks to RFE's exports to India exceeding unity in almost all years.
Next, we turn our attention to RFE's import potential presented in table 4. In the RFE sample, actual imports outstrip projected ones in six of the ten years, while in the total sample the indices are generally higher and exceed unity in every year of the sample. There is no clear increasing or decreasing tendency over time in either of the samples but estimates vary across countries. Judging by the RFE sample benchmark, imports from RFE's Northeast Asian neighbors reach potential levels only for 3-4 years and mostly in the period 2012-2014. In the last three years, all three countries underperform, which is also reflected in the average indices for the entire sample period. Adding Russia's trade with the world changes the picture dramatically. Against the new benchmark, imports from China exceed the predicted mean levels by a factor of 2 on average and do not drop below 1,5 in any given year, although a decline can be detected in the last three years. Japan and Korea, on the other hand, exhibit dismally low levels of 70% below potential on average.
As with exports, no single EU country dominates the results for the RFE sample. But in the total sample, indices, which are much higher and never dip below unity, are largely driven by imports from Finland, and to a lesser extent Poland and the UK. Similarly, the levels for CIS do not match the projected level in most years in the RFE sample, but are extremely high in the total sample, which is mostly due to imports from Belarus. For the remaining countries in the sample, the switch of benchmark reduces dramatically trade potentials. In the ПЭ № 4 2018 RFE sample, Brazil and Taiwan achieve higher import potentials, propelling levels for the group to above unity for the later years of the sample period. In the total sample, imports from the US are the only ones consistently exceeding unity in all years. The last part of the analysis deals with RFE's trade potential in intranational trade with Russia. Besides the RFE sample and the total sample, we now also include the Russia sample and show the results in table 5.
The Russia sample establishes Russia's trade with the world as a benchmark, allowing us to assess how well RFE is integrated with the rest of Russia relative to Russia's main trading partners. Russia's imports in columns 3 and 4 of table 5 represent RFE's export performance. Although RFE and the rest of the world reach the export potential in most years, the former exhibits higher indices, especially since 2013. In terms of imports, the results in columns 1 and 2 suggest that RFE does not perform as well because the rest of the world exceeds unity in seven out of the ten years, almost twice as much as RFE.
The RFE and total samples reveal very similar patterns. RFE's export performance with Russia declines steadily over the years, from a high of 1,5 in 2008 to a low of 0,3 in 2017. The rest of the world exhibits exactly the reverse tendency, rising from 0,3-0,4 at the start of the sample period to 1,5-1,8 ten years later. These numbers signal that RFE's integration via exports to Russia has been ПЭ № 4 2018 diminishing over time, dropping below potential in 2012-2014. This process was countered by deeper export relations with the rest of the world, which contributed to attaining and exceeding the potential level in 2013-2014. The pattern for imports is similar, although the change is less linear. The index for Russia declines in both samples but matches potential in 2015-2016 before dropping again. Imports from the rest of the world attain unity in 2012 and continue growing in the RFE sample, while they are consistently above unity throughout the sample period in the total sample.

CONCLUSION
Over the past decade, the Russian government has embarked on an ambitious program of economic development in RFE, envisioning the transformation of the region into a hub for trade with the Asia-Pacific area. On the one hand, RFE with its abundance of natural resources and its geographic proximity has some prerequisites to become integrated with the dynamics markets of Northeast Asia. On the other hand, its dwindling labor force, small market size, and onerous bureaucratic procedures are likely to discourage cross-border trade and investment. The question then becomes what RFE's potential for trade integration with its neighbors is and whether the region's effective trade flows live up to this potential. Another concern is the effect of a deeper integration with the Asia Pacific on RFE's trade relations with the rest of Russia. The paper addresses these issues by conducting an empirical investigation of RFE's trade flows with its major trading partners over the period 2008-2017. The trade potential is calculated on the basis of mean predicted values from a gravity model using three samples that offer different perspectives. Actual trade flows are then evaluated relative to the potential and the resulting index is analyzed for various years and sets of countries, allowing us to explore patterns and tendencies of RFE's exports and imports.
When viewed in isolation, as a separate entity trading with countries around the world, including Russia, RFE's export performance with respect to Northeast Asia shows consistent improvement over time. While initially exports fall short of the potential, they exceed it in the last few years of the sample period. In the more realistic scenario, where Russia is added to the sample as a trading entity in its own right alongside RFE, the pattern is similar, although there are differences between the three major foreign trading partners. Exports to China get increasingly closer to but almost never reach the benchmark, whereas exports to Japan and Korea surpass it in nearly all years. For most other destinations, RFE exports rarely exceed the potential, and when they do, the dynamics are usually not consistent over time.
The results for imports paint a different picture. By the standards of the RFE sample, imports from the three Northeast Asian countries overperform only over the period 2012-2014. But once the sample broadens to include Russia's trade with the world, RFE imports from China are almost 2-3 times larger than the potential, while those from Japan and Korea barely exceed 30% of this benchmark. Imports from other countries, like Finland and Belarus, exhibit considerably higher efficiency.
The estimates of the gravity model show unequivocally that there is a strong home bias in trade relations between RFE and the rest of Russia. However, regardless of the sample, RFE exports exceed their potential only until 2012-2013, after which they drop sharply. This tendency is a mirror image of the changes in RFE exports to the rest of the world. Imports from Russia underperform in most years of the sample, in contrast to those from the rest of the world.
Based on the findings of the paper, we can draw several conclusions. First, RFE exports to Northeast Asia have intensified over the period 2008-2017, allowing the region to surpass its potential, although there seems to be room to grow with respect to China. The Russian government could facilitate crossborder trade by further reducing non-tariff barriers and improving transnational ПЭ № 4 2018 infrastructure links. Second, the deepening integration with Northeast Asia has been achieved at the expense of trade links with the rest of Russia. This might appear worrisome, given the geostrategic importance of RFE for Russia. At the same time, it might simply reflect the fact that RFE's natural resource exports are increasingly diverted to the Asia Pacific, which is more efficient than to ferry them to Western Russia, where they might end up being re-exported to Europe. Similarly, it might be more efficient for RFE to import from China than from more distant parts of Russia. Third, imports from Japan and Korea are far below potential, although these two countries can play a key role in promoting the economic development of RFE.