1. Introduction
Growth in global import demand from developed countries has declined in recent years. The share of the EU-28 countries in global imports of goods declined from 18.4% to 14.8% between 2002 and 2016, similarly the share of the United States declined from 24.9% to 17.6% for the same period (Eurostat, Reference Head and Ries2018). Such a significant decline in the relative share of the high-income countries has encouraged exporting countries to seek new markets. However, uncertainty arising from information asymmetry is one of the major trade barriers in new destinations, most of which are developing countries. Language barriers, bureaucratic procedures, security concerns all contribute to the sunk cost a firm has to take into account when entering into a new market. Thanks to improvements in information and transportation technologies, those asymmetries across trade partners have declined over time. Nevertheless, countries still continue to increase the number of their high cost foreign missions (embassies, consulates, export promotion agencies etc.), mainly to support economic relations as well as to maintain the consular affairs or political interests (Rose, Reference Rose2007). Most embassies provide direct assistance to exporters through commercial and economic officers in areas such as translating the local legal framework, export opportunities, and even arranging introductions to local business persons.
There exists only a handful of studies that directly examine the causal relationship between costly foreign missions and bilateral trade across countries. Among those, Rose (Reference Rose2007) is one of the most relevant to the research question in this study. He carried out a cross-sectional analysis of the annual average bilateral trade from 22 large source countries to 200 destination countries for the years 2002–2003 to investigate if there is a causal relationship between the presence of an embassy and trade. Since the cross-sectional analysis does not control for country-pair fixed effects, he uses a wide range of control variables. He also tries to address reverse causality, employing instrumental variables that may correlate with the presence of foreign missions, including oil reserve capacity and the desirability index of destination country. The validity of these instruments is questionable, as they are likely to be correlated with trade volume as well, but Rose concludes that embassy effect on exports is positive and statistically significant with a magnitude ranging between 6% and 10%.
Segura-Cayuela and Vilarrubia (Reference Segura-Cayuela and Vilarrubia2008) also investigate the trade impact of foreign missions through a cross-sectional bilateral trade analysis, following a similar method to Rose (Reference Rose2007). However, the authors focus on the source of the impact by considering the extensive and intensive margins of the causal effect. They find that the presence of a foreign service increases exports by around 11–18% and that this increase originates from the formation of trade links (extensive margin).
Afman and Maurel (Reference Afman, Maurel, Bergeijk and Brakman2010) perform panel data analysis using new foreign mission openings in Eastern Europe after the dissolution of the Soviet Block. They specifically focus on the pair-wise trade between 26 OECD countries and 30 transition countries for three observation years 1995, 2000, and 2005, excluding within-group trade. The resulting impact is positive and high in magnitude (a new mission increases trade by around 40%); however, its statistical significance disappears when controlling for country pair and time fixed effects. Extracting causal interpretations from this analysis is not trivial since there are many other factors that may directly impact the trade volume of those transition countries during the observation period 1995–2005.
Head and Ries (Reference Moons and Bergeijk2010) carry out a single country analysis by investigating the impact the Canadian trade missions had on Canadian exports. The authors run panel data regressions for various treatment time spans (1–4 years). They handle reverse causality by controlling for unobserved characteristics with destination fixed effects and including the lagged dependent variable among the control variables. The study finds no statistically significant effect on Canadian exports.
Another single country analysis by Creusen and Lejour (2013) looks at the determinants of the entry decision of new exporter firms using the international trade transactions of Dutch firms between 2002 and 2008. They find a stronger effect such that the presence of foreign missions stimulates both the entry decision and volume of trade by 5–20%. In line with the literature on the role of experimentation in exporting (e.g., Albornoz et al., Reference Albornoz, Calvo Pardo, Corcos and Ornelas2012), the adjustment process revealed by the data confirms that some firms exit after the first trial, while those who find staying in the new market profitable increase their trade volume over time.
Finally, a meta-analysis by Moons and Bergeijk (Reference Creusen and Lejour2013) compares 32 empirical studies on the impact of economic diplomacy on trade and investment, including embassies, consulates and other diplomatic facilities, investment and export promotion offices, trade and state visits. They conclude that primary studies conducted on a single country basis (panel data of a single country) will in general show a lower significance of the coefficient of interest and studies tend to generate more significant results when the dependent variable is exports rather than imports or foreign direct investment.
The majority of the above-mentioned empirical studies use cross-sectional bilateral trade data due to lack of variation overtime. However, unless a valid instrumental variable approach is employed, cross-sectional analysis is subject to simultaneity bias because of the reverse causality between the size of trade and presence of foreign missions. The case of Turkey, in this sense, is unique as it allows for panel data analysis thanks to significant variation in its foreign embassy network over a relatively short time period. Furthermore, research of the literature indicates that there are no studies that investigate the impact of foreign missions from an emerging country perspective. The example of Turkey may provide an insight for other developing countries that need to allocate their already limited resources efficiently across various policy options.
Turkey has made a significant change in its foreign policy objectives since the beginning of the twenty-first century. The main goal of this change was to develop a multi-dimensional and more proactive foreign policy, which in part led to the expansion of relations with long-time neglected Sub-Saharan Africa, Latin America, and the Asia-Pacific regions. Accordingly, Turkey declared 2005 as the ‘Year of Africa’ and prepared a strategic plan drafting an opening policy towards Africa. Likewise, the year 2006 was declared as the ‘Year of Latin America and the Caribbean’, and in line with these developments, Turkey also put into effect, in recent years, policies targeting the Asia-Pacific region. As a result of these expansion policies, Turkey opened 39 new foreign embassies in the aforementioned regions between 2008 and 2015. Table 1 shows the list of those countries and the year of embassy opening.
List of the Countries with New Embassies by Opening Year
The purpose of this paper is to estimate whether Turkey gained positive returns in terms of export opportunities by answering the following questions: Does the existence of an embassy impact exports to (imports from) that country? Does this effect vary across differentiated versus non-differentiated products? If there is an impact on exports (imports), what are the roles of the extensive margin (number of export varieties based on the 6-digit HS 2002 classification) and intensive margin (volume effect)? And, finally, does the number of exporting firms increase?
The estimation is performed in a panel difference-in-differences setting by introducing the presence of an embassy as a control variable to the standard gravity equation. A country-specific time trend is also added as a control variable in order to deal with the possible pre-existing export (import) growth trends. The estimation results suggest that the presence of an embassy increases exports to that country by 30% while having no material impact on imports. The increase in exports is driven mainly by the exports of differentiated goods and is largely due to the volume effect (intensive margin). Finally, the number of exporting firms rises by almost 8%. The results are robust to various modifications in the baseline specification and sample size.
2. Empirical Strategy
A panel data difference-in-differences estimation method is employed to estimate the trade outcomes. Following is the estimating equation for embassy impact:
$$lnY_{it} = \alpha E_{it} + \beta W_{it} + \rho _iI_iT + \gamma _t + \mu _i + + \varepsilon _{it}$$where lnY it is the natural logarithm of the consumer price index (CPI) adjusted export (import) value or the number of export varieties, based on the 6-digit HS 2002 classification, from Turkey to country i at time t, E it is the indicator of the presence of a Turkish embassy in the partner country i at time t, W it is a vector of the constant term and time variant explanatory variables, including the natural logarithm of CPI adjusted gross domestic product (GDP), an indicator of a state visit to country i from the Turkish Government at the President level, the presence of a Turkish development agency, and the existence of a direct flight from Turkey to country i. Footnote 1ρ i’s indicates the effects of the country-specific time trend (I iT), γ tand μ i are time and country fixed effects, and εit is the unobserved error term. The key parameter is α, which represents the percentage change in the outcome variable due to presence of an embassy.
The main issue that arises in such settings is the reverse causality between exports and the opening of an embassy in a country. If an embassy is opened because of a pre-existing export growth trend, the estimated embassy coefficient will be upward-biased. This should be less of a concern in the case of Turkey because opening of 39 new embassies in just a few years was in large related to a new foreign policy targeting the countries in Sub-Saharan Africa, Latin America, and Asia-Pacific regions as discussed in the introduction. However, I carried out the following exercise to check if reverse causality exists. I regressed the natural logarithm of exports on a set of dummies corresponding to +1, +2, +3, +4, +5 years since the opening and –1, –2, –3, –4, –5 years before the opening of an embassy together with country fixed effects and GDP. The goal of this exercise was to see if export growth after the opening of the new embassies represented a significant departure from the pre-existing trends. Figure 1 shows the coefficients and confidence intervals for the before and after dummies. The departure from the pre-existing level after embassy opening is visually evident. All dummies representing the years after embassy opening are positive and statically significant, whereas the dummies representing the years before embassy opening are negative. This clear shift in level suggests that reverse causality is not likely to impact our estimates. Nevertheless, much of the robustness analysis below will focus on whether reverse causality exists.
Before and after analysis of the embassy impact
2.1 Data and Summary Statistics
The export and import data at the 6 digit and country levels are obtained from the Turkish Statistical Institute (Turkstat). The number of exporting firms by country is retrieved from the Entrepreneurship Information System of the Ministry of Industry and Technology. The export and import data are further categorized as homogenous goods and differentiated goods using the 2007 version of Rauch (Reference Rauch1999) classification of goods. Annual GDP values are obtained from the UNstats database (United Nations Statistics Division, 2018). Country specific development indices are gathered from the World Bank World Development Indicator (World Bank, 2018). Historical data on the presence of embassies and consulates and information on the regulatory process were obtained from the Prime Minister's Office in Turkey. Information on the other types of foreign diplomacy institutions was collected from the annual reports of the related organizations.
Table 2 presents the summary statistics of various outcome and control variables for the treatment and control countries before and after the treatment period. A country is defined as a treatment country if a new Turkish embassy was opened during the observation period. The remaining countries are considered as the control group but divided into two sub-groups. The first consists of countries that never had a Turkish embassy and the second group consists of countries that already had an embassy before 2008.
Summary statistics
Descriptive evidence from the simple log mean comparison in Table 2 suggests that the exports to the treatment countries increased more than the exports to the control countries between 2008 and 2016. Divergence is larger between the treatment countries and the countries that never had a Turkish embassy. Also, the differential increase in the exports of differentiated goods and in the number of export varieties (extensive margin) is greater, which is expected since the associated information asymmetry is greater for those types of goods. However, descriptive evidence also indicates a differential increase in the GDP of the treatment countries relative to the GDP of control countries. Thus, a causal relationship between the mentioned trade outcomes and the presence of embassies can only be argued after controlling for other time-varying factors that may also impact the outcomes, such as GDP, presence of direct international flights, and state visits.
3. Results
3.1 Export Results
Table 3 shows the baseline regression results from the specification in equation (1). The first column presents the estimation result of a Poisson maximum likelihood estimation when the level of exports is regressed on the natural logarithm of GDP and the embassy, time, and country dummies. The dependent variable is the natural logarithm of exports and the estimations are from ordinary least square specifications. Each column differs in terms of the set of control variables. The lagged dependent variable in column (3), region specific trends in column (4), country specific time trends in column (5), time varying foreign diplomacy measures other than the embassies in column (6), and finally time varying country development indices such as political stability in column (7) are introduced as additional control variables. The embassy effect is stable and statistically significant across all specifications. Column (7) represents the richest set of control variables but the sample size drops substantially since the development indices are not available for some countries. Coefficients on the development indicators are all not statistically significant probably because these indicators are reflected in GDP values. Therefore, the column (6) is preferred as the baseline specification in order to keep the sample size large as much as possible. The coefficient of interest in the preferred specification is statistically significant at the 5% level, which implies that the presence of an embassy in the partner country increases exports to that country by almost 30%. Other foreign diplomacy factors are all statistically insignificant.
Embassy Impact on the Log of Total Exports
Notes: Clustered standard errors by country in parenthesis *** p < 0.01, ** p < 0.05, * p < 0.1. Column (1) presents estimates from a Poisson pseudo-maximum Likelihood estimation whereas the rest of the columns are from OLS estimations. All variables are annual aggregates between 2006 and 2016. Countries receiving small Turkish exports that are small (less than a thousand dollar annually) are dropped from the sample in OLS estimates. All monetary variables are deflated using US Consumer Price Index. The sixth column is the preferred specification.
Table 4 presents the regression results of the preferred specification (column 6 of Table 3) when exports are categorized as differentiated versus homogenous goods according to the Rauch (Reference Rauch1999) classification. The outcome variables in the first three rows are the natural logarithm of all exports, differentiated exports, and homogenous exports, respectively. Regression results show that the impact of embassy opening is statistically significant for differentiated goods exports and the magnitude of impact is larger than the impact on overall exports. On the other hand, there seems to be no statistically significant impact on homogenous goods exports. This is consistent with the fact that the information asymmetry associated with differentiated goods is greater than that for homogenous goods.
Decomposition of the embassy impact
Notes: Clustered standard errors by country in parenthesis *** p < 0.01, ** p < 0.05, * p < 0.1. All regressions are run on the same set of control variables from the preferred specification in Table 3 column (6). The total amount of variety represents the total number of export industries at 6 digit HS2002 classification by partner country
In the next three rows, I carried out a simple decomposition exercise to investigate if the change in exports originated from the variety effect (extensive margin) or the volume effect (intensive margin). Rows 4–6 in Table 4 show the results of regressions when the dependent variable is replaced with the total number of export varieties based on the 6-digit HS 2002 classification, which will give the share of the extensive margin in the overall export surge.Footnote 2 The results indicate that the presence of an embassy has a statistically significant impact on export varieties with a magnitude of 9.9%. This implies that approximately one-third of the embassy impact on total export value is due to the extensive margin and the rest is due to volume effect. The extensive margin share of differentiated goods exports is in line with the total exports as shown in row (6). However, the variety effect dominates in homogenous goods exports even though the impact on value of total homogenous goods exports is not statistically significant. This suggests that opening an embassy helps homogenous goods exports by introducing new varieties even if the associated information asymmetry is lower for those goods.
Finally, row (7) in Table 4 shows the estimated coefficient when the number of exporting firms is regressed on the same set of variables. The coefficient implies that the presence of an embassy in the partner country increases the number of exporting firms to that country by almost 8%. This result can be interpreted such that the larger the role of the intensive margin in explaining the embassy impact is partly because of the increase in the number of entrant firms.
3.2 Robustness Checks
Table 5 represents the results of various robustness checks when the observation period and the sample of countries used in the regressions are reorganized. The goal in undertaking this exercise was to check whether the result was driven by a particular year or group of countries. The first row is the original model. The second and third rows show estimates when the observation period was extended or shortened. Rows (4) through (7) represent results when the high-income countries, countries with very small population, countries that had an embassy prior to 2008, and countries that never had an embassy as of 2015 are dropped from the original sample, respectively. In all specifications, the results are robust to the original model in terms of the sign and significance level but the magnitude changes slightly.
Robustness checks for the overall embassy impact
Notes: Clustered standard errors by country in parenthesis *** p < 0.01, ** p < 0.05, * p < 0.1. Each row represents a regression that differ only in sample. All regressions are run on the same set of control variables from the preferred specification in the Table 3, column (6).
I further run a placebo test by changing the observation period from 2006–2016 to 2000–2010 while keeping the treatment variable (embassy dummy) the same, meaning that the embassy dummy at year t for this exercise reflects the presence of an embassy at year t + 6. The results are all statistically insignificant both for extensive and intensive margins.
3.3 Import Results
A new Turkish embassy may help to reduce the information asymmetry not only for Turkish exporters but also for importers and exporters from the partner country. In addition, 28 countries out of 38 countries with a new Turkish embassy did also open their embassies in Turkey simultaneously or in the following years due to reciprocity. Table 6 represents the results when the imports by Turkey are regressed on the same set of control variables in the preferred specification in export estimation (column 6 in Table 3). The sign of the embassy coefficient is positive but not statistically significant. This is probably due to the fact that most of the countries with new embassies are small and low income countries, the exports of which are mainly composed of homogenous goods that are less prone to information asymmetry. Indeed, when homogenous goods imports are the outcome variable (column 3), the coefficient becomes significantly larger and the t-value improves but not enough to make the coefficient statistically significant at the 10% level. Controlling for reciprocal embassy opening does not alter the results.
Embassy impact on imports
Notes: Clustered standard errors by country in parenthesis *** p < 0.01, ** p < 0.05, * p < 0.1. All regressions are run on the same set of control variables from the preferred specification in the Table 3, column (6).
4. Conclusion
This study analyzes the impact of the presence of embassies on trade outcomes by exploiting the significant variation in Turkey's foreign mission network between 2008 and 2015. Departing from the existing studies, this study employs a panel difference-in-differences strategy that exploits the size of the variation over a short time period. After controlling for country-specific time trends and standard gravity covariates, I found that the presence of an embassy increases exports by 30% and has no statistically significant impact on imports. The increase in the level of exports comes from differentiated goods exports and is largely explained by the volume (intensive margin) effect. The extensive margin (export varieties) channel explains only one-third of the rise in exports. The greater role of the intensive margin in explaining embassy impact is partly due to the increase in new entrant firms.
Acknowledgements
I thank Devashish Mitra, Mary E. Lovely, Perry Singleton, and seminar participants at Syracuse University. I would also like to thank the anonymous referees for their insightful guidance. The views expressed here are of the author and do not necessarily reflect those of the Central Bank of the Republic of Turkey. The author declares that he has no conflict of interest.
