ij,t

= 1 if a PTA is in force between i and j at year t,

0 otherwise.

It is important to note that, due to the fixed effects structure described in sec- tion 3, the analysis only captures the effects of PTAs that came into force during the period under study (1995-2020). For Catalonia, this includes 59 new PTAs that were implemented during this time frame. Table A1 in the Appendix lists the coun- tries with which Catalonia established a new PTA, as well as the year in which each agreement was enacted.

Bilateral Investment Treaties

Bilateral Investment Treaties (BITs) are agreements primarily aimed at enhancing investment between the signatory countries or jurisdictions. However, recent research has shown that BITs can also have significant effects on trade flows (see Heid and Vozzo, 2020), making them relevant for inclusion as explanatory variables in this study.

The data on BITs is sourced from the United Nations Conference of Trade and Development (UNCTAD) Investment Policy Hub. Similar to the treatment of PTAs, a BIT is considered to include Catalonia as a signatory if Spain is a signatory. BITs are also modeled as binary variables:

ij,t

= 1 if a BIT is in force between i and j at year t,

0 otherwise.

It is important to emphasize that the empirical specification used in this study only captures the effects of Bilateral Investment Treaties (BITs) that came into

force during the analysis period (1995–2020). For Catalonia, this includes 48 BITs. A detailed list of the countries that entered into new BITs with Catalonia, along with the year each agreement came into force, is provided in Table A2 in the Appendix.

A potential concern when including both PTAs and BITs as explanatory vari- ables is the possibility of overlap between the two. In fact, for 22 countries, both a PTA and a BIT with Catalonia came into force during the sample period. How- ever, it is worth noting that for none of these cases did the PTA and BIT come into effect in the same year. This temporal separation helps to mitigate concerns about confounding effects, allowing the study to more accurately isolate the impact of each type of agreement on trade flows.

Accio

Accio is a Catalan governmental agency whose primary mission is to enhance the competitiveness of Catalan firms by promoting innovation, supporting internation- alization, and attracting foreign investment.

For the purposes of this study, the presence of Accio delegations will be captured using a binary variable:

Accio_ij,t

= 1 if there is a Accio delegation between i and j at year t,

0 otherwise.

Of the full network of 41 Accio delegations, this study is only able to estimate the impact of 13 delegations that were opened during the analysis period (1995–2020). These newly established delegations allow for a before-and-after comparison, helping to isolate the effect of Accio on trade flows. A complete list of the countries where new Accio delegations opened, along with the corresponding years, is provided in Table A3 in the Appendix.

Gravity and other controls

The gravity dataset is sourced from CEPII, a widely-used resource for obtaining key gravity model controls such as distance between trading partners, GDP, contiguity (whether countries share a border), and diplomatic disagreement. These variables are mainly used as controls.

In addition, OECD membership is included as a proxy for distinguishing between developed and developing countries, allowing for an examination of the heterogene- ity in trade effects based on the level of economic development. Furthermore, the product classification system from Rauch (1999) is used to differentiate between homogeneous and differentiated goods.

2.3 Descriptive Statistics

This section presents the summary statistics of the dependent variable and the ex- planatory variables, as shown in Table 1.

The dependent variable, the trade flows between country i and j¹, in sector k at year t, denoted as X_ijk,t, has an average value of 222 thousand euros. However, it exhibits a large variance due to the significant number of zero trade flows in many observations. This high proportion of zero flows justifies the use of a count-data model for the analysis.

The main variables of interest, PTA_ij,t, BIT_ij,t and Accio_ij,t, are binary. Their mean values are 0.30, 0.32, and 0.16, respectively, reflecting the proportion of obser- vations where these policies are in place.

Note: PTA_ij,t, BIT_ij,t, Accio_ij,t, OECD_i, Contiguity_ij and Differentiated Product_k are binary variables; GDP_i, t is expressed in thousands of US dollars ; Trade flows (X_ijk,t) is expressed in thousands of euros; Distance_ij is expressed in kilometers.

3 Empirical analysis

The main empirical specification is based on the structural gravity model as de- scribed in Anderson and Van Wincoop (2003). A key aspect to highlight is that the dependent variable, representing trade flows, is a count variable. Consequently, the

¹Note that by definition, either i or j is always Catalonia

estimation employs a Pseudo-Poisson model to account for the nature of the data. Additionally, given the high-dimensional index structure of the dependent variable, the model includes fixed effects to control for unobserved heterogeneity across coun- tries and products.

The baseline regression is specified as follows:

Xijk,t = exp(PTAij,tβPT A + BITij,tβBIT + Accioij,tβAccio + Diplomaticij,tβDiplomatic

+ β_XX_i,t + β_zZ_j,t + η_i + γ_j + ν_k + ϵ_ijk,t), (1)

where X_ijk,t are trade flows from country i to country j in sector k at year t. The variables PTA_ij,t, BIT_ij,t, Accio_ij,t and Diplomatic_ij,t represent the main policy variables of interest, which are expected to influence trade flows. X_i,t and Z_j,t capture control variables that account for country-specific characteristics that may vary over time.

All parameters β are regression coefficients reflecting the impact of the respective variables. The model incorporates fixed effects denoted by η_i for the origin country, γ_j for the destination country, and ν_k for the product-specific effects. Finally, ϵ_ijk,t is an idiosyncratic disturbance term.

It is important to remark that the current study examines trade flows exclusively from a single country perspective, specifically focusing on Catalonia as the origin for exports and the destination for imports. This unique framework implies that some of the traditional gravity based fixed effects either cannot be included or are already nested in the fixed effects of equation 1.

For instance, when conducting a regression on Catalan exports, it holds that i = Catalonia for all i. This implies the following: (1) it is not necessary to include i-specific fixed effects; (2) it is impossible to include destination-time fixed effects and, at the same time, obtain coefficients on the variables of interest; (3) the country- pair fixed effect that is often used in the literature is equivalent to the inclusion of j-specific fixed effects.

Given these structural characteristics, the parameters β in equation 1 capture the time variation of the variables of interest. In other words, this translates to a before- and-after comparison within the same destination country (for Catalan exports) or origin country (for Catalan imports).

4 Results

This section presents the key findings of the study. It begins with the baseline results, which provide an overall view of the effectiveness of the trade-promoting tools under consideration. Following this, some results related to country and product heterogeneities are introduced, offering deeper insights into how these tools perform across different contexts.

The analysis of country and product-level heterogeneities helps to unravel the underlying mechanisms of the trade-promoting policies. By understanding how these tools vary in effectiveness depending on the characteristics of the trading partner (such as their level of development) and the nature of the products (differentiated or homogeneous), the study provides a better understanding of the mechanisms behind these policies.

4.1 Baseline

This section presents the baseline results derived from estimating equation 1. Ini- tially, a naive approach is employed in Table 2, where the equation is estimated without incorporating any fixed effects. While this approach may introduce bias in some of the coefficients (as noted by Felbermayr and Yotov, 2021), it remains a valuable exercise. The absence of fixed effects allows for the estimation of effects for time-invariant variables, which would otherwise be absorbed when fixed effects are included. This provides an initial understanding of the relationships at play before refining the model with more robust methods.

Table 2 is organized into three columns. The first column reports results for all trade flows (Catalan imports and exports combined), while the second and third columns focus exclusively on Catalan imports and exports, respectively.

Due to the omission of fixed effects, Table 2 includes results for time-invariant controls like Distance and Contiguity. As expected, Distance has a negative impact on trade, while Contiguity positively affects trade, both in line with gravity theory and existing literature. Additionally, the GDP of both the origin country (i) and the destination country (j) positively influences trade flows, reinforcing the idea that larger economies trade more.

The table also highlights the role of diplomatic factors and trade policies: Diplo- matic disagreement acts as a deterrent to trade, while PTAs and Accio delegations promote trade. Interestingly, the results for Bilateral Investment Treaties (BITs) show a negative coefficient, suggesting that BITs might decrease trade. However, this is likely due to the naive approach used here, which is susceptible to omitted

Robust standard errors are reported in parentheses.

• p < 0.1, ** p < 0.05, *** p < 0.01

included.

In Table 3, the focus is on time-variant trade policies, with the results suggesting that PTAs positively impact overall trade, particularly Catalan exports. This is evident from the positive and highly significant PTA coefficients, with the coefficient in the exports column being larger than that in the imports column, indicating that PTAs have a stronger effect on exports. Similar patterns are observed for BITs and new Accio delegations, which also show positive effects on trade flows.

The coefficients in this table should be understood in a before-and-after context,

²In the imports column, only the country of origin fixed effect is included, as adding destination fixed effects would be redundant since Catalonia is the sole destination. Similarly, in the exports column, only destination fixed effects are used.

Robust standard errors are reported in parentheses.

• p < 0.1, ** p < 0.05, *** p < 0.01

comparing the impact of these policies over time. Since all variables are binary indi- cators, the coefficients can be directly interpreted as semi-elasticities. For instance, the 0.226 coefficient for Accio in the exports column suggests that opening a new Accio delegation in a specific country is associated with a 22.6% increase in exports to that country. A similar interpretation applies to the coefficients for PTAs and BITs.

Interestingly, the magnitudes of the coefficients for traditional trade and invest- ment policy instruments (PTAs and BITs) align with those found in previous studies that report global averages (see Heid and Vozzo, 2020; Nagengast and Yotov, 2023). This consistency reinforces the validity of the findings and suggests that these policies have a comparable effect on trade in the Catalan context.

Overall, Table 3 suggests that PTAs, BITs, and Accio delegations all have a positive effect on trade. The results highlight their effectiveness in promoting both imports and exports, with a particularly strong impact on Catalan exports.

In the next two sections, country and product heterogeneities are examined to bet- ter understand the mechanisms behind the success of these treaties and delegations. By exploring how these tools perform differently depending on the characteristics of the trading partner and the nature of the goods being traded, the analysis will

provide deeper insights into why and how these policies work.

4.2 Country heterogenties

This section investigates country-level heterogeneities, examining how the three trade- promoting tools (PTAs, Accio delegations, and BITs) may have varying effects de- pending on the level of development of the trading partner. The main idea is that these tools could influence trade differently depending on whether the trading part- ner is more or less developed. To capture this, the level of development is measured by whether the trading partner is a member of the OECD.

Table 4 presents the heterogeneous effects of the trade-promoting tools by in- teracting each tool with the OECD membership of the trading partner. The most critical components of Table 4 are, therefore, these interaction terms, which should be interpreted relative to the excluded category, i.e., non-OECD members.

For PTAs, the results reveal a negative and highly significant coefficient, indicat- ing that PTAs are more effective in boosting trade when the trading partner is not an OECD member. In other words, PTAs appear to generate a larger increase in trade with non-OECD countries.

As for BITs, the results do not show any significant heterogeneity based on the development level of the trading partner, suggesting that BITs have a uniform effect regardless of whether the partner is an OECD member or not.

Finally, the Accio delegations exhibit a notable asymmetry: they have a stronger effect in promoting Catalan exports to OECD countries, while they have a greater impact on imports from non-OECD countries. This suggests that Accio delegations are particularly useful for facilitating exports to developed countries, but they play a different role when it comes to imports.

4.3 Product heterogenties

This section explores product-level heterogeneities, focusing on how the three trade- promoting tools, namely, PTAs, Accio delegations, and BITs, may exert varying effects depending on certain product characteristics. The central premise is that these tools may influence trade differently based on whether a product is differentiated or homogeneous.

Specifically, the product differentiation classification proposed by Rauch (1999) is employed to distinguish between differentiated and non-differentiated products. The underlying idea is that differentiated products possess unique characteristics, which complicate the assessment of their quality or fair prices. In contrast, homogeneous

products are those with established reference prices or those traded on organized exchanges, making their evaluation straightforward.

Table 5 presents the results for this exercise by means of interaction terms. The interaction terms are particularly interesting and a close examination reveals that both PTAs and Accio delegations show positive and significant coefficients. This indicates that these two tools play a crucial role in enhancing trade, especially for differentiated goods.

Interestingly, the coefficients related to new Accio delegations are substantially larger than those for PTAs. This could be explained by the nature of Accio delega- tions, which, unlike PTAs, do not modify tariffs or introduce new trading regulations. Instead, they function as key providers of information. As a result, their influence is particularly pronounced in the trade of differentiated goods, where access to detailed information is essential.

Lastly, Table 5 shows no significant heterogeneity in the effects of BITs based on product characteristics, suggesting that BITs do not exhibit a differentiated impact across various types of goods.

5 Conclusions

In conclusion, this study provides valuable insights into the effectiveness of trade- promoting tools such as Preferential Trade Agreements (PTAs), Bilateral Investment Treaties (BITs), and Accio delegations for the Catalan economy. The results indicate that all three instruments have a positive impact on trade flows.

The study’s findings are consistent with gravity theory and prior research, re- inforcing the validity of the results and the relevance of trade-promoting tools in contemporary economic interactions. The coefficients associated with PTAs and BITs align closely with global averages reported in previous studies, suggesting that the Catalan context reflects broader trends observed in international trade.

Moreover, the exploration of product and country-level heterogeneities shows the mechanisms underlying the effectiveness of these trade policies. The findings suggest that PTAs are particularly beneficial when engaging with non-OECD countries, high- lighting their role in reducing trade barriers in less developed contexts. Conversely, Accio delegations appear to significantly enhance exports to OECD countries.

Furthermore, the analysis of differentiated goods reveals critical insights into how trade-promoting tools function in relation to product characteristics. The results indicate that PTAs and Accio delegations have a more pronounced effect on trade flows for differentiated products compared to homogeneous ones. This is particularly

important, as differentiated goods often present challenges in terms of quality assess- ment and pricing, making the provision of information and reduced trade barriers essential for successful market entry.

In summary, this research highlights the significant role of PTAs, BITs, and Accio delegations in enhancing trade. By understanding how these tools function across different contexts, policymakers can better design and implement trade agreements and initiatives that effectively boost economic activity. Future research could expand on these findings by exploring additional variables or contexts, further enriching the discourse on international trade policy.

Table 4: Catalan trade determinants: Trading partner OECD membership

Robust standard errors are reported in parentheses.

• p < 0.1, ** p < 0.05, *** p < 0.01

Robust standard errors are reported in parentheses.

• p < 0.1, ** p < 0.05, *** p < 0.01

References

Anderson, James E and Eric Van Wincoop, “Gravity with gravitas: A solution to the border puzzle,” American economic review, 2003, 93 (1), 170–192.

Bagir, Yusuf Kenan, “Impact of the presence of embassies on trade: Evidence from Turkey,” World Trade Review, 2020, 19 (1), 51–60.

Baier, Scott L and Jeffrey H Bergstrand, “Determinants of free trade agree- ments,” Journal of international Economics, 2004, 64 (1), 29–63.

Egger, Peter H and Gerard Masllorens, “Deep Trade Agreements and Firm Ownership in GVCs,” The World Bank Economic Review, 2024, p. lhae003.

, Katharina Erhardt, and Gerard Masllorens, “Backward versus forward integration of firms in global value chains,” European Economic Review, 2023, 153, 104401.

Felbermayr, Gabriel and Yoto V. Yotov, “From theory to policy with gravi- tas: A solution to the mystery of the excess trade balances,” European Economic Review, 2021, 139, 103875.

Heid, Benedikt and Isaac Vozzo, “The international trade effects of bilateral investment treaties,” Economics Letters, 2020, 196, 109569.

Hofmann, Claudia, Alberto Osnago, and Michele Ruta, “Horizontal depth: a new database on the content of preferential trade agreements,” World Bank Policy Research Working Paper, 2017, (7981).

Nagengast, Arne and Yoto V Yotov, “Staggered difference-in-differences in grav- ity settings: Revisiting the effects of trade agreements,” 2023.

Rauch, James E, “Networks versus markets in international trade,” Journal of international Economics, 1999, 48 (1), 7–35.

Rose, Andrew K, “The foreign service and foreign trade: embassies as export promotion,” World Economy, 2007, 30 (1), 22–38.

Appendix

A Data Tables

Table A1: New PTAs with Catalonia during 1995-2020 by country

Year Country 1999 Brazil

2000 Canada

2003 Egypt

2012 Colombia

2012 South Korea

2014 Ghana

2014 Peru

2015 Israel

2015 Panama

2017 Croatia

2017 Kenya

2017 Netherlands

2017 Serbia