Temporary extra-regional linkages and export product and market diversification

ABSTRACT While considerable attention has been paid towards how some long-lasting extra-regional linkages such as foreign direct investments facilitate interregional knowledge transfer, less is known about the role of temporary extra-regional linkages formulated via trade fairs. This paper seeks to fill this gap by focusing on the China Import and Export Fair and examining the relationship between temporary extra-regional linkages forged by trade fairs and regional export dynamics. In doing so, we also speak with the literature on ‘temporary clusters’. Our second contribution is to differentiate export market and product diversification, and to show distinct effects of temporary extra-regional linkages on those two.


INTRODUCTION
The evolutionary economic geography (EEG) literature has stressed that regions tend to develop new industries that are closely related to pre-existing industries, since knowledge spillovers and resource sharing are more likely to occur between related industries with cognitive proximity (Boschma, 2005;Boschma et al., 2013;Neffke et al., 2011). This idea of related diversification or 'regional branching' indicates that regional industrial diversification can be seen as a path-dependent process (Boschma & Martin, 2007). More recently, EEG studies have moved beyond this endogenous process of related diversification and acknowledged that, in some cases, regions may diversify into new industries that are unrelated to regional preexisting industrial profile (MacKinnon et al., 2019). One key enabling factor of unrelated diversification is extraregional linkages, which may provide complementary assets and resources, such as knowledge and investments that are not available locally (Binz et al., 2016;Trippl et al., 2018). Such linkages could unfold in the form of trade flows, labour flows, investment flows and joint research and development (R&D) projects (Binz et al., 2016;Trippl et al., 2018;Zhu et al., 2017).
While plenty of empirical studies have examined how those different forms of extra-regional linkages facilitate long distance knowledge diffusion and resource exchange, less attention has been paid to the role of some temporary extra-regional linkages formulated by trade fairs. Extraregional linkages such as investment flows and joint R&D projects are often more long-lasting and stable, and demand a large amount of resources (Schoenberger 1997, Gertler, 2001Coe & Bunnell, 2003, Morgan, 2004. Hence, it is difficult to forge such linkages (Martin et al., 2018). In contrast, extra-regional linkages formed in the process of trade fairs seem to be more unstable and temporary. This idea of temporary extra-regional linkages echoes with the strand of literature that sees trade fairs as 'temporary professional gatherings' or 'temporary clusters' (Maskell et al., 2006;Torre, 2008;Comunian, 2017;Rallet & Torre, 2009;Power & Jansson, 2008).
Since the 1950s, international trade fairs (ITFs) have developed rapidly. Countries all over the world, such as France, Germany, Singapore and China, have all hosted several ITFs every year, and private firms have become the main organizers of ITFs (Maskell et al., 2006). Suppliers attend such ITFs to present their new and most advanced products as well as to examine their competitors' products, while buyers hope to find suppliers that can meet their needs. Firms spend considerable time and money in those events, to identify potential partners and obtain the latest information on markets and technologies. In this sense, ITFs serve as a forum for knowledge diffusion and information exchange, which may further trigger new discussions and ideas (Maskell et al., 2006). They are also a crucial platform for firms to meet the right people and forge new partnership. Since most ITFs happen on a regular basis, often annually, firms are also able to enhance their relations with existing partners via recurring faceto-face interactions during ITFs (Power & Jansson, 2008). This is important for the maintenance of longterm and stable cooperation between partners, especially those that are geographically far away from each other.
Although the impact of ITFs on exports has been explored by the literature on international trade (Pfeiffer et al., 1998), its role as some form of extra-regional linkages as well as how they facilitate regional export diversification does not draw enough scholarly attention in the EEG literature. This paper seeks to fill this gap and investigates whether ITFs can act as extra-regional linkages and enable regions to achieve related/unrelated diversification. In doing so, we also speak with the current literature on the role of some more long-lasting, stable extra-regional linkages (i.e., investment flows) in regional export diversification (Binz et al., 2016;Trippl et al., 2018;Zhu et al., 2017), but move beyond by bringing temporary extraregional linkages into the forefront. Specifically, we focus on one of China's most important and representative trade fairs: China Import and Export Fair (CIEF) (or Canton Fair), which has been held in spring and autumn every year since 1957 in Guangdong province, and analyse the possible relationship between temporary extra-regional linkages forged by ITFs and regional export diversification. This is our main contribution. Furthermore, we differentiate two types of export diversification: product diversification capturing regions' diversification into new products and market diversification that occurs when regions enter new export markets. Empirical results show that temporary extra-regional linkages have distinct effects on those two. This is our second contribution.
The rest of this paper is organized as follows. The next section reviews the literature on (un)related diversification and temporary extra-regional linkages. Section 3 introduces our data and research design. Section 4 provides some background information on China's trade fairs as well as its export dynamics. Section 5 presents empirical results. The last section concludes the paper.

(UN)RELATED DIVERSIFICATION AND
TEMPORARY EXTRA-REGIONAL LINKAGES 2.1. Product and market relatedness New industries do not start from scratch in a region; instead, regions rely on their pre-existing capabilities to develop new industries (Elekes et al., 2019). This idea of path-dependent regional industrial diversification has been first testified at the country level. Hidalgo et al. (2007) have confirmed that countries often develop a comparative advantage in new export products that are related to the country's pre-existing export baskets. Likewise, Neffke et al. (2011) have examined industrial diversification at the regional level and found that a new industry is more likely to emerge in a region if it is technologically related to regional pre-existing industries. Relatedness between industries/products has been seen as a key explanatory variable of regional industrial dynamics, resulting in a process of related diversification or 'regional branching' (Frenken & Boschma, 2007).
Such a branching process also exists in regional export dynamics, since regions are more likely to export new products that are related to pre-existing regional export profiles (Boschma, 2017). Regions often develop new export products based on their pre-existing capabilities (Aw et al., 2007;Ursic et al., 1984). It is much easier to draw resources and knowledge from related export products while developing new products, since related products demand similar manufacturing skills and exporting knowledge. On the supply side, related products often contain similar technologies and skills, and thus capabilities a region has manufacturing certain products can be used in the development of some new related products (Content & Frenken, 2016). On the demand side, export markets of related products are likely to have similar features in terms of distribution channels, consumer tastes, and laws and regulations (Roberts & Tybout, 1999;Albornoz et al., 2016). Hence, exporting experience and distribution networks of pre-existing products can be easily adjusted and re-used to export related new products. In short, it is less costly and risky for exporters to diversify into related new products, than into random new products.
In addition to product relatedness, another key explanatory factor of export diversification is market relatedness. Exporters often search for new export markets in two ways. First, the standard gravity model has pointed out that exporters tend to enter new markets that are geographically close and culturally similar to the country of origin (Eaton & Kortum, 2002;Helpman et al. 2008;Head & Mayer, 2014). However, some studies have found that the observed spatial correlation is much greater than what the standard gravity model alone could predict, and more attention has then been directed towards the second way of market diversification that is based on the extended gravity model (Morales et al., 2011(Morales et al., , 2019. This model argues that exporters can search for new export markets from their existing export market networks (Chaney, 2014). While the gravity model stresses relatedness between home market and new export markets, the extended gravity model emphasizes relatedness between exporters' existing export markets and new markets.
Some terms, such as 'sequential exporting' and 'remote search', have been developed to describe the extended gravity model and reveal how an exporter's existing export markets affect which new markets it will enter subsequently (Albornoz et al., 2012;Defever et al., 2015;Chaney, 2014). Entering new export markets means that exporters may need to modify their products to customize them to consumer preferences, laws and regulations in new markets (Morales et al., 2011(Morales et al., , 2019. Some other costs incurred in this process include time and resources spent in searching for new employees and local distributors. Temporary extra-regional linkages and export product and market diversification Chaney (2014) has referred to those adaptation costs as 'informational barriers', and shown that exporters are likely to enter new markets that are similar to their existing export markets since they can use their knowledge on existing export markets to overcome such barriers. When markets are different but correlated, exporters tend to enter new markets that are similar to their existing ones to reduce risks and adaptation costs (Albornoz et al., 2012). Given the idea of market relatedness, the expansion of exporters' markets is likely to unfold in a path-dependent way (Guo, Zhu, & Boschma, 2020). These arguments lead to the following hypothesis: Hypothesis 1: Regions are more likely to enter new products/ markets that are related to their pre-existing products/markets.

ITFs as temporary extra-regional linkages
More recent EEG studies have acknowledged that, regional economic development is not completely an endogenous process of related diversification dependent on what capabilities are available in this region beforehand (Hidalgo et al., 2007;Neffke et al., 2011;Boschma et al., 2012); it is also shaped by another process of unrelated diversification, that is, how regions develop new industries that are unrelated to regional pre-existing industrial structures (Boschma, 2017;Boschma & Capone, 2015;Pinheiro et al., 2022). Unrelated diversification is often seen in open regional economies with extra-regional linkages, which facilitate interregional knowledge transfer and contribute to the renewal of local capabilities (Hassink, 2005). This also echoes with the idea of 'global pipelines' for knowledge diffusion across regional borders proposed by Bathelt et al. (2004). The idea is that regions need to maintain diverse overlaps between actors inside and outside the region (Sydow et al., 2010), and forge various kinds of connections with actors beyond the region (Bathelt et al., 2004;Boschma & Iammarino, 2009). Such extraregional linkages enable regions to move beyond path dependence, and achieve disruptive changes and unrelated diversification (Bathelt et al., 2004;Sydow et al., 2010;Bathelt & Li, 2014).
Empirical studies have subsequently examined various types of extra-regional linkages. Foreign direct investments (FDIs) can support interregional knowledge exchange in many ways, such as the formulation of forward and backward linkages between regions, competition and demonstration effects among regions, the possibility for domestic firms to recruit skilled employees released from foreign firms, and knowledge spillovers between foreign and domestic firms (Poncet & Starosta de Waldemar, 2013;Zhu & Fu, 2013). Some empirical studies have confirmed that FDIs enable regions to make long jumps to new products that are unrelated to regional pre-existing industrial profiles (Lo Turco & Maggioni, 2016;Zhu et al., 2017). Furthermore, non-local firms can also breathe new air into a region and trigger unrelated diversification, given their connections with the outside (Neffke et al., 2018). Elekes et al. (2019) have also found that foreign firms act as the key driver of unrelated diversification and structural change in Hungary.
While a large body of literature has shown the importance of extra-regional linkages such as FDIs and foreign firms, less is known about if long distance knowledge transfer can be also facilitated by some temporary extraregional linkages, such as those formulated by trade fairs. The former are often more long-lasting, but requires plenty of time and resources spent in searching for proper locations and needed workers, bargaining with local partners and governments, and investing in fixed assets (Morisset, 2000;Basu & Srinivasan, 2002). In contrast, it is much easier to formulate temporary extra-regional linkages via ITFs. Recent studies have emphasized that ITFs are not merely a marketing event where buyers and suppliers meet with each other, but rather an important platform that supports knowledge spillovers and information exchange over long distance (Borghini et al., 2004(Borghini et al., , 2006Maskell et al., 2006). Bathelt and Zeng (2014) have pointed out that ITFs can stimulate vertical and horizontal interactions. Vertical interactions between buyers and suppliers take place when those two groups communicate with each other about recent changes in trends, technologies and customer preferences as well as ideas for future products. Furthermore, by attending ITFs, firms not only reinforce their connections with existing partners, but also seek to identify new partners and establish new connections (Backhaus 1992;Meffert, 1993). Vertical interactions are one key important source for suppliers to improve their products. Horizontal interactions bring together competitors that do not often meet with each other. At ITFs, firms have the chance to know more about their competitors' products and strategies, and then modify their own accordingly (Borghini et al., 2006). This process of learning by observing during the ITFs allow firms to see the most advanced product designs and technologies, and to compare their own products with others (Sharland & Balogh, 1996;Borghini et al., 2006). In short, ITFs as temporary extra-regional linkages can bring knowledge on new products and technologies as well as information about potential partners in new markets, which are necessary for exporters to develop new products and penetrate into new markets.
However, the effects of temporary extra-regional linkages on product and market diversification are not expected to be the same. Marketing products is exporters' main motives to participate in ITFs (Shereni et al., 2021). Potential benefits include the introduction of products to a large number of buyers, retention of existing partners, identification of new buyers and reliable partners, and observing new trends (Nayak, 2019). The main goal of exporters attending ITFs is thus to increase their sales by establishing connections with new partners, especially those in new markets (Lilien & Grewal, 2022). Bonama (1983) has pointed out two types of benefits firms can gain by participating in ITFs: selling and non-selling benefits. Selling benefits are defined as benefits stemming from selling products to new partners, while non-selling benefits contain monitoring competitors' strategies and exhibiting new products (Wang et al., 2017). Nonetheless, all activities during ITFs are expected to translate into increasing sales, especially by establishing new market linkages and penetrating into new markets (Seringhaus & Rosson, 1998).
One other goal of participating in ITFs is to know recent trends and competitors' products (Borghini et al., 2006). However, it is still difficult for firms to quickly absorb new technologies or develop new products in response to their competitors' new products, since knowledge contained in new products and technologies, especially high-tech ones, cannot be easily comprehended during the short period of ITFs (Gann & Salter, 2000). Furthermore, the development of new products demands a new set of assets and competences (e.g., skilled workers and infrastructure), which may be unavailable in the locality. In contrast, it is much easier for exporters to find new partners in new markets at ITFs, and then tap into new markets they are not familiar with beforehand. Hence, ITFs provide firms an opportunity to establish connections with partners far away and then enable the latter to enter new markets that are unrelated to their existing export markets, whereas their effect on firms' diversification into new unrelated products is expected to be limited.
Furthermore, buyers participating in ITFs search for suppliers in two ways. First, they identify new suppliers who can offer the products they need and discuss potential contracts (Backhaus 1992;Meffert, 1993). Second and more frequently, they rely on their existing suppliers to reduce risks and uncertainties (Ford, 1980;Handfield & Bechtel, 2002). It is argued that most firms participating in ITFs seek to strengthen connections with existing partners via face-to-face communication and some informal dinner meetings during ITFs (Maskell et al., 2006). In both cases, buyers tend to go directly to suppliers with experiences and capabilities of exporting the products they want or at least some related products. They may also require suppliers to make some modifications and improvements. In this sense, suppliers are sometimes pushed to implement some related product diversification after meeting with buyers during ITFs. However, buyers rarely ask their suppliers to diversify into some new unrelated products, which demand capabilities suppliers do not possess. Since unrelated product diversification takes plenty of resources and is extremely risky, a wise strategy would be to search for new suppliers that can provide the target products immediately, rather than collaborate with incapable suppliers and wait for their unrelated product diversification. Based on those arguments, we hypothesize that, Hypothesis 2a: Regions with many exporters participating in ITFs are more likely to enter new markets that are unrelated to their existing export markets.
Hypothesis 2b: Regions with many exporters participating in ITFs are more likely to develop new products that are related to their existing products.

Data
We use China's export data during 2003-2016 compiled by the Chinese Customs Trade Statistics (CCTS). The dataset reports all merchandise transactions passing through Chinese customs and includes firm information (e.g., firm name, address and ownership), eight-digit Harmonized System (HS) code, export/import value and quantity, destination and origin of exports/imports. We exclude trading firms and intermediaries that do not manufacture products.
This paper focuses on China's most important and representative trade fairthe CIEF, to examine the effect of ITFs on export product and market diversification. The CIEF was held in spring and autumn every year since 1957 in Guangzhou. During the last few decades, it has developed dramatically with increasing number of purchasers and rising trading volume (see Table A1 in Appendix A in the supplemental data online). We use data on firms participating in the CIEF during 2012-14 compiled by the China Foreign Trade Center. This database contains every exhibitor's name, main products, booths and some other basic information. We match those products to four-digit HS code.

Product and market relatedness
We calculate product relatedness using the co-occurrence approach proposed by Hidalgo et al. (2007). The product relatedness indicator between product i and j is defined as the minimum of the pairwise conditional probabilities of a region exporting a product given that it also exports the other.
where exp c,i,t is the export value of product i of region c in year t. Region c is seen as having a revealed comparative advantage (RCA) in product i if RCA c,i,t is above 1. The rationale behind the product relatedness indicator is that if two products are often co-exported by same regions, they are likely to demand similar input factors, infrastructure and technologies. Those two products can thus be seen as being related. We then calculate a product density indicator defined as the average relatedness between a product and the regional pre-existing productive structure. This density indicator developed by Hidalgo et al. (2007) captures the extent to which product i is related to the export structure Temporary extra-regional linkages and export product and market diversification of region c: where C j,c,t takes the value of 1 if region c has an RCA in product j in year t, and 0 otherwise. If product i is related to many products in which region c already has an RCA, the relatedness between product i and region c's export productive structure is strong, and the product density indicator is high. Similarly, we use the co-occurrence approach to calculate the market relatedness indicator between export market m and n, which is measured as the minimum between the conditional probability of exporting to market m given region c exporting to market n, and the conditional probability of exporting to market n given region c exporting to market m.
where exp c,m,t denotes the export value of region c to market m in year t. If RCA c,m,t is above 1, region c is defined as having an RCA in market m in year t. The idea is that if regions often export to both markets simultaneously, those two markets are likely to be similar, in terms of customer preferences, institutions and cultures. We then calculate the market density indicator defined as the average relatedness between a market and regional pre-existing export market networks.
Mdensity m,c,t = n C n,c,t × ∅ m,n,t n ∅ m,n,t .
where C n,c,t takes the value of 1 if region c has an RCA in market n in year t, and 0 otherwise. If market m is related to many markets to which region c has already exported, the relatedness between market m and region c's export market networks is strong, and the market density indicator should be high.

Model specifications
The CIEF is held in Guangzhou in Guangdong province. Firms far away from Guangzhou are less likely to participate in the CIEF due to the long geographical distances and high levels of transportation costs (Figure 3). To avoid such a selection bias, we only choose firms in 21 prefecture-level cities in Guangdong province, that is, the province where the host city is. Our sample includes 239 countries as export markets and 1219 four-digit HS products. To examine the effect of product/market relatedness on a region's diversification into new product-market combinations (Hypothesis 1), we estimate the following equation: Entry c,i,m,t+2 = a + b 1 Pdensity c,i,t + b 2 Mdensity c,m,t + b 3 Control c,i,m,t + m c,i,m,t + 1 c,im,t (7) The dependent variable Entry c,i,m,t+2 is a dummy, taking the value of 1 if city c's export value in product i and market m is 0 in year t but above 0 in year t + 2, and 0 otherwise. This indicator captures if city c's export of product i to market m is a new product-market combination. The Probit model is used. We also include product dummies at the four-digit level, city dummies, and year dummies. We divide the 239 countries in 15 big areas (see Figure A1 in Appendix A in the supplemental data online) and include area dummies in our model. We then investigate if firms participating in ITFs affects a region's export product/market diversification (Hypothesis 2), by estimating the following equation: where Fair c,i,t is the number of firms participating in the CIEF that export product i in year t in city c. We also include the interaction between Fair and Pdensity/Mdensity. A positive and significant coefficient of the interaction term indicates that firms participating in the CIEF enhances regions' reliance on product/market relatedness, while a negative and positive one means a weaker effect of product/market relatedness. In the former case, regions rely more on product/market relatedness, and path-dependent related diversification becomes more dominant. In contrast, in the latter case, firms participating in the CIEF enables regions to diversify into less related products/markets, leading to more path-breaking unrelated diversification. Furthermore, some control variables are also included. We control the gross domestic product (GDP) per capita of city c (PGDP), GDP per capita of market m (CGDP), and the distance between China and the largest city in market m (Dis), since these are the factors stressed by the standard gravity model. The instability of diplomatic relations between China and other countries also affects China's exports to the latter. We calculate the difference between the scores of China and other countries on six World Governance Indicators (WGI) to construct variable Ins. The larger Ins is, the more unstable the diplomatic relationship is. We also control city c's fiscal expenditure on science and technology (R&D) and education (Edu), since innovation and human capital both affect the city's export dynamics. The definitions and data sources of all variables are reported in Table A2 in Appendix A in the supplemental data online.

CIEF AND CHINA'S EXPORT DYNAMICS
In the past decade, China's exhibition activities have developed consistently. The number of exhibitions and their area have both expanded dramatically, leading to increasing amount of export value (Figure 1). Among them, the CIEF is the most important one with the longest history, the largest scale, the most diverse products, and the largest number of buyers from a variety of regions inside and outside China. Despite considerable changes in China's economic development and international trade during the last few decades, the CIEF is still seen as one of the most crucial events for Chinese firms and global buyers to meet each other (Bathelt & Zeng, 2015;Jin & Weber, 2013). It is not only a marketing event where firms search for potential partners and meet with their existing partners, but also a temporary space for learning and knowledge diffusion between actors that are far away from each other . Until 2018, the cumulative export value generated by the CIEF has been around US$132.7 billion, with approximately 8.42 million global buyers. Recently, the average exhibition area of every CIEF is 1.19 billion m 2 , with nearly 25,000 exhibitors and approximately 200,000 global buyers from more than 210 countries/regions ( Figure 2). In 2018, there were 59,647 booths and 24,947 exhibitors at the CIEF, covering more than 90% of 1219 four-digit HS products. Figure 3 shows the share of exhibitors from different provinces of China. We can see that the CIEF is so influential that firms from all over China have come to participate. However, the majority of exhibitors come from coastal China, and  Temporary extra-regional linkages and export product and market diversification the number of exhibitors decreases from Southeast China to North and West China.
Meanwhile, China's exports have also developed rapidly in this period. Figure 4 shows that during 2003-16, the total number of exporters in China has increased from 90,996 to 267,987. There are around 20,000-30,000 new exporters emerging in China every year. Though the entry rate of new exporters decreases in the 2010s, the development of China's exports has already been remarkable. In this process, Chinese exporters have diversified into many new products and markets. For instance, Guo et al. (2022) found that since 2006, Chinese firms have established more than one million new firmproduct-market export combinations every year. It is thus necessary to examine China's export product and market diversification in this period.   Figure 5 shows the kernel density distribution of product and market relatedness. The kernel distribution of product relatedness is left skewed, indicating that only a small number of products are closely related. More than 95% of product relatedness indicators are below 0.2, which is in line with existing studies (Hidalgo et al., 2007;Boschma et al., 2011). Similarly, the distribution of market relatedness is also left skewed. This illustrates that most export markets are weakly related. Given this, China's quick development is likely to be characterized by both related and unrelated diversification.
To identify the effect of the CIEF on China's export dynamics, we compare the export value of products that are exhibited at the CIEF and export value of products that are not exhibited at the CIEF in our study area, that is, Guangdong province. Figure 6 shows that the export values of products exhibited at the CIEF accounted for more than 76% of the total during 2014-16. The CIEF is expected to play a critical role in China's export dynamics. Table A3 in Appendix A in the supplemental data online suggests that there is no serious issue of multi-collinearity. To alleviate the heteroscedasticity, we adopt the logarithm of continuous variables. We first estimate equation (7) to examine the effect of product/market relatedness on a region's diversification into new product-market combinations. Empirical results are reported in Table 1. The coefficients of Pdensity and Mdenisty are both positive and significant. This confirms that regions are more likely to diversify into new products/markets that are related to their pre-existing products/markets. Hence, export product and market diversification is a path-dependent process, which is consistent with some existing studies on China's export dynamics (He et al., 2018). This also confirms Hypothesis 1. Furthermore, the coefficient of Pdensity is significantly smaller than that of Mdensity, indicating that export market diversification is much more path- Figure 5. Kernel density distribution of (a) product and (b) market relatedness. Figure 6. Export values of products that are exhibited and not exhibited at the China Import and Export Fair (CIEF).

Correlation analysis in
Temporary extra-regional linkages and export product and market diversification dependent than product diversification. It is much easier to use knowledge and experience in one market to tap into a new market, than to use knowledge and resources in one product to develop a new product. The development of new products, especially high-tech ones, demands new capabilities and assets, which may not be available in the region. Our findings echo with Fafchamps et al. (2007) who have also stressed that the development of exports is associated more with market learning that depends on export experience in other markets, than with product learning that depends on experience in manufacturing products.
The parameter of CGDP is positive and significant, indicating that exporters are more likely to target at developed markets with greater market potential. This is consistent with the standard gravity model (Eaton & Kortum, 2002;Helpman et al., 2008). The parameter of PGDP is negative and significant, suggesting that developed regions are less likely to enter new products and markets. Developed regions have already occupied many products and markets, and there are not many new markets/products left for them to enter. Furthermore, exporters are less likely to enter new markets that do not have stable diplomatic relationships with China. Investments in R&D and education are expected to boost regional innovative capacity and human capital, and are thus a key driving force of export diversification. Counterintuitively, the parameter of Dis is positive and significant, which is in discordance with the standard gravity model. One possible reason is that cities in Guangdong have a long history of exporting, and now their new target markets are mostly far away. Table 1. Empirical results of export product/market diversification.
(1) (2) (3) (4)  To investigate if firms participating in ITFs affects a region's export product/market diversification (Hypothesis 2), we include Fair and its interaction terms with Pdensity and Mdensity in our models. Econometric results are reported in Table 2. The coefficient of Fair is positive and significant in all models, implying that attending ITFs has enabled exporters to diversify into new products/markets. Existing studies have seen ITFs as an important platform that supports knowledge spillovers and information exchange between buyers and suppliers as well as between competitors (Borghini et al., 2004(Borghini et al., , 2006Maskell et al., 2006;Bathelt & Zeng, 2014). Our results show that ITFs as a form of temporary extra-regional linkages bring external knowledge and information into regions, and support interregional learning by providing face-to-face communication opportunities (Boschma, 2005;Power & Jansson, 2008;Bathelt et al., 2017). For instance, according to the report of China Economic Weekly, most of the interviewed firms acknowledged that ITFs are an important channel for them to know their competitors' strategies and products, and to build up connections to new markets (China Economic Weekly, 2018). Similarly, based on a survey with small and medium-sized firms in the UK, Westhead et al. (2002) have also found that the main reason that those firms entered the international market is orders made by global buyers, and hence ITFs represent a crucial platform to get in touch with the latter.
More importantly, the coefficient of Fair × Pdensity is significant and positive, while that of Fair × Mdensity is significant and negative. This confirms Hypothesis 2. It implies that exporters participating in ITFs are more likely to enter new related products, while participating in ITFs enable them to enter new unrelated markets. As shown in Figure 7, firms participating in the CIEF enhances the effect of product relatedness on export product diversification. Regions rely more on product relatedness, and pathdependent related product diversification becomes more dominant. In contrast, Figure 7 shows that firms participating in the CIEF weakens the effect of market relatedness. In this case, regions rely less on market relatedness and diversify into less related markets, resulting in more path-breaking unrelated market diversification.
On the one hand, it is often difficult for firms to fully understand and absorb new technologies in their competitors' products, especially high-tech ones, even if they see those products at ITFs. It is even more difficult for firms to develop new products based on their observation during the short period of ITFs (Gann & Salter, 2000), which often demands new capabilities and assets that may be unavailable in the locality. Hence, horizontal interactions during ITFs do not necessarily lead to unrelated product diversification. Furthermore, buyers participating in ITFs often go directly to suppliers that can supply the products they need (Schuldt & Bathelt, 2011). Sometimes, buyers may ask their suppliers to make some modifications and to develop new related products. However, they rarely push suppliers to supply a product that is unrelated to the latter's existing export profiles, since such unrelated product diversification is costly and risky. A better idea is to find new suppliers that can provide this new product. In short, vertical interactions during ITFs are likely to result in related product diversification. On the other hand, it is much easier for exporters to meet some buyers in new markets during ITFs, since one main goal of ITFs is to bring faraway firms together (Lilien & Grewal, 2022). Exporters attending ITFs may thus establish connections with buyers in new markets that are unrelated to their existing export market networks.
In Table 3, we seek to study if the effects of ITFs on export product/market diversification varies across space. Although Guangdong is one of China's most developed provinces in China, there still exists a considerable regional disparity in economic development. For instance, Shenzhen's GDP was as high as CNY1600 billion in 2015, while that of Meizhou was only CNY88.5 billion. We divide cities in Guangdong province into four quartiles according to per capita GDP, and re-run our estimations for each quartile separately. The first quartile includes the least developed cities, while the fourth quartile contains cities with the highest levels of per capita Temporary extra-regional linkages and export product and market diversification GDP. Empirical results for four quartiles are reported in models 1-4 in Table 3. First, the coefficient of Fair × Pdensity in model 1 is significantly greater than that in other three models, implying that exporters attending the ITFs tend to focus more on related diversification in underdeveloped regions. In those regions, exporters are mostly small and medium-sized, specialized suppliers and they have to follow large global buyers' orders. They are more likely to offer pre-existing products, and develop some new related products with minor modifications and improvements.
Furthermore, the coefficient of Fair × Mdensity in models 2 and 3 is significantly greater than that in other models. It implies that participating in ITFs is less likely to trigger unrelated market diversification in the most and least developed regions (models 1 and 4). ITFs provide marketing opportunities and rich market information for participating enterprises. Through ITFs, firms can find partners in new markets to tap into new markets (Menon & Edward, 2014). However, firms in developed regions may have already entered the majority of export markets and thus have fewer new markets to enter. They may also have other channels to access market information and establish connections with global buyers (Wojnicka-Sycz, 2018;Zhou et al., 2019). Hence, the effect of ITFs on export market diversification is negligible in the most developed regions. On the other hand, firms in the least developed regions often lack the learning and absorptive capacity to comprehend market information at ITFs, and may be incapable of winning the trust of new partners (Banerjee & Chau, 2004;Hennart, 2005). The effect of ITFs is thus less evident in model 1.
As a robustness check, we re-estimate our models using different definitions of the dependent variable (Entry) and Fair. First, a firm's exports of a product to a market may fluctuate, resulting in some unstable export market/product diversification. For instance, many exporters adopt a 'trial-and-error' strategy, by investing a small amount of resources in a new product/market, and then decide whether to diversify into this new product/market based on its export performance (Albornoz et al., 2012). Hence, we adopt a stricter definition of Entry, to avoid the influence of export fluctuation. Specifically, if the export value of product i in city c to market m in year t is 0 but above 0 in year t + 1 and t + 2, Entry is defined as 1, and 0 otherwise. Second, we define Fair as a dummy, which takes the value of 1 if there are firms that export product i in city c attending the CIEF in year t and otherwise 0. Empirical results are reported in Table  A4 in Appendix A in the supplemental data online. Our main findings remain.

CONCLUSIONS
The EEG literature has stressed relatedness between products as a key explanatory factor in regional export dynamics, and argued that it not only advances the growth of existing products through agglomeration externalities but also contributes to the emergence of new related products (Hidalgo et al., 2007;Neffke et al., 2011;Boschma and Capone, 2015). More recent studies have, however, acknowledged that regional development is not completely an endogenous process of related diversification, and some extra-regional linkages may enable regions to deviate from existing growth paths and diversify into new unrelated products (Bathelt et al., 2004;Bahar et al., 2014). While many have focused on the role of some long-lasting extra-regional linkages such as FDIs, less attention has been directed towards the importance of temporary extra-regional linkages such as ITFs. In this paper, we seek to fill in this gap and show that ITFs as temporary extraregional linkages play a decisive role in regional export dynamics. This is our first contribution. Although ITFs have been seen as 'temporary professional gatherings' or 'temporary clusters' (Maskell et al., 2006;Torre, 2008;Torre, 2008;Comunian, 2017), its role as temporary extraregional linkages as well as how it facilitates export diversification remains understudied. In this paper, we differentiate two types of export diversification: export product and market diversification, and study the distinct effects of ITFs on them. This is our second contribution. Empirical results show that exporters participating in ITFs are more likely to follow a development trajectory of related product diversification, while participating in ITFs enables firms to diversify into new markets that are unrelated to their pre-existing export market networks. Furthermore, participating in ITFs is less likely to trigger unrelated market diversification in the most developed regions, which have already occupied many markets, and the least developed regions, which often lack the capabilities to implement unrelated market diversification.
One main limitation of our paper is its focus on one main trade fair and exporters in one province, due to the unavailability of data on other trade fairs. Future research with better data could examine the effect of temporary extra-regional linkages on export product/market at the national level as well as in other countries, to testify the generality of our findings. Furthermore, our paper hopes to show that countries should invest in trade fairs to formulate temporary extra-regional linkages connecting domestic firms with global buyers. Interregional knowledge spillovers facilitated by temporary extra-regional linkages may bring new information and knowledge about remote export markets that are unrelated to regional preexisting export market networks, enabling regions to diversify into new export markets they are unable to enter on their own. Policy-makers should be also aware of the potential effect of trade fairs on export product diversification. Unlike some long-lasting extra-regional linkages such as FDIs that can trigger path-breaking unrelated product diversification, temporary extra-regional linkages tend to reinforce path-dependent related diversification, and do not allow regions to jump further into high-end, new products.

DISCLOSURE STATEMENT
No potential conflict of interest was reported by the authors.