Bio-based and Applied Economics 8(3): 261-277, 2019 ISSN 2280-6180 (print) © Firenze University Press ISSN 2280-6172 (online) www.fupress.com/bae Full Research Article DOI: 10.13128/bae-9447 Export propensity and intensity in the wine industry: a fractional econometric approach Anthony MAcedo*, João Rebelo, SofiA GouveiA Department of Economics, Sociology and Management (DESG), Centre for Transdisciplinary Development Studies (CETRAD), University of Trás-os-Montes and Alto Douro (UTAD), Quinta de Prados, 5001-801 Vila Real, Portugal Abstract. Using export market shares as a measure of international competitiveness, this paper studies wine exports in terms of propensity and intensity. Based on data for the period 1999-2014, a two-part fractional regression model is applied. The results suggest that for importing countries GDP per capita, their own wine production, and EU membership have a positive effect on the probability of importing wine but tend to evolve inversely to market shares, as taste for variety becomes more important. Additionally, export propensity is positively affected by regional trade agreements, common language, similarity of religious culture, wine production in the exporting country, and the exporting country being from the Old World, while export intensity is boosted by common language and wine production in the exporting country. Bilat- eral distance has a negative effect on both margins of trade. Keywords. globalisation, international trade, market share, fractional regression model. JEL codes. F14, L66. 1. Introduction International competitiveness has become a topic of growing interest over the last dec- ades. The concept of competitiveness is multifaceted and multidimensional (De Grauwe, 2010), and researchers with such different backgrounds as economics, politics, management and history have all studied this concept. In economic literature, the roots lie in the interna- tional economic theories of Adam Smith, David Ricardo and their followers. A definition of competitiveness, as given by the European Commission is the ability “to sustainably produce and sell goods and services on a given market, in such a way that buyers prefer these goods to those offered by competitors” (European Commission, p. 12, 2014). On the other hand, the Organisation for Economic Co-operation and Development suggests it is the “ability of companies, industries, regions, nations, and supranational regions to generate, while being and remaining exposed to international competition, relatively high factor income and factor employment levels on a sustainable basis” (Hatzichronologou, p.20, 1996). *Corresponding author: anthonym@utad.pt 262 Anthony Macedo, João Rebelo, Sofia Gouveia These definitions suggest that competitiveness can be observed from different points, namely at a company, sector, or national level. While ratios of financial profitability are a predominant indicator of performance for companies and the sectors of which they are com- posed, a high-performing country is one with positive trade balance (Cardebat, 2019). This means that, although there is no “national decision” regarding trade in the sense of a central- ized decision, the “national decision” is the sum of the decisions made by its companies. The well-known export market share is a simple but informative measure of competi- tiveness that allows direct trading positions to be established and was used, for example, by Banterle and Carraresi (2007), Wijnands et al. (2008), and Carraresi and Banterle (2015). Generally, this indicator is calculated by dividing the exports of a country by the total exports of a trading area and it is a measure of the degree of importance of a country, henceforth called country i, within the total exports of that trading area. The market share is given by a fractional variable bound by 0 and 1, being 0 if country i does not export and being 1 when all exports of the area are made by country i. In econometric modelling the fractional nature of this dependent variable provides limi- tations, in particular with linear specifications, where predicted values can be outside the boundaries [0, 1], resulting in meaningless outcomes. Therefore, the fractional regression model (FRM) developed by Papke and Wooldridge (1996, 2008) seems to be a preferable approach. Moreover, by considering a two-part model (2P-FRM) the aim is to estimate two different effects of explanatory variables: on the one hand, the effect on the decision of two countries establishing an international trade relationship (export propensity) and, on the other hand, the effect on the decision about how much to trade (export intensity). Following a similar logic, Bouët et al. (2017) estimated Cognac export propensity and intensity using the Heckman’s procedure to correct a sample selection issue. Comparatively, the 2P-FRM has the advantage of not requiring an exclusion restriction (Ramalho et al., 2011). The global intensification of international trade have been accompanied by a wave of research estimating macroeconomic determinants mainly through the gravity framework to explain trade flows in value or volume (Bergstrand 1989; McCallum 1995; Anderson and van Wincoop 2003; Silva and Tenreyro 2006; Anderson and Yotov 2012; Cirera et al. 2016). Inspired by Newton’s law of gravity, the gravity model sets out to explain trade between an ex- porting country and an importing country based on their economic masses and the distance between them. It follows that economic masses (generally represented through the GDP) are expected to positively influence trade, while bilateral distance is expected to have a negative effect. Over time, researchers have added more variables to the model and some works high- light the importance of studying the impact of trade agreements, transport costs, purchasing power, and cultural proximity, among other determinants, on international trade patterns.1 Wine is a good example of a product increasingly globalised during the last five decades, presenting diversified geography of production and consumption and fierce competition be- tween countries and even among local wine regions. Studies on the global alcohol markets, and more specifically wine, have been multiplying in the last decades. Anderson et al. (2018) presented historic facts about alcohol consumption and highlighted the study from Holmes and Anderson (2017) to state the more recent trend of convergence in national consumption 1 However, generally, such analyses do not take into consideration the contribution from the exporting countries to total imports in the destination country using, for example, market shares. 263Export propensity and intensity in the wine industry patterns of alcohol as a result of globalisation. Despite social, political, and fiscal differences among countries, consumption levels have decreased in “traditional” consuming countries (e.g. France, Italy, and Portugal) and increased in other countries without a tradition of wine consumption (e.g. China, Russia, and USA) (Smith and Mitry, 2007; Dal Bianco et al, 2013). Additionally, other topics were also studied such as the potential market implications of Brexit (Anderson and Wittwer, 2017), the emergence of Asia in the beverage market (Ander- son, 2019), and the impact of climate change on wine industry (Anderson, 2017; Ashenfelter and Storchmann, 2016a, 2016b). A complementary strand of the literature deals with micro- economic factors of competitiveness. For example, Bargain et al. (2018) use a qualitative ap- proach to discuss key comparative advantages of 16 wine-producing countries, and Ugaglia et al. (2019) dedicate several chapters to the study of the industrial organization of Old and New World wine producing countries. Based on the gravity model, but considering the fractional nature of the explained vari- able, which leads to the application of the FRM, this research aims to contribute to the de- bate and provide further insight into the dynamics of international trade. To the best of our knowledge, there are no published studies that use the FRM to analyse the competitiveness of an industry. Specifically, a 2P-FRM is applied to data on wine exports from the main fifteen producing countries to 193 partner countries, between 1999 and 2014, taking into account explanatory variables inspired by the gravity equation. After this introduction, the paper is organized as follows: Section 2 presents the model framework and the data used; Section 3 presents the results and discussion; and Section 4 concludes. 2. Material and methods Given that the variable of interest y, i.e. market share, is a proportion defined and observed only in the interval 0 ≤ y ≤ 1, an approach capable of dealing with a bounded and fractional re- sponse variable is required. Standard linear specifications can become inconsistent since they do not guarantee predicted values within the boundaries [0,1]. Following the seminal papers of Papke and Wooldridge (1996, 2008), the FRM is recommended since it presents advantages in relation to linear methods or other common solutions of the literature such as logit, probit, Tobit and Heckman sample selection model. Unlike the FRM, all these methods do not guar- antee predictions within the meaningful interval [0,1] (Ramalho et al., 2011). Comparing in particular with the Heckman model, the FRM presents also the advantage of not requiring an exclusion restriction, which is useful in empirical practice and, following Schwiebert (2018), avoids severe biases due to the imposition of an incorrect restriction. The FRM is a non-linear model that does not require transformations for values at the boundaries, accounting for the non-linearity in the data, while being fully robust under generalized linear model assump- tions (Gallani et al. 2015). Observations at the extremes of the distribution are included based on the assumption that E(Y|X) = G(Xβ), where fitted values are guaranteed within the unit interval by 0 ≤ G(∙) ≤ 1. Papke and Wooldridge (1996) suggest the quasi-maximum likelihood (QML) method for estimations of β, which is based on the maximization of the Bernoulli log- likelihood function LL(β) ≡ y log[G(Xβ)] + (1 - y) log[1 - G(Xβ)]. Furthermore, if there is a high concentration of observations at the boundary 0, Ra- malho et al. (2011) advise that it is better to consider a two-part model. The 2P-FRM is 264 Anthony Macedo, João Rebelo, Sofia Gouveia constituted of a binary model for the discrete component (0 or 1) and a fractional model for the continuous component. Choosing between one- or two-part models depends on the in- terpretation of the zeros based on the existence or not of two decision mechanisms for zeros and positive values [Ramalho et al. (2011) also suggest a P test to compare the two models]. In international trade, by the estimation of a two-part model, it is presumed that a country as representing all of its firms has two distinct decisions to make: the first is whether to establish a trade relationship with another country (export propensity) and the second concerns the amount to be traded (export intensity). If none of the companies from a country i decides to export to a country j (e.g., because costs are too high for potential profit), then the decision of country i is to not export to country j. Therefore, in the 2P-FRM of this work, the first part estimates the factors influencing the probability of the wine of country i being imported by a certain country j and it can be defined as 0 for shareijt = 0 shareijt* = { (1) 1 for shareijtε(0,1) Pr(shareijt* = 1|Xijt) = E(shareijt*|Xijt) = F(Xijt β1) (2) where shareijt is the market share of country i’s exports as a proportion of the total of country j’s imports in year t, F(∙) is a non-linear conditional mean specification and β1 is a vector of coefficients for the covariates in Xijt. The second part of the model considers only the observations of (1) where country i’s wine was imported, i.e. positive outcomes, to estimate the factors influencing the magnitude of a market share of country i in country j. It can be represented as E[shareijt|Xijt, shareijt ∈(0,1)] = M(Xijt β2) (3) where in M(Xijt β2) the regressors Xijt are the same as the first part (despite not being re- quired) and M(∙) is also a non-linear conditional mean specification but not necessarily the same specification as F(∙). Hence, following Ramalho et al. (2011), E[shareijt|Xijt] can be defined by E[shareijt|Xijt] = M(Xijt β2) ∙ F(Xijt β1) (4) But to correctly interpret the estimated coefficients of non-linear econometric models it is advisable to estimate average and total partial effects. As xijt is a covariate of vector Xijt, an average partial effect (APE) will be computed for the first part of the model to estimate the effect of xijt on the probability of a good (in this case wine) of country i being imported by a certain country j: ∂ Pr(shareijt* = 1|Xijt) = β1f(xijt β1) (5) ∂xijt 265Export propensity and intensity in the wine industry Similarly, another APE will be computed for the second part to estimate the effect of xijt on the magnitude of a non-zero market share of country i in country j: ∂E[shareijt|Xijt, shareijt ∈(0,1)] = β2m(xijt β2) (6) ∂xijt Finally, a total partial effect (TPE) will be computed to estimate the effect of xijt on the magnitude of any (including zero) market share of country i in country j: ∂E[shareijt|Xijt] ∂M(Xijt β2) ∂F(Xijt β1) = F(Xijt β1) + = M(Xijt β2) (6) ∂xijt ∂xijt ∂xijt Inspired by the literature on the gravity model (Bergstrand 1989; McCallum 1995; Ander- son and van Wincoop 2003; Silva and Tenreyro 2006; Anderson and Yotov 2012; Cirera et al. 2016), the covariates that could influence the wine market share are: geographic distance be- tween countries i and j (distij); GDP per capita of importer j in year t (gdppcjt); wine produced in countries i and j in year t-1 (respectively, prodit-1 and prodjt-1); exporter i being an Old World country (oldi); the annual average exchange rate between the currencies of countries i and j in year t (erijt); importer j’s European Union (EU) membership status in year t (eujt); the existence in year t of regional trade agreements (RTA) between countries i and j (rtaijt); the same official lan- guage in countries i and j (langij); and countries i and j sharing common religion beliefs (religij).2 As regards the expected sign of the explanatory variables, distance should have a negative effect on the probability of wine trade and on market shares as this is a proxy for transport costs. On the other hand, the existence of a regional trade agreement between two countries should reduce trade costs and, consequently, may have a positive effect. The same is expected for the effect of wine production in the importing country because more wine produced means possessing a greater stock to export. Exporting countries from the Old World may also present some advantage in wine trade due to their experience. Sharing a common lan- guage or common religion beliefs should also have a positive effect in both parts of the model because it represents higher cultural proximity. Moreover, the sign of the effect provoked by an explanatory variable may not be the same in the first and second parts. For example, the GDP per capita of importing countries represents purchasing power, which should increase the probability of trade and quantity traded but can have a different impact on market shares. In fact, a hypothesis to be tested in the results section is that higher purchasing power may lead importing countries to search for differentiation, therefore spreading their imports across more countries. 2 The effect of tariff and non-tariff measures on wine trade have also been studied using the gravity model. The literature shows that specific tariffs may be a deterrent to trade (Dal Bianco et al., 2016; Dal Bianco et al., 2017) or not have a significant effect on certain wines (Gouveia et al., 2018; Macedo et al., 2019, 2020). However, there is an ongoing debate about non-tariff measures as the papers published on this subject relatively recent and scarce (Dal Bianco et al., 2016; Santeramo et al., 2019). Dal Bianco et al. (2016) find that only some non-tariff measures present a significant negative effect on wine exports, while Santeramo et al. (2019) find more of a positive impact of some non-tariff measures in wine imports. Although recognizing the relevance of this issue, it was felt that it deserves a deeper analysis, which is beyond the scope of this paper, i.e. the application of the FRM to international wine trade. 266 Anthony Macedo, João Rebelo, Sofia Gouveia Similarly, wine production and EU membership of importing countries may have a posi- tive impact on the probability of trade because it may represent a higher degree of cultural openness to wine consumption, however that may also result in a taste for variety that would have a negative impact on market shares. Regarding the exchange rate, depreciation of an exporter i is expected to have a positive impact on market shares, because exports of country i become cheaper for importers j, while the converse will be true if exporter i’s currency undergoes appreciation. However, as sug- gested by Chaney (2016), the effect on the probability of country i exporting to j may be the opposite because, in the presence of fixed costs and liquidity constraints, a depreciation will also mean that the value of domestic assets abroad decreases and, therefore, foreign markets become less accessible to some firms of country i whilst appreciation will have the opposite effect. These two effects of exchange rate are not incompatible considering a two-part model. The first part of the model should be more sensitive to fixed costs constraints because it refers to the factors affecting the probability of trade. On the other hand, the second part of the model should not be sensitive to fixed costs constraints because it only considers existent trade relationships. Studying French wine exports, Cardebat and Figuet (2019) also identified that variations in exchange rates can lead to quality sorting, in the sense that higher-quality wines are less sensitive to exchange rate movements. The model is applied to the fifteen main wine producers3 (Argentina, Australia, Chile, China, France, Germany, Greece, Hungary, Italy, Portugal, Romania, Russia, Spain, South Af- rica and USA), focusing on 193 trade partners, all of which represent around 89% of world bottled wine exports during the period 1999-2014. Regarding the explained variable, data for wine exports of the main exporting countries and total imports of each destination country are from the COMTRADE database4 in US dollars. The computation of the market share in this work measures the weight that an exporting country i has on total exports to a certain country j (or, in other words, on the total imports of a certain country j). About the explana- tory variables, the sources are: World Development Indicators (WDI) database for GDP per capita in current US dollars and nominal exchange rate in local currency unit per US dollar (used to compute the exchange rate of exporting countries per 1 currency unit of importing countries); International Organisation of Vine and Wine (OIV) for wine production data; Cardebat (2019) for Old World countries5; EU official website6 for EU membership dummy variable; and Gravity database from Centre d’Études Prospectives et d’Informations Internation- ales (CEPII) for bilateral distance in kilometres weighted by population, religious proximity7, dummy for common official language and dummy for regional trade agreements. Descriptive statistics of all variables considered in the paper are present in Table A.1 in Appendix. 3 In this paper the sample considers the main wine producers instead of the main exporters to limit bias. Data on worldwide wine trade is not completely trackable and, consequently, they do not distinguish export from re-export. For that reason, countries such as United Kingdom, Switzerland and Hong Kong appear among the main wine exporters in COMTRADE database without producing relevant quantities of wine. 4 Harmonised system codes starting by 220421 5 Contrary to Cardebat (2019), in this study Hungary is included in the group of Old World countries for having a historically relevant wine industry (Luptak et al., 2016). 6 Website: https://europa.eu/. 7 Index from Disdier and Mayer (2007) calculated by adding the products of the relative proportions of Catholics, Protestants and Muslims in the exporting and importing countries. Higher values in this index mean sharing more common religion beliefs. 267Export propensity and intensity in the wine industry 3. Results and discussion In order to assess the robustness of the results, the 2P-FRM is estimated assuming four alternative non-linear conditional mean specifications (cauchit, logit, probit, and loglog) for F(∙) and M(∙). Based on RESET and Goodness-of-functional-form (GOFF) tests (Ramalho et al. 2011), the results suggest adopting the logit specification in both parts of the model (see Table A.2 in Appendix).8 Therefore, the results of the estimations with logit are presented in Table 1.9 Columns (1) and (3) refer to the coefficients estimated for the export propensity (equations 1 and 2) and the export intensity (equation 3), respectively. Table 1 also presents the estimations of two APEs for each explanatory variable: in column (2) the effect on the 8 RESET and GOFF tests to the specification also reject the one-part model, which corroborate with the results of the P test suggested by Ramalho et al (2011). Additionally, estimations of the 2P-FRM with a sample of 154 exporting countries (the maximum number for which information is available) were attempted but the specification was rejected by RESET and GOFF tests. However, the results are fairly similar. The only differences in signs and significance of coefficients estimated are that, in the 1st part, religious proximity is not statistically significant and, in the 2nd part, exchange rate is not significant while RTA is positive and significant. These results are available upon request. 9 For robustness of analysis, estimates were also made by splitting the sample into two sub-periods: before and after the financial crisis of 2008. The results do not indicate any marked differences between the two sub-periods (Table A.3 in Appendix). Table 1. 2P-FRM estimations, Average Partial Effects and Total Partial Effect. Variables (1) (2) (3) (4) (5) Export propensity Export intensity TPE β1 APE β2 APE GDP pc importer (log) 0.349*** 0.057*** -0.167*** -0.014*** -0.003** (0.029) (0.004) (0.032) (0.003) (0.001) EU importer 1.037*** 0.170*** -0.292*** -0.024*** 0.000 (0.145) (0.024) (0.105) (0.009) (0.005) Production importer (t-1) (log) 0.127*** 0.021*** -0.054*** -0.004*** -0.001 (0.012) (0.002) (0.012) (0.001) (0.001) Production exporter (t-1) (log) 0.824*** 0.135*** 0.661*** 0.054*** 0.038*** (0.047) (0.007) (0.066) (0.005) (0.003) Old World exporter 0.772*** 0.127*** -0.066 -0.005 0.007 (0.082) (0.013) (0.117) (0.010) (0.005) Exch. rate (log) -0.069*** -0.011*** 0.038** 0.003** 0.001 (0.013) (0.002) (0.016) (0.001) (0.001) RTA 0.611*** 0.100*** 0.071 0.006 0.010** (0.107) (0.017) (0.102) (0.008) (0.004) Distance (log) -0.278*** -0.046*** -0.620*** -0.051*** -0.029*** (0.060) (0.010) (0.054) (0.005) (0.003) 268 Anthony Macedo, João Rebelo, Sofia Gouveia probability of wine of country i being imported by a certain country j (equation 5); and in column (4) the effect on the magnitude of a non-zero market share of country i’s wine in country j (equation 6). Column (5) includes the TPE (equation 7). Time effects are consid- ered through yearly dummy variables (omitted due to space considerations), and standard errors account for intra-group correlation. The results suggest that all parameters estimated in the 1st part of the model (propensity to export) are statistically significant. By observing the APEs, it is also possible to confirm the expected signs for all explanatory variables. It is estimated that the distance between two countries has a negative effect on the probability of wine of an exporting country i being imported by a certain country j. A depreciation of the exporter’s currency in relation to an importer’s currency has a negative effect on the export propensity, which following Chaney Variables (1) (2) (3) (4) (5) Export propensity Export intensity TPE β1 APE β2 APE Common language 1.486*** 0.244*** 1.469*** 0.121*** 0.080*** (0.138) (0.022) (0.106) (0.009) (0.005) Religious proximity 1.644*** 0.270*** -0.141 -0.012 0.014** (0.192) (0.031) (0.149) (0.012) (0.007) Constant -9.317*** -1.961*** (0.650) (0.750) Observations 44,313 22,671 Pseudo R2 0.341 0.328 Time effects’ significance 44.76*** 97.11*** [0.000] [0.000] RESET 0.205 2.476 [0.651] [0.116] GOFF1 1.049 1.093 [0.306] [0.296] GOFF2 1.294 0.216 [0.255] [0.642] Note: Robust standard errors in parentheses; *** p<0.01, ** p<0.05, * p<0.1; Figures in [ ] indicate p-values; Time dummies included but not reported; GOFF1 and GOFF2 are goodness-of-functional-form tests; GDP pc is the per capita gross domestic product of the importer; EU is European Union membership (or not) of the importer; wine produced in period t-1 in importing and exporting countries are represented by Production importer and Production exporter, respectively; Old World is a dummy variable coded 1 if the exporting country is considered an Old World country in wine trade and 0 otherwise; Exch. Rate is the annual average exchange rate between the currencies of importing countries and exporting countries; RTA is a dummy variable coded 1 if there is a regional trade agreement between the exporter and the im- porter; Distance is the geographical distance between the exporter and the importer; Common langua- ge is a dummy variable coded 1 if importer and exporter share the same official language; and Religious proximity is an index measuring common religion beliefs. Source: Authors’ computation. 269Export propensity and intensity in the wine industry (2016) is an expected result in the presence of fixed costs and liquidity constraints. On the other hand, GDP per capita of the importers, wine production of both trade partners, the exporting country being an Old World country, EU membership of importing countries, re- gional trade agreements established between both countries, and cultural proximity aspects, such as religion and language, present a positive effect. In the 2nd part of the model most of the covariates have a statistically significant effect, the exceptions being Old World exporting countries, regional trade agreements and religious proximity. The APEs indicate that the market share of wine from an exporting country i in a certain country j is negatively affected by GDP per capita, wine production, EU membership of importing countries and bilateral distance. However, the market share is positively affected by wine production in the exporting country, common language between trade partners and depreciation of the exporter’s currency in relation to an importer’s currency. It is noticeable that only three explanatory variables have APEs with similar signs in both 2nd and 1st parts: distance, wine production in the exporting country, and common language between trading partners. Regarding the other variables, as expected, the effect of deprecia- tion of the exporter’s currency becomes positive in the 2nd part, as a result of wine becoming cheaper for importers. The effect of GDP per capita of importing countries goes from posi- tive to negative, confirming the hypothesis that higher purchasing power leads importing countries to search for variety, therefore ranging across more countries to obtain their im- ports. A negative effect is also caused by higher wine production in importing countries and importer’s EU membership, because these variables seems to indicate higher cultural open- ness to wine consumption but also a taste for variety, which has a negative impact on market shares. Regional trade agreements, religious proximity, and Old World exporting countries have a positive impact on propensity to trade but have not a significant effect on the intensity of trade, meaning that historical, cultural, and commercial relationship are advantages in market entrance. The TPEs show the effect of each covariate on the magnitude of any market share (in- cluding zero) of country i’s wine in country j. This can be interpreted as a global view of the effect of explanatory variables in both parts of the model. Therefore, the results suggest that, overall, per capita GDP and distance have a negative impact on market shares, while wine production of exporting countries, regional trade agreements, common language, and reli- gious proximity have a positive impact. As far as it is known, there are no fully comparable results in wine trade literature, as the dependent variable is usually not the market share. The closest comparison can be made with studies focused on determinants of wine exports, in which the dependent variable is usually the value or volume of exports. Most of these works also suggest that bilateral distance has a negative effect on trade (Castillo et al. 2016; Dal Bianco et al. 2016; Lombardi et al. 2016), common language facilitates commercial relationship (Castillo et al. 2016; Dal Bianco et al. 2016; Lombardi et al. 2016; Gouveia et al. 2018), regional trade agreements enhance trade (Dascal et al. 2002; Castillo et al. 2016), and wine production in exporting countries creates an export stimulus (Dascal et al. 2002; Agostino and Trivieri 2014; Dal Bianco et al. 2016). With regard to the effects estimated for GDP per capita, wine production, and EU member- ship of importing countries, they are not comparable to such studies due to the difference in the nature of the dependent variable. 270 Anthony Macedo, João Rebelo, Sofia Gouveia 4. Conclusion Competition may be within domestic markets but, in a globalised economy, competitive- ness is more and more dependent on the ability of industries to trade at an international level. One of the measures of competitiveness is the export market share, allowing the establishment of direct trading positions. However, when studying international competitiveness it is impor- tant to distinguish export intensity from export propensity, as some determinants can present opposite effects in these two respective measures, suggesting that caution should be exercised in policies aiming to improve export propensity because those same policies can lead to dete- rioration in export intensity, and vice-versa. These aspects call for the use of a 2P-FRM. It is inferred through the results that strategic decisions of firms in the wine sector should vary according to the objective of the boards. In the case of managerial boards aiming to en- ter new markets, they should focus on importing countries with high purchasing power, EU membership, and high levels of wine production. These are characteristics of countries with a tradition of wine consumption, which means greater openness to try new wines. Besides that, exploring markets with cultural (religious and linguistic) and commercial (trade agree- ments) proximity seems to increase trade propensity. On the other hand, corporate decisions should also be concerned about exchange rates and costs underlying distant markets. A different strategy should be adopted for managerial boards aspiring to increase market shares. In fact, markets with high purchasing power, EU membership, and high levels of wine production become less attractive because their taste for diversification in wine consumption may limit market shares. Therefore, focus should be turned to markets with less tradition of wine consumption, despite the challenge of surpassing greater barriers to entry. This chal- lenge can be overcome, for example, through the establishment of trade agreements and tak- ing advantage of cultural proximity. The market share is a simple and informative measure of international competitiveness; nevertheless, in future research, it would be interesting to compare different measures to test the robustness of the results. Also, the search for explanatory variables not yet considered in the literature should be done, since their eventual omission can lead to endogeneity and to the subsequent correction of the econometric model. In terms of methodology, including country-pair fixed effects in the 2P-FRM, to treat neglected individual heterogeneity, or con- sidering the Heckman fractional model in development, by Schwiebert (2018), could bring new insights to the nature of international competitiveness in the wine sector. 5. Acknowledgements This paper benefits from the comments of three anonymous referees. The usual disclaim- er applies. 6. Funding This work is supported by the project VINCI – Wine, Innovation and International Competitiveness, under the operation number SOE3/P2/F0917, FEDER – Interreg SUDOE, and national funds, through the FCT – Portuguese Foundation for Science and Technology under the project UIDB/SOC/04011/2020. 271Export propensity and intensity in the wine industry 7. References Agostino, M. and Trivieri, F. (2014). Geographical Indication and Wine Exports. An Empirical Investigation Considering the Major European Producers. Food Policy 46: 22-36. Anderson, J.E. and van Wincoop, E. (2003). Gravity with Gravitas: A Solution to the Bor- der Puzzle. The American Economic Review 93(1): 170-192. Anderson, K. (2017). How Might Climate Changes and Preference Changes Affect the Competitiveness of the World’s Wine Regions? Wine Economics and Policy 6(1): 23-27. Anderson, K. (2019). Asia’s Emergence in Global Beverage Markets: The Rise of Wine. Wine Economics Research Centre Working Paper No. 0219, University of Ade- laide. Anderson, K., Meloni, G. and Swinnen, J. (2018). Global Alcohol Markets: Evolving Con- sumption Patterns, Regulations, and Industrial Organizations. Annual Review of Resource Economics 10(1): 105-132. Anderson, K. and Wittwer, G. (2013). Modeling Global Wine Markets to 2018: Exchange Rates, Taste Changes, and China’s Import Growth. Journal of Wine Economics 8(2): 131-158. Anderson, K. and Wittwer, G. (2017). U.K. and Global Wine Markets by 2025, and Impli- cations of Brexit. Journal of Wine Economics 12(3): 221-251. Anderson, J.E. and Yotov, A.V. (2012). Gold Standard Gravity. NBER Working Paper No. 17835. Ashenfelter, O. and Storchmann, K. (2016a). Climate Change and Wine: A Review of the Economic Implications. Journal of Wine Economics 11(1): 105-138. Ashenfelter, O. and Storchmann, K. (2016b). The Economics of Wine, Weather, and Cli- mate Change. Review of Environmental Economics and Policy 10(1): 25-46. Banterle, A. and Carraresi, L. (2007). Competitive Performance Analysis and European Union Trade: The Case of the Prepared Swine Meat Sector. Acta Agriculturae Scandi- navica Section C – Food Economics 4(3): 159-172. Bargain, O., Cardebat, J.M., Chiappini, R. and Laffitte, C. (2018). Long-Term Prospects of the International Wine Trade. Journal of Wine Economics 13(4): 461–468. Bergstrand, J.H. (1989). The Generalized Gravity Equation, Monopolistic Competition, and Factor Proportions Theory in International Trade. The Review of Economics and Statistics 71(1): 143-153. Bouët, A., Emlinger, C. and Lamani, V. (2017). What Determines Exports of Luxury Prod- ucts? The Case of Cognac. Journal of Wine Economics 12(1): 37-58. Cardebat, J.-M. (2019). Introduction: Does a National Model Exist Which Favors Trade Performance? In Ugaglia, A., Cardebat, J.-M. and Corsi, A. (eds.), The Palgrave Handbook of Wine Industry Economics, Palgrave Macmillan: 455-461. Cardebat, J.-M. and Figuet, J.M. (2019). The Impact of Exchange Rates on French Wine Exports. Journal of Wine Economics 14(1): 71-89. Carraresi, L. and Banterle, A. (2008). Agri-food Competitive Performance in EU Coun- tries: A Fifteen-year Retrospective. International Food and Agribusiness Management Review 18(2): 37-62. 272 Anthony Macedo, João Rebelo, Sofia Gouveia Castillo, J.S., Villanueva, E.C. and García-Cortijo M.C. (2016). The International Wine Trade and its New Export Dynamics (1988-2012): A Gravity Model Approach. Agri- business 32(4): 466-481. Chaney, T. (2016). Liquidity Constrained Exporters. Journal of Economic Dynamics and Control 72: 141-154. Cirera, X., Foliano, F. and Gasiorek, M. (2016). The Impact of Preferences on Developing Countries’ Exports to the European Union: Bilateral Gravity Modelling at the Prod- uct Level. Empirical Economics 50(1): 59-102. Dal Bianco, A., Boatto, V. and Caracciolo, F. (2013). Cultural Convergence in World Wine Consumption. Revista de la Facultad de Ciencias Agrarias 45(2): 219-231. Dal Bianco, A., Boatto, V. L., Caracciolo, F. and Santeramo, F.G. (2016). Tariffs and Non- tariff Frictions in the World Wine Trade. European Review of Agricultural Economics 43(1): 31-57. Dal Bianco, A., Estrella-Orrego, M., Boatto, V., and Gennari, A. (2017). Is Mercosur Pro- moting Trade? Insights from Argentinean Wine Exports. Spanish Journal of Agricul- tural Research 15(1): 1-10. Dascal, D., Mattas, K. and Tzouvelekas, V. (2002). An Analysis of EU Wine Trade: A Gravity Model Approach. International Advances in Economic Research 8(2): 135- 147. Disdier, A.C. and Mayer, T. (2007). Je T’aime, Moi Non Plus: Bilateral Opinions and Inter- national Trade. European Journal of Political Economy 23(4): 1140-1159. European Commission (2014). Study on the Competitiveness of European Wines. Consu- lenti per la Gestione Aziendale – COGEA. Gallani, S., Krishnan, R. and Wooldridge, J.M. (2015). Applications of Fractional Response Model to the study of Bounded Dependent Variables in Accounting Research. Har- vard Business School Working Paper No. 16-016. Gouveia, S., Rebelo, J. and Lourenço-Gomes, L. (2018). Port Wine Exports: A Gravity Model Approach. International Journal of Wine Business Research 30(2): 218-242. Hatzichronoglou, T. (1996). Globalisation and Competitiveness: Relevant Indicators. OECD Science, Technology and Industry Working Papers No. 1996/5. Holmes, A.J., and Anderson, K. (2017). Convergence in National Alcohol Consumption Patterns: New Global Indicators. Journal of Wine Economics 12(2): 117-148. Lombardi, P., Dal Bianco, A., Freda, R., Caracciolo, F. and Cembalo, L. (2016). Develop- ment and Trade Competitiveness of the European Wine Sector: A Gravity Analysis of Intra-EU Flows. Wine Economics and Policy 5(2): 50-59. Luptak, M., Boda, D. and Szucs, G. (2016). Profitability and Capital Structure: An Empiri- cal Study of French and Hungarian Wine Producers in 2004-2013. Business Systems Research 7(1): 89-103. Macedo, A., Gouveia, S. and Rebelo, J. (2019). Does Wine Quality Have a Bearing on Exports? AGRIS on-line Papers in Economics and Informatics 11(4): 49-59. Macedo, A., Gouveia, S. and Rebelo, J. (2020). Horizontal Differentiation and Determi- nants of Wine Exports: Evidence from Portugal. Journal of Wine Economics (forth- coming). McCallum, J. (1995). National Borders Matter: Canada-U.S. Regional Trade Patterns. The American Economic Review 85(3): 615-623. 273Export propensity and intensity in the wine industry Papke, L.E. and Wooldridge, J.M. (1996). Econometric Methods for Fractional Response Variables with an Application to 401(k) Plan Participation Rates. Journal of Applied Econometrics 11(6): 619-632. Papke, L.E. and Wooldridge, J.M. (2008). Panel Data Methods for Fractional Response Variables with an Application to Test Pass Rates. Journal of Econometrics 145(1): 121-133. Ramalho, E., Ramalho, J. and Murteira, J. (2011). Alternative Estimating and Testing Empirical Strategies for Fractional Regression Models. Journal of Economic Surveys 25(1): 19-68. Santeramo, F.G., Lamonaca, E., Nardone, G. and Seccia, A. (2019). The Benefits of Coun- try-Specific Non-Tariff Measures in World Wine Trade. Wine Economics and Policy 8(1): 28-37. Silva, J.S. and Tenreyro, S. (2006). The Log of Gravity. The Review of Economics and Statis- tics 88(4): 641-658. Smith, D.E. and Mitry, D.J. (2007). Cultural Convergence: Consumer Behavioral Changes in the European Wine Market. Journal of Wine Research 18(2): 107-112. Schwiebert, J. (2018). A Sample Selection Model for Fractional Response Variables. Uni- versity of Lüneburg Working Paper No. 382. Ugaglia, A., Cardebat, J.-M., and Corsi, A. (2019). The Palgrave Handbook of Wine Indus- try Economics. Palgrave Macmillan. Wijnands, J., Bremmers, H., van der Meulen, B. and Poppe, K. (2008). An Economic and Legal Assessment of the EU food Industry’s Competitiveness. Agribusiness 24(4): 417-439. 274 Anthony Macedo, João Rebelo, Sofia Gouveia Appendix Table A.1. Descriptive statistics. Variables Obs. Mean Med. Std. Dev. Min Max Market share 44313 0.1 0.0 0.1 0.0 1.0 Distance 44313 7823.0 7687.4 4467.6 204.8 19539.5 GDPpc importer 44313 11905.0 3831.5 17791.7 102.6 119225.4 RTA (yes=1; 0 otherwise) 44313 0.2 0.0 0.0 1.0 Exch. rate 44313 23.5 0.3 109.0 0.0 2320.0 EU importer (yes=1; 0 otherwise) 44313 0.1 0.0 0.0 1.0 Production exporter (t-1) 44313 16285.7 10007.0 15483.5 1762.0 60535.0 Production importer (t-1) 44313 1359.8 0.0 6004.3 0.0 60535.0 Old World exporter (yes=1; 0 otherwise) 44313 0.5 0.0 0.0 1.0 Common language (yes=1; 0 otherwise) 44313 0.1 0.0 0.0 1.0 Religious proximity 44313 0.2 0.0 0.2 0.0 1.0 Note: For the binary variables RTA, common language, Old World exporter, and EU membership of the importer the mean represents the percentage of observations equal to one. Source: Authors’ computa- tion. 275Export propensity and intensity in the wine industry Table A.2. Alternative functional form specifications of the 1st and 2nd parts of 2P-FRM estimations. Variables Export propensity Export intensity Cauchit β1 Probit β1 Loglog β1 Cauchit β2 Probit β2 Loglog β2 GDPpc importer (log) 0.398*** 0.200*** 0.211*** -0.386*** -0.084*** -0.061*** (0.034) (0.017) (0.018) (0.086) (0.016) (0.012) EU importer 0.962*** 0.641*** 0.826*** -0.707 -0.153*** -0.114*** (0.207) (0.082) (0.123) (0.467) (0.051) (0.036) Production importer (t-1) (log) 0.131*** 0.077*** 0.078*** -0.167*** -0.026*** -0.018*** (0.014) (0.007) (0.008) (0.055) (0.006) (0.004) Production exporter (t-1) (log) 0.850*** 0.487*** 0.595*** 1.095*** 0.351*** 0.274*** (0.059) (0.027) (0.032) (0.338) (0.030) (0.021) Old World exporter 0.825*** 0.460*** 0.495*** -0.288 -0.035 -0.026 (0.099) (0.048) (0.051) (0.456) (0.056) (0.039) Exch. rate (log) -0.084*** -0.039*** -0.034*** 0.086* 0.021*** 0.017*** (0.017) (0.007) (0.007) (0.051) (0.008) (0.005) RTA 0.642*** 0.355*** 0.461*** 0.180 0.038 0.030 (0.128) (0.063) (0.081) (0.257) (0.051) (0.037) Distance (log) -0.359*** -0.145*** -0.190*** -1.199*** -0.314*** -0.231*** (0.067) (0.037) (0.040) (0.138) (0.027) (0.021) Common language 1.490*** 0.886*** 0.988*** 2.399*** 0.794*** 0.642*** (0.161) (0.082) (0.107) (0.246) (0.058) (0.050) Religious proximity 1.839*** 0.936*** 1.132*** -0.128 -0.076 -0.058 (0.242) (0.112) (0.138) (0.318) (0.076) (0.058) Constant -9.322*** -5.629*** -5.916*** -0.355 -1.285*** -0.962*** (0.838) (0.376) (0.436) (3.527) (0.355) (0.251) Observations 44,313 44,313 44,313 22,671 22,671 22,671 Pseudo R2 0.338 0.341 0.338 0.305 0.325 0.318 Time effects’ significance 45.40*** 44.97*** 52.53*** 39.34*** 107.03*** 116.64*** [0.000] [0.000] [0.000] [0.001] [0.000] [0.000] RESET 75.262*** 16.840*** 193.217*** 775.004*** 19.202*** 115.371*** [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] GOFF1 17.701*** 18.047*** n.a. 530.827*** 21.229*** n.a. [0.000] [0.000] [0.000] [0.000] GOFF2 91.675*** 7.477*** 212.317*** 307.573*** 19.215*** 114.896*** [0.000] [0.006] [0.000] [0.000] [0.000] [0.000] Note: Robust standard errors in parentheses; *** p<0.01, ** p<0.05, * p<0.1; Figures in [ ] indicate p-values; Time dummies included but not reported; n.a. = not applicable. Source: Authors’ computation. 276 Anthony Macedo, João Rebelo, Sofia Gouveia Ta b le A .3 . 2 P- FR M e st im at io n s sp lit ti n g t h e sa m p le in p re - a n d p o st -fi n an ci al c ri si s o f 2 00 8. V ar ia bl es Pe ri od 1 99 9- 20 08 Pe ri od 2 00 9- 20 14 Ex po rt p ro pe ns ity Ex po rt in te ns ity T PE Ex po rt p ro pe ns ity Ex po rt in te ns ity T PE β 1 A PE β 2 A PE β 1 A PE β 2 A PE G D P pc im po rt er ( lo g) 0. 34 7* ** 0. 05 7* ** -0 .1 63 ** * -0 .0 14 ** * -0 .0 03 * 0. 34 5* ** 0. 05 7* ** -0 .1 73 ** * -0 .0 13 ** * -0 .0 04 ** (0 .0 29 ) (0 .0 04 ) (0 .0 32 ) (0 .0 03 ) (0 .0 01 ) (0 .0 35 ) (0 .0 05 ) (0 .0 38 ) (0 .0 03 ) (0 .0 02 ) EU im po rt er 1. 12 2* ** 0. 18 4* ** -0 .2 48 ** -0 .0 21 ** 0. 00 4 0. 93 9* ** 0. 15 5* ** -0 .4 02 ** * -0 .0 31 ** * -0 .0 07 (0 .1 55 ) (0 .0 25 ) (0 .1 12 ) (0 .0 10 ) (0 .0 05 ) (0 .1 53 ) (0 .0 25 ) (0 .1 18 ) (0 .0 09 ) (0 .0 05 ) Pr od uc tio n im po rt er ( t- 1) ( lo g) 0. 12 2* ** 0. 02 0* ** -0 .0 58 ** * -0 .0 05 ** * -0 .0 01 0. 13 7* ** 0. 02 3* ** -0 .0 47 ** * -0 .0 04 ** * -0 .0 01 (0 .0 12 ) (0 .0 02 ) (0 .0 13 ) (0 .0 01 ) (0 .0 01 ) (0 .0 14 ) (0 .0 02 ) (0 .0 13 ) (0 .0 01 ) (0 .0 01 ) Pr od uc tio n ex po rt er ( t- 1) ( lo g) 0. 83 3* ** 0. 13 6* ** 0. 61 0* ** 0. 05 3* ** 0. 03 7* ** 0. 82 8* ** 0. 13 7* ** 0. 75 4* ** 0. 05 8* ** 0. 04 0* ** (0 .0 49 ) (0 .0 07 ) (0 .0 66 ) (0 .0 06 ) (0 .0 03 ) (0 .0 53 ) (0 .0 08 ) (0 .0 75 ) (0 .0 06 ) (0 .0 03 ) O ld W or ld e xp or te r 0. 73 9* ** 0. 12 1* ** -0 .0 30 -0 .0 03 0. 00 8 0. 81 5* ** 0. 13 5* ** -0 .1 17 -0 .0 09 0. 00 4 (0 .0 84 ) (0 .0 13 ) (0 .1 23 ) (0 .0 11 ) (0 .0 05 ) (0 .0 91 ) (0 .0 15 ) (0 .1 21 ) (0 .0 09 ) (0 .0 05 ) Ex ch . r at e (l og ) -0 .0 58 ** * -0 .0 10 ** * 0. 03 1* 0. 00 3* 0. 00 1 -0 .0 82 ** * -0 .0 14 ** * 0. 05 1* ** 0. 00 4* ** 0. 00 1* (0 .0 13 ) (0 .0 02 ) (0 .0 17 ) (0 .0 01 ) (0 .0 01 ) (0 .0 14 ) (0 .0 02 ) (0 .0 17 ) (0 .0 01 ) (0 .0 01 ) R TA 0. 62 0* ** 0. 10 1* ** 0. 13 9 0. 01 2 0. 01 4* ** 0. 60 9* ** 0. 10 1* ** -0 .0 12 -0 .0 01 0. 00 6 (0 .1 36 ) (0 .0 22 ) (0 .1 20 ) (0 .0 10 ) (0 .0 05 ) (0 .1 09 ) (0 .0 18 ) (0 .1 10 ) (0 .0 08 ) (0 .0 05 ) D is ta nc e (l og ) -0 .2 70 ** * -0 .0 44 ** * -0 .5 32 ** * -0 .0 46 ** * -0 .0 26 ** * -0 .2 95 ** * -0 .0 49 ** * -0 .7 60 ** * -0 .0 58 ** * -0 .0 35 ** * (0 .0 61 ) (0 .0 10 ) (0 .0 57 ) (0 .0 05 ) (0 .0 03 ) (0 .0 67 ) (0 .0 11 ) (0 .0 62 ) (0 .0 05 ) (0 .0 03 ) C om m on la ng ua ge 1. 45 7* ** 0. 23 8* ** 1. 44 0* ** 0. 12 5* ** 0. 08 0* ** 1. 52 2* ** 0. 25 1* ** 1. 49 8* ** 0. 11 5* ** 0. 07 8* ** (0 .1 43 ) (0 .0 22 ) (0 .1 07 ) (0 .0 10 ) (0 .0 05 ) (0 .1 51 ) (0 .0 24 ) (0 .1 14 ) (0 .0 09 ) (0 .0 05 ) R el ig io us p ro xi m ity 1. 59 0* ** 0. 26 0* ** -0 .1 67 -0 .0 14 0. 01 3* * 1. 74 2* ** 0. 28 8* ** -0 .1 03 -0 .0 08 0. 01 4* * (0 .1 90 ) (0 .0 30 ) (0 .1 51 ) (0 .0 13 ) (0 .0 07 ) (0 .2 26 ) (0 .0 36 ) (0 .1 59 ) (0 .0 12 ) (0 .0 07 ) C on st an t -9 .4 09 ** * -2 .2 40 ** * -9 .2 85 ** * -1 .6 41 ** (0 .6 80 ) (0 .7 82 ) (0 .7 29 ) (0 .8 27 ) O bs er va tio ns 27 ,4 08 13 ,5 01 16 ,9 05 22 ,6 71 277Export propensity and intensity in the wine industry V ar ia bl es Pe ri od 1 99 9- 20 08 Pe ri od 2 00 9- 20 14 Ex po rt p ro pe ns ity Ex po rt in te ns ity T PE Ex po rt p ro pe ns ity Ex po rt in te ns ity T PE β 1 A PE β 2 A PE β 1 A PE β 2 A PE Ps eu do R 2 0. 34 5 0. 30 2 0. 33 3 0. 37 8 T im e eff ec ts ’ s ig ni fic an ce 28 .1 9* ** 56 .9 9* ** 18 .4 4* ** 40 .8 4* ** [0 .0 01 ] [0 .0 00 ] [0 .0 02 ] [0 .0 00 ] R ES ET 0. 02 0 2. 04 4 0. 12 9 0. 46 0 [0 .8 88 ] [0 .1 53 ] [0 .7 19 ] [0 .4 97 ] G O FF 1 0. 02 6 1. 16 9 0. 86 4 0. 11 7 [0 .8 73 ] [0 .2 80 ] [0 .3 53 ] [0 .7 32 ] G O FF 2 0. 06 1 0. 41 4 1. 14 5 0. 02 3 [0 .8 04 ] [0 .5 20 ] [0 .2 85 ] [0 .8 81 ] N o te : R o b u st s ta n d ar d e rr o rs in p ar en th es es ; * ** p < 0. 01 , * * p < 0. 05 , * p < 0. 1; F ig u re s in [ ] i n d ic at e p- va lu es ; T im e d u m m ie s in cl u d ed b u t n o t re p o rt ed .