Description of the variables and previous literature.
In this section we will describe the variables that have been analyzed in this work and present some of the previous analysis done for each of them. Farther information about the variables and the data sources are reported in the appendix.
The analysis of both, goods and services exports have been carried out in this work. Most of the analyses on trade flows have focus on goods more than on services. Some of the causes of the relative limited number of studies on services could be: i) lack of comparable data2; ii) the share of services exports over total world’s trade keeps around 20 % of total international trade, despite its increase in absolute terms; iii) the modest share that services represent in international trade favors that they have been considered as non-tradable goods. Among the limited number of references analyzing trade in services, Ceglowski (2006) and Kimura and Lee (2006) analyzed the complementarity between trade in goods and services for the OECD countries using the gravity model. More recently, the gravity model has been used to study the international trade of services in other works as Kandilov and Greennes (2012).
As for trade flows, there are a large number of references analyzing the factors affecting capital flows and stocks. Within the bilateral flows linked with capital, we analyzed outward FDI stocks and outflows (using OECD database and national sources) and Portfolio Equity Stocks and Portfolio Equity Long Term Debt from the IMF - Coordinated Portfolio Investment Survey (CPIS).
The ambiguous theoretical role of distance on FDI has favored the proliferation of studies analyzing bilateral FDI using a gravity equation (Egger, 2008). On the one hand, a positive effect is expected when firms locate in other markets to overcome high trade costs and to be able to supply more remote places (horizontal FDI). However, empirical analyses have found a large negative coefficient for distance. This could be explained by a ‘dominant’ role of vertical FDI, that exploits international factors price differentials creating a global production network, or by the existence of high fixed costs increasing with distance. In this respect, for example, Kleinert and Toubal (2010) developed a theory-based gravity model for foreign affiliates.
There are a large number of works analyzing to what extent different factors affect other capital flows. First, given the ‘weightless’ nature of capital, one can expect that distance might not act as a detriment to bilateral capital flows since transactions costs for capital are not directly related with geographical distance. In fact, if investors look for more heterogeneous portfolios, they might want to invest in more distant countries, given that business cycles are correlated with distance (Clark and van Wincoop, 2001). In this regard, Portes and Rey (2005) analyzed the effect of information costs and its relation with distance, noting that the gravity model performs for capital flows at least as well as it does for trade flows. They noted that this negative relation between distance and international portfolio equity is related with information frictions. Aviat and Coeurdacier (2007) studied the complementarities between trade in goods and portfolio assets, controlling for some administrative characteristics as the existence of tax treaties, the proximity of legal systems or the existence of fiscal agreements. In Aggarwal et al. (2012) the authors analyzed the role of culture including Hofstede’s individualism, masculinity, power distance and uncertainty avoidance dimensions, finding that cultural similarities can help to offset the negative effect of distance. Daude and Fratzscher (2008) analyzed the role of information frictions and institutions quality on the main capital flows: FDI stocks; portfolio investment (debt and equity) and loans.
The logic of gravity models also applied to people mobility. In this work we are considering three different types of people flows: migration, incoming tertiary students and tourists’ arrivals. The gravity model has been applied in international migration, among others, in Mayda (2010), Anderson (2011) and Ortega and Peri (2013), all of them suggesting a negative relation between distance and migration, since distance increases the cost of migrate. As a measure of medium-term indicator for people mobility we are including the number of international tertiary students. The number of references analyzing the factors affecting international students’ mobility is scarcer. Among them: Rodriguez –Gonzalez et al. (2011) used a gravity model to evaluate the Erasmus program and Bessey (2012) studied whether international students arriving to Germany follow a similar pattern than international migration. As a short term indicator of people movement, we are including tourists’ arrivals. The gravity model approach has been used by Khadaroo and Seetanah (2008) to study the effect of infrastructures in tourism flows and by Keum (2010) to test the Linder hypothesis on tourism flows.
Finally, regarding information flows, the analysis of this type of flows is more limited given the lack of data. Within the bilateral flows that can be considered as information flows, we are including the exports of printed publications (exports under the code 49 in the Harmonized System from UNComtrade), the number minutes of international outgoing phone calls and the number of patents registered in the patent office by inventors from other countries. First, the factors affecting printed publications trade flows and other type of cultural flows has been studied in some recent references analyzing the effect of cultural distance in trade flows. To this regard, Disdier et al. (2010) used trade flows on cultural goods to build a proxy for cultural proximity between countries, finding a positive association between trade in cultural goods and total trade flows. Meng (2011) use a gravity model to estimate the factors affecting international trade in China’s cultural goods, differentiating between antiques, musical instruments, recorded media, paintings, other visual arts, craft, jewellery, photography, books, newspapers, other printed matter, film and video and architecture. A negative impact of distance was obtained in most of the cases but for books, newspapers and other printed material the estimates of distance, although vary depending on the type of good, is not very high: non-significant coefficient in books; in newspapers trade the coefficient was significant just at the 10% and the coefficient for distance in other printed materials was significant but lower in magnitude than for other type of goods -22%. Lastly, regarding patenting data, we are using data on the number of patents applications in each patent office by country of origin from the WIPO database. Archontakis and Varsakelis (2011) have used this database to study the factors that influence the flow of US patents to 27 OECD countries using a gravity model for the period 1995–2005, finding ambiguous results for the effect of distance. Finally, to the best of our knowledge there is not any reference analyzing the effect of distance on the minutes of international phone calls through gravity models.
The gravity model and the CAGE Framework
The gravity model was first used to analyze trade flows by Tinbergen (1962), by analogy with Newton’s law, but theoretical models at that time relied on differences in technology (Ricardian model) or in factor endowments (Heckscher-Ohlin model) across countries to explain trade patterns. However, its empirically well performance and stability have favored the appearance of theoretical models that provide a basis for the gravity equation (Anderson, 1979; Bergstrand, 1985 and 1989; Krugman, 1980; Anderson and van Wincoop, 2003; Helpman et al. 2008… among others). The good empirical performance and the theoretical basis benefited the proliferation of a large number of references analyzing the impact of different factors on trade flows through a gravity equation.
The gravity model is the framework used to analyze the effect of different factors in all kind of interactions in space as a function of the sizes of both countries and the distance between them. Specifically, flows between two countries are expected to increase with the size of both countries, while distance tends to lowered them. Ghemawat (2001) described ‘distance’ as a multidimensional construct that comprises cultural, administrative, geographical and economic aspects. Eq. 1 summarizes the basic gravity equation where the intensity of the connection between country o and country d in year t () is a linear function of a set of variables classified according to the 4 dimensions of distance: cultural distance , administrative distance , geographical distance and economic distance .
In order to control for the unobservable heterogeneity of the countries and for the specific characteristics of the time period analyzed, origin, destination and year fixed effects are included (. Anderson and van Wincoop (2003) noted that bilateral trade flows do not depend just on bilateral trade barriers but also on trade barriers across all trading partners. Previous works have controlled this interdependence including in the model some remoteness variable3. In this case, as suggested by Feenstra (2004), the ‘multilateral resistance terms’ have been controlled using country-specific fixed effects. A similar concept has been defined for migration by Bertoli and Moraga (2011) and by Anderson (2011). In general, we can assume that there are country–specific unobservable characteristics that have to be taken into account; to avoid bias in the coefficients estimates. Eq.1 describes a general structural gravity model considering four distance categories:
Errors are clustered by country-pair to avoid potential heteroskedasticity problems and to relax the assumption of independence within observations for the same country-pair during the period studied. Finally, it is important to note that in this work, we are assuming that different flows are independent between them. Other works have analyzed the interrelations between different type of flows analyzing their complementarity, but the study of the way in which all the flows interact is out of the scope of this project.