Preliminary and Incomplete Abstract



Download 1,67 Mb.
Page4/10
Date conversion08.12.2018
Size1,67 Mb.
1   2   3   4   5   6   7   8   9   10

A general specification for the gravity model


In this section, we follow Head and Mayer (2013) that identified some variables typically used in the gravity models for trade flows to choose a set of variables that we think that could have an impact in most of the flows that we are analyzing. Head and Mayer (2013) make a meta-analysis taking Disdier and Head (2008) as a starting point. They consider 159 papers and more than 2500 estimates from those published in top 5 journals, the Journal of International Economics and the Review of Economics and Statistics, from 2006 to 2012 together with some other papers that estimates the trade costs elasticity.

Within the variables identified, geographical distance and sharing a common border are the ones most commonly analyzed in the literature. The distance variable that we are using takes city-level data to assess the geographic distribution of population (in 2004) in each region. Then, distance between two countries is based on bilateral distances between the biggest cities of those two countries, weighting inter-city distances by the share of the city in the overall country’s population4. Another variable that is commonly used in gravity models is a dummy variable identifying whether both countries share the same official language, which is a proxy of cultural similarity. A dummy variable that identifies countries that have ever had a colonial linkage is also included as a measure of historical proximity. Among the variables related with administrative characteristics, Head and Mayer (2013) noted that one of the policy variables that have been more broadly studied on gravity models for trade flows is the effect of trade agreements or multilateral blocs. The literature has tried to measure the effect of Trade Agreements in general (RTA), and of some of them in particular (CUSA/NAFTA or the EU). We include two different variables depending on the type of flow: a dummy variable that captures the effect of any trade agreement in force for merchandise and services exports and a dummy variable that captures the effect of some regional blocs: MERCOSUR, ASEAN, CARICOM, EU, NAFTA and GCC. Both variables have been retrieved from the CEPII dataset and have been updated using data on the WTO.

Head and Mayer (2013) also mentioned that the effect of sharing the same currency has been largely studied in the literature. We are not including this variable for two reasons: i) this variable is included in the meta-analysis of Head and Mayer (2013) for the coefficients estimates of gravity model for trade flows, but we doubt that it could be an additional factor across all type of flows; ii) Head and Mayer (2013) recognized that the results obtained in the literature for this variable have been subject to controversy and the authors pointed that the differences in the estimates come from differences in the set of fixed effects and on the estimation procedures used. As cited by Head and Mayer (2013), Rose (2000) found that trade flows triple when countries share a common currency, while Baldwin (2006) noted that the literature has found an average effect of 30% and Santos–Silva and Tenreyro (2006) found no trade effect for the euro. Finally, Baldwin and Taglioni (2007) found a significant coefficient of -0.09. In fact, when we included this variable with the rest of the variables, we obtained a negative and significant sign around -0.4 for merchandise trade (these results are not included here, but are available upon request).

As a measure of the total size of the countries, we include the product (logged) of the gdps or populations of both countries. Table 4 in Head and Mayer (2013) includes estimates coefficients of gdp of the exporting and importing countries separately. However, the number of references that obtain an estimate for these two variables through a structural gravity model is very limited, because it is not possible to identify this effect when exporter and importer dummy variables are used. In this case for trade and capital flows we are using the product of the gdps and for information and people flows we are using the product of the population of both countries.

Finally, we are including an additional variable that we consider that it is relevant to explain the interactions between countries: the disparities in pc income. We have defined this variable as the ratio between the maximum and the minimum of the pc income for each country-pair. For trade flows, the Linder hypothesis (Linder, 1961) suggests that countries of similar income per capita should trade more intensely with one another. Also for FDI, Fajgelbaum et al. (2011) provided evidence on the disproportional response of FDI to differences in per capita income. Lucas (1990) noted that capital does not flow from development to developing countries contrary to the expected by classical economic theory. This observation has been named the Lucas paradox. Finally, according to Borjas (1987, 1991) the migration decision depends, among other things, on how a potential migrant perceives his or her gains if decide to migrate from a labor market to another country. The migration decision thus, depends according to Borjas’ theory on the average difference in wages across countries and on whether the worker’s abilities and other human capital can fit into the destination country labor.

Table A1 in the appendix show a comparison between the coefficient estimates that we obtain for merchandise exports and the median coefficients obtained using structural gravity equations in Head and Mayer (2013). Despite some differences, the results seem to be quite consistent with the median results found in the literature.



    1. Results

Table 1 shows the results for the gravity model for the 12 types of flows that we are analyzing and the coefficient estimates for each of the variables included in the general specification. In general, it is notable that despite the fact that we are using almost the same specification of all types of flows, the goodness of fit of the models is around 70%, reaching almost 90% in services exports, international tourists’ arrivals and outgoing phone calls. The coefficient for common official language is significant almost across the board, obtaining a positive coefficient estimate in all the cases, with the only exception of services exports and portfolio long term debt –the first one with a non - significant coefficient and the second one with a positive estimate, but significant just at the 10%- . In general, the effect found for information and people flows is higher than for the rest of the flows, with the larger effect obtained in the case of printed publications exports, which seems reasonable given the nature of these flows.

The colonial linkage variable is significant in all the cases, with two exceptions: portfolio equity assets and long term debt. The non-significant effect on portfolio equity assets and long term debt is consistent with most of the findings in Daude and Fratzscher (2008). Among the significant cases, the effect is positive with larger impacts for long and medium term people flows (emigration and international students) and for FDI flows and stocks. As potential causes to explain the larger impacts in these types of flows, it can be pointed the cultural proximity between colonies and colonizers or the existence of lower administrative barriers for countries with colonial linkages. The regional trade agreement or regional bloc variable is significant at least at the 5% in 7 cases, all of them with a positive impact, with the exceptions of emigrant intensities and of patenting activity. The variables in which a positive impact is found are: merchandise and services exports, FDI outflows, portfolio long term debt and international tourists’ arrivals. The impact in these cases is quite similar, although lower for merchandise exports than for the rest of the cases.

The geographical distance variable is significant across the board and in all the cases acting in detriment of the magnitude of the flows, as in trade models characterized by “iceberg” trade costs (Samuelson, 1952). The effect of this variable is larger for people and information flows. This can be related with the need of keep some face-to-face contact for social interactions and with the increasing cost of trade printed publications to large distances. Capital flows are the ones that are less affected by geographical distance, what can be explained by their ‘weightless nature’. However, it is surprising that the magnitude of the coefficients for these flows is still slightly below -1, but it is consistent with the findings in the literature (i. e. Daude and Fratzscher (2008) obtained in their baseline estimation a coefficient of -0.676 for portfolio equity and -0.808 for long term debt). Portes and Rey (2005) suggested that distance is a proxy for frictions due to information and transaction costs. Within the trade pillar, the effect for merchandise exports is similar to the one found for people and information flows, while the effect for services is more similar to the one found for some capital flows. The similarity between trade in services and FDI can be explained by the fact that, according to the GATS5, FDI is one of the modes that could be used to trade services. This relatively lower effect of distance on trade in services can also be explained by the nature of some services as consultancy, call centers, programmers…, which can be traded without facing transportation costs.

Sharing a common border is somehow related with the geographical distance variable, getting a positive and significant impact at least at the 5% in five out of the twelve types of flows analyzed. Consistently with the results found for the geographical distance variable, these 5 flows are emigration intensities, international tourists, international phone calls, merchandise exports and international patents, all of them, except patents are among the ones that get a higher impact also for distance.



Regarding the variable of the ratio of pc income, it is significant in eleven out of the twelve flows analyzed. The only exception is found for merchandise exports, in which a non-significant coefficient is found. Hallak (2010) demonstrated that although the Linder hypothesis is rejected when using aggregate data, it is confirmed using sectoral disaggregated data, noting that Linder explicitly argued that the link between income and

Table 1. General gravity model

 

 

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

(10)

(11)

(12)

CAGE dimension

Variables

Merchandise exports

Services exports

FDI outward stocks

FDI outflows

Portfolio equity assets

Portfolio long term debt

Intl Students Arrivals

Emigration intensities6

Intl Tourists Arrivals

Outgoing phone calls

Printed publications exports

Intl. Patenting activity

Cultural

Common official language 

0.784***

0.118

0.622***

0.436***

0.586***

0.207*

1.164***

0.801***

0.923***

1.002***

1.821***

0.634***

(0.0596)

(0.115)

(0.141)

(0.135)

(0.169)

(0.112)

(0.0858)

(0.0522)

(0.101)

(0.0865)

(0.0887)

(0.0740)

Administrative

Colonial link 

0.730***

0.808***

1.337***

1.127***

0.384

0.172

1.381***

1.628***

0.484**

0.627***

1.437***

0.690***

(0.130)

(0.200)

(0.192)

(0.165)

(0.261)

(0.167)

(0.160)

(0.154)

(0.201)

(0.119)

(0.163)

(0.111)

Trade agreement 

0.258***

0.460***































(0.0608)

(0.107)































Regional bloc







0.221

0.406***

0.294*

0.619***

0.101

-0.858***

0.450**

-0.107

0.0757

-0.441***







(0.175)

(0.157)

(0.158)

(0.124)

(0.102)

(0.112)

(0.191)

(0.100)

(0.124)

(0.102)

Geographical

Distance (logged) 

-1.553***

-0.899***

-1.198***

-0.955***

-1.001***

-0.829***

-1.147***

-1.329***

-1.532***

-1.141***

-1.935***

-0.266***

(0.0365)

(0.0611)

(0.0766)

(0.0722)

(0.0795)

(0.0545)

(0.0467)

(0.0307)

(0.0637)

(0.0521)

(0.0509)

(0.0452)

Share a common border 

0.478***

0.257

-0.0416

-0.120

0.382*

-0.116

0.185

1.419***

0.633***

0.415***

0.205

0.534***

(0.158)

(0.172)

(0.216)

(0.215)

(0.231)

(0.168)

(0.153)

(0.152)

(0.224)

(0.104)

(0.165)

(0.119)

Economic

Ratio of pc income (max/ min) -logged

-0.0191

0.408***

-0.239***

-0.302***

-0.440***

-0.561***

-0.143***

-0.168***

-0.121***

0.191***

-0.270***

-0.0668***

(0.0148)

(0.0483)

(0.0830)

(0.0698)

(0.0682)

(0.0499)

(0.0281)

(0.0127)

(0.0296)

(0.0256)

(0.0221)

(0.0258)

Product of gdps (logged) 

0.646***

0.638***

0.273***

0.187

0.592***

0.165*



















(0.0405)

(0.0732)

(0.0899)

(0.125)

(0.113)

(0.0890)



















Product of populations (logged) 



















0.529

0.751***

1.521***

1.081***

1.534***

3.381***



















(0.322)

(0.0507)

(0.362)

(0.262)

(0.515)

(0.525)

 

Constant

-5.821***

-8.319**

-0.226

3.426

-27.59***

1.386

-2.162

-9.390***

-33.00***

-29.14***

-25.85

-108.2***

 

 

(2.071)

(3.535)

(4.595)

(6.446)

(5.942)

(4.492)

(10.79)

(1.835)

(12.04)

(8.776)

(15.99)

(16.63)

 

Observations

55,386

7,690

11,241

11,961

15,233

12,769

20,193

8,316

8,881

11,284

30,109

8,726

 

Adjusted R-squared

0.783

0.871

0.766

0.699

0.743

0.789

0.737

0.790

0.891

0.898

0.704

0.829

Robust standard errors in parentheses, clustered by country – pair. *** p<0.01, ** p<0.05, * p<0.1

the quality of the demand operates within, more than between sectors. Among the significant cases, the impact is negative in nine out of the eleven significant cases, meaning that countries with similar levels of income tend to interact more. The exceptions are services exports and minutes of outgoing phone calls. In absolute value, the higher value of the coefficients is found for the capital flows, which contradicts the Lucas paradox (Lucas, 1990).



    1. Adding specific variables

Finally, regarding the variables capturing the size of the economies, for trade and capital flows the product of gdps is significant at least at the 5% in all the cases except for FDI outflows and portfolio long term debt. When the product of populations is used for people and information flows also a positive and significant sign is obtained for all the cases except for international students. The non-significant impact in this case can suggest that, while a higher population in the countries of origin will tend to enlarge the magnitude of the flows, it could also be the case, that then, the competitiveness between foreign and national universities increases, reducing the probability to go to study abroad.

In this section we add some more variables to the general specification of the gravity model to capture specific characteristics of each type of flow and to figure out whether the goodness of fit of the model increases when new variables are included. The tables showing the results described here can be found in the appendix.



Trade flows.

As a way to measure cultural similarities in a more complete way than including just a dummy for the countries that share the same official language, an index measuring the similarities between the share of population that follow each religion has been computed. This index complements the dummy of official language and seems to have a positive impact in trade flows.

As mentioned in The Global Competitiveness Report (2012-2013)7, sophisticated businesses are conductive to higher efficiency in the production of goods and services. Business sophistication concerns two elements: the quality of a country’s overall business network and the quality of individual firms’ operations and strategies. Then, the positive coefficient estimate found for the ratio of the business sophistication index between the exporter and the importer captures how countries that score higher in the Business Sophistication sub-index of the Global Competitiveness Report will have an advantage to export more to countries scoring lower in this index. As a complement of this variable it has been included the ratio between the Economic Complexity Index of both countries. This index captures the fact that the goods produced by countries scoring high, are unlikely to be produced by a large number of countries. When this variable has been included in the regression, the disparities in per capita income has been substituted by the ratio between the per capita income of the exporter over the importer, due to the correlation between them. A positive and significant coefficient is obtained for the ratio of the ECI, suggesting that trade flows between country pairs with larger differences in terms of economic complexity will be larger, because of the existence of an a certain competitive advantage. Finally, when the common currency variable is included, a negative coefficient estimate is obtained. As explained previously in this work, the results found in the literature regarding the effect of having the same currency on trade flows have been controversial and depend on the method of estimation and sample used.

In the case of services exports, the dummy related with the coincidence of the same official language has been substituted by another less restrictive variable and that focus on the amount of people that actually speak the same language in both countries, independently on whether this is an official language or not. Then, a dummy that takes the value of 1 when both countries have at least a 9% of the population that speak the same language is included. Non-significant impact is found for the common currency dummy variable and for the ratio of the business sophistication index, but a positive impact of the ratio of the economic complexity index with a positive effect of income disparities. The religion similarity index has also been tested in this case but in any of the different specification we have get a significant impact, so because of parsimony it is not included in the reported specifications. Similarly, the variable describing situations in which both countries share a border has also been drop from this set of regressions.



Capital flows

As it was suggested by La Porta et al. (1997) and confirmed by Aviat and Coeurdacier (2007), Lane and Milesi-Ferreti (2008) or Daude and Fratzscher (2012), similarities between the legal systems of two countries might reduce information costs increasing capital flows. In addition, it might also reflect some kind of cultural proximity, as it was pointed out by Aggarwat et al. (2012). In general, we can expect a positive coefficient estimate for this variable. As an additional measure of cultural proximity, the religion similarity index has been included, obtaining a positive and significant coefficient.

The magnitude of the FDI flows and stocks can also be related with social institutions, defense of the private property and with contract enforcement. In this respect, the rule of law sub-index from the World Governance Indicators reflects perceptions of the extent to which agents have confidence in and abide by the rules of society, and in particular the quality of contract enforcement, property rights, the police, and the courts, as well as the likelihood of crime and violence (higher values meaning more confidence in the rule of society). We have included the disparities between the score obtained by the source country relative to the score of the target country. Then, if both countries have similar scores, the ratio tends to be close to 1; while values of the ratio above 1 represent situations in which the individuals from the source country have more confidence in the rules of society than the individuals from the target country. A negative and significant coefficient is obtained for FDI outward stocks while non-significant for the FDI outflows.

Since, firms located in one country have to accomplish the labor laws and regulations in the target country, it is included the disparities between the score of the source and the target country on the hiring regulation and minimum wage from the component of Labor Market Regulations of the Economic Freedom of the World Annual Report from the Fraser Institute. Countries with higher difficulty of hiring are given lower scores in the sub-index. Then, the ratio between the index in the source and the target country will be above one if it is easier to hire in the source than in the target country and below one if it is more difficult to hire in the source than in the target country. A non- significant impact is obtained in this variable to explain FDI outward stocks, but a negative and significant sign is obtained in flows, suggesting that countries where labor regulations are more restrictive will have incentives to locate in countries where it is easier to hire.

For FDI stocks, a positive sign is obtained in the ratio between the pc income of the source and the target country, reflecting that the magnitude of the FDI stocks increases when the source target has a higher pc income than the target country, although this relation is not significant for the FDI outflows. Finally, an index composed by the differences of the shares of agriculture, industry and services in each economy has been included. This index that captures the differences in the structures of the economic activity of both countries obtains a negative coefficient estimate.

For portfolio equity assets and long term debt, the product of the gdp has been substituted by the product of the market capitalization in US$ as a measure of the economic size. In respect to the results found for portfolio equity assets and long term debt, although the effect of having the same official language has small impact in comparison to the effect found for other kind of flows, other cultural measures as the religion similarity index and the fact that both countries have the same legal origin, have a large impact in this kind of flows. In fact, the common official language dummy variable has been dropped from the estimates on portfolio long term debt, since it was significant just at the 10% and turns to be non-significant when the rest of the cultural variables were included. These factors can reduce the information costs of investing in a different country. Some additional variables that could reinforce this effect have been tested, with a non-significant impact. For example, sharing a common currency minimizes the exchange rate risk and reduces the transaction costs in financial flows; differences in the rule of law or in the political stability are also included. None of these variables have a significant impact in portfolio equity assets, but having the same currency and the differences in the rule of law obtain significant coefficient estimates for portfolio long term debt.

The impact of capital controls in the magnitude of the bilateral portfolio equity assets has been controlled using the sum of the values of the Investment Freedom sub-index of the Index of Economic Freedom from the Heritage Foundation for both countries. The Investment Freedom index evaluates a variety of restrictions that are typically imposed on investment. A given amount of points are deducted from the ideal score of 100 for each of the restrictions found in a country’s investment regime. Higher values of the index are related with more open countries. A positive and significant sign is obtained for portfolio equity assets and long term debt.

People flows

Although the different people flows are driven by very different factors depending on the type and aim of travelling to another country, it seems reasonable to argue that variables linked with cultural similarities and with social networks are expected to play a major role in people flows in comparison to other kind of interactions.

Starting with the results for emigration intensities, the religion similarity index has been included to measure the impact of culture on the geographical pattern of emigration. It has also been included the existence of some ethnic similarities captured by introducing the lagged migration stocks. Since it could be expected that people thinking in where to migrate have more information regarding the countries in which they already have a social network created by past migration flows, past migration intensities between the two countries can affect the more recent migration relations. An additional reason is that, in some cases, the existence of a social network can facilitate the administrative procedure to access to the host country. Then, the bilateral emigration stocks in 1960 (Özden et al, 2011) has been included to capture the effect of potential social linkages in the host country. However, this variable also captures part of the effect of the rest of the variables, reducing the impact of all of them.

Regarding income disparities, the negative coefficient of the variable that we were using in the general specification (max/min) suggests that migration tend to be between countries with similar income levels. It is expected that people decide to migrate when the potential economic gains in the host country surpasses the costs to migrate. Then, there are incentives to move from countries with lower pc income to countries with a higher pc income. Then, a negative effect is expected for the ratio of the pc income of the home-country relative to the one of the host-country. Finally, a negative sign is obtained for the Difference in Economic Structure index. This can be related with a lower probability for the individuals to find an employment in a place with economic structures very different to their home countries. Then, there is a certain trend to migrate to similar countries in terms of economic activities, but where the individual is expected to have a higher income.

Regarding the incoming international tertiary students, cultural distance seems to play a central role with large coefficients in the common official language and in the religion similarity index. In addition, a higher difference between the public expenditure per student of the destination and origin country seems to attract students from the countries in which the public expenditure is lower to the countries with a higher public expenditure. Larger flows are expected between countries with similar income levels. Similar specifications including the ratio of the per capita income of the destination country relative to the origin of the student has also been tested, but not included, since its levels of significance was lower.

The last type of people flow analyzed is the number of tourists’ arrivals. In this case, it is included a dummy variable that is coded as 1 if a visa is required to individuals living in the country of origin to visit the destination country for stays of 30 days or fewer and 0 if not. Visa requirements are not necessarily symmetric. For example, people from Ecuador need a visa to come into Spain, but Spaniards do not need a visa to go to Ecuador. The negative sign obtained for this variable is according to the expected and shows how visa requirements have a negative impact on tourists’ arrivals. When this variable is included, the regional bloc variable loses its significance. This could be explained by the fact that regional blocs use to have agreements on people movements, so the effect that regional blocs could have on tourists’ arrivals is captured by the visa requirement variable. The visa requirement variable is preferred to the regional bloc dummy, since it is a more accurate proxy to the administrative barriers that people have to overcome when travelling abroad. As a measurement of the relative attractiveness of one country, the ratio between the number of World Heritage Sites (WHS) of the destination and the origin is included. This variable accounts for the relative attractiveness of the destination country of the trip in respect to the country of origin of the tourist. Then, if one country has a large number of WHS in respect to another country (large values of the ratio), the former country will be more attractive to be visited by people living in the latter. On the other hand, people living in countries with a large number of WHS (that will tend to obtain lower values of the ratio) could choose to visit any other place within their home countries, instead of travelling abroad. A positive sign is obtained, according to the expected. Regarding economic distance, it seems that the larger the income differences between the destination country and the origin of the tourist, the lower the number of tourists arriving at that country. In general, this can be explained because these income differences can be related with a lower purchasing power of the tourist travelling to countries with higher income per capita. A non-significant impact of the income disparities (max/min) measuring the differences in both countries. Although not reported here, the effect of having a similar religious structure and of having the same currency has been tested. However, non- significant impact was found for none of them.



Information flows

To capture the effect of cultural distance in printed publications, the religion similarity index and a dummy variable for countries with at least 9% of the population speaking the same language are included. In addition, the ICT use of both countries is included, confirming that the higher the use ICT in both countries, the easier will be to have access to the information from cheaper and faster ways than by printed publications. Contrary, for outgoing phone calls, the use of ICT seems to have a positive (impact), although significant just at the 10% and the religion similarity index loses its significance when the dummy for common official language was substituted by the dummy that captures the effect of at least 9% of the population speaking the same language.



Regarding patent activities, the logged disparities of the innovation pillar sub-index from the Global Competitiveness Report has been included. Large values of this ratio describe a situation in which the both countries show very different performance in terms of competitiveness linked with the capacity to innovate and the protection of the intellectual property protection. From the results obtained it seems that the higher the disparities in the competitiveness due to innovation, the less intense the international patent activity between both countries. The similarity in the structure of the population following the same religions has been included to capture similar attitudes to innovation of different cultures, but the effect of this variable is very limited. Regarding the negative effect of the regional bloc variable the effect can be due to the existence of regional offices and especially to the European Patent Office, as mentioned before. The hypothesis here is that the European Patent Office competes with the rest of the national European offices. Then, the regional bloc dummy variable has been divided in two different variables: one that captures the effect of the regional blocs, with the exception of the European Union, and a second one that captures that since a supra-national organism exists in Europe, the level of international patent activity can be underestimated for countries belonging to the European Union (EU27). The results for these two variables confirm this hypothesis, obtaining a positive sign for the dummy variable capturing the effect of regional blocs (except the EU) and a negative impact when both countries belong to the EU. Finally, the number of international patent activity reduces as long as the difference in income per capita between the country where the patent office is located and the country origin of the inventor increases.

Table 2 shows the variables that have been used in the preferred customized specification for each type of flow, the number of observations used in each regression and the adjusted R2 obtained with the standard set of variables and with the customized set of variables. The set of variables and the goodness of fit of the regression with the customized set of variables consider the preferred specifications shadowed in the tables in the appendix. In general, the goodness of fit

Table 2. Summary table. Customized regressions. Variables and goodness of fit.

 

Standard gravity equation

Customized set of variables

Type of flow

N. of obs

Adj. R2

N. of obs

Adj. R2

New variables included

Variables included that were part of the standard gravity

Variables that were part of the standard equation that are excluded

Merchandise exports

55,386

0.783

48,377

0.802

Religion similarity index

Common official language

-

Ratio of business sophistication

Colonial linkage

 

 

Trade Agreement

 

 

Distance

 

 

Share a common border

 

 

Product of gdps

 

 

Ratio of pc income (max/min)

 

Services exports

7,690

0.871

7,670

0.871

Common language (>9%)

Colonial linkage

Common official language

Ratio of business sophistication

Trade Agreement

Share a common border

 

Distance

 

 

Product of gdps

 

 

Ratio of pc income (max/min)

 

FDI outward stocks

11,241

0.766

9,858

0.782

Religion similarity index

Common official language

Regional Bloc

Common legal origin

Colonial linkage

Share a common border

Ratio of the rule of law(source/target)

Distance

Ratio of pc income (max/min)

Ratio of pc income (source/target)

Product of gdps

 

Difference of Economic Structure Index

 

 

FDI outflows

11,961

0.699

10,535

0.703

Religion similarity index

Common official language

Regional Bloc

Common legal origin

Colonial linkage

Share a common border

Ratio of the hiring regulations (source/target)

Distance

Ratio of pc income (max/min)

Difference of Economic Structure Index

Product of gdps

 

Ratio of pc income (source/target)

 

 

Portfolio equity assets

15,233

0.743

14,322

0.759

Religion similarity index

Common official language

Colonial linkage

Common legal origin

Distance

Regional Bloc

Investment Freedom

Ratio of pc income (max/min)

Share a common border

Market capitalization

 

Product of gdps

Portfolio long term debt

12,769

0.789

12,131

0.801

Religion similarity index

Distance

Common official language

Common legal origin

Ratio of pc income (max/min)

Colonial linkage

Investment Freedom

 

Regional Bloc

Common currency

 

Share a common border

Ratio of the rule of law(source/target)

 

Product of gdps

Product of market capitalization

 

 

Table 2. Cont.

 

Standard gravity equation

Customized set of variables

Type of flow

N. of obs

Adj. R2

N. of obs

Adj. R2

New variables included

Variables included that were part of the standard gravity

Variables that were part of the standard equation that are excluded

Emigration intensities

8,316

0.790

4,837

0.810

Religion similarity index

Common official language

Regional Bloc

Emigration 1960

Colonial linkage

Ratio of pc income (max/min)

Ratio pc income (origin / destination)

Distance

 

Difference of Economic Structure Index

Share a common border

 

 

Product of populations

 

Intl. Tertiary students

20,193

0.737

12,612

0.738

Religion similarity index

Common official language

Regional Bloc

Ratio of public expediture per student

Colonial linkage

Share a common border

 (dest / origin)

Distance

 

 

Product of populations

 

 

Ratio of pc income (max/min)

 

Intl. Tourists' arrivals

8,881

0.891

8,881

0.894

Visa Requirements

Common official language

 

 

Colonial linkage

 

 

Regional Bloc

 

 

Distance

 

 

Share a common border

 

 

Product of populations

 

 

Ratio of pc income (max/min)

 

Printed Publications exports

30,109

0.704

27,765

0.716

Religion similarity index

Colonial linkage

Common official language

Common language (>9%)

Distance

Regional Bloc

Product of ICT use

Product of populations

Share a common border

 

Ratio of pc income (max/min)

 

Outgoing phone calls

11,284

0.898

11,284

0.898

Religion similarity index

Colonial linkage

Regional Bloc

Common language (>9%)

Distance

Common official language

 

Share a common border

 

 

Product of populations

 

 

Ratio of pc income (max/min)

 

Patenting activity

8,726

0.829

8,654

0.838

Religion similarity index

Common official language

Regional Bloc

Ratio of innovation pillar

Colonial linkage

Ratio of pc income (max/min)

Regional bloc (except EU)

Distance

 

EU

Share a common border

 

Ratio pc income (patent office / origin)

Product of populations

 

of the model (according to the adjusted R2) slightly increases when a more specific set of variables are included in the regressions, instead of the standard set of variables. However, this increment is certainly lower than will be expected. The highest gains are for: emigration intensities, merchandise exports, FDI outward flows and portfolio equity assets, but none of the customized specifications are able to increase the adjusted R2 in more than 0.02.
1   2   3   4   5   6   7   8   9   10


The database is protected by copyright ©sckool.org 2016
send message

    Main page