Tradability of Output and the Current Account:

An empirical investigation for Europe

Roman Stöllingerǂ

Abstract

In this paper we put forward the hypothesis that increasing specialisation in the production

of non-tradable output has a negative impact on the current account balance. To test this

tradability hypothesis empirically we proceed in two steps. Firstly, we develop a tradability

index which captures specialisation patterns with regard to the tradability of output.

Secondly, we embed the tradability index into an empirical current account model for the

full sample of European countries. We find strong evidence for a positive relationship

between the current account balance and the tradability index in both in the short and in

the long run. The relationship is stronger for emerging economies in Europe than for

developed countries. This finding has an important policy implication: the anxieties about

‘de-industrialisation’ in many parts of Europe seem justified because the resulting loss of

export capacity increases the risk of external imbalances.

Keywords: current account, tradability index, tradable goods, structural change, value

added exports.

JEL Codes: F41, F32, F10, F14

Version: April 2016

Acknowledgement: Research for this paper was financed by the Jubilee Fund of Oesterreichische Nationalbank

(Project No. 15291). Financial support provided by Oesterreichische Nationalbank for this research is gratefully

acknowledged. The author would like to thank Vladimir Gligorov, Mario Holzner, Michael Landesmann, Leon

Podkaminer and Robert Stehrer for very insightful discussions and suggestions. Thanks also goes to the

participants at the 7th FIW Research Conference in Vienna, the Conference on Competitiveness, Capital Flows

and Structural Reforms in Brno and the 18th Göttinger Workshop Internationale Wirtschaftsbeziehungen for

helpful comments and suggestions. The author is particularly indebted to Alexandra Bykova for very valuable

research assistance.

ǂ

1

Tradability of Output and the Current Account:

An empirical investigation for Europe

1 Introduction

Current account imbalances in Europe are a recurring issue that attracts the concern of policy-

makers and the interests of academics alike (e.g. Gaulier and Vicard, 2012; Lane and Pels, 2012;

Ca’Zorzi et al., 2012; Blanchard and Giavazzi, 2002). Among the various potential determinants of the

current account this paper will focus on role of the production structure. We put forward the

hypothesis that the size of the more tradable sectors relative to the sectors producing less tradable

output is an important determinant of a country’s current account position. To test this “tradability

hypothesis” we develop an indicator that we label tradability index (TI)1. This index allows us to

examine the relationship between countries’ sectoral structure and the current account position in a

more systematic and comprehensive way. This is because the TI reflects an economy’s entire

economic structure and in particular how tradable the output that a country produces is. The first

components of the TI are the sector-specific ratios between the global value added exports (Johnson

and Noguera, 2012) and industry-specific global value added. These sector-specific ratios, the

tradability scores, are then weighted by the countries’ value added shares and summed up to arrive

at the TI. The tradability index can be interpreted as the expected export openness given a country’s

production structure. By construction, the TI does not contain a country’s own trade flows and

therefore eliminates an important source of endogeneity that plagues the use of conventional

openness measures in empirical work. We argue that the TI is sufficiently exogenous to be used

directly in a regression framework for testing the tradability hypothesis.

The tradability hypothesis is linked to the debate about the de-industrialisation in Europe. A shrinking

manufacturing sector (in relative terms but in several European countries also in absolute terms) is a

widespread phenomenon in many European countries. Although the structural shift out of the

manufacturing sector has been and continues to be quite uneven across Member States with the

result that manufacturing activities are increasingly concentrated in a ‘Central European

Manufacturing Core’ (Stöllinger and Stehrer, 2014; IMF, 2013), the decline of manufacturing is a

European-wide concern. One reason for this concern is that a relatively smaller manufacturing

sector, which is the main tradable goods producing sector in most European countries, may put

pressure on the current account. In addition, if some areas are exempted from the general trend, as

it is the case with the members of the Central European Manufacturing Core, the resulting structural

divergence will give rise to growing external imbalances within Europe. Countries with shrinking

manufacturing sectors can be expected to end up with current account deficits while countries with

more robust manufacturing sectors will tend to run current account surpluses.

1 Zeugner (2013) uses a measure of country-specific value added in trade to evaluate the tradability of sectors in the context

of calculating unit value costs.

2

The use of the TI for studying the relationship between the production structure and the current

account has three main advantages. First of all, instead of focusing only on the manufacturing sector

or any other sector that is suspected to be particularly important for the external balance, the TI is a

comprehensive measure which reflects a country’s entire economic structure. Therefore the TI is

capable of capturing basically all structural phenomena such as de-industrialisation (respectively the

lack of industrialisation) or booms in the construction sector (as analysed in Gehringer, 2015).

Second, the TI is a more direct way of testing the tradability hypothesis than the value added share of

manufacturing which commonly serves as measure for the tradable sector (for a recent example see

Ehmer, 2014). This is because the TI is constructed using a value added based measure for exports.

The value added exports represent the amount of value added produced by an industry in one

particular country but consumed in another. The ratio between value added exports and value added

is therefore the natural proxy for the tradability of output. The TI ensures that one actually captures

the effect of the production structure, while a the positive relationship between the share of

manufacturing and the current account position may also signal other features of the sector such as

high increasing returns to scale. The TI is also appropriate from a theoretical perspective because the

essential distinction in the context of the tradability-current-account-nexus is not between the

manufacturing sector and the services sector but between the tradable and the non-tradable sector.

Third, the TI avoids applying a dichotomous classification of sectors into tradable and non-tradable

sectors (as for example in De Gregorio et al., 1994) which requires the choice of a discretionary

threshold. In contrast, the TI incorporates the fact that in reality there is no strict distinction between

tradables and non-tradables but rather gradual differences in the tradability of output produced in

the various sectors of the economy.

We use the TI for testing econometrically the effect of the tradability of output on the current

account position in a truly European-wide context. Our sample comprises 46 European countries

which are observed over the period 1995-2014. This means that our sample goes well beyond the

euro area which has attracted a lot of interest in the recent current account literature (see e.g.

Blanchard and Giavazzi, 2002; Gaulier and Vicard, 2012). A major advantage of a European wide

sample is that it results in a sufficiently large number of observations for a cross-country analysis.

The contribution of this paper to the literature is twofold. Firstly, we develop the tradability index

which makes use of the World Input-Output Database (WIOD). Secondly, we establish this tradability

index as a new determinant of the current account in a European context.

The rest of the paper is structured as follows. Section 2 discusses some of the related literature.

Section 3 explains the construction of the tradability index and presents some descriptive results.

Section 4 discusses the empirical model and the data while section 5 contains the estimation results.

Section 6 concludes with some policy implications of our findings.

2 Related Literature

There is no shortage of explanations for the re-occurring current account imbalances. Authors have

related these imbalances to fiscal policy and budgetary discipline (e.g. Schnabl and Wollmershäuser,

2013), productivity shocks (Cova et al., 2009; Fournier and Koske, 2010; Coricelli and Wörgötter,

2012) and structural policies such as wage policies (Kerdrain et al., 2010). Other contributions

highlight the development of financial markets (Mendoza et al., 2009), the degree of financial

3

integration (Blanchard and Giavazzi, 2002) or the existence of safe assets available to people with

savings (Caballero et al. ,2008) as being decisive for current account imbalances.

Another potential explanation is the production structure and changes thereof. The role of an

economy’s economic structure on the current account is, however, far from being universally

accepted. A major reason for this is that, according to the absorption approach (Alexander, 1952),

the current account is ultimately determined by the difference between savings and investment (as

the mirror image of the capital account). The inter-temporal approach to the current account

(Obstfeld and Rogoff, 1996), an extension of the absorption approach, emphasises macroeconomic

fundamentals as the determinants of the current account. In this framework, current account deficits

and surpluses are basically the results of either shocks to the economy or differences in expected

growth rates which result in shifts of consumption between periods. This typically leaves little room

for trade policies, trade openness or specialisation patterns for influencing the current account. With

regard to economic structures, one important aspect in this strand of the literature is the distinction

between a tradable goods producing sector and a non-tradable goods producing sector. (see e.g.

Ostry and Reinhart, 1991; Obstfeld and Rogoff, 1996; Ferrero et al., 2010). This distinction is essential

with regards to the tradability hypothesis. The theoretical predictions regarding the relationship

between the relative development of the tradable and the non-tradable sector and the current

account balance are generally ambiguous though and depend inter alia on the type of the shock and

the elasticity of intertemporal substitution. The most relevant scenario for our empirical investigation

is a wealth-neutral shift in the economic structure from one sector to another where the sectors

differ with regard to their tradability. In the simple case of an endowment economy this can be

modelled as an increase in the supply of the non-tradable good by one unit and a simultaneous

decrease in the supply of the tradable good by one unit. This scenario differs from a positive

productivity shock which in an intertemporal current account model would induce a current account

deficit as consumers will want to consume a part of the additional wealth already now due to

consumption smoothing. This effect is switched off by simultaneously increasing the output of the

non-tradable sector and decreasing the output of the tradable sector. In this scenario, a wealth-

neutral shift towards the non-tradable sector leads to a current account deficit only if the elasticity of

substitution is sufficiently small. The reason is an income effect which arises from a decrease in the

price level (because the price of the non-tradable good declines as the sector expands assuming

constant consumption shares). As a result the effective (i.e. consumption based) interest rate

increases, because postponing consumption until later now not only benefits from the interest

payment but also from the fact that for the same income more of the non-tradable good can be

consumed2. In principle, this income effect tends to cause a current account surplus. But if the

elasticity of substitution is sufficiently small, a strong desire for consumption smoothing will make

households shift forward consumption again, causing, ceteris paribus, a current account deficit.

Hence, the tradability hypothesis put forward in this paper is in line with the standard intertemporal

model of the current account only if the intertemporal elasticity of substitution is sufficiently small.

More recently, researchers have proposed a series of current account models that explicitly

incorporate production and export specialisation. Jin (2012) develops a model which combines inter-

temporal trade with intra-temporal factor-proportions-based trade. This combination introduces a

composition effect into the inter-temporal current account model which arises in addition to (and

2 Given the Cobb-Douglas consumption function, there is no intratemporal substitution effect between the tradable and

the non-tradable good. For further discussion of these issues see for example Harms (2008).

4

works against) the traditional convergence effect3 in current account models. The composition effect

suggests that the current account balance depends (among other things) on the capital intensity of

the output produced in a country. Since in a Heckscher-Ohlin framework richer (more capital-

abundant) countries specialise in the production of capital-intensive goods, investment will be higher

in these countries and they tend to run current account deficits. As a result capital will flow from

capital-poor to capital-rich countries, a fact that is observable for example in US-China relations.

Nedoncelle work with is a trade-cost augmented partial equilibrium version of the model by

Jin (2012). He tests for a joint effect of capital intensity and trade costs on the current account. The

main finding is that a reduction in trade costs worsens the current account of countries specialised in

the production of capital-intensive goods. These results can be directly related to the tradability

hypothesis forwarded in this paper. The composition effect runs counter the tradability hypothesis if

the tradable sector is characterised by higher capital intensity than the services sector. In the

opposite case, the composition effect and the tradability hypothesis would alter the current account

balance in the same direction.

Baraterri (2014) extends the model by Obstfeld and Rogoff (2000) in which all goods are tradable but

the goods are differentiated by the level of associated trade costs. In this model factor endowments

drive specialisation and the resulting comparative advantages which in turn matter for the current

account. Assuming completely specialised countries this model predicts that countries specialising in

services (such as the US) tend to run current account deficits. The reason for this is faster trade

liberalisation in the manufacturing sector than in the services sector (interpreted as the sector that

produces less-tradable output).4 The main inter-temporal mechanism in this model is the

asymmetrical timing of trade policies which affects saving decisions. More precisely, countries that

specialise in the production of manufacturing goods (and hence importing services) will postpone

consumption of imported services in expectation of future reductions in trade costs of services. This

way, households can benefit from lower prices of services in the future.5 Empirically Baraterri (2014)

finds that countries with revealed comparative advantages in services tend to run current account

deficits for the period after 1995.

The empirical literature referring to the inter-temporal approach to the current account has followed

two directions. One direction that uses various estimation techniques to establish evidence in favour

of the baseline model and another direction that attempts to identify short, medium or long-run

determinants of the current account which emerge from a broader class of models (see Bussière et

al., 2006). This paper belongs to the second strand of the literature and as such follows the approach

of Debelle and Faruqee (1996) and Chinn and Prasad (2003). This has the advantage that also other

key determinants can be controlled for. For example, we will incorporate the convergence effect

which is one of the testable predictions that the inter-temporal current account model delivers. It

predicts that countries which are below their steady state equilibrium (e.g. poorer countries or

countries experiencing a positive productivity shock) will run current account deficits because the

rate of return on capital will be above the (world) interest rate (Obstfeld and Rogoff, 1995).

Particularly in a European context, where capital still flows in ‘the right direction’, the (relative)

3 The convergence effect relates to the theoretical prediction that capital flows towards countries the effective capital-

labour ratio is relatively lower. In other words, countries with lower GDP per capita will borrow internationally against future growth, part of which is already consumed today. Therefore low income countries are expected to run current account deficits.

4 The relationship between specialisation in services and current account deficits is found to be robust only after 1995.

5 This result depends on the assumption that the inter-temporal rate of substitution is large (greater than 1).

5

income level is an important control variable. Blanchard and Giavazzi (2002) suggests that

particularly in the EU current account positions are increasingly related to countries’ income per

capita, i.e. that the convergence effect gains importance. This convergence effect or stages of

development effect is also argued for by Debelle and Faruqee (1996). However, their cross-section

estimations for fails to find statistical evidence for the convergence effect in a sample of industrial

countries in the non-linear specification.6 Chinn and Prasad (2003) do find a convergence effect in

their full sample but not for the subsample of developing countries. They also test, for the effect of

trade openness on the current account but fail to find a significant effect in almost all specifications.

There in a large number of additional determinants of the current account that have been tested in

the literature and which we will also include in our empirical model such as the dependency ratio

(e.g. Chinn and Prasad, 2003; Ca’Zorzi et al., Lane and Pels, 2012) and the real exchange rate

(Brissimis et al., 2010) will be discussed in detail in the next section.

Ehmer (2014) and Gehringer (2015) are to recent examples of papers that directly test for the impact

of the particular sector shares on the current account. Based on a sample of euro area countries,

Ehmer (2014) reports a positive effect of the share of the manufacturing sector in GDP on the current

account. Gehringer (2015) finds a worsening effect of an expanding construction sector on the

current account for the peripheral crisis-stricken euro area. In contrast, she her results do not

suggest a significant relationship between the share of the manufacturing sector and the current

accounts in any subset of EU countries.

3 Tradability Index

A natural benchmark for the tradability of goods and services is how much they are actually traded

(De Gregorio et al., 1994)7. De Gregorio et al. (1994) consider a sector as tradable if more than 10

percent of total output is exported. We will depart from this approach by switching from a

dichotomous classification of sectors into either tradable or non-tradable to a continuous measure of

sectors’ tradability. This gradual approach gives due credit to the fact that basically all goods and

increasingly also services are potentially tradable though to a different extent. Hence, in our

empirical model we will replace the dual distinction between tradable sectors and non-tradable

sectors with a continuous ‘tradability score’ specific to each sector.

The tradability score can be obtained from industry level information on value added and either

gross exports or value added exports (Johnson and Noguera, 2012). Both measures of exports will be

used but our preferred metric of the tradability score is based on the value added exports (VAX)

concept. The methodological expositions will therefore outline the construction of the tradability

scores and the tradability index based on value added exports. The first step for the derivation of the

tradability index is therefore the calculation of the value added exports at the industry-country level.

For this we follow the trade in value added concept in Johnson and Noguera (2012) and the

6 The authors explain the lack of significance with potential multicollinearity between income per capita and the capital-

output ratio which is also included in the regression. 7 An alternative approach to capture the tradability of goods (or sectors) is to look at tariffs or trade barriers more

generally. The difficulty is that the magnitude of such trade barriers is hard to identify. While the trade costs for merchandise can be estimated with gravity models (see e.g. Anderson and Wincoop, 2004), this approach is harder to implement for services.

6

expositions in Stehrer (2012). Intuitively, the value added export of a particular industry and country

is the value added created by that country and industry but absorbed in other countries.8

The calculation of the value added exports will be based on the most detailed data that is available to

us, i.e. the input-output table of the World Input-Output Database (WIOD), which comprises 40

countries (plus the rest of the world) and 35 industries. Once the industry and country-specific VAX

have been retrieved they are summed up over all countries to obtain global VAX data. In a next step

we aggregate the VAX of the 35 industries to 14 broad sectors9. The reason for aggregating to 14

sectors is that for the countries in our sample it is hard to find more disaggregated production data.10

Hence the tradability scores are ultimately calculated for 14 broader sectors. Finally, the global VAX

as well as the global value added data are aggregated over the available time period, i.e. 1995-2011.

By summing up over time we make the implicit assumption that the tradability of output does not

change over time. Formally, we arrive at the global tradability score ( ) by calculating the ratio

between the global sector-level value added exports ( ) and the sector-level value added ( ).

where the subscript t indexes time and j indexes countries.

The resulting tradability scores for the 14 sectors are shown in Figure 1. The figure presents both the

tradability score based on value added exports (dark blue bars) and the tradability score based on

gross exports (light blue bars). Both rankings are very intuitive. In the tradability score based on value

added exports (TSvax) mining and manufacturing emerge as the sectors producing by far the most

tradable output with a tradability score of 0.51 and 0.41 respectively. They are followed by the

transport and communication sector and the agricultural sector. At the bottom of the ranking are the

services sectors health and public administration which are both characterised by a very low

tradability score amounting to 0.006 and 0.014 respectively.

8 See Appendix for the methodological details of calculating the value added exports.

9 For the list of the resulting 14 sectors and the corresponding NACE Rev.1 and NACE Rev.2 industry codes see Appendix.

10 Another factor influencing the choice of the 14 sectors is that the calculation of the tradability score requires data

based on both NACE Rev. 1 and NACE Rev. 2 sections. Since the sample period comprises years for which only either NACE Rev. 1 data or NACE Rev. 2 data are available, the only solution is to add up the broader aggregate. For example, while NACE Rev. 1 distinguishes between Agriculture (A) and Fisheries (B), NACE Rev. 2 does not. Hence, the project uses Agriculture and Fisheries as one sector. For details see Appendix.

7

Figure 1: Global tradability scores (TS) of sectors

Note: TSvax = tradability score based on value added exports. TSx = tradability score based on gross exports

Source: WIOD, author’s own calculations.

The tradability scores based on gross exports (TSx) are by definition higher. Nevertheless, the ranking

is very similar though with some important differences. The major difference is that mining and

manufacturing switch position compared to the value added exports based scores. The reason is that

manufacturing is characterised by intensive trade in intermediates which leads to double counting of

in gross export flows. This double counting is corrected for in the value added based approach.

Certainly, if one considers a good that is crossing borders (i.e. is exported) several times as being

more tradable than a good that crosses borders only once, then the gross exports based tradability

scores is preferable. However, we have a clear preference for the tradability scores based on value

added exports (TSvax). This is because we believe that (i) no exported good should enter export term

more than once (i.e. a good cannot be more than 100% traded) and (ii) only domestically value added

should enter export term which is also methodologically correct because they are related to value

added.

The gravity literature emphasises that a particular product may face higher trade costs (which

typically includes various types of costs such as tariffs, non-tariff measures and transportation costs)

in one country than in another. This implies that the tradability scores of sectors may differ across

countries. Despite some potential differences across countries, it seems very likely that if the output

of industry A (say the manufacturing industry) is more tradable than that of industry B (say the health

sector) then it seems acceptable to assume that this is also the case in other countries. To be sure,

we run Spearman rank correlation tests between the country level industry rankings by their

tradability score and the global industry ranking just discussed. This is done for all European

countries covered in the WIOD database (Table 1).

0.00

0.20

0.40

0.60

0.80

1.00

trad

abili

ty s

core

TSvax TSx

8

Table 1: Spearman's rank correlation coefficients of European country's tradability scores with the global tradability score (time invariant)

country Code ρ country code ρ

Austria AT 0.9560 ***

Italy IT 0.9780 ***

Belgium BE 0.9912 ***

Lithuania LT 0.7275 ***

Bulgaria BG 0.9209 ***

Luxembourg LU 0.8330 ***

Cyprus CY 0.9253 ***

Latvia LV 0.9165 ***

Czech Republic CZ 0.8198 ***

Malta MT 0.5341 **

Germany DE 0.9868 ***

Netherlands NL 0.9516 ***

Denmark DK 0.9868 ***

Poland PL 0.9473 ***

Spain ES 0.9560 ***

Portugal PT 0.9692 ***

Estonia EE 0.9341 ***

Romania RO 0.8945 ***

Finland FI 0.9429 ***

Russia RU 0.8857 ***

France FR 0.9736 ***

Slovakia SK 0.8681 ***

United Kingdom UK 0.9121 ***

Slovenia SI 0.9516 ***

Greece EL 0.9516 ***

Sweden SE 0.9516 ***

Hungary HU 0.9121 ***

Turkey TR 0.9077 ***

Ireland IE 0.6659 ***

Note: ρ is the spearman rank correlation coefficient, ***, **, and * indicate p-values being statistically significant at the 1%, 5% and 10%

level respectively.

Source: WIOD, author’s own calculations.

Not surprisingly we find a very strong rank correlation that exceeds 0.90 on average. The reason why

the rank correlation is not perfect is that the differences in the tradability score between the sectors

with very low scores are small. Hence, the ranking is not completely identical across countries. The

second reason is that that commodity-producing countries differ from the rest of the countries in the

sample insofar as the tradability score of their mining sector exceeds that of the manufacturing

sector whereas in all other countries the opposite is true.

Also worth mentioning is the case of Malta which has by far the lowest Spearman rank correlation.

The reason for this is not entirely clear but one reason may be that the availability and hence the

quality of the input-output data for Cyprus and Malta is not as good as for the other European

countries.11 The correlation coefficient is also below average in the case of Ireland which may be

related to the country’s role as a location for foreign headquarters. Given this information, we will

also perform the econometric analysis excluding Malta and Ireland.

The global tradability score is the first element in the calculation of the tradability index. To obtain

the tradability index the tradability scores are weighted with each country’s sectoral value added

shares, i.e. the sector-specific value added (

) over total value added (

). Note that the

value added shares vary over time and over countries. The tradability index is retrieved by summing

up the weigthed tradability scores over all industries i. Formally, the TI of country j in any year t is

calculated as:

11

See Timmer (2012) for details.

9

The advantage of the tradability index is that it reflects the entire composition of production of each

country. This makes it an interesting summary variable for an investigation of the nexus between the

tradability of output and the current account. Note that the tradability index, , is country and time

specific and that the variance over time is coming uniquely from the changes in the respective

country’s sector composition of value added. This is because we chose the tradability scores to be

time invariant.

Figure 2 presents the ranking of the 46 countries in the sample according to the value added exports

based tradability index. There is quite some variation in the index across countries ranging from

0.278 for Azerbaijan to a mere in 0.131 in Cyprus. Next to Azerbaijan mainly other oil and commodity

exporters are found at the top of the ranking. At the bottom of the ranking one finds a number of EU

countries which are Cyprus Greece, Luxembourg and France. Among the EU countries, Romania and

the Czech Republic have the highest tradability indices. The sample average of the tradability index is

0.176 which is about the value found for Lithuania. A country with a TI of 0.176 implies that, given its

production structure, this country is expected to export 17.6% of its value added to other countries.

The Appendix shows that the country ranking for the TI based on gross exports is similar though not

identical to the ranking in Figure 2. The Spearman rank correlation coefficient between the two

alternative TI measures, when based on countries’ year averages, is 0.91512.

Figure 2: Tradability index across countries, value added based (average 1995-2014)

Note: Tradability index based on value added exports.

Source: WIOD, author’s own calculations.

While we emphasised that the TI should be interpreted as the predicted openness of a country given

its production structure, we should equally stress that these predictions are in most cases nowhere

close to the respective country’s actual export openness. This is not surprising given that a country’s

trade openness does not only depend on its economic structure. Rather it depends for example also

strongly on country size as smaller countries tend to be more open economies. A major advantage of

the tradability index is that it is by construction independent of country size. The reason is that it is is

derived using a global sectoral measure for trade openness, i.e. the tradability scores. Moreover, the

tradability index also depends much less on trade policy than the actual trade openness. Indirectly,

12

The Spearman rank correlation between the yearly TI measures is similarly high amounting to 0.909.

sample mean TIvax

0.12

0.16

0.20

0.24

0.28

0.32

0.36

AZ

KZ

NO

R

U

BY

UA

R

O

TR

CZ

SK

IE

RS SI

B

G

HU

P

L M

D

AM

FI

D

E LT

MK

EE

H

R

SE

XK

A

T G

E IT

LV

CH

N

L B

E D

K

BA

U

K

MT

AL

PT ES

IS

M

E FR

LU

EL

C

Y

trad

abili

ty in

dex

(V

AX

bas

ed)

10

countries’ trade policies may show up though due to the resulting specialisation in production. But

the latter is exactly what we intend to capture with our tradability index. Importantly, the tradability

index of a country does not reflect its own exports which would be a problem for the econometric

model. The only thing the tradability index reflects is the share of global value added exports in

output. Hence, in contrast to export openness, which reflects many country characteristics such as

country size, trade policy and the success of the latter, the tradability index is to a large extent

purged from these influences and therefore basically reflects a country’s specialisation in the

production of tradable output.13 To demonstrate how different the measures actually are, Figure 3

shows the correlations between the tradability index and the export openness.

Figure 3: Tradability of Output and actual trade openness across countries, (average 1995-2014)

Note: Graph shows value added exports based tradability and export openness based on gross exports. Luxembourg is not shown for

reasons of scale but it is reflected in the fitted line.

Source: WIOD, World Bank (WDI), author’s own calculations.

There is basically no relationship between the actual openness of a country and its predicted openness given the production structure, i.e. the tradability index. Hence the two measures are very different and for our research question at hand, the tradability index is obviously the more informative indicator. Moreover, as just explained, it has in our view a more straightforward interpretation.

13

One may argue that the tradability index reflects changes in the global stance to trade policy because it includes the global value added exports to vale added ratio. However, since we take the average over time of this ratio, only a long term average of the world’s stance towards ‘globalisation’ all that is included in the tradability score.

AL AM

AT AZ

BA

BE

BG

BY

CH

CY

CZ

DE

DK

EE

EL

ES

FI

FRGE

HR

HU

IE

IS

IT

KZ

LT

LV

MD

ME

MK

MT

NL

NO

PL

PT

RO

RS

RU

SE

SI

SK

TR

UA

UK

XK

0.2

.4.6

.8

expo

rt o

pen

ne

ss

.12 .16 .2 .24 .28tradability index

11

4 Empirical Model and Data

4.1 Empirical Model

In the discussion of the tradability index we emphasised that it was unaffected by each country’s

own exports and to a large extent also to the country’s trade policy. This is a great advantage for the

econometric analysis because this strongly reduces the reverse causality running from the current

account to the economic structure and hence to the tradability account. For this reason we choose a

simple econometric approach that consists of using the tradability index directly as an explanatory

variable in our current account model.

In addition to the tradability index as our main variable of interest the empirical model encompasses

a large number of control variables which have been identified by the literature as determinants of

the current account. In our choice of control variables we draw heavily on the contributions by

Debelle and Faruqee (1996), Chinn and Prasad (2003) and Ca’Zorzi et al. (2012).

The general econometric approach is then to regress the current account balance expressed in per

cent of GDP, , on the tradability index (TI) and a set of control variables. Our first approach is to

investigate the long run effects of the tradability of output on the current account. To this end we

follow Debelle and Faruqee (1996) by taking the average of all variables over the sample period and

run a pure cross-country regression of the form:

(1)

where is a vector of control variables, is the error term and j is the country index. The tradability

hypothesis predicts to be positive.

In addition we also exploit the panel structure of our data which has the advantage that a part of the

potential omitted variable bias can be eliminated by including country and time effects. Since both

the current account series and the tradability index series are for most countries integrated of order

one and we are unable to detect a co-integration relationship, we estimate the panel in first

differences. These results should therefore be interpreted as short term effects. The regression

equation thus becomes

(2)

where the subscript indicates the time index and and denote country and time effects

respectively.

We will present and discuss the cross-section model and the panel model based on estimates using

the value added exports based tradability index. As a robustness check some results for the gross

export based TI along with the traditional approach of using the share of manufacturing as the proxy

for the tradability of output will be shown. Furthermore, we will equally run the models for several

subsets of countries.

12

4.2 Data

Our analysis covers basically the whole of Europe resulting in a sample of 46 countries14. Hence, the

countries are a mix of developed and emerging countries as well as economies in transition. Since

there is a debate to what extent transition has already been accomplished (see e.g. Shleifer and

Treisman, 2014) in a number of former Socialist countries, these countries will, together with Turkey

and the countries of former Yugoslavia, be referred to as ‘emerging Europe’. The sample period

generally stretches from 1995 to 2014 though the fact that the sample comprises the whole of

Europe implies that the sample will be slightly unbalanced as for countries that have gained

independence more recently, such as Montenegro, data is not available back until 1995.15

The primary sources for the current account and the sector-level value added data are the

wiiw Annual Database (wiiw ADB) and Eurostat. An additional data source is the OECD SNA database.

For the countries covered by neither of these databases we turn to the IMF International Financial

Statistics (IMF IFS) and the IMF World Economic Outlook Database (IMF WEO). For the sector-level

value added data an additional data source is the United Nations SNA database. Taken together this

yields in a highly balanced panel. Several countries are covered by two or more databases and the

data series are not always entirely identical.16 As a general rule we use the data source with the most

recent methodology. Apart from this, we use the wiiw ADB as our preferential source of data. If a

country is not covered in the wiiw ADB, data from Eurostat is used. The next choice is the OECD SNA,

followed by the IMF databases and the WDI. The details concerning data sources and data availability

for both the current account data and the TI are found in the Appendix.

As explained in the previous section, the calculation of the tradability index based on value added

exports also requires global input-output data which we get from the World Input-Output Database

(WIOD).

The current account developments of European countries are very heterogeneous. There are a

number of countries which run persistent current account surpluses during the entire sample period

(e.g. Germany, the Netherlands, Sweden) or permanent deficits such as several of the Balkan

countries. Over time, there are a number of countries which by inspection would fit the tradability

hypothesis of this paper. For example, Germany, with its strong manufacturing sector, has been

running a permanent and increasing current account surplus while in France and Britain – both

countries which are reported to have experienced a de-industrialisation – the deficits are persistent

and on the rise.17 Another interesting feature of the data is that the current account positions of

many deficit countries have declined considerable and in some cases abruptly after 2008. A further

observation is that the current account balances appear to be more volatile in commodity exporting

countries (e.g. Ukraine and Russia) than those of the other European countries. We mention this

because due to its high tradability score, the mining sector is strongly influencing the tradability

index of commodity exporters.

Figure 4 gives a first indication of the relationship between the tradability of output and the current

account position in Europe. Clearly a positive relationship is emerging between the two variables

14

Exceptions are Lichtenstein, Monaco, San Marino and the Vatican. See Appendix for the list of countries. 15

See Appendix for details on data availability. 16

Some discrepancies between data sources are explained by reporting in different SNA series (e.g. ESA 1995 vs. ESA 2010) or reporting according to different generations of the Balance of Payment Manuals.

17 For country specific developments of the current account, the reader is referred to the Appendix.

13

(panel a). This positive relationship also prevails when Azerbaijan and Montenegro, two potential

outliers, are excluded from the sample (panel b). This relationship is what we are going to investigate

econometrically, taking into account the effects of other determinants of the current account.

Figure 4: Current account positions and the tradability of output, 1995-2014

(a) Full sample (b) excluding AZ and ME

Note: Tradability index based on value added exports. Right panel excludes Azerbaijan (AZ) and Montenegro (ME).

Source: WIOD, wiiw ADB, Eurostat, IMF IFS, IMF WEO, OECD SNA database, World Bank (WDI), author’s own calculations.

In the following these control variables will be discussed briefly.

Relative per capita income. Stages of development models predict that poorer countries will borrow

from abroad to finance their catch-up process, i.e. they are borrowing against an expected higher

income stream in the future. This implies that poorer countries tend to run current account deficits.

In an inter-temporal framework this is captured by the convergence effect.18 The proxy used will be

GDP per capita expressed relative to the average of the sample (rel gdpcap). GDP per capita will also

enter in quadratic form (rel gdpcap sq) in order to capture the possibility that this income effect

levels off as countries grow richer and approach the average level of GDP per capita. The theoretical

predictions of stages of developments models let us expect a positive sign for the income level and a

negative sign for the quadratic term as the convergence effect typically levels off as a country grows

richer. The data sources are wiiw ADB and Eurostat as well as the United Nations Commission for

Europe (UNECE) for Armenia, Azerbaijan, Georgia and Moldova.

Real GDP growth. Related to the convergence effect it is expected that countries with higher growth

rates (gdp growth) tend to run current account deficits. The results on the GDP growth variable in the

literature concerning GDP growth are mixed though. Nevertheless we are bound to expect a negative

sign as growing incomes will be, at least partly, partly spent on foreign goods leading to an increase

in imports and a deterioration of the current account. The data for real GDP growth is taken from

wiiw ADB, Eurostat, national sources and the World Bank’s WDI database.

18

This requires considering poorer (and less capital-abundant) countries as countries that are further away from their steady state which further implies that they also grow faster.

AL

AL

AL

ALALAL

ALAL

ALAL

ALAL

AL

ALAL

ALAL

ALALAL

AM

AMAMAM

AMAM

AM

AMAM

AMAMAM

AM

AMAM

AM

AMAMAMAM

ATATATATATATAT

ATATATATATATAT

ATATATATATAT

AZ

AZ

AZ

AZ

AZ

AZAZ

AZ

AZAZ

AZ

AZ

AZ

AZ

AZ

AZAZ

AZ

AZ

AZ

BABA

BA

BABA

BABA

BABA

BA

BABA

BABA

BABA

BEBEBEBEBEBEBEBEBEBEBEBEBE

BEBE

BE

BEBEBEBEBG

BG

BG

BG

BGBGBG

BG

BGBG

BG

BG

BGBG

BG

BGBGBG

BG

BY

BYBYBY

BYBYBY

BYBYBY

BY

BY

BYBY

BYBY

BY

BY

BY

BY

CHCHCHCH

CHCH

CHCH

CHCHCHCH

CH

CH

CH

CH

CHCHCH

CY

CYCY

CY

CY

CYCYCY

CYCYCYCY

CY

CY

CY

CY

CYCY

CYCYCZ

CZCZ

CZCZCZCZCZCZ

CZ

CZCZCZ

CZCZCZ

CZCZCZCZ

DEDEDEDEDEDEDE

DEDE

DEDEDEDE

DEDEDEDEDEDEDE

DKDKDKDK

DK DKDKDKDKDKDKDKDK

DKDKDKDKDK

DKDK

EE

EE

EEEE

EEEEEE

EEEEEE

EE

EEEE

EE

EE EEEE

EEEEEE

ELELELELEL

ELELELELELEL

EL

ELEL

ELELEL

ELELEL

ESESESESESESESESES

ESESESESES

ESESES

ESESES

FIFIFIFIFI

FIFIFI

FIFI

FIFIFIFIFIFI

FIFIFIFIFRFRFRFRFRFRFRFRFRFRFRFRFRFRFRFRFRFRFRFR

GE

GEGE

GEGEGEGEGE

GEGE

GE

GE

GEGE

GEGEGE

GE

GE

GE

HRHR

HR

HRHR

HRHR

HRHRHR

HRHRHR

HR

HR

HRHRHRHR

HUHUHU

HUHUHUHUHU

HUHUHUHUHUHU

HUHUHU

HUHUHUIEIE IE

IEIEIE IEIEIEIE

IEIEIEIE

IE

IE IEIE

IEIE

IS

ISIS

IS

IS

IS

IS

IS

IS

IS

IS

IS

IS

ISIS

IS

IS

ITITITITITITITITITITITITIT

ITITITITITITIT

KZKZKZ

KZ

KZ

KZ

KZKZ

KZKZ

KZKZ

KZ

KZ

KZ

KZ

KZ

KZKZKZ

LTLTLT

LTLT

LTLTLT

LTLTLT

LT

LTLT

LTLT

LT

LT

LTLT

LULULULU

LU

LU

LULU

LU

LULULULU

LULULULULULULU

LV

LVLV

LVLV

LV

LVLVLV

LVLV

LVLV

LV

LV

LV

LVLVLVLV

MD

MD

MD

MD

MDMD

MDMD

MD

MD

MD

MD

MDMD

MDMD

MD

MDMDMD

MEME

MEME

ME

ME

ME

ME

ME

ME

MEME

MEME

MKMK

MKMK

MK

MK

MK

MK

MKMK

MK

MK

MK

MKMKMKMK

MTMT

MTMTMT

MT

MT

MT

MTMT

MTMT

MTMT

MTMTMT

MTMTMT

NLNLNL

NLNLNLNLNL

NLNLNLNLNLNL

NLNLNLNLNLNL

NO

NONO

NO

NO

NONONONO NO

NONO

NO

NO

NONONONO

NONO

PL

PLPLPL

PLPL

PLPLPL

PL

PLPLPLPL

PLPLPL

PLPLPL

PT

PTPT

PTPT

PTPTPT

PTPT

PTPTPTPT

PTPT

PT

PT

PT

RORO

RORO

RORORO

RORO

RORORO

RORO

RO RORORO

RORO

RSRSRS

RS

RS

RS

RS

RS

RSRS

RSRS

RSRS

RSRS

RSRS

RURU

RURU

RU

RU

RURURU

RURU

RU

RURURURURU

RURU

RUSESESESESESESESESESESESESESE

SESESESESESE

SISISISISISI

SISISISISISI

SISI

SISISISI

SISI

SK

SKSKSK

SKSK

SKSKSK

SKSKSKSK

SK

SKSKSK

SKSK

SKTRTRTR

TRTR

TR

TRTRTR

TRTRTRTRTR

TR

TR

TR

TRTRTR

UAUAUAUA

UAUAUA

UAUA

UA

UA

UAUA

UA

UAUA

UAUAUA

UAUKUKUKUK

UKUKUKUKUKUKUKUKUKUKUKUK

UKUKUKUK

XKXK

XK

XKXK

XK

XKXK

XK

-.4

-.3

-.2

-.1

0.1

.2.3

.4

cu

rren

t acco

unt in

% o

f G

DP

.12 .16 .2 .24 .28 .32 .36tradability index

AL

AL

AL

ALAL

AL

AL

AL

ALAL

AL

AL

AL

ALAL

AL

AL

ALAL

AL

AM

AMAMAM

AMAM

AM

AMAM

AMAMAM

AM

AM

AM

AM

AMAM

AMAM

ATATATATAT

ATAT

ATATATATATAT

AT

ATATATATATAT

BA

BA

BA

BA

BA

BABA

BABA

BA

BABA

BABA

BA

BA

BEBEBEBEBEBE

BEBE

BEBEBEBEBE

BEBE

BE

BEBEBEBE

BG

BG

BG

BG

BGBG BG

BG

BGBG

BG

BG

BG

BG

BG

BG

BGBG

BG

BY

BY

BYBY

BY

BYBY

BYBY

BY

BY

BY

BY

BY

BY

BY

BY

BY

BY

BY

CHCH

CHCHCH

CH

CHCH

CH

CHCH

CH

CH

CH

CH

CH

CH

CHCH

CY

CYCY

CY

CY

CY

CYCYCY

CYCY

CY

CY

CY

CY

CY

CY

CYCYCY

CZ

CZCZ

CZCZ

CZCZCZCZCZ

CZCZ

CZ

CZCZCZ

CZCZCZ

CZ

DEDEDEDEDEDE

DE

DEDE

DEDEDE

DEDEDE DEDE

DEDEDE

DKDK

DKDK

DK DK

DKDK

DKDKDKDK

DKDK

DK

DKDKDKDK

DK

EE

EE

EE

EE

EEEE

EE

EE

EEEE

EE

EEEE

EE

EEEEEE

EE

EEEE

ELELEL

ELEL

ELELELEL

ELEL

EL

ELEL

ELEL

EL

EL

ELELESESES

ES

ESESES

ESES

ES

ESES

ESES

ESESES

ES

ESES

FIFIFI FIFI

FIFIFI

FIFI

FIFIFI

FIFIFI

FIFIFIFI

FRFRFRFRFR

FRFRFRFRFRFRFRFRFRFRFRFRFRFRFR

GE

GE

GE

GEGEGEGEGE

GE

GE

GE

GE

GE

GE

GE GE

GEGE

GE

GE

HR

HR

HR

HRHR

HRHR

HRHR

HRHRHR

HR

HR

HR

HRHRHR

HR

HUHU HU

HUHU

HU

HUHU

HUHU

HUHUHUHU

HUHUHU

HU

HU

HUIEIE IE

IEIEIE IE IEIEIE

IE

IEIEIE

IE

IE IEIE

IEIE

IS

ISIS

IS

IS

IS

IS

IS

IS

IS

IS

IS

IS

ISIS

IS

IS

ITITITIT

ITITITITITITIT

ITITIT

IT

ITIT

ITITIT

KZ

KZKZ

KZ

KZ

KZ

KZ

KZ

KZ

KZ

KZKZ

KZ

KZ

KZ

KZ

KZ

KZKZ

KZ

LTLTLT

LTLT

LTLTLT

LTLTLT

LT

LT

LT

LT

LT

LT

LT

LT

LT

LULULULU

LU

LU

LULU

LU

LULU

LULU

LULU LULULULULU

LV

LVLV

LVLV

LV

LVLV

LV

LVLV

LVLV

LV

LV

LV

LVLVLVLV

MD

MD

MD

MD

MD

MD

MDMD

MD

MD

MD

MD

MDMD

MDMD

MD

MD

MDMD

MKMK

MKMK

MK

MK

MK

MK

MK

MK

MK

MK

MK

MK MKMK

MK

MTMT

MTMT

MT

MT

MT

MT

MTMT

MTMT

MTMT

MT

MT

MT

MT

MTMT

NLNL

NL

NLNL

NLNLNL

NL

NLNL

NL

NL

NL

NL

NL

NL

NLNLNL

NO

NONO

NO

NO

NONO

NONO NO

NONO

NO

NO

NONO

NONO

NONO

PL

PL

PLPL

PLPL

PLPL PL

PL

PLPL

PLPL

PLPLPL

PL

PLPL

PT

PT

PTPT

PT

PTPT

PTPT

PT

PTPTPT

PT

PTPT

PT

PT

PT

RO

RORO

RO

RORO

RO

RO

RO

RORO

RO

RO

RO

RO RORORO

RORO

RS

RSRS

RS

RS

RS

RS

RS

RSRS

RS

RS

RSRS

RSRS

RSRS

RU RU

RURU

RU

RU

RU

RURU

RURU

RU

RURU

RU RURU

RU

RURUSESESESESESE

SESE

SESESE

SESESE

SE SESESESE

SE

SISISISI

SISI

SISI

SI

SISISI

SISI

SI SI SI

SI

SISI

SK

SKSKSK

SKSK

SKSK

SK

SKSK

SK

SKSK

SKSKSK

SKSK

SK

TRTRTR

TRTR

TR

TR

TR

TRTRTR

TRTRTR

TR

TR

TR

TR

TR

TR

UAUAUA UA

UAUAUA

UA

UA

UA

UA

UA

UA

UA

UAUA

UA

UAUA

UA

UKUKUKUK

UKUKUKUKUKUKUK

UKUKUKUKUK

UK

UKUK

UK

XKXK

XK

XK

XK

XK

XKXK

XK

-.25

-.2

-.15

-.1

-.05

0

.05

.1.1

5.2

.25

cu

rren

t acco

unt in

% o

f G

DP

.12 .16 .2 .24tradability index

14

Dependency ratios. The age structure of the population may affect the current account through the

savings rate. Old and young people do not earn income and therefore cannot save.19 Therefore,

countries with higher dependency ratios can be expected to save less and we expect a negative sign

for the impact of the dependency ratio on the current account position. Following Blanchard and

Giavazzi (2002) the overall dependency ratio, comprising old and young people, is defined as the

ratio of population to the labour force. The dependency ratios (dep ratio) are taken from the World

Bank’s World Development Indicators.

Financial depth. Domestic credit (dom cred) provided by private banks relative to GDP is one of the

proxies for financial depth as suggested by Rajan and Zingales (1998). We will fall back on this

indicator because private credit booms are often associated with housing bubbles or other bubbles in

non-tradable sectors. The presumption here is that in periods of rapidly expanding domestic credit,

funds are predominantly channelled into non-tradables. This will drive up prices of non-tradables,

leading to an appreciation of the real exchange rate and consequently to a worsening of the current

account balance. At the same time, there is also the view that financial deepening may increase the

savings rate (e.g. Chinn and Prasad, 2003) which tends to improve the current account. The effect of

domestic credit on the current account is therefore ambiguous.

Government balance. In the absence of full Ricardian equivalence, public expenditure and taxation

can be expected to affect the current account balance. In case of a budget deficit (surplus), the

government uses up (adds to) domestic savings, which tends to worsen the current account balance

(see for example Debelle and Faruqee, 1996). This is the famous twin-deficit hypothesis. In

overlapping generation models, a positive correlation between the government balance and the

current account arises because budget deficits lead to a shift of income forward from future to

present generations. The government balance (gov bal) enters our empirical model expressed in per

cent of GDP. As with our main variables, the data for the government balance comes from various

sources, mainly the wiiw ADB and Eurostat supplemented with data from national sources in the case

of Armenia, Azerbaijan, Moldova and Georgia and in the latter case also from the CIS database.

Net foreign asset position. The net foreign asset (NFA) position is tightly connected to the current

account position through the fact that the current account balance equals the trade account plus the

(positive or negative) return in the stock of NFA. Therefore the NFA is an important initial condition

for future current account balances because countries with large negative NFAs have to pay interest

(or dividends) on the assets owned by foreigners which contributes negatively to the current account

balance. This suggests a positive relationship between the current account position and the NFA

position. However, from an inter-temporal perspective, the current account should be balanced in

the steady state, implying that there is no relationship between the current account and the NFA

position. However, if countries grow over time, it is possible that current account deficits are

persistent and equal to the nominal growth rate times the NFA position. The NFA position will also

enter the empirical model in relative terms, i.e. expressed as a percentage of GDP (nfa) and we

expect a positive sign20. We use the net international investment positions (IIP) as reported by the

IMF IFS. For the more distant years for which IIP are not reported we used the estimates from the

19

They need to draw down on their wealth or rely on their parents respectively. 20

Ca’Zorzi et al. (2012) argue that the expected effect is ambiguous because in the intertemporal approach to the current account heavily indebted countries (i.e. those with already high NFAs) should not be able to continuously run current account deficits.

15

External Wealth of Nations database assembled by Lane and Milesi-Ferretti and described in Lane

and Milesi-Ferretti (2007).

Euro area membership. Given the strong interest in current account imbalances in the euro area, the

empirical model will include a dummy variable for the members of the euro area (EA MS). This

dummy variable is expected to be positive as it reflects monetary and financial integration.

Oil exporter. This is a dummy variable taking the value one for oil exporting countries and zero

otherwise. We define as oil exporters countries with more than 20% export revenues stemming from

oil exports.21 This threshold defines Azerbaijan, Kazakhstan, Norway and Russia as oil exporters. With

the inclusion of a dummy for oil exporters we attempt to exclude the possibility that we are only able

to detect a positive relationship between the current account balance and the tradability index

because of the high export proceeds of oil exporting countries.

Table 2 provides an overview over the dependent variable and the explanatory variables. Two things

are worth mentioning here. First, Europe on average has run a current account deficit over the

sample period (though this is an unweighted average) and the variability of the current account

position is relatively high with deficits and surpluses reaching 50% (Montenegro in 2008) and 35%

(Azerbaijan in 2008) of GDP respectively. In contrast, the variability of the value added based TI is

relatively low, also compared to the two alternative structural indicators which are the gross exports

based TI and the value added share in manufacturing.

Table 2: Summary statistics

Variable explanantion Obs. Mean Std. Dev. Min Max

Dependent variable:

ca current account balance in % of GDP 904 -0.0265 0.0807 -0.4975 0.3548

Main explanatory variable:

TIvax tradability index, value added exports based 890 0.1751 0.0279 0.1234 0.3566

TIx tradability index, gross exports based 890 0.2310 0.0483 0.1157 0.3802

shmanuf share of sector value added in total value added

890 0.1655 0.0563 0.0424 0.3312

control variables:

rel gdpcap GDP per capita at purchasing power parity (PPP) relative to sample average

893 1.0000 0.6438 0.1103 3.3841

rel gdpcap sq. squares relative GDP per capita at PPP. 893 1.4140 1.7215 0.0122 11.4524

gdp growth real growth rate of GDP 908 0.0334 0.0547 -0.1510 0.8602

gov bal government balance in % of GDP 886 -0.0231 0.0409 -0.3230 0.1870

nfa net foreign asset position in % of GDP 860 -0.3280 0.7111 -6.6215 1.4866

dep ratio dependency ratio 910 2.1742 0.4648 1.6700 6.4600

dom cred domestic credit in % of GDP 854 0.8547 0.6648 -0.0600 3.8000

21

We rely on the export of crude oil (HS Code 2709). The data is obtained from UN Comtrade database accessed via the World Bank’s WITS download tool.

16

We close the discussion of the data with the variance decomposition of the data (Table 3).

Table 3: Decomposition of variance into between and within components

variable between within between (BSS) within (WSS) total (TSS) BSS TSS

variability sum of squares share

ca 0.0624 0.0553 3.2569 2.6221 5.8790 0.5540 0.4460

TIvax 0.0256 0.0116 0.5789 0.1134 0.6923 0.8363 0.1637

TIx 0.0436 0.0225 1.6489 0.4268 2.0758 0.7944 0.2056

shmanuf 0.0518 0.0236 2.3492 0.4718 2.8210 0.8327 0.1673

gov bal 0.0285 0.0300 0.7250 0.7569 1.4819 0.4892 0.5108

nfa 0.6493 0.4370 278.9750 155.4459 434.4210 0.6422 0.3578

dep ratio 0.6011 0.0965 188.3063 8.0446 196.3509 0.9590 0.0410

dom cred 0.5740 0.3471 279.6073 97.3508 376.9581 0.7417 0.2583

gdp growth 0.0212 0.0518 0.4033 2.3151 2.7184 0.1484 0.8516

rel gdpcap 0.6621 0.1093 359.6310 10.1102 369.7411 0.9727 0.0273

What is remarkable is that while the variance of the current account position is almost evenly split

between the between and the within dimension, the variability of almost all explanatory variables is

to a very large extent due to the between variability. This is particularly true for our main

independent variable, the value added exports based TI, for which the between component accounts

for more than 80%. The only exception among the explanatory variables to the predominance of the

between components is the growth rate of GDP whose variability is to 85% explained by the within

component.

5 Results

We report several sets of results. A first set of results (6.1) tries to capture the long run relationship

and is based on the sample average for each country. This leaves us with a sample of 46

observations. The second set (6.2) exploits the panel dimension. The relationship is estimated in first

differences using annual data and the results are to be interpreted as the short term relationship. In

addition, we report some panel results for sub-samples of countries (6.3). Finally, we compare the

results for the value added based TI with those using the gross exports based TI and the share of

manufacturing in value added as the main explanatory variable respectively.

5.1 Cross-section Results

The results for the model in equation (1) are shown in Table 4. We interpret these as the long term

relationship between the explanatory factors and the current account over the entire sample period

which stretches over 20 years.

Specification (1) starts off with a univariate regression of the tradability index (VAX based) on the

current account position. This yields a coefficient of 0.61 which is, however, only significant at the

10% level. In terms of magnitude, this result suggests that a 1 percentage point increase in the

tradability of output (say from 0.18 to 0.19) improves the current account balance in percent of GDP

by 0.61 percentage points. This may be regarded as a relatively large effect but it should be

17

considered that a one percentage point change in the tradability index is actually associated with a

relatively large change in the production structure. Assume that the tradability index increases from

0.18 – which is approximately the sample average – to 0.19 due to shift of production towards

manufacturing. We can now ask by how much the manufacturing sector must expand in order to

arrive at this 1 percentage point change in the TI. The tradability score of the manufacturing sector is

about 0.4. If we assume that the initial share of the manufacturing sector in GDP was 20% and that

resources were shifted out of a sector with an tradability score of about 0.18 (i.e. agriculture), the

share of the manufacturing sector needs to increases by approximately 5 percentage points

(≈(0.40-0.19) x ((0.20+0.25)/2), i.e. to 25% of GDP, in order to accomplish a 1 percentage point

increase in the tradability index. In any case, the positive coefficient of the tradability index is a first

indication that the tradability hypothesis holds.

Table 4: Long term cross-section regression

Dependent variable: Current Account Position in % of GDP

(1) (2) (3) (4) (5) (6)

TIvax 0.6124* 1.4642*** 0.8844*** 0.7380*** 0.5481** 0.5447**

(0.3302) (0.2370) (0.1962) (0.1662) (0.2157) (0.2105)

gdp growth -0.9815*** -0.8700*** -0.8759*** -0.9244*** -0.9503***

(0.3310) (0.2418) (0.2069) (0.2267) (0.2148)

rel gdpcap 0.0648*** 0.0558*** 0.0497*** 0.0494** 0.0533**

(0.0216) (0.0153) (0.0174) (0.0196) (0.0215)

rel gdpcap sq. 0.0027 -0.0023 -0.0020 -0.0017 -0.0015

(0.0055) (0.0037) (0.0038) (0.0045) (0.0044)

govbal 0.1756 0.2806* 0.1854 0.1210

(0.1657) (0.1472) (0.1583) (0.1785)

nfa 0.0488*** 0.0527*** 0.0532*** 0.0535***

(0.0085) (0.0082) (0.0088) (0.0084)

dep ratio -0.0135*** -0.0133*** -0.0126***

(0.0038) (0.0037) (0.0036)

domcred -0.0033 -0.0061 -0.0061

(0.0084) (0.0095) (0.0096)

oil 0.0269 0.0285

(0.0216) (0.0215)

EA MS -0.0123

(0.0145)

Observations 46 46 46 46 46 46

R-squared 0.0623 0.7592 0.8957 0.9124 0.9177 0.9198

R-squared adj. 0.0409 0.736 0.880 0.893 0.897 0.897

F-test 3.441 42.15 102.7 106.8 90.80 76.64

Note: TIvax = tradability index based on value added exports. EA MS=euro area members. Robust standard errors in parentheses. ***, **, and * indicate statistical significant at the 1%, 5% and 10% level respectively. All regressions are based on actual sample averages used in the regression. All regressions include a constant.

In specification (2) the growth rate and GDP per capita are added as control variables. Note first, that

this addition improves greatly the explanatory power of the regression with the (adjusted) coefficient

of determination going up to 0.76. Specification (2) shows that by controlling for the growth rate and

GDP per capita, both the size and the magnitude of the tradability index goes up considerably. The

coefficient of the growth rate itself is negative (-0.98) confirming the hypothesis that higher growth

18

rates tend to worsen the current account balance because of increased import demand. This may be

due to both increased imports of consumption goods (as household shift forward consumption with

expected higher incomes in future) and of investment goods (needed to sustain the growth process).

The fact that the inclusion of the growth rate and the GDP per capita improves the significance of the

tradability index is very reassuring. It is reassuring because these two determinants should capture

the wealth effect described above so that the tradability index really picks up the structural effect,

i.e. the impact of pure changes in the composition of output on the current account. While the

coefficient of the growth rate is statistically significant at the 1% level in our model, it does not

belong to the most likely determinants of the current account according to the Bayesian model

selection approach of Ca’Zorzi et al. (2012).

Specification (2) also adds the relative GDP per capita, including a quadratic term. The estimated

coefficient is highly significant and has the expected positive sign. This is as expected by the stages of

development model (e.g. Calderon et al., 2002) as it suggests that richer countries are closer to their

steady state, more capital intensive and are therefore providers of capital to poorer countries. This

result is in line with those of Debelle and Faruqee (1996) and Chinn and Prasad (2003) who also find a

positive effect of relative income in their cross-section results. In terms of magnitude of the

coefficients, the 0.065 we find are in a similar range as those in Chinn and Prasad (2003) who find a

coefficient of 0.11 in their full sample. In general, the positive coefficient of the GDP per capita

variable that we obtain confirms the conclusion in Blanchard and Giavazzi (2002) and Lane and Pels

(2012) that in Europe capital continues to flow from rich to poor countries (though their sample is

limited to the euro area and the EU plus EFTA respectively). What we cannot establish, however, is

that this positive effect of income per capita diminishes as countries grow richer because the

quadratic term turns out to be not statistically significant.

For the government balance (specification 3) we do not obtain a significant coefficient reflecting the

outcome in Debelle and Faruqee (1996) for their global sample. Hence, there is no evidence for the

twin deficit hypothesis in a European context in the long run. On the contrary, Chinn and Prasad

(2003) do find that government deficits go together with current account deficits in both their full

sample results and the results for industrial countries. The government balance is also one of the

explanatory variables that are quite commonly included the models resulting from the selection

criteria in Ca’Zorzi et al. (2012).

Another important control variable is the net foreign asset position whose estimated coefficient is

highly statistically significant and also quite large in magnitude amounting to 0.05. This implies that

countries with positive NFA positions tend to have higher current account surpluses respectively

more moderate deficits. The rationale behind this is that countries with accumulated positive foreign

assets earn interest on these assets. This result would also suggest that it is quite possible that

countries are able to persistently run (more or less moderate) current account deficits. This is also

what can be observed in the country-specific time series of the current account balance as there is

quite a large number of countries that run either persistent surpluses or persistent deficits.

The next controls we add are the dependency ratio and the domestic credit (specification 4). For the

former we find a negative coefficient (-0.014) that is significant at the 1 percent level. This reflects

the fact that children and retired people do not save – or at least have lower savings rates than the

active part of the population. In terms of magnitudes, however, the effect is not very large. In fact it

19

is much smaller than those found for example by Ca’Zorzi et al. (2012) who distinguish between the

dependency ratio of old people and that of young people.

Less relevant according to our cross-section model is the amount of domestic credit. This mirrors the

result in Chinn and Prasad (2003) for their industrial countries sample though they find an effect in

their analysis for their full sample of countries.

Finally, we include dummy variables for oil exporting countries and the members of the the euro

area. While the euro area dummy turns out to have little impact on the results, but effect of the oil

exporter dummy is noteworthy. While not statistically significant itself, the oil exporter dummy

reduces the size of the coefficient of the TI considerably – from 0.74 in specification (4) to 0.55 in

specification (5). This is not entirely surprising given that a high endowment with oil resources is an

important structural feature of an economy. This effect is so to speak “deducted” from the structural

effect that the tradability index is capturing. What is important to see, is that the tradability index is

also capable of explaining some of the variation in the current account apart from the particularities

of oil exporting countries. In other words, we can basically rule out that the positive relationship

between the tradability index and the current account is an artefact caused by oil exporters.

To summarise, we find that – in line with the tradability hypothesis – the tradability index emerges as

a determinant of the current account which is a result that holds throughout all specifications.

5.2 Panel Results

We now turn to the short term model of the current account in equation (2) which exploits the panel

structure of our data. The results are summarised in Table 5.

We start with “pooled” estimations, i.e. the panel is estimated without country effects. Time fixed

effects, however, are included (specifications 1 & 2). One difference to the long term model is that

we omit the growth rate as explanatory variable because the model is estimated in differences and

so the changes in the income per capita already captures the potential impact of economic growth

on the current account.

The most interesting result for us is of course the positive and statistically significant coefficient of

the tradability index. In the model including control variables, this coefficient amounts to 3.06 which

is more than 3 times larger than what was found for the cross-section regression. The reason for this

huge coefficient is that the economic structure is a slow moving variable so that short term changes

tend to be very small. Therefore a one unit change in the TI, which corresponds to a one percentage

point change, would be an extremely large movement.

The result concerning both the statistical significance and the large magnitude of the TI coefficient

holds throughout specifications with different estimation methods. These include random effects,

fixed effects and generalized method of moments (GMM) using the Arellano-Bond estimator

(Arellano and Bond, 1991). The GMM approach has the advantage that it allows controlling for

dependencies between the current and lagged values of the dependent variable which may be

relevant in the case of current account data as well.

For deciding the appropriate estimator we perform a Hausman test which does not reject the use of

the random effects model at conventional levels of significance. However, with a p-value of 0.13 the

20

pass of the test for consistency is not overwhelming and so we also report the fixed effects model

and the GMM specification in Table 5.

Table 5: Panel regression in first differences

Dependent variable: ΔCurrent Account Position in % of GDP

(1) (2) (3) (4) (5) (6) (7) (8)

(POOLED) (POOLED+) (RE) (RE+) (FE) (FE+) (XTA) (XTA+)

ΔTIvax 2.8548*** 2.9541*** 2.8548*** 2.9541*** 2.8910*** 2.9329*** 3.0567*** 3.1521***

(0.4532) (0.5299) (0.6678) (0.7880) (0.6590) (0.7650) (0.2548) (0.2908)

Δrel gdpcap -0.1512 -0.1512 -0.2350 -0.1564

(0.0992) (0.1213) (0.1422) (0.1157)

Δrel gdpcap sq 0.0340 0.0340 0.0509 0.0354

(0.0263) (0.0329) (0.0349) (0.0316)

Δgov bal -0.0183 -0.0183 -0.0071 -0.0323

(0.0918) (0.0898) (0.0880) (0.0681)

Δnfa (t-1) -0.0184*** -0.0184*** -0.0179*** -0.0232***

(0.0059) (0.0059) (0.0058) (0.0071)

Δdep ratio 0.0263 0.0263 0.0419 0.0464

(0.0286) (0.0312) (0.0423) (0.0418)

Δdom cred -0.0156** -0.0156** -0.0157** -0.0158*

(0.0064) (0.0075) (0.0077) (0.0084)

Δca (t-1) -0.0482 -0.0528

(0.0335) (0.0370)

country fixed effects no no no no yes yes yes yes

time fixed effects yes yes yes yes yes yes yes yes

Observations 841 711 841 711 841 711 753 653

R-squared 0.2481 0.2833 0.2500 0.2787

R-squared adj. 0.231 0.258 0.233 0.253

rho 0 0 0.1213 0.0298

F test 6.238 6.776 7.382 26.14

Number of id 46 46 46 46 46 46

Note: POOLED= pooled panel regression; RE= random effects regression; FE=fixed effects regression; XTA=Arellano-Bond fixed effects estimator. TIvax = tradability index based on value added exports. Robust standard errors in parentheses. ***, **, and * indicate statistical significant at the 1%, 5% and 10% level respectively. All regressions include a constant and time fixed effects.

When running a joint F-test for statistical significance of the country fixed effects we find that the

country fixed effects can as well be omitted. Hence, it is equally appropriate to interpret the results

of the “pooled” model without country fixed effects but including time effects. Likewise, in the

random effects model the variance due to the country dimension is 0 indicating the appropriateness

of the pooled model as well. In addition to the econometric arguments in favour of a pooled panel

regression approach Chinn and Prasad (2003) as well as Ca’Zorzi et al. (2012) also support the pooled

specification because the explanatory variables capture more of the cross-country variation in the

data much of which would potentially be captured by the country effects. Given that the majority of

the variability in our data comes from the between dimension (i.e. cross-section) we are also

sympathetic with this argument. In the end, however, the choice of estimator has no effect on our

results as there are only very minor differences in the coefficients.

Sticking to specification (2) we find that changes in the TI are a significant determinant of the current

account also in the short term. The estimated coefficient of 2.95 appears to be very large but also

21

here it has to be borne in mind that shifts in the composition of output from year to year are rather

small so that a 1 percentage point change in the TI must be considered to be very large.

In general, however, the fit of our short term model for the current account seems to be less good

than the long term version. Apart from the tradability index, only changes in the domestic credit and

the net foreign asset position are suggested to affect the current account. Regarding the latter, the

negative coefficient of the net foreign asset position is a bit of a puzzle. Since switches from being a

current account surplus country to being a deficit country or vice verse are relatively rare in our data,

one would rather expect that changes in the foreign asset position of the previous year should be

positively associated with changes in the current account balance. So at this stage we are unable to

offer a plausible explanation for this result.

In contrast, the negative sign obtained for the provision of domestic credit is as expected. It reflects

the supposition that times of rapid expansion of domestic credit are associated with booms in the

housing sector, the financial sector or other non-tradable sectors which will drive up the real

exchange rate and thus lead to a deterioration of the current account.

Qualitatively, all results are confirmed by the Arellano-Bond estimator (Arellano and Bond, 1991).

The GMM approach allows controlling for the correlation of the current account balances of a

country over time. When performing the estimation in first differences, however, this is less of an

issue as indicated by the coefficient of the lagged current account position which is not statistically

significant.

Overall we find a very robust relationship between the tradability of an economy’s output and the

current account balance also for the sort run although in general our short term model of the current

account seems to be less convincing.

5.3 Results by country groups

We no want to find out whether our results hold for individual groups of countries. Therefore we re-

estimate our model separately for developed and emerging European countries. Another sub-group

will be the EU respectively the euro area members. A final group we test are the Central, Eastern and

South Eastern European (CESEE) countries which comprise basically the Balkan countries, the former

Communist countries and Turkey. This sensitivity analysis is done for both the long term and the

short term relationship where for the latter we report the results of the pooled model.

As shown in Table 6. the results – and in particular the result regarding the tradability index – also

hold for most sub-samples. The only exceptions are the euro area members if taken by themselves

and the emerging markets. To some extent this lack of statistical significance may also be due to the

fact that the number of observations gets relatively small if we only take sub-samples of countries.

What is worth mentioning, however, is that a part of the impact of the TI on the current account is

indeed due to oil resources. This is less of an issue for the developed countries in Europe but it

matters for Emerging Europe. In fact, if oil is controlled for, the impact of the tradability index on the

current account remains positive but it is not statistically significant anymore. As mentioned before,

however, we believe that oil the production of oil is also a part of the composition of output that we

want to capture with our tradability index.

22

Table 6: Long term cross-section regression, various subsamples

Dependent variable: Current Account Position in % of GDP

(1) (1') (2) (2') (3) (3') (4) (5) (6) (7) (8)

(full sample) (developed) (emerging) (EU) (EA) (CESEE) (ex AZ & ME) (ex IE&LT&MT)

TIvax 0.7380*** 0.5481** 0.4910* 0.7640** 0.6784* 0.3521 0.6117** 0.3199 0.7491*** 0.7834*** 0.7922***

(0.1662) (0.2157) (0.2606) (0.2795) (0.3263) (0.3881) (0.2652) (0.3153) (0.1729) (0.2259) (0.1854)

gdp growth -0.8759*** -0.9244*** -0.8484** -1.0141*** -0.9302 -1.1571** -0.6315** -0.4372 -0.9259*** -0.8091*** -0.8378***

(0.2069) (0.2267) (0.3095) (0.2848) (0.5591) (0.4263) (0.3006) (0.5045) (0.2763) (0.2315) (0.2293)

rel gdpcap 0.0497*** 0.0494** 0.0652** 0.0701*** 0.0933 0.0351 0.1015*** 0.1124* 0.1063** 0.0507*** 0.0473**

(0.0174) (0.0196) (0.0230) (0.0199) (0.2326) (0.1527) (0.0304) (0.0525) (0.0447) (0.0174) (0.0181)

rel gdpcap sq. -0.0020 -0.0017 -0.0055 -0.0068 -0.0197 0.0304 -0.0140* -0.0164 -0.0597 -0.0018 -0.0017

(0.0038) (0.0045) (0.0052) (0.0048) (0.2455) (0.1560) (0.0072) (0.0121) (0.0370) (0.0038) (0.0038)

gov bal 0.2806* 0.1854 0.4258*** 0.6447*** 0.1434 -0.1662 0.6051*** 0.3054* 0.1716 0.3252** 0.2472

(0.1472) (0.1583) (0.1464) (0.1682) (0.5592) (0.3948) (0.1955) (0.1661) (0.4453) (0.1554) (0.1518)

nfa (t-1) 0.0527*** 0.0532*** 0.0386*** 0.0362*** 0.0661*** 0.0656*** 0.0279* 0.0582*** 0.0670*** 0.0426*** 0.0533***

(0.0082) (0.0088) (0.0044) (0.0036) (0.0117) (0.0079) (0.0161) (0.0164) (0.0079) (0.0059) (0.0081)

dep ratio -0.0135*** -0.0133*** -0.0088*** -0.0081** -0.0285 -0.0149 0.0037 -0.0623 -0.0132*** -0.0115*** -0.0122***

(0.0038) (0.0037) (0.0026) (0.0029) (0.0409) (0.0384) (0.0290) (0.0382) (0.0035) (0.0032) (0.0037)

dom cred -0.0033 -0.0061 -0.0039 -0.0038 -0.0143 -0.0315 -0.0106 -0.0131 0.0092 -0.0018 0.0027

(0.0084) (0.0095) (0.0129) (0.0118) (0.0699) (0.0453) (0.0092) (0.0091) (0.0598) (0.0089) (0.0094)

oil 0.0269 -0.0589** 0.0525**

Observations 46 46 27 27 19 19 28 19 27 44 43

R-squared 0.9124 0.9177 0.9314 0.9438 0.9075 0.9446 0.8913 0.9549 0.8910 0.8908 0.9161

R-squared-adj. 0.893 0.897 0.901 0.914 0.834 0.889 0.845 0.919 0.843 0.866 0.896

F-test 106.8 90.80 129.2 . 48.20 588.0 150.4 342.6 100.7 112.8 128.1

Note: TIvax = tradability index based on value added exports. Robust standard errors in parentheses. ***, **, and * indicate statistical significant at the 1%, 5% and 10% level respectively. All regressions are based on actual sample averages used in the regression. All regressions include a constant.

23

I few neglect the oil issue, the results by country groups suggest that the tradability of output is more

important for the external balance for emerging countries than for developed countries. The reason

for this may be that components of the current account which are not (or only indirectly) linked to

the tradability of output such as payments of factor incomes play a larger role in developed

countries. Another reason may be that industrial countries only require a smaller industrial base

which is increasingly interlinked with and performs an important carrier function for services by

absorbing a large amount of services which are exported only indirectly via manufactures.

A large coefficient for the TI is found for the CESEE countries. For this country group the oil issue also

plays a role as with Russia, Azerbaijan and Kazakhstan it includes three oil exports. Another

contributing factor to this result for the CESEE region are the countries of the Southern Balkan which

tend to run large current account deficits while having great difficulties with building up

manufacturing capacities.

Interestingly, the government balance turns out to be statistically significant for the industrialised

European countries and the EU member states. Hence, for these countries, the twin deficit

hypothesis seems to have some relevance. Most other control variables are very robust across the

various sub-samples. This is particularly true for the net foreign assets and the relative GDP per

capita and the growth rate.

To conclude, we take a look at specifications (7) and (8) which relate to the full sample but excludes

outliers regarding the (descriptive) current account-tradability nexus (Azerbaijan and Montenegro)

and the countries which are outliers regarding the tradability scores of sectors (Ireland, Lithuania,

Malta) respectively. The particularities of these countries do not seem to drive the main results.

Next we turn to the short term regressions distinguishing between the same sub-groups (Table 7). By

and large, the short term analysis by country group confirms the result from the full sample. The

tradability index is again positive and statistically significant in all subsamples except for the EU. The

pattern that the tradability of output is more important for Emerging Europe than for the developed

countries also holds true in the short term. A new finding is that the growth effect, i.e. changes in the

GDP per capita, is relevant for the current account mainly for the developed countries in Europe.

Moreover, the negative effect of the expansion of domestic credit on the current account balance

which is found in the full sample is mainly due to the experiences in the CESEE region. Finally, it is

worth mentioning that in the short run model at least the magnitude of the coefficient to the TI is

sensitive to the outliers in the sample, i.e. Azerbaijan and Montenegro (specification 7).

24

Table 7: Panel regression (pooled) in first differences, various subsamples

Dependent variable: ΔCurrent Account Position in % of GDP

(1) (2) (3) (4) (5) (6) (7) (8)

(full sample) (developed) (emerging) (EU) (EA) (CESEE) (ex AZ ME) (ex IE LT MT)

ΔTivax 2.9541*** 1.0635* 3.5405*** 0.6242 1.2452* 3.2916*** 1.7055*** 3.1450***

(0.5299) (0.5886) (0.5836) (0.4940) (0.6380) (0.5594) (0.3848) (0.5324)

Δrel gdpcap -0.1512 -0.4422*** 0.2194 -0.4078*** -0.2788*** -0.0063 -0.2325*** -0.1227

(0.0992) (0.1062) (0.3073) (0.0906) (0.1027) (0.2578) (0.0890) (0.0987)

Δrel gdpcap sq. 0.0340 0.1147*** -0.1150 0.0977*** 0.0548** -0.0275 0.0619*** 0.0260

(0.0263) (0.0278) (0.2054) (0.0238) (0.0236) (0.1417) (0.0232) (0.0261)

Δgov bal -0.0183 0.0417 -0.1052 -0.0517 -0.0894* -0.1336 0.0430 -0.0249

(0.0918) (0.1038) (0.1461) (0.0546) (0.0495) (0.0922) (0.0869) (0.1234)

Δnfa (t-1) -0.0184*** -0.0215*** -0.0215 -0.0092 0.0047 -0.0329** -0.0218*** -0.0186***

(0.0059) (0.0077) (0.0132) (0.0108) (0.0108) (0.0143) (0.0060) (0.0057)

Δdep ratio 0.0263 -0.0187 0.1513* 0.0461 -0.0611 0.0384 0.0029 0.0353

(0.0286) (0.0240) (0.0862) (0.0461) (0.0705) (0.0324) (0.0264) (0.0301)

Δdom cred -0.0156** -0.0173*** -0.0176 -0.0082 -0.0019 -0.0507** -0.0138** -0.0210*

(0.0064) (0.0053) (0.0189) (0.0052) (0.0079) (0.0219) (0.0060) (0.0110)

Observations 711 415 296 444 187 424 690 674

R-squared 0.2833 0.2508 0.3906 0.2016 0.2825 0.3661 0.2065 0.3111

R-squared-adj. 0.258 0.205 0.337 0.156 0.196 0.328 0.178 0.286

F-test 6.776 4.581 4.337 3.689 . 4.938 6.627 7.065

Note: TIvax = tradability index based on value added exports. Robust standard errors in parentheses. ***, **, and * indicate statistical significant at the 1%, 5% and 10% level respectively. All regressions include a constant and time fixed effects.

25

5.4 Comparing structural indicators

Another interesting comparison is between the one between the various measures of tradability of

output and their effects on the current account. For this purpose we compare the long run effects on

the current account of the value added exports based tradability index (TIvax) with that of the gross

exports based TI (TIx)and also the simple share of manufacturing in value added.

Table 8 demonstrates that all three structural indicators deliver similar qualitative results. Also, the

control variables show very robust patterns across the three structural indicators.

Table 8: Cross section regressions, different tradability measures

Dependent variable: Current Account Position in % of GDP

TIvax TIx Manufacturing share

(1) (2) (3) (1’) (2’) (3’) (1’) (2’) (3’)

TIvax 0.6124* 0.7380*** 0.5481**

(0.3302) (0.1662) (0.2157)

TIx 0.4080* 0.2506** 0.2355**

(0.2343) (0.1098) (0.0951)

sh manuf 0.2038 0.0735 0.1954**

(0.2092) (0.0802) (0.0841)

gdp growth -0.8759*** -0.9244*** -0.5831** -0.8295*** -0.4634 -0.7312***

(0.2069) (0.2267) (0.2236) (0.2210) (0.2816) (0.2280)

rel gdpcap 0.0497*** 0.0494** 0.0462** 0.0380* 0.0613** 0.0336

(0.0174) (0.0196) (0.0223) (0.0219) (0.0265) (0.0243)

rel gdpcap sq. -0.0020 -0.0017 -0.0016 0.0017 -0.0064 0.0028

(0.0038) (0.0045) (0.0052) (0.0051) (0.0061) (0.0057)

gov bal 0.2806* 0.1854 0.4203*** 0.1870 0.4626*** 0.1992

(0.1472) (0.1583) (0.1366) (0.1509) (0.1459) (0.1490)

nfa 0.0527*** 0.0532*** 0.0545*** 0.0525*** 0.0584*** 0.0527***

(0.0082) (0.0088) (0.0085) (0.0087) (0.0093) (0.0088)

dep ratio -0.0135*** -0.0133*** -0.0157*** -0.0126*** -0.0191*** -0.0128***

(0.0038) (0.0037) (0.0042) (0.0036) (0.0048) (0.0036)

dom cred -0.0033 -0.0061 -0.0055 -0.0038 -0.0148 -0.0031

(0.0084) (0.0095) (0.0100) (0.0102) (0.0106) (0.0107)

oil 0.0269 0.0469*** 0.0625***

(0.0216) (0.0140) (0.0121)

Observations 46 46 46 46 46 46 46 46 46

R-squared 0.0623 0.9124 0.9177 0.0803 0.8959 0.9185 0.0283 0.8827 0.9176

R-squared adj. 0.0409 0.893 0.897 0.0594 0.873 0.898 0.00622 0.857 0.897

F-test 3.441 106.8 90.80 3.033 103.0 89.94 0.950 83.94 86.13

Note: TIvax = tradability index based on value added exports. Tx= tradability index based on gross exports. shmanuf = value added share of manufacturing. Robust standard errors in parentheses. ***, **, and * indicate statistical significant at the 1%, 5% and 10% level respectively. All regressions are based on actual sample averages used in the regression. All regressions include a constant.

When comparing the magnitude of the coefficients of the tradability indices on the one hand and the

share of manufacturing on the other hand one has to bear in mind, that the variation in the former

tends to be smaller because it reflects the whole structure of the economy. As mentioned before a

one percentage point increase in the manufacturing share would trigger a much smaller change in

the TI.

26

Another interesting aspect is the fact that dummy variable for oil exporters is not significant in the

specification with the value added exports based TI. One interpretation for this result is that the

latter captures better the impact of structural features on the current account (such as countries

being strong oil exports) than the other two indicators. Certainly, one could also make the argument

that it is preferable is isolate the particularities of oil production. In any case, what Table 8 shows is

that the positive association between the tradability of output does not depend on the particular

choice of the index, whether it is value added exports or gross exports based. The use of the share of

the manufacturing sector in GDP also yields similar results but we would strongly argue that the

tradability index we have developed is a much more comprehensive and hence more informative

indicator for the tradability-current account nexus.

6 Conclusions

We claim to have identified a new important empirical determinant for the current account. This

determinant is the tradability index which summarises the tradability of the output a country

produces. Importantly, this indicator is not constructed using the respective country’s exports.

Rather it is based on global value added exports to value added ratios weighted by the respective

country’s sectoral structure. As such the tradability index is the predicted export openness given a

country’s economic structure.

We find robust evidence for the tradability hypothesis in a sample of 46 European countries over the

period 1995-2014 in both the short and the long run. The result holds irrespective of whether a

cross-section (based on averages) or a panel is estimated. The results are robust across various

European country groups such as Emerging Europe, the European industrialised countries or the

CESEE regions.

This result is highly relevant for economic policy and the debate about external imbalances and the

phenomenon of de-industrialisation in Europe. The fact that, on the other hand the tradability of

output is a key determinant of the current account and on the one hand many European countries

are experiencing a structural shift towards services and hence relatively less tradable output

increases the risk of external imbalances. At least this is true for countries whose main tradable

sector is manufacturing. Though rather intuitive, this structural problem of increasingly service-

oriented economies is much too often neglected. One reason is that in the orthodox view the current

account is ultimately determined by purely macroeconomic factors, i.e. by the saving decisions by

households that are optimising their consumption path over time. In contrast, our empirical results

suggest that the tradability of output is an important structural feature of the economy which is

highly relevant for the current account balance and external stability.

For the EU our results imply that the increasing diverging structural developments, in particular with

regards to manufacturing activities, in the Central European Manufacturing Core (Stehrer and

Stöllinger, 2014; IMF, 2013) and the rest of the EU member states must be expected to foster

external imbalances within the Union. This calls for a solution which à la long can only consist in a

comprehensive reform of the current fiscal framework, either by significantly strengthening the

cohesion efforts or by introducing an internal transfer mechanism of some sort.

27

Our results also indicate that the current account is a complicated matter co-determined by a

plethora of factors which are often tightly intertwined. For example, Rodrik (2012) suggests that the

manufacturing sector unconditionally serves as an accelerator for economic development. Economic

development implies higher growth. According to this manufacturing convergence hypothesis we

should observe that countries which shift towards manufacturing grow faster. Faster growth,

however, tends to create current account deficits. In contrast, our tradability hypothesis suggests

that a move towards tradable activities such as manufacturing should improve the current account

position. This is why in the empirical application we made a great effort to control for a wide range of

factors that are equally relevant for the current account position.

We see this analysis as the beginning of more intensive research in this area. The natural extension of

this paper is to take the test of the tradability-current account nexus to the global level. While there

will be serious data constraints, for OECD countries, several South and South East Asian countries as

well as South American countries the required data would be available. A main topic that we have

not taken on board in this paper is the role of the real exchange rate. Since trade theories suggest

that relative prices determine specialisation pattern, the exchange rate could be responsible for a

part of the effect of our tradability index on the current account. Therefore one further route will be

to integrate the real exchange rate in the current account model. Another interesting exercise would

be an analysis of the tradability-current account nexus for selected countries for which longer time

series are available (such as France or Hungary). This would make it possible to explore the dynamics

of the relationship between the current account and the tradability index.

28

Literature

Arellano, M., Bond, S. (1991) Some tests of specification for panel data: Monte Carlo evidence and an application to

employment equations, Review of Economic Studies, 58, pp. 277–297.

Alexander, S. S. (1952) Effects of a devaluation on a trade balance, IMF Staff Papers, 2, pp. 263-278.

Anderson, J.E., van Wincoop, E. (2004) Trade Costs, Journal of Economic Literature, 42, pp. 691-751.

Barattieri, A. (2014) Comparative advantage, service trade, and global imbalances, Journal of International

Economics, 2014, 92(1), pp.1-13.

Blanchard, O., Giavazzi, F. (2002) Current Account Deficits in the Euro Area: The End of the Feldstein-Horioka Puzzle?,

Brookings Papers on Economic Activity, 2, pp. 147-186.

Brissimis, S.N., Hondroyiannis G., Papazoglou, C., Tsaveas, T.N., Vasardani A.M. (2010) Current account determinants

and external sustainability in periods of structural change. Economic Change and Restructuring, 45(1), pp. 71-95.

Bussière, M., Fratzscher, M., Müller, G. (2006) Current account dynamics in OECD and new EU member states: an

intertemporal approach. Journal of Economic Integration 21 (3), pp. 593-618.

Caballero, R.J., Farhi, E., Gourinchas, P.-O. (2008). An equilibrium model of ‘global imbalances’ and low interest rates.

American Economic Review 98, pp. 358–393.

Calderon, C., Chong, A., Loayza, N. (2002) Determinants of current account in developing countries, Contributions to

Macroeconomics 2, pp. 1-35.

Ca’Zorzi, M., Chudik, A., Dieppe, A. (2012) Thousands of Models, one Story. Current Account Imbalances in the Global

Economy, ECB Working Paper Series, No. 1441, June 2012.

Cova, P., Pisani, M., Batini, N., Rebucci, A. (2009) Global Imbalances: the Role of Non-Tradable Total Factor

Productivity in Advanced Economies, IMF working paper 09/63.

Chinn, M. D., Prasad, E. S. (2003) Medium-term determinants of current accounts in industrial and developing

countries: an empirical exploration, Journal of International Economics 59, 47-76.

Debelle, G., Faruqee, E. (1996) What determines the current account? A cross-sectional and panel approach. IMF

Working Paper No. 58.

De Gregorio, J., Giovannini, A., Wolf, H.C. (1994) ‘International Evidence on Tradables and Nontradables Inflation,

European Economic Review, 38, 1225-1244.

Ehmer, P. (2014) The impact of diverging economic structure on current account imbalances in the euro area,

Witten/Herdecke University Working paper, 27/2014.

Ferrero, A., Gertler, M., Svensson, L. (2010) ‘Current Account Dynamics and Monetary Policy’, in: Galí, J. and M.

Gertler (eds.), International Dimensions of Monetary Policy, University of Chicago Press.

Fournier, J., Koske, I. (2010) A Simple Model of the Relationship Between Productivity, Saving and the Current

Account, OECD Economics Department Working Papers, No. 816.

Gaulier, G., Vicard. V. (2012) Current account imbalances in the euro area: competitiveness or demand shock?,

Quarterly selection of articles - Bulletin de la Banque de France, 27, pp. 5-26.

Gehringer, A. (2015) New Evidence on the Determinants of Current Accounts in the EU, Empirica, 42(4), pp. 769-793.

Harms, P. (2008) Internationale Makroökonomik, Mohr Siebeck, Tübingen.

IMF (2013) IMF Multi-Country Report. German-Central European Supply Chain – Cluster Report, IMF Country Report,

No. 13/263, August.

Jin, K. (2012) Industrial Structure and Capital Flows, American Economic Review, 102(5), 2111-2146.

Johnson, R.C., and Noguera, G., (2012), Accounting for Intermediates: Production Sharing and Trade in Value Added,

Journal of International Economics, 86(2), 224-236.

29

Kerdrain, C., Koske, I., Wanner, I. (2010) The Impact of Structural Policies on Saving, Investment and Current

Accounts, OECD Economics Department Working Papers No. 815.

Lane, P. R., Pels, B. (2012) Current Account Imbalances in Europe, CEPR Discussion Papers, 8958.

Mendoza, E.G., Quadrini, V., Rios-Rull, J.-V. (2009) Financial integration, financial deepness, and global imbalances.

Journal of Political Economy, 117 (3), pp. 371–416.

Nedoncelle, C. (2014) Trade costs and the current account, Document de travail, Université de Lille, No. 2014-44.

Obstfeld, M., Rogoff, K. (1996) The intertemporal approach to the current account, Handbook of International

Economics, in: Grossman, G. M., Rogoff, K. (ed.), Handbook of International Economics (ed. 1), Elsevier, volume 3,

chapter 34, pp.1731-1799.

Obstfeld, M., Rogoff, K. (2000) The six major puzzles in international macroeconomics: is there a common cause?,

NBER Working Paper 7777.

Rajan, R., Zingales, L. (1998) Financial dependence and growth, American Economic Review, 88, pp. 559-586.

Rodrik, D. (2012) Unconditional Convergence in Manufacturing, Quarterly Journal of Economics, 128(1), pp. 165-204.

Schnabl, G. Wollmershäuser, T. (2013) Fiscal Divergence and Current Account Imbalances in Europe, CESifo Working

Paper Series, No. 4108.

Shleifer, A., Treisman, (2014) Normal Countries: The East 25 Years After Communism, Foreign Affairs, 93(6).

Stehrer, R., Stöllinger, R. (2015) The Central European Manufacturing Core: What is Driving Regional Production

Sharing?, FIW-Research Reports 2014/15 N° 02, February 2015.

Stöllinger, R., Foster-McGregor, N., Holzner, M., Landesmann, M., Pöschl, J., Stehrer, R. (2013) A ‘Manufacturing

Imperative’ in the EU – Europe’s Position in Global Manufacturing and the Role of Industrial Policy, wiiw Research

Reports, No. 391.

Timmer M (2012) The World Input–output Database (WIOD): contents, sources and methods. Version 0.9, April,

2012.

Vogel, L. (2011) Structural reforms and external rebalancing in the euro area: a model-based analysis, European

Economy Economic Paper, no. 443.

Zeugner, S. (2013) Tradable vs. Non-Tradable: An Empirical Approach to the Classification of Sectors, Mimeo,

23 October 2013.

30

Appendix 1: List of countries

Abbreviation Country Country Category

AL Albania Emerging

AM Armenia Emerging

AT Austria Developed

AZ Azerbaijan Emerging

BY Belarus Emerging

BE Belgium Developed

BA Bosnia and Herzegovina Emerging

BG Bulgaria Emerging

HR Croatia Emerging

CY Cyprus Emerging

CZ Czech Republic Developed

DK Denmark Developed

EE Estonia Developed

FI Finland Developed

FR France Developed

GE Georgia Emerging

DE Germany Developed

EL Greece Developed

HU Hungary Developed

IS Iceland Developed

IE Ireland Developed

IT Italy Developed

KZ Kazakhstan Emerging

LV Latvia Developed

LT Lithuania Emerging

LU Luxembourg Developed

MK Macedonia Emerging

MT Malta Emerging

MD Moldova Emerging

ME Montenegro Emerging

NL Netherlands Developed

NO Norway Developed

PL Poland Emerging

PT Portugal Developed

RO Romania Emerging

RU Russia Emerging

RS Serbia Emerging

SK Slovakia Developed

SI Slovenia Developed

ES Spain Developed

SE Sweden Developed

CH Switzerland Developed

TR Turkey Emerging

UA Ukraine Emerging

UK United Kingdom Developed

XK Kosovo Emerging

Note: The distinction between ‘Developed’ and ‘Emerging’ mirrors the categorisation of European countries as ‘Advanced’ and

‘Emerging and Developing’ by the IMF as of April 2014.

31

Appendix 2: List of sectors for the calculation of the tradability index

Number Sector NACE Rev. 1 NACE Rev. 2

1 Agriculture, hunting and forestry + Fishing A+B A

2 Mining and quarrying C B

3 Manufacturing D C

4 Electricity, gas and water supply E D+E

5 Construction F F

6 Wholesale, retail trade, repair of motor vehicles etc. G G

7 Hotels and restaurants H I

8 Transport, storage + Communication I H + J

9 Financial intermediation J K

10 Real estate, renting and business activities K L+M+N

11 Public administration, defence, compuls.soc.security L O

12 Education M P

13 Health and social work N Q

14 Other community, social and personal services +

Private households with employed persons O+P R+S+T

32

Appendix 3: Methodology for calculating value added exports and the tradability scores

Deriving the tradability score requires the calculation of the value added exports (VAX) at the

industry-country level. This appendix illustrates the basic input-output methodology to calculate the

VAX, including a 3-country, 2-sector example.

Following the trade in value added concept in Johnson and Noguera (2012) and the expositions in

Stehrer (2012) we require three components in order to calculate the value added exports. For any

country r, these components are the value added requirements per unit of gross output, ; the

Leontief inverse of the global input-output matrix, ; and the final consumption vector, . Both

vectors as well as the Leontief inverse have an industry dimension . The industry index is omitted in

order to facilitate the exposition.

Country r’s value added coefficients are defined as

. The value added coefficients

are arranged in a diagonal matrix of dimension 1435 x 1 (40 countries x 35 industries). This matrix

contains the value added coefficients of country r for all industries along the diagonals. The

remaining entries of the matrix are zero because the interest here is with the value added created in

country r.

The second element is the Leontief inverse of the global input-output matrix, where

denotes the coefficient matrix. In the WIOT the coefficient matrix (and hence the Leontief matrix) is

of dimension 1435 × 1435 which contains the technological input coefficients of country r in the

diagonal elements and the technological input coefficients of country r’s imports (from a column

perspective) and exports (from a row perspective) in the off-diagonal elements.

The final building block is the global final consumption vector. This vector is also industry specific and

if of dimension 1435 × 1. Most importantly, for our purposes, final consumption must be split into

separate blocks indication the origin of the consumed goods though within the elements in the

column vector. As usual, each row is associated with one source of the final demand. The full

consumption vector, , in the 3-country one sector case has the form

where the subscript J indicates that the vector comprises the consumption of all countries . The

typical element of this vector contains the final consumption from all possible sources. For example,

the element captures the value of final goods that country 3 demands from country r. Since the

idea of value added exports is that it comprises only value added that is created in one country but

absorbed in another, the final demand from country r itself needs to be eliminated for the calculation

of country r’s VAX. Therefore we will work with an adjusted final demand vector,

in which

country r’s final demand (i.e. the first column in the above matrix) is set to zero. Country r’s value

added exports can then be calculated as

(1)

where are the sector specific value added exports of country r to all partner countries.

33

To illustrate this, we illustrate the matrices in the three countries–two sector case, where country r

acts as the model country and we label the industries with m (for manufacturing) and s (for services).

Equation (1) then has the following form:

The coefficients in the Leontief matrix represent the total direct and indirect input requirements of

any country in order to produce one dollar worth of output for final demand. For example, the

coefficient indicates the input requirement of country r’s services sector from country r’s

manufacturing sector for producing one unit of output. Likewise the coefficient indicates

country r’s input requirement in the manufacturing sector supplied by country 3’s manufacturing

sector.

The resulting elements in this example,

and

are the total value added exports of

country r’s manufacturing respectively services sector to all other sectors of all partner countries.

The VAX are not only calculated for country r but for all 40 countries plus the rest of the world.

Hence, the final step needed to arrive at the global industry-level VAX is to sum up the VAX of all

countries for each individual sector i. Dividing the global industry-specific VA by the corresponding

industry-specific value added yields the tradability score by sector.

34

Appendix 4: Tradability index across countries, gross exports based (average 1995-2014)

Figure A.1: Tradability index across countries, based on gross exports (average 1995-2014)

Note: Tradability index based on gross exports.

Source: WIOD, author’s own calculations.

sample mean TIx

0.12

0.16

0.20

0.24

0.28

0.32

0.36 B

Y A

Z C

Z R

O

IE

UA

R

U

SI

FI

SK

KZ

HU

TR

D

E R

S N

O

SE

MK

LT

P

L A

T B

G

CH

M

D

EE

AM

IT

H

R

BE

MT

DK

LV

ES

P

T G

E X

K

NL

UK

B

A

FR

IS

LU

ME EL

A

L C

Y

trad

abili

ty in

dex

(gr

oss

exp

ort

s b

ased

)

35

Appendix 5: Data and data availability: Current Account and TI, 1995-2014

Current Account: data sources and data availability

reporter country data source data availability switch BMP5 to BMP6

AL Albania wiiwADB 1995-2014 2013

AM Armenia IMF WEO/IMF IFS 1995-2014 2005

AT Austria EurostatBOP 1995-2014 1995

AZ Azerbaijan IMF WEO/IMF IFS 1995-2014 2005

BA Bosnia and Herzegovina wiiwADB 1998-2014 2007

BE Belgium EurostatBOP 1995-2014 2003

BG Bulgaria wiiwADB 1995-2014 2007

BY Belarus IMF WEO/IMF IFS

wiiw ADB 1995-1999

2000-2014 2000

CH Switzerland OECD 2005-2014 2000

CY Cyprus EurostatBOP 1995-2014 2008

CZ Czech Republic wiiwADB 1995-2014 1995

DE Germany EurostatBOP 1995-2014 1995

DK Denmark EurostatBOP 1995-2014 2005

EE Estonia wiiwADB 1995-2014 1995

EL Greece EurostatBOP 1995-2014 2004

ES Spain EurostatBOP 1995-2014 1995

FI Finland EurostatBOP 1995-2014 1995

FR France EurostatBOP 1995-2014 1999

GE Georgia IMF WEO

IMF WEO/IMF IFS 1995-1996 1997-2014

2005

HR Croatia wiiwADB 1995-2014 2000

HU Hungary wiiwADB 1995-2014 1995

IE Ireland EurostatBOP 1995-1997 1998-2014

1998

IS Iceland EurostatBOP 1995-2014 1995

IT Italy EurostatBOP 1995-2014 1995

KZ Kazakhstan IMF WEO wiiwADB

1995-1999 2000-2014

2005

LT Lithuania wiiwADB 1995-2014 2004

LU Luxembourg EurostatBOP 1995-2014 2000

LV Latvia wiiwADB 1995-2014 2000

MD Moldova IMF WEO/IMF IFS 1995-2014 2005

ME Montenegro wiiwADB 2001-2014 2013

MK Macedonia wiiwADB 1995-2014 1998

MT Malta EurostatBOP 1995-2014 2013

NL Netherlands EurostatBOP 1995-2014 2004

NO Norway EurostatBOP 1995-2014 2012

PL Poland wiiwADB 1995-2014 2004

PT Portugal EurostatBOP 1995-2014 1996

RO Romania wiiwADB 1995-2014 1999

RS Serbia wiiwADB 1997-2014 2007

RU Russia wiiwADB 1995-2014 2000

SE Sweden EurostatBOP 1995-2014 1995

SI Slovenia wiiwADB 1995-2014 1995

SK Slovakia wiiwADB 1995-2014 2008

TR Turkey wiiwADB 1995-2014 2010

UA Ukraine wiiwADB 1995-2014 2005

UK United Kingdom EurostatBOP 1995-2014 1997

XK Kosovo IMFWEO wiiw ADB

2000-2003 2004-2014

2004

36

Tradability index: data sources and data availability

reporter country data source years industry structure

AL Albania UN

wiiwADB 1995-2013

NACE Rev.1 1995-2007 NACE Rev.2 2008-2014

AM Armenia UN 1995-2013 NACE Rev.1 1995-2008 NACE Rev.2 2009-2014

AT Austria Eurostat 1995-2013 NACE Rev. 2 1995-2014; ESA2010

AZ Azerbaijan UN 1995-2013 NACE Rev.1 1995-2004 NACE Rev.2 2005-2013

BA Bosnia and Herzegovina UN

wiiwADB wiiwADB

1999-2014 NACE Rev.1 1995-2007 NACE Rev.1 2008-2004 NACE Rev.2 2005-2014

BE Belgium Eurostat 1995-2014 NACE Rev. 2 1995-2014; ESA2010

BG Bulgaria UN

wiiwADB wiiwADB

1995-2013

NACE Rev.1 1995 NACE Rev.1 1996-1999 NACE Rev.2 2000-2013 ESA 2010

BY Belarus UN

wiiwADB 1995-2014

NACE Rev.1 1995-1999 NACE Rev.1 2000-2014

CH Switzerland UN

Eurostat 1995-2013

NACE Rev.1 1995-1997 NACE Rev.2 1997-2013 ECA 2010

CY Cyprus Eurostat 1995-2014 NACE Rev. 2 1995-2014; ESA2010

CZ Czech Republic wiiwADB 1995-2014 NACE Rev. 2 1995-2014; ESA2010

DE Germany Eurostat 1995-2014 NACE Rev. 2 1995-2014; ESA2010

DK Denmark Eurostat 1995-2014 NACE Rev. 2 1995-2014; ESA2010

EE Estonia wiiwADB 1995-2014 NACE Rev. 2 1995-2014 ESA 1995: 1995-1999 ESA2010: 2000-2014

EL Greece Eurostat 1995-2014 NACE Rev. 2 1995-2014; ESA2010

ES Spain Eurostat 1995-2014 NACE Rev. 2 1995-2014; ESA2010

FI Finland Eurostat 1995-2014 NACE Rev. 2 1995-2014; ESA2010

FR France Eurostat 1995-2014 NACE Rev. 2 1995-2014; ESA2010

GE Georgia UN 1995-2014 NACE Rev.1 1995-2014

HR Croatia wiiwADB 1995-2013 NACE Rev. 2 1995-2014; ESA2010

HU Hungary wiiwADB 1995-2014 NACE Rev. 2 1995-2014; ESA2010

IE Ireland Eurostat 1995-2014 NACE Rev. 2 1995-2014; ESA2010

IS Iceland Eurostat 1997-2013 NACE Rev. 2 1997-2013; ESA2010

IT Italy Eurostat 1995-2014 NACE Rev. 2 1995-2014; ESA2010

KZ Kazakhstan

UN UN

wiiwADB wiiwADB

1995-2014

NACE Rev. 1 1995-1997 NACE Rev. 1 1998 NACE Rev. 1 1998-2005 NACE Rev. 2 2006-2014

LT Lithuania wiiwADB 1995-2014 NACE Rev. 2 1995-2014; ESA2010

LU Luxembourg Eurostat 1995-2014 NACE Rev. 2 1995-2014 ESA1995: 1995-1999 ESA2010: 2000-2014

LV Latvia wiiwADB 1995-2014 NACE Rev. 2 1995-2014; ESA2010

MD Moldova UN 1995-2014 NACE Rev.1 1995-2014

ME Montenegro wiiwADB 2000-2014 NACE Rev. 1: 2000-2009 NACE Rev. 2: 2010-2014

MK Macedonia wiiwADB 1997-2013 NACE Rev. 1: 1997-1999 NACE Rev. 2: 2000-2013

MT Malta Eurostat 1995-2014 NACE Rev. 2 1995-2014; ESA2010

NL Netherlands Eurostat 1995-2014 NACE Rev. 2 1995-2014; ESA2010

NO Norway Eurostat 1995-2014 NACE Rev. 2 1995-2014; ESA2010

PL Poland Eurostat wiiwADB

1995-2014 NACE Rev. 2 1995-2014 ESA1995: 1995-2001 ESA2010: 2002-2014

PT Portugal Eurostat 1995-2013 NACE Rev. 2 1995-2013; ESA2010

RO Romania wiiwADB 1995-2014 NACE Rev. 2 1995-2014; ESA2010

RS Serbia wiiwADB 1995-2014 NACE Rev. 2 1995-2014; ESA2010

RU Russia UN

wiiwADB 1995-2014

NACE Rev.1 1995-2001 NACE Rev.1 2002-2014

SE Sweden Eurostat 1995-2014 NACE Rev. 2 1995-2014; ESA2010

SI Slovenia wiiwADB 1995-2014 NACE Rev. 2 1995-2014; ESA2010

SK Slovakia wiiwADB 1995-2014 NACE Rev. 2 1995-2014; ESA2010

TR Turkey UN

wiiwADB 1995-2014

NACE Rev.1 1995-1997 NACE Rev.1 1998-2014

UA Ukraine UN

wiiwADB wiiwADB

1995-2014 NACE Rev.1 1995-1999 NACE Rev.1 2000 NACE Rev.2 2001-2014

UK United Kingdom Eurostat 1995-2014 NACE Rev. 2 1995-2014 ESA1995: 1995-1997 ESA2010: 1998-2014

XK Kosovo wiiw ADB 2006-2014 NACE Rev. 1 2006-2007 NACE Rev. 2 2008-2014

37

Current Account in % of GDP series

reporter 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014

AL 1.459 -1.888 -15.168 -2.533 -4.026 -4.694 -6.961 -9.437 -6.924 -5.783 -8.978 -6.572 -10.529 -15.696 -15.351 -11.321 -13.219 -10.203 -10.896 -12.864

AM -17.147 -20.176 -20.484 -21.995 -16.940 -15.778 -10.442 -6.221 -6.181 -2.204 -2.529 -2.396 -7.357 -14.225 -16.484 -13.622 -10.440 -9.962 -7.595 -7.292

AT -2.746 -2.777 -2.563 -1.870 -2.266 -0.707 -0.798 2.106 1.554 2.083 2.268 3.311 3.824 4.518 2.613 2.867 1.639 1.490 1.949 1.971

AZ -16.576 -29.310 -23.108 -31.881 -13.091 -3.182 -0.906 -12.330 -27.774 -29.823 1.263 17.633 27.256 35.475 22.974 28.425 26.450 21.799 16.633 13.769

BA -7.406 -9.438 -7.107 -12.855 -17.521 -19.106 -16.121 -16.614 -7.639 -9.079 -13.784 -6.410 -6.040 -9.475 -8.850 -5.494 -7.819

BE 5.310 4.916 5.426 5.080 4.966 3.937 3.317 4.369 3.528 3.269 2.084 1.920 1.950 -0.998 -1.077 1.764 -1.073 -0.054 -0.222 -0.216

BG -1.372 1.600 9.315 -0.218 -4.629 -5.324 -5.395 -2.309 -5.189 -6.211 -11.273 -16.991 -23.680 -21.882 -8.336 -0.875 0.916 -0.260 1.826 1.159

BY -13.543 -3.558 -6.094 -6.678 -1.596 -4.020 -4.372 -2.248 -2.588 -5.159 1.519 -3.758 -6.662 -8.212 -12.508 -15.071 -9.531 -2.939 -10.458 -6.853

CH 6.010 6.479 8.602 8.524 10.030 12.375 8.568 8.950 13.249 15.213 14.024 14.949 10.759 2.959 7.994 14.816 7.646 10.320 11.148 8.823

CY -2.096 -4.735 -4.379 2.828 -1.623 -4.977 -3.043 -3.480 -2.080 -4.585 -5.353 -6.233 -10.685 -15.647 -7.743 -10.655 -3.955 -5.629 -4.451 -4.565

CZ -2.272 -6.106 -5.821 -1.923 -2.272 -4.442 -4.859 -5.116 -5.716 -4.238 -0.949 -2.138 -4.309 -1.896 -2.314 -3.650 -2.119 -1.567 -0.529 0.619

DE -1.190 -0.658 -0.510 -0.714 -1.419 -1.746 -0.363 1.888 1.406 4.441 4.613 5.680 6.748 5.595 5.736 5.624 6.087 6.801 6.372 7.275

DK 0.674 1.424 0.599 -0.839 1.870 1.372 3.048 2.418 3.362 2.935 4.258 3.153 1.384 2.668 3.321 5.728 5.742 5.689 7.132 7.719

EE -3.939 -8.126 -11.120 -8.636 -4.287 -5.409 -7.107 -11.137 -12.922 -12.008 -8.742 -14.981 -15.003 -8.706 2.545 1.807 1.335 -2.434 -0.104 1.028

EL -2.085 -3.129 -3.349 -2.552 -3.431 -7.426 -6.955 -6.241 -6.297 -7.734 -8.877 -11.596 -15.228 -15.125 -12.367 -11.435 -10.006 -3.833 -2.044 -2.124

ES -0.932 -0.727 -0.722 -1.670 -3.298 -4.400 -4.391 -3.738 -3.883 -5.586 -7.494 -8.990 -9.648 -9.251 -4.281 -3.921 -3.180 -0.231 1.509 0.983

FI 3.987 3.780 5.095 5.151 5.174 7.548 8.060 8.224 4.642 5.800 3.000 3.800 3.800 2.200 1.900 1.243 -1.779 -1.934 -1.654 -0.941

FR 0.676 1.271 2.674 2.688 3.385 1.182 1.507 1.174 0.867 0.440 -0.018 0.036 -0.299 -0.953 -0.831 -0.836 -1.032 -1.194 -0.806 -0.926

GE -18.292 -12.574 -14.383 -7.616 -7.078 -5.785 -6.152 -6.392 -9.581 -6.945 -10.838 -15.369 -19.573 -21.964 -10.581 -10.276 -12.748 -11.679 -5.742 -10.521

HR -6.439 -4.597 -10.467 -5.364 -6.369 -2.189 -2.918 -7.061 -5.898 -4.050 -5.182 -6.503 -7.145 -8.783 -5.100 -1.085 -0.706 -0.048 1.017 0.846

HU -3.395 -3.719 -4.054 -7.155 -7.889 -8.488 -5.842 -6.339 -8.013 -8.563 -6.999 -7.038 -7.105 -7.088 -0.805 0.279 0.749 1.770 3.986 2.260

IE 2.473 2.715 2.336 0.779 0.244 0.616 0.176 0.247 0.492 -0.098 -3.303 -4.924 -6.071 -5.777 -4.142 -0.755 -1.155 -1.537 3.096 3.614

IS 0.187 -2.502 -2.364 -7.258 -7.186 -10.310 -4.644 1.187 -4.919 -9.877 -16.013 -23.237 -13.708 -25.946 -10.251 -6.679 -5.401 -4.063 5.730 3.322

IT 2.100 2.900 2.600 1.700 0.800 -0.300 0.100 -0.298 -0.642 -0.385 -0.946 -1.562 -1.452 -2.866 -1.939 -3.476 -3.079 -0.430 0.891 1.910

KZ -1.281 -3.570 -3.606 -5.532 -1.034 2.003 -6.273 -4.158 -0.884 0.777 -1.814 -2.469 -7.985 4.684 -3.574 0.936 5.090 0.490 0.352 2.635

LT -9.167 -8.618 -9.697 -11.548 -10.959 -5.905 -4.668 -5.089 -6.696 -7.600 -7.238 -10.514 -15.047 -13.346 2.074 -0.330 -3.867 -1.179 1.542 3.580

LU 11.471 11.075 10.669 9.401 8.312 12.537 8.401 9.468 6.562 11.923 11.099 10.021 9.843 7.669 7.340 6.785 6.183 6.061 5.677 5.497

LV -0.331 -4.693 -5.308 -8.994 -8.626 -4.714 -7.486 -6.466 -7.744 -12.205 -11.743 -20.722 -20.721 -12.361 8.099 2.363 -2.844 -3.271 -2.385 -1.980

MD -5.872 -11.324 -14.257 -19.723 -5.792 -7.467 -1.808 -1.191 -6.569 -1.775 -7.557 -11.339 -15.249 -16.112 -8.214 -7.524 -12.139 -8.738 -6.358 -7.088

ME -15.373 -12.864 -6.761 -7.175 -16.635 -31.342 -39.497 -49.755 -27.852 -22.726 -17.560 -18.471 -14.472 -15.205

MK -6.709 -7.705 -7.676 -7.822 -1.785 -2.732 -6.345 -9.483 -3.871 -7.904 -2.433 -0.427 -6.910 -12.732 -6.754 -2.025 -2.509 -3.164 -1.646 -0.806

MT -9.827 -10.675 -5.193 -5.436 -3.077 -12.128 -3.633 2.280 -2.878 -3.704 -6.515 -6.585 -1.916 -1.052 -6.578 -4.658 -2.441 1.322 3.607 2.977

NL 5.917 4.923 6.175 3.032 3.634 1.907 2.442 2.487 5.217 6.763 6.091 7.897 5.989 4.068 5.806 7.351 9.111 10.789 10.982 10.571

NO 3.453 6.701 6.183 0.637 6.023 15.196 15.031 12.493 12.111 12.447 15.984 16.795 13.689 17.217 12.865 11.959 13.335 12.426 10.218 9.709

PL 0.599 -2.041 -3.609 -3.983 -7.357 -5.999 -3.127 -2.819 -2.537 -5.458 -2.628 -4.021 -6.352 -6.711 -4.018 -5.387 -5.168 -3.714 -1.274 -2.020

PT -0.133 -4.388 -6.110 -7.594 -8.854 -10.802 -10.435 -8.487 -7.166 -8.329 -9.882 -10.673 -9.739 -12.127 -10.422 -10.148 -6.002 -2.000 1.400 0.549

RO -4.712 -6.911 -5.852 -6.904 -4.085 -3.744 -5.536 -3.387 -5.831 -8.340 -8.627 -10.357 -13.828 -11.806 -4.853 -5.072 -4.947 -4.789 -1.069 -0.451

RS -4.583 -2.469 -2.367 -0.626 2.043 -3.924 -7.189 -13.122 -8.425 -9.642 -18.585 -21.143 -6.628 -6.842 -10.938 -11.588 -6.124 -5.957

RU 2.220 2.768 -0.020 0.081 12.567 17.474 10.455 7.961 7.697 9.908 11.044 9.322 5.552 6.247 4.113 4.422 5.101 3.309 1.558 2.879

SE 3.178 3.332 3.904 3.640 3.936 3.953 4.732 4.482 6.595 6.294 6.467 8.260 8.863 8.596 5.885 6.003 6.885 6.627 6.707 5.724

SI -0.443 0.064 0.109 -0.669 -3.235 -2.782 0.039 0.870 -0.810 -2.690 -1.792 -1.831 -4.128 -5.317 -0.561 -0.119 0.190 2.584 5.634 6.989

SK 2.560 -9.082 -8.245 -8.695 -4.718 -3.406 -8.167 -7.766 -5.812 -7.653 -8.331 -7.688 -5.192 -6.473 -3.458 -4.724 -4.964 0.944 1.959 0.132

TR -1.379 -1.340 -1.386 0.746 -0.415 -3.719 1.961 -0.298 -2.484 -3.609 -4.451 -6.068 -5.830 -5.435 -1.975 -6.200 -9.697 -6.149 -7.881 -5.852

UA -3.113 -2.659 -2.662 -3.094 5.250 4.737 3.567 7.219 5.559 10.279 2.839 -1.447 -3.530 -6.794 -1.428 -2.136 -6.043 -7.851 -8.770 -3.487

UK -1.147 -0.744 0.146 -0.091 -2.523 -2.243 -2.031 -2.069 -1.694 -1.834 -1.249 -2.301 -2.509 -3.615 -3.036 -2.809 -1.694 -3.298 -4.492 -5.108

XK -20.158 -5.837 -6.688 -8.074 -7.152 -8.240 -7.245 -6.185 -11.871 -9.193 -11.715 -13.677 -7.514 -6.371 -7.849

38

Tradability index series

reporter 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014

AL 0.185 0.164 0.159 0.161 0.159 0.156 0.152 0.148 0.146 0.148 0.148 0.149 0.150 0.143 0.142 0.149 0.154 0.157 0.160 0.163

AM 0.214 0.210 0.205 0.200 0.200 0.187 0.187 0.180 0.176 0.179 0.172 0.157 0.150 0.144 0.149 0.158 0.167 0.166 0.166 0.166

AT 0.167 0.166 0.168 0.167 0.168 0.169 0.170 0.169 0.167 0.167 0.168 0.169 0.170 0.168 0.163 0.164 0.165 0.165 0.164 0.163

AZ 0.225 0.219 0.214 0.188 0.215 0.258 0.270 0.267 0.261 0.262 0.314 0.341 0.357 0.349 0.311 0.322 0.326 0.305 0.287 0.270

BA

0.159 0.154 0.155 0.154 0.157 0.159 0.160 0.161 0.163 0.161 0.157 0.160 0.161 0.160 0.162 0.160

BE 0.170 0.169 0.171 0.170 0.167 0.167 0.166 0.164 0.162 0.162 0.161 0.159 0.159 0.155 0.149 0.151 0.149 0.148 0.147 0.146

BG 0.199 0.201 0.202 0.191 0.182 0.170 0.175 0.175 0.178 0.176 0.176 0.179 0.178 0.173 0.171 0.171 0.179 0.180 0.176 BY 0.217 0.223 0.225 0.221 0.219 0.210 0.203 0.204 0.202 0.207 0.208 0.204 0.203 0.208 0.200 0.196 0.216 0.210 0.195 0.194

CH 0.166 0.165 0.168 0.167 0.166 0.165 0.166 0.166 0.165 0.165 0.166 0.168 0.170 0.170 0.165 0.165 0.165 0.163 0.163 CY 0.142 0.141 0.140 0.140 0.139 0.138 0.137 0.135 0.132 0.131 0.129 0.126 0.125 0.123 0.124 0.126 0.125 0.125 0.127 0.128

CZ 0.192 0.193 0.196 0.194 0.193 0.196 0.197 0.191 0.189 0.193 0.193 0.196 0.196 0.193 0.186 0.188 0.192 0.192 0.192 0.196

DE 0.176 0.173 0.175 0.176 0.175 0.177 0.177 0.176 0.176 0.177 0.177 0.180 0.181 0.179 0.170 0.176 0.178 0.177 0.176 0.176

DK 0.161 0.161 0.163 0.160 0.161 0.167 0.165 0.164 0.161 0.162 0.164 0.165 0.164 0.162 0.152 0.155 0.157 0.158 0.157 0.156

EE 0.178 0.178 0.181 0.175 0.173 0.177 0.179 0.178 0.177 0.174 0.171 0.169 0.167 0.164 0.161 0.170 0.173 0.172 0.171 0.171

EL 0.153 0.151 0.146 0.145 0.146 0.147 0.145 0.145 0.141 0.139 0.141 0.136 0.139 0.140 0.134 0.135 0.137 0.138 0.140 0.139

ES 0.160 0.160 0.162 0.161 0.159 0.159 0.158 0.156 0.155 0.153 0.151 0.150 0.149 0.147 0.142 0.144 0.146 0.146 0.147 0.147

FI 0.185 0.181 0.183 0.187 0.187 0.191 0.190 0.188 0.184 0.182 0.180 0.181 0.183 0.178 0.163 0.165 0.163 0.156 0.156 0.156

FR 0.155 0.154 0.154 0.155 0.154 0.155 0.153 0.151 0.149 0.148 0.146 0.144 0.144 0.142 0.138 0.138 0.138 0.138 0.138 0.137

GE 0.188 0.194 0.183 0.176 0.177 0.176 0.175 0.175 0.176 0.168 0.166 0.159 0.150 0.144 0.142 0.148 0.151 0.150 0.154 0.154

HR 0.188 0.177 0.177 0.173 0.172 0.177 0.177 0.173 0.171 0.173 0.169 0.169 0.167 0.165 0.164 0.166 0.168 0.170 0.167 HU 0.177 0.179 0.184 0.185 0.183 0.181 0.180 0.176 0.175 0.177 0.177 0.180 0.180 0.177 0.174 0.179 0.182 0.183 0.183 0.184

IE 0.184 0.183 0.189 0.193 0.195 0.193 0.198 0.202 0.190 0.182 0.176 0.171 0.171 0.170 0.181 0.184 0.189 0.183 0.181 0.179

IS

0.164 0.160 0.155 0.153 0.158 0.152 0.147 0.146 0.138 0.139 0.137 0.146 0.152 0.159 0.159 0.154 0.153 IT 0.176 0.175 0.174 0.175 0.173 0.172 0.170 0.169 0.166 0.165 0.163 0.163 0.164 0.162 0.155 0.157 0.158 0.157 0.157 0.158

KZ 0.238 0.236 0.201 0.200 0.215 0.233 0.226 0.222 0.222 0.223 0.225 0.219 0.214 0.231 0.224 0.236 0.235 0.232 0.224 0.222

LT 0.175 0.174 0.172 0.168 0.166 0.176 0.178 0.176 0.177 0.180 0.181 0.177 0.170 0.168 0.168 0.178 0.183 0.185 0.181 0.179

LU 0.158 0.155 0.156 0.156 0.154 0.153 0.151 0.149 0.149 0.148 0.146 0.145 0.148 0.144 0.135 0.138 0.137 0.136 0.135 0.133

LV 0.181 0.182 0.185 0.172 0.165 0.166 0.170 0.170 0.166 0.165 0.162 0.155 0.149 0.146 0.152 0.164 0.161 0.161 0.159 0.157

MD 0.205 0.198 0.189 0.180 0.180 0.186 0.188 0.180 0.182 0.180 0.177 0.172 0.167 0.165 0.160 0.164 0.168 0.165 0.168 0.167

ME

0.162 0.170 0.166 0.160 0.157 0.152 0.152 0.143 0.140 0.136 0.139 0.138 0.134 0.136 0.137

MK

0.181 0.179 0.177 0.154 0.151 0.149 0.149 0.145 0.152 0.152 0.154 0.154 0.151 0.157 0.163 0.158 0.158 MT 0.173 0.170 0.170 0.171 0.171 0.176 0.163 0.160 0.159 0.154 0.152 0.153 0.153 0.154 0.147 0.147 0.147 0.146 0.140 0.138

NL 0.169 0.170 0.168 0.167 0.164 0.166 0.166 0.161 0.160 0.160 0.161 0.162 0.162 0.162 0.153 0.154 0.156 0.157 0.157 0.154

NO 0.199 0.210 0.211 0.191 0.202 0.236 0.228 0.216 0.215 0.223 0.235 0.242 0.231 0.242 0.214 0.218 0.227 0.225 0.220 0.214

PL 0.190 0.185 0.184 0.180 0.178 0.177 0.172 0.171 0.175 0.183 0.180 0.181 0.180 0.179 0.176 0.175 0.178 0.179 0.177 0.178

PT 0.165 0.166 0.165 0.163 0.161 0.157 0.156 0.153 0.151 0.149 0.147 0.148 0.149 0.147 0.143 0.146 0.146 0.148 0.148 RO 0.215 0.219 0.218 0.209 0.201 0.200 0.206 0.203 0.195 0.199 0.196 0.195 0.188 0.183 0.184 0.194 0.196 0.191 0.193 0.189

RS 0.200 0.192 0.186 0.192 0.194 0.202 0.200 0.187 0.181 0.176 0.176 0.175 0.173 0.173 0.173 0.175 0.177 0.180 0.184 0.181

RU 0.211 0.217 0.214 0.215 0.222 0.223 0.212 0.201 0.198 0.212 0.220 0.218 0.213 0.208 0.196 0.200 0.206 0.192 0.191 0.190

SE 0.176 0.173 0.175 0.176 0.175 0.177 0.174 0.171 0.170 0.170 0.170 0.170 0.170 0.167 0.159 0.166 0.165 0.161 0.159 0.157

SI 0.189 0.188 0.189 0.190 0.188 0.186 0.186 0.185 0.185 0.184 0.182 0.181 0.181 0.176 0.168 0.171 0.174 0.176 0.179 0.181

SK 0.199 0.194 0.186 0.188 0.183 0.189 0.193 0.184 0.187 0.189 0.189 0.187 0.187 0.183 0.168 0.176 0.178 0.178 0.176 0.179

TR 0.202 0.197 0.196 0.206 0.198 0.196 0.194 0.193 0.194 0.193 0.193 0.192 0.190 0.189 0.185 0.187 0.190 0.187 0.186 0.187

UA 0.228 0.219 0.213 0.218 0.229 0.215 0.205 0.205 0.203 0.200 0.204 0.203 0.202 0.202 0.191 0.193 0.196 0.192 0.188 0.189

UK 0.177 0.178 0.174 0.170 0.167 0.167 0.163 0.158 0.155 0.153 0.152 0.151 0.149 0.150 0.146 0.147 0.148 0.147 0.149 0.147

XK

0.144 0.142 0.169 0.170 0.171 0.164 0.165 0.167 0.164

39