Current accounts in the European Union and the sectoral influence: an empirical assessment.
Gehringer, Agnieszka
INTRODUCTION
The investigation of the main causes of the global economic
meltdown in the aftermath of the recent financial crisis led to focus on
the role that current account imbalances might have played in
instigating the negative developments (Obstfeld and Rogoff, 2009). The
discussion on current account imbalances intensified also in Europe
(Knedlik and von Schweinitz, 2012).
The question of what determines the current account positions and
current account imbalances in Europe has started to attract attention in
the economic literature only in the recent years. Different studies have
brought to light specific evidence on the European--and more precisely
on the euro area--economy (Rahman, 2008; Jaumotte and Sodsriwiboon,
2010; Belke and Dreger, 2013; Ca' Zorzi and Rubaszek, 2012;
Gehringer, 2015). Whereas the contributions concerning the euro area are
not lacking, only limited effort has been made to examine the European
Union (EU) in a unified framework, while at the same time accounting for
the heterogeneity across the EU (Gehringer, 2015).
The present paper extends the evidence concerning the EU. In this
framework, a central contribution of the paper concerns the current
account determinants themselves. Although, for completeness, the
standard determinants of the current account identified in the past
investigations--as for instance Chinn and Prasad (2003)--are included,
the main focus of the empirical investigation is on the new determinants
that refer to the relevance of sector-level activities in the creation
of the economy-wide value added. A distinction is made in the first
place between the aggregate manufacturing sector, business services and
non-business services. In the second place, a more detailed sectoral
perspective is adopted in order to understand the influence exercised
especially by the construction sector and renting
activities--representing, respectively, the least and the most
productive non-manufacturing sectors. In this way, the relative
intensity and economic importance of different sectors is better
understood, and the direction and magnitude of their impact on the
current account positions can be assessed.
From the theoretical perspective, Giavazzi and Spaventa (2010)
argue that if the impact on the current account, even if transiently
negative, comes from tradable sectors or sectors with high productivity,
there is not much reason to be worried about the intertemporal budget
constraint. Indeed, the accumulated external debts could be repaid with
the future gains from trade. More precisely, in modeling the
intertemporal budget constraint, Giavazzi and Spaventa (2010)
distinguish between tradable and non-tradable goods, and observe that
low productivity in both sectors implies a more stringent intertemporal
budget constraint. Especially critical is seen investment in
non-tradables, as it can be treated analogously to current consumption.
Financing of such investment through the accumulation of external debt
might imply problems with the debt service in the future.
In the current analysis, both tradability and productivity issues
are accounted for. Accordingly, whether a sector has a truly negative
impact on the current account should be considered together with its
economic potential in terms of productivity. In this sense, the most
serious concerns arise over the construction sector. Indeed, it is not
only non-tradable, but also low-productive. Given that the activity in
the sector was booming alongside the deterioration in the current
account positions in the EU, its negative role should not be undermined.
The results, indeed, confirm the negative association between the
construction sector and the current account balance. On the contrary,
business services--that are among the most productive service activities
exercised a positive impact on current account positions. This could
have occurred both directly, as these sectors are among tradables, and
indirectly, by sustaining the activity in other tradable sectors.
This is the first contribution investigating the sectoral impact on
current accounts in such a broad sense. In the past literature, separate
studies dedicated some attention to housing market influence on current
account (Aizenman and Jinjarak, 2009; Gete, 2009; Punzi, 2007; Punzi,
2013). Generally, these studies confirm that a construction investment
boom enhances current account deterioration. Apart from this specific
evidence, the related literature says nothing about the contribution of
other sectors to the explanation of the current account. The present
paper fills this gap.
The paper is structured as follows. The next section recalls the
literature on the standard as well as the new set of sectoral
determinants of the current account. The section 'Data issues and
descriptive statistics' opens the empirical investigation with the
description of the facts and developments in the sectoral structures of
the European economy, followed by the discussion of the estimation
strategy in the section 'Estimation strategy'. Subsequently,
estimation results are presented (section 'Results from OLS panel
estimations') and strengthened by two important sensitivity checks
(see section 'Sensitivity analysis'). Finally, the last
section summarizes the paper.
STANDARD DETERMINANTS AND THE INFLUENCE OF THE SECTORAL COMPOSITION
ON THE CURRENT ACCOUNT
Before the theoretical discussion regarding the novel contribution
of the paper concerning the influence of the sectoral composition on the
current account, it is useful to briefly summarize the standard
determinants of the current account position.
There is more than one approach to deal with forces determining the
current account. From these different approaches, a standard set of
variables has emerged. Among the most influential works in this context
are the papers by Debelle and Faruqee (1996), Calderon et al. (2002),
Chinn and Prasad (2003), Bussiere et al (2006) and De Santis and
Luhrmann (2008). This strand of the literature plays a crucial role in
understanding the medium-term dynamics of current accounts in a broad
context of both developing and industrialized economies.
The main baseline current account model refers to the intertemporal
approach, initially proposed by Sachs (1981) and Buiter (1981), and
subsequently elaborated by Obstfeld and Rogoff (1995a) as well as
Gandolfo (2001). Under the permanent income hypothesis and rational
expectations, the current account is viewed as an outcome of
forward-looking, perfectly smoothing consumption and investment
decisions.
On the basis of such theoretical underpinning, the most relevant
determinants of current accounts normally include factors that through
savings and/or investment decisions impact the current account. The
customary set of determinants includes the initial stock of net foreign
assets (NFA), government budgets both variables expressed in relative
terms to GDP--relative income, demographic variables (young- and old-age
dependency ratios) and real exchange rate. (1)
Regarding NFA, Chinn and Prasad (2003) observe that, especially for
industrialized countries, a significantly large initial stock of NFA is
associated with large current account surpluses. On the other hand,
however, a high initial level of indebtedness might suggest that the
economy will try to prevent the insolvency episode, so that its current
account position should improve.
The theoretical explanation of the link between the government
balance and the current account, although dependent on the degree to
which consumers react in accordance with the Ricardian equivalence,
implies a positive relationship between both variables. An
increase/decrease in the government balance would make more/less
national savings available, with the consequent
improvement/deterioration of the current account. (2)
The relative income variable refers to the stage of development
assumption, according to which a less-developed country normally runs
current account deficits that will be repaid with future current account
surpluses, once the economy reaches a pattern of development typical of
advanced economies.
The two demographic variables express relative population
dependency of the young and old generations, respectively. The higher
this dependency the lower are the savings and the worse the outcome in
terms of the current account balance.
Concerning real exchange rate, its theoretical link to the current
account has more recently emerged in the context of the new open economy
macroeconomics models (eg, Obstfeld and Rogoff, 1995b). (3) The current
account deficits may reflect a loss in competitiveness, which is
mirrored in the appreciation of the real exchange rate (Arghyrou and
Chortareas, 2008; Belke and Dreger, 2013). This is because real exchange
rate appreciation could redirect demand from domestic to foreign
markets.
Such standard determinants should play a role independently of a
geographic context, but also in specific country groupings. Indeed, some
authors undertook the effort to explain current account positions in
different sub-samples of countries, with a number of works already
dedicated to the euro area (Jaumotte and Sodsriwiboon, 2010; Belke and
Dreger, 2013). When dealing with particular country groupings, it is
advisable to extend the set of standard variables to consider more
specific factors. Accordingly, Glick and Rogoff (1995) argue that
focusing not only on global but also on country-specific events is by no
means irrelevant. Such factors might be important in determining
country-specific productivity dynamics, with further consequences on the
current account. They find that the current account responds in a highly
sensitive way even to small changes in the degree of mean reversion
observable for country-level productivity development. (4)
Considering the EU context of the paper, there are specific
country-level and regional characteristics worth investigation. First,
the process of economic and monetary integration should play a role in
determining the current account. It has often been argued that the
institution of a supranational community and the contemporaneous
elimination of economic and financial barriers via the common market
could and should translate into more favorable conditions to get finance
(Lane and Milesi-Ferretti, 2008; Jaumotte and Sodsriwiboon, 2010;
Gehringer, 2013). Specifically, the creation of the European Monetary
Union and the introduction of the single currency considerably reduced
external constraint, removed the exchange rate risk and permitted
interest rates to become almost insensitive to domestic developments. As
a consequence, the increasing dispersion in current accounts should not
be viewed negatively, as it would derive from more efficient allocation
of resources and would be of a temporary nature. Once the economies
running current account deficits develop and establish a strong tradable
sector, they should be able to repay accumulated external debt.
Although the macroeconomic variables that form the standard set of
current account determinants are crucial, they are not exhaustive in
explaining the current account developments. Indeed, it is crucial to
consider other factors that pertain to the composition and quality of
economic activity. The analysis by Giavazzi and Spaventa (2010) is
insightful here. They examine conditions under which the intertemporal
budget constraint of an open economy can be fulfilled. On the basis of a
simple analytical model, their main conclusion is that so long as an
economy, even if actually running current account deficits, avoids
inefficient allocation of resources in excessive consumption and/or in
investments in non-tradable sectors, current account deficits are a
natural consequence of the growth process. Two issues need more
specification here. This general conclusion should not imply that any
investment in non-tradables is undesirable, as it would undermine a
crucial role played by some non- or at least less-tradable service
sectors. Moreover, although the dichotomy of tradables/non-tradables is
an indisputable matter of interest in examining the conditions of the
current account position of an economy, it still provides scarce
indication on the sectoral efficiency and its economic importance for
the rest of the economic system. The assessment of the impact of a
sector on the current account requires considering its productivity
dynamics, and thus its growth potential. Indeed, even a non-tradable
sector might have a positive and long-lasting impact on the current
account, if it stimulates domestic growth through the productivity
channel.
This is particularly the case of some service sectors.
Specifically, the so-called knowledge-intensive business services (in
short, KIBS) play a crucial complementary role in sustaining
efficiency-driven economic growth, both directly, as the drivers of
innovations, and indirectly, supporting the rest of the economic
activity. In this sense, it has been argued that such services play a
pivotal role as facilitators or even as co-producers of economy-wide
innovation (den Hertog, 2000; Macpherson, 2008). Apart from KIBS, other
less knowledge-intensive but business-near services, like transport and
storage or post and telecommunication, are also crucial due to the
strong linkages and interactions with the rest of the economy. Without
such efficiently functioning services, the activity of other sectors--be
they tradable or non-tradable would be considerably jeopardized. (5)
On the other hand, the case of the construction sector is quite
problematic. As will become clear in the next section, the sector is
among the ones with the lowest productivity dynamics. Moreover, it is
not only non-tradable, but it is highly dependent--at least relative to
the other non-tradable sectors--on the imports of intermediate inputs.
(6) Consequently, housing booms and periods of intensified activity in
construction might have disruptive consequences on the intertemporal
budget constraint. (7)
Regarding the housing market, Gete (2009) shows in a theoretical
model that shocks to the demand for housing contribute to trade deficits
through intensified imports of tradable goods. Other studies confirming
a link between housing market dynamics and current accounts include
Punzi (2007) and Punzi (2013). More precisely, an exogenous upward shift
in housing demand, due to the shift of the domestic preferences toward
housing, would imply a reallocation of labor from other tradable and
more productive sectors. Such an increase in housing production implies
the opportunity cost of lower production of tradable goods. This cost,
nevertheless, can be diminished in an open economy context, thanks to
the possibility of importing consumer tradable goods (Gete, 2009). The
trade balance deteriorates due to intensified imports of consumable
goods. Moreover, given that domestic production of tradable goods
diminishes, the trade balance is further aggravated through diminishing
exports.
DATA ISSUES AND DESCRIPTIVE STATISTICS
The database constitutes a panel referring to 20 EU members in the
period 1995-2009. The time span considered is the longest possible based
on data availability issues. In the estimation framework, data
transformed into 5-year non-overlapping averages are applied. This is
done to account for cyclical influences. The transformation, however,
has a possible disadvantage of reducing the number of observations and
thus the bulk of information contained in annual data. For that reason,
in the sensitivity analysis, I validate the baseline model by re-running
all specifications on annual data.
Statistical information on the dependent variable (current
account), government balance (gov balance), the initial value of net
foreign assets [NFA], as well as both demographic dependency ratios
(oldjage and young_age), is taken from Eurostat. Both the current
account and government balance variables are expressed in percentage of
GDP. The NFA variable is given by the difference between foreign assets
and liabilities and set in percentage of GDP. Old-age dependency ratio
expresses the share of the old-age population (65 +) over the young
population (15-64). With the aim of overcoming possible
multicollinearity problems, the young-age dependency ratio is expressed
in terms of the population growth rate. The relative income measure,
obtained as real GDP per capita of each country to real GDP per capita
of the US both in purchasing power parity, is based on data from the
World Economic Outlook. (8) The real effective exchange rate is an index
(2000= 1.0)--taken from AMECO macroeconomic database--based on unit
labor cost and measuring performance relative to the rest of 35
industrial countries, with double export weights. (9) Finally, the
variables measuring the relative sectoral economic incidence have been
calculated for each sector as a share of this sector's value added
in total value added. The sectoral data to calculate these ratios are
taken from OECD STAN Database for Structural Analysis. (10)
Regarding the sectors' composition, a distinction is made
between an aggregate category of manufacturing sector and two service
sectors groups, BUSS and non-BUSS, according to the explanation offered
in Appendix Al. In addition, in estimations regarding the whole EU, the
two service categories are split into more disaggregated sectors. In
particular, within the category of BUSS, the influence on the current
account coming from its three components, namely, electricity, gas and
water supply; financial intermediation and renting activities is
considered. The category of non-BUSS is replaced by construction sector,
and wholesale and retail trade.
In addition, more in-depth analysis of the sectoral influence that
might differ between four roughly homogenous country groups within the
EU is performed. For that reason, in separate estimations, apposite
interaction terms constructed as a product between the specific
country-group dummy (core, GUPS, East and non-euro) and each of the
sectoral variables [BUSS, Non-BUSS, construction and renting) are
included. In particular, the core euro area includes Austria, Belgium,
Finland, France, Germany, the Netherlands; GUPS refers to Greece,
Ireland, Italy, Portugal, Spain; to the East group belong the Czech
Republic, Estonia, Hungary, Poland, Slovakia, Slovenia; finally, the
noneuro group is composed of Denmark, Sweden and the United Kingdom.
Tradability, productivity and the sectoral dependence on imports
Before turning to the main estimation framework, it is useful to
shortly review two features characterizing any sectoral system,
tradability and productivity. This analysis will be crucial in
understanding the forthcoming empirical exercise.
It is indisputable that the division between tradable versus
non-tradable sectors is of central importance to the economic theory.
(11) This notwithstanding, there has been little empirical investigation
concerning this division, mostly due to the insufficient availability of
data. The conventional classification between tradable and non-tradable
sectors was usually corresponding to the commonly applied distinction
between manufacturing sectors belonging to the first group and services
to the second group. But this distinction is losing its significance,
given the growing importance of services for international trade.
As argued by De Gregorio et al. (1994), a natural benchmark for
tradability is given by the degree to which a certain item is actually
traded. Consequently, they adopt the ratio between total exports of a
sector across all 14 OECD countries taken into analysis to the total
output of that same sector, with the threshold of more than 10 per cent
to classify a sector as tradable. They admit that this procedure is
sensitive to the arbitrary choice of the threshold. (12)
Following the methodology by De Gregorio et al. (1994), I calculate
a measure of tradability separately for four groups of countries within
the EU. In general, almost all service sectors--with the exception of
transport and storage, as well as financial intermediation for GIIPS,
electricity, gas and water supply for non-euro, and renting activities
for core and East--are limitedly involved in international trade, and
thus can be labeled as non-tradables. Conversely, the aggregate class of
manufacturing sectors easily overcomes the 10 per cent benchmark and
could be treated as tradable. Thus, from this short analysis it can be
concluded that the traditional distinction between
tradables-manufacturing and non-tradables services can be confirmed (for
more details, see Appendix A2).
From the above analysis still little can be said about the quality
and the magnitude of the sectoral impact on the current account. This is
mainly because the measure of tradability merely says whether a sector
is involved in the international trade through relatively high export
rate. More precisely, as already stated in the previous section, even if
a sector is non-tradable, its economic importance for the economic
system, deriving particularly from its growth potential, should mitigate
the worries about its negative impact on the current account. On the
other hand, if a low-productive sector dominates, or at least
significantly influences economic activity of a country, sustainability
of the intertemporal budget constraint might be impaired. This influence
on the current account derives from the fact that higher investment in a
low-productive sector attracts investment away from more productive
activities, thus it directly deteriorates both current and future
current account positions.
It is often the case that comparing data on tradability and
productivity growth, sectors performing well in terms of productivity
dynamics are also those stronger involved in international trade (for
details on tradability and productivity growth data, see Appendix A2).
This is particularly true for manufacturing sectors and also for
business-related activities. Conversely, especially with regard to
construction, low or even negative productivity growth combined with low
tradability should be a matter of concern for an economy involved in a
construction boom. If an intensified investment in sectors, such as
construction, leads to lower-tradable goods production, the resulting
current account deterioration is a matter of concerns. The opposite is
valid for the BUSS category, which is dominated by tradable and
productive sectors. Consequently, current investment in BUSS, even if
contributing to the deterioration in the current account position, might
be crucial to create the necessary business environment, with
long-lasting positive returns generated by sectors profiting from BUSS.
Thus, the potential for the future repayment of actual net liabilities
would be increased.
The validity of this conclusion clearly depends on the relative
size of investment made in a particular sector. More precisely, in
sectors like construction, with low-productivity growth combined with
non-tradability, intensified and to some extent excessive investment
activities might have contributed to deteriorating current account
positions in some European economies. This can be seen from the analysis
of data reported in Figure 1, where sector-level contribution to the
growth of value added is reported. Consequently, in the forthcoming
estimation framework, and keeping in mind the tradability and
productivity performance of each sectoral category, I will use the
sectoral contribution to the total value added as a main explanatory
variable in order to empirically assess the sectoral impact on the
current accounts.
From Figure 1, the picture for construction and for the wholesale
and retail trade sector is particularly striking, as it shows the great
importance of these sectors in the value-added growth, especially for
GUPS and for the Eastern European countries. This growth, extensively
based on a construction boom, might have been possible for the GUPS
countries, thanks to more favorable financing conditions after the euro
adoption. For Eastern European countries, the prospect of the EU
accession and successful institutional convergence to the EU norms in
the pre-accession period led to intensified investment flows from abroad
that were increasingly allocated in housing and other private services
(like wholesale and retail trade), often tightly linked to the
construction sector.
[FIGURE 1 OMITTED]
ESTIMATION STRATEGY
Following past empirical contributions, I first apply pooled OLS on
data transformed into 5-year non-overlapping averages. This strategy was
often used in this framework in order to overcome possible cyclicality
problems with annual data (Calderon et al, 2002; Chinn and Prasad,
2003). I am also aware of possible endogeneity problems of my
specifications, especially regarding the government budget variable.
(13) In this context, when considering specifications with
country-groups variables, I apply the principal component analysis
(PCA). This method should overcome the problem by replacing the standard
set of determinants with only two components. Specifically, the impact
of the standard determinants is accounted for in the two principal
components rather than directly, which is the main source of
endogeneity. As an additional sensitivity check, IV estimations of the
main specifications are performed. Finally, another suitable method to
deal with endogeneity concerns would be through generalized methods of
moments methodology that, however, requires a sufficient number of
groups, surpassing the number of instruments --a condition that is
difficult to satisfy here, even by applying methods to reduce the number
of instruments. As a consequence, when interpreting the results from the
pooled OLS estimations, it is crucial to bear in mind that these
represent associations rather than causal relations.
The baseline specification for all estimations assumes the
following form:
[ca.sub.kt] = [[beta].sub.1] + [[beta]'.sub.2] [X.sub.kt] +
[[beta]'.sub.3][Z.sub.kt] + [[tau].sub.t] + [[epsilon].sub.kt] (1)
where [ca.sub.kt] is current account in percentage of GDP in
country k at time t. Vector [X.sub.kt] includes the standard
determinants of the current account. Moreover, vector [Z.sub.kt]
contains a set of explanatory variables referring, alternatively, to the
sectoral importance in the generation of value added, to interaction
terms of these variables with the country groups' dummies or to the
two euro-country-groups interaction terms [euro*core and euro*GUPS).
Finally, [[tau].sub.t] and [[epsilon].sub.kt] refer to time dummies and
to idiosyncratic error term, respectively. (14)
RESULTS FROM OLS PANEL ESTIMATIONS
On the basis of equation 1, different specifications according to
the pooled OLS method are estimated. The results are reported in Tables
1 and 2. The first specification in Table 1 considers only the standard
determinants of the current account. The second one adds two dummy
variables, euro*core and euro*GIIPS, being one for core, or
alternatively, GUPS countries after 1999 and null otherwise. Columns 3
and 4 investigate more precisely the sectoral composition of production
and its influence on the current account. In particular, Column 3
considers an aggregate of all manufacturing sectors (manu), in addition
to business services (BUSS) and non-business services [non-BUSS).
Moreover, given the focus lying in disentangling
country-group-specific impact generated by the sectoral variables,
further regressions are run, the results of which are summarized in
Table 2. These regressions are based on the similar baseline
specification as in equation 1, where the standard set of current
account determinants are replaced with two principal components obtained
from PCA on such standard determinants. More precisely, PCA permits to
reduce the number of the explanatory variables, without losing relevant
information that is compressed now in the two components. (15) In
addition to the principal components, in each specification, a set of
interaction terms between one of the sectoral variables--manufacture,
BUSS, non-BUSS, construction and renting--and a specific country-group
dummy variable is considered.
Among the standard determinants of the current account, government
balance signs significantly positive influence, giving a strong support
to the twin deficit hypothesis. (16) Also NFA and relative income
variable significantly influence the current account. These results are
in line with the past investigations, especially those explicitly
concerning the subgroup of industrialized countries (see, for instance,
Chinn and Prasad, 2003). Finally, the lack of evidence obtained for
young- and old-dependency ratios confirm the previous results (Chinn and
Prasad, 2003; Chinn and Ito, 2007).
When controlling for the euro adoption, a significantly negative
influence for GIIPS and a positive but insignificant evidence for the
core euro area is obtained. This might support the hypothesis that with
the introduction of the euro, less-developed countries of the newly
created Eurozone faced better financing conditions that contributed to
higher current account deficits. As argued in the previous discussion,
this might be to a certain extent considered as a natural and desirable
consequence of the intensified European monetary integration process.
Considering, however, that those resources went not only to finance
profitable investment, but also extensively current consumption and
investment in sectors being among the least productive, such current
account deficits should be viewed with caution.
Supportive to this general statement are the results from Table 2,
where in Columns 3 and 4 the coefficients corresponding to construction
and non-BUSS in GIIPS are significantly negative. Precisely, a 1
percentage point increase in the contribution of the construction and
non-BUSS sectors to the overall value-added creation implies
deterioration in the current account balance by 0.7 and 0.5 percentage
point, respectively. It is worth noting that those results might have
more in common than only the sign of the coefficient. Indeed, the
intensified activities in the construction sector were provoking growing
activities also in some other less-productive non-business services, in
particular in wholesale and retail trade, and real estate. Finally,
among non-BUSS there is also the category of community services, and in
particular public administration sector that especially in the Southern
EU members was source of not always efficient investment decisions.
A similar conclusion could be drawn for the core euro area and for
East, at least regarding the developments in the construction sector.
They resulted in a strongly negative impact on current accounts.
Nevertheless, the influence of non-BUSS in the core euro area appeared
to be positive, whereas in East the variable did not report any
significant evidence. Finally, regarding renting activities, they
appeared to produce a positive impact in the core euro area and negative
in GIIPS. But due to the high-productivity performance and an important
economic role of this sector, the negative sign in GIIPS should not
provoke concerns.
Finally, no particular conclusion can be drawn regarding non-euro
country group. Indeed, neither of the coefficients appeared to
significantly contribute to the explanation of their current accounts.
SENSITIVITY ANALYSIS
In the main specifications, country fixed effects are excluded in
order to permit the cross-country variation in the explanatory variables
to contribute to the explanation of the current account patterns in the
EU. Indeed, for the majority of the covariates, the between-group
variability was significantly contributing to the overall standard
deviation. Accounting explicitly for country-specific effects would thus
subtract from the explanatory power of the variables included in the
baseline specifications, both EU-aggregate and country-group specific.
To check this hypothesis, in Columns 1-4 of Table 3, the results from
the fixed-effects estimations analogous to the previously discussed
pooled regressions (in Table 1) are shown.
As already anticipated, the results reported in Table 3 show that
especially for the standard determinants of the current account their
explanatory power became weaker. Also in the case of the sector-specific
influence for some variables, like manufacturing and finance, no
statistically significant influence could be confirmed. But the previous
conclusions regarding construction, as well as the euro variable for
GIIPS, are corroborated.
Finally, the use of non-overlapping averages is desirable under the
potential cyclicality of the data. On the contrary, the inclusion of
annual data increases their frequency, and thus the number of
observations. In order to verify the influence of the choice of data
frequency on the baseline results, the analogous specifications as in
Table 1 on annual data are re-estimated.
Columns 5-8 of Table 3 summarize the results that are rather
supportive to the outcome of the baseline analysis. Generally, the sign
and the strength of the influence remained unchanged for all variables
and independently of the specification. This confirms the findings in
the previous section.
CONCLUSIONS
The recent developments in the current accounts in the EU and more
precisely the excessive accumulation of external debt in the pre-crisis
period (Lane and Milesi-Ferretti, 2012) assume importance in the
political and scientific discussion. As imbalances within the EU surge,
questions regarding their sources as well as the limits of such
developments are increasing. The relevance of the issue has been
dampened by the rebalancing of current accounts in the more recent
years, but such rebalancing is illusive, as the causes of the past
imbalances have not been properly addressed yet.
The aim of the paper was to provide new insights to the factors
determining the current account in the EU. The results offer sustain to
some of the standard determinants, with the fiscal balance variable
reporting the strongest evidence. But the main focus of the analysis was
on new sector- and country-group-specific effects. Jointly with the
descriptive analysis of the sectoral tradability and productivity
dynamics of the past decades, the aim was to assess the current account
development in a more qualitative way. Precisely, given that there are
sectors, particularly construction, characterized by non-tradability and
low productivity, it was crucial to verify whether their strong role in
the creation of value added would have negative impact on the current
account balances.
The results based on the implementation of such a new set of
sector-specific explanatory variables suggest that the deterioration in
the current account positions in the EU could be to a large extent
explained by strong activity in the construction sector. A closer look
at the within-EU heterogeneity permits to assess that this result is
driven by GIIPS, East, core euro area, but not by non-euro members. The
evidence regarding GIIPS goes hand in hand with the construction boom
and the subsequently bursting real estate bubble. But also the evidence
regarding the core euro area and Eastern Europe, even if not accompanied
by the burst of an asset bubble, should lead to reflections on the
quality of external debts accumulated in the past decades. Consequently,
this finding points to some ill-conditioned developments of the current
account positions.
Finally, negative evidence could be also confirmed for BUSS and
non-BUSS services. Some concerns remain regarding non-BUSS services,
especially on the ground of their relatively low-productivity dynamics.
Instead, given that BUSS services revealed positive productivity growth
rates in the past decades, their qualitative impact on the current
account should not create much concerns.
Acknowledgements
The author is thankful to Menzie Chinn, Gian Maria Milesi-Ferretti,
Alberto Bagnai, the participants of the Conference 'Intra-European
Imbalances, Global Imbalances, International Banking, and International
Financial Stability' and of the INFER Annual Conference 2013 for
their insightful comments and suggestions to the previous version of the
paper. The major part of research on the paper has been done when the
author was working at the University of Gottingen, Germany.
REFERENCES
Arghyrou, MG and Chortareas, G. 2008: Real exchange rate and
current account imbalances in the euro area. Review of International
Economics 16(4): 747-764.
Aizenman, J. and Jinjarak, Y. 2009: Current account patterns and
national real estate markets. Journal of Urban Economics 66(2): 75-89.
Bagnai, A. 2006: Structural breaks and the twin deficits
hypothesis. International Economics and Economic Policy 3(2): 137-155.
Balassa, B. 1964: The purchasing-power parity doctrine: A
reappraisal. Journal of Political Economy 72(6): 584-596.
Belke, A and Dreger, C. 2013: Current Account Imbalances in the
Euro Area: Does Catching up Explain the Development? Review of
International Economics 21(1): 1-6.
Buiter, WH. 1981: Time preference and international lending and
borrowing in an overlapping generations model. Journal of Political
Economy 89(4): 769-797.
Bussiere, M, Fratzscher, M and Muller, G.J. 2006: Current account
dynamics in OECD countries and in the new EU member states: An
intertemporal approach. Journal of Economic Integration 21(3): 593-518.
Ca' Zorzi, M, Chudnik, A and Dieppe, A. 2012: Thousand of
models, one story: Current account imbalances in the global economy.
Journal of International Money and Finance 31 (6): 1319-1338.
Ca' Zorzi, M and Rubaszek, M. 2012: On the empirical evidence
of the intertemporal current account model for the euro area. Review of
Development Economics 16(1): 95-106.
Calderon, CA, Chong, A and Loayza, N. 2002: Determinants of current
account deficits in developing countries. Contributions to
Macroeconomics 2(1): 1-31.
Chinn, MD and Ito, H. 2007: Current account balances, financial
development and institutions: Assaying the world 'saving
glut'. International Journal of Money and Finance 26(4): 546-569.
Chinn, MD and Prasad, ES. 2003: Medium-term determinants of current
accounts in industrial and developing countries: An empirical
exploration. Journal of International Economics 59(1): 47-76.
De Gregorio, J, Giovannini, A and Wolf, HC. 1994: International
evidence on tradables and nontradables inflation. European Economic
Review 38(6): 1225-1244.
De Santis, R and Luhrmann, M. 2008: On the determinants of net
international portfolio flows: A global perspective. Journal of
International Money and Finance 28(5): 890-901.
Debelle, G and Faruqee, H. 1996: What determines the current
account? A cross-sectional and panel approach. IMF Working Paper No.
96/58, International Monetary Fund.
Den Hertog, P. 2000: Knowledge intensive business services as
co-producers of innovation. International Journal of Innovation
Management 4(4): 491-528.
Gandolfo, G. 2001: International finance and open-economy
macroeconomics. Springer: New York.
Gehringer, A. 2013: Growth, productivity and capital accumulation:
The effects of financial liberalization in the case of European
integration. International Review of Economics and Finance 25: 291-309.
Gehringer, A. 2015: New evidence on the determinants of current
accounts in the EU. Empirica: Journal of European Economics 42(4):
769-793.
Gete, P. 2009: Housing markets and current account dynamics.
University of Chicago, dissertation.
Giavazzi, F and Spaventa, L. 2010: Why the current account matters
in a monetary union. CEPR Discussion Paper No. 8008, Centre for Economic
Policy Research.
Glick, R and Rogoff, K. 1995: Global versus country-specific
productivity shocks and the current account. Journal of Monetary
Economics 35(1): 159-192.
Jaumotte, F and Sodsriwiboon, P. 2010: Current account imbalances
in Southern euro area. IMF Working Paper No. WP/10/139, International
Monetary Fund.
Kahn, MS and Knight, MD. 1983: Determinants of current account
balances of non-oil developing countries in the 1970s: An empirical
analysis. IMF Staff Papers 30(4): 819-842.
Kim, S and Roubini, N. 2008: Twin deficit or twin divergence?
Fiscal policy, current account and real exchange rate in the U.S..
Journal of International Economics 74(2): 362-383.
Knedlik, T and von Schweinitz, G. 2012: Macroeconomic imbalances as
indicators for debt crises in Europe. Journal of Common Market Studies
50(5): 726-745.
Lane, PR and Milesi-Ferretti, GM. 2008: The drivers of financial
globalization. American Economic Review 98(2): 327-332.
Lane, PR and Milesi-Ferretti, GM. 2012: External adjustment and the
global crisis. Journal of International Economics 88(2): 252-265.
Macpherson, A. 2008: Producer service linkages and industrial
innovation: Results of a twelve-year tracking study of New York State
manufacturers. Growth and Change 39(1): 1-23.
Obstfeld, M and Rogoff, K. 1995a: The intertemporal approach to the
current account. In: Grossman, GM and Rogoff, K (eds). Handbook of
International Economics, Volume 3. Elsevier: North-Holland,
pp.1731-1799.
Obstfeld, M and Rogoff, K. 1995b: Exchange rate dynamics redux.
Journal of Political Economy 103(3): 624-660.
Obstfeld, M. and Rogoff, K. 2009: Global imbalances and the
financial crisis: Products of common causes. CEPR Discussion Paper No.
7606.
Punzi, M.T. 2007: The link between housing prices and current
account deficit: A study of 10 OECD countries. Boston College, mimeo.
Punzi, M.T. 2013: Housing market and current account imbalances in
the international economy. Review of International Economics 21 (4):
601-613.
Rahman, J. 2008: Current account developments in new member states
of the European Union: Equilibrium, excess, and EU-phoria. IMF Working
Paper No. 92, International Monetary Fund. Sachs, JD. 1981: The current
account and macroeconomic adjustment in the 1970s. Brooking Papers of
Economic Activity 1981(1): 201-268.
Samuelson, PA. 1964: Theoretical notes on trade problems. Review of
Economics and Statistics 46(2): 145-154.
APPENDIX
A1: Classification of service sectors between BUSS and non-BUSS
For the purpose of the empirical investigation, I distinguish all
service sectors in two groups according to the 'supplier's
role criterion'. To apply this criterion, I calculated the ratio
between the sum of the intermediate inputs delivered to the other
sectors over the total output produced by that sector for each service
sector in each analyzed country and in each year. In the second step, I
computed average ratios over time and between countries (electricity,
gas and water supply--0.410; construction--0.278; wholesale and retail
trade--0.165; transport and storage--0.463; post and
communication--0.578; financial intermediation--0.567; real
estate--0.076; renting activities--0.671; community services--0.134).
Finally, I established 0.3 as a threshold value: the value of the
average ratio higher/lower than 0.3 classified a sector as a BUSS/
non-BUSS sector. To calculate those ratios, I used World Input-Output
Database (WIOD) annual input-output tables.
A2: Tradability and productivity of sectors in four country groups
Applying the methodology by De Gregorio et al. (1994), tradability
of industrial sectors has been calculated, as reported in Table Al.
Although a commonly adopted distinction between
manufacturing-tradable and services-non-tradable holds, one observation
is due here.
Within country groups, a certain degree of variability in the
calculated measure of tradability can be observed, as expressed by the
standard deviation. This confirms the previous intuition that it is
difficult to generalize the quality of tradable or non-tradable sectors,
as significant differences for particular sectors might sometimes occur
across countries. This is, for instance, the case for the renting
activities that in France and Germany would be clearly considered as
non-tradable (5.8% and 6.4%, respectively), whereas it is tradable in
all the remaining countries of the core euro area. Another example
refers to the construction sector, which in Hungary and Poland appears
to be tradable (14.7% and 14.9%, respectively), whereas it is
non-tradable in the rest of the East.
Table A1: Tradability of sectors by country groupings in the EU
Averages over 1996-2009
Core East
Manufacture 99.0 (1.3) 96.7 (10.5)
Electricity, gas and water supply 8.0 (7.1) 4.8 (2.9)
Construction 3.3 (3.5) 7.2 (6.0)
Wholesale and retail trade 5.8 (4.1) 8.1 (6.5)
Transport and storage 25.7 (12.3) 27.5 (9.6)
Post and telecommunication 9.6 (6.2) 7.3 (1.9)
Financial intermediation 9.3 (6.8) 6.5 (2.3)
Real estate 0.3 (0.3) 0.9 (0.4)
Renting activities 16.2 (7.9) 12.9 (4.9)
Public administration and defence 1.2 (0.9) 1.2 (0.8)
Education 0.8 (1.0) 0.7 (0.7)
Social 0.2 (0.1) 1.7 (1.7)
Other service activities 4.1 (3.6) 6.5 (4.7)
BUSS 16.1 (7.6) 16.4 (5.7)
Non-BUSS 3.1 (2.2) 4.5 (2.4)
Averages over 1996-2009
GUPS Non-euro
Manufacture 98.8 (1.2) 84.9 (35.8)
Electricity, gas and water supply 1.2 (0.7) 12.3 (3.4)
Construction 1.3 (1.0) 0.3 (0.0)
Wholesale and retail trade 6.1 (6.7) 3.2 (0.3)
Transport and storage 29.0 (22.1) 62.1 (26.2)
Post and telecommunication 4.8 (1.9) 6.9 (1.7)
Financial intermediation 14.1 (21.6) 5.0 (1.0)
Real estate 0.5 (0.5) 0.1 (0.0)
Renting activities 7.2 (2.9) 8.3 (2.4)
Public administration and defence 0.5 (0.4) 0.6 (0.1)
Education 0.7 (0.8) 0.1 (0.0)
Social 0.3 (0.4) 0.1 (0.0)
Other service activities 2.6 (1.3) 2.9 (0.7)
BUSS 13.5 (8.3) 22.3 (4.4)
Non-BUSS 1.6 (0.7) 1.7 (0.3)
Note: Standard deviation in parentheses. BUSS are classified as
electricity, gas and water supply; transport and storage; post
and telecommunication; financial intermediation; and renting
activities. Non-BUSS refers to the remaining service activities.
The values reported for BUSS and non-BUSS express averages for
the respective groups.
Source: Own calculations based on WIOD
Table A2: Average Total factor and Labor productivity growth
rates by sectors of activity in the EU country groupings (in %)
Averages over 1996-2009 (3)
Core
TFP Labor Deflator
productivity
3.4 2.9 1.2
Manufacture 3.2 3.3 2.3
Electricity, gas and water
supply -0.3 0.1 2.7
Construction 1.5 0.9 1.3
Wholesale and retail trade 1.6 1.5 1.4
Business services (BUSS)
Averages over 1996-2009 (3)
East
TFP Labor Deflator
productivity
6.4 7.3 2.7
Manufacture 4.8 2.8 10.9
Electricity, gas and water
supply -1.9 1.0 6.8
Construction 2.9 4.2 5.1
Wholesale and retail trade 3.5 4.0 6.5
Business services (BUSS)
Averages over 1996-2009 (3)
GUPS
TFP Labor Deflator
productivity
0.9 2.3 2.6
Manufacture 0.1 2.5 2.2
Electricity, gas and water
supply -0.1 -0.1 4.6
Construction 0.3 1.1 3.0
Wholesale and retail trade 0.9 2.1 3.2
Business services (BUSS)
Averages over 1996-2009 (3)
Non-euro
TFP Labor Deflator
productivity
4.1 3.1 -0.1
Manufacture 0.7 1.3 2.8
Electricity, gas and water
supply -0.7 -0.3 4.2
Construction 2.1 1.6 1.6
Wholesale and retail trade 2.3 2.0 1.1
Business services (BUSS)
(a) Due to the data availability regarding total factor
productivity (TFP) growth, for East averages are calculated over
1999-2008, for non-euro TFP growth over 1997- 2007 and for GUPS
over 2001-2009. Data on the productivity growth in the other
service sectors, especially in the community services, were not
available. In addition to the productivity measures, for each
country group, I report the growth rates of the price deflator,
in order to verify whether the price dynamics might be justified
with the productivity increases.
Source: TFP and labor productivity growth - Own calculations
based on OECD STAN Database for Structural Analysis (Productivity
by Industry database). Price deflator growth rates refer to
either total output or value-added deflator, depending on data
availability for each country. Data on price deflation are based
on OECD STAN Database
(1) This set of variables is by no means exhaustive, but summarizes
well the usual factors included in the empirical investigations
concerning specifically industrialized countries, similarly as in
Debelle and Faruqee (1996) and Chinn and Prasad (2003). There are also a
number of studies, for instance, the panel investigation by Calderon et
al (2002) or a pooled longitudinal estimation by Kahn and Knight (1983)
regarding exclusively developing countries, in which also more specific
variables are included.
(2) In this sense, the Ricardian equivalence implies that the
perfectly foreseeing consumers anticipate the future increase in taxes
deriving from the current increase in the government expenditures and
answer with higher current savings. This implies no effect of the
variation in the government budget on the current account balance.
(3) Provided that nominal prices are fixed in the producer country
and the exchange rate pass-through is complete, the
'expenditure-switching' effect is valid in the Obstfeld and
Rogoff (1995a, b) Redux model.
(4) Recalling 'thousands of models' that are supposed to
tell 'one story', Ca' Zorzi et al. (2012) make a valuable
methodological effort to evaluate different current account models, and
find that in the period preceding the actual financial turmoil the
fundamentals possessed limited explanatory power over current accounts.
(5) It would be preferable to include in the empirical
investigation an explicit category of KIBS. Given, however, the problems
with data availability, I apply an ad-hoc classification of services.
More precisely, I distinguish between BUSS and non-BUSS, as explained in
the next section and in Appendix At.
(6) The ratio of imports directed toward the sector over its total
output accounts on ca 0.12 for construction and 0.07 for services.
(7) Ideally, the focus should be on construction of residential
buildings, kept separately from civil engineering. Indeed, the latter is
responsible for the creation of infrastructures that play a crucial role
for the general business activities. Unfortunately, data are not
available at this level of disaggregation, so that in the forthcoming
empirical analysis 1 refer to the entire construction sector. At the
same time, however, the anecdotal evidence of the past decades confirms
that major movements in the construction sector were due to residential
construction.
(8) The US is chosen as a benchmark, given its leading economic
position among industrialized countries.
(9) Real effective exchange rates are quoted using the indirect
quotation, meaning that an increase (reduction) of the index denotes a
real appreciation (depreciation).
(10) Given my focus on productivity, in alternative specifications,
I expressed the sectoral variables in terms of total factor productivity
(TFP) (levels and growth). These variables were mostly insignificant.
This might be due to the fact that by considering only productivity
might be insufficient, as it provides limited information on the
relative economic importance of a given sector. A VA-share variable is
more insightful in this respect. Consequently, if interpreted together
with the analysis of data on sectoral productivity, this variable should
provide the meaningful qualitative interpretation of the influence on
current account.
(11) As confirmed, for instance, by the formal analyses of Balassa
(1964) and Samuelson's (1964).
(12) The choice is also arbitrary to the extent that it considers
only exports and not also imports to determine tradability. Whereas in
some strands of the literature it might be reasonable to account for
imports as well (in order to know whether the price of a product/service
is set domestically or rather in the world markets), in the present
context it is more straightforward to concentrate on exportable. This is
because if a sector is tradable in the sense of being exportable, its
contribution to rebalancing the current account will be clearly
positive.
(13) See the discussion in Chinn and Prasad (2003) on problems with
finding the appropriate instruments for the government budget variable.
(14) The recent economic crises was accounted for in a separate
specification by introducing a dummy variable (equal to 1 for the period
after 2008 and 0 otherwise), but it was never significant.
(15) The two principal components are those with the highest
loading factors, the first with the value of 2.2 and the second 1.0. The
two components together manage to explain over 60% of variability of the
original variables, whereas the first factor contributed already with
40%. I validated the PCA with the Kaiser-Mexer-Olkin's measure of
sampling adequacy that supported the overall suitability of the data set
with a value of 0.6.
(16) The literature is still inconclusive on the validity of the
hypothesis. See, for instance, Bagnai (2006) and Kim and Roubini (2008).
ALBERTO BAGNAI (1) & CAMELIA TURCU (2)
(1) Department of Economics, Gabriele D'Annunzio University,
viale Pindaro 42, Pescara 65127, Italy.
E-mail: alberto.bagnai@unich.it
(2) University of Orleans, LEO (Laboratoire d'Economie
d'Orleans) UMR CNRS 7322, Rue de Blois, Orleans BP 2637, France.
E-mail: camelia.turcu@univ-orleans.fr
AGNIESZKA GEHRINGER
Flossbach von Storch Research Institute, Ottoplatz 1, 50679
Cologne, Germany.
Table 1: Pooled OLS on 5-year non-overlapping averages
with time dummies--Baseline specifications
Dependent variable CA/GDP (in %)
(1) (2) (3) (4)
gov balance 0.675 0.578 0.572 0.768
(0.229) ** (0.169) *** (0.181) ** (0.172) ***
NFA 0.054 0.037 0.056 0.032
(0.017) *** (0.017) ** (0.014) *** (0.015) **
relative income 0.155 0.104 0.176 0.162
(0.038) *** (0.036) ** (0.037) *** (0.039) ***
old-age dep. -0.167 0.043 0.051 -0.283
(0.119) (0.121) (0.135) (0.129) **
young-age dep. -4.499 3.078 -4.038 3.933
(2.593) * (3.121) (2.765) (2.272)
reer -5.681 -4.333 -5.231 -5.655
(3.732) (3.791) (3.504) (2.859) *
manu 0.177 0.544
(0.333) ** (0.319) **
BUSS -1.92
(0.913) **
Non-BUSS -2.175
(1.020) **
construction -1.291
(0.297) ***
wholesale 0.223
(0.214)
electr, gas, 0.691
water (0.563)
finance -1.213
(0.323) ***
renting 0.353
(0.245)
euro * core 1.927
(1.148)
euro * GIIPS -4.572
(1.626) ** 57
Number of 58 58 58
observations
[R.sup.2] 0.760 0.815 0.807 0.877
Note: Robust standard errors are reported in parenthesis. ***, ** and *
refer to 1%, 5% and 10% significance level. Time dummies are included.
Table 2: Pooled OLS on 5-year non-overlapping
averages: Sector-country-groups specification
Manufacturing Business-related services
pc1 3.316 pc1 2.871
(0.498) *** (0.482) ***
pc2 0.325 pc2 0.042
(0.392) (0.390)
manu * core 0.740 BUSS * core 0.331
(0.245) ** (0.202)
manu * GIIPS -0.256 BUSS * GUPS -0.939
(0.244) (0.289) **
manu * East 0.805 BUSS * East 0.124
(0.288) ** (0.358)
manu * non-euro 0.493 BUSS * non-euro -0.090
(0.294) (0.246)
Number of 58 Number of 58
observations observations
[R.sup.2] 0.781 [R.sup.2] 0.770
Non-business services Construction
pc1 2.951 pc1 3.112
(0.488) *** (0.421) ***
pc2 0.201 pc2 -0.121
(0.375) (0.377)
Non-BUSS * core 0.271 constr * core -0.173
(0.140) * (0.276) **
Non-BUSS * GIIPS -0.462 constr * GIIPS -0.726
(0.147) ** (0.189) ***
Non-BUSS * East 0.204 constr * East -0.521
(0.243) (0.256) **
Non-BUSS * non-euro 0.049 constr * non-euro -0.223
(0.145) (0.287)
Number of 58 Number of 58
observations observations
[R.sup.2] 0.778 [R.sup.2] 0.787
Renting activities
pc1 3.176
(0.490) ***
Pc2 0.120
(0.391)
rent * core 0.247
(0.093) **
rent * GIIPS -0.332
(0.170) *
rent * East 0.329
(0.240)
rent * non-euro 0.094
(0.129)
Number of 57
observations
[R.sup.2] 0.756
Note: Robust standard errors are reported in parenthesis. ***, **
and * refer to 1%, 5% and 10% significance level. Time dummies are
included.
Table 3: Fixed-effects regressions on non-overlapping 5-year
averages and pooled OLS estimations on annual observations
FE estimations
(1) (2) (3) (4)
gov balance 0.527 0.625 0.275 0.767
(0.292) * (0.237) ** (0.290) (0.218) **
NFA 0.055 0.011 0.050 0.025
(0.021) ** (0.020) (0.018) ** (0.022)
relative income -0.004 -0.159 -0.057 -0.036
(0.117) (0.099) (0.109) (0.122)
old-age dep. -0.089 0.403 -0.264 -0.229
(0.335) (0.312) (0.312) * (0.316)
young-age dep. -0.805 1.312 3.319 2.082
(0.996) (2.462) (2.741) (3.336)
manu -0.646 -0.123
(0.697) (0.771)
BUSS -0.917
(1.825)
Non-BUSS -1.123
(2.025) ***
construction -1.996
(0.473) ***
wholesale -0.923
(0.815)
electr, gas, water 1.615
(1.109)
finance 0.126
(0.843)
renting 1.543
(0.668) **
euro * core 0.948
(1.509)
euro *GIIPS -3.160
(1.815) ***
Number of 58 58 58 58
observations
[R.sup.2]--overall 0.450 0.045 0.060 0.479
Pooled OLS
(5) (6) (7) (8)
gov balance 0.556 0.464 0.502 0.583
(0.122) *** (0.105) *** (0.111) *** (0.118) ***
NFA 0.049 0.033 0.050 0.038
(0.012) *** (0.011) ** (0.012) *** (0.012) **
relative income 0.164 0.118 0.174 0.168
(0.022) *** (0.019) *** (0.023) *** (0.023) ***
old-age dep. -0.133 0.050 0.104 -0.222
(0.069) * (0.073) (0.079) (0.078) **
young-age dep. -4.485 1.769 -3.883 2.968
(1.503) ** (1.526) (1.537) ** (1.486) **
manu 0.335 0.405
(0.220) (0.192) **
BUSS -1.227
(0.548) **
Non-BUSS -1.889
(0.575) **
construction -1.260
(0.227) ***
wholesale 0.016
(0.128)
electr, gas, water 0.350
(0.343)
finance -1.083
(0.184) ***
renting 0.180
(0.139)
euro * core 2.191
(0.646) **
euro *GIIPS -4.316
(0.838) ***
Number of 288 288 279 268
observations
[R.sup.2]--overall 0.618 0.678 0.647 0.713
Note: Robust standard errors are in parentheses. ***, ** and *
refer to 1%, 5% and 10% of significance level, respectively.
