The Impact of the Global Recession in Europe The Role of International Trade1

Catherine Keppel
Wirtschaftuniversitt Wien
Institute for Fiscal and Monetary Policy

Julia Wrz
Oesterreichische Nationalbank
Foreign Research Division

Abstract:
We assess empirically the marginal impact of export orientation for the
severity of the growth downturn in 2009 while controlling for a host of
country characteristics such as structural features, macroeconomic
imbalances, previous growth record, and the like. We employ an iterative
robust regression technique using a sample of 38 mostly European countries
to quantify the direct contribution of trade to the crisis response with a
special focus on Eastern European economies. We find that greater export
orientation is related to a stronger growth downturn in our sample. This
effect is only revealed when we allow for an interaction term between
export orientation and domestic GDP structure. Moreover, we find strong
evidence that the domestic structure of value added (in particular a high
share of industry in GDP) has compounded a countrys vulnerability during
the recent crisis. We further find support for the effects of overheating:
previous high growth rates are systematically linked to a stronger growth
downturn in all specifications. Finally, a negative influence from previously
high debt ratios on the output response during the crisis is confirmed.
JEL-codes: F14, F15, O52
Keywords: trade collapse; industrial structure; Central, Eastern and
Southeastern Europe

This version: February 26 2010

Paper prepared for the SUERF/BWG Conference and Special OeNB East Jour Fixe Contagion and Spillovers
New Insights from the Crisis, held on February 12 2010 at the OeNB Vienna. The authors thank Jarko Fidrmuc
for helpful comments and suggestions.

1. Introduction
Since the unfolding of the recent financial crisis, world exports and imports dropped dramatically.
While global GDP is forecast has shrunk by xx% (according to the IMF World Economic Outlook
04/2010), world trade volume has fallen by xx% in 2009. The figures are similar for all major
industrialized regions in the world and some emerging markets. Hence, trade has certainly been one
of the most severely hit economic activities. The question arises to what extent this strong and
immediate trade response has propagated the crisis throughout the globe in 2009 and to what
extent trade openness has reinforced the impact of the crisis for individual countries.
In other words, the remarkably large drop in trade is central in the discussion on how a local financial
crisis could turn into a global economic crisis. The academic discussion seems to agree that only the
combination of a number of distinct shocks could have led to the sudden spread of the crisis and
cause a global recession of this magnitude (see for example Escaith, 2009; McKibbin and Stoeckl,
2009; Rose and Spiegel, 2009, to name just a few). Starting from the bursting of the housing bubble
in the US in mid 2007, households in the US experienced a sharp reduction in wealth which, coupled
with changing perceptions of risk by households and businesses, implied a demand shock in the US.
All else equal, this would have triggered two offsetting effects on the global economy: a shortfall in
external demand for countries which are heavily dependent on the US market on the negative side
and rising investment opportunities in response to low returns on investment in the US on the
positive side (McKibbin and Stoeckel, 2009).
In the recent crisis the latter effect did not materialize, because the reappraisal of risks was just as
global in scope as was the shortfall in demand. In particular, emerging markets which in theory could
have profited from worsening investment conditions in the US (and other highly developed
countries) were negatively affected by changing investor risk perceptions. Hence in the current crisis,
the two effects reinforced instead of offsetting each other. In that way, international trade acted as a
compounding factor not only transmitting but also magnifying the effects of the US crisis. All this has
led to a worsening in investment climate, deteriorating credit conditions and falling domestic
demand in particular for durable and investment goods.
But how important in quantitative terms was trade in the global transmission of the crisis? In this
paper, we try to quantify this effect and argue that the direct contribution of trade to deepening the
economic downturn should not be overstated. Only the combination of structural characteristics and
highly open markets have implied a strong output response in the countries which are most heavily
exposed to the sectors contracting most strongly during the crisis. We start by summarizing the
dramatic developments in world trade in 2009 in Section 2. Section 3 reviews the existing related
literature and discusses the mechanisms of crisis transmission through trade. We identify basically
three factors of theoretical importance in shaping the outcome of the crisis. Two demand-side
explanations are related to the rising importance of global supply chains and structural differences
between the domestic and external sector of the economy. A supply-side explanation relates to
problems in trade financing. Section 4 uses an econometric model to estimate the importance of
trade, structural characteristics and financial uncertainties for the incidence of the crisis in a crosssection of 38 predominantly European countries. Section 5 concludes.

2. Recent Developments in World Trade

The drop in international trade flows observed in early 2009 was unprecedented and largely
unexpected. Figure 1 shows real growth of exports and imports for the world, the Euro Area, the
United States, and for Central and Eastern Europe from August 2004 to August 2009. The decline in
trade volumes started in May 2008 and lasted for roughly one year. By May 2009, global trade
volume has dropped by almost 20% annually in real terms.
A similar picture emerges when looking at different world regions. US imports were down by 21%,
Euro Area imports by 18%. Exports of both regions contracted by 20% and 19% respectively. For
Central and Eastern Europe, the contraction in trade has been particularly strong. Exports from
Central and Eastern European countries fell by 23%, while the regions imports contracted by as
much as 28%.
Figure 1

Real Merchandise World Trade, Aug. 2004 - Aug. 2009

Exports in billion USD

Imports in billion USD

1100
1000
900

100

1100

100

90

1000

90

80

900

80

800

800

70

70
700

700
60

600

60

600

500

50

500

50

400

40

400
300

40

300

170

170

250

150
130

200

250

150
130

200

110

110

150

150
90

90

70
100
Aug- Feb- Aug- Feb- Aug- Feb- Aug- Feb- Aug- Feb- Aug04
05
05
06
06
07
07
08
08
09
09

70
100
Aug- Feb- Aug- Feb- Aug- Feb- Aug- Feb- Aug- Feb- Aug04
05
05
06
06
07
07
08
08
09
09

World imports (l.s.)

Euro Area (l.s.)

CEE (r.s.)

United States (r.s.)

World exports (l.s.)

CEE (r.s.)

Euro Area (l.s.)

United States (r.s.)

Source: CPB Trade Monitor.

At the sectoral level, mineral fuels, crude materials, manufactured goods and machinery and
transport equipment experienced the most severe drops in trade. Figure 2 displays three-month
moving averages of import growth for different sectors in OECD countries. OECD imports in these
categories recorded dramatic declines between 11% and 15% month-on-month at the height of the
trade collapse (from December 2008 to January 2009). In May 2009, when global trade volumes
reached the trough, these categories were between 49% (mineral fuels) and 34% (machinery and
transport equipment) below their previous levels in annual terms. However, these were also the
categories which showed the strongest signs of recovery in mid-2009.
As mentioned above, trade started to decline in May 2008 and turned negative first in October 2008
when the collapse of Lehman and the consequent re-assessment of risks has added a second,
decisive shock to the global economy in addition to the US-demand shock. The trough in trade
growth was reached already in January 2009, positive import growth, however, was only reached
again in July 2009.
Much of the nominal trade collapse depicted in Figure 2 is of course due to price effects, which
becomes visible in the steep decline in OECD fuel imports shown in the second panel of figure 2. The
weakness in US demand for mineral fuel may also have a supply-chain component since petroleum
3

and natural gas are major inputs into chemicals and plastics, which are in turn intermediate inputs
for many other industrial sectors (Ferrentino and Larsen 2009).
Figure 2

OECD Import growth by product categories, based on 3-month moving averages

month-on-month growth in %

month-on-month growth in %
10

10

-5

-5

-10

-10

-15

-15

-20
2008M1

2008M4

2008M7

2008M10

2009M1

2009M4

Mineral fuels, lubricants and related materials

Crude materials, inedible, except fuels
Manufactured goods
Machinery and transport equipment
Beverages and tobacco
Source: OECD.

2009M7

-20
2008M1

2008M4

2008M7

2008M10

2009M1

2009M4

2009M7

Commodities and transactions, n.e.c

Chemicals and related products, n.e.s
Animal and vegetable oils, fats and waxes
Miscellaneous manufactured articles
Food and live animals

The casual inspection of the data suggests that trade may have the potential to act as an important
transmission channel in particular for countries heavily exposed to those manufactured exports
experiencing the sharpest declines. For example, the Central, Eastern and Southeast European
countries (CESEE) are strongly specialized in exports of machinery and transport equipment. Recent
trade data for the CESEE EU member states reveal that indeed their exports of automobiles and
related parts declined strongly already in the fourth quarter of 2008 and thus before most of these
countries showed a GDP-reaction to the financial crisis. The initial drop was particularly strong in
exports of automobiles and related parts, followed by a substantial reduction in capital goods
exports in early 2009 (see Figure 3).

Figure 3

Absolute Change In Exports By End-U se Categories for CESEE EU Member States

change to previous year in ppts of GDP
4.00

2.00

0.00

-2.00

-4.00

-6.00

-8.00

-10.00
2007Q1

2007Q2

2007Q3

2007Q4

2008Q1

2008Q2

2008Q3

Food, Feeds, & Beverages

Fuels, etc.

Capital Goods

Automotive vehicles, etc.

Consumer Goods

Other Goods

2008Q4

2009Q1

2009Q2

2009Q3

Industrial Supplies

Source: Eurostat.

By all standards, the trade developments in 2009 represented a dramatic shock for all export
oriented countries throughout the world. In the next section, we will separate conceptually different
mechanisms through which this external demand shock could be transmitted to domestic output
shocks in individual countries.

3. The role of trade in crisis transmission

The discussion in the literature on openness and vulnerability to shocks is not new, the findings may
often seem, however, to be contradictory at first sight. For example, Cavallo and Frankel (2008) find
robust evidence that even when controlling for endogeneity between trade and growth openness
to trade reduces the vulnerability to crises. Edwards (2004) shows that trade integration can lead to
fewer sudden stops in capital inflows. Martin and Rey (2006) find that emerging markets showing
increased financial integration without increased trade openness are at higher risk to financial crises since profits and dividends relate to output volatility.
The contrasting view, namely that trade propagates shocks, is presented inter alia very recently by
Fidrmuc and Korhonen (2009) who point towards greater synchronisation of shocks in the presence
of stronger trade ties. Their analysis is based on East Asian business cycles over the period 19902008. Haile and Pozo (2008) confirm this view with respect to currency crises. Both studies conclude
that openness is one factor which renders a country prone to sudden stops of capital inflows. To
summarize, countries that are more integrated are also more at risk to be affected by external
shocks.
The apparent puzzle between the conflicting findings presented above is easily resolved when
disentangling differences in the definition of a crisis as being of either domestic or external origin.
On the one hand, greater openness will have a positive impact when the crisis is of domestic origin.
There is even evidence that openness can help to avoid domestic crises, for it spurs financial
(Aizenman, 2008) and general regulatory, economic and legal reform (Cavallo and Frankel, 2008). On
the other hand, greater openness by definition helps to propagate external shocks across countries
5

which are highly integrated. This is also often considered to be an explanation why most crises
remain regional in scope (Caramazza, Ricci and Salgado, 2000).
Assigning the centre of the current crisis to the housing bubble in the US, all other countries have
been confronted with an external shock during 2007/2008. It is still unclear, to which extent this USbased crisis has then been transmitted by the financial channel as opposed to a real (i.e. trade)
channel. Looking at currency crises, Caramazza et al. (2000) identify weak output growth and to a
lesser extent external imbalances as important factors for the vulnerability of emerging and
developing countries. Trade and financial linkages play a role only when coupled with weak current
account balances. Also for industrial countries they find that fundamentals play an important role,
however, their importance differs with unemployment playing a greater role here. Haile and Pozo
(2008) also confirm these findings and stress the importance of fundamentals and highlight the
possibility of contagion through the trade channel, when fundamentals are weak.
Basically, three rather distinct channels emerge, which could explain the fast propagation of the crisis
across the globe. Two of these offer a demand-side explanation, while the third one works through
supply shortages. All three channels imply also an indirect role for international trade in crisis
transmission:
1. Rising importance of global value-chains
2. Structural differences between domestic and external sector
3. Problems in (trade) financing
The first two channels imply that the demand contraction in the US is transmitted across trading
partners with different intensity depending on structural characteristics of these countries. The third
channel is qualitatively different and describes a supply shock, i.e. production outside the US is
impeded due to deteriorating credit conditions.
3.1. Rising importance of global value-chains
The interconnectedness of modern production chains provides for the first channel. The US demand
shock is transmitted via international trade, often within a firm but nevertheless across countries,
through strongly integrated global supply chains, also referred to as fragmentation of production,
outsourcing or offshoring. The elasticity of trade to changes in GDP has been explored by many
empirical studies. While the literature provides different estimates depending on the model being
used, there is a clear consensus that the elasticity of trade to GDP seems to have increased over
recent years in response to the higher degree of vertical specialisation we see today. Irwin (2002)
estimates the elasticity of real world trade to real world income to have increased from around 2 in
the 1960s and 1970s to 3.4 in 1990. In a rather simple empirical framework, Freund (2009) finds this
elasticity to be around 3.5 today. Yi (2008) and Tanaka (2009) also attribute the significant increase in
the elasticity of trade to income to greater fragmentation of production. The high degree of vertical
integration we see today would however imply that the countries which are similarly integrated in
global value chains are hit to the same extent. However, this is not what we observed in practise.
Escaith (2009) attributes also a prominent role in the spreading of the 2008/09 crisis to trade effects.
He argues that global production networks aggravated the downturn in the real sector because
global supply chains involve a high degree of stock-building and hence financing. The drop in final
demand and the drying up of credit markets consequently resulted in the aggravation of the initial
effect in each link in the chain and thereby exacerbated the economic downturn. By the same logic,
economic recovery should also work its way back the supply chain in a similar way.
However, increased vertical specialization and supply chains might be only part of the explanation for
the strong trade reaction. As several authors point out (Francois and Woerz, 2009; Freund, 2009), a
6

part of the unproportionally strong drop in trade is due to simple accounting reasons. Since GDP is a
value added measure but trade data are measured on a gross value basis, any decline in GDP
necessarily triggers a relatively higher decline in trade. Double counting over countries adds to this.
Hence, the contribution to crisis-transmission by trade potentially remains considerably smaller than
the observed volatility in trade flows. A large decline in trade could reflect a much smaller decline in
global value added if production is done across countries at the margin and - as demand falls international production chains break down. For example, Porsche in April announced that it is
reducing production in Finland, but upholding German production (New York Times, 4 April 2009).
Firms may tend to source relatively more from home country suppliers during times of crisis.2
3.2. Structural differences between domestic and external sector
A second demand-side transmission channel via trade is to be found in structural differences
between the domestic sector and the external sector of an economy. For developed countries, about
80% of trade is manufactures, while they constitute only 20% of domestic GDP. This structural
argument offers an explanation for a different crisis-impact on countries depending on their specific
trade and GDP structure. Thus, not only openness as such matters for a countrys vulnerability to a
severe external demand shock, but it is important whether the country is particularly exposed to
those sectors which suffer the strongest decline in demand.
A casual inspection of available data underlines this argument. Figure 3 contrasts the differences in
GDP versus trade structure for the EU25 (as of 2004). The left panel of figure 3 shows a weighted
average for all EU members which acceded prior to 2004 (EU15), while the right hand side panel
gives the same data for the Central and Eastern European countries which acceded in 2004 (CEE EU
member states). The differences are striking. First of all, we see a clear distinction between the
industrial structure of the external sector (exports and imports) and domestic value added in each
region. The EU15 exhibit a typical structure for highly industrialized countries with 66% of domestic
income generated by the service sector in 2005. The manufacturing sector accounted for 30% in
domestic value added but as much as 80% of trade. The respective figures for the CEE EU member
states manufacturing sector were 43% and 89% respectively in 2005, the share of services in
domestic value added was at 48%. Thus, the importance of manufacturing in trade is much greater
than in domestic GDP for both regions, hence explaining the steep drop in trade during the peak of
the crisis.

As a word of caution, there is the danger that protectionist policies kick in when global GDP drops sharply and
this sharp drop is exacerbated through a decline in trade. However the same protectionist policies, while
arriving too late to shield a country from the downturn may also impede the country to profit from the global
upswing again. The Great Depression of the 1930s illustrated very impressively this policy mistake which has
led to a prolonged period of recession. (see Almunia et al. 2009)

Figure 4

Structure of Value Added and T rade in 2005

EU 15

CEE EU Member States

100%

100%
17.5

90%

80%

80%
12.1

60%

50%

50%

40%

40%

20%

66.6

3.2

11.3

72.7

73.7

3.8

63.0
30%

39.4

20%

26.9

10%
0%

15.6

70%

66.2

60%

30%

12.8

48.1

14.0
70%

9.9
90%

22.0

10%
8.7

3.7

1.9

2.9

Value Added

Exports

Imports

primary

other manufacturing

transport equ.

0%

services

Value Added

primary

1.8

2.2

Exports

Imports

other manufacturing

transport equ.

services

Source: EUKLEMS database (Timmer et al. 2007).

These figure further illustrates a second important difference between the two regions: The catchingup economies of Central and Eastern Europe show a higher share of manufacturing not only in trade
but also in domestic value added. A tenth of this is transport equipment, almost 4% of GDP and more
than 15% of exports for the region as a whole. This share is however higher for the three largest
economies in the region. According to recent Eurostat figures, exports of automobiles and related
parts accounted for 22% in the Czech Republic, 21% in Hungary and almost 23% in Poland in 2008.
Countries with high trade shares in industries and sectors that were hit especially hard in the crisis
were strongly affected by the crisis as well. During the 2008/09 downturn, the manufacturing
industries have been among the ones that have been hit the hardest. For Central and Eastern Europe,
the automobile sector along with increased risk perceptions by international investors was
particularly relevant for the deep impact of the crisis in 2009.
3.3. Problems in (trade) financing
Finally, financial crises may affect different sectors not only on the demand side but also through
different financing needs. Iacovona and Zavacka (2009) show that sectors which are highly
dependent on external financing needs suffer more from a banking crisis than other sectors. Auboin
(2009) assigns an important role to problems with financing in general and hence also with trade
finance concerning the trade collapse in the recent crisis. He stresses that credit for trade while
holding up well compared to other forms of credit until autumn 2008 - was subject to a tremendous
reassessment of risk and scarce liquidity in late 2008 as shown by the increase in spreads for credit
for emerging and developing countries. This is in line with Chauffour and Farole (2009) who underline
the role of trade finance in the recent trade collapse and give a number of reasons why trade finance
declines particularly strongly during times of financial crises: First, markets may encounter difficulties
in (re)assessing the associated risks, second, market participants face asymmetric information and
increased risks of default, third, financial intermediaries have to respond to scarce liquidity and
needs of recapitalization and finally, political factors might play a role as well. Furthermore, it might
be more difficult to assert ones claims across borders during times of crisis. Alun (2009) emphasizes
that trade is usually financed via short-term credit lines which are more easily cut back in times of
crises. He shows that the simultaneous halt of capital inflows and banking crises significantly lowers
imports in emerging economies.
8

4. Quantifying the importance of trade for the economic downturn

In this section we try to quantify the direct contribution of trade to the crisis response with a special
focus on Eastern European economies. We use regression analysis to assess empirically the marginal
impact of export orientation on the severity of the growth downturn in 2009 while controlling for a
host of country characteristics such as structural features, macroeconomic imbalances, previous
growth record, etc. In particular, this allows us to shed some light on the relative importance of the
two demand-side explanations given earlier and to assess empirically to which extent openness as
such has implied an increased economic weakness during the crisis.
Our econometric model identifies how the vulnerability of individual countries to common shocks is
determined by national characteristics of each country. Our approach is similar to Rose and Spiegel
(2009) in two ways: First, as in Rose and Spiegel (2009) we undertake a cross-section analysis
(determined by the uniqueness of the event) using iterative robust regression. Secondly, we also do
not attempt to determine the timing or causes of the crisis, but the different impact of the crisis
across countries. However, we depart in two important ways from the analysis by Rose and Spiegel
(2009): First of all, while they incorporate changes in real GDP, the stock market, country credit
ratings and exchange rates into a single measure of crisis incidence, we do not construct a latent
variable. In contrast, we consider some of the components which they incorporate into their crisis
measure as determinants for the crisis response. We focus here on the output response to the
crisis, thus our measure of the severity of the crisis is the downturn in real GDP growth between the
realized growth rate in 2007 and the projected growth rate in 2009. Second, we also apply a different
timing for the causes and crisis-response than in Rose and Spiegel (2009). Although many countries
already exhibited negative growth in the final quarter of 2008, most countries continued to grow
throughout 2008. However, the majority of countries is expected to show a real GDP decline in 2009.
Further, in particular emerging countries and the countries in Central and Eastern Europe, which are
of particular interest to us, entered into recession with a roughly 1 quarter delay compared to major
industrial countries. Forecasts for 2010 are again somewhat brighter, even if the recovery may turn
out to be unassertive. Hence, 2009 appears to be a good choice for the trough of the crisis. By the
same reasoning, we calculate the full impact of the crisis as the absolute change in real growth
between 2007, the last year before the crisis struck, and the real growth forecast for 2009 rather
than the growth difference 2007-2008 as in Rose and Spiegel (2009).
4.1 Data description
In our model we relate the decline in output growth between 2007 and 2009 to a set of country
specific characteristics relating to potential transmission channels and a range of other control
variables. The dependent variable the absolute decline in real GDP growth - is calculated as the
absolute change in real GDP growth between the average real GDP growth over 2007-2008 and
recent forecasts for the 2009 growth rate (and 2010 for robustness checks). The latter are taken from
IMF and Consensus Forecasts and date from November 2009. Thus, the growth downturn is
measured in percentage points with high values indicating a strongly negative crisis response.
We identify a range of macroeconomic preconditions arising from the literature review in Section 3
and which may have played a role in determining the severity of the crisis across countries. The
direct impact of openness or trade is captured by various trade ratios (exports to GDP ratio,
intermediate or consumption goods as a share of total exports, exports to the EU as a share of total
exports) whereby we concentrate on total exports to GDP in most specifications. Structural aspects
are captured by the share of industry in total value added. We further control for a certain lack of
investor confidence already prior to the crisis rather crudely through the change in nominal exchange
rates or alternatively the change in CDS spreads between 2006 and 2008. This can also be
9

interpreted as an indication of economic weakness (or perceptions thereof) before the full impact of
the crisis stroke. Apart from these prime variables of interest, we also control for a range of
alternative factors. Previous growth is included to control for effects from overheating. We further
control for weak fundamentals by including external debt stocks in relation to GDP. Finally, we also
include regional integration dummies. A detailed description of the variables used is given in
appendix table A1.
In order to minimize the number of missing observations and to even out short-term fluctuations on
some variables, we calculate a three-year average for all explanatory variables over the period 20062008. This approach also avoids endogeneity problems, since all explanatory variables are lagged
compared to the short-term reaction of GDP on changes in the external environment.3 For the
exchange rate and CDS spreads, we use the percentage change over the period 2006-2008 instead of
average levels, since we want to capture strong revaluations and thus already existing vulnerabilities
in these variables prior to the crisis, rather than relative standing of countries as reflecting a potential
initial weaknesses.4
We base our analysis on a sample consisting of 38 OECD and European countries (see appendix table
A2). Our data comes from various sources. We combine data from the IMF International Financial
Statistics and the World Banks World Development Indicators with data from Eurostat, national
sources and Datastream in order to obtain detailed information on trade structure and changes in
risk perception. A first inspection of the data shows an apparent but rather weak positive
relationship between the share of exports in GDP and the output response to the crisis. As
mentioned before, our dependent variable is defined such that a large positive value implies a strong
decline in GDP growth, hence in the figure below a positive correlation is interpreted as a reinforcing
effect of trade openness on the crisis response.

We also run two robustness checks replacing our dependent variable by the decline between 2008 and 2009
as well as 2007-2010. The results are presented in the appendix.
4
It lies outside the framework of our empirical model to capture transmission channels, and in particular the
financial channel, during the unfolding of the crisis, due to the simultaneity of events. The time period between
the large and sudden re-assessments of risks in late 2008, leading to sharp increases in credit costs, and the
outbreak of the crisis in individual countries is too short to be captured by the macro-economic data which we
are using in our model.

10

Figure 5

Crisis Response in 2009 and Trade O penness 2006-2008

30

growt h reversal 07-09

25

20

15

10

0
0.2

0.4

0.6

0.8

1.2

1.4

1.6

1.8

T rade O penness

Sources: Authors' Calculations based on Eurostat, UN, IMF, Consensus Economics.

4.2. The econometric model

Our econometric model is rather simple and the basic specification is summarized in equation 1.

yi = + 1 xratioi + 2 vad _ ind i + 3eri + 4 avgrowthi + 5 debt i + i

In earlier versions of the model, we included a range of additional control variables. Since they all
turned out to be insignificant and were without impact on the remaining variables in the model, we
dropped them in view of severe degree of freedom constraints in the small sample. The following
variables were additionally used in earlier specifications: Macroeconomic imbalances were captured
by the current account (and trade) balance as a ratio of GDP. The magnitude of domestic demand,
which may influence the effect of external demand as transmitted through trade, was proxied for by
country size in terms of GDP and population. We also constructed a measure of domestic demand as
domestic consumption plus imports minus exports. From the growth accounting framework we used
gross fixed capital formation as a share of GDP. According to the convergence hypothesis, poorer
countries experience faster growth. Hence, we also included per capita GDP at purchasing power
parities in an initial specification. Finally, given the importance of global production chains and the
lack of suitable data to capture outsourcing, we included the inward FDI stock as a ratio of GDP as
well. 5 As mentioned, all these factors were without influence on the growth downturn and on the
remaining variables and were dropped in the final specification.
We use iteratively re-weighted least squares to eliminate the influence of outliers and to deal with
the violation of fundamental assumptions in the empirical model related to our small sample size6.
5

We would like to emphasize that, although our original model might at first sight resemble a neoclassical
growth model, we are not trying to explain long-run determinants of economic growth here. Rather we want to
identify stylized facts of the short-run output reaction to the crisis. Put differently, we identify the
characteristics shared by countries prior to the outbreak of the crisis and which implied a particularly strong
decline in output.
6
This form of estimation is implemented in Stata through the command rreg.

11

This uses a Maximum-Likelihood estimator, whereby in a preliminary step, possible influential

outliers are detected and assessed based on Cooks (1977) distance statistic. Two types of iterations
are performed subsequently whereby observations are weighted such that influential outliers obtain
lower weights. First, Huber (1981) iterations are performed, attaching non-zero weights to all
observations. In the next step, up to 1000 bi-weight iterations are applied, whereby all cases with
residuals are down weighted and cases with large residuals obtain zero weight. In most cases,
convergence occurred relatively fast after about 20 iterations in total. This sequencing ensures fast
convergence (from the Huber iterations) while also minimizing problems with extreme outliers (due
to the bi-weight iterations). (Yaffee 2002)
4.3. A Weak Case for Strong Trade Effects
Table 1 shows the results of our regression and compares the importance of the three main variables
of interest: trade openness (to proxy for the importance of increased integration of production) as
measured by the export to GDP ratio (xratio), differences in domestic structure of value added
measured by the share of industry in total value added (vad_ind), and potential macro economic
weaknesses prior to the crisis as captured by strong devaluations of the currency (er). We find no
evidence for a significant direct effect of increased export orientation on the GDP downturn in the
short run. However, we find rather strong evidence for the structural-difference hypothesis in terms
of both, statistical and economic significance. Those countries with a higher share of industry in
domestic value added are hit considerably stronger by the crisis. The growth downturn for countries
with a 1% greater share of industry in domestic GDP is predicted to be increased by roughly 0.4 pps.
This would imply for example that the Slovak Republic, with 40% of total value added generated in
the industrial sector, would have seen a drop in real GDP by 6.8 pps beyond the drop recorded in the
UK, where industry accounts for only 23% of GDP. These 6.8 pps difference, which is explained solely
by structural differences between the two economies, is in fact rather close to the observed
difference of 8 pps (-15 pps in the Slovak Republic versus -7 pps growth downturn in the UK). Finally,
we also find no significant direct effect from strong exchange rate fluctuations of the currency reflecting less stable macroeconomic conditions - prior to the crisis on the severity of the growth
downturn. Along the same line, we also do not see a significant effect in either direction from
participating in a fixed currency arrangement (a point which we turn to in section 4.4).

12

Table 1: Determinants of the Growth Downturn in 2009.

Variable
xratio
vad_ind
er
avgrowth
debt

(1)
-0.003
0.89
0.427 ***
0.00
0.074
0.20
0.511 ***
0.01
0.002 ***
0.00

(2)
0.019
0.47

(3)

(4)

0.354 ***
0.00

0.830 ***
0.00
0.001
0.56

0.487 ***
0.01
0.002 ***
0.01

-0.034
0.62
0.872 ***
0.00
0.001
0.48

xratio*vad_ind
xratio*er
constant
N
2
R
F
Prob > F

-5.721 ***
0.01
38
0.69
14.0 ***
0.00

4.953 ***
0.00
38
0.32
5.4 ***
0.00

-3.644 *
0.08
38
0.60
17.3 ***
0.00

5.510 ***
0.00
38
0.37
6.7 ***
0.00

(5)
0.216
0.03
0.705
0.00
0.107
0.34
0.652
0.00
0.002
0.01
-0.009
0.02
-0.004
0.09
-13.000
0.01
38
0.71
10.7
0.00

**
***

***
***
**
*
***

***

Note: The dependent variable is the percentage point change in real GDP growth between 2007 and
the projected growth rate for 2009; p-values are reported below each coefficient; *(**)[***] indicate
significance at the 10% (5%) [1%] level or below.
Columns 2-4 in table 1 show that our main variables of interest are robust to in- or exclusion of the
remaining two variables. Nevertheless, we further tested for possible multiplicative relationships
between export orientation and the remaining two main variables of interest. Controlling for
interactions between the export ratio and the weight of industry in GDP in column (5) brings out
more clearly the importance of either factor in aggravating the growth downturn, while at the same
time the two factors jointly seemed to have ameliorated the situation. Put differently, greater export
orientation has helped to soften the adverse effects arising from the dominance of the industrial
sector which was hit particularly strongly in during the crisis. However, this dampening effect is very
limited in scope, given the statistically significant (at the 2% level), but economically rather small
coefficient on the first interaction term. The second interaction between the export ratio and
exchange rate developments also shows a similar, economically small and statistically only weakly
significant, dampening impact on the growth downturn.7
Let us now turn to the remaining control variables in our model. There is clear evidence for the
overheating-hypothesis: countries with a previous fast growth performance (avgrowth) showed a
significantly greater contraction in GDP during the crisis. This effect is highly significant and robust
across different specifications. The coefficient of 0.6 implies that countries which on average grew by
1 pp more over the period 2006-2008 experienced a decline in GDP growth of 0.6 pps beyond the
average in the period 2007-2009. Finally, also the debt to GDP ratio (debt) emerges as a robust and
significant determinant of the growth reaction to the crisis in the sense that starting from a higher
debt ratio has clearly added to a countrys vulnerability in 2009.
The results are more or less robust to using an alternative timing for the calculation of the growth
downturn (see appendix table A3). In particular our results are not sensitive to measuring the growth
downturn by the change in GDP growth rates between 2008-2009 rather than 2007-2009. The same
7

We also included in a different specification the interaction term between the exchange rate and the industry
share in GDP, which turned out to be insignificant and without impact on the results.

13

factors emerge as being relevant, apart from the debt to GDP ratio, which looses its significance in
the very short term. On the other hand, strong currency depreciation just prior to the outbreak of
the crisis seemed to have intensified the growth downturn in the short run. The coefficient of roughly
0.2 in the appendix table implies that a 1% greater depreciation over the 2006-2008 period translates
into a 0.2 pp further decrease in real GDP on average in the very short-run (2008-2009). This may
reflect less investor confidence in countries with rather volatile exchange rate developments already
before the crisis as well as the fact that strong devaluations before the crisis may have limited the
scope for further devaluation during the crisis, which have improved nominal competitiveness in the
short run.
In this context, it is interesting to see some qualitative differences arising when looking at the
medium-term growth response. The last column of appendix table A3 uses the change in real GDP
growth between 2007 and the forecast for 2010 as the dependent variable. Neither the export ratio,
nor the share of industry in GDP or any of the interaction terms between the two show a significant
effect here. In contrast, now strong currency devaluation seems to have smoothed the GDP
contraction. This implies that in the medium term, depreciation and hence improvements in nominal
competitiveness have helped to bolster the impact of the crisis on GDP. However, one should bear in
mind that the dependent variable in this last specification in the appendix is subject to considerable
uncertainty, since the 2010 forecasts rely on information up until September 2009 plus on certain
external assumptions. Therefore, we decided to focus on the more short term output response using
forecasts for 2009 only which are already based on data for the first three quarters of 2009.
Table 2: Controlling for Different Regions
Variable
xratio

EU-27
0.198
0.05
vad_ind
0.680
0.00
er
0.123
0.28
avgrowth
0.674
0.00
debt
0.002
0.01
xratio*vad_ind
-0.008
0.03
xratio*er
-0.004
0.11
EU
0.591
0.51
EU10

**
***

***
***
**

EU-10
0.204
0.04
0.690
0.00
0.083
0.45
0.723
0.00
0.006
0.05
-0.009
0.02
-0.004
0.15

BAL
**
***

***
**
**

CEE
0.218
0.02
0.727
0.00
0.099
0.32
0.744
0.00
0.006
0.03
-0.009
0.01
-0.005
0.05

**
***

***
**
***
**

***

***
**
**
*

14.441 ***
0.00

CEE

N
2
R
F
Prob > F

**

0.845
0.39

BAL

constant

0.206
0.04
0.691
0.00
0.104
0.34
0.600
0.01
0.006
0.04
-0.009
0.02
-0.004
0.09

-13.000 ***
0.01
38
0.72
9.1 ***
0.00

-13.000 ***
0.00
37
0.74
9.9 ***
0.00

-14.000 ***
0.00
37
0.90
30.6 ***
0.00

1.188
0.27
-13.000 ***
0.00
37
0.74
9.9 ***
0.00

Note: The dependent variable is the percentage point change in real GDP growth between 2007 and
the projected growth rate for 2009; p-values are reported below each coefficient; *(**)[***] indicate
significance at the 10% (5%) [1%] level or below; see appendix table A1 for explanation of regional
dummies.

14

Controlling for regional integration clubs (table 2) does not alter the results of our baseline model
(last column in table 1). Controlling for membership in the EU or the effect of being located in
Eastern Europe, brings out the effect of most factors more clearly: the coefficients on domestic
economic structure, previous growth and debt ratios all increase marginally. This is particularly true
when we include a dummy for the three Baltic states, which showed by far the worst growth
performance, or in the framework of our model: the largest crisis response.
4.4. Robustness of the Results
Finally, we performed a series of robustness checks, using alternative measures to capture the effect
of financial weakness and trade. We continue to include the dummy for the Baltic states given their
extremely bad growth performance in 2009, which seems to be largely unrelated to the explanatory
variables in our model.8 In the first column in table 3 we use the growth in CDS spreads (in percent)
to control for early indication of financial trouble instead of changes in the exchange rate. The results
suggest that the reappraisal of risks has had a decisive influence for the impact of the crisis,
amounting to half a pp of additional loss in GDP growth following a one percent stronger rise in CDS
spreads. In columns 2 and 3 we alternatively use a dummy variable for Euro area members and fixed
exchange rate regimes instead of exchange rate movements. The results are again robust, however
the exchange rate dummies themselves are not statistically significant. Thus, we cannot identify any
effect (neither negative nor positive) from being a member of the euro zone or having a fixed
exchange rate regime.
In the last three columns of table 3, we add different measures of trade structure to our model. In
column 4 we include the share of consumer goods in total exports, while columns 5 and 6 use the
share of intermediate goods and goods shipped to the EU instead. The results are again highly robust
when controlling for sectoral and regional trade structure in addition openness to trade as in our
baseline specification. One new result emerges with respect to the commodity structure of trade: A
large share of consumption goods in total exports seemed to have had a cushioning effect during the
crisis. Contrary to our expectations, there is no evidence for a significantly negative influence arising
from global production chains given that the coefficient on the intermediate goods share fails to be
significant. Finally, export orientation to the EU also had no systematic effect for the growth
downturn.
To summarize, the sensitivity checks show that our results are rather robust to alternative
specifications concerning the financial channel and to refining the trade channel. Domestic output
structure appears as one of the strongest and quantitatively speaking most important determinant of
a countrys vulnerability in the recent crisis. Further, we find string evidence for an aggravating
impact (in terms of subsequent output loss) from previous high growth rates (i.e. overheating).
Finally, the negative influence from previously high debt ratios on the output response during the
crisis is confirmed.

The three Baltic countries show average values of export orientation, export structure and the share of
industry in GDP. They all have a fixed exchange rate regime. Their previous growth performance was mixed,
only Latvia showed very strong growth, coinciding with rather low debt ratio. The other two countries had
elevated debt ratios. All three countries clearly stand out with respect to the increase in CDS spreads between
2006 and 2008.

15

Table 3: Alternative Financial and Trade Measures

CDS
xratio
vad_ind

0.229 ***
0.01
0.462 ***
0.00

er
avgrowth
debt
exports*vad_ind
exports*financial 1)
BAL
cds

0.712
0.00
-0.008
0.13
-0.009
0.00
-0.015
0.01
16.55
0.00
0.505
0.00

***

***
***
***

EURO
FIX-ER
Exp_int
0.059
0.032
0.163
0.47
0.64
0.08
0.476 ***
0.403 ***
0.646
0.00
0.00
0.00
0.113
0.27
0.797 ***
0.871 ***
0.676
0.00
0.00
0.00
0.005 *
0.006 **
0.005
0.06
0.02
0.04
-0.003
-0.002
-0.007
0.36
0.50
0.04
-0.044
-0.077
-0.004
0.55
0.18
0.08
13.48 ***
14.25 ***
14.367
0.00
0.00
0.00

*
***

***
***
**
*
***

Exp_cons
0.188
0.02
0.626
0.00
0.033
0.72
0.816
0.00
0.007
0.00
-0.008
0.01
-0.004
0.04
15.136
0.00

**
***

***
***
***
*
***

***

***
***
***
*
***

1.017
0.67

FIX-ER

1.683
0.36

exp-int

0.054
0.17

exp-cons

-0.066 *
0.07

exp-eu27

N
R2
F
Prob > F

**

***

EURO

constant

Exp_EU27
0.216
0.02
0.697
0.00
0.062
0.58
0.774
0.00
0.006
0.02
-0.009
0.01
-0.004
0.07
14.410
0.00

-5.378
0.12
36
0.93
42.6 ***
0.00

-7.333 **
0.05
37
0.90
30.5 ***
0.00

-5.372 *
0.10
37
0.92
40.1 ***
0.00

-14.000 ***
0.00
37
0.90
26.6 ***
0.00

-9.448 ***
0.02
37
0.92
36.1 ***
0.00

-0.019
0.50
-12.000 ***
0.01
37
0.90
27.3 ***
0.00

Note: The dependent variable is the percentage point change in real GDP growth between 2007 and
the projected growth rate for 2009; p-values are reported below each coefficient; *(**)[***] indicate
significance at the 10% (5%) [1%] level or below.
1)
The interactions terms are built with the respective variables for financial factors used in each
specification and thus vary across columns.

16

5. Conclusions
World trade has reacted strongly to the global recession. Several reasons explain the particularly
strong response of trade to the decline in GDP. Besides accounting and price effects, increased
vertical specialisation and global supply chains boosted the drop of trade figures in the aftermath of
the financial crisis. While on the one side the decline in GDP had an impact on imports and exports
worldwide, the collapse of trade on the other side may also have aggravated the decline in growth
and helped to spread the crisis quickly and profoundly across regions. Due to sudden and severe
financing constraints demand dropped steeply in particular for durable consumption and investment
goods, thus causing a recession which fell hard on manufacturing and within the manufacturing
sector on intermediate goods. This in turn affected the functioning of global supply chains, making a
role for international trade as a transmission channel of the crisis. As a consequence, regions with
high trade and output share in these sectors are expected to be hit particularly hard during the crisis.
In this paper we present some empirical evidence to verify the above line of argument and to
quantitatively assess the importance of different factors for the impact of the crisis within a
predominantly European context. We focus on a sample of 38 countries, consisting of EU members,
advanced OECD members and a number of Eastern and South-Eastern European countries. We
employ robust regression techniques in the cross-section (owing to the uniqueness of the event) to
quantify the effect of trade openness on the reversal of GDP growth, thereby controlling for a range
of additional determinants such as structural features of the economies, exchange rate
developments, previous growth performance and external debt ratios to account for macroeconomic
imbalances at large. Thus, we do not attempt to determine the timing or causes of the crisis. Our
econometric model rather identifies how the vulnerability of individual countries to the common
shock occurring in late 2008 is determined by national characteristics of each country.
Our analysis suggests that greater export orientation is related to a stronger growth downturn in our
sample but the effect is only revealed when allowing for interaction terms between the export ratio
and the structure of GDP. In quantitative terms, the marginal effect of export openness is nonnegligible but still considerably lower than the impact of domestic economic structure on the growth
reversal. In contrast to the conditional findings for openness, we find quite convincing evidence that
the domestic structure of value added matters strongly for a countrys vulnerability during the recent
crisis. Those countries with a higher share of domestic value added in the industrial sector as
opposed to more service-based economies were suffering more severe GDP shortfalls. While the
predominance of the industrial sector for a countrys output response in the crisis is independent of
other characteristics of the economy, the trade effect only materializes when interacted with
structural characteristics of the economy. Thus, an important finding relates to the observation that
the trade shock could only result in adverse growth dynamics when coupled with an appropriate
economic structure. Hence, only the combination of different characteristics allow us to explain
regional differences in the impact of the crisis.
Another highly robust result gives support to the overheating hypothesis: previous high growth rates
are systematically linked to a stronger growth downturn in all specifications. Further, a negative
influence from previously high debt ratios on the output response during the crisis is confirmed
reflecting increased vulnerability from the fiscal side. Our results support the view that a higher level
of indebtedness prior to the crisis implied severe limitations to the room for necessary countercyclical measures during the crisis.
Sensitivity checks show that our results for are rather robust and in particular do not depend on the
geographical composition of trading partners and only to a limited extent on the sectoral
composition of trade. Countries which are specialized in consumption goods suffered relatively
weaker declines in output growth, while a specialization on intermediate goods failed to yield a
17

statistically significant result. However, all these robustness checks suggest that the case for a strong
role for trade alone in aggravating the crisis remains limited.
The robust correlation between the importance of industry and manufacturing and the severity of
the crisis response raises of course another, highly policy relevant question: Did the crisis reveal a
fundamental structural change, or in other words: Are the industries which suffered the strongest
declines in the recent crisis going to recover or will they loose importance in the near future? This
would of course call for a completely different set of long-term, structural policy responses, rather
than short-term interventions in the reaction to the crisis.
Finally, since we find only weak evidence that openness to trade has reinforced the impact of the
crisis on the growth performance in the short-run, it remains an open question by how much trade is
able to spur growth in the upswing. In the presence of global supply chains, open borders are of
course an important pre-requisite to guarantee the functioning of economic activity and to assure a
prosperous development in all countries. However, hopes that trade will pull the world out of
recession may be unwarranted as long as credit markets are not fully functioning and global demand
for goods continues to remain below average.

18

References
Aizenman J. (2008), On the hidden links between financial and trade opening, Journal of
International Money and Finance 27 (2008) 372-386.
Almunia M., A. Bntrix, B. Eichengreen, K.ORourke, G. Rua (2009), From Great Depression to Great
Credit Crisis: Similarities, Differences and Lessons, Paper presented at the 50th Economic Policy
Panel Meeting, Tilburg, October, 2009
Alun T. (2009), Financial Crises and Emerging Market Trade, IMF Staff Position Note 09/04
Auboin M. (2009) Boosting the availability of trade finance in the current crisis: Background analysis
for a substantial G20 package, Policy Insight No. 35, June 2009, Centre for Economic Policy Research
Caramazza F., L. Ricci and R. Salgado (2000), Trade and Financial Contagion in Currency Crises, IMF
Working Paper 55/2000.
Cavallo E., J. Frankel (2008), Does openness to trade make countries more vulnerable to sudden
stops, or less? Using gravity to establish causality, Journal of International Money and Finance 27
(2008) 1430-1452.
Chauffour J-P., T. Farole (2009), Trade Finance in Crisis Market Adjustment or Market Failure?,
Policy Research Working Paper 5003, July 2009, The World Bank.
Cook, R. D. (1977), "Detection of Influential Observations in Linear Regression", Technometrics,
American Statistical Association 19 (1) 1518.
Edwards S. (2004), Financial Openness, Sudden Stops and Current Account Reversals NBER Working
Paper 10277
Escaith H. (2009), Trade Collapse, Trade Relapse and Global Production Networks: Supply Chains in
the Great Recession, MPRA Paper No. 18433.
Ferrentino M., A. Larsen, (2009), Transmission of the global recession through US trade, VoxEU,
November 2009.
Fidrmuc J., I. Korhonen (2009), The impact of the global financial crisis on business cycles in Asian
emerging economies, BOFIT Discussion Papers 11/2009.
Francois J., J. Woerz (2009), Follow the Bouncing Ball Trade and the Great Recession Redux
Freund, C. (2009), The Trade Response to Global Downturns. Historical Evidence, World Bank
Working Papers 5015, August 2009.
Haile F., S. Pozo (2001), Currency crisis contagion and the identification of transmission channels,
International Review of Economics and Finance 17 (2008) 572588.
Huber, P. (1981), Robust Statistics, John Wiley & Sons, New York, 153-199.
Iacovona and Zavacka, (2009), Banking Crises and Exports: Lessons from the Past, World Bank
Policy Research Working Paper No. 5016, August 2009
IMF World Economic Outlook 10/2009
19

Irwin, D. (2002) Long-Run Trends in World Trade and Income World Trade Review, 1, pp 89-100.
Martin P., H. Rey (2006), Globalization and Emerging Markets: With or Without Crash?, NBER
Working Paper 11550.
McKibbin, W.J., A. Stoeckel (2009), Modelling the Global Financial Crisis. Centre for Applied
Macroeconomic Analysis, The Australian National University, Working Paper 25/2009.
Dougherty C. (2009), Porsche Finds Fortune From Unlikely Outsourcing, New York Times,
04.04.2009
Rose A., M. Spiegel (2009), Cross-Country Causes and Consequences of the 2008 Crisis: International
Linkages and American Exposure, NBER Working Paper 15358.
Tanaka, K. (2009) Trade collapse and international supply chains: Evidence from Japan VoxEU.org,
May 2009.
Yaffee, R. (2002) Robust Regression Analysis: Some Popular Statistical Package Options ITS
Statistics, Social Science and Mapping Group, New York State University, downloaded on Dec 23 2009
from: http://www.nyu.edu/its/statistics/Docs/RobustReg2.pdf
Yi, K-M. (2009) The Collapse of Global Trade: The Role of Vertical Specialization in The Collapse of
Global Trade, Murky Protectionism, and the Crisis: Recommendations for the G20, edited by R.
Baldwin and S. Evenett, a VoxEU.org publication.

20

Appendix
Appendix Table A1: List of Variables and Data Sources
Variable name

Description

dependent variable
gdp0709
Absolute difference in real GDP growth 2007 to real GDP growth forecasts for 2009, in pp
gdp0809
gdp0710

Absolute difference in real GDP growth 2008 compared to real GDP growth forecasts for
2009, in pp
Absolute difference in real GDP growth 2007 compared to real GDP growth forecasts for
2010, in pp

Source
Eurostat/ Consensus Economics

Eurostat/ Consensus Economics

trade factors
xratio

Exports of goods and services (% of GDP)

exp_cons
exp_eu27

Export share of consumption goods plus motor spirit and passenger motor cars (in % of total Eurostat/ IMF
exports)
Export share to EU-27 (in % of total exports)
Eurostat/ IMF

World Development Indicators

exp_int

Export share of intermediate goods (in % of total exports)

Eurostat/ IMF

Industry, value added (% of GDP)

World Development Indicators

5 year Credit Default Swaps

Nominal exchange rate versus Euro

Datastream
Eurostat/ IMF

structural factors
vad_ind
financial factors
cds
2
er

additional control variables

avgrowth
debt

Average real GDP growth 2006-2008

External debt, incl. intercompany loans (% of GDP)

Eurostat/ IMF
Eurostat/ IMF

regional dummies
BAL

Dummy variable for Estonia, Latvia, Lithuania

CEE

Dummy variable for Albania, Bosnia and Herzegovina, Bulgaria, Belarus, Czech Republic,
Estonia, Croatia, Hungary, Lithuania, Latvia, Montenegro, Macedonia, Poland, Romania,
Serbia, Russia, Slovenia, Slovakia, Ukraine

EU

Dummy variable for the EU-27 countries

EU10

Dummy variable for Poland, Slovakia, Slovenia, Czech Republic, Hungary, Estonia, Latvia,
Lithuania, Bulgaria, Romania

FIX-ER

Dummy variable for countries with fixed exchange rates (Austria, Belgium, Bulgaria, Bosnia
and Herzegovina, Cyprus, Germany, Estonia, Spain, Finland, France, Greece, Ireland, Italy,
Luxembourg, Lithuania, Latvia, Macedonia, Malta, Montenegro, The Netherlands, Portugal,
Slovakia, Slovenia)

EURO

Dummy variable for Euro Area 16

Note: All control variables represent averages over the period 2006-2008, except of the exchange
rate and CDS, where the growth rate between 2006 and 2008 was used.
1
Growth of CDS spreads between 2006 and 2008; data on 5y CDS spreads were not available for
Albania, Bosnia and Herzegovina, Belarus, Switzerland, Macedonia, Norway, and the United Stated,
for these countries, CDS spreads were assumed to equal zero.
2
Growth of the nominal exchange rate of national currency vis-a-vis the euro between 2006 and
2008, a positive number indicates a depreciation of the national currency vis-a-vis the euro.

21

Appendix Table A2: List of Countries.

Albania

Greece

Romania

Austria

Hungary

Russia

Belarus

Ireland

Serbia

Belgium
Bosnia and
Herzegovina

Italy

Slovak Republic

Japan

Slovenia

Bulgaria

Latvia

Spain

Croatia

Lithuania

Sweden

Czech Republic

Luxembourg

Switzerland

Denmark

Macedonia

Turkey

Estonia

Netherlands

Ukraine

Finland

Norway

United Kingdom

France

Poland

United States

Germany

Portugal

Appendix Table A3: Sensitivity of results to differences in the timing of measured growth downturn.
Variable
xratio
vad_ind
er
xratio*vad_ind
xratio*er
avgrowth
debt
constant
N
2
R
F
Prob > F

2007-2009
0.216
0.03
0.705
0.00
0.107
0.34
0.65
0.00
0.00
0.01
-0.009
0.02
-0.004
0.09
-13.000
0.01
38
0.71
10.7
0.00

**
***

***
***
**
*
***

***

2008-2009
0.237
0.02
0.524
0.00
0.196
0.09
0.88
0.00
0.00
0.21
-0.008
0.03
-0.002
0.39
-11.000
0.01
37
0.75
12.4
0.00

**
***
*
***

**

***

***

2007-2010
-0.054
0.34
0.073
0.43
-0.114
0.08
0.81
0.00
0.01
0.00
0.002
0.42
0.002
0.19
-1.942
0.44
37
0.83
20.4
0.00

*
***
***

***

Note: The dependent variable is the percentage point change in real GDP growth; p-values are
reported below each coefficient; *(**)[***] indicate significance at the 10% (5%) [1%] level or below.

22