The Effect of the 2007 Crisis on Developing Countries

(1)

The Effect of the 2007 Crisis on Developing

Countries

Master Thesis International Economics and Business

University of Groningen, Faculty of Economic and Business

January 2015

Author: Supervisors:

(2)

Abstract

This paper explores the effect of the 2007 crisis on developing countries by looking at GDP co-movement among developing and developed countries during 2007to 2010. This paper also examines the decoupling theory of developing countries from developed countries and the effect that the crisis of 2007 had on the propagation mechanisms, like trade and financial integration of developing countries.

(3)

I. Introduction

………... 3

II. Literature review

………..4

II.1. Crisis………...4

II.2. Globalization………..6

II.3. Propagation mechanisms………...7

II.3.1. Trade links………..8

II.3.2. Financial links………....8

II.4. Hypotheses………...9

III. Research Methodology

………10

III.1. Methods to measure interdependence………...10

III.1.1. Static correlation analysis………..10

III.1.2. Dynamic correlation analysis……….11

III.2. Measure of propagation mechanisms

………12

III.2.1. Trade links………..12

III.2.2. Financial links………13

III.3. Model………14

III.4. Data………17

III.5. Roadblocks………..20

III.5.1. Normality of the errors………...20

III.5.2. Heteroskedasticity………..20

III.5.3. Autocorrelation………..21

III.5.4. Collinearity………21

III.5.5. Omitted variables………...22

IV. Results

...22

IV.1. Results regarding hypothesis 1………..22

IV.2. Results regarding hypothesis 2………..23

IV.3. Results regarding hypothesis 3………..26

IV.4. Results regarding roadblocks………28

V. Conclusion

……….29

References

………31

(4)

I. Introduction

The global economy has changed dramatically since the 1980s. The globalization that was started after World War II accelerated considerably since the mid -198os due to two main factors. The first factor involves technological advances that have contributed to the decrease of the costs of communication, transportation and computation. The other factor is the increasing liberalization of trade and capital market (Soubbotina, 2000). According to Soubbotina (2000) globalization has boosted economic growth in developing countries through specialization and technology spill over. Developing countries as a group has become more important in world trade,they account for one-third of world trade and a quarter of world output (IMF, 2001). Some of the developing countries like India and China play an important role in world economy and global growth.

However the globalization in recent decades has also increased cross border economic

interdependence. It is logical to expect that greater openness to trade and financial flows would make a country more susceptible to external crisis and increase the output co-movement among participating countries. One example of this crisis is the recent global economic crisis that started in the U.S.

This financial turmoil that began in the summer of 2007 with a mortgage related crisis in the United States got transformed into a global crisis that reached security markets and banking systems of several developed countries around the world. The collapse of Lehmann Brothers in September 2008 led the developed countries into another depression. Although the developed countries were the center of this crisis nevertheless the crisis could have an affect on the developingcountries as well.

The impressive output growth rate of some developing countries like China and India has fueled the debate on the issue of decoupling of developing countries from developed countries. Looking at figure 1 it is clear that India continued with its impressive output growth rate even after the burst of crisis in 2008. China’s output growth rate had a slight decline after the crisis. However they both have an increasing output growth rate in 2009.

Figure1), Source; IMF world Economic Outlook Database 2011

Gross Domestic Product

0 2.000 4.000 6.000 8.000 10.000 12.000 14.000 16.000 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 Year P e rc e nt c ha ng e G D P

(5)

The quick recovery of these developing countries might suggest that they have been able to soften the blow of the crisis by the increase of their economic convergence with other developing countries and reducing their interaction with the developed countries.

According to economic literature crisis is transmitted through trade and financial linkages from one country to another country (Frankel and Rose, 1998; Otto et al., 2001; Imbs, 2004). With the dramatic changes in the world economy and increase of trade and financial integration among different countries and specially developing countries this crisis may perhaps spread fast among the developing countries.

Considering the increasing importance of developing countries in the world economy it is logical to examine the effect of the recent crisis on these countries.

This paper examines the quarterly Gross Domestic Product (GDP) of 33 developed and 35 developing countries from before the crisis of 2007 and after the crisis to detect the effect of the crisis that was originated in developed countries on the developing countries. Unfortunately China is missing in this study due to lack of data availability.

The aim of this paper is to investigate the output co-movement among developing countries1 with developed countries2 before the 2007 crisis and after the crisis to see whether there is a change in their output co-movement after the crisis. The second object of this paper is to explore the GDP co-movement of developing countries with other developing countries to see whether their regional co-movement of GDP is higher than their GDP co-movement with developed countries. The last object is regarding the channels through which the crisis is

transmitted to another country: have these channels been affected by the crisis or not? These are the subjects that will be discussed further in this research paper.

II. Literature review

II.1. Crisis

The recent global economic crisis is the deepest and most complex crisis since the Great Depression. Perhaps the universal consequence of a U.S. based shock is what has drawn the comparison with the Great Depression more than the severity of the recession itself. The crisis started with the collapse of the sub-prime mortgage market in United States after the burst of a major housing boom in 2007 and subsequent fall in the value of mortgaged backed securities (Bordo, 2008). This led to another crisis in the U.S. shadow banking system which spread to European countries through the drying up of interbank liquidity.

While the Federal Reserve, European Central Bank, and the Bank of England all provided liquidity in the fall of 2007, the collapse of the Lehman Brothers deteriorated the situation and

1_{Argentina, Belarus, Bolivia, Botswana, Brazil, Brunei Darussalam, Bulgaria, Chile, Colombia, Costa Rica,}

Croatia, Georgia, Guatemala, Hungary, India, Indonesia, Jamaica, Jordan, Kyrgyz Republic, Latvia, Lithuania, Malaysia, Mauritius, Mexico, Morocco, Peru, Philippines, Poland, Romania, Russian Federation, South Africa, Thailand, Tunisia, Turkey, Ukraine

2_{Australia, Austria, Belgium, Canada, Cyprus, Czech Republic, Denmark, Estonia, Finland, France, Germany,}

(6)

the credit crunch led to a recession in the U.S. and Europe. In the early stages, most experts believed that mainly developed countries will be affected by the crisis. However as the crisis progressed, the developing countries also were affected through various transmission channels like trade, commodity prices, capital flows and remittances (Mishkin, 2010).

According to Imbs (2010) the shock of this crisis that originated in United States was

transmitted first and mostly among developed countries, while the developing countries were relatively not affected until 2008 after which the shock was slowly transmitted to the

developing countries as well.

Antonakakisa (2012) researched the degree of business cycles synchronization among U.S. and other G7 countries during four different fundamental globalization periods in the world

economy. First period: 1880 till 1913 the classical Gold Standard relatively free trade and capital mobility. Second period: 1920 till 1939 the Great Depression with trade and capital controls. Third period: 1950 till 1973 the Bretton Woods era of fixed but adjustable exchange rates and limited capital mobility. Fourth period: after 1973 with floating exchange rates, increasing trade and capital integration. According to this research paper the U.S. recessions were negatively associated with business cycle co-movement among U.S and other G7 countries during the classical Gold Standard period, which could indicate decoupling among G7 countries in that period.

The U.S. recessions that occurred during the 1920-1939 and the Bretton Woods era had no significant effect on the business cycle synchronization among G7 countries. However the business cycle co-movement among G7 countries with U.S increased during its recessions only after the break of Bretton Woods fixed exchange rates.

Antonakakisa and Scharler (2012) find a strong increase in output correlation among G7 countries during the 2007-2009 recessions in the United States which was rather unusual since the output growth rate did not increase in synchronization during the previous recessions. Looking at historical events there is an increase in output co-movements that corresponds to global or regional crisis. Such as second oil shock in 1979 and the recession in U.S. and Europe that started in 1980; the Latin American debt crisis in the early to mid-1980s; the “Black

Friday” stock market crash in 1987; the U.S. recession in 1990-91; the Exchange Rate

Mechanism (ERM) crisis and European recession in 1992; the tequila, Asian, and Russian crisis in mid –to late 1990s, and the dot-com burst in 2000 followed by a U.S. recession. All these events were either financial in nature or were associated with downturn in U.S. or Europe except for the 1979 oil shock (IMF, 2013).

The increase in output correlation after crisis could be explained by trade and financial linkages because they could transmit country-specific shock to other countries (Frankel and Rose, 1998; Imbs, 2004).

A second explanation could be common shocks simultaneously affecting many countries at roughly the same time-such as a sudden increase in financial uncertainty or a “wake-up call” that elicited a change in investors’ perception of the world (Goldstein, 1998).

(7)

II.2. Globalization

Globalization is generally understood as the process of diffusion of goods, services, capital, technology and human resources across national borders. According to Dreher (2006) globalization not only has a significant affect on human well-being but it also increases integration and interdependence of all countries and regions that participate in the world economy.

During the globalization period the volume and the nature of the trade flows have significantly changed due to liberalization of trade policies around the world and fast decrease of

transportation and communication costs.

According to Akin and Kose (2007) since 1986 and the liberalization of capital account regimes there has been an increase of international financial flows that even overshadows the trade flows.

The global economy has been through profound changes during the past two decades. The trade and financial linkages between developed countries and developing countries have become stronger and some of the developing countries have become more important in the global economy since they now account for a substantial share of the world output. Looking at figure 2 and 3 there is a rising line for both exports and financial flows for developing countries since 1982. Especially financial flows were rising considerably until the crisis of 2007.

Volume of exports of goods and services

-15.000 -10.000 -5.000 0 5.000 10.000 15.000 20.000 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 Years V ol um e of E x po rt s P e rc e na tge c ha ng e

──World Volume of exports of goods and services ──Advanced economies Volume of goods and services

──Emerging and developing economies Volume of goods and services

(8)

Net private financial flows

-100000 0 100000 200000 300000 400000 500000 600000 700000 800000 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005 2007 2009 2011 Years U .S .do ll a rs B il li on s

──Emerging and developing economies Private Financial flows, net U.S. dollars Billions

According to economic theory globalization in recent decades has led to an increase of cross-border economic interdependence and therefore convergence of business cycle fluctuations. The increase of trade and financial flows make economies more susceptible to external shocks. Since the channels through which the shocks could be transmitted to other countries have expanded an increase of fluctuation in the economy of the developed countries could more easily and more rapidly be transmitted to developing countries and consequently increase the output co-movement of developing countries with the developing countries.

During a few years after the crisis the impressive growth rate of some of the developing countries like China and India fed the fuel for the heated debate about the decoupling of these countries from the developed countries. Some researchers like Kim et al. (2011) believe that these countries have been able to better deal with the crisis that originated in the developed countries due to their increase of regional ties which is in line with the decoupling hypothesis. The decoupling hypothesis suggests that there is a decoupling of economies from developed economies and an increase of convergence among regional economies.

According to Kose et al. (2008) in detecting common cycles across countries, the level of economic development and integration into the global trade and financial markets is more important than the geographical proximity. Therefore to test the hypothesis regarding the decoupling of developing countries from the developed countries, the grouping of the countries will be based on their level of integration and also the regional division.

II.3. Propagation mechanisms

(9)

II.3.1. Trade links

The output of a country could be affected by crisis in another country with which it has trade linkages. The literature gives an ambiguous view about the effect of trade integration on output movement. It is not very clear whether higher trade will lead to higher degree of output co-movement or not and that is since trade integration affects co-co-movement through various channels. First, the spillover of the aggregate demand shock through trade has a tendency to make business cycles more correlated among countries. For example, an investment or consumption boom in one country could increase demand for import and therefore act as a positive demand shock for a trading partner. Second, trade flows could encourage greater specialization of production which according to Krugman (1993) will result in less synchronization of business cycles. In case of specialization of production there is an asymmetric industry composition among trading partners and if business cycles are driven mostly by industry-specific shocks, the different compositions of industries will contribute to less synchronization. Third, Frankel and Rose (1998) argue against Krugman (1993) and claim that if intra- industry trade is more prominent than inter-industry trade, business cycles will be more positively correlated as trade integration increases. Frankel and Rose (1998) use thirty years of data for twenty industrial countries and come to the conclusion that the more countries trade with each other the more highly correlates their business cycles are. Lastly, increase trade could require more coordinal fiscal as well as monetary policies, which could lead to

synchronization of policy shock among trading countries. Subsequently business cycles become more correlated since movements of outputs are also driven by coordinated policy shocks. There have been many empirical studies about the effect of trade as a propagation mechanism but with different results. For example, Otto et al. (2001) examine bilateral output growth correlation among 17 OECD countries and find that bilateral trade plays a significant role in explaining bilateral growth correlations among these countries. Their finding is consistent with the earlier work of Frankel and Rose (1998) that higher bilateral trading intensities are

associated with higher output growth correlations. Imbs (2004) finds that trade has an effect on business cycles synchronization but a sizeable portion of this effect is through intra-industry trade. The link between inter-industry trade and business cycle correlation is much smaller in magnitude but the pattern of specialization drives business cycle correlations other things being equal, two countries with a similar economic structure are significantly more correlated. Economic research gives an ambiguous view about the effect of increased trade linkages on cross-country output co-movement, for example trade linkages generate both demand- and supply-side spillovers across countries, which could lead to a higher degree of business cycle co-movement. However stronger trade linkages could result to increased specialization of production across countries and if sector specific shocks are dominant, then co-movement of output across countries could plunge.

II.3.2. Financial links

(10)

goods in these countries and therefore increase output co-movement among these countries. Financial integration could imply a negative output correlation due to facilitation of resource transfer from a country with a negative shock to a country with a positive shock. Financial integration could also result to a reduction of cross-country output correlation by encouraging specialization of production through the redistribution of capital in a way that is consistent with countries’ competitive advantage, see Krugman (1993).

However, many empirical papers show a positive relation between financial integration and output synchronization across countries, like Imbs (2004).

A few of the empirical researches concerning the effect of financial integration on output co-movement across countries will be stated next. Imbs (2004) looks at the bilateral correlation of business cycles among 24 countries in the 80s and 90s and came to the conclusion that the countries with stronger financial linkages are also more synchronized however these countries are also more specialized economies which consequently have a less synchronized cycles. Financial integration could have a positive effect on output correlation due to imperfect

information; liquidity constraints and investors’ herd behaviour that could make them withdraw capital from many countries all together. Kim et al. (2007) observe that Asia-Pacific countries during the 1997-1998 Asian financial crisis show positive business cycle correlation after the shock to their capital flows. Otto, Voss and Willard (2001) also find that OECD countries with strong FDI linkages have more similar cycles. On the other hand Kalemli-Ozcan et al. (2009) find a strong negative effect of banking integration on output synchronization. They come to this conclusion using a panel estimate to examine the dynamics of financial integration and business cycle synchronization across 150 industrial countries during 1978-2006, conditional on global shocks and county-pair heterogeneity. However Kalemli-Ozcan et al. (2009) in their paper look only at banking integration and not the other forms of financial integration like foreign direct investment or portfolio investment.

Based on the above literature the next hypotheses are formulated.

II.4. Hypothesis

H1: There is a higher GDP co-movement of developing countries with the developed countries after the crises

H2: There is a higher GDP co-movement among emerging countries from the same region than among these emerging countries with the developed countries

(11)

III. Research Methodology

Empirical studies that focus on business cycle co-movement usually use GDP growth rate, domestic consumption growth rate, domestic investment growth rate, employment rate and inflation to measure the strength of the relationship between business cycles and the factors that influence them.

GDP growth rate is usually used in empirical studies because it is a better indicator of the economy in general.

III.1. Methods to measure interdependence

The literature offers different methods to measure various aspects of interdependence. Here are a number of most used measures: correlation, co-integration, panel, vector auto regression (VAR), and dynamic factor analysis. The correlation or co-movement analyses are widely used measure in the interdependence literatures. There are two types of correlation analysis: static and dynamic.

Static correlation analyses include simple correlation and trend filter correlation. Dynamic correlation analysis is Dynamic Conditional Correlation (DCC)-GARCH.

III.1.1. Static correlation analysis

Simple correlation

The simple correlation is used to measure the overall co-movements and provides the basic framework for the assessment of interdependence. By dividing the covariance of the two variables by the product of their standard deviations the simple correlation will be obtained. The mathematical presentation of the simple correlation between two random variables X and Y are:

X and Y are two variables of which the relationship is to be evaluated, μX and μY are expected values for X and Y respectively, σX and σY are their standard deviations, and E is the expected value operator (Li, Zhang and Willett, 2011).

Correlation analysis is used by many researchers to test for interdependence. Baig and Goldfajn (1999) used correlation analysis for Thailand, Malaysia, Indonesia, Korea and the Philippines to test for evidence of contagion between the financial markets of these Asian countries through out the 1997-1998 crisis. According to their paper there was a significant increase of correlation for currency and sovereign spread but the correlations for equity market had mixed results. Otto, Voss and Willard (2001) use bilateral growth correlation among a sample of 17 OECD countries to identify a number of variables that can explain those correlations. They find that higher bilateral trading is associated with higher output growth correlation.

(12)

development in one country to those in another country (interdependence) but also those developments themselves. Another disadvantage of simple correlation is that they can not distinguish long run relationships like trends from short run movement around these trends.

Trend-filter correlation

There are two types of trend-filter correlations: linear and non–linear.

By removing the effects of medium or long term trends the trend-filtered correlation find the correlation on detrended data.

Linear trend-filter correlation use OLS regression to estimate a linear trend-line. The linear trend-filtered correlation is the correlation of deviations from the linear trends of two variables. Non-linear de-trending analysis uses the Hodrick-Prescott filter usually to detrend a

macroeconomic time series, which fits the time series with a sum of a linear trending plus a cyclical component. To detrend the macroeconomic time series the fitted curve is subtracted from the original time series (see Hodrick and Prescott, 1997).

Gomez, Torgler and Ortega (2013) use both linear and non-linear Hodrick-Prescott trend filter correlation to investigate the business cycle co-movement across countries and regions and their results show that the synchronisation in regional growth patterns has been the driving force of interdependence instead of synchronisation of the world economy. They also find out that the world crisis increases the global co-movement dramatically.

There are advantages and disadvantages in using trend-filter correlations. Although they can remove the effects of trends from business cycles HP filter correlation puts more weight on the observations at the end of the series.

Static correlations are in general less capable to capture high frequent time varying or dynamic characteristics of the co-movement which are usually shown in fast –changing financial

markets.

III.1.2. Dynamic correlation analysis

Dynamic correlations provide time –varying correlations between economic variables. There are different types of dynamic correlation: Dynamic Conditional Correlations-GARCH (DCC-GARCH) and Time Varying Coherence Function (TVCF) will be explained further in this section.

Dynamic Conditional Correlations-GARCH

DCC-GARCH, which was developed by Engle (2002), takes the volatility or heteroscedasticity and autocorrelation of the variables into account and produces a time-varying calculation of correlations. This process is done in two stages. First, univariate GARCH models are fitted for each variable and then it uses the transformed residuals resulting from the first stage to estimate the dynamic conditional correlation.

(13)

Although DCC-GARCH is more robust than the static correlation methods, particularly for financial variables which are very volatile but its estimations could be greatly affected by the outliers. During the calm periods of the financial markets this model fits the actual behaviour of the markets rather well but it breaks down during crisis when the markets shows larger changes than predicted by the normal distribution.

Time Varying Coherence Functions

Time Varying Coherence Functions is based on time varying coherence to identify endogenously structural changes in the co-movement process.

Essaadi and Boutahar (2008) apply this method to GDP growth rate of the US and the UK to capture the changes in economic business cycles over time and also to see how synchronisation between US and UK business cycles changed from 1960 to 2006.

The main advantage of TVCF is that they can detect both co-movement dynamics in different cycles but also tests if these countries tend to be more synchronized or not.

As mentioned before there are other methods to measure interdependence. For example dynamic factor analysis, Vector Auto-regression analysis which are better able to deal with dynamic aspects of correlation among economies but they are usually used in long time series or variables that are highly volatile. Considering the fact that the data used in this paper is over a short time period and is only looking at the effect of one particular crisis and also the

complexity of the dynamic correlation analysis which is too far reached for this paper therefore the simple correlation is chosen for this research paper.

III.2. Measures of propagation mechanisms

III.2.1. Trade links

There are different proxies for trade intensity some researcher use only the export to GDP ratio, some use the import to GDP ratio or a combination of export and import to GDP ratio.

Frankle and Rose (1996) use the next proxies to measure bilateral trade intensity in their empirical paper.

wx

ijt = Xijt / (Xi,t + Xj,t)

wm

ijt = Mijt / (Mi.t+ Mj.t)

wt

ijt= (Xijt+ Mijt) / ( Xi.t+ Xj.t+ Mi.t+ Mj.t)

Xijt denotes total nominal exports from country i to country j during period t; Xi.t denotes total

global exports from country i; and M denotes imports.

(14)

Kwanho and Chan (2006) define the trade integration between a pair of countries, (i, j) by normalising trade which they define as exports plus imports between these two countries by the Sum of world trade made by the pair.

Where

x

ijt (

x

jit) denotes total nominal exports from country i ( j) to country j (i ) during period

t;

m

ijt (

m

jit) denotes total nominal imports from country j (i) to country i ( j ) during period t;

and Xit (

X

jt) and Mit (

M

jt) denote total global exports and imports for country i ( j) during

period t.

According to IMF (2009) the importance of trade linkages in transmitting crisis from developed countries to developing countries could be measured by the export to developed countries divided by domestic GDP of the developing country.

From theoretical papers it is not clear which one of the trade intensity measures is optimal. Frankel and Rose (1996) conduct their test with import, export and also import and export as a measure of trade intensity but their results were not sensitive to the exact measure of trade intensity. This was expected since the three different measures are highly positively inter-correlated.

Considering the finding of IMF (2009) and Frankel and Rose (1996) finding regarding insensitive nature of the exact measure of trade intensity, the exports from each developing country to developed countries is used to measure the trade intensity in this paper. To measure how important these trade and more specifically export is for the economy of the developing country, the export from each developing country to developed countries is divided by the GDP of the developing country. The following equation will illustrate this measurement.

Export importance =

X

it /

GDP

it

X

it stands for nominal export from developing country i to developing countries during period t.

GDP

it stands for nominal GDP of developing country i during period t.

This equation is used to measure the export importance for each year from 2004 to 2010 for each developing country. The trade intensity of each developing country during 2004 to 2007 is then the average of export importance for each country for 2004 to 2007. The trade intensity for 2008 to 2010 is also the average of export importance for that period.

Some researchers use gravity variables or other instrument to account for other variables that could affect trade between countries. However, the focus of this paper is on the output correlations among countries for a short period of time; therefore gravity variables or other variables that are time-invariant or persistence over time like the structure of production or distance between countries are not included in this research paper.

III.2.2. Financial links

(15)

capital flows which give a better picture of country’s financial integration. One of the measures of financial openness is the sum of gross stock of foreign asset and liabilities as a ratio to GDP. Otto et al (2003) use three different classes of investment which are Foreign Direct Investment (FDI), trade in equity and trade in long term bonds.

Forbes (2001) uses the private capital inflows to GDP ratio as a measure of financial integration.

Gross private capital flows are the sum of the absolute values of direct, portfolio, and other investment inflows and outflows recorded in the balance of payments financial account, excluding changes in the assets and liabilities of monetary authorities and general government. This indicator is calculated as a ratio to GDP.

Due to data restrictions only one investment class is tested here, the portfolio investment which is a combination of both equity and debt investment. The transmission of shocks through portfolio investment is straightforward. If capital is mobile between two economies then a change in the saving and investment decision in one affects the price and the availability of financial assets in the other economy, which could lead to more business cycle synchronization. It is difficult to construct a meaningful measure of bilateral portfolio investment flows between developing and developed economies, therefore the integration intensity of a economy is measures by total portfolio investment liabilities in that developing economy.

According to Kose et al. (2003) the G-7 countries account for more than two-third of all private capital flows and recently an even a greater fraction of flows to developing countries comes from G-7 countries.

By scaling the portfolio investment liabilities of developing country by the country’s GDP, this measure will indicate the importance of portfolio investment liabilities to total production in each economy.

Portfolio importance=

PIL

it /

GDP

it

PIL

it stands for nominal portfolio investment liabilities of the developing country i during

period t.

GDP

it stands for nominal GDP of developing country i during period t.

This equation is used to measure the portfolio importance for each year during 2004 to 2010 for each developing country. The financial integration intensity of each developing country for 2004 to 2007 is then the average of portfolio importance for each country for 2004 to 2007. The financial integration intensity for 2008 to 2010 is also the average of portfolio importance during that time period.

III.3. Model

The first and the second hypotheses use Pearson correlation coefficient and the third hypothesis uses a panel regressions model which is a model for pooling time-series and cross-sectional data.

(16)

(



a



d) (1)

The

y

a is the quarterly percentage growth of GDP of a developed country (a) for the time

period of 2004(Q1) to 2007(Q4).

The

y

d is the quarterly percentage growth of GDP of a developing country (d) for the time

period of 2004(Q1) to 2007(Q4).



ad is the correlation between a pair of GDP of a developed country (a) and the GDP of a

developing country (d) over the time period of 2004(Q1) to 2007(Q4).



ad is obtained by

dividing the covariance of the two variables which are

y

a and

y

d by the product of their

standard deviations.

The same model is also used for the period 2008(Q1) to 2010(Q4) to calculate the pair

correlation between a developed country and a developing country.

For the first hypothesis total average of the pair wise correlations between developed and developing countries is calculated for periods 2004(Q1) to 2007(Q4) and 2008(Q1) to 2010(Q4).

If the average correlation of developed countries with the developing countries for the time period of 2008(Q1) to 2010(Q4) has not increased than the first hypothesis could be rejected. Of

course this is not a very sophisticated method of testing and there are better method to test the correlation but these methods are unfortunately too complicated and above my knowledge. To answer the second hypothesis the developing countries are divided in different regional groups of emerging countries and the pair wise GDP correlation among these emerging

(



e



g) (2)

(17)

This equation is used to calculate the pair wise GDP correlation of two emerging countries from the same regional group for the time periods of 2004(Q1) to 2007(Q4) and 2008(Q1) to

2010(Q4).



ad represents the pair wise correlation of one developing country with one developed country.



eg represents the pair wise correlation among one emerging country which is a sub group of

developing countries with another emerging country from the same regional group.

To test the decoupling hypothesis the average of pairwise correlation of an emerging country with the developed countries is calculated. Next step is to calculate the average of pairwise correlation of the same emerging country with other emerging countries from the same region. The emerging countries could be considered as decoupled from the developed countries if their average correlation with other emerging countries from the same regional group is higher than their average correlation with the developed countries.

To see whether the crisis has had an effect on the propagation mechanisms (trade and financial integration) a regression model is used with a dummy variable distinguishing the years after the crisis from the years before the crisis.

y

ad

ad represents the error term and is assumed to be independently and identically distributed

with mean zero.

The variables used in the equation three are explained before but here is a summary of the step by step process of calculation of these variables.

First the pairwise correlation of real GDP of a developing country with a developed country is calculated using equation (1) for the period 2004(Q1) to 2007(Q4) and 2008(Q1) to 2010(Q4).

Then for each developing country its pair wiseGDP correlations with every developed country are added together and is divided by the number of developed countries which is 33 for the period of 2004(Q1) to 2007(Q4) and the same process is repeated for 2008(Q1) to

2010(Q4)period as well. The results represent two sets of average GDP correlation for each

developing country with developed countries for two periods which is identified as

y

ad in the

equation(3).

(18)

calculated for 2004 to 2007 and 2008 to 2010. These calculations result in two sets of average trade intensity for each developing country which is represented by

T

ad in the equation (3).

Third, for each developing country the annual nominal portfolio investment liabilities of that country are divided by its nominal GDP of the same year. This calculation is completed for each developing country during 2004 to 2010.After that for each developing country an average of this scaled annual portfolio investment liabilities is calculated for 2004 to 2007 and 2008 to 2010. These calculations produce two sets of average financial integration intensity for each developing country which is represented by

F

ad in equation (3).

The third hypothesis is rejected if



3=0which means that the crisis had no effect on the

propagation channels trade and financial integration.

III.4. Data

Growth rate of Gross domestic Product (GDP) is chosen as the mean of identifying the common cycle in economies due to its advantage of being a natural measure of economic performance. GDP is a better measure of overall economical performance than investment or consumption.

For the first two hypothesis the pair wise correlations of quarterly real GDP between developed countries and developing countries between 2004(Q1) to 2007(Q4) and 2008(Q1) to 2010(Q4)

are calculated using data from the International Financial Statistics 2011 (IFS) of International Monetary Fund (IMF) data source. The GDP is a real GDP percent which measured the unit change of GDP from a year before with a country specific base year. The real GDP percent is chosen because it shows the changes in each economy and in this way the size of economies and their economical growth situations are also accounted for.

Due to lack of data availability of trade and financial integration on quarterly and real terms basis for the third hypothesis nominal annual GDP is used to scale trade and financial

integration in developing countries. The nominal annual GDP is based on current prices in U.S dollar.

The data on export is from Direction of Trade statistics 2011 ofInternational Monetary Fund (IMF) data source starting from 2004 to 2010.The export is expressed in millions of US Dollar. The portfolio investment is also expressed in millions of US dollar using the data from Balance of Payments statistics 2011 of International Monetary Fund (IMF) data source.

For the first hypothesis the sample is divided in two time periods of 2004Q1to 2007Q4 and 2008Q1to 2010Q4. This division is made to see whether there is a higher correlation among developing countries with the developed countries after the crisis. The end period 2010Q4 is chosen because the focus of this paper is on the effect of the crisis therefore data consist only until a few years after the crisis.

(19)

The developing countries are; Argentina, Belarus, Bolivia, Botswana, Brazil, Brunei

Darussalam, Bulgaria, Chile, Colombia, Costa Rica, Croatia, Georgia, Guatemala, Hungary, India, Indonesia, Jamaica, Jordan, Kyrgyz Republic, Latvia, Lithuania, Malaysia, Mauritius, Mexico, Morocco, Peru, Philippines, Poland, Romania, Russian Federation, South Africa, Thailand, Tunisia, Turkey, Ukraine.

This classification is based on World Economic Outlook April 2011 which divides the world into two major groups: advanced economies, and emerging and developing economies. This classification is not based on strict criteria, economic or otherwise, and it has evolved over time.

The group of emerging and developing economies consist of 150 countries but due to lack of data availability this group has only 35 representatives in this research paper. Also to make it more clear and convenient the groups are named developed and developing in this paper. To test for the decoupling hypothesis the emerging countries are chosen as a subgroup of the developing countries and their pair wise GDP correlation of the countries from the same regional emerging group will be compared to their GDP correlation with the developed countries to see whether there is a possibility of decoupling from developed countries and convergence among emerging countries. This grouping is made because the emerging countries are more integrated in the world economy than the rest of the developing countries and

therefore based on the finding of Kose et al. (2008) they should be investigated separately from the less integrated countries from the developing group when one is detecting common cycle across countries. Due to lack of data on the less integrated countries from the developing countries it is not possible to consider them as a separate group and investigate them as well. Based on the decoupling literature the regional integrations are more important for these

countries than their integrations with developed countries. Therefore the emerging countries are also divided into different regional subgroups. This division is according to the regional

breakdowns of the emerging and developing economies by IMF The World Economic Outlook April 2011. The regions are Central and Eastern Europe (CEE), Commonwealth of Independent States (CIS), Developing Asia, Latin America and the Caribbean (LAC), Middle East and North Africa (MENA), and Sub- Saharan Africa (SSA).

For the second hypothesis the following subgroups are created.

Central and Eastern Europe (CEE) group contains Hungary, Poland, Romania and Turkey. Commonwealth of Independent States (CIS) group includes Russia.

Developing Asia group consists of India, Indonesia, Malaysia, Philippines and Thailand. Latin America and the Caribbean (LAC) group include Argentina, Brazil, Chile, Colombia, Mexico and Peru.

Middle East and North Africa (MENA) consists of Morocco. Sub- Saharan Africa (SSA) contains South Africa.

Groups (CIS), (MENA) and (SSA) are excluded from further research due to lack of countries in those groups.

This research contains observations on multiple entities (countries), where each entity is observed at two points in time therefore a panel data analysis is chosen.

According to Baltagi (2005) there are some benefits using panel data instead of cross section or time series analysis. Only a few of these benefits will be mentioned next.

(20)

 Panel data give more informative data, more variability and less collinearity among variables.

 Panel data are better suited to study the dynamics of adjustment

Panel data could control for factors that vary across entities but not vary over time, could cause omitted variable bias if they are omitted and factors that are unobserved or unmeasured

therefore can not be included in the regression model.

Based on the available data and benefits of panel analysis this method is chosen for this research paper.

A panel data regression differs from regular time series and cross section regression since it has a double subscription on its variables;

y

it

=



0 +

μ

i is time-invariant and it explains any individual specific effect that is not

included in the regression whereas



it varies with individuals and time and can be thought of as

the usual disturbance in the regression Baltagi (2005).

In this research paper panel regression analysis will control for countries heterogeneity and time –invariant omitted variables like distance or other factors mentioned in gravity model that could have an effect on trade or financial integration of a country.

There are two models: fixed effect model and random effect model. The fixed effect model explores the relation between the dependent and independent variables within an entity which could be country, individuals, company etc.. Each entity has its own individual characteristics that may or may not have an influence on the dependent variable. The fixed effect model assumes that there is something within the entity that could impact or bias the dependent variable or the independent variables which needed to be controlled for. The fixed effect model also assumes that the time-invariant characteristics are unique to the entity and should not be correlated with other entity’s characteristics. Each entity is different, for that reason the entity’s error term and the constant which confines the entity’s individual characteristics should not be correlated with the others. However the random effect model assumes that the variation across entities is random and uncorrelated with the dependent or independent variables included in the model. The random effect model assumes that there is no correlation between the entity’s error term and the independent variables.

(21)

The null hypothesis in Hausman test is that there is no correlation among the error term and the independent variables and therefore the random effect model should be chosen. The alternative hypothesis is that there is a correlation and therefore the fixed effect model should be chosen

III.5. Roadblocks

III.5.1. Normality of the errors

It is an assumption of the regression model that the errors and therefore the dependent variables are normally distributed. The assumption of normally distributed errors could be verified by looking at the histogram of the residuals, however the formal test is the Jarque-Bera test for normality which is based on two measures: skewness and kurtosis. Skewness refers to how symmetric the residuals are around zero. A skewness of zero says that the residuals are perfectly symmetric. The kurtosis refers to the ‘peakedness’ of the distribution. A normal distribution has the kurtosis value of 3 (see Hill, Griffiths and Judge, 2001).The Jarque-Bera statistic is given by

JB= T/6 [S² + (k-3)²/4]

S stands for skewness, k stands for kurtosis and T stands for the number of observations. When residuals are normally distributed theJarque-Bera statistic has a chi-squared distribution with 2 degrees of freedom. Therefore the hypothesis of normally distributed errors is rejected if the calculated value of the statistic exceeds the critical value 5,99 selected from the chi-squared distribution with 2 degrees of freedom (see Hill, Griffiths and Judge, 2001)

III.5.2. Heteroskedasticity

Heteroskedasticity is a common problem when using cross-sectional data. Heteroskedasticity means that the variances for all observations are not the same. The existence of

(22)

(T1- K) and (T2- K) degrees of freedom. T1 and T2 are the number of observations in each

sub-sample (see Hill, Griffiths and Judge, 2001)

Heteroskedasticity has consequences for the regression model. One is that the standard errors are incorrect under heteroskedasticity. In this case ignoring heteroskedasticity and continuing using incorrect standard errors tends to overstate the precision of estimation; the confidence intervals are narrower than they should be.

The White’s heteroskedasticity consistent standard errors could be used to avoid computing incorrect interval estimates or incorrect values for test statistics when there is

heteroskedasticity.

Another consequence of heteroskedasticity is that the least square estimator is no longer the best estimator and there is another estimator with a smaller variance. In this case the best linear unbiased estimator is the Generalized Least Squares estimator. By applying Generalized Least Squares the variables are transformed in order to convert the heteroskedastic error model into a homoskedastic error model and this transformation does not change the meaning of the

coefficients in the model. It is expected that the Generalized Least Squares have lower standard errors than the least squares because it is a better estimation in the presence of

heteroskedasticity however it also puts more restrictions on the model because more assumptions about σ² will be included in the regression model.

After conducting both White’s standard error test and the Generalizes Least Squares the one with the lowest standard errors will be chosen.

III.5.3. Autocorrelation

One of the assumptions in linear regression model is that the errors are not correlated. Cross-sectional data are generated by random sampling which implies that the error terms for different observations will be uncorrelated. However using time-series data where the observations follow a natural ordering through time, there is the possibility that the successive error could be correlated with each other. Autocorrelation means that the current error term not only contains the effect of the current situation but also the carryover from previous happenings. To detect autocorrelation the Durbin-Watson test could be used. The existence of autocorrelation has the same consequences for the regression model as heteroskedasticity.

III.5.4. Collinearity

There could be some collinearity between trade openness and financial integration; however it is not expected to be an exact collinearity. According to Hill et al(2008) only exact collinearity is a violation of least squares assumptions therefore if the estimated equation has the

coefficients which are precisely estimated and significant and have the expected signs and magnitudes than there is no need to recognize or mitigate collinearity. Just to make sure that the correlation between trade openness and financial integration is not too high, a sample

(23)

III.5.5. Omitted variables

According to the literature there are other factors that could be involved in transmitting of crisis from one country to the other but trade and financial linkages are the most prominent. Other explanatory variables could be monetary and fiscal policies of the countries, specialization, exchange rate volatility etc. Some of these variables are found to be significant and some not. Nevertheless the most significant were trade and financial linkages. Thanks to the limited data available it is not possible to test all these different explanatory variables so the most important one are chosen however if the sample was larger other variables could have helped to improve the model.

IV. Results

IV.1. Results concerning hypothesis 1

The pair wise correlations are calculated using statistics program SPSS 22.

The pair wise correlations of each developing country with each developed country are calculated for time periods of 2004Q1-2007Q4 and 2008Q1-2010Q4. Table 2 in the appendix contains these pair wise correlations for 2004Q1-2007Q4 and table 3 in the appendix holds the pair wise correlations of 2008Q1-2010Q4.

Table 1contains the average of pair wise GDP correlations of each developing country with the developed countries for both time periods and also the average difference of these two time periods.

The average GDP correlation of each developing country with the developed countries for the time period of 2008 to 2010 is higher than for the period of 2004 to 2007, except for Kyrgyz Republic. On average the correlation of developing countries with the developed countries has increased by 0.54 points after the crisis.

The first hypothesis states that there is a higher GDP co-movement between developing countries and the developed countries after the crisis which is true except for Kyrgyz Republic which has a higher average GDP correlation with developed countries before the crisis.

However all the other developing countries have a higher average GDP correlation with developing countries after the crisis of 2007.

Antonakakis and Scharler found an increase of 0.2 points of GDP growth rate correlations among G7 countries during the 2007-2009 recessions in the United States (2012).

(24)

Table 1: Average pair wise GDP Correlation of Developing Countries with Developed Countries

Country 2004Q1-2007Q4 2008Q1-2010Q4 After the crisis - before the crisis Average

Argentina -0,11 0,81 0,92 Belarus 0,09 0,71 0,62 Bolivia -0,02 0,52 0,54 Botswana -0,07 0,53 0,60 Brazil 0,07 0,78 0,71 Brunei Darussalam -0,02 0,20 0,22 Bulgaria -0,03 0,53 0,56 Chile -0,12 0,78 0,90 Colombia 0,14 0,77 0,63 Costa Rica 0,25 0,77 0,52 Croatia 0,26 0,67 0,41 Georgia 0,11 0,73 0,62 Guatemala 0,14 0,58 0,44 Hungary -0,02 0,81 0,83 India 0,07 0,20 0,13 Indonesia -0,02 0,79 0,81 Jamaica 0,19 0,73 0,54 Jordan -0,1 0,38 0,48 Kyrgyz Republic 0,18 -0,03 -0,21 Latvia 0,2 0,80 0,60 Lithuania 0,18 0,78 0,60 Malaysia 0,09 0,77 0,68 Mauritius 0,25 0,45 0,20 Mexico 0,16 0,81 0,65 Morocco -0,06 0,06 0,12 Peru 0,09 0,74 0,65 Philippines 0,12 0,76 0,64 Poland 0,19 0,72 0,53 Romania 0,05 0,56 0,51 Russian Federation 0,25 0,81 0,56 South Africa 0,07 0,76 0,69 Thailand 0,11 0,72 0,61 Tunisia 0,27 0,39 0,12 Turkey -0,09 0,72 0,81 Ukraine 0,08 0,84 0,76 0,54

The second hypothesis states that there is a higher GDP co-movement among emerging countries from the same region than among these emerging countries and developed countries.

(25)

Table 4a: Average GDP Correlation of Central & Eastern Europe (CEE) with Developed countries period 2004Q1-2007Q4

Looking at the table 4a shows that the average of GDP correlation among the Central and Eastern Europe (CEE) themselves and also their average GDP correlation with the developed countries for period 2004Q1-2007Q4. Only Poland among these countries has a higher average GDP correlation with developed countries than with the other emerging countries in (CEE) group.

On average the (CEE) group of countries have0,067 points higher correlation among themselves than with the developed countries.

Table 4b: Average GDP Correlation of Central & Eastern Europe (CEE) with Developed countries period 2008Q1-2010Q4

Table 4b contains the average GDP correlation of the same group of countries but for the period of 2008Q1-2010Q4. For this time period Hungary and Romania have a higher average GDP correlation with developed countries than with the countries from (CEE) group. Poland that had a higher correlation with developed countries in the period 2004Q1-2007Q4 now has a higher correlation with other countries of (CEE) group.

On average the (CEE) group of countries for the period of 2008Q1-2010Q4 have 0,016 points higher correlation with developed countries than with the countries from (CEE) group.

Table 5a: Average GDP Correlation of Developing Asia with Developed countries period 2004Q1-2007Q4

(26)

Table 5a shows the average GDP correlations of Developing Asia countries among themselves and also their average GDP correlation with the developed countries for period 2004Q1-2007Q4.

In this group only Philippines has a higher average correlation with the other emerging countries from the Developing Asia group than with the developed countries.

On average the countries from the Developing Asian group have 0,127 points higher correlation with developed countries than with the other countries from the Developing Asia countries for period 2004Q1-2007Q4.

Table 5b: Average GDP Correlation of Developing Asia with Developed countries period 2008Q1-2010Q4

Table 5b contains the average GDP correlations of the same group but for the period 2008Q1-2010Q4. For this time period India and Thailand have higher correlations with the other countries from the Developing Asia group than with the developed countries. Philippines that had a higher correlation with other Developing Asia countries than with developed countries for period 2004Q1-2007Q4 now has a higher correlation with developed countries than with the other countries from the Developing Asia countries.

On average the countries from the Developing Asia group have 0,053 points higher correlation with the developed countries than with the countries from the Developing Asia group.

Table 6a: Average GDP Correlation of Latin America & the Caribbean (LAC) with Developed countries period 2004Q1-2007Q4

Table 6a shows the results of average GDP correlations among the countries of Latin America and the Caribbean (LAC) themselves and also their average GDP correlation with the

developed countries for the period 2004Q1-2007Q4. In this group only Mexico has a higher average correlation with developed countries than with the other countries from the (LAC) group.

(27)

Table 6b: Average GDP Correlation of Latin America & the Caribbean (LAC) with Developed countries period 2008Q1-2010Q4

Table 6b has the results of the average correlation of the same group of countries for the period 2008Q1-2010Q4. In this period all the countries have higher correlations with other countries from the (LAC) group than with the developed countries. Even Mexico has now higher correlation with other countries from the (LAC) group than with the developed countries. On average the countries from the (LAC) group have 0,084 points higher correlation with other countries from the (LAC) group than with the developed countries.

The countries from the Latin America and Caribbean group have higher GDP correlations among themselves than with the developed countries for both time periods. The countries from this group have converged among themselves.

The countries from the Developing Asia group have higher GDP correlations with developed countries than with the other countries from the Developing Asia group for both time periods. Therefore it could be assumed that there is a higher GDP co-movement among these countries and developed countries than with the other countries from the Developing Asia group.

The countries from the Central and Eastern Europe (CEE) group have a higher GDP correlation among the countries from the (CEE) group than with developed countries for the period

2004Q1-2007Q4 but they have a higher GDP correlations with the developed countries for the period of after the crises thus 2008Q1-2010Q4.

In conclusion it is not possible to say that the emerging countries are decoupled from the developed countries. Each group of countries needs to be examined separately and each time period could have different results for the same group of countries. Therefore there is no

definitive answer to the question considering the decoupling of the emerging countries from the developed countries and convergence of these emerging countries among themselves.

The last hypothesis states that the crisis of 2007 which started in United States of America has had an effect on the propagation mechanisms: trade and financial integration in developing countries. In a nutshell H3 states that



3 ≠ 0. The alternative hypothesis states that the crisis has

had no effect on the propagation mechanisms and therefore



3 = 0.

(28)

The results of the fixed effects regression can be found in table 7 in the appendix and the results of random effect regression model are in table 8 in the appendix. Table 9 displays a summary of the results of these two models.

Table 9: The results summary of the fixed and random effects model using EViews

Variable Fixed Effects Random Effects

C 0.162194 0.037571 T -146.2676 152.1484 F -270.6364 162.9589 Dummy 0.542976 0.557408 Random Effects(Cross) 0.0000 Idiosyncratic Random 1.0000 R-squared 0.863853 0.725761 Adjusted R-squared 0.705014 0.712491 S.E. of regression 0.178347 0.176072

Sum squared resid 0.954230 1.922090

F-statistic 5.438566 54.69339

Prob(F-statistic) 0.000005 0.000000

Durbin-Watson stat 2.792816 1.928546

As has been mentioned before to decide which one of these models is best suited for our data, the Hausman test is conducted and the results of this test can be seen in table 10 in the

appendix. The null hypothesis in the Hausman test states that there is no correlation among unobserved characteristics and the explanatory variables. The null hypothesis can be rejected if the P< 0,05 on 95% significant level which means that there is a correlation among unobserved characteristics and the explanatory variables. Therefore the fixed effect model is the best option. Looking at table 10 it states a warning that the estimated cross-section random effects variance is zero, thus there is no evidence of individual effects. With such result, it appears the model is not efficient for a computation of Hausman test variance. Therefore the random effect model does not appear to be the best option here. Based on this result the fixed effect model is chosen as the preferred panel regression model.

Table 11: Statistic summary of fixed effect model

Variable Coefficient Std. Error t-Statistic Prob.

C 0.162194 0.250093 0.648537 0.5216

T -146.2676 1225.820 -0.119322 0.9058

F -270.6364 768.7698 -0.352038 0.7273

DUM 0.542976 0.055052 9.862984 0.0000

Table 11 shows a summary of fixed effect model. According to these results the only coefficient which is significant is the DUM coefficient. The DUM coefficient in this table stands for



3 in the

y

ad

c

(0.250093) (1225.820) (768.7698) (0.055052)

Wald Test:

Equation: Untitled

Test Statistic Value df Probability t-statistic 9.862984 30 0.0000 F-statistic 97.27846 (1, 30) 0.0000 Chi-square 97.27846 1 0.0000 Null Hypothesis: C(4)=0

Null Hypothesis Summary:

Normalized Restriction (= 0) Value Std. Err.

C(4) 0.542976 0.055052

Restrictions are linear in coefficients.

IV.4. Results concerning roadblocks

The normality of the errors was one of the issues mentioned in the roadblock section. The histogram normality test can be seen as figure 4 in the appendix section. The Jarque-Bera test was conducted by EViews and result is 5,35 which is less than the critical value of 5,99 therefore the null hypothesis of normally distributed errors can not be rejected.

The heteroskedasticity test was not conducted since EViews does not have a heteroskedasticity test for panel data analysis.

(30)

will be near zero. If there is negative correlation, the statistic will lie somewhere between 2 and 4. The 2,79 test result here indicates a negative serial correlation but it is not alarming.

To test for the collinearity in the data a correlation test was conducted between the explanatory variables. The correlation results can be seen in table 14 in the appendix. The pairwise

correlation between trade and financial integration is 0.54. Although there is some collinearity between trade openness and financial integration it is not an exact collinearity and therefore it is not in violation of the least squares assumption.

The omitted variables obstacle is controlled by the fixed effect model therefore that is not an issue here.

V. Conclusion

The crisis of 2007 was originated in U.S. but soon it spread to other developed countries and developing countries as well. One of the major players in the global economy today is China. China had a slight decrease of output growth rate after the crisis in 2008 but it started to increase again till 2010 but it has been decreasing since then. China has not been able to reach its impressive output growth rate level of before the crisis. Another developing country that plays a major role in world economy today is India. India on the other hand had been enjoying an increasing output growth rate between 2008 till 2011 but it had a decreasing output growth rate for 2011 and 2012. After 2012 its output growth rate has been increasing again but India has not been able to reach its highest output growth rate level of 2010 again. Thus developing countries with really impressive output growth rate like China and India were also affected by the crisis of 2007 however the signs were not visible simultaneously.

Looking at the results of first hypothesis the average pairwise GDP correlation of developing countries with the developed countries before the crisis is only 0,09 but after the crisis it increases to 0,63. During the crisis and uncertain economic situation there is more volatility in the economies therefore it is natural that the correlation among countries would increase during the crisis period however an increase of 0,54 point of GDP correlation between developing countries and the developed countries is unexpected. Specially when Antonakakis and Scharler found only an increase of 0.2 points of GDP growth rate correlations among G7 countries during the 2007-2009 recessions in the United States (2012). The crisis in the developed world has an effect on the economy of the developing countries.

Regarding the decoupling of emerging countries from the developed world the results were less straightforward. The countries from Latin America and the Caribbean were more converged among themselves before and after the crisis. The crisis has only increased their convergence level. For the Central and Eastern Europe group the crisis has led to an increase of

co-movement between emerging countries from this group with the developed countries. The countries from the Developing Asia have also higher correlation with the developed countries but their correlation is less strong after the crisis than before the crisis. However it was clear that every regional emerging country group was different and further investigations are necessary to reach a conclusion regarding the decoupling of emerging countries from the developed countries.

The Effect of the 2007 Crisis on Developing Countries