Business cycle synchronization in the asean degree and driving forces

A growing literature has also shortlisted three candidates for the determinants of business cycle synchronization: trade intensity, similarity of industrial structures, and financial int

UNIVERSITY OF ECONOMICS ERASMUS UNIVERSITY ROTTERDAM

VIETNAM – THE NETHERLANDS PROGRAMME FOR M.A IN DEVELOPMENT ECONOMICS

BUSINESS CYCLE SYNCHRONIZATION IN THE ASEAN:

DEGREE AND DRIVING FORCES

BY

NHU DINH HIEP

MASTER OF ARTS IN DEVELOPMENT ECONOMICS

HO CHI MINH CITY January 2018

UNIVERSITY OF ECONOMICS ERASMUS UNIVERSITY ROTTERDAM

HO CHI MINH CITY INSTITUTE OF SOCIAL STUDIES VIETNAM THE NETHERLANDS

VIETNAM – THE NETHERLANDS PROGRAMME FOR M.A IN DEVELOPMENT ECONOMICS

BUSINESS CYCLE SYNCHRONIZATION IN THE ASEAN:

DEGREE AND DRIVING FORCES

A thesis submitted in partial fulfilment of the requirements for the degree of

MASTER OF ARTS IN DEVELOPMENT ECONOMICS

ABSTRACT

Business cycle synchronization can be understood as the symmetric movements of business fluctuations over a timeframe In the event of economic shocks, the degree of business cycle synchronization will indicate if a country can rely on its economic policy to overcome the shocks or it needs the policy coordination from other countries Business cycle synchronization is perceived as one of the irreplaceable criteria to assess the degree of regional economic integration The ASEAN is currently at the stage which can be considered to resemble the early stages of the European Union Most ASEAN countries are also members of the Asia-Pacific Economic Cooperation (APEC) forum, so the concept of business cycle synchronization becomes more and more important Empirically, there are few studies that incorporated the term “business cycle synchronization” in the analysis of a potential currency union in ASEAN However, such a study that can explore the matter directly is hard to find This study is conducted to evaluate the degree of the business cycle synchronization in the 10 ASEAN economies from 2000-2016 Additionally, the study expects to uncover the driving forces that lead to the synchronization of business cycles in the region To overcome the limitation of the old measuring technique and data constraints, this study makes use of the new measure of business cycle synchronization using Abiad, et al (2013)’s instantaneous quasi-correlation measure

A growing literature has also shortlisted three candidates for the determinants of business cycle synchronization: trade intensity, similarity of industrial structures, and financial integration These three variables are tested for their explaining powers in a regression model

To reduce the possibility of the endogeneity problem, control variables (the product of log GDP, the product of log population and per capita GDP difference) are added to the model to control for omitted-variables bias Also, lagged values of right-hand side variables are used to prevent the reverse causality To purge the autocorrelation and heteroskedasticity existed in the model, the Feasible Generalized Least Squares estimator is the main estimation method in this study (OLS and Random-effects results are used as references)

The calculation using new measure signifies that the degree of business cycle synchronization in ASEAN is moderate from 2000-2016 The level of synchronization reaches

to the highest points in 2009, which was the time when the effects of global financial crisis took place Correlations of business cycles among ASEAN-5 members are generally higher than among the CLMV group, or among the whole region

The regression-based analysis from this study also presents some key findings Bilateral trade intensity and the similarity of industrial sectors are positively correlated with the synchronization of business cycles Their explaining powers are robust and consistent across models The mutual capital control is likely to weaken the degree of synchronization Capital control is not significant in OLS estimate, fairly significant in Random-effects model, but highly significant in FGLS model

In summary, empirical evidence from this study underlines the three positive determinants of cycle synchronization: (i) trade integration; (ii) industrial similarity; and (iii) financial integration As such, relevant policy implications are also discussed

Financial integration JEL Classification: C21, E32, F15

ACKNOWLEDGEMENTS

I would like to seize this chance to express my gratitude to Doctor Vo Hong Duc for his patience and sharp critiques while supervising me Without his valuable comments, I could never complete my thesis

Special thanks are also due to Vo The Anh for his assistance in econometric techniques

CONTENTS

Abstract 3

Acknowledgements 5

Contents 7

Tables & Figures 7

CHAPTER 1: INTRODUCTION 8

1 Problem statements 8

2 Research objectives and questions 10

3 Scope of the study 10

4 Structure of the thesis 11

CHAPTER 2: LITERATURE REVIEW 12

1 The literature review 12

2 Review of empirical studies 15

CHAPTER 3: RESEARCH METHODOLOGY 25

1 Measurement methods 25

2 Estimation technique 26

3 Variable measurement and data 29

3.1 Business cycle synchronization 29

3.2 Trade integration 30

3.3 Industrial similarity 31

3.4 Financial integration 32

3.5 Other variables 32

CHAPTER 4: RESEARCH RESULTS 34

1 Business cycle synchronization in ASEAN 34

2 The driving forces of business cycle synchronization in ASEAN 37

2.1 Descriptive statistics 37

2.2 Regression results 40

2.3 Discussions 42

CHAPTER 5: CONCLUSIONS AND POLICY IMPLICATIONS 44

1 Conclusions 44

2 Policy implications 46

3 Limits of the study 47

REFERENCES 48

TABLES

Table 1: Summarized specification link test result

Table 2: Ramsey RESET test using powers of the fitted values of QCORR

Table 3: Test of overidentifying restrictions: fixed vs random effects

Table 4: Tests of heteroskedasticity and autocorrelation

Table 5: Descriptive statistics of variables

Table 6: Correlation matrix among variables

Table 7: OLS, Random Effect and FGLS regression results

FIGURES

Figure 1: The mechanisms of business cycle synchronization

Figure 2: The dynamics of business cycle synchronization in ASEAN (2000-2016)

Lucas (1977) defined business cycles as the recurrent fluctuations of macroeconomic aggregates around trend Business cycle synchronization means that similar movements in countries' business cycles exist over time Within the context of this thesis, these terms

“business cycle synchronization”, “correlation of business fluctuations”, “cycle co-movement”

or “symmetric business cycles” are used synonymously Furthermore, the term “business cycle synchronization” should not necessarily be confused with “economic convergence” as the latter refers to the catch-up effect when the poor or less developed countries have faster growth rates than the rich ones, narrowing down the gaps between their economies

The degree of cycle synchronization is of great importance because it shows the necessity

of policy coordination among integrated economies If economic shocks are purely specific, then national fiscal and monetary policies can help the country get back to the original equilibrium However, if the shocks are common, or one national shock spreads beyond borders, then uniform or cooperative policy interventions should be more effective This is not only essential for the ASEAN region but also for any economic bloc such as the MERCOSUR, the South Asia region, the ANZCERTA, or the East Asia region Business cycle synchronization becomes a key measure of economic integration (Dorrucci, et al 2002) ASEAN is one of the most vibrant economic regions in the world However, the ASEAN regional integration process is very different from the European countries the 20th century While EU established its single market twenty years ago, it took ASEAN almost twenty three years (from 1992 to 2015) to finalize the ASEAN free trade area (with the formation of the

country-AEC) The profound differences in culture, history or political systems as well as the current level of economic development between the ASEAN old members and new ones could be one

of the reasons for the ASEAN economies are difficult to retrace the footsteps from the EU.Thus, some skeptical arguments arise in response to the true economic integration among members within the ASEAN Against all odds, the ASEAN members are now making efforts

to strengthen economic and financial cooperation regionally and globally Out of 10 ASEAN members, 7 of them (Brunei Darussalam, Indonesia, Malaysia, the Philippines, Singapore, Thailand, and Vietnam) also belong to the Asia-Pacific Economic Cooperation (APEC) forum Finding new driving forces for regional economic integration is one of the key commitments

of the APEC leaders in the latest Da Nang Declaration 2017 in Vietnam A question is raised,

‘What are the driving forces?’ Clear insights of business cycle synchronization now become integral to answer the question convincingly

Extensive research, such as empirical papers by Frankel and Rose (1997, 1998), Baxter and Kouparitsas (2005), a series of papers by Imbs (particularly Imbs and Wacziarg, 2003; Imbs 2004, 2006), or Calderon, Chong and Stein (2007) has indicated some impacts of trade intensity, specialization and financial integration on business cycle synchronization Böwer and Guillemineau (2006) which studied European countries solely also shortlisted some key determinants of business cycle synchronization Similar papers which include an Asian context (Kumakura, 2006; Park and Shin, 2009; Duval, 2014) also pointed out possible factors that drive a business cycle synchronization

The concept of business cycle synchronization in ASEAN was mentioned in several papers that examined the suitability of a currency union in ASEAN, such as Bayoumi, Eichengreen and Mauro (2000) or Bacha (2008) However, studies on the determinants of business cycle synchronization in the ASEAN region exclusively are limited or virtually non-existent One of the biggest obstacles is that the traditional estimation of business cycle synchronization does not allow econometric regressions to be properly conducted due to the lack of data for interested countries Normally, the calculation of business cycle synchronization is based on the Pearson correlation of actual or de-trended time series for each country pair over a window period This requires either a large amount of country pairs or many rolling windows to maintain the required number of observations Nevertheless, the statistics data for ASEAN countries are limited, especially for the less developed members

2 Research objectives and questions

Analysis of the business cycle synchronization is important and hotly debated for the ASEAN region In response to this debate this study is conducted to achieve the following two key objectives

- First, an assessment of the degree and dynamics of business cycle synchronization can

be used as a measurement of the regional economic integration With a high degree of business cycle synchronization, regional policymakers are more likely to respond with

a common policy and they cooperate with each other to stabilize economic shocks This implies ASEAN is a real emerging powerhouse The goal is to unfold the current status

of business cycle synchronization in ASEAN’s economies

- Second, understanding the driving forces of business cycle synchronization will play a

crucial role in forming sustainable governance for the ASEAN region Each country member will adjust the economy system respectively to enhance the synchronization The second objective of this study is to discover the determinants that spur the level of synchronization

In order to achieve those stated objectives, this thesis attempts to provide the answers to the following questions:

- At what level is the synchronization of business cycles in the ASEAN?

- Which factors drive business cycle synchronization, to be named the driving forces in

the ASEAN?

3 Scope of the study

This thesis will observe the synchronization of national business cycles among 10 ASEAN countries They are: ASEAN-5 (Singapore, Thailand, the Philippines, Malaysia, and Indonesia), CLMV (Cambodia, Lao PDR, Myanmar, and Vietnam), and Brunei Darussalam Each country will pair up with the other 9 countries to make a total of 45 country pairs Due the limited availability of data, the investigation of the determinants of business cycle synchronization is confined to the period 2000-2016

4 Structure of the thesis

The structure of this thesis is organized as follows Following this Introduction, Chapter, Chapter 2 provides the theoretical framework and a brief review of relevant empirical studies

on the topic Estimation methods and data requirements are presented in Chapter 3 Chapter 4 provides a summary of the empirical findings and discussions, followed by the concluding marks and policy implications in Chapter 5

CHAPTER 2:

LITERATURE REVIEW

1 The literature review

The concept of business cycle synchronization was first mentioned in the optimum currency area (OCA) literature that was pioneered by Mundell (1961) Mundell (1961) identified that asymmetry of shocks is one of the key sources that push up the cost of adopting

a single currency Business cycle synchronization or the symmetry of economic cycles is identified as one precondition of the formation a currency union McKinnon (1963) added to the OCA literature by emphasizing the importance of the openness of the economies The more the country is integrated to the world, the more benefits it can enjoy from the flexible exchange regime A common currency may be beneficial to countries that share close economic ties He postulated trade intensity or integration as a prerequisite criterion for the candidates of a currency union Kenen (1969) cited sector diversification as a precondition He underscored that the more diversified the economy structure is, the less likely asymmetric shocks occur As less diversified economies are usually small economies

The second contribution of Kenen (1969) is the argumentation of fiscal transfers: In the event that there are region-specific asymmetric shocks, redistribution of financial resources from the rich countries to the poor ones might be effective especially when market-based mechanisms are absent However, this precondition requires a certain level of politics integration and is subject to controversy because wealthy nations may be unwilling to bail out less developed ones Besides the works of by Mundell (1961), McKinnon (1963), and Kenen (1969), the OCA literature also acknowledges the contributions of Ingram (1969) for the inclusion of the degree of financial integration as one OCA criterion, and Fleming (1971) for the addition of similarity of inflation rates

The consensus on the important determinants of business cycle synchronization was initiated in the studies by Frankel and Rose Frankel and Rose (1997) contributed to the OCA literature by introducing the hypothesis of endogenous OCA properties The modern approach

to OCA relies upon the assumption that the currency union automatically creates conditions for its optimality Potential entrants to a currency union may satisfy the OCA criteria after joining the union (ex post), although they did not fulfill them initially (ex ante) Frankel and Rose (1998) argued that more integration might result in more trade due to reduced trade

barriers and more trade might lead to more correlated business cycles This gave rise to the first candidate of business cycle synchronization determinants: trade integration However, Krugman (1993) showed an opposing view in that higher trade integration will bring higher possibility of asymmetric shocks due to higher specialization of industrial activities He reasoned that lower transaction costs increase the economies of scale, and thus entail higher concentration on production In such a case, trade intensity induces lower level of business cycle synchronization

The next possible explanation for business cycle synchronization is the similarity in industrial specialization across concerned economies If two countries share similar economic sectors, they are expected to cope with similar sector-specific shocks This connection has been highlighted in a string of papers by Imbs from 1998 to 2004, especially Imbs and Wacziarg (2003) and Imbs (2004) Notwithstanding, empirical studies on the impact of the patterns of specialization on business cycle synchronization present conflicting evidence

Financial integration is theoretically considered as the third factor that may affect the extent of business cycle synchronization Like trade intensity, there exist the ambiguous views

on the impact of financial integration Financial integration is the third major field of determinants Kalemli-Ozcan et al (2003) presumed that financial integration should dilute the extent of business cycle synchronization through the channel of specialization They stated that

a high degree of financial integration is likely to be associated with greater sectoral specialization and more asynchronous business cycles On the contrary, Kose et al (2003) argued that financial integration raises international spillovers of economic fluctuations, and therefore amplifies business cycle synchronization

The direct and indirect channels of business cycle synchronization can be portrayed in Figure 1 Both trade integration and financial integration have direct and indirect mechanisms

to influence the degree of co-movement of business cycles

Figure 1: The mechanisms of business cycle synchronization

Source: Adopted from Imbs (2004)

In Figure 1, the relationships through the direct channel are presented by the coefficients

could predict this relationship Most empirical papers cited the works of Baxter (1995) Conversely, specialization of industrial sectors or the dissimilarity of sectors is expected

countries produce the same types of goods, they are likely to experience the same specific shocks, and the reverse is true Kraay and Ventura (2007) reasoned that comparative advantage is the incentive that rich countries specialize in industries that require new technologies, while poor countries stick to labour-intensive techonologies Such asymmetry in the production patterns can cause business cycles asynchronous

sector-The direct effect of financial integration is, however, uncertain (which leaves the sign of

banks may increase lending funds in unaffected countries and decrease lending in those which are hit hard This may spur GDP correlations However, because of imperfect information, investors may respond to negative shocks by withdrawing their capitals simultaneously everywhere Thereby, financial integration can actually decrease output correlations

Simultaneously, through the indirect channel, both trade integration and financial are

difference in trading prices can be a good motivation for specialization Financial linkages may

help investors to shift their capital flows to speculative projects, enabling economies to sharpen their comparative advantages by specialization Because specialization will lower cycle synchronization, trade integration and financial integration are assumed to decrease co-movement of business cycles as well

2 Review of empirical studies

Frankel and Rose (1998) was the first empirical paper to discover the positive linkage of trade intensity and business cycle synchronization By reasoning the endogeneity property of the OCA criteria stemming from Frankel and Rose (1997), their study focused on the first two OCA criteria which are the correlation of business cycles and trade intensity From their theoretical perspectives, trade ties may either increase or decrease the cycle synchronization, which becomes the motivation for them to conduct an empirical investigation

Their sample includes a group of 21 industrialized economies around the world from

1959 through 1993 This may be subject to sample selection bias because the list does not include developing countries

As one of the early studies that deals with business cycle synchronization, there is lack

of optimal measurement and estimation methods They introduced a very simple regression equation with correlation of business activities being the explained variable, and bilateral trade intensity being the sole explanatory variable To measure correlations of business activities, four activities are used: real GDP, industrial production, employment and the unemployment rate They also applied 4 different detrending techniques to these four business activities, one

of which is the well-known Hodrick–Prescott filter This results in a total of 16 different measures of business cycle synchronization To measure trade intensity, they have 2 different kinds of measures: bilateral trade over total trade, and bilateral trade over GDP There are actually total 32 regressions in their study As the index of business cycle correlations is calculated over a period of time, their time frame from 1959 to 1993 is divided in to 4 periods

As such, the panel dataset is constructed with 210 country pairs and the sample size is 840 Acknowledging the omitted-variables bias, Frankel and Rose (1998) used instrumental variables to instrument for the trade intensity variables The instruments are: graphic distance

in logarithm form, common border dummy, and common language dummy, which comes from the trade literature After considering all the regression results, they came to a conclusion that bilateral trade intensity brings more correlated business cycles

Imbs (2004) was a continued study from a series of papers by the same author from

1998 He extended the list of determinants of business cyle synchronization by adding specialization and financial integration He was also the first that discussed a complex relationship among the three possible factors that may affect cycle synchronization As depicted in the Figure 1, Imbs (2004) assumed that beside the direct impacts, trade integration and financial integration have indirect impacts (through specialization) on cycle co-movemet Since the idea was new in the literature, he tested his view empirically by studying 24 countries with 276 country pairs from the 1980s and 1990s Most of the countries in the list are industrialized economies where the data are available

Based on the relationship of the three factors, his study involves a system of simultaneous equations: (i) business cycle correlation is regressed on bilateral trade intensity, bilateral financial integration, specialization index, and exogenous vectors, (ii) trade intensity is regressed on sector specilization, and exogenous vectors (iii) sector specialization is regressed

on financial integration, trade integration and exogenous vectors, and (iv) financial integration

is regressed on some exogenous vectors To address the endogenity problem, he applied the stage least square method which includes the usage of instrumental variables

3-In Imbs (2004), the dataset is a cross-section one with 276 obervations The index of business cycle synchronization is calculated by detrending the time series of quarterly GDP data using Baxter-King filter The index is calculated for the period of 20 years Sector specialization index is calculated by the the sum of absolute value of sector share differences between a country pair (averaged by the time frame) The trade variables are similar to those from Frankel and Rose (1998) The financial integration variable is calculated by the indices

of capital restrictions and risk-sharing

The paper concluded that business cycles synchronization is driven by trade (positively), specialization (negatively) and financial integration (positively) Besides, he also found that trade has no significant impact on specialization, while the impact of financial integration on specialization is positive

Imbs (2006) reexamined the effect of financial integration on cycle co-movement The theoretical framework is similar to Imbs (2004) in which financial integration afftects output fluctuations through direct and indirect channels He stated that goods atrade and specialization are two possible indirect channels Based on that, his empirical estimation adopted a simultaneous equation approach In the main equation that connects business cycle

synchronization and its determinants, the author used the variable of similarity of insdustrial sectors instead of the specialization Other equations are in general similar to those in Imbs (2004)

The study used a cross-section dataset of 41 countries with core-periphery pairs The time frame is from 1960-2000 Imbs (2006) used the Hodrick–Prescott filter to detrend GDP data as

a step of correlation calculation One important note is that he separated GDP correlations from consumption correlations, which means that there are two measures of business cycle synchronization in the study The papers also used a wider measurement of financial integration Apart from the indices in Imbs (2004), he also used Dennis Quinn’s measure of

on the IMF’s Coordinated Portfolio Investment Survey data Actually, the CPIS estimations only cover portfolio investment data To address the endogeneity problem, Imbs (2006) employed both IV and GMM estimators

The empirical results reaffirmed those in Imbs (2004) that financial integration leads to synchronization Particularly, financial integration is positively correlated with goods trade and specialization

Baxter and Kouparitsas (2005) conducted an in-depth extreme bound analysis of factors that may affect the business cycle correlations The purpose of the paper is to detect any relationship between business cycle correlations and economic variables Beside the variables which come from the literature, many explanatory variables and gravity variables are included

in the model to test the robustness: bilateral trade, similarity in industrial structure, trade similarity, factor endowments, currency union dummy, graphical distance, common language, bilateral population variable and others

The estimation method was based on multivariate cross-sectional analyses using OLS estimator with or without two dummy variables to capture the fix effects of country pairs Baxter and Kouparitsas (2005) analyzed a dataset of over 100 countries (5670 country pairs), including both developed and developing countries from the period 1970-1995 The index of cycle correlations is calculated from the correlation of cycle components after using Baxter-King filter to detrend annual GDP series They also provides various measures of explanatory variables to test the robustness, especially the calculations of similarity of industrial structure, export goods basket and import goods basket are different from Imbs (2004)

Baxter and Kouparitsas (2005) confirmed that trade intensity is robustly correlated with business cycle synchronization However, other findings contradicted previous studies Similarity in industrial structure is not a significant factor The variable for financial integration, which is a dummy variable whether countries belong to a currency union in this study, is also insignificant They also found no robust links between all gravity variables included in the model except for graphical distance

Bower and Guillemineau (2006) carried out extreme-bounds analysis similar to Baxter and Kouparitsas (2005) to determine the determinants of business cycle co-movement in 12 European countries from 1980-2004 The goal is to find more potential candidates to the determinants, rather than focusing on the traditional ones Consequently, apart from the traditional variables like trade intensity, specialization or similarity of industrial structure, financial integration, the paper introduced more explanatory variables including policy and structure indicators They are: bilateral bank liability flows, real short-term interest rate differentials, nominal exchange rate volatility, fiscal deficit differentials, price competitiveness differentials, stock market differentials, trade union membership differentials, employment protection differentials, relative population and graphical distance

Like Baxter and Kouparitsas (2005), the estimation methods of Bower and Guillemineau (2006) relied on cross-sectional analyses The two authors applied pooled OLS regressions with Newey-West standard errors to address for heteroskedasticity and auto-correlation in the residuals

The dataset contains 66 country pairs over the 1980-2004 period Two sub-periods are from 1980-1996 and from 1997-2004 They employed Baxter-King filter on annual real GDP series in the calculation of the dependent variable (i.e, business cycle synchronization of a country pair) The authors also provided various measurements for explanatory variables to test for robustness The proxy for financial integration variable in this study is the bilateral bank flows (bilateral assets and bilateral liabilities) They also separate trade specialization from economic specialization, together with more detailed product specialization

The findings in Baxter and Kouparitsas (2005) confirmed the role of bilateral trade intensity (measured either by the ratio of bilateral trade to total trade or GDP) However, the paper doubted the candidacy of specialization (either trade specialization or economic specialization) In fact, trade specialization and economic specialization are only robust or quasi-robust in some sub-periods, but not in the whole period of the sample Out of other “new”

gravity variables, the paper shortlisted some key determinants of synchronization: price competitiveness differentials, the fiscal deficit differentials and the stock market differentials

Calderon, Chong and Stein (2007) analyzed the impacts of trade and specialization on cycle comovement with an application on 147 countries from 1965-2004 Out of 147 countries, there are 23 industrialized countries and 124 developing countries The study also included an analysis of the issue with a sample of Latin American countries vis-à-vis China and India With a focus on trade, the baseline regression equation in Calderon et al (2007) included

a variable for intra-industry They reasoned that specialization in industrial sector may lead to intra-industry trade, and thus may not necessarily cause asynchronous output correlations as (a result of industry-specific shocks) However, the role of financial integration in the literature

is absent in the equation

Calderon et al (2007) divided the time frame into 4 sub-periods (1965-1974, 1975-1984, 1985-1994, and 1995-2004) As such, the dependent variable is calculated by the correlations

of cycle components of real GDP (using Baxter-King filter) every 10 years The measurement

of bilateral intensity and industrial similarity is generally similar to previous studies The “new” variable, intra-industry trade intensity, is calculated based on the Grubel-Lloyd index

The study applied least square estimation method with fixed effects for country pairs and also for country groups To deal with endogeneity issues, they utilized instrumental variable estimation methods with instruments that come from the trade literature (for example, graphical distance, common border dummy, geographic remoteness, population density, colonial origin, common language dummy)

Calderon et al (2007) confirmed that trade intensity and similar structure of production are two key factors that spur cycle co-movement in the full sample analysis However, in the sample of Latin American countries vis-à-vis China and India, they found that the dissimilarity

of production structure or trade structure is positively related with output correlations

Kumakura (2006) focused on the role of trade on business cycle co-movement in 13 Asia Pacific economies from 1984–2003 Those countries are: Australia, China, India, Japan, Korea, New Zealand, Taiwan, the United States and ASEAN-5 (Indonesia, Malaysia, the Philippines, Singapore, and Thailand)

The regression model in Kumakura (2006) is simple with busines cycle synchronization regressed on trade intensity, similarirty index of commodity structures, capital movement

correlations There is one unique variable in the model of Kumakura (2006) is the index of the world electronics cycle According to the author, some East Asian countries rely heavily on the electronics sector

The dataset in Kumakura (2006) is a cross-sectional one with 78 country pairs The

GDP time series The author also designed unique measurements for trade intensity, and specialization Regard the finanical integration in the literature, he used the correlation of capital inflows as a share of GDP

To address the endogeneity concerns, his estimation strategy is the IV regressions He tested instruments for the trade variable (graphical distance, common border dummy, common language) and for the variable of capital inflows correlations (the gaps in the per capita GDP, the minimum per capita income) The final outcome of Kumakura (2006) is that trade induces cycle synchronization In addition, the specialization of those countries in electronics industry also drives co-movements

Inklaar et al (2008) again tested empirically the position of trade in relationship with business cycle synchronization The authors focused on a sample of 21 OECD countries from 1970–2003 The study came up with a system of equations that is partly similar to those in Imbs (2004): (i) synchronization is regressed on trade, specialization, fiscal policy, monetary policy, exchange rate variability; (ii) trade is regressed on specialization, fiscal policy, monetary policy, exchange rate policy and gravity variables; (iii) specialization is regressed on trade, financial integration and gravity variables The estimation strategy of Inklaar et al (2008)

is the multivariate model, which is estimated by the OLS and 3-stage least square methods This is different from previous studies which relied on the use of instrument variables

As the time horizon is divided in 11-year period, the dataset of Inklaar et al (2008) is a panel data with 3 periods (with total 630 observations) The cycle synchronization index is calculated by using the Baxter-King filter on annual GDP and industrial production data (with exception for Australia, New Zealand and Switzerland, they used the reported quarterly production data) Unlike other studies which keep the original values of correlations, they converted those values so that they are not bounded from the range of -1 to 1

To calulate the similarity of fiscal policy, they use the correlation of budget deficits which are cyclically adjusted Similarly, the similarity of monetary policy is based on the correlation of short-term interest rates The authors contributed that trade intensity,

specialization and policy congruence (i.e the similarity of monetary and fiscal policy) have a significant impact on business cycle synchronization

Dées and Zorell (2012) studied the impacts of trade and financial integration on the correlation of business fluctuations Their whole sample covers 56 countries from 1993-2007 This sample includes many emerging economies, particularly in Central and Eastern Europe along with others from OECD

Inspired from both the theoretical and empirical frameworks of Imbs (2004, 2006), their model contains a system of four equations: (i) synchronization is regressed on trade, finance, specialization and exogenous variables; (ii) trade is regressed on finance, specialization and exogenous variables; (iii) specialization is regressed on trade, finance and exogenous variables; (iv) finance is regressed on a group of exogenous variables

Similar to previous studies with a system of equations, Dées and Zorell (2012) also applied three-stage least squares method to cope with the concern of endogeneity Also, with a large sample with maximum of 964 country pairs, the authors divided it into two sub-samples: OECD countries and Europeean countries

The data used in Dées and Zorell (2012) is a cross-sectional one The measurement of business cycle synchronization is based on the Hodrick–Prescott detrending filer The author also calculated using the Baxter-King filter, Christiano-Fitzgerald filter, and simple year-on-year log-differences technique to test for robustness Measurements of other explanatory

proxy for financial integration In general, we can say Dées and Zorell (2012) replicated the study of Imbs (2004, 2006) with updated dataset

The major findings of the study confirmed the candidacy of trade integration, similar patterns of sectors, and financial integration as determinants of cycle co-movement The only difference is that financial integration promotes the similarity in sectoral specialization, hence

it indirectly causes business cyles to be synchronous

The study of Kalemli-Ozcan, Papaioannou and Peydró (2013) focused on the impact of financial Regulation and financial globalization on synchronization As previous empirical papers that mainly examine cross-sectional variation, the methodology of Kalemli-Ozcan et al (2013) focuses on changes over time, i.e time-varying bilateral financial data Their study analyzed the database of 18 rich countries (with 150 country pairs) from 1978 to 2006 To be

specific, they used the banking data from the BIS International Locational Banking Statistics Database which contain the information of bilateral external assets and liabilities They divided into 2 measures of banking integration: total assets and liabilities over the sum of populations, and total assets anfliabilities over sum of GDPs of two involved countries

The authors also adopted two measures for synchronization which are not based on the colloquial detending techniques The first one is the negative value of the absolute value of real GDP growth (in logarithm form) differences between two countries The second measure is more complex which involves the residuals of the regression of real GDP growth on fixed effects and time effects

However, the intial estimation model is simple with banking integration, trade and specialization as the explanatory variables for synchronization They then contructed an intrumental variable estimation In the first stage of this, banking integration is regressed on the variable of harmonization of financial legislation and the variable of the flexibility of exchange rate regime In another words, they use those variables to instrument for the banking integration variable in the baseline OLS estimates The results of the 2 stage least squares estimation suggest that banking integration lowers the level of cycle co-movement

Xie, Cheng and Chia (2013) investigated the determinants of output co-movement in 12 Asia-Pacific countries for two periods of 1984–1996 and 1997-2007 The list of countries include: ASEAN-5 (Indonesia, Malaysia, the Philippines, Singapore and Thailand), South Korea, Hong Kong, Japan, China, and the 3 important trade partners (Australia, New Zealand and the United States)

Like the work of Dées and Zorell (2012), Xie et al (2013) based their study mainly on Imbs (2004) The authors considered that policy coordination might be as good as a determinant

of business cycle synchronization, so they added this variable into the system of simultanous equations in Imbs (2004) Therefore, the first equation of the system is as folows: synchronization is regressed on trade, specialization, finance and exogenous variables Policy coordination is one of those exogenous variables Policy coordination consists of two variables: the similarity of fiscal policy, and the similarity of monetary policy The estimation method is also similar They applied the same the three-stage least-square method as a solution for the possible endogenity issue

The measurement of the dependent variable is based on the detrending filters In their study, they use three different filters on the annual GDP series As the correlation of cyclical

components takes the value from -1 to 1, they followed Inklaar et al (2008) to transform the correlation values As for the financial integration variable, they used the index for net foreign assets positions The paper concluded that they identified the positive relationship of trade integration and financial integration with cycle co-movement, whereas specialization is the negative factor

Duval, Cheng, et al (2014) and Duval, Li, et al (2016) are the two papers that used the new measurement method of business cycle synchonization introduced in (Abiad, et al (2013) The new measure is called the instantaneous quasi-correlation measure and will be discussed more in details later in this thesis

Duval et al (2014) focused on the relationship of trade integration in 63 economies during 1995–2012 Out of these 63 countries, there are 34 developed economies (7 of which in Asia) and 29 emerging countries (8 of which in Asia) Trade integration in this study is segregated into trade intensity, intra-industry trade, trade specialization correlation, and vertical trade integration Their main model is constructed as follows: synchronization is regressed on trade variables (which are the four detailed variables above), financial integration and policy synchronization variables

To settle the issue of possible endogeneity problem, they employed instrumental variables for the variable of trade intensity and the variable of vertical trade integration The instruments used in the study are the product of the real outputs, graphical distance, WTO membership dummy, trade cooperation index and import tariffs, real per capita GDP difference

Since Duval et al (2014) used a new measure of business cycle synchronization which can be calculated for each year, the dataset is a panel one The paper concluded that trade intensity measure in value added term are the positve factor of business cycle synchronization Duval et al (2016) studied the relationship beteen valued-added trade and synchronization of business cycles This is actually an updated version of the working paper Duval (2014) but the model is specified differently The difference lies in the use of control variables Instead of using policy coordination variables, they use gravity variables that include similarity in production structures, product of log GDP, product of log population, and the absolute difference in log PPP GDP per capita To be more clear, synchronization is regressed

on trade intensity, banking integration, and a host of control variables

For fear that some omitted variable may be correlated with the trade intensity in the model, they applied instrumental variables The instruments are: bilateral tariffs, and the preferential trade agreements The latter is an index constructed by using binary codes for existing provisions in the preferential trade agreements

Duval et al (2016) drew the sample of Duval et al (2014) which covers 63 countries from 1995–2012 The dependent variable is the instantaneous quasi-correlation index They use only bilateral trade intensity as the proxy for trade integration The measure is similar to those in Duval et al.2014 but they converted it into logarithm form For banking integration variable, they also converted into logarithm form They simply put the neagtive sign to the specialization measure in Imbs (2004) to make the similarity index Duval et al (2016) confirmed the findings in Duval et al.(2014) that value-added trade entensity spurs business cycle correlations In a addition, we are able to see that banking integration is negatively correlated with synchronization, but similarity of industrial structure is not statistically significant

In summary, trade intensity is likely to be consistently and positively correlated with the synchronization of economic cycles The impact of specialization in most studies is negative However, the impact of financial integration varied across studies

CHAPTER 3:

RESEARCH METHODOLOGY

1 Measurement methods

This thesis aims to analyze the business cycle synchronization among the ASEAN economies

As the total number of country pairs is limited to 45, which is fairly small compared to most other studies, the choice of measurement method is crucially important to ensure econometric regressions feasible Therefore, the first task is to select the appropriate proxies for all key variables

Business cycle synchronization: Traditionally, in most previous studies, business cycle

synchronization (BCS) index is approximated by first detrending output time series using a band-pass filter, such as the Hodrick and Prescott or the Baxter and King filters When the cycle components are isolated, a Pearson correlation will determine the level of synchronization Another way is to apply the Pearson correlation on the output gaps The disadvantage of those methods is that BCS index has to be calculated over a period of time For the ASEAN context, the number of countries is limited to 10, and the data, especially for less developed members, are not adequate The usage of quarterly data is also difficult due to the lack of reported statistics data

Abiad, et al (2013) proposed an inuitive way to calculate BCS index, which becomes the

quasi-correlation measure The QCORR index can be calculated using a pair of annual growth rates, together with the mean value and the standard deviation The details of this method are presented in the section of variable measurement (3.1) The biggest advantage of this measure

is that this allows the calculation of the QCORR index can be done at any point in time rather than a time interval It is also noted that the values using this measure are not necessarily bounded between -1 and 1

Trade integration: There are two main ways to measure trade integration Trade openness is

fraction of bilateral trade over GDP, while bilateral trade intensity is a fraction of bilateral trade

as a share of total trade with the world This thesis chooses the latter way for easier interpretation