Aloisio Campelo, Jr. (IBRE/FGV)

Renata Carvalho Silva (IBRE/FGV)

Marco Malgarini (IBRE/FGV & ANVUR)

Johanna Plenk (IFO)

34th International Symposium of Forecasting, Rotterdam, June 29th, 2014

Synchronisation of Economic Cycles in Latin America: Evidence from the World Economic Survey

Outline

1. Motivations of the study

2. The data

3. Methodology

4. Results

5. Conclusions

Motivation – Theoretical background

• An increase in trade and financial linkages is often considered to lead to increasing cross-border interdependance and global convergence of business cycle fluctuations

• However, increasing productive specialisation and the emerging of large economies like India and China as global players may conversely imply less, and not more, convergence of global business cycles

• Discriminating among the two different possibilities is essentially an empirical question

Motivation – Theoretical background

• Most existing convergence literature has concentrated on industrialised economies (Bagliano and Morana, 2010; Artis et al., 2004; Mody et al., 2007)

• In a study concerning also emerging and other developing economies, Kose et al. (2012) find evidence of declining inter group convergence and increasing intra group convergence in the period 1985-2008

Motivation – Theoretical background

• For Latin America, Caporali and Girardi (2012) find that the trade channel is particularly important

• Similarly, Cesa-Bianchi et al. (2011) recognize a growing relevant role for the Chinese economy on Latin America

• In both cases, the consequences are:

• Increasing intra-group synchronisation, especially for LA countries

• Decreasing inter-group synchronisation among LA countries and industrialised countries

• However, evidence available until now is limited to the period just after the start of the financial crisis

Motivations – Research questions

• Our aim is to build on this literature, looking at:

• Intra group synchronization:

– Has the degree of synchronization among Latin American economies increased or decreased in the period 1989-2013?

– The pattern observed for LA countries is similar to what happened to more industrialized countries (Eurozone, the G8) or other developing economies (the Asian group)?

• Inter group synchronization:

– Are Latin American economies more or less synchronized now than in the past with respect to older industrialized countries?

– Has overall business cycle synchronization increased or decreased in the last twenty years?

The Data

• Most existing studies on business cycle convergence are based on trended data

• Business cycle components should hence be extracted with an appropriate filter in order to evaluate some measure of convergence/synchronisation

• The choice of the appropriate filter is however largely arbitrary, and may have a relevant impact on the results obtained, especially towards the end of the sample, when results are particuarly important from a policy perspective

• Moreover, conventional quantitative statistics as GDP or IP are particularly difficult to interpret towards the end of the sample also because they are usually (often strongly) revised

The Data

• In this paper, we use data stemming from the World Economic Survey, performed by IFO on a quarterly basis since 1989

• The survey is based on interviews with approximately 1,200 experts based in 120 countries worldwide

• Survey participants may range from a minimum of 4 for small countries to up to 50 for large ones

• Panel members are rather heterogeneous in terms of their professional occupation, ranging from multinational executives, members of economic research institutions and of national and international chambers of commerce and trade

The Data

• Survey questions are referred to current and expected country overall situation (six months ahead), trade flows and expectations on inflation and interest rate

• Individual replies are transferred to an ordinal scale that ranges from one (negative) to nine (positive), where five is neutral

• Individual replies are then aggregated for each country as a simple arithmetic mean.

• The Ifo Economic Climate Indicator (ECI) for each country is then calculated as the arithmetic average of economic assessment and expectations for the next six months.

The Data

• In the paper we focus on intra-group and inter-group convergence for:

• Latin America: Argentina, Brazil, Chile, Colombia, Mexico, Peru, Uruguay and Venezuela.

• Euro zone: Austria, Belgium, Finland, France, Germany, Greece, Ireland, Italy, Netherlands, Portugal and Spain.

• Asian countries: China, India, Indonesia, Japan, Philippines, South Korea, and Thailand

• In evaluating global convergence, we will also add the consideration of the remaining G-8 countries (Canada, Russia, United Kingdom, United States), for a total of 30 countries representing over 75% of world GDP

The Data

The Data

• Advantages of the use of IFO survey data (see also Gayer and Weiss, 2006):

• They do not require the extraction of a cyclical component because they do not contain a long-term trend (table)

• They are based on the same methodology and calculated by the same institution across a very large and diversified group of countries

• They are usually available in advance with respect to standard quantitative indicators

• They are not revised • They are highly correlated with hard data, both

considering: • A growth cycle approach • A growth rate cycle approach

The Data

Country Mean Median Standard Deviation

ADF test

Significance #Lags #obs

Argentina 4.93 5.00 1.38 ** 4 95

Brazil 5.48 5.70 1.20 ** 4 95

Chile 6.24 6.40 1.18 *** 4 95

Colombia 5.06 5.22 1.18 ** 4 95

Mexico 5.05 5.10 0.99 *** 4 95

Peru 5.88 5.95 1.36 *** 0 58

Uruguay 5.28 5.00 1.80 ** 0 95

Venezuela 3.72 3.65 1.36 ** 0 99

Latin American countries 5.07 5.20 0.72 *** 4 95

Austria 5.48 5.50 1.16 *** 4 95

Belgium 5.23 5.30 1.13 *** 4 95

Finland 5.69 6.00 1.51 *** 0, with drift 99

France 4.89 4.95 1.12 *** 4 95

Germany 5.43 5.40 1.12 *** 4 95

Greece 4.76 4.60 1.57 ** 0, with drift 99

Ireland 5.91 6.30 1.64 * 0 99

Italy 4.72 4.75 1.05 *** 4 95

Netherland 5.50 5.45 1.19 ** 4 95

Portugal 4.47 4.60 1.25 ** 0 99

Spain 4.68 5.00 1.31 *** 0, with drift 99

Eurozone 5.19 5.20 0.96 *** 4 95

China 5.22 5.23 0.95 *** 4 95

India 5.99 6.10 1.37 *** 4 95

Indonesia 5.47 5.70 1.29 ** 4 95

Japan 5.13 5.25 1.28 *** 4 95

Philippines 5.56 5.50 1.64 *** 4 95

South Korea 5.19 5.00 1.28 *** 4 95

Thailand 5.34 5.40 1.31 ** 4 95

Asia countries 5.06 5.10 0.82 *** 4 95

Canada 5.55 5.75 1.15 ** 4 95

Russia 4.48 4.40 1.69 *** 0, with drift 99

United Kingdom 5.04 5.10 1.19 ** 4 95

USA 5.36 5.50 0.91 ** 4 95

G8 5.13 5.30 0.77 *** 4 95

Cross correlation ECI and GDP growth

Countries

Lead

t=0

Lag

Maximum correlation t-4 t-3 t-2 t-1 t+1 t+2 t+3 t+4

Austria 0.34 0.47 0.60 0.71 0.75 0.68 0.51 0.28 0.10 0.75

Belgium 0.05 0.24 0.36 0.48 0.47 0.34 0.23 0.11 0.03 0.48

Canada 0.44 0.57 0.68 0.72 0.66 0.52 0.36 0.25 0.20 0.72

China 0.18 0.24 0.30 0.43 0.45 0.29 0.09 -0.13 -0.26 0.45

Finland 0.41 0.55 0.68 0.75 0.78 0.76 0.70 0.59 0.48 0.78

France 0.24 0.43 0.61 0.76 0.77 0.69 0.58 0.40 0.26 0.77

Germany 0.00 0.24 0.49 0.67 0.72 0.63 0.48 0.26 0.04 0.72

Greece 0.69 0.71 0.74 0.74 0.72 0.73 0.71 0.68 0.66 0.74

India 0.18 0.39 0.70 0.84 0.83 0.77 0.52 0.24 0.02 0.84

Indonesia 0.52 0.61 0.76 0.74 0.66 0.59 0.38 0.27 0.28 0.76

Ireland 0.64 0.73 0.79 0.83 0.87 0.80 0.76 0.67 0.58 0.87

Italy 0.36 0.53 0.63 0.71 0.70 0.61 0.50 0.35 0.20 0.71

Japan 0.02 0.23 0.44 0.55 0.55 0.41 0.25 0.13 0.08 0.55

Netherlands 0.38 0.52 0.64 0.72 0.71 0.64 0.55 0.42 0.28 0.72

Philippines 0.20 0.34 0.49 0.56 0.59 0.50 0.38 0.22 0.02 0.59

Portugal 0.62 0.69 0.78 0.82 0.79 0.71 0.58 0.47 0.42 0.82

Russia 0.15 0.36 0.61 0.81 0.85 0.74 0.56 0.33 0.14 0.85

South Korea -0.02 0.16 0.38 0.53 0.56 0.43 0.18 -0.05 -0.23 0.56

Spain 0.83 0.90 0.94 0.93 0.88 0.82 0.75 0.67 0.60 0.94

Thailand 0.26 0.44 0.51 0.63 0.64 0.53 0.47 0.26 0.10 0.64

United Kingdom 0.47 0.60 0.70 0.73 0.66 0.56 0.46 0.34 0.24 0.73

United States 0.39 0.53 0.66 0.70 0.67 0.58 0.44 0.32 0.24 0.70

Argentina 0.15 0.33 0.51 0.61 0.66 0.62 0.51 0.41 0.28 0.66

Brazil -0.15 0.04 0.31 0.47 0.50 0.38 0.26 0.14 0.06 0.50

Chile 0.39 0.59 0.70 0.67 0.63 0.54 0.30 0.06 -0.15 0.70

Colombia 0.44 0.57 0.67 0.75 0.78 0.72 0.56 0.31 0.02 0.78

Mexico -0.04 0.20 0.41 0.56 0.63 0.53 0.40 0.20 -0.02 0.63

Peru 0.28 0.41 0.58 0.76 0.78 0.67 0.55 0.39 0.35 0.78

Uruguay 0.54 0.64 0.75 0.81 0.80 0.73 0.65 0.57 0.51 0.81

Venezuela 0.10 0.26 0.38 0.45 0.51 0.43 0.40 0.41 0.36 0.51

Mean 0.30 0.45 0.60 0.69 0.69 0.60 0.47 0.32 0.20

Median 0.30 0.46 0.63 0.72 0.70 0.62 0.49 0.31 0.20 Diapositiva 17

Cross correlation ECI and HP-filtered GDP

Countries

Lead t=0

Lag Maximum correlation

t-4 t-3 t-2 t-1 t+1 t+2 t+3 t+4

Austria 0.58 0.58 0.52 0.44 0.33 0.16 -0.06 -0.29 -0.46 0.58

Belgium 0.43 0.53 0.54 0.52 0.37 0.18 0.00 -0.19 -0.30 0.54

Canada 0.38 0.39 0.37 0.28 0.13 -0.06 -0.24 -0.36 -0.39 0.39

China -0.13 -0.06 -0.01 -0.01 0.03 -0.06 -0.29 -0.54 -0.58 0.03

Finland 0.43 0.47 0.48 0.43 0.33 0.19 0.03 -0.13 -0.24 0.48

France 0.40 0.47 0.50 0.49 0.39 0.20 -0.01 -0.23 -0.35 0.50

Germany 0.41 0.51 0.58 0.59 0.46 0.24 0.00 -0.25 -0.43 0.59

Greece 0.05 0.09 0.13 0.11 0.07 0.05 0.00 -0.07 -0.11 0.13

India 0.32 0.47 0.57 0.54 0.43 0.24 -0.07 -0.38 -0.54 0.57

Indonesia 0.21 0.23 0.23 0.09 0.02 -0.14 -0.41 -0.48 -0.47 0.23

Ireland 0.51 0.51 0.48 0.38 0.30 0.14 0.00 -0.17 -0.33 0.51

Italy 0.32 0.41 0.41 0.35 0.21 0.04 -0.11 -0.28 -0.39 0.41

Japan 0.52 0.57 0.64 0.63 0.49 0.26 0.04 -0.13 -0.23 0.64

Netherlands 0.55 0.56 0.54 0.46 0.31 0.11 -0.09 -0.28 -0.42 0.56

Philippines 0.23 0.30 0.35 0.29 0.22 0.05 -0.11 -0.26 -0.43 0.35

Portugal 0.17 0.20 0.22 0.19 0.10 -0.04 -0.23 -0.41 -0.47 0.22

Russia 0.42 0.47 0.54 0.57 0.46 0.24 -0.04 -0.29 -0.43 0.57

South Korea 0.26 0.36 0.44 0.47 0.38 0.12 -0.21 -0.45 -0.56 0.47

Spain 0.36 0.35 0.30 0.20 0.06 -0.09 -0.23 -0.33 -0.39 0.36

Thailand 0.30 0.35 0.36 0.29 0.21 -0.02 -0.12 -0.34 -0.44 0.36

United Kingdom 0.45 0.46 0.43 0.35 0.20 0.01 -0.17 -0.31 -0.38 0.46

United States 0.49 0.48 0.43 0.34 0.21 0.03 -0.18 -0.32 -0.40 0.49

Argentina 0.56 0.60 0.59 0.53 0.44 0.25 0.00 -0.19 -0.32 0.60

Brazil 0.05 0.22 0.34 0.47 0.43 0.29 0.02 -0.15 -0.21 0.47

Chile 0.48 0.55 0.46 0.27 0.07 -0.10 -0.32 -0.48 -0.62 0.55

Colombia 0.42 0.39 0.33 0.28 0.15 -0.04 -0.24 -0.44 -0.62 0.42

Mexico 0.35 0.46 0.50 0.51 0.41 0.21 -0.01 -0.19 -0.35 0.51

Peru 0.27 0.28 0.37 0.45 0.39 0.19 -0.01 -0.21 -0.27 0.45

Uruguay 0.38 0.35 0.33 0.29 0.18 0.03 -0.14 -0.29 -0.35 0.38

Venezuela 0.55 0.60 0.59 0.52 0.43 0.26 0.11 -0.02 -0.15 0.60

Mean 0.37 0.42 0.43 0.39 0.28 0.10 -0.10 -0.28 -0.39

Median 0.41 0.46 0.45 0.43 0.32 0.13 -0.08 -0.29 -0.40 Diapositiva 15

Cross correlation ECI and BK-filtered GDP

Countries

Lead

t=0

Lag

Maximum correlation t-4 t-3 t-2 t-1 t+1 t+2 t+3 t+4

Austria 0.57 0.61 0.59 0.51 0.37 0.16 -0.08 -0.31 -0.49 0.61

Belgium 0.43 0.55 0.60 0.54 0.38 0.18 -0.03 -0.18 -0.27 0.60

Canada 0.49 0.55 0.55 0.48 0.33 0.12 -0.09 -0.22 -0.27 0.55

China -0.20 -0.09 0.01 0.06 0.01 -0.13 -0.36 -0.57 -0.66 0.06

Finland 0.53 0.60 0.64 0.62 0.55 0.41 0.26 0.09 -0.07 0.64

France 0.42 0.51 0.55 0.52 0.40 0.21 -0.02 -0.23 -0.38 0.55

Germany 0.40 0.56 0.66 0.64 0.50 0.26 -0.02 -0.29 -0.51 0.66

Greece -0.52 -0.49 -0.45 -0.45 -0.47 -0.50 -0.54 -0.57 -0.60 -0.45

India 0.34 0.62 0.86 0.91 0.80 0.45 0.01 -0.45 -0.80 0.91

Indonesia 0.35 0.43 0.40 0.29 0.10 -0.19 -0.52 -0.76 -0.81 0.43

Ireland 0.72 0.70 0.65 0.56 0.41 0.24 0.04 -0.16 -0.32 0.72

Italy 0.27 0.37 0.42 0.36 0.22 0.04 -0.15 -0.30 -0.40 0.42

Japan 0.56 0.67 0.70 0.64 0.49 0.28 0.03 -0.19 -0.29 0.70

Netherlands 0.56 0.59 0.56 0.47 0.33 0.15 -0.06 -0.26 -0.42 0.59

Philippines 0.34 0.52 0.61 0.57 0.46 0.27 0.01 -0.25 -0.45 0.61

Portugal 0.17 0.30 0.40 0.41 0.31 0.06 -0.18 -0.37 -0.45 0.41

Russia 0.56 0.67 0.79 0.80 0.67 0.42 0.07 -0.26 -0.50 0.80

South Korea 0.26 0.38 0.50 0.53 0.40 0.13 -0.21 -0.49 -0.65 0.53

Spain 0.33 0.31 0.25 0.12 -0.04 -0.20 -0.34 -0.45 -0.53 0.33

Thailand 0.46 0.51 0.52 0.44 0.29 0.06 -0.19 -0.38 -0.52 0.52

United Kingdom 0.48 0.51 0.48 0.39 0.24 0.05 -0.14 -0.30 -0.39 0.51

United States 0.55 0.57 0.53 0.42 0.25 0.03 -0.17 -0.33 -0.43 0.57

Argentina 0.66 0.71 0.70 0.62 0.48 0.29 0.08 -0.12 -0.28 0.71

Brazil -0.05 0.23 0.46 0.58 0.52 0.29 -0.04 -0.35 -0.54 0.58

Chile 0.61 0.64 0.57 0.42 0.21 -0.03 -0.24 -0.46 -0.62 0.64

Colombia 0.66 0.69 0.65 0.51 0.28 0.02 -0.25 -0.54 -0.78 0.69

Mexico 0.37 0.54 0.66 0.66 0.53 0.28 -0.01 -0.29 -0.47 0.66

Peru 0.19 0.27 0.34 0.37 0.30 0.13 -0.09 -0.29 -0.39 0.37

Uruguay 0.27 0.30 0.34 0.37 0.29 0.15 -0.01 -0.16 -0.28 0.37

Venezuela 0.58 0.64 0.63 0.55 0.43 0.28 0.13 -0.01 -0.15 0.64

Mean 0.39 0.48 0.52 0.47 0.34 0.13 -0.10 -0.31 -0.46

Median 0.45 0.55 0.57 0.52 0.38 0.15 -0.07 -0.29 -0.45 Diapositiva 11

Methodology

• Business cycle co-movements are often analyzed in terms of the Pearson correlation coefficient

• More recently, analysis has been extended to the frequency domain, using the concepts of dynamic correlation and cohesion firstly introduced by Croux et al. (2001)

• However, with Croux’ dynamic correlation and cohesion index all information about time dependence of co-movements is lost, while in this paper we are particularly interested in the evolution of synchronization of business cycles over time

• Hence, in this paper our measure of co-movement is based on the evolution over time of bilateral correlations among the countries considered in the analysis, and on the derivation of a cohesion index in the time domain

Methodology

• More precisely, we consider both the mean and standard deviation of bilateral correlation coefficients and we name them as “mean cohesion” and “standard deviation cohesion”

• In fact, an increase in mean correlation can also go together with higher dispersion in the distribution of correlation coefficients, implying lower instead of higher synchronization of business cycles (Gayer and Weiss, 2006)

• The procedure we adopt is the following:

• First, we calculate the n*(n-1)/2 centered bilateral correlation coefficients ri,j among the n countries of a given group over a 3-years (13 quarters) window

• Then we calculate the weighted average and weighted standard deviation of ri,j in each point in time, considering countries GDPs’ in PPP as weights (wi):

• Weighted average (mean cohesion) = 𝑤𝑖𝑤𝑗𝜌𝑖,𝑗𝑖≠𝑗

𝑤𝑖𝑤𝑗𝑖≠𝑗

• Weighted standard deviation (std. dev. cohesion) = 𝑤𝑖𝑤𝑗(𝜌𝑖,𝑗−𝑚𝑒𝑎𝑛 𝑐𝑜ℎ𝑒𝑠𝑖𝑜𝑛)2𝑖≠𝑗

𝑤𝑖𝑤𝑗𝑖≠𝑗

• Finally, we evaluate the evolution of mean and standard deviation cohesion over time

Research questions answered - intra group synchronisation

Q: Has the degree of synchronization among Latin American economies increased or decreased in the period 1989-2013?

A: We can identify four periods after a first “assessment” of survey data (1990-1995)

• 1996-2004: substantial stability of mean synchronisation around 0,40 – very similar to the one existing among emerging Asian economies, but much lower than in the Eurozone. In that period, std. dev decreases as well

• 2004-2006: mean cohesion decreases and std. dev increases

• 2007-2010 (max globalisation period): mean cohesion peaks at 0.8, std dev decreases at 0.1

• 2010-2012 (financial crisis): cohesion returns on the pre-globalisation, 1990’s, levels

Intra-group mean Cohesion: Latin American and Asian Economies

Intra-group standard deviation: Latin American and Asian

Economies

Research questions answered - intra group synchronisation

Q: The pattern observed for LA countries is similar to what happened to more industrialized countries (Eurozone) or other developing economies (the Asian group)?

A: The pattern observed for LA countries is very similar to that observed for the fast developing Asian countries; however, the Eurozone remain much more integrated, with higher mean cohesion, and lower standard deviation of bilateral correlation coefficients, even if the financial crisis has implied also in this case a fall in cohesion

Business cycle cohesion in the Eurozone

Research questions answered – inter group synchronisation

Q: Are Latin American economies more or less synchronized now than in the past with respect to older industrialized countries?

Has overall business cycle synchronization increased or decreased in the last twenty years?

A: At the global level:

• Mean cohesion has followed a path quite similar to that observed for LA and Asian economies

• Stad. Dev cohesion has been much more stable, with a fall in he period of max globalisation followed by a return on the average levels of the past

Global synchronisation

Conclusions and further research

• Survey data from the IFO WES may be very useful in the business cycle synchronisation analysis among industrial and developing countries

• According to a measure of coherence based on bilateral correlations coefficients in recent years a big increase in cohesion in the period of max globalisation has been followed by a return on previous levels in the aftermath

of the crisis

• This has been true both looking at intra-group and inter-group coherence

Conclusions and further research

• In the future, analysis can be extended:

• To other countries for which the survey is performed (Non EU countries in Europe, other Asian, African countries)

• To the frequency domain, concentrating on the business cycle frequency band and evalutating measures of dynamic correlation a’ la’ Croux (2011)

• Trying to look simoultaneously at the evolution of cohesion over time and across frequencies using wavelet analysis (Rua-Lopez, 2012)