Business Cycles in South America and East Asia : An Introduction

We study the business cycle component of the per capita gross domestic product of ten South American and six East Asian countries during the period 1970-2000. Both sets of countries display an average volatility roughly equal to twice the U.S. one. Average persistence in each of these areas is similar to the U.S. The cyclical phases usually do not coincide temporally across countries. Key-words: business cycles, South America, East Asia Received in 11/09/2005; revised in 20/09/2005; accept in 11/11/2005. * Jorge Reis Sandes Mestrado em Economia IBMEC, Adress: Avenida chile 100 5o andar, centro 20031170 Rio de Janeiro – RJ – Brazil Email: sandes@vetor.com.br Telephone: (21) 37479731 ** Luis Gustavo Cubas Vivanco IBMEC/RJ lvivanco@terra.com.br *** Alexandre Barros Cunha: Doutor em Economia Universidade Federal do Rio de Janeiro Adress: Av. Pasteur 250, sala 119, Urca, 22290240 Rio de Janeiro – RJ – Brazil Email: abcunha13@globo.com Telephone: (21) 39385275


I. MEANINGS OF THEORIES
he essay Time to build and aggregate fluctuations by Kydland and Prescott (1982) had a profound impact on macroeconomic theory.The relevance of their groundbreaking work was recognized when they won the 2004 Nobel Prize for economics.Kydland and Prescott (1982) implemented a methodology that subsequently became standard.They constructed a dynamic and stochastic general equilibrium model.Then they used numerical methods to compute the solution to this model, that is, mathematical relations that described the temporal path of the endogenous variables as functions of the parameters, the initial conditions and the realizations of the random variables.Finally, they compared the statistical properties of the series generated by the model with equivalent series in the real world.
The methodology employed by Kydland and Prescott has become quite popular among researchers in the area of macroeconomics.It has been used to study a wide range of questions.For example, Mendonza (1991) and Backus, Kehoe and Kydland (1992) used it to study economic cycles in models of open economies.Chari, Kehoe and McGrattan (2000) utilized similar procedures to verify whether monetary shocks in a general equilibrium model with rigid prices are able to generate economic cycles with persistence similar to those observed in the United States.
For the method Kydland and Prescott to be implemented, there evidently must be information on the statistical properties of the macroeconomic series.For this reason, Kydland and Prescott (1990, p. 3) affirmed that "reporting facts -without assuming that the data are generated by some type of probabilistic model -is an important scientific activity." Various texts have been dedicated to the statistical properties of macroeconomic variables.For example, Backus and Kehoe (1992) documented some characteristics of series, such as gross national product (GNP), aggregate investment and price level for countries of the Organization for Economic Cooperation and Development (OECD) in the period 1850-1986. Baxter and Stockman (1989), using a sample of forty-nine countries, studied whether the volatility and correlation of some macroeconomic variables are dependent on the exchange rate regime.Ambler, Cardia and Zimmermann (2004) used econometric techniques to estimate the correlations between diverse macroeconomic variables of twenty industrialized countries.
This essay fits in the line of research described in the paragraph above.It has the modest objective of documenting some properties of the business cycles in ten South American countries (Argentina, Bolivia, Brazil, Chile, Colombia, Ecuador, Paraguay, Peru, Uruguay and Venezuela), six East Asian countries (Singapore, South Korea, Hong Kong, Indonesia, Malaysia and Thailand), and for comparison purposes, the United States, in the period 1970-2000.
We reached several conclusions in this work.We found that the cyclical fluctuations in East Asia and South America are more volatile than those observed in the United States.More specifically, the standard deviation of the cyclical component of per capita GDP (Gross Domestic Product)1 of developing countries is, on average, twice that in the United States.
On average, the cyclical swings in South America and East Asia are slightly more persistent than in the U.S. In some countries, however (such as Argentina and Hong Kong), the cyclical variations are less persistent than in the U.S.
Except for Brazil and Hong-Kong, the business cycles in South American and East Asian countries have a low correlation with those in the United States.The respective mean correlation coefficients for those two regions were 0.04 and -0.07.These statistics are quite distant from the equivalent ones found by Backus and Kehoe (1992) for nine developed countries.
We also carried out an exercise to date the cyclical business phases.In other words, we checked year-by-year whether each of the countries studied was in recession or expansion.We concluded that there is no propensity for the countries to show the same timing in their business cycles.This finding is valid even when the exercise is restricted to countries of a single continent.
The rest of this work is organized as follows.In Section 2 we present a simple procedure to implement the popular Hodrick-Prescott (HP) filter.In Section 3 we analyze the properties of economic cycles of the countries in the sample.We present our conclusions in Section 4. After the references, the Appendix contains graphs that depict the behavior of per capita GDP in the nations included in the sample.

II. THE HODRICK-PRESCOTT FILTER
A temporal series yt can be decomposed into trend components T and cycle components T C y C .Obviously, the original series and its components must satisfy the equality There are various ways to perform the decomposition mentioned above.One of the most popular is to use the Hodrick-Prescott (HP) filter, which we did in this work.
The HP filter was first presented in a working paper by Hodrick and Prescott (1980).It took 17 years for this work to be published in an updated version -Hodrick and Prescott (1997).
Despite the delay in official publication of the original work, the HP filter became extremely popular among researchers studying economic cycles.Kydland and Prescott (1982), Backus and Kehoe (1992) and Ellery, Gomes and Sachsida (2002) are typical examples of works that have used this filter.
We present here a procedure to compute the HP filter.Let yt be the natural logarithm of the real per capita GDP of some country.Given N observations of this variable, its trend T y t is then chosen so as to minimize the function It can be seen that that because it is a sum of squares, F must be non-negative.
The function F has two components.The first is given by the deviation of the observed product yt from the trend y T .The second corresponds to the variation in growth of the trend the number of its citizens working abroad versus the number of foreign workers within its borders.Despite these differences, the two measures are sufficiently close that the comparisons made here between our results and those of authors using GNP are still valid.T .For the first summation to be nil, the trend must be equal to the effective product.For the second summation to be nil, the first difference of the trend must be constant.Hence, if the first difference of the per capita product is not constant, it will not be possible for both summations to be nil.The arguments developed in the preceding paragraphs illustrate the underlying procedure of the HP filter.The trend T must be chosen so as to obtain a compromise between two targets: (1) a small difference between the observed product and the trend; and (2) smooth growth of the trend.The smoothing parameter l defines the relative weight attributed to the two targets.The greater the value of l is, the smoother the growth of the trend will be.
Following the procedure employed by Backus and Kehoe (1992) and Ellery, Gomes and Sachsida (2002), we adopted in this essay a value of 100 for the parameter l.Although there is no consensus among authors, this is the value usually utilized for data with annual frequency.A detailed discussion of the choice of the value of l can be found in Ravn and Uhlig (1997).

III. ECONOMIC CYCLES IN SOUTH AMERICA AND EAST ASIA
In this work we used the annual per capita GDP series for the United States, ten South American countries (Argentina, Bolivia, Brazil, Chile, Colombia, Ecuador, Paraguay, Peru, Uruguay and Venezuela) and six from East Asia (Singapore, South Korea, Hong Kong, Indonesia, Malaysia and Thailand).The sample covers the period 1970-2000.
We obtained data on the GDP of each country from the CD International Financial Statistics from the IMF.Information on the population of the various countries was available at the Internet page of the U.S. Census Bureau.Based on these data, we constructed the per capita GDP series for each country.
The principal factor determining the choice of the data's frequency, the period and the countries analyzed was availability.The use of monthly or quarterly data and/or the extension of the sample study period would have meant excluding some countries.Likewise, the inclusion of more countries would have meant the reduction of the period studied.We sought a compromise between detail and scope of the sample, choosing the period and group of countries mentioned at the start of this section.
We present in the Appendix two graphs for each of the countries studied.The first graph of each pair plots the trajectory of the logarithm of the real per capita GDP yt and its trend y T .The second depicts the evolution of the cyclical component y C .
The United States, Colombia, Uruguay and the East Asian countries presented nearly linear trends, unlike the other South American countries.The trend curves for Argentina, Bolivia and Peru exhibited sinusoidal behavior.Brazil, Ecuador and Paraguay had growth trends until 1980 and stagnation thereafter.The Chilean economy was stagnant until 1980 and ascendant from 1981 onward.Finally, Venezuela had a rising trend until 1977, descended from 1978 until 1987 and was flat from 1988 onward. 2  A quick analysis of the graphs of the cyclical components of per capita GDP suggests that the statistical properties of the series C are not homogenous among the countries studied.For example, the cycles are apparently more persistent in Bolivia than in Argentina.Such heterogeneity is documented in the discussion that follows.
Among the many attributes of business cycles, three are usually studied in more detail.These are volatility, persistence and degree of comovement of the variables.The volatility of an economy's cycles is normally measured by the standard deviation of y C .The persistence is measured by the correlation coefficient of C with C t -1 .Finally, the degree of comovement of the cyclical component of per capita GDP with some variable of interest is given by the correlation coefficient of C with this other series.
Table 1 contains the statistics on volatility and persistence of the business cycles for each country studied, along with the degree of comovement of these national economies with the United States'.The lines referring to South America and East Asia present the simple arithmetic means of the respective regions. 2 The long-term growth differential among countries of South America and East Asia has been the subject of various studies.We suggest that readers wanting a better understanding of this question consult De Gregorio andLee (2003) andFerreira, Pessoa andVeloso (2004) The average volatility of South America (4.36%) is slightly higher than in East Asia (3.94%).Both means are markedly greater than the volatility in the United States (2.13%).
Among all the developing countries included in the sample, the smallest volatility observed is in Colombia (2.45%).This figure exceeds the volatility for the United States, so each South American and East Asian country individually has less volatile cycles than in the United States.
The greater volatility of business cycles in developing countries becomes even more evident when the results obtained in this study are compared with those of Backus and Kehoe (1992).These authors documented the statistical properties of business cycles in ten developed countries (West Germany, Australia, Canada, Denmark, United States, Italy, Japan, Norway, United Kingdom and Sweden).Particularly, for periods in the neighborhood of 1950-1985, 3 the volatility of these countries' GNP was equal to 2.06%, with the highest figure observed being 3.11% for Japan.Except for Columbia and the United States, all the countries studied here had volatility above 3.11%.
A marked characteristic of Table 2 is the lack of homogeneity of the cyclical phases among the countries.Only in one of the years studied (1982) were all the countries in the same phase (recession).
It is possible that this anomalous outcome of uniform recession in 1982 is related to three shocks occurring in that period.In January 1979 the Fundamentalist Revolution overthrew the Shah of Iran and in September that year the war between Iran and Iraq broke y BBR, Braz.Bus.Rev. (Engl.ed., Online), Vitória, v. 2, n. 2, Art.6, p. 172 -186, jul.-dez.2005 www.bbronline.com.bry

Table 2 -
Phases of economic cycles