Testing Business Cycles Asymmetry in Central and Eastern European Countries

doi:10.4236/me.2012.36091

Paper Menu >>

Journal Menu >>

Modern Economy, 2012, 3, 713-717

http://dx.doi.org/10.4236/me.2012.36091 Published Online October 2012 (http://www.SciRP.org/journal/me)

Testing Business Cycles Asymmetry in Central

and Eastern European Countries

Viorica Chirila, Ciprian Chirila

Faculty of Economics and Business Administration, University Alexandru Ioan Cuza, Iaşi, Romania

Email: [email protected], [email protected]

Received July 11, 2012; revised August 25, 2012; accepted September 18, 2012

ABSTRACT

The idea of business cycles asymmetry is not new in economic theory. According to business cycles asymmetry, a

country’s economy behaves differently during economic growth periods as compared to economic recession periods.

The results achieved by business cycles asymmetry testing are far from unanimous: some are positive, others are nega-

tive. Business cycles asymmetry has major econometric implications: business cycles cannot be modeled using linear

models. This paper aims to test business cycles asymmetry in Central and Eastern European Countries, where few

business cycles analyses, and especially business cycles asymmetry researches, have been conducted. The industrial

production index was considered when testing business cycles asymmetry. We estimated business cycles using the

Hodrick-Prescott filter and Mills’ test of asymmetry. Mira’s test was also employed to test results reliability. According

to our results, business cycles in Central and Eastern European countries are not asymmetric.

Keywords: Business Cycles; Asymmetry; Hodrick-Prescott Filter

1. Introduction

Business cycles research enjoys a cyclical evolution itself.

Papers on this topic are published mostly at times of

economic recession. The explanation seems simple and it

relies on the researchers’ intent to forecast future eco-

nomic recessions likely to have a major negative impact

on economy, which in its turn influences the population’s

standard of living. The empirical characteristics of busi-

ness cycles are vital in business cycles modeling and

forecast. Therefore, special attention is paid to empirical

characteristics. The concept of asymmetry as a business

cycle characteristic is not new, yet converging results

have not been achieved so far. Authors such as Mitchell

[1], Keynes [2] Burns and Mitchell [3], and Hicks [4]

have mentioned business cycles asymmetry in their eco-

nomic theory papers.

Business cycles asymmetry roughly means that econ-

omy behaves differently during economic growth periods

than during economic recession periods. Boldwin [5]

argues that asymmetric business cycles only occur in the

case where recessions and expansions are not mirror im-

ages of each other. Business cycles asymmetry refers, on

the one hand, to the fact that decrease due to economic

recession is more abrupt than increase during economic

growth periods and, on the other hand, to the fact that the

minimum value reached at times of economic recession

is greater in absolute value than the peak reached during

economic growth. Siechel [6] defines the first type as

steepness asymmetry and the second as deepness asym-

metry and reckons that they may occur either simultane-

ously or separately. These types of asymmetry are also

called transversal and longitudinal asymmetry [7] or un-

conditional and conditional asymmetry [8].

Business cycles asymmetry has serious implications

on their econometric modeling: Business cycles cannot

be described by linear models. Deepness asymmetry en-

ables specialists to capture business cycles using a model

involving asymmetric price adjustments (as positive de-

mand shocks have greater relative negative impact on

output than positive shocks, which have less impact on

output). Steepness asymmetry allows capturing business

cycles by an asymmetric costs model, which relies on the

assumption that production may decrease very rapidly,

yet its increase is much slower.

The outcome of business cycles asymmetry testing is

different. Nefci [9], Falk [10] and Mills [11] did not get

positive results when testing business cycles asymmetry

in industrial production. On the other hand, there are

studies supporting the existence of asymmetry in a num-

ber of economic series by Ramsey and Rothman [7], An-

dreano and Savio [12] and Stanca [13].

Speight tests business cycles asymmetry [14] on a

sample of 16 OECD countries considering their volume

of industrial production, as he thinks that this variable

“displays as much cyclical variation as possible”, based

V. CHIRILA, C. CHIRILA

714

on available data. The analyzed period is 1961:1-1994:4

for most countries, except for Spain, Greece and OECD

aggregates. It uses Sichel’s methodology (1993), as well

as Newey’s and West’s corrections [15] with two parzen

windows: T/4 and T/3. Although negative asymmetry is

present in very many countries, deepness asymmetry is

significant, considering an up to 10% risk, only for Ger-

many, Japan, Sweden and UK. He also achieves signifi-

cant steepness asymmetry for Japan, Sweden and UK,

taking an up to 10% risk.

Business cycles asymmetry has not been tested for

Central and Eastern European countries. Therefore, our

study is designed to fill this gap. The following countries

were included in our business cycles asymmetry analysis:

Bulgaria, Croatia, Czech Republic, Estonia, Latvia, Li-

thuania, Hungary, Poland, Romania, Slovenia and Slova-

kia. The analyzed period was 1998.1-2011.3. We used

the Hodrick-Prescott filter [16] to estimate business cy-

cles, and the Mills test [17] and Mira test [18] to test

asymmetry. The two estimation and testing variants al-

lowed us to check the reliability of the reported results.

The rest of the paper is structured as follows: the sec-

ond section includes a synthetic presentation of, on the

one hand, the methods employed to estimate business

cycles and, on the other hand, the methods devoted to

business cycles asymmetry testing; the third section de-

scribes the data used, whereas the fourth section reveals

the reported results. This paper ends with a set of conclu-

sions.

2. Methodology

When testing business cycles asymmetry, the cyclical

component should be estimated first and the asymmetry

tests should be conducted afterwards. Business cycles

estimation relies on the general assumption that an un-

seasonable variable may be decomposed in three com-

ponents, namely the trend, cycle and random components.

There are several methods applied to exclude the trend

component. Nevertheless, none of them has been de-

clared as the best variable trend exclusion method so far.

Canova [19] provides a well-structured detailed presenta-

tion of these methods. It is important to say that the pre-

vious studies proved that the trend determination method

may influence the results.

Some of these trend exclusion methods consider the

assumption according to which the variable only includes

the trend and cycle components:

ttt

yxc

where: t

is the non-stationary trend component and

t is the cyclical stationary component, which is trend-

dependent.

 

-1-1 -2

ttttt

cgggg





In our paper, we decided to determine the cyclical

component using the Hodrick-Prescott filter.

Even if the Hodrick-Prescott filter was very much

criticized by Rand and Tarp [20], it is also the most used

in business cycles analysis. Therefore, we will also use it

in our study. By means of the Hodrick-Prescott filter, the

trend is determined by minimizing the expression:





 





 (1)

where:









ln ln

cytyt



ln 1

gyt





1ln

, , ,







2ln 1

gyt

*

and



—the long-term trend of the

variable y.

The most frequent value used for the parameter



the case of quarterly data is 1600.

To test the existence of the cyclical component for a

time series we use the Ljung-Box test. The tested hy-

potheses are the following: the null hypothesis H0 pre-

supposes that the variable is a white noise and the alter-

native hypothesis H1 presupposes that the variable is

autocorrelated. The test statistics is calculated according

to the relation:



QTT Ti





 

 (2)

For asymmetry testing purposes, we will consider the

test proposed by Mills [17], which modifies the test pro-

posed by Sichel [6] by a Newey-West adjustment of

variance for lack of normality.

The test relies on the following asymmetry coefficient

3/2



(3) 



where



is the moment j of the cyclical component of

the series. If the sample is large and the component is

normal and independently distributed, the variance of the

estimated asymmetry coefficient would be equal to 6/T.

Yet, since these assumptions are not observed, we calcu-

late variance S as follows:



653

169 935







 





(4)

where:







is the kurtosis coefficient (Jaba, 2001),

T is the sample volume.

Mills [17] adjusts the variance as follows

SS jj













 (5)



where:

is the autocorrelation coefficient j of the variable



V. CHIRILA, C. CHIRILA 715

3/2



, 11

 calculates the weights and

2/9

4100









The statistical test employed is asymptotically stan-

dard normal



 (6)

and it determines whether asymmetry is significantly ne-

gative. According to the null hypothesis, business cy-

cles have no deepness asymmetry, whereas according to

the alternative hypothesis, business cycles do have deep-

ness asymmetry.

In order to test results reliability, we applied Mira’s

alternative test [18], which is calculated as follows:

z (7)





where:

tmed

cc med t

44DDE

, is the median c















 

4/5

1/2 2TT

 



 

 











4/5

1/5

1/2 TT

DT c

















ttmed

EccIc c

T

 



cc

Just like Mills’ test [17], Mira’s test [18] tests whether

asymmetry is significantly negative. According to the

null hypothesis, business cycles have no deepness asym-

metry, whereas according to the alternative hypothesis,

business cycles do have deepness asymmetry. In order to

test steepness asymmetry existence, the first variable

difference t is considered instead of the variable t.

The first difference enables us to test whether the deep

series decreases are more considerable and less common

than series increases.

3. Empirical Results

We considered the actual industrial production index to

estimate business cycles in Central and Eastern European

countries. Industrial production is a poly-cyclical vari-

able and it is one of the most commonly used variables,

although there is no evidence supporting its asymmetric

character. Asymmetry testing is more recommended in

industrial production rather than in GDP, since the latter

is a more comprehensive variable, which may have

counter-cyclical components.

The quarterly data were taken from the Eurostat data-

base and the time period considered was dependent on

the availability of the data in this database. We found

data for the analyzed countries (Bulgaria, Croatia, Czech

Republic, Estonia, Latvia, Lithuania, Hungary, Poland,

Romania, Slovenia, Slovakia), whose common registra-

tion period was 1998.1-2011.3. We employed X 12

ARIMA for data deseasoning.

To estimate economic cycles, we used the Hodrick-

Prescott filter described in the paragraph above.

Tables 1 and 2 show the results of the statistical tests

for deepness and steepness asymmetry. The asymmetry

indicator is negative for Bulgaria, Czech Republic, Esto-

nia, Latvia, Lithuania, Hungary, Slovenia and Slovakia

and positive for Croatia, Poland and Romania. The busi-

ness cycles of these countries would be characterized by

deepness asymmetry if the resulting asymmetry indica-

tors were significantly negative. The results of Mills’ [17]

and Mira’s [18] asymmetry tests do not support this.

Therefore, the business cycles of Central and Eastern

European countries are not characterized by deepness

asymmetry.

In order to test steepness asymmetry, we first calcu-

lated the asymmetry indicator, yet, this time, for the first

business cycles values difference for each analyzed

country. In this case, almost all the countries in the sam-

ple have negative asymmetry indicator, with the excep-

tion of Croatia and Romania. Mills’ [17] and Mira’s [18]

tests show no evidence of any significant asymmetry,

hence they do not support the presence of steepness

asymmetry.

Table 1. Tests results for deepness asymmetry.

S(c)

Bulgaria –0.310 –1.373 –0.182

Czech Republic –0.063 –0.198 0.040

Croatia 0.294 1.579 0.089

Estonia –0.747 –3.611 –0.039

Latvia –0.356 –1.492 0.062

Lithuania –0.447 –1.517 –0.084

Hungary –0.271 –1.291 –0.034

Poland 0.014 0.069 –0.024

Romania 1.182 1.331 0.060

Slovenia –0.056 –0.183 0.035

Slovakia –0.068 –0.176 0.121

Remark: The business cycles were estimated using the Hodrick-Prescott

filter of the Eviews 7 software.

V. CHIRILA, C. CHIRILA

716

Table 2. Tests results for steepness asymmetry.

S() c

Bulgaria –0.115 –0.429 –0.111

Czech Republic –0.514 –1.236 0.011

Croatia 0.450 1.443 0.046

Estonia –0.724 –1.116 –0.035

Latvia –0.726 –1.735 –0.114

Lithuania –0.413 –2.143 –0.106

Hungary –0.251 –0.658 –0.115

Poland –0.592 –2.414 –0.019

Romania 0.385 0.264 –0.097

Slovenia –0.611 –1.633 –0.074

Slovakia –0.806 –2.434 –0.034

Remark: The business cycles were estimated using the Hodrick-Prescott

filter of the Eviews 7 software.

4. Conclusion

The concept of business cycles asymmetry is not new in

economic theory. The results of business cycles asym-

metry testing are both positive and negative. The asym-

metric nature of business cycles has been tested espe-

cially in developed countries, where large series of data

on macroeconomic indicators are available. The Central

and Eastern European countries started to embrace mar-

ket economy in 1989. Consequently, the macroeconomic

indicators series recorded in accordance with the re-

quirements of the European Union are much smaller.

This accounts for the relatively small number of papers

devoted to business cycles in Central and Eastern Euro-

pean countries. We preferred the industrial production

index to test steepness asymmetry and deepness asym-

metry. According to the results of Mills’ test [17], ap-

plied to determine the two types of asymmetry, they are

absent in the Central and Eastern European countries.

This lack of asymmetry is also supported by the results

of Mira’s test [18]. Business cycles asymmetry in the

Central and Eastern European countries should be reana-

lyzed in the future, when a larger data sample and further

macroeconomic indicators are available.

5. Acknowledgements

This work was cofinanced from the European Social

Fund through the Sectorial Operational Programme Hu-

man Resources Development 2007-2013, project number

POSDRU/1.5/S/59184 “Performance and excellence in

postdoctoral research in Romanian economic science do-

main”.

REFERENCES

[1] W. C. Mitchell, “Business Cycles: The Problem and Its

Setting,” NBER, New York, 1927.

[2] J. M. Keynes, “The General Theory of Employment, In-

terest and Money,” Macmillan, London, 1973.

[3] A. F. Burns and W. C. Mitchell, “Measuring Business

Cycles,” National Bureau of Economic Research, New

York, 1946.

[4] J. Hicks, “A Contribution to the Theory of the Trade Cy-

cle,” Clarendon, Oxford, 1950.

[5] M. D. Boldin, “Should Policy Makers Worry about Asym-

metries in the Business Cycle?” Studies in Nonlinear Dy-

namics and Econometrics, Vol. 3, No. 4, 1999, pp. 203-

220. doi:10.2202/1558-3708.1049

[6] D. E. Sichel, “Business Cycle Asymmetry: A Deeper

Look,” Economic Inquiry, Vol. 31, No. 2, 1993, pp. 224-

236. doi:10.1111/j.1465-7295.1993.tb00879.x

[7] J. B. Ramsey and P. Rothman, “Time Irreversibility and

Business Cycle Asymmetry,” 1993.

http://econ.as.nyu.edu/docs/IO/9389/RR93-39.pdf

[8] A. Peiró, “Are Business Cycles Asymmetric? Some Euro-

pean Evidence,” Applied Economics, Vol. 36, No. 4, 2004,

pp. 335-342. doi:10.1080/00036840410001674268

[9] S. N. Nefci, “Are Economic Time Series Asymmetric over

the Business Cycle?” The Journal of Political Economy,

Vol. 92, No. 2, 1984, pp. 307-328. doi:10.1086/261226

[10] B. Falk, “Further Evidence on the Asymmetric Behavior

of Economic Time Series over the Business Cycle,” Jour-

nal of Political Economy, Vol. 94, No. 5, 1986, pp. 1096-

1109. doi:10.1086/261423

[11] T. C. Mills, “Business Cycle Asymmetries and Non-Li-

nearities in UK Macroeconomic Time Series,” Ricerche

Economiche, Vol. 49, No. 2, 1995, pp. 97-124.

doi:10.1016/0035-5054(95)90018-7

[12] M. S. Andreano and G. Savio, “Further Evidence on Busi-

ness Cycle Asymmetries in G7 Countries,” Applied Eco-

nomics, Vol. 34, No. 7, 2002, pp. 895-904.

doi:10.1080/00036840110060984

[13] L. Stanca, “Asymmetries and Non-Linearities in Italian

Macroeconomic Fluctuations,” Applied Economics, Vol.

31, No. 4, 1999, pp. 483-491.

doi:10.1080/000368499324192

[14] A. E. H. Speight, “Testing for Business Cycle Asymme-

try: Some International Evidence,” Applied Economics

Letters, Vol. 4, No. 10, 1997, pp. 603-606.

doi:10.1080/758533282

[15] W. Newey and K. West, “A Simple Positive Semi-Defi-

nite Heteroskedasticity and Autocorrelation Consistent

Covariance Matrix,” Econometrica, Vol. 55, No. 3, 1987,

pp. 703-708. doi:10.2307/1913610

[16] R. Hodrick and E. C. Prescott, “Postwar Business Cycles:

An Empirical Investigation,” Journal of Money, Credit

and Banking, Vol. 29, No. 1, 1997, pp. 1-16.

doi:10.2307/2953682

[17] T. C. Mills, “Business Cycle Asymmetry and Duration

Dependence: An International Perspective,” Journal of

V. CHIRILA, C. CHIRILA

717

Applied Statistics, Vol. 28, No. 6, 2001, pp. 713-724.

doi:10.1080/02664760120059246

[18] A. Mira, “Distribution-Free Test for Symmetry Based on

Bonferroni’s Measure,” Journal of Applied Statistics, Vol.

26, No. 8, 1999, pp. 959-972.

doi:10.1080/02664769921963

[19] F. Canova, “Detrending and Business Cycle Facts,” Jour-

nal of Monetary Economics,” Elsevier, Vol. 41, No. 3,

1998, pp. 475-512.

[20] J. Rand and F. Tarp, “Business Cycles in Developing

Countries: Are They Different?” World Development, Vol.

30, No. 12, 2002, pp. 2071-2088.

doi:10.1016/S0305-750X(02)00124-9