Frequency domain analysis of consumer confidence, industrial production and retail sales for selected European countries/Vartotoju pasitikejimo, pramonines gamybos ir mazmenines prekybos dazniu analize pasirinktose europos salyse.

Mermod, Asli Yuksel ; Dudzeviciute, Gitana

1. Introduction

Consumer confidence / sentiment is commonly described as a leading economic indicator. In its simplest sense, such an indicator is defined as any economic statistic, which possesses information on the current and future path of an economy, Tvaronaviciene et al. (2009). According to the surveys of Tvaronaviciene and Grybaite (2007), Tvaronavicius and Tvaronaviciene (2008), such statistics receive widespread attention from experts, investors and business and financial press as economic agents may amend consumption / investment strategies depending on the pattern of leading indicators. Therefore, public and /or private institutions in many developed /emerging countries have constructed consumer confidence indices (CCI) to measure and disseminate the latest stance of consumer attitudes (1).

The analysis of consumer confidence advocates the positive relationship between consumer optimism and the future path of consumption expenditures. Among others, Carroll et al. (1994), Bram and Ludvigson (1998), Hiifner and Schroder (2002) and Kwan and Cotsomitis (2006) provide support for the link between changes in consumer attitudes and personal consumption expenditures. However, the bulk of the literature mainly focuses on developed countries and the expectations-consumption channel where consumer confidence is modeled as strictly exogenous (2). Recently, Gomes (2007) emphasizes the inherent characteristics of endogenous growth models that rely on the optimization problem of a consumption utility maximizing representative agent. In such a theoretical setting, economic agents are expected to increase (decrease) their propensity to consume in expansionary (recessionary) periods. Hence, an increase in consumer confidence should lead to an increase in total retail sales and economic growth given that the survey responses are unbiased and there is no attrition problem.

The originality of this study is twofold: First, the research object of our study is to investigate the direct link between consumer confidence, economic growth and retail sales for the case of several countries, including both developing and developed countries. Second, as research methodology, we use spectrum analysis tools such as causality in frequency domain and spectral variance decompositions. We employ the consumer confidence indexes, industrial production (as a proxy for economic growth) and retail sales (as a proxy for the consumer expenditures) to provide insights into the transmission mechanism of changes in the consumer confidence, the response of domestic production and retail sales in selected countries. Our analysis would also shed light to the differences in this transmission mechanism between developed and developing countries.

The second section of this paper includes a brief literature survey on consumer confidence. Section three explains the data of our analysis. In section four, we introduce the methodology of empirical analysis, followed by section five, where we will present and explain the empirical findings. Section six will conclude with some remarks for further research.

2. Literature survey

There are two distinctive categories of literature on consumer confidence. The first could be termed as conventional with its focus on the predictive ability of consumer confidence while searching an answer to the well-known question: "Does consumer sentiment accurately forecast household spending?" Among others, Acemoglu and Scott (1994); Carroll et al. (1994); Fan and Wong (1998); Kwan and Cotsomitis (2004) constitute some of this orthodox approach. The second category includes studies that employ anything outside the orthodox realm (Among others, see Flavin 1991; Alessie, Lusardi 1997; Batchelor, Dua 1998; Souleles 2004).

The orthodox approach argues that improvements in consumer sentiment stimulate consumption growth in the short run. Therefore, the starting point for these studies is to obtain the goodness-of-fit values from regressions of the growth of various measures of household spending on lagged values of consumer confidence using the following equation:

[DELTA]log([C.sub.t]) = [[alpha].sub.0] + [n.summation over (i=1)] [[beta].sub.i][S.sub.t-i] + [[epsilon].sub.t], (1)

where [C.sub.t] denotes consumption at time t, and [S.sub.t] shows the CCI at time [t.sup.3]. Next they test the predictive ability of the sentiment while adding a vector of so-called control variables to the right-hand side (4). Hence, the model becomes:

[DELTA] log ([C.sub.t]) = [[alpha].sub.0] + [n.summation over (i=1)] [[beta].sub.i][S.sub.t-i] + [gamma][Z.sub.t-1] + [[epsilon].sub.t], (2)

where [Z.sub.t-1] denotes a vector of other variables at time (t-1). This approach builds on the canonical permanent income (or life-cycle) hypothesis which postulates that consumers' decisions depend on their expectations of their future incomes. Thus, if consumer confidence is high, then consumer expenditures should be high simultaneously and in the near future.

On the other hand, an unconventional study by Batchelor and Dua (1998) tests the rationality of the economic forecasters' predictions through the proposed stable relationship between the Blue Chip economic indicators and the CCI. They show that consumer confidence is successful in predicting the 1991 recession but would not have performed as well in other times. Moreover, Souleles (2004) employs household-level data that from the Michigan Survey of Consumer Attitudes and Behavior. His results show that households' expectations are biased as forecast errors by individuals do not average out even over a sample period of 20 years.

There is no consensus on the usefulness of consumer confidence as a leading economic variable, either. Garner (1991); Roberts and Simon (2001) and Desroches and Gosselin (2002) conclude that the link between aggregate consumer expectation index and changes in future consumer sales activity is rather weak. Others like Throop (1992); Huth et al. (1994); Otoo (1999); Nahuis (2000); Eppright et al. (2003) and Jansen and Nahuis (2003) support consumer confidence in predicting changes in total consumer expenditures and demonstrate the link between confidence and financial market variables. Recently, there have been some skeptical studies like Dominitz and Manski (2004) which question the methods used in the preparation of consumer confidence indices and Van Oest and Franses (2008) which cautions on the interpretation of movements in consumer confidence.

The previous literature on the relationship between consumer confidence, domestic demand, and different variables of interest has not been conclusive. Consumer confidence can be considered as a quick and relatively inexpensive measure that operates as a proxy for consumer spending. Tvaronaviciene and Kalasinskaite (2010) in their surveys agree, that in emerging markets there is hardly any data for personal consumer expenditures except GDP whereas economic growth is well measured by industrial production,. We believe that households incorporate the signals from the production figures (which are released earlier than personal consumption expenditures) into their decision making process. Hence, we propose that the link between consumer confidence and economic growth should provide valuable information for policy makers, market participants and households.

Theoretically, we follow Matsusaka and Sbordone (1995) which finds a significant relationship between the Michigan Index of Consumer Sentiment and GDP growth. They conclude that consumer confidence indices are able to forecast the evolution of economic activity when their coincident nature is taken into account and that a number of data-coherent parameter restrictions are imposed. Methodologically, we enhance Gelper et al. (2007), the first study in the consumer confidence literature to decompose Granger causality in the time domain, by performing a spectral density analysis in the frequency domain.

3. Data

Our data includes monthly industrial production index (IP), CCI and retail sales (RS) of various countries in order to test the relationship between the growth, consumer confidence and consumer expenditures. All series are obtained from countries' national statistical institutes, and seasonally adjusted IP and RS series are gathered. For all series, both log transformed and year-on-year changes are considered. The variables and descriptions are given in Table 1.

Table 1 shows all the variables and its descriptions that are used for our research.

Depending on the availability of data the time period ranges from 1980 to 2010. Selected countries and corresponding time periods are summarized in Table 2. The data is obtained from OECD Statistics Database (5)

Table 2 shows selected European countries that we analyze in our research. As it's seen from the data, some statistics start from year 1980 whereas for some of the data starting period is 1995. The reason is that some countries started to make CCI evaluations after 1995 and they do not have any statistical data before concerning this object.

4. Causality tests in time and frequency domain

The Granger causality tests indicate whether the past changes in x (y) have an impact on current changes in y (x) over a specified time period. Nevertheless, these test results can provide results on causality over all frequencies. On the other hand, Geweke's linear measure of feedback from one variable to another at a given frequency can provide detailed information about feedback relationships between growth and consumer confidence over different frequency bands. Even though frequency decompositions are generally investigated for neurophysiologic studies, it is important to address how the causality changes with frequency. This measure would enable us to quantify what fraction of total power at frequency [omega] of growth (consumer confidence index) is attributed to consumer confidence index (growth). Besides, studies such as Yildirim and Tastan (2009) show that the significance and / or direction of the Granger causality can change after adopting the causality test in frequency domain.

By using a Fourier transformation to VAR (p) model for x and y series, the Geweke's measure of linear feedback from y to x at frequency [omega] is defined as (6):

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]. (3)

If [[absolute value of [[psi].sub.12]([e.sup.-i[omega]])].sup.2] = 0, then the Geweke's measure will be zero, then y will not Granger cause x at frequency co. Breitung and Candelon (2006) present this test by reformulating the relationship between x and y in VAR equation:

[x.sub.t] = [[alpha].sub.1][x.sub.t-1] + .... + [[alpha].sub.p][x.sub.t-p] + [[beta].sub.1][y.sub.t-1] + ... + [[beta].sub.p][y.sub.t-p] + [[epsilon].sub.1t]. (4)

The null hypothesis tested by Geweke, [M.sub.y [right arrow] x] ([omega]) = 0 , corresponds to the null hypothesis of [H.sub.0] : R([omega])[beta] = 0 where [beta] is the vector of the coefficients of y and

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII].

Breitung and Candelon (2006) simplify the Geweke's null hypothesis so that a usual F-statistics can be used to test causality in frequency domain. Therefore, this study uses Breitung and Candelon (2006) version of Geweke (1982).

5. Empirical findings

5.1. Causality tests in time domain

Before conducting Granger causality tests in frequency domain, the causality tests are conducted in time domain. Table 3 summarizes the results for log-level and year-to-year growth rate specifications when CCI is considered vis-a-vis IP and RS. The lag orders are selected based on Akaike Information Criteria. We check both cases of a deterministic trend and no deterministic trend while employing the unconditional Granger causality analysis.

We have significance at 5% level for 57 cases out of 192 with almost no difference between trend and no trend cases. The group of Germany, France and Portugal has a total of 32 cases of causality whereas Czech Republic, Sweden and United Kingdom each have 2 and Italy has none. Hence, it is not possible to argue for the existence of causality depending on different levels of per capita income. On the other hand, Y-O-Y specification has a slight edge of 32 to 25, signaling a longer term perspective could better capture the dynamics of the relationships.

More important is the CCI-IP and CCI-RS pairings. We observe causality for 31 cases between CCI and IP compared to 26 between CCI and RS. CCI-RS pairing seems to work better under the Y-O-Y specification. Significant majority of the causality cases are unidirectional links between the pairings while we observe only 4 cases of bi-directional causality. These are CCI-IP for Germany and France in both log-level and Y-O-Y cases. 22 of the uni-directional cases are from CCI to IP/RS whereas 19 of them are from IP / RS to CCI. Overall, CCI causes IP in 14 cases and RS in 16 cases whereas IP causes CCI in 17 and RS causes CCI in 10 cases.

These results show that the consumers somewhat incorporate the past growth information as enhancing their expectations. Therefore, it is not possible to disregard that the agents in the economy are rational and use available growth prospect of the economy to form their expectations. Besides, the past changes in consumer confidence seem to slightly affect the growth of economy because the consumer confidence is a direct measure of their propensity to consume, wealth conditions and perceptions of the economic situation.

Nonetheless, the simple Granger causality methodology leaves lots of questions unanswered as it fails to deliver results that should emphasize the link between CCI-IP and CCI-RS pairings in a strong manner.

Next, we check whether a numeraire CCI would enhance our results by employing German CCI (CCIG) vis-a-vis other country's CCI (CCIX), IP (IPX) and RS (RSX). The results are in Table 4.

We observe some support for the dominance of German consumer confidence especially in pairings with consumer sentiment and industrial production indices of other highly developed European countries like France, Italy and United Kingdom. Nonetheless, it is not possible to advocate German consumer confidence index as the main leading indicator of European household behavior with respect to the time domain Granger causality analysis.

5.2. Causality test in frequency domain

Table 5 summarizes the Granger causality tests in frequency domain where 79 cases show the existence of causality out of 192 (7). This is simply an improvement of 38.6 per cent on time domain analysis. There is also a significant difference between low frequency (long-run) and seasonal frequency (short-run) cases as low frequency has 23 cases more than seasonal frequency.

When we consider CCI-IP and CCI-RS pairings, the superiority of frequency domain causality analysis becomes obvious. There are 42 cases between CCI and IP compared to 31 in time domain causality and 37 cases between CCI and RS compared to only 26 in time domain causality. There is a slight edge of 41 to 38 for log-level specification compared to Y-O-Y.

Significant majority of the causality cases are unidirectional links between the pairings. Nonetheless, we observe 15 cases of bi-directional causality, another drastic improvement from time domain case. Overall, CCI causes IP in 17 cases and RS in 13 cases whereas IP causes CCI in 25 and RS causes CCI in 24 cases.

The frequency domain results with respect to countries groups also differ from the time domain case as only Greece and Italy have less than 5 cases of causality. Therefore, we strongly believe that this is a sign of coherence between European emerging and developed countries.

These results underline the rationality of consumers as they gather significant information from several resources about production and / or sales and use it while forming their expectations as well as understanding the current stance of the economy. Besides, the past changes in economic growth and retail sales seem to affect the consumer sentiment through wealth conditions and perceptions of the economic situation. Therefore, we argue that there is a strong causal link between CCI-IP and RS.

Next, we perform our numeraire CCI exercise by checking the causality from German CCI ([CCI.sub.G]) to other country's CCI ([CCI.sub.X]), IP ([IP.sub.X]) and RS ([RS.sub.X]). The results are in Table 6.

Using frequency domain analysis again improves our results with 53 cases of causality out of 132, an improvement on the case of only 29 cases in time domain results. We observe support for the dominance of German consumer confidence in CCI and IP pairings with no specific pattern for countries. Moreover, we have causality in [CCI.sub.G]-[RS.sub.X] pairings, an outcome which we failed to obtain employing the time domain technique. Hence, it is possible to advocate German consumer confidence index as the main leading indicator of European household behavior with respect to the frequency domain Granger causality analysis (8).

Last, we need to emphasize the improvement that Breitung and Candelon (2006) test offers with respect to the simple Granger causality in time domain. The main reason for such this superiority depends on the notion that Granger causality considers an average measure to test the causality whereas the Geweke's approach decomposes the causality at each frequency.

6. Conclusions

This study assesses the link between consumer confidence, economic growth and retail sales for a group of 12 European nations employing simple part of the spectral density analysis. Our contribution is three-fold. First, while most of the previous studies analyze the expectations-consumption channel, we examine the dynamic nature of expectations-production channel as well. It is possible to argue that emerging markets could experiment business cycles at shorter horizons with respect to an industrialized economy, a factor, which results in different links between sentiment, growth and sales.

Secondly, we calculate Granger causality tests in both time and frequency domain and measure the forecasting power of the CCI at different forecasting horizons. Our empirical findings show that variations in consumer confidence mainly concentrate over seasonal frequencies. Besides, we observe significant feedbacks from consumer confidence to economic growth over seasonal frequencies as well as low frequencies. Hence, we conclude that consumer sentiment remains a useful predictor of growth for both short and long time horizons.

Thirdly, German CCI stands as the leading economic indicator for European area as we observe its effect on economic growth and retail sales of other countries for both short and long time horizons.

Consequently this study presents an analysis of the link between consumer confidence, economic growth and retail sales by the breakdown of variance over main frequency bands and causality in the time and frequency domain analysis. The empirical methodology we employ yields new and interesting additional insights into the causal relationship. For further research, our methodology can be applied to test the forecasting performance of leading economic and financial indicators like Business Tendency Surveys, Consumption Index and Wholesale Confidence Index.

doi: <DO>10.3846/16111699.2011.599406</DO>

References

Acemoglu, D.; A. Scott, A. 1994. Consumer Confidence and Rational Expectations: Are Agents Beliefs Consistent with the Theory?, Economic Journal 104: 1-19. doi:10.2307/2234671

Alessie, R.; Lusardi, A. 1997. Saving and Income Smoothing: Evidence from Panel Data, European Economic Review 41: 1251-1279. doi:10.1016/S0014-2921(96)00023-2

Balkyte, A.; Tvaronaviciene, M. 2010. Perception of competitiveness in the context of sustainable development: facets of "Sustainable competitiveness", Journal of Business Economics and Management 11(2): 341-365. doi:10.3846/jbem.2010.17

Basdas, U.; Celik, S. 2010. Frequency Domain Analysis of Consumer Confidence and Industrial Production in Turkey, Paper presented at 69th International Atlantic Economic Society Conference, Prague, March 24-27, 2010.

Batchelor, R.; Dua, P. 1998. Improving Macro-economic Forecasts: the Role of Consumer Confidence, International Journal of Forecasting 14: 71-81. doi:10.1016/S0169-2070(97)00052-6

Bram, J.; Ludvigson, S. 1998. Does Consumer Confidence Forecast Household Expenditures? A Sentiment Index Horse Race, Federal Reserve Bank of New York Economic Policy Review 4: 59-78.

Breitung, J.; Candelon, B. 2006. Testing for Short and Long-run Causality: a Frequency Domain Approach, Journal of Econometrics 132: 363-378. doi:10.1016/j.jeconom.2005.02.004

Carroll, C. D.; Fuhrer, J. C.; Wilcox, D. W. 1994. Does Consumer Sentiment Forecast Household Spending? If So, Why?, American Economic Review 84: 1397-1408.

Desroches, B.; Gosselin, M. A. 2002. The Usefulness of Consumer Confidence Indexes in the United States, Bank of Canada Working Paper No. 2002-22.

Dominitz, J.; Manski, C. F. 2004. How Should We Measure Consumer Confidence?, Journal of Economic Perspectives 18: 51-66. doi:10.1257/0895330041371303

Eppright, D. R.; Arguea, N. M.; Huth, W. L. 2003. Aggregate Consumer Expectation Indexes as Indicators of Future Consumer Expenditures, Journal of Economic Psychology 19: 215-235. doi:10.1016/S0167-4870(98)00005-1

Fan, C. S.; Wong, P. 1998. Does Consumer Sentiment Forecast Household Spending? The Hong Kong Case, Economics Letters 58: 77-84. doi:10.1016/S0165-1765(97)00247-4

Flavin, M. 1991. The Joint Consumption /Assets Demand Decision: a Case Study in Robust Estimation, NBER Working Paper No. 3802, Cambridge, MA.

Garner, A. 1991. Forecasting Consumer Spending: Should Economists Pay Attention to Consumer Confidence Surveys?, Federal Reserve Bank of Kansas City Economic Review 76: 57-71.

Gelper, S.; Lemmens, A.; Croux, C. 2007. Consumer Sentiment and Consumer Spending: Decomposing the Granger Causal Relationship in the Time Domain, Applied Economics 39: 1-11. doi:10.1080/00036840500427791

Geweke, J. 1982. Measurement of Linear Dependence and Feedback between Multiple Time Series, Journal of American Statistical Association 77: 304-324. doi:10.2307/2287238

Gomes, O. 2007. Consumer Confidence, Endogenous Growth and Endogenous Cycles, MPRA Paper No. 2883.

Gunes, H.; Uzun, S. 2010. Differences in Expectation Formation of Consumers in Emerging and Industrialized Markets, Paper presented at 69th International Atlantic Economic Society Conference, Prague, March 24-27, 2010.

Hosoya, Y. 1991. The Decomposition and Measurement of the Interdependence Between Second-order Stationary Process, Probability Theory and Related Fields 88: 429-444. doi:10.1007/BF01192551

Huth, W. L.; Eppright, D. R.; Taube, P. M. 1994. The Indexes of Consumer Sentiment and Confidence: Leading or Misleading Guides to Future Buyer Behavior, Journal of Business Research 29: 199-206. doi:10.1016/0148-2963(94)90004-3

Hufner, F. P.; Schroder, M. 2002. Forecasting Economic Activity In Germany-How Useful Are Sentiment Indicators?, Centre for European Economic Research Discussion Paper No. 02-56, September.

Jansen, W. J.; Nahuis, N. J. 2003. The Stock Market and Consumer Confidence: European Evidence, Economics Letters 79: 89-98. doi:10.1016/S0165-1765(02)00292-6

Katona, G. 1960. The Powerful Consumer. New York: McGraw Hill.

Kwan, A. C. C.; Cotsomitis, J. A. 2004. Can Consumer Attitudes Forecast Household Spending in United States? Further Evidence from the Michigan Surveys of Consumers, Southern Economic Journal 71: 136-144. doi:10.2307/4135316

Kwan, A. C. C.; Cotsomitis, J. A. 2006. The Usefulness of Consumer Confidence in Forecasting Household Spending in Canada: a National and Regional Analysis, Economic Inquiry 44: 185-197. doi:10.1093/ei/cbi064

Matsusaka, J. G.; Sbordone, A. M. 1995. Consumer Confidence and Economic Fluctuations, Economic Inquiry 33: 296-318. doi:10.1111/j.1465-7295.1995.tb01864.x

Nahuis, N. J. 2000. Are Survey Indicators Useful for Monitoring Consumption Growth? Evidence from European Countries, Working Paper 2000-08 (June), Monetary and Economics Policy Department, De Nederlandsche Bank, Amsterdam.

Otoo, M. W. 1999. Consumer Sentiment and the Stock Market, Finance and Economics Discussion Paper 1999-60, Federal Reserve Board, Washington, D.C, November.

Roberts, I.; Simon, J. 2001. What do Sentiment Surveys Measure?, Economic Research Reserve Bank of Australia, Research Discussion Paper.

Roos, M. W. M. 2008. Willingness to Consume and Ability to Consume, Journal of Economic Behavior and Organization 66: 387-402. doi:10.1016/j.jebo.2006.03.008

Souleles, N. S. 2004. Expectations, Heterogeneous Forecast Errors, and Consumption: Micro Evidence from the Michigan Consumer Sentiment Surveys, Journal of Money, Credit and Banking 36: 39-72. doi:10.1353/mcb.2004.0007

Throop, A. W. 1992. Consumer Sentiment: Its Causes and Effects, Federal Reserve Bank of San Francisco Economic Review 1: 35-59.

Tvaronaviciene, M.; Grybaite, V. 2007. Impact of FDI on Lithuanian economy: insight into development of main economic activities, Journal of Business Economics and Management 8(3): 285-290. ISSN 1611-1699 (Thomson ISI Web of Science).

Tvaronaviciene, M.; Grybaite, V.; Tvaronaviciene, A. 2009. If Institutional Performance Matters: Development Comparisons of Lithuania, Latvia and Estonia, Journal of Business Economics and Management 10(3): 271-278. ISSN 1611-1699 (Thomson ISI Web of Science). doi:10.3846/1611-1699.2009.10.271-278

Tvaronaviciene, M.; Kalasinskaite, K. 2010. Whether globalization in form of FDI enhances national wealth: empirical evidence from Lithuania, Journal of Business Economics and Management 11(1): 3-18. ISSN 1611-1699 (Thomson ISI Web of Science). doi:10.3846/jbem.2010.01

Tvaronavicius, V.; Tvaronaviciene, M. 2008. Role of fixed investments in economic growth of country: Lithuania in European context, Journal of Business Economics and Management (9)1: 57-65. ISSN 1611-1699 (Thomson ISI Web of Science). doi:10.3846/1611-1699.2008.9.57-64

Van Oest, R.; Franses, P. H. 2008. Measuring Changes in Consumer Confidence, Journal of Economic Psychology 29: 255-275. doi:10.1016/j.joep.2007.10.001

Yildirim, N.; Tautan, H. 2009. Capital Flows and Economic Growth across Spectral Frequencies: Evidence from Turkey, Turkish Economic Association Discussion Paper 2009/2.

Asli Yuksel Mermod [1], Gitana Dudzeviciute [2]

[1] Department of Business Administration, Marmara University, FEAS, Istanbul, Turkey

[2] Department of Economics and Management of Enterprises, Vilnius Gediminas Technical University, Sauletekio al. 11, LT-10223 Vilnius, Lithuania

E-mails: [1] asliyuksel@marmara.edu.tr (corresponding author); [2] gitana.dudzeviciute@vgtu.lt

Received 26 January 2011; accepted 27 March 2011

(1) The first survey of consumer attitudes has been in the United States by the University of Michigan in the 1940s. Katona (1960) is cited as the seminal study for the concept and measurement of consumer confidence.

(2) Roos (2008) incorporates Katona's theory into a standard model of intertemporal utility maximization by allowing for a time-varying preference parameter which is exogenous to the consumer and determined by the social environment.

(3) Consumption variable used is the total real personal consumption expenditures. It is usually partitioned into categories as durables, non-durables and services.

(4) As Carroll et al. (1994) state, "... the choice of which other variables to include in the equation is inherently somewhat arbitrary".

(5) We do not include Austria, Belgium, Poland and other smaller economies due to improper and / or shorter data series.

(6) For details of the computation of the measure, see Geweke (1982) and Breitung and Candelon (2006).

(7) No plots are given to save space. However, they are available from the corresponding author upon request.

(8) As our focus is not the sign between the variables, the co-spectrum analysis is not conducted. Here, our main interest is to quantify the causality between confidence and economic growth / retail sales, and, if available, to show how the frequency affects this causality.

Asli Yuksel MERMOD. Dr Asli Yuksel Mermod is a Professor of Finance in the Department of Business Administration at Marmara University, Istanbul Turkey. She teaches also as visiting finance professor at Webster University in Geneva-Switzerland and Bahcesehir University-Istanbul. Research areas cover Bank Management, Consumer Confidence Indices, Socially Responsible Investing, Ethical Finance, Brand Equity, Financial Services Marketing. She teaches International Banking, International Finance, Project Finance, Bank Management, Principles of Finance and Financial Markets and Institutions courses for undergraduates and Bank Funds Management, Project Finance and Management, Risk management in Banking for MBA, Ph.D. and Executive MBA classes.

Gitana DUDZEVICIUTE. Associate Professor, Doctor of social science (economics), Department of Enterprise Economics and Management, Vilnius Gediminas Technical University (Lithuania). Research arears cover economic growth, finance, banking sector development, marketing.

Table 1. The variables and their descriptions

Variable Name    Description

IP               Logarithm of seasonally adjusted industrial
                   production index
IP_Y-O-Y         Year-to-year growth rate of seasonally adjusted
                   industrial production index
RS               Logarithm of seasonally adjusted retail sales
RS_ Y-O-Y        Year-to-year growth rate of seasonally adjusted
                   retail sales
CCI              Logarithm of consumer confidence index
CCI_ Y-O-Y       Year-to-year growth rate of consumer confidence index

Table 2. Selected countries and time periods

Country           Start      End       Number of Observations

Czech Republic    Jan-96     May-10    173
Denmark           Dec-80     May-10    354
France            Dec-80     May-10    354
Germany           Dec-80     May-10    354
Greece            Jan-85     May-10    305
Hungary           Feb-93     May-10    208
Italy             Jan-90     May-10    245
Netherlands       Dec-80     May-10    354
Portugal          Jan-90     May-10    245
Spain             Jan-95     May-10    185
Sweden            Oct-95     May-10    176
United Kingdom    Dec-80     May-10    354

Table 3. Granger causality in time domain for all countries

Case [right     Log-level
arrow]

Variables       CCI [right   IP           IP [right    CCI
[right arrow]   arrow]                    arrow]

Country         T            NT           T            NT

Czech           0.536        0.441        3.775#       3.288#
Republic        (0.658)      (0.724)      (0.012)#     (0.022)#

Denmark         0.219        0.340        2.745#       1.266
                (0.883)      (0.796)      (0.043)#     (0.286)

France          3.732#       3.700#       6.001#       5.960#
                (0.012)#     (0.012)#     (0.001)#     (0.001)#

Germany         4.329#       4.257#       3.893#       3.383#
                (0.005)#     (0.006)#     (0.009)#     (0.018)#

Greece          0.233        0.737        1.205        0.941
                (0.873)      (0.531)      (0.308)      (0.421)

Hungary         1.627        1.434        0.507        0.380
                (0.184)      (0.234)      (0.678)      (0.768)

Italy           0.506        0.570        1.551        1.857
                (0.678)      (0.635)      (0.201)      (0.137)

Netherlands     1.923        1.775        1.169        0.256
                (0.125)      (0.152)      (0.321)      (0.857)

Portugal        0.226        0.267        3.981#       4.736#
                (0.878)      (0.849)      (0.008)#     (0.003)#

Spain           3.572#       2.791#       1.792        1.537
                (0.015)#     (0.041)#     (0.149)      (0.205)

Sweden          1.864        1.635        0.413        0.306
                (0.138)      (0.183)      (0.744)      (0.821)

United          1.121        1.121        1.110        1.103
Kingdom         (0.341)      (0.341)      (0.345)      (0.348)

Case [right     Log-level
arrow]

Variables       CCI [right   RS           RS [right    CCI
[right arrow]   arrow]                    arrow]

Country         T            NT           T            NT

Czech           0.376        0.098        0.799        0.484
Republic        (0.770)      (0.961)      (0.496)      (0.693)

Denmark         1.068        0.876        1.263        0.937
                (0.363)      (0.454)      (0.287)      (0.423)

France          1.326        1.238        4.150#       4.252#
                (0.266)      (0.296)      (0.007)#     (0.006)#

Germany         3.032#       2.968#       0.611        0.678
                (0.029)#     (0.032)#     (0.608)      (0.566)

Greece          2.663#       2.889#       1.387        1.411
                (0.048)#     (0.036)#     (0.247)      (0.240)

Hungary         2.064        1.963        2.101        2.146
                (0.106)      (0.121)      (0.101)      (0.096)

Italy           0.419        0.404        1.039        0.997
                (0.740)      (0.750)      (0.376)      (0.395)

Netherlands     5.746#       5.7633       1.325        1.149
                (0.001)#     (0.001)#     (0.266)      (0.329)

Portugal        0.820        0.861        4.553#       4.133#
                (0.484)      (0.462)      (0.004)#     (0.007)#

Spain           0.773        0.850        0.485        0.501
                (0.510)      (0.468)      (0.693)      (0.682)

Sweden          2.042        1.670        2.107        1.610
                (0.110)      (0.175)      (0.101)      (0.189)

United          0.399        0.341        1.713        1.877
Kingdom         (0.753)      (0.796)      (0.164)      (0.133)

Case [right     Y-O-Y
arrow]

Variables       CCI [right   IP           IP [right    CCI
[right arrow]   arrow]                    arrow]

Country         T            NT           T            NT

Czech           0.469        0.471        2.597        2.648
Republic        (0.704)      (0.703)      (0.054)      (0.051)

Denmark         1.093        1.115        3.023#       2.808#
                (0.352)      (0.343)      (0.030)#     (0.040)#

France          4.914#       5.007#       7.002#       7.113#
                (0.002)#     (0.002)#     (0.000)#     (0.000)#

Germany         6.906#       6.929#       4.053#       4.017#
                (0.000)#     (0.000)#     (0.007)#     (0.008)#

Greece          1.298        1.253        1.252        1.268
                (0.275)      (0.291)      (0.291)      (0.286)

Hungary         2.401        2.752#       2.132        2.475
                (0.069)      (0.044)#     (0.098)      (0.063)

Italy           1.271        1.253        1.811        1.780
                (0.284)      (0.291)      (0.145)      (0.151)

Netherlands     2.583        2.637#       1.528        1.532
                (0.053)      (0.050)#     (0.207)      (0.206)

Portugal        0.197        0.309        3.110#       3.709#
                (0.898)      (0.819)      (0.027)#     (0.012)#

Spain           6.501#       6.246#       1.890        1.744
                (0.000)#     (0.000)#     (0.132)      (0.158)

Sweden          1.919        1.933        1.573        1.157
                (0.129)      (0.127)      (0.198)      (0.328)

United          1.097        0.910        2.030        1.606
Kingdom         (0.351)      (0.437)      (0.109)      (0.188)

Case [right     Y-O-Y
arrow]

Variables       CCI [right   RS           RS [right    CCI
[right arrow]   arrow]                    arrow]

Country         T            NT           T            NT

Czech           0.604        0.603        1.151        1.108
Republic        (0.613)      (0.614)      (0.331)      (0.348)

Denmark         1.018        1.008        1.481        1.469
                (0.385)      (0.389)      (0.220)      (0.223)

France          0.568        0.557        10.950#      11.087#
                (0.637)      (0.644)      (0.000)#     (0.000)#

Germany         2.926#       2.908#       0.698        0.668
                (0.034)#     (0.035)#     (0.554)      (0.572)

Greece          2.937#       2.919#       0.892        0.894
                (0.034)#     (0.034)#     (0.446)      (0.445)

Hungary         5.600#       5.507#       2.545        2.308
                (0.001)#     (0.001)#     (0.058)      (0.078)

Italy           1.774        1.762        2.553        2.476
                (0.153)      (0.155)      (0.056)      (0.062)

Netherlands     6.788#       6.788#       1.869        1.871
                (0.000)#     (0.000)#     (0.135)      (0.134)

Portugal        1.191        1.165        4.021#       3.745#
                (0.314)      (0.324)      (0.008)#     (0.012)#

Spain           0.929        0.831        0.312        0.311
                (0.428)      (0.478)      (0.816)      (0.817)

Sweden          2.779#       2.815#       1.629        1.623
                (0.043)#     (0.041)#     (0.185)      (0.186)

United          1.072        1.071        2.967#       2.966#
Kingdom         (0.361)      (0.361)      (0.032)#     (0.032)#

Note: p-values are given in the brackets. T stands for the case with a
deterministic trend and NT for the case with no deterministic trend.
Bold values indicate significance at 5% level

Note: # with Bold values indicate significance at 5% level

Table 4. Granger causality in time domain for German CCI

Case [right     Log-level
arrow]

Variables       [CCI.sub.G]   [CCI.sub.X]   [CCI.sub.G]   [IP.sub.X]
[right arrow]   [right                      [right
                arrow]                      arrow]

Country         T             NT            T             NT

Czech           0.385         0.494         2.090         3.373
Republic        (0.536)       (0.483)       (0.150)       (0.067)

Denmark         4.341#        3.781         0.561         0.151
                (0.038)#      (0.053)       (0.455)       (0.697)

France          0.588         0.581         6.235#        6.822#
                (0.444)       (0.446)       (0.013)#      (0.009)#

Greece          1.910         2.284         1.024         1.450
                (0.168)       (0.131)       (0.312)       (0.229)

Hungary         1.109         1.116         10.525#       12.101#
                (0.294)       (0.292)       (0.001)#      (0.001)#

Italy           5.952#        5.9723#       6.425#        6.645#
                (0.015)#      (0.015)#      (0.012)#      (0.010)#

Netherlands     2.389         2.425         0.641         3.107
                (0.123)       (0.120)       (0.424)       (0.079)

Portugal        4.958#        4.685         0.069         0.120
                (0.027)#      (0.031)#      (0.792)       (0.729)

Spain           0.078         0.034         5.075#        5.091#
                (0.780)       (0.854)       (0.025)#      (0.025)#

Sweden          1.673         1.692         21.219#       23.056#
                (0.198)       (0.195)       (0.000)#      (0.000)#

United          0.188         0.181         10.859#       10.480#
Kingdom         (0.665)       (0.671)       (0.001)#      (0.001)#

Case [right     Log-level                   Y-O-Y
arrow]

Variables       [CCI.sub.G]   [RS.sub.X]    [CCI.sub.G]   [CCI.sub.X]
[right arrow]   [right                      [right
                arrow]                      arrow]

Country         T             NT            T             NT

Czech           0.003         0.001         3.352         3.347
Republic        (0.959)       (0.998)       (0.069)       (0.069)

Denmark         0.301         0.239         3.744         3.758
                (0.584)       (0.625)       (0.054)       (0.053)

France          0.042         0.033         0.919         0.883
                (0.837)       (0.855)       (0.338)       (0.348)

Greece          3.730         3.373         0.117         0.145
                (0.054)       (0.067)       (0.733)       (0.704)

Hungary         0.667         0.604         0.772         0.802
                (0.415)       (0.438)       (0.381)       (0.372)

Italy           1.568         1.317         8.014#        8.051#
                (0.212)       (0.252)       (0.005)#      (0.005)#

Netherlands     0.001         0.077         0.715         0.705
                (0.992)       (0.782)       (0.398)       (0.402)

Portugal        0.040         0.114         6.709         6.836
                (0.842)       (0.735)       (0.010)#      (0.009)#

Spain           0.602         0.386         0.285         0.256
                (0.439)       (0.535)       (0.594)       (0.613)

Sweden          1.415         1.416         0.875         0.875
                (0.235)       (0.235)       (0.351)       (0.351)

United          0.657         0.749         0.403         0.406
Kingdom         (0.418)       (0.387)       (0.526)       (0.524)

Case [right     Y-O-Y
arrow]

Variables       [CCI.sub.G]   [IP.sub.X]    [CCI.sub.G]   [RS.sub.X]
[right arrow]   [right                      [right
                arrow]                      arrow]

Country         T             NT            T             NT

Czech           2.211         2.263         0.022         0.042
Republic        (0.138)       (0.134)       (0.882)       (0.837)

Denmark         0.387         0.397         1.065         1.061
                (0.534)       (0.529)       (0.303)       (0.304)

France          5.308#        5.346#        0.163         0.060
                (0.022)#      (0.021)#      (0.687)       (0.807)

Greece          1.654         1.703         1.086         1.120
                (0.199)       (0.193)       (0.298)       (0.291)

Hungary         5.649#        5.805#        2.432         2.463
                (0.018)       (0.017)#      (0.121)       (0.118)

Italy           6.525#        6.648#        0.434         0.463
                (0.011)#      (0.010)       (0.511)       (0.497)

Netherlands     1.38          1.366         0.003         0.003
                (0.241)       (0.243)       (0.954)       (0.956)

Portugal        1.762         2.003         0.809         0.945
                (0.185)       (0.158)       (0.369)       (0.332)

Spain           2.722         2.691         0.092         0.025
                (0.100)       (0.102)       (0.762)       (0.875)

Sweden          13.661#       13.745#       0.001         0.011
                (0.000)#      (0.000)#      (0.987)       (0.917)

United          3.705         3.551         0.279         0.281
Kingdom         (0.055)       (0.060)       (0.597)       (0.596)

Note: p-values are given in the brackets. T stands for the case with a
deterministic trend and NT for the case with no deterministic trend.
Bold values indicate significance at 5% level

Note: # with Bold values indicate significance at 5% level

Table 5. Granger causality in frequency domain for all countries

Case [right       Log-level
arrow]

Variables         CCI [right         IP [right
[right arrow]     arrow] IP          arrow] CCI

Country               LF       SF        LF       SF

Czech Republic        Y         Y        NF       NF
Denmark               NF       NF        Y        NF
France                Y         Y        Y         Y
Germany               Y        NF        Y         Y
Greece                NF       NF        Y        NF
Hungary               NF       NF        Y        NF
Italy                 NF       NF        NF       NF
Netherlands           NF       NF        Y        NF
Portugal              Y         Y        NF       NF
Spain                 Y        NF        Y         Y
Sweden                NF       NF        Y        NF
United Kingdom        NF       NF        Y        NF

Case [right       Log-level
arrow]

Variables         CCI [right         RS [right
[right arrow]     arrow] RS          arrow] CCI

Country               LF       SF        LF       SF

Czech Republic        NF       NF        Y        NF
Denmark               NF       NF        Y        NF
France                Y         Y        NF       NF
Germany               NF       NF        Y         Y
Greece                NF       NF        Y        NF
Hungary               NF        Y        NF        Y
Italy                 NF       NF        Y        NF
Netherlands           NF       NF        Y         Y
Portugal              Y         Y        Y        NF
Spain                 NF       NF        Y        NF
Sweden                NF       NF        Y         Y
United Kingdom        Y         Y        Y        NF

Case [right       Y-O-Y
arrow]

Variables         CCI [right         IP [right
[right arrow]     arrow] IP          arrow] CCI

Country               LF       SF        LF       SF

Czech Republic        NF        Y        NF       NF
Denmark               NF        Y        Y         Y
France                Y         Y        Y         Y
Germany               NF       NF        Y         Y
Greece                NF       NF        Y        NF
Hungary               NF       NF        Y        NF
Italy                 NF       NF        NF       NF
Netherlands           NF       NF        Y        NF
Portugal              Y         Y        NF       NF
Spain                 Y         Y        Y         Y
Sweden                Y        NF        Y        NF
United Kingdom        NF       NF        Y        NF

Case [right       Y-O-Y
arrow]

Variables         CCI [right         RS [right
[right arrow]     arrow] RS          arrow] CCI

Country               LF       SF        LF       SF

Czech Republic        NF       NF        Y        NF
Denmark               NF       NF        Y        NF
France                Y         Y        Y        NF
Germany               NF       NF        Y        NF
Greece                NF       NF        Y        NF
Hungary               NF        Y        NF        Y
Italy                 NF       NF        Y        NF
Netherlands           NF       NF        Y         Y
Portugal              Y        NF        Y        NF
Spain                 NF       NF        NF       NF
Sweden                NF       NF        NF       NF
United Kingdom        Y         Y        NF       NF

Note: Y denotes significance at 5% level and NF stands for No
feedback. LF denotes Low Frequency which is higher than 18 months with
0 < [omega] < 0.35 and SF stands for Seasonal Frequency which is for
the period 2 months to 18 months with 0.35 < [omega] < [pi].

Table 6. Granger causality in frequency domain for German CCI

Case [right       Log-level
arrow]

Variables         [CCI.sub.G]   [CCI.sub.X]   [CCI.sub.G]   [IP.sub.X]
[right arrow]     [right                      [right
                  arrow]                      arrow]

Country           LF            SF            LF            SF

Czech Republic    NF            NF            Y             Y
Denmark           Y             Y             NF            NF
France            Y             Y             NF            NF
Greece            NF            NF            NF            Y
Hungary           NF            NF            Y             Y
Italy             NF            Y             NF            Y
Netherlands       Y             Y             Y             Y
Portugal          NF            Y             NF            NF
Spain             Y             NF            NF            Y
Sweden            Y             NF            Y             Y
United Kingdom    NF            NF            Y             Y

Case [right       Log-level                   Y-O-Y
arrow]

Variables         [CCI.sub.G]   [RS.sub.X]    [CCI.sub.G]   [CCI.sub.X]
[right arrow]     [right                      [right
                  arrow]                      arrow]

Country           LF            SF            LF            SF

Czech Republic    NF            NF            NF            NF
Denmark           NF            NF            Y             Y
France            NF            NF            Y             Y
Greece            NF            NF            NF            NF
Hungary           NF            NF            NF            NF
Italy             NF            NF            Y             Y
Netherlands       Y             NF            Y             Y
Portugal          Y             NF            Y             Y
Spain             NF            NF            Y             NF
Sweden            NF            NF            Y             NF
United Kingdom    NF            NF            Y             NF

Case [right       Y-O-Y
arrow]

Variables         [CCI.sub.G]   [IP.sub.X]    [CCI.sub.G]   [RS.sub.X]
[right arrow]     [right                      [right
                  arrow]                      arrow]

Country           LF            SF            LF            SF

Czech Republic    NF            NF            NF            NF
Denmark           NF            NF            Y             NF
France            Y             Y             NF            NF
Greece            NF            Y             NF            NF
Hungary           Y             Y             NF            NF
Italy             NF            Y             NF            NF
Netherlands       Y             Y             Y             NF
Portugal          NF            NF            NF            NF
Spain             NF            Y             NF            NF
Sweden            Y             Y             NF            NF
United Kingdom    Y             Y             NF            NF

Note: Y denotes significance at 5% level and NF stands for No
feedback. LF denotes Low Frequency which is higher than 18 months with
0 < [omega] < 0.35 and SF stands for Seasonal Frequency which is for
the period 2 months to 18 months with 0.35 < [omega] < n.