Alternative roles of consumer confidence in forecasting consumption: Evidence from European countries

Transcription

1 Alernaive roles of consumer confidence in forecasing consumpion: Evidence from European counries Jana ZÁVACKÁ * Deparmen of Sysems Engineering, Faculy of Economics, VŠB-Technical Universiy of Osrava, Sokolská ř. 33, Osrava, Czech epublic. Absrac The predicive power of he consumer senimen indicaor o forecas consumpion spending has already been acceped, bu he heoreical explanaion is sill vague. Moreover, some sudies have shown ha, afer adding oher macroeconomic variables o he forecasing model, he predicive power of he consumer confidence indicaor becomes limied. This paper proposes o disinguish beween wo predicive channels of confidence daa for consumpion spending. The firs, he income channel, should represen he abiliy of confidence daa o predic he expeced fuure income of consumers. The predicive power of confidence hrough he second channel should serve for he approximaion of consumers inenions o consume. According o his disincion, I propose o use he daa from iniial survey quesions separaely raher han using one composie consumer confidence indicaor. In his way he predicive power of confidence daa can be beer clarified. I use he daa from he European Commission Consumer Survey and, using a sample of 7 counries, confirm ha higher predicive power of confidence daa could be achieved by using separae confidence daa raher han a composie indicaor. Keywords Consumer confidence, consumer confidence indicaor, forecasing, precauionary savings. JEL Classificaion: C53, E1 * jana.zavacka1@vsb.cz This paper was funded by he Suden Gran Compeiion of VŠB-TU Osrava wihin he projec SP013/91, wihin he Operaional Programme Educaion for Compeiiveness Projec No. CZ.1.07/.3.00/0.096 and wihin he programme of Suppor of esearch and Developmen in he Moravian-Silesian egion in Published by VŠB-TU Osrava. All righs reserved. E-CEEI, Volume 19: (016). ISSN (Prin), (Online) doi: /cerei

2 134 Ekonomická revue Cenral European eview of Economic Issues 19, 016 Alernaive roles of consumer confidence in forecasing consumpion: Evidence from European counries Jana ZÁVACKÁ 1. Inroducion The predicive power of senimen indicaors for shorerm predicions of consumpion spending has already been invesigaed for several decades. Sill, he resuls are slighly mulivalen. On he basis of Michigan s Index of Consumer Senimen (ICS), Carrol e al. (1994) confirmed he saisical significance of he senimen indicaor for forecasing consumpion spending in he USA. The same resuls have also been confirmed, for example, by Acemoglu and Sco (1994), Bram and Ludvigson (1998), Souleles (001) and Throop (199). Al-Eyd e al. (009) and Cosomiis and Kwan (006) reached he conclusion ha he role of he consumer senimen indicaor in he predicion of fuure consumpion spending is, afer conrolling for oher measurable macroeconomic variables, a leas limied. However, Dees and Brinca (013) saed ha, while he evidence is overall raher mixed, mos auhors seem o, a leas, find a significan saisical relaionship beween confidence measures and economic variables, curren and fuure. Lahiri e al. (01) re-examined he exising empirical models of consumpion and consumer confidence, obaining a resul in favour of consumer confidence s significance. A possible explanaion for he raher mulivalen resuls abou he significance of consumer confidence in predicing consumpion spending could be ha his significance varies wih ime or he business cycle. This idea was suppored by Bachelor and Dua (1998), who saed ha using he confidence indicaor would have been helpful in forecasing he recession in 1991, bu oherwise he predicive power is weak. Dees and Brinca (013) used hreshold models and argued ha he imporance of confidence increased when only large changes in confidence were considered. Haugh (005) and Howrey (001) saed ha he confidence indicaor was significan in predicing recessions or recoveries. According o Taylor and McNabb (007), he consumer senimen is procyclical and could be used o deermine he urning poins of he business cycle. The imporance of he confidence indicaor in modelling consumpion spending is also decreased by he fac ha srong heoreical suppor for i is lacking. The sudies invesigaing he predicive power of confidence indicaors have mainly been empirical. One of he aemps o explain he significance of consumer senimen for he predicion of consumer spending was made by Carrol e al. (1994), who used consumer senimen as a proxy for consumers expecaions abou heir fuure income sream. In line wih he life cycle hypohesis, Carrol e al. (1994) rejeced he idea ha he predicive power of consumer confidence for consumpion spending was based only on income expecaions. They obained he same resul by allowing for rule-of-humb consumers consisen wih he Campbell Mankiw model. The inerpreaion of predicive power of confidence indicaors hus remains mainly inuiive. The main goal of his aricle is o ry o exend he heoreical inerpreaion of he predicive power of he consumer confidence indicaor. In accordance wih Carrol e al. (1994), I aribue par of he predicive power of consumer confidence o expecaions abou a fuure income sream. In addiion, I inroduce he second possible channel of he influence of confidence on consumpion spending, called willingness o spend. 1 This channel represens he moivaion of consumers o spend or raher save heir income according o he acual and expeced economic condiions. I invesigaed he predicion power of he confidence indicaor wih respec o hese wo differen channels using he daa of he Join Harmonised EU Programme of Business and Consumer Surveys. Consisen wih Dreger and Kholodilin (013), I did no use he consruced consumer indicaor bu raher he daa from he iniial survey quesions (hereinafer referred o as confidence quesions). Differenly from Dreger and 1 This is de faco an analogy o he marginal propensiy o consume bu is esimaed by confidence indicaors.

3 J. Závacká Alernaive roles of consumer confidence in forecasing consumpion: Evidence from European counries 135 Kholodilin (013), I did no choose confidence quesions or heir weighs in he model purely based on he empirical analysis bu raher proposed o disinguish among confidence quesions wih respec o differen predicive channels. To suppor his approach, I esimaed firs he predicive power of he consumer confidence indicaor consruced by he European Commission in he model and hen he predicive power of he combined daa from he chosen confidence quesions. On he basis of he empirical esimaions, I concluded in line wih Dreger and Kholodilin (013) ha he predicive power of confidence daa for consumpion spending could be increased by a suiable choice of confidence quesions and heir weighs in he model. The aricle is organized as follows. In Secion I inroduce wo possible channels of confidence predicive power and describe he economeric models for empirical esing. The daa and empirical resuls are presened in Secion 3. In Secion 4 I conclude. See he Appendix for a more deailed descripion of he survey mehodology of confidence daa.. Mehodology Firsly, I will inroduce and heoreically explain he wo possible channels of confidence s predicive power. Furher, he model soluion for hese channels is presened..1 Two channels of he predicive power of confidence I formulaed he consumpion spending equaion in line wih he raional expecaion permanen income hypohesis (Hall, 1978). This hypohesis is based on he implemenaion of he concep of raional expecaions ino he permanen income hypohesis (Friedman, 1957). The permanen income heory posulaes ha only changes in income ha are perceived by he consumer as permanen lead o consumpion spending adjusmens in he long erm. The lifeime sream of incomes is esimaed by raional expecaions. These expecaions should be formed no only according o he pas levels of esimaed variables bu also wih respec o all he achievable informaion on he par of a consumer. In his sense he consumpion spending could be formed as he level of previous consumpion spending adjused for he changes in consumer income expecaions. If hese changes were unexpeced, hen he consumpion spending would be a random-walk process (random-walk hypohesis (Hall, 1978)). However, Flavin (1993) found significan excess sensiiviy of consumpion spending o curren changes in income; he changes in consumpion seemed o be larger han he expeced changes in lifeime income. Thus, he random-walk hypohesis was shaken. Here he role of consumer confidence in providing informaion abou consumers expecaions arises. Carrol e al. (1994) confirmed using he Universiy of Michigan s Index of Consumer Senimen ha his indicaor has predicive power for forecasing fuure consumpion spending while conrolling for pas values of consumpion spending. In his way he random-walk hypohesis was rejeced. Furher, using he Campbell Mankiw model, hey rejeced in mos cases ha he predicive power of confidence would ac only hrough he income channel. They proposed ha consumer confidence could also be considered as a measure of uncerainy. Accordingly, changes in consumer confidence could reflec he precauionary savings proposed by Carroll e al. (199). The confidence indicaor would be correlaed posiively wih pas values of consumpion growh bu negaively wih is curren value. Thus, he second channel for he predicive power of confidence indicaor was inroduced. I would like o emphasize he difference beween hese wo channels. I propose o divide he process of decision making abou consumpion spending ino wo seps. In he firs sep, he consumer will beer or worse ry o esimae his/her curren and fuure income. In he second sep, according o his/her expeced financial possibiliies, he consumer will decide abou his consumpion and savings. The sudies ha explained consumpion according o expeced fuure income buil he expecaions abou consumpion only according o he firs sep, assuming ha a higher expeced income implies higher consumpion spending. However, as Carrol e al. (1994) already menioned, even a posiive fuure expecaion does no necessarily mean equal consumpion growh (precauionary moives). This idea is in fac he consideraion of he second sep in he consumer decision. Le me call his second channel, hrough which he consumpion decision is formed in he second sep, he willingness o spend. In his sep he decision abou consumpion spending is influenced by all he oher addiional facors excep he expeced fuure income. One of hese facors, uncerainy, has already been inroduced by Carrol e al. (1994), moivaing consumers o engage in precauionary saving. As anoher facor I can consider he neighbourhood, presened by Duesenberry (1949). According o his I can sar o polemize he exen o which he consumer is able o mainain his financial plan, wih he influence of his/her neighbourhood and supply-side incenives, bu for simpliciy I will no consider his impac.

4 136 Ekonomická revue Cenral European eview of Economic Issues 19, 016 hypohesis, he consumpion spending of an individual is influenced by he consumpion spending of his neighbours. I imagine ha here could be addiional facors influencing consumer decisions abou consumpion, such as he family background, risk awareness, he life siuaion and so on, which have a posiive or negaive impac on consumpion. Alhough I do no know all he facors and heir impacs on consumer spending, I do no in fac necessarily need o know hem. I is enough for me o know wha he final decision of a consumer will be regarding consumpion spending wih respec o hese facors. In he survey daa, here are also confidence quesions asking consumers abou heir inenions relaing o major purchases or savings. Hence, he confidence daa could be used no only for approximaing consumer expecaions bu also for predicing consumer inenions.. The consumpion growh equaion On he basis of hese wo channels, I can model he consumpion growh as c c E y WS (1) C, where,, are consans, 0 1 of fuure income growh, WS Ey 1 is an expecaion is he willingness o spend, represening consumer inenions, and C is an error erm. I furher propose he model according o he confidence quesions from he Join Harmonised EU Programme of Business and Consumer Surveys (EC, 016) published by he European Commission. 3 The firs wo quesions from he survey focus on he financial siuaion of he consumer. Because he firs quesion asks abou he change wih respec o a pas period, I decided o use as a proxy for consumers expecaions abou he fuure sream of incomes he daa from survey quesion, which is How do you expec he financial posiion of your household o change over he nex 1 monhs? Because he daa abou consumpion are available wih a quarerly frequency and he survey quesion is asking for a comparison wih he las year, I also include in he equaion he las four values of income growh and define he expeced growh of fuure income () Y YY Y 1 i i i i E y cci y, Y where YY i, i are model parameers, y is he growh i of income, cci i is he average value of he response Y o confidence quesion a ime i and is an error erm. The willingness o spend should represen he inenions of consumers o spend heir income. There are more quesions in he survey ha could be considered in his case. In confidence quesion 8, consumers are asked if hey hink ha i is he righ momen for people o make major purchases. Alhough a consumer may answer ha i is he righ momen (because he prices for furniure, elecronic devices, ec. are low or i is he period of sales), he/she may no acually be planning o make major purchases. Thus, he answers do no necessarily reflec he consumers inenions. I propose o use he daa from survey quesion 9, which is Compared o he pas 1 monhs, do you expec o spend more or less money on major purchases (furniure, elecrical/elecronic devices, ec.) over he nex 1 monhs? Analogous o quesion, he quesion asks he consumer abou he comparison wih he level in a previous year. For ha reason I ake ino accoun he previous levels of consumpion growh and define he willingness o spend: (3) CO9 COC CO i i i i WS cci9 c. cci is he weighed average response o quesion 9 a ime i and is an error erm. CO 9 COC, are parameers, i i 9 i CO I can furher consider using more confidence quesions for his channel. In quesion 10 consumers are asked abou he righ momen o save; here a similar reasoning o quesion 8 could be applied. Anoher possible opion is quesion 11, in which he consumers are asked how likely hey are o save money in he nex 1 monhs. This quesion is very similar o quesion 1, when hey are asked if hey are now saving any money. I ried o perform he esimaions for boh quesions and he resuls were very similar; hence, I preferred quesion 1, which could be more realisic as he consumers are asked how hey really behave now. Therefore, I would like o presen an alernaive version of he definiion of by adding he daa from survey quesion 1, which is Which of hese saemens bes describes he curren financial siuaion of your household? The possible answers o his quesion are: we are saving a lo (+ +), we are saving a lile (+), we are jus managing o make ends mee on our income (=), we are having o draw on our savings ( ), we are running ino deb ( ) and don know (N). This confidence quesion should represen consumers moivaion for saving. WS, 3 All he confidence quesions wih heir possible answers are in he Appendix.

5 J. Závacká Alernaive roles of consumer confidence in forecasing consumpion: Evidence from European counries 137 CO9 CO1 9 1 i i i i WS cci cci 4 i1 c COC CO i i. (4) CO1 are parameers and i cci1 i is he weighed average response o quesion 1 a ime i. I would also be possible o consider he quarerly confidence quesions 13 15, bu hese quesions are focused on more expensive purchases, like cars, houses, home improvemens and so on. These purchases are very ofen financed by loans or morgages and from he economic poin of view are ofen considered as an invesmen. I was worried abou various inerpreaions of he answers; hence, I leave hese quesions for fuure invesigaion. By subsiuing () and (3) ino equaion (1), I obained CO9 COC 9 0 i i i i c cci c Y YY cci y i i i i, (5) where is an error erm. I used his equaion as model M1 for esing he predicive power of confidence indicaors for forecasing consumpion spending. By analogy, I subsiued () and (4) ino equaion (1) and obained he second model, M: CO9 CO i i i i c cci cci 4 COC Y YY c cci y i i i i i i i1. (6) Boh model M1 (5) and M (6) are very similar o he models ha were esed for example by Al-Eyd e al. (009) and Carrol e al. (1994) for he predicive power of consumer confidence indicaor. Al-Eyd e al. (009) and Carrol e al. (1994) used in heir models he log-difference of consumpion, which is an approximaion of consumpion growh. The European Commission wihin he Join Harmonised EU Programme of Business and Consumer Surveys has also published a proposed indicaor for consumer confidence called he consumer confidence indicaor. This indicaor is he average response from quesions, 4, 7 and 11 4 and is usually used in sudies conduced using EU daa. To compare he effeciveness of he daa from specific survey quesions raher han he consumer confidence indicaor, I also esimaed he generally used model M0 in he form COF COC 0 i i i i c cof c 4 i1 y. YY i i (7) COF are model parameers and cof i is he consumer i confidence indicaor a ime i. The common predicive power of confidence indicaors in he iniial models is calculaed by comparison wih heir common sub-model: (8) COF YY 0 i i i i c cof y. The predicive power is expressed as an increase in he coefficien of deerminaion (incremenal ) caused by adding confidence indicaors o he model wih respec o he coefficien of deerminaion from he baseline model (9). 3. Empirical resuls The predicive power of confidence indicaors is firs considered wihin he correlaion analysis and hen by he model esimaions. 3.1 Daa For he main macroeconomic variables I used he quarerly daa from he Naional Accouns published by Eurosa. I couned he consumpion growh and income growh from he final consumpion expendiure of households and he gross domesic produc (GDP) a marke prices. Boh series were chain linked volumes (005) expressed in million unis of naional currency, adjused seasonally and by working days. The confidence daa were from he Join Harmonised EU Programme of Business and Consumer Surveys published by he European Commission. Due o harmonizaion, he daa are comparable wihin he EU counries. The daa for confidence quesions cci, cci9, cci 1 and he confidence indicaor cof were ransformed from monhly o quarerly daa by average. A deailed descripion abou he mehodology of he consumer survey daa can be found in he Appendix. Due o he differen lenghs of available ime series for EU member counries, I decided o use he daa of 7 EU member counries: Ausria (1996Q 015Q4), Belgium (1995Q 015Q4), Bulgaria (001Q 015Q4), Croaia (005Q 015Q4), Cyprus (001Q 015Q4), he Czech epublic (1996Q 015Q4), Denmark (1995Q 015Q4), Esonia (1995Q 015Q4), Finland (1995Q4 y c 4 The exac definiion of his indicaor can be found a he end of he Appendix.

6 138 Ekonomická revue Cenral European eview of Economic Issues 19, Q4), France (1995Q 015Q4), Germany (1995Q 014Q3), Greece (1995Q 015Q4), Hungary (1995Q 015Q4), Ialy (1996Q 015Q4), Lavia (001Q 015Q4), Lihuania (001Q 015Q4), Luxembourg (00Q1 015Q4), Mala (00Q4 015Q4), he Neherlands (1996Q 016Q1), Poland (00Q 015Q4), Porugal (1995Q 015Q4), omania (001Q 015Q4), Slovakia (1999Q 015Q4), Slovenia (1996Q1 015Q4), Spain (1995Q 015Q4), Sweden (1995Q4 015Q4) and he Unied Kingdom (1995Q 015Q4). Like Al-Eyd e al. (009) and Dreger and Kholodilin (013), he esimaions were performed for each counry separaely. 3. Correlaion analysis For he beginning of he invesigaion of he possible predicive power of confidence quesions for consumpion spending, I conduced a correlaion analysis. I esimaed he correlaion coefficiens among he model variables in he logic of he wo channels inroduced in (1). For he income channel,, he correlaion among income growh, y quesion and confidence indicaor esimaed. The resuls are displayed in Table 1. cci Ey 1 confidence cof was I can observe from he esimaions ha he correlaion beween income growh and he consumer confidence indicaor as well as he correlaion beween income growh and confidence quesion are in some cases lower; however, hey are significan a he 10% level in 5 of 7 counries. The correlaion in mos cases is slighly higher wih consumer confidence indicaor which could be explained by wo poins. Firsly, his indicaor as an aggregae already conains daa from cof, cci, focusing on he consumers expecaions abou heir fuure financial posiion. Secondly, he growh of income is approximaed in he model by he growh of he GDP and he consumer indicaor also conains he daa abou he confidence of consumers in he general economic siuaion in he counry in he nex year (quesion 4), which is oriened owards he aggregae level. Therefore, he slighly bigger difference in he correlaion coefficien could also be explained by he beer maching wih he growh of he GDP. cof However, in he case of he income channel, we are acually ineresed no in he growh of he GDP bu in he growh of consumer incomes. We can see ha he differences in he esimaions of hese wo correlaion coefficiens are mainly small. Thus, I proposed o use for he approximaion of he income channel he daa from cci, which are, according o he formulaion of quesion, direcly focused on consumers expecaions abou heir fuure financial siuaion. Alhough he correlaion coefficien is smaller, using daa from for he income channel and daa from oher confidence quesions focused on he predicive power hrough he second channel separaely could increase he predicive power of confidence daa in he model. cci Table 1 Conemporaneous correlaion beween income growh and confidence daa Counry cof cci Ausria (0.000) 0.1 (0.061) Belgium (0.000) 0.31 (0.004) Bulgaria (0.000) (0.000) Croaia (0.00) 0.49 (0.001) Cyprus (0.000) (0.019) Czech epublic (0.000) 0.37 (0.003) Denmark 0.37 (0.003) (0.095) Esonia 0.18 (0.048) (0.098) Finland 0.63 (0.000) (0.000) France (0.000) (0.000) Germany (0.001) 0.33 (0.034) Greece (0.000) 0.69 (0.000) Hungary (0.000) 0.58 (0.000) Ialy (0.000) (0.000) Lavia 0.63 (0.000) 0.60 (0.000) Lihuania (0.000) (0.000) Luxembourg 0.41 (0.074) (0.69) Mala 0.1 (0.111) 0.31 (0.097) Neherlands (0.000) (0.000) Poland 0. (0.104) 0.66 (0.049) Porugal 0.61 (0.000) (0.000) omania (0.000) (0.00) Slovakia (0.001) 0.59 (0.034) Slovenia (0.000) (0.000) Spain (0.000) (0.000) Sweden (0.000) 0.10 (0.364) Unied Kingdom (0.000) (0.000) p-values are repored in he parenheses; boldface indicaes significance a he 10% level. The esimaions of correlaion coefficiens for he second channel WS beween consumpion growh, confidence quesions 9 and 1 and he confidence indicaor are presened in Table. I can see ha he correlaion is significan a he 10% level in 5 of 7 counries in he case of consumer confidence indicaor cof, in counries for he confidence quesion 9 and in 11 counries for he confidence quesion cci 1. In every counry, excep for Ausria and Luxembourg, here is a leas one of hese correlaions wih cci

7 J. Závacká Alernaive roles of consumer confidence in forecasing consumpion: Evidence from European counries 139 confidence quesion cci 9 or cci 1 ha is saisically significan (10% level). Table Conemporaneous correlaion beween growh of consumpion and confidence daa Counry cof cci9 cci1 Ausria 0.11 (0.38) (0.681) (0.463) Belgium 0.47 (0.04) (0.736) (0.646) Bulgaria 0.55 (0.000) (0.438) (0.60) Croaia (0.000) 0.51 (0.000) (0.15) Cyprus 0.5 (0.087) (0.000) 0.34 (0.01) Czech epublic (0.000) 0.38 (0.035) (0.085) Denmark 0.83 (0.010) 0.87 (0.008) (0.8) Esonia 0.30 (0.036) (0.001) (0.005) Finland 0.67 (0.016) (0.00) (0.197) France (0.001) (0.004) (0.337) Germany (0.074) 0.34 (0.033) (0.966) Greece (0.000) (0.000) 0.83 (0.009) Hungary 0.70 (0.000) (0.000) (0.000) Ialy 0.56 (0.000) 0.43 (0.031) (0.000) Lavia (0.000) (0.003) (0.458) Lihuania 0.67 (0.000) (0.015) (0.007) Luxembourg (0.0) (0.637) (0.914) Mala 0.53 (0.067) (0.64) 0.33 (0.015) Neherlands 0.63 (0.000) (0.000) (0.00) Poland (0.000) 0.56 (0.059) 0.08 (0.840) Porugal (0.000) (0.000) (0.764) omania (0.011) (0.003) 0.09 (0.87) Slovakia (0.000) (0.000) (0.73) Slovenia 0.61 (0.00) (0.00) (0.156) Spain (0.000) (0.000) (0.000) Sweden 0.6 (0.018) 0.43 (0.000) (0.669) Unied (0.000) 0.65 (0.000) 0.39 (0.030) Kingdom p-values are repored in he parenheses; boldface indicaes significance a he 10% level. Alhough he correlaion coefficien is more ofen saisically significan in he case of he confidence indicaor cof, i is no always higher. The explanaion for his could be similar o he case of he correlaion wih he growh of income. Again, he confidence indicaor by iself is an aggregaion of he daa from four confidence quesions, which was proposed o have predicive power for consumpion growh. Thus, he predicive power of his aggregae is bigger han he predicive power from he daa of only one quesion. Noneheless, we can see from Table ha he correlaion is also high for cci and cci 1. We can 9 herefore use he combinaion of he daa from hese wo quesions o approximae he influence hrough he second channel and separae he confidence informaion ino hese wo channels. 3.3 egression resuls Because of he possible auocorrelaion among he residuals, I used he robus mehod o esimae all he models, M0 (7), M1 (5) and M (6). I calculaed he increase in (incremenal ) caused by adding confidence quesions or he confidence indicaor o he model in he form of M0, M1 and M. Furher, I esed he join significance of he confidence quesions or he confidence indicaor in he model wih he F-es (his corresponds o he saisical significance of he incremenal ). The regression resuls are repored in Table 3. We can see from he resuls ha he incremenal from he models M1 and M are in mos cases (for M1 all excep Hungary, Lavia and Sweden; for M all cases) higher han he incremenal from M0. This means ha he predicive power of confidence daa is greaer when using he daa from he confidence quesions separaely han when using he aggregaed consumer confidence indicaor This is in line wih he resuls ha Dreger and Kholodilin (013) obained on an empirical basis. According o he join saisical significance, i is more appropriae o use M0 han M1 or M only in he case of Hungary. The spurious resuls abou he predicive power of confidence indicaors for forecasing consumpion could be suppored by he fac ha he confidence indicaors were, in he case of model M0, joinly saisically significan a he 10% level only in 7 of he 7 counries. This could be increased o 13 of he 7 counries by using model M1 and o 16 by using model M. Firsly, his suppors he hypohesis ha confidence daa have predicive power o forecas consumpion spending. Secondly, i suppors he idea ha he daa from confidence quesions could be separaed ino wo influence channels he income channel and he willingness-o-spend channel in he model. By his separaion no only can higher predicive power of confidence daa be achieved bu also he heoreical explanaion of heir use in he model can be more undersandable. cof.

8 140 Ekonomická revue Cenral European eview of Economic Issues 19, 016 Table 3 Predicive power of confidence daa: incremenal Counry M0 M1 M Ausria 0.09 (0.684) (0.417) (0.47) Belgium 0.05 (0.544) (0.30) (0.386) Bulgaria (0.687) (0.453) (0.47) Croaia (0.1) (0.084) 0.15 (0.09) Cyprus (0.944) (0.08) (0.060) Czech epublic 0.07 (0.10) (0.10) (0.5) 0.10 Denmark (0.009) 0.00 (0.031) 0.17 (0.059) Esonia (0.056) (0.149) (0.006) Finland (0.467) 0.61 (0.000) (0.000) France (0.160) (0.049) (0.067) Germany (0.319) 0.1 (0.041) (0.081) Greece (0.568) 0.18 (0.60) (0.78) Hungary (0.009) (0.649) (0.0) Ialy (0.16) (0.16) (0.144) Lavia (0.330) (0.959) (0.90) Lihuania (0.60) (0.3) (0.) Luxembourg (0.77) (0.101) 0.51 (0.006) Mala (0.616) (0.06) (0.090) Neherlands (0.149) (0.008) (0.011) Poland (0.868) 0.67 (0.008) 0.30 (0.006) Porugal (0.005) (0.00) (0.000) omania (0.198) 0.93 (0.09) (0.39) Slovakia (0.15) (0.00) 0.33(0.003) Slovenia (0.393) (0.136) (0.11) Spain (0.004) 0.6 (0.000) 0.31 (0.000) Sweden 0.13 (0.010) (0.075) 0.9 (0.097) Unied 0.06 (0.041) (0.000) 0.18 (0.000) Kingdom p-values of he join significance of all confidence indicaors in he model (from F-es) are repored in he parenheses. The boldface indicaes significance a he 10% level. obus esimaions were used. 4. Conclusion The main goal of his aricle was o exend he heoreical suppor for using confidence daa o forecas consumpion spending and o propose an alernaive use of confidence daa o increase heir predicive power for consumpion spending. I sae ha consumers decision abou heir consumpion is deermined no only by heir expeced fuure income bu also by oher facors like uncerainy abou he fuure, he influence from heir neighbours and so on. In his way I disinguish beween wo channels for he predicive power of confidence daa for consumpion. The firs, he income channel, represens he abiliy of confidence daa o predic he expecaions abou fuure income. The second channel is called willingness o spend and represens he moivaion of consumers o spend his expeced income. Through his channel he confidence daa are used for he approximaion of consumers inenions regarding consumpion. By emphasizing he difference beween hese wo channels, I proposed a model for consumpion growh based on he daa from various confidence quesions raher han one aggregaed confidence indicaor. By using he daa from he Join Harmonised EU Programme of Business and Consumer Surveys, I chose he daa from separae survey quesions for boh domains and esimaed he growh of consumpion. For comparison, I also esimaed he model of consumpion growh based only on pas values of consumpion growh and income growh and anoher exended o include he consumer confidence indicaor (he official aggregaed indicaor for confidence). Using he daa from 7 EU member counries, I confirmed he predicive power of confidence daa in 17 counries. This predicive power was increased by using separae confidence quesions wih he excepion of one case. The alernaive use of confidence daa no only could improve consumpion growh predicion bu could also serve o invesigae he impac of oher facors on consumpion spending or explain he excess sensiiviy of consumpion o income growh in some periods. In addiion, using confidence quesions separaely raher han in heir aggregaed form could increase he heoreical inerpreaion for is use. The confidence daa could be disinguished furher among hose ha represen informaion abou consumers expecaions (abou heir fuure income, he general economic siuaion, he fuure price or unemploymen evolvemen) and indicaors ha are more informaive abou consumers inenions and behaviour (planned major purchases, savings, building a house, ec.). A fuure more deailed sudy of hese confidence daa could be considered. eferences AL-EYD, A., BAELL,., DAVIS, E.P. (009). Consumer confidence indices and shor-erm forecasing of consumpion. The Mancheser School 77: ACEMOGLU, D., SCOTT, A. (1994). Consumer confidence and raional expecaions: are agens' beliefs consisen wih he heory? The Economic Journal 4: BATCHELO,., DUA, P. (1998). Improving macroeconomic forecass: The role of consumer confidence. Inernaional Journal of Forecasing 14:

9 J. Závacká Alernaive roles of consumer confidence in forecasing consumpion: Evidence from European counries 141 BAM, J., LUDVIGSON, S.C. (1998). Does consumer confidence forecas household expendiure? A senimen index horse race. Economic Policy eview 4: CAOLL, CH.D., HALL,.E., ZELDES, S.P. (199). The buffer-sock heory of saving: Some macroeconomic evidence. Brookings papers on economic aciviy : CAOLL, CH.D., FUHE, J.C., WILCOX, D.W. (1994). Does consumer senimen forecas household spending? If so, why? The American Economic eview 84: COTSOMITIS, J.A., KWAN, A.C.C. (006). Can consumer confidence forecas household spending? Evidence from he European commission business and consumer surveys. Souhern Economic Journal 7: DEES, S., BINCA, P.S. (013). Consumer confidence as a predicor of consumpion spending: Evidence for he Unied Saes and he euro area. Inernaional Economics 134: DEGE, CH., KHOLODILIN, K.A. (013). Forecasing privae consumpion by consumer surveys. Journal of Forecasing 3: DUESENBEY, J.S. (1949). Income, saving, and he heory of consumer behavior. Cambridge, MA: Harvard Universiy Press. FLAVIN, M. (1993). The excess smoohness of consumpion: idenificaion and inerpreaion. The eview of Economic Sudies 60: FIEDMAN, M. (1957). A heory of he Consumpion Funcion: A Sudy by he Naional Bureau of Economic esearch. Princeon: Princeon Universiy Press. HALL,.E. (1978). Sochasic implicaions of he life cycle-permanen income hypohesis: Theory and evidence. Journal of Poliical Economy 86: HAUGH, D.L. (005). The influence of consumer confidence and sock prices on he Unied Saes business cycle, CAMA Working Papers, No. 3/005. Ausralia: ANU Universiy. HOWEY, E.P. (001). The predicive power of he index of consumer senimen, Brookings Papers on Economic Aciviy 1: LAHII, K., MONOKOUSSOS, G., ZHAO Y. (01). Forecasing consumpion: The role of consumer confidence in real ime wih many predicors. Discussion papers 1-0: 1 9. New York: Universiy a Albany. SOULELES, N.S. (001). Consumer senimen: Is raionaliy and usefulness in forecasing expendiureevidence from he Michigan micro daa. Naional bureau of economic research, No. W8410. Cambridge, MA: Naional Bureau of Economic esearch. TAYLO, K., MCNABB,. (007). Business cycles and he role of confidence: Evidence for Europe. Oxford Bullein of Economics and Saisics 69: THOOP, A.W. (199). Consumer senimen: Is causes and effecs. Economic eview-federal eserve Bank of San Francisco 1: Addiional sources EC (EUOPEAN COMMISSION) (016). The join harmonised EU programme of business and consumer surveys Available a www: <hp://ec.europa. eu/economy_finance/db_indicaors/surveys/mehod_g uides/index_en.hm>. Appendix The Join Harmonized Consumer Survey The measuremen of confidence indices in he EU follows he Join Harmonised EU Programme of Business and Consumer Surveys (016) designed by he European Commission. The consumer survey is based on 1 monhly quesions and 3 quarerly quesions, organized around 4 opics: he households financial siuaion, he general economic siuaion, savings and inenions wih regard o major purchases. All he quesions are qualiaive, usually wih 6 answer alernaives (+ +, +, =,,, N). For each quesion he aggregae balance is calculaed. The aggregae balance is he difference beween posiive and negaive answers expressed in he percenage poins of he oal answers, wih double weighs on he exremes (weigh 1: + +, ; weigh 1/: +, ; weigh 0: =, N). The daa are also available in a seasonally adjused form. Monhly quesions: Q1: How has he financial siuaion of your household changed over he las 1 monhs? I has + + go a lo beer; + go a lile beer; = sayed he same; go a lile worse; go a lo worse; N don know. Q: How do you expec he financial posiion of your household o change over he nex 1 monhs? I will ge a lo beer; + ge a lile beer; = say he same; ge a lile worse; ge a lo worse; N don know.

10 14 Ekonomická revue Cenral European eview of Economic Issues 19, 016 Q3: How do you hink he general economic siuaion in he counry has changed over he pas 1 monhs? I has go a lo beer; + go a lile beer; = sayed he same; go a lile worse; go a lo worse; N don know. Q4: How do you expec he general economic siuaion in his counry o develop over he nex 1 monhs? I will ge a lo beer; + ge a lile beer; = say he same; ge a lile worse; ge a lo worse; N don know. Q5: How do you hink ha consumer prices have developed over he las 1 monhs? They have + + risen a lo; + risen moderaely; = risen slighly; sayed abou he same; fallen; N don know. Q51: If quesion 5 was answered by 1,, 3 or 5: By how many per cen do you hink ha consumer prices have gone up/down over he pas 1 monhs? (Please give a single figure esimae). Consumer prices have increased by... % / decreased by... %. Q6: By comparison wih he pas 1 monhs, how do you expec ha consumer prices will develop in he nex 1 monhs? They will + + increase more rapidly; + increase a he same rae; = increase a a slower rae; say abou he same; fall; N don know. Q61: If quesion 6 was answered by 1,, 3 or 5: By how many per cen do you expec consumer prices o go up/down change in he nex 1 monhs? (Please give a single figure esimae). Consumer prices will increase by... % / decrease by... %. Q7: How do you expec he number of people unemployed in his counry o change over he nex 1 monhs? The number will increase sharply; + increase slighly; = remain he same; fall slighly; fall sharply; N don know. Q8: In view of he general economic siuaion, do you hink ha now i is he righ momen for people o make major purchases, such as furniure, elecrical/elecronic devices, ec.? + + yes, i is he righ momen now; = i is neiher he righ momen nor he wrong momen; no, i is no he righ momen now; N don know. Q9: Compared o he pas 1 monhs, do you expec o spend more or less money on major purchases (furniure, elecrical/elecronic devices, ec.) over he nex 1 monhs? I will spend + + much more; + a lile more; = abou he same; a lile less; much less; N don know. Q10: In view of he general economic siuaion, do you hink ha now is...? + + a very good momen o save; + a fairly good momen o save; no a good momen o save; a very bad momen o save; N don know. Q11: Over he nex 1 monhs, how likely is i ha you will save any money? + + very likely; + fairly likely; no likely; no a all likely; N don know. Q1: Which of hese saemens bes describes he curren financial siuaion of your household? + + we are saving a lo; + we are saving a lile; = we are jus managing o make ends mee on our income; we are having o draw on our savings; we are running ino deb; N don know. Quarerly quesions (January, April, July and Ocober): Q13: How likely are you o buy a car over he nex 1 monhs? + + very likely; + fairly likely; no likely; no a all likely; N don know. Q14: Are you planning o buy or build a home over he nex 1 monhs (o live in yourself, for a member of your family, as a holiday home, o le, ec.)? + + yes, definiely; + possibly; probably no; no; N don know. Q15: How likely are you o spend any large sums of money on home improvemens or renovaions over he nex 1 monhs? + + very likely; + fairly likely; no likely; no a all likely; N don know. The European Commission also publishes a monhly composie indicaor he consumer confidence indicaor. This indicaor is calculaed as a simple arihmeic average of seasonally adjused balances (expressed in percenage poins) of he answers o quesion Q, Q4, Q7 and Q11.