Myopia, Liquidity Constraints, and Aggregate Consumption: The Case of Greece

Transcription

1 Volume 7, Number, June 00 Myopia, Liquidiy Consrains, and Aggregae Consumpion: The Case of Greece Konsaninos Drakos In his paper we invesigae he behaviour of aggregae consumpion in Greece. In paricular, we documen he empirical failure of he Life-Cycle/Permanen Income Hypoheses. The empirical findings imply ha his failure is due o he presence of Liquidiy Consrains raher han Myopia. I. Inroducion Modelling consumpion has araced grea ineres, boh on a heoreical as well as an empirical level. Tradiionally, he benchmark heoreical framework for aggregae consumpion has been given by he Life Cycle Hypohesis (LCH) developed by Modigliani and Ando (963) and he Permanen Income Hypohesis (PIH) developed by Friedman (957). A series of empirical papers has documened he basic heoreical framework s inabiliy o accoun for he observed consumpion dynamics (Flavin (98), Abel (988), Campbell and Mankiw (990), Bayoumi (993), Shea (995) o name jus a few). Two explanaions have been posulaed for his failure. One explanaion is ha agens may be unable o borrow and lend freely; ha is hey may be liquidiy-consrained (Zeldes (989)). Anoher explanaion is ha agens are myopic and herefore fail o opimise in an ineremporal framework (Runkle (99)). In he presen paper we explore he dynamic behaviour of aggregae consumpion in Greece. Recen sudies on consumpion in Greece have repored a failure of he LCH/PIH (Peleologos and Georganelis (99), Apergis e al. (000)). In paricular, Apergis e al. (000) focus on he effecs of deregulaion in he Greek moneary secor on aggregae consumpion. Using Granger causaliy ess hey conclude ha moneary decisions significanly affeced consumpion decisions only in he period following he deregulaion. In paricular, he repor a significan increase boh in he shor and he long run sensiiviy of consumpion o money supply changes following he deregulaion, while in conras he corresponding sensiiviies wih respec o income changes have decreased. The presen sudy is closer o he spiri of Peleologos and Georganelis, (99) who explore he validiy of he LCH/PIH by esing wheher consumpion responds o anicipaed changes in income, where anicipaed income is proxied by an ARMA model. They employ aggregae daa Deparmen of Economics, Universiy of Essex, Wivenhoe Park, Colcheser, CO 43 SQ, UK. Tel: , kdrako@essex.ac.uk. The auhor is graeful o an anonymous referee whose insighful commens have improved considerably he paper. Any remaining errors or ambiguiies are he auhor s responsibiliy. 97

2 sampled annually for he period and show ha he LCH/PIH fails since (i) consumpion does respond boh o anicipaed and unanicipaed changes in income, and (ii) oher macroeconomic variables such as inflaion and is variabiliy seem o affec consumpion. The conribuion of he presen paper is wofold. Firs, given ha no sudy has explored he sources of he LCH/PIH rejecion we aemp o shed ligh on his issue by considering failures due o Myopia or Liquidiy Consrains. Second, we exend he sudy of Peleologos and Georganelis, (99) by employing more recen daa (including informaion from he 90s) and also by proxying anicipaed changes in income by linearly projecions on informaion available a he ime of decision making (essenially using Insrumenal Variables) insead of using an ad hoc ARMA model. The remainder of he paper is organised as follows. Secion II reviews he heoreical background. Secion III discusses he economeric mehodology employed. Secion IV presens he daase employed. Secion V discusses he esimaion and he empirical resuls. Finally, Secion VI concludes. II. LCH/PIH, Myopia and Liquidiy Consrains: A Brief Review The LCH/PIH assers ha agens base heir curren consumpion decisions on heir expecaions abou heir lifeime income, raher han heir curren income. This dissociaion beween curren consumpion and curren income coupled wih he Random Walk Hypohesis (Hall (978)) implies ha only surprises in permanen income should affec curren consumpion, once lagged consumpion is conrolled for (Blinder and Deaon (985)). More formally he benchmark model considers a represenaive agen who is assumed o maximise he expeced value of a ime-separable lifeime uiliy funcion. In each period, he agen chooses consumpion ( C ) and porfolio shares ( ) in order o: w j, max subjec o: k T E U + k k = 0 + δ ( C ) M = ( A + k ) w j, + k ( + rj, + k ) + Y+ k C k, A + k + j=, () C 0, () +k A 0 (3) T wih k = 0,..., T, where U ( ) is he one period uiliy funcion, C real consumpion in period, δ is he 98

3 DRAKOS: MYOPIA, LIQUIDITY CONSTRAINTS, AND AGGREGATE CONSUMPTION rae of ime preference. E is he expecaion operaor condiional on informaion available a ime, T is he end of agen s horizon, A is he real end-of-period financial (non-human) wealh of he agen in period (afer receiving income and consuming), is he ex pos real afer-ax reurn on he j h asse beween periods and +, w j, is he share of end-of-period wealh held in he form of asse j, M is he number of available asses, and Y is he real disposable labour income of he agen in period. As discussed in Zeldes (989), analyic soluions o his problem when income is sochasic canno in general be derived. Perurbaion argumens can be used, however, o derive a se of firs-order condiions (Euler equaions) ha are necessary for an opimum. In paricular, hese condiions should guaranee ha an agen should be unable o increase expeced uiliy by reshuffling consumpion across periods. In algebraic form his condiion resuls in: U ( C ) U = E ( C )( + r ) + j, + δ =,..., T, j =,..., M,, where U denoes he parial derivaive of U wih respec o C. If expecaions are raional his leads o: U ( C )( + r ) U + j, = + e j, + ( C )( + δ), (5) where e j, + is an expecaional error orhogonal o elemens belonging o he informaion se a ime. This Euler equaion should hold wih respec o any asse available including he riskless asse. Moving away from he benchmark now, under myopia, consumpion should rack curren income. Therefore, any deviaion from he LCH/PIH due o myopia should be symmeric. Consumpion should exhibi equal sensiiviy o predicable income increases and decreases (Shea (995)). Alonji and Siow (987) have shown ha he presence of liquidiy consrains implies an asymmery in consumpion behaviour. In paricular, liquidiy consrains imply ha during periods of low income agens canno susain heir opimal consumpion plan by borrowing. Thus, since liquidiy consrains impede borrowing bu no saving, consumpion should be more srongly correlaed wih predicable income increases han declines (Shea (995)). III. Economeric Mehodology Following Campbell and Mankiw (990), and Shea (995) one can es he LCH/PIH by running he following OLS regression (all variables are in naural logarihms): (4) r j, 99

4 c = µ + λ y + βr + u, (6) where c is consumpion growh, y difference operaor, is expeced income growh, sands for he r is he expeced real ineres rae and u is a well-behaved disurbance erm. Under he LCH/PIH, predicable income movemens should no affec consumpion, conrolling for he reurn for saving. Thus, under he LCH/PIH, λ should equal zero provided y and r are measured using informaion available a. Under myopia, consumpion should rack income. In oher words, consumpion should respond symmerically o predicable deviaions of income from is mean. Under liquidiy consrains, consumpion should respond more srongly o predicable increases han decreases; liquidiy consrains do no cause he Euler equaion beween adjacen periods o fail if opimal fricionless consumpion growh exceeds expeced income growh (Shea (995)). Therefore, by slighly modifying model 6 one can es for he presence of liquidiy consrains or myopia. In paricular, his can be esed by running he following OLS regression: ( POS ) y + λ ( NEG ) y + βr u c = µ + λ +, (7) where POS is a dummy variable for periods in which y > 0, and NEG is a dummy variable for periods in which y < 0. Under he LCH/PIH boh λ and λ should equal zero. Under myopia he λ s should be posiive, significan and equal. Wih liquidiy consrains, λ should be significanly posiive, and should be significanly greaer han λ. Table A summarises he relevan esable hypoheses in erms of he parameers of Equaion (7), derived under he validiy of alernaive models. IV. Daa Table A Model Tesable Hypoheses LCH/PIH λ = λ = 0 Myopia (i) λ = λ, (ii) λ > 0, λ > 0 Liquidiy Consrains (i) λ > 0, (ii) λ > λ The daase consiss of real disposable income (GDP less axes) and real consumpion a 995 consan prices. The (ex pos) real ineres rae was consruced as he difference beween he spo nominal ineres rae for medium erm deposis (3 o monhs) and he change in he Consumer price Index (CPI). The daa were sampled annually from 960 o 999 and were colleced from he Minisry of Naional Economy of Greece and he IMF s CD-ROM. The following figures depic he ime series plos of real income and real consumpion (lef) and heir firs differences (righ). 00

5 DRAKOS: MYOPIA, LIQUIDITY CONSTRAINTS, AND AGGREGATE CONSUMPTION log(c) log(y) Firs difference in log(c) Firs difference in log(y) The ime series plos of he level of series reflec heir apparen non-saionariy and given heir comovemen one migh srongly expec ha hey are linked by a long-run relaionship (coinegraed). However, he presen analysis is raher more ineresed wih heir shor run behaviour as given by heir firs differences, which effecively also guaranee saionariy. The ime series behaviour of changes in real consumpion seems o rack ha of real income. The sample correlaion beween he wo series is 0.76 (p-value 0.00). V. Esimaion and Empirical Resuls. Esimaion Mehodology As i becomes apparen, he heory involves he predicable componen of income growh and real ineres rae, for which unforunaely, direcly measured daa is no available. Thus, one has o device a way of approximaing he expeced componens of he variables. Following Cambell and Mankiw (990) and Shea (995) a way of circumvening his apparen difficuly is o esimae he parameers of Equaions (6), and (7) by Insrumenal Variables (IV). Following he lieraure, we use various informaion ses as insrumens, such as lagged values of income growh, real ineres rae, and consumpion growh. Essenially, we projec ex pos changes on he informaion se available o consumers a he ime of he decision making and hence exrac he predicable componen.. Empirical Resuls In order o es he validiy of he LCH/PIH we firs esimae model (6) where we resric negaive and posiive income changes o exer a symmeric impac (if any) on consumpion. We esimae he model described in Equaion (6) by (i) OLS and (ii) IV obaining:. In all esimaed models we accoun for he 973 oil shock by he inclusion of an impulse dummy variable.. Uni Roo ess are no repored for space conservaion reasons. However, hey are available upon reques. 0

6 Table Esimaion Resuls For: c = µ + λ y + β r + δ( OIL ) + ε Model (OLS) Model (IV) Model 3 (IV) Model 4 (IV) µ (4.9) *** (.83) (.53) ** (.7) ** λ (8.38) *** (4.09) *** (4.56) *** (4.89) *** β E E 3 (.53) ** (0.0) ( 0.5) ( 0.0) R F-sa *** 4.6 *** 6.44 *** 6.0 *** LM() ARCH() Sargan s Insrumen Validiy Tes Noes: (i) Model is esimaed by OLS, Models -4 Esimaion Mehod Two-Sage Leas Squares, (ii) Model insrumen lis: hree lags of income growh, Model 3 insrumen lis: hree lags of income and ineres rae changes, Model 4 insrumen lis: hree lags of income, ineres rae changes and consumpion growh (iii) Sample: , (iv) Whie Heeroskedasiciy-Consisen Sandard Errors & Covariance, (v) OIL sands for an impulse dummy variable capuring he 973 oil shock, (vi) Two (Three) aserisks denoe significance a he 5% (%) level. The parameer of ineres (λ) when we employ sandard OLS is significanly differen from zero (-sa 8.38) suggesing a value for he uncondiional income elasiciy of consumpion d log( C) ( ε C, Y = ) of This finding implies ha movemens in income affec d log( Y ) consumpion growh. However, in order o es he implicaions of he LCH/PIH one needs o focus on predicable movemens in income. We are enabled o do so by virue of he models esimaed by IV also repored in Table. The IV esimaion resuls sugges ha he elasiciy is beween 0.6 and 0.74 and is significan a all convenional levels. Thus, predicable movemens in real income growh do exhibi significan explanaory power over consumpion growh evidence which is agains he validiy of he LCH/PIH. In order o invesigae he source of he LCH/PIH rejecion, ha is wheher myopia or liquidiy consrains are o blame, we use model (7) where we allow negaive and posiive changes in income growh o exer an asymmeric impac on consumpion. However, before we repor he esimaion resuls one poin is worh menioning. As i becomes apparen from he ime series plos of real consumpion and real income he Greek economy has been on an expansionary rajecory over he pas four decades. This creaes an economeric problem since he acual number of negaive changes in income is very small and migh bias our economeric findings. In fac, real income has shrunk only in six occasions (ou of 39 daa 0

7 DRAKOS: MYOPIA, LIQUIDITY CONSTRAINTS, AND AGGREGATE CONSUMPTION poins, approximaely 5% of he sample size). 3 So, bearing his in mind which migh affec he saisical credibiliy of he analysis, 4 one has o be cauious in he inerpreaion of he empirical resuls. The esimaion yields he following resuls presened in Table. Table Esimaion Resuls For: c = µ + λ( POS ) y + λ ( NEG ) y + βr + δ( OIL ) + ε Model (OLS) Model (IV) Model 3 (IV) Model 4 (IV) µ (.98) *** (.43) (.40) (.) λ (6.60) *** (.37) *** (3.9) *** (4.38) *** λ 0.83 (.36) (0.8) 0.40 (0.59) ( 0.07) β 0.00 (.70) ** 0.E 3 (0.04) 0.76E 3 ( 0.0) 0.508E 4 (0.003) R F-sa 0.7 *** 6.79 *** 8.70 *** 7.94 *** LM() ARCH() Sargan s Insrumen Validiy Tes F-sa H 0 : λ = λ = 0 H : λ 0, λ ***.03 ***.57 ***.39 *** Noes: (i) Model is esimaed by OLS, Models -4 Esimaion Mehod Two-Sage Leas Squares, (ii) Model insrumen lis: hree lags of income growh, Model 3 insrumen lis: hree lags of income and ineres rae changes, Model 4 insrumen lis: hree lags of income, ineres rae changes and consumpion growh (iii) Sample: , (i v) Whie Heeroskedasiciy-Consisen Sandard Errors & Covariance, (v) OIL sands for an impulse dummy variable capuring he 973 oil shock, (vi) Two (Three) aserisks denoe significance a he 5% (%) level. Moving o he hypoheses of ineres, in all cases we are able o rejec he join hypohesis (LCH/PIH) ha predicable negaive and posiive componens of income growh do no affec consumpion growh ( λ = λ 0 ). However, i should be noed ha individually i is = only he coefficien of posiive income growh ha drives he rejecion. This highlighs he problem caused by he relaively infrequen occurrence of negaive income growh in he sample, as discussed above, which resuls o very high sandard errors for he coefficien of 3. For comparison purposes i is worh menioning ha Shea (995) esimaed a similar model using various samples of US consumpion where he fracion of negaive changes in consumpion is beween.6 percen and 4.7 percen. 4. The auhor is graeful o an anonymous referee for poining his ou. 03

8 negaive income growh. In erms of he absolue values of he esimaed parameers consumpion is more responsive o predicable posiive changes in income, finding which is consisen wih he case of liquidiy consrains. Furhermore, he behaviour of consumpion under he case of Myopia is no observed in he sample since he requiremen ha λ > 0, λ > 0 is no saisfied given he saisical insignificance 5 of λ. In conras, he daa seem o be consisen wih he predicions of he case of Liquidiy Consrains. In paricular, he robus finding of a significanly posiive elasiciy of consumpion o posiive predicable changes in income ( λ 0 ) and he finding > of an insignifican sensiiviy o predicable negaive changes in income ( λ 0 ) provide evidence for he presence of Liquidiy Consrains. All in all, he empirical findings poin o a clear rejecion of he LCH/PIH as far as he Greek case is concerned, a conclusion which is in accordance o he resuls repored in Paleologos and Georganelis (99) and Apergis e al. (000). Moreover, we provide furher evidence as o he possible reasons for he LCH/PIH s empirical failure, which is due o Liquidiy Consrains. Given he relaively underdeveloped financial marke in Greece unil he lae 980 s, wih ineres raes being heavily regulaed by moneary auhoriies and also credi allocaion decisions being heavily conrolled, he case for Liquidiy Consrains is even sronger. I was afer he deregulaory acions adoped in 988 which led o a liberalised financial secor. In fac, he presen sudy can be viewed as complemenary o Apergis e al. (000) who explicily sudied he effecs of deregulaion on aggregae consumpion and concluded ha indeed here is a link beween he wo. VI. Conclusion Using aggregae daa on consumpion for Greece he paper invesigaed he reasons ha underlie he rejecion of he LCH/PIH. Allowing for poenial asymmeries in consumpion s response o predicable income, he empirical findings sugges ha he rejecion of he LCH/PIH is driven by he presence of Liquidiy Consrains. Fuure research would be more informaive if micro daa were employed in order o focus on he behaviour of households. However, for he momen, he unavailabiliy of such daase precludes such an economeric analysis. = References Abel, A. (988), Consumpion and Invesmen, NBER Working Paper No 580. Alonji, J., and A. Siow (987), Tesing he Response of Consumpion o Income Changes wih (noisy) Panel Daa, Quarerly Journal of Economics, 0, Ando, A., and F. Modigliani (963), The Life-Cycle Hypohesis of Saving: Aggregae Implicaions and Tess, American Economic Review, 53, Again recall ha his coefficien is esimaed very imprecisely (high sandard errors). 04

9 DRAKOS: MYOPIA, LIQUIDITY CONSTRAINTS, AND AGGREGATE CONSUMPTION Apergis, N., E. Varelas, and K. Velenzas (000), Money Supply, Consumpion and Deregulaion: The Case of Greece, Applied Economics Leers, 7, Bayoumi, T. (993), Financial Deregulaion and Consumpion in he Unied Kingdom, Review of Economics and Saisics, 75, Blinder, A., and A. Deaon (985), The Time Series Consumpion Funcion Revisied, Brookings Papers in Economic Aciviy,, Campbell, J., and G. Mankiw (990), Permanen Income, Curren Income, and Consumpion, Journal of Business Economics and Saisics, 8, Deaon, A. (99), Saving and Liquidiy Consrains, Economerica, 59(5), -48. Flavin, M. (98), The Adjusmen of Consumpion o Changing Expecaions abou Fuure Income, Journal of Poliical Economy, 89, Friedman, M. (957), A Theory of he Consumpion Funcion, Princeon Universiy Press, Princeon. Hall, R. (978), Sochasic Implicaions of he Life Cycle -Permanen Income Hypohesis: Theory and Evidence, Journal of Poliical Economy, 86, Paleologos, J., and S. Georganelis (99), The Raional Expecaions Approach o he Sudy of Consumpion-Income Dynamics: The Case of Greece , Rivisa Inernaionale di Scienze Economiche e Commerciali, 39, Runkle, D. (99), Liquidiy Consrains and he Permanen Income Hypohesis, Journal of Moneary Economics, 7, Shea, J. (995), Myopia, Liquidiy Consrains, and Aggregae Consumpion: A Simple Tes, Journal of Money, Credi, and Banking, 7(3), Whie, H. (980), A Heeroskedasiciy-Consisen Covariance Marix Esimaor and a Direc Tes for Heeroskedasiciy, Economerica, 48(4), Zellner, A. (96), An Efficien Mehod of Esimaing Seemingly Unrelaed Regressions and Tes for Aggregaion Bias, Journal of he American Saisical Associaion, 57, Zeldes, S. (989), Consumpion and Liquidiy Consrains: An Empirical Invesigaion, Journal of Poliical Economy, 97(),