The current issue and full text archive of this journal is available at www.emeraldinsight.com/1086-7376.

htm

Stock market and aggregate consumption asymmetry: evidence from Malaysia

Mansor H. Ibrahim and Muzafar Shah Habibullah
Faculty of Economics and Management, University Putra Malaysia, Serdang, Malaysia
Abstract
Purpose The purpose of this paper is to analyze the inuences of real share prices on aggregate consumption for Malaysia with the focus on whether there is asymmetry in the long-run relation of the two variables. Design/methodology/approach The paper species aggregate consumption to depend on real income and real share prices. Alternatively, imposing long-run budget constraint, the paper species the relation between aggregate consumption and real share prices as ratio to real income. Then, it applies an asymmetric cointegration and error correction modeling. Findings The cointegration tests indicate the presence of a long-run relation between consumption-income ratio and share price-income ratio. More interestingly, while changes in share prices exert short-run causal inuences on Malaysias private consumption, evidence is found for the adjustments of consumption income ratio to the long-run equilibrium path only when it is above its long-run value. The paper interprets the nding as suggesting downward revisions in the consumption patterns when there are adverse shocks in share prices and, accordingly, supports the existence of especially negative wealth effect for Malaysia. Research limitations/implications Owing to data limitations, the paper relies on aggregate consumption and aggregate income data. It acknowledges that the sum of non-durable consumption and ow-of-services from durable purchases and labor income are more appropriate measures of, respectively, consumption and real income. Originality/value The ndings have important implications for understanding consumption behavior in a developing country and can provide insight on the effectiveness of monetary policy. Keywords Consumption, Stock markets, Malaysia, Share prices, Modelling Paper type Research paper

Aggregate consumption asymmetry 19

1. Introduction Rapid growth and increasing volatility of stock markets in recent years and their relations to the economy have captured a great deal of attention from both nancial economists and policymakers. While the attention has been predominantly on causal relations between share prices and such macroeconomic variables as real output, exchange rates, and money supply, several recent studies have also focused on the role the stock markets play in private consumption behavior. Theoretically, on the basis of the life-cycle hypothesis, movements in stock prices may exert impacts on private consumption through what is termed as the wealth effect. According to the hypothesis, consumption is proportional to lifetime resources, i.e. expected labor income over the lifetime and value of presently held assets. Thus, as noted by Liu and Shu (2004), to the extent that the stock price movements reect changes in the asset values and accordingly wealth, they should anticipate changes in consumption.

Studies in Economics and Finance Vol. 27 No. 1, 2010 pp. 19-29 q Emerald Group Publishing Limited 1086-7376 DOI 10.1108/10867371011022957

SEF 27,1

20

While the life-cycle hypothesis predicts a direct relation between consumption and wealth, the effects of share prices on consumption are not unambiguous. Share price movements reect changes in both asset returns and volatility or risk. These, as argued by Bonser-Neal and Dewenter (1999), lead to indeterminate relations between share prices and savings or consumption. The effect of asset returns depends on the relative strength of the well-known substitution effect and income effect. The increase in stock returns makes current consumption less attractive (substitution effect) but, as mentioned, it also raises wealth and accordingly consumption (income effect). Meanwhile, whether increasing uncertainty or risk affects savings and thus consumption positively or negatively depends crucially on the coefcient of relative risk aversion (Rothschild and Stiglitz, 1971). Thus, with these indeterminate theoretical predictions, the relation between share prices and consumption is open to empirical investigation. Empirically, many studies have evaluated the stock market wealth effect for advanced markets using either household data or aggregate time series data. The results for the US seem to indicate at best its modest inuence on consumption, as reected by the estimated marginal propensity to consume out of stock market wealth to be around 0.03 (see Poterba, 2000 and references therein). Boone et al. (1998) further note that the stock market wealth effect tends to be weaker in especially continental European countries. Recently, Liu and Shu (2004) evaluate the causal link between real private consumption and real share prices for the advanced markets of Asia Hong Kong, Japan, Singapore, South Korea, and Taiwan. The long-run relationship between the two variables is documented for all countries except South Korea. While documented causal patterns differ across countries, the inuences of the stock markets tend to be more visible in the short run than in the long-run. Cutler (2005) estimates the marginal propensity to consume out of nancial wealth for Hong Kong to be around 0.02 and 0.04, which is in line with those estimated for developed countries. In a related study on 16 emerging markets by Bonser-Neal and Dewenter (1999), however, the relation between nancial development and saving is not robust depending on the inclusion of outlying countries. Some researchers have also emphasized consumption asymmetry, which may arise from loss aversion and capital market imperfections or liquidity constraints (Patterson, 1993; Shirvani and Wilbratte, 2000; Apergis and Miller, 2006). More specically, using an error correction model that allows for short run asymmetric responses of consumption to changes in the stock market prices, Apergis and Miller (2006) document supportive evidence for stronger effect of bad news (i.e. negative stock returns) as compared to that of good news (i.e. positive stock returns). Stevens (2004) contends that heightened return volatility during market downturns may lead to asymmetric consumption adjustments. As he argues, the increasing information asymmetry and subsequent decline in lending and borrowing during periods of market downturns creates hysteresis in consumer spending. Examining the US data, he nds quicker consumption adjustments to its long-run value during boom periods. The present paper attempts to contribute to existing literature by examining empirically the relation between aggregate private consumption and stock prices for an emerging market, Malaysia. More specically, in line with the recent focus on consumption asymmetry, we evaluate whether Malaysias private consumption adjusts asymmetrically towards its long-run value via threshold and momentum-threshold cointegration and error-correction modeling due to Enders and Siklos (2001). As in other emerging markets, Malaysia experienced drastic increase in its share prices during the

rst-half of the 1990s. The Kuala Lumpur Composite Index (KLCI) increased by more than 100 percent from 1990 to 1996. However, the index succumbed to the Asian crisis, recording a drastic drop from 1,238 points in 1996 to 586.1 in 1998. Post crisis, the market has again exhibited an uptrend pattern but faced various global uncertainties. Against this background, it would be interesting to see whether stock market swings over the past decades have any impact on private consumption. We believe that the ndings have important implications on stabilization policies and would also be relevant for other emerging Asian countries that have witnessed similar patterns of stock price uctuations. The rest of the paper is structured as follows. In Section 2, we outline the empirical approach. Data and estimation results are presented in Section 3. Finally, Section 4 summarizes the main nding and provides concluding remarks. 2. Empirical approach 2.1 Long-run relation The relation between aggregate consumption and the stock market is normally founded on the life-cycle hypothesis of Ando and Modigliani (1963), which postulates a proportional relation between consumption and expected lifetime resources from labor income and wealth. More specically, given lifetime budget constraint and the assumptions of no bequests and time-separable logarithmic utility function, the optimal level of consumption depends linearly on the sum of initial wealth and total labor income over the lifetime (Starr-McCluer, 2002). Given that shares are a part of wealth, the link between consumption and share prices can thus be established. To begin, the following standard consumption is specied: Model 1 : ct a0 a 1 y t a2 s t 1 t 1

Aggregate consumption asymmetry 21

where ct, yt, and st denote real aggregate consumption, real income, and real share prices, respectively. All variables are in natural logarithm. As mentioned in Byrne and Davis (2003), expressing the variables in natural logarithm instead of level form is necessary to ensure difference stationarity of income. The positive coefcient of real income, i.e. a2 . 0, is generally taken as supportive evidence for the long-run stock market wealth effect. In the present analysis, we re-specify equation (1) to conform to the consumers lifetime constraint along the line of Barrel and Davis (2007). As noted by Barrel and Davis (2007), in steady state, consumption to income ratio is linked to wealth to income ratio. We believe that this is more theoretically appealing as equation (1) is generally referred as the long-run consumption behavior. In the light of this, we have: Model 2 : c 2 yt b0 b1 s 2 yt 1t 2

It is worth mentioning that specication equation (2) is also in line with Boone et al. (1998) and Hall and Patterson (1992). More specically, the use of consumption income and share price income ratios is implied by the presence of target consumption and asset levels embodying a unit income elasticity. It should be noted that Model (2) has a statistical advantage in that, by reducing the estimated dimension, robust evidence can be uncovered (Maki and Kitasaka, 2006). Moreover, it avoids the need to approximate higher frequency population data using the available yearly population estimates, the process that may induce measurement

SEF 27,1

error and bias in the estimation. Based on these considerations, we take Model (2) as our preferred model. Still, we also estimate Model (1) for comparative purposes. 2.2 Threshold cointegration tests Econometrically, assume that the variables in the two equations are non-stationary integrated of Order 1 or I(1), the presence of their long-run relations or validity of equations (1) and (2) as long-run or cointegrating equations requires stationarity of the error term. In testing for cointegration, we adopt the threshold cointegration approach due to Enders and Siklos (2001). The approach is in line with possible asymmetric consumption dynamics in moving towards the long-run path as specied in equations (1) and (2). As highlighted by Apergis and Miller (2006), consumption may respond asymmetrically to good and bad news due to loss aversion and capital market imperfections. However, in their analysis, allowance is made only for short-run asymmetric consumption adjustments. In the present analysis, we evaluate asymmetric adjustments of consumption towards its long-run value via an error-correction mechanism. Basically, the Enders and Siklos (2001) asymmetric cointegration test is based on the following equation: D1t r1 I t 1t21 r2 1 2 I t 1t21
k X i1

22

li D1t2i ut

where D is the rst-difference operator and It is the Heaviside indicator specied to depend on either the level or changes of error terms, respectively, as in equations (4) and (5) below: ( 1 if 1t21 $ 0 It 4 0 if 1t21 , 0 ( 1 if D1t21 $ 0 It 5 0 if D1t21 , 0 The lagged rst-differenced terms are included to whiten the noise process. Accordingly, the order k is specied such that the error terms are serially uncorrelated. In equation (3), the change in 1t is allowed to depend on whether the last-period error term, normally termed shock, is positive or negative. Equations (3) and (4) are termed as the threshold autoregressive (TAR) model. Meanwhile, if it is believed that the shock has a tendency to show greater momentum in moving in one direction than in the other, specication equation (5) would be more appropriate as the indicator function (Enders and Dibooglu, 2001). Equation (3) with indicator function equation (5) is referred as the momentum-threshold autoregressive (M-TAR) model. The stationarity of 1t is achieved when 2 2 , (r1, r2) , 0. The test for this stationarity or cointegration for both TAR and M-TAR models is carried out using the F-test for the null-hypothesis of no cointegration, H0: r1 r2 0. The F-test statistics, however, has a non-standard distribution and is denoted as F by Enders and Siklos (2001). In the presence of asymmetric cointegration, the null-hypothesis H0: r1 r2 can be tested using the standard F-statistics. The evidence in support of asymmetric adjustment is indicated when both H0: r1 r2 0 and H0: r1 r2 are rejected.

2.3 Threshold error-correction models In the presence of asymmetric cointegrating relationship, we employ threshold error-correction models to capture asymmetric consumption dynamics. To illustrate, the threshold error-correction model corresponding to Model (2) is written as: Dc 2 yt f0 f1 I t 1t21 f2 1 2 I t 1t21 l Z t
k X i1

Aggregate consumption asymmetry 23

u1 Dc 2 yt2i

k X i1

u2 Ds 2 yt2i v2t

where Z is a vector of exogenous dummy variables, which may include the Asian crisis dummy and seasonal dummies, and other variables are as dened above. The parameter f1 captures the adjustment of consumption-income ratio when it is above its long-run value as given by the current level of stock market prices. Meanwhile, parameter f2 measures the response of consumption-income ratio when it is below its long-run value. Finally, the signicance of at least one u1, . . . , uk indicates short-run inuences of stock prices on consumption. 3. Results We employ quarterly data from 1991.Q1 to 2007.Q2. The consumption variable is the nal consumption expenditure of the private sector. Gross domestic products (GDP) are used to represent the level of income. The end-of-the-quarter value of the KLCI is used to represent share prices. These data are deated using the consumer price index to obtain their corresponding real gures. All data are obtained from Monthly Statistical Bulletin (various issues) published by Bank Negara Malaysia, i.e. Malaysias Central Bank. These variables are rst checked for their unit root properties using the standard augmented Dickey-Fuller (ADF) and Kwiatkowski-Phillips-Schmidt-Shin (KPSS) unit root tests. While the ADF test has non-stationarity null-hypothesis, the KPSS test states stationarity as the null-hypothesis. The results of these tests are given in Table I. As may be noted from the table, all variables can be generally classied as non-stationary in level but stationary in rst difference. That is, they belong to an I(1) process. 3.1 Long-run relation Given that all variables are I(1), we proceed to TAR and M-TAR cointegration tests. For comparison, we also conduct the traditional two-step cointegration test due to Engle and Granger (1987). Table II presents the results of these tests. The Engle-Granger cointegration test reveals no long-run relation between consumption
Level Variables c y s c2y c2s ADF 20.904 21.862 22.785 21.030 22.607 KPSS 0.157 * * 0.133 * * * 0.114 0.290 * 0.119 * * * ADF 23.914 * * 23.899 * * 24.257 * 23.221 * * * 24.247 * First difference KPSS 0.103 0.205 * * 0.073 0.085 0.058

Note: The signicance of *1, * *5, and * * *10 percent, respectively

Table I. ADF and KPSS unit root tests

SEF 27,1
Model EG

TAR F-stat H0: r1 r2 0

F-stat H0: r1 r2 3.750 4.016 2.799

M-TAR F-stat F-stat H0: r1 r2 0 H0: r1 r2 5.599 6.452 7.91 6.88 6.39 5.32 4.609 4.016 2.799

24

(1) 21.399 3.861 (2) 22.792 5.963 Critical values of TAR and M-TAR statistics (1) 5% 7.79 10% 5.71 (2) 5% 6.33 10% 5.22

Table II. Cointegration tests

Note: The values correspond to n 50 and lag 4 Source: TAR and M-TAR critical values for Model (1) are from Enders and Dibooglu (2001) while corresponding values for Model (2) are from Enders and Siklos (2001)

and its determinants for both models. Similarly, we nd no evidence for the long-run relation between aggregate private consumption, real income, and real share prices using TAR and M-TAR cointegration test. Interestingly, the evidence for asymmetric cointegration between the variables is uncovered when they are expressed as ratios of income. Moreover, the null-hypothesis H0: r1 r2 is rejected at 10 and 5 percent, respectively, for TAR and M-TAR models. These results indicate that consumption-income ratio adjusts asymmetrically depending on whether it is above or below its long-run value. Statistically, they lend support to the use of asymmetric cointegration test to uncover the long-run relation between consumption and stock market wealth. In other words, imposing symmetric adjustments as in the Engle-Granger test can be misleading. The long-run relation between consumption-income ratio and share price-income ratio is estimated to be: c 2 yt 20:254 0:118s 2 yt : 7

It should be noted that the coefcient of share price income ratio is signicant at even 1 percent signicant level. We also estimate the long-run relation by including the Asian crisis dummy variable. However, it is insignicant and, accordingly, omitted from the regression. As will be noted later, the Asian crisis does not seem to affect consumption even in the short-run. From equation (7), the estimated stock market wealth elasticity of consumption is estimated to be 0.12 percent. Thus, in the Malaysian context, consumers are quite sensitive to changes in real share prices in the long-run, supporting the presence of stock market wealth effect. 3.2 Threshold error-correction models Given the presence of asymmetric cointegration between private consumption expenditure and its determinants, we estimate threshold and momentum-threshold error correction models to assess its dynamics in the short run. In the estimation, we set the maximum lag order of the rst-differenced terms to eight. Then, we trim the lag order down if the last lag is found to be insignicant using 10 percent signicant level. We also include Asian crisis dummy and seasonal dummies in the short-run equations. Tables III and IV present the results, respectively, for TAR and M-TAR models. The models are

 (cy)t = 0.055 0.039Crisist 0.460It e t1 0.111(1It) e t-1 0.213(cy)t1 (0.000) (0.031) (0.005) (0.473) (0.092) 0.033(cy)t2 0.242(cy)t3 + 0.037(sy)t1 0.007(sy)t2 (0.793) (0.009) (0.218) (0.822) 0.016(sy)t3 0.079(sy)t4 (0.594) (0.005) Adjusted-R2 = 0.8408 JB = 1.079 (0.583) RESET = 1.200 (0.279)
20

Aggregate consumption asymmetry 25

AIC = 4.0514 LM(1) = 0.819 (0.366) LM(3) = 4.352 (0.226)

1.6

SC = 3.6016 ARCH(1) = 0.812 (0.367) ARCH(3) = 3.305 (0.347)

1.2 10 0.8 0 0.4 10

0.0

20 98 99 00 01 02 03 04 05 06

0.4 98 99 00 01 02 03 04 05 06

CUSUM

5% significance

CUSUM of squares

5% significance

Note: RESET, Ramseys mis-specication test

Table III. Threshold error correction model

diagnosed such that robust evidence can be drawn. As may be noted from both tables, the estimated models are free from problems of non-normality, misspecication, auto-correlation and autoregressive conditional heteroskedascity (ARCH) effects as indicated by the Jargue-Bera (JB) test for normality, Ramseys test for model misspecication, Breusch-Godfrey Lagrange multiplier (LM) test for serial correlation, and Engles test for ARCH effect. It is also pleased to note that both models are structurally stable as reected by cumulative sum and cumulative sum squared statistics. Note that, based on Akaike information criterion and Schwartz criterion, the momentum-threshold error correction model is preferred to the threshold error correction model. While the estimated coefcients of lagged changes in share price income ratio indicate quite different short-run inuences, the two models suggest quick adjustments of consumption income ratio once it is above its long-run value. Likewise, the two models provide no evidence for consumption income adjustment when it falls below its long-run value. Using the M-TAR model as an illustration for the above-equilibrium case, we have: Dc 2 yt K 2 0:341c 2 yt21 0:254 2 0:118s 2 yt21 ; ^ for 1t21 . 0 8

where K represents other terms (i.e. constant, dummies, and lagged changes in consumption-income and share price income ratios) in equation (6). The estimated

SEF 27,1

(cy)t = 0.070 0.021Crisist 0.341It e t1 0.175(1It) e t-1 0.400(cy)t1 (0.000) (0.247) (0.005) (0.298) (0.003) +0.018(cy)t2 0.188(cy)t3 + 0.094(sy)t1 +0.026(sy)t2 (0.884) (0.038) (0.005) (0.389)

26

0.029(sy)t3 0.093(sy)t4 + 0.056(sy)t5 +0.070(sy)t6 (0.290) (0.001) (0.047) (0.012) Adjusted-R2 = 0.8556 JB = 1.166 (0.558) RESET = 1.555 (0.223)
20

AIC = 4.1791 LM(1) = 0.746 (0.388) LM(3) = 2.304 (0.512)

1.6

SC = 3.6509 ARCH(1) = 1.460 (0.227) ARCH(3) = 4.298 (0.231)

1.2 10 0.8 0 0.4 10

0.0

Table IV. Momentum-threshold error correction model

20 98 99 00 01 02 03 04 05 06

0.4 98 99 00 01 02 03 04 05 06

CUSUM

5% significance

CUSUM of squares

5% significance

error-correction coefcient suggests that 34 percent of last-period deviation of consumption-income ratio from its long-run value will be corrected by consumption-income ratio adjustment. In other words, any shocks including adverse shocks in stock market wealth tend to lead to downward revision in the aggregate consumption. This pattern of adjustment reects the presence of stock market wealth effect for Malaysia. Namely, the income effect from a drop in stock market wealth tends to dominate. As noted, the adjustment of consumption-income ratio is found to be insignicant when it is below its long-run value. Plausibly, since our modeling is not able to distinguish between the channels through which stock market wealth inuences consumption, it may simply reect equal but conicting income and substitution effects. Moreover, viewed together with the signicant adjustments for the aforementioned above-equilibrium case, the results seem to be in line with the asymmetric information content of stock prices for future economic activity arising from business cycle asymmetries (Neftci, 1984; Domian and Louton, 1997). More specically, it has been noted that stock price increase does not necessarily anticipate higher level of future real activity. By contrast, stock price decline tends to predict future recession reasonably well. In the case of Malaysia, Ibrahim (1996) documents evidence for strong causal link from market downturns to real output. Meanwhile, the link between market upturns and real activity is relatively weak. Finally, despite different adjustment patterns to the long-run equilibrium, share prices do exert short run inuences on private consumption behavior. From our

preferred model, the M-TAR error correction model, the coefcient sum of lagged changes in share price income ratio is positive and signicant. Indeed, there is quite a long lagged impact of changes in share prices, which is found to be up to six quarters. It should also be noted that the crisis does not seem to signicantly affect aggregate consumption. Perhaps, the drop in share prices during the Asian crisis has already picked up the adjustments in consumption and, thus, no independent inuence of the episode other than that from the adverse nancial shock emerges. 4. Conclusion The present paper analyzes empirically the aggregate consumption behavior of an emerging market, Malaysia. The main focus is on whether share price movements play any role in the aggregate consumption behavior in both the short run and long-run. In the analysis, we take note of possible asymmetric adjustments of aggregate consumption to its long-run value by using asymmetric cointegration tests and asymmetric error correction modeling developed by Enders and Siklos (2001). The presence of asymmetric cointegration and asymmetric adjustments of consumption is found. Accordingly, as an empirical contribution, any long-run test on Malaysias consumption function imposing symmetric error correction can be misleading. In the long-run, there is a positive relation between aggregate consumption income ratio and share price income ratio. However, in the face of deviations from the long-run equilibrium, consumption income ratio takes the burden of making adjustment when it is only above its long-run value. Indeed, the adjustment seems to be quick with 34 percent of the deviations corrected the next period. Stated in the context of our analysis, a drop in share prices exerts downward revisions in the aggregate consumption. Meanwhile, no consumption adjustment to its long-run value is observed when the stock prices increase. Finally, consumption tends to react to share price changes in the short run. Thus, our analysis suggests the presence of stock market wealth effect which exerts, as noted, interesting dynamics on the aggregate consumption behavior. From a policy point of view, the traditional Keynesian consumption theory is normally viewed to be relevant for understanding the effectiveness of scal stabilization policies. The life cycle and permanent income hypotheses of consumption have however added the relevance to monetary policy since its conduct may inuence nancial wealth including stock market wealth. In the Malaysian context, the signicant role of monetary policy as captured by changes in either money supply or interest rate in inuencing stock market behavior seems to be robust across studies. These include Habibullah and Baharumshah (1996), Ibrahim (1999), and Ibrahim and Aziz (2003). Thus, to the extent that monetary policy affects share prices, monetary policy can inuence real activity through consumption behavior in the short run. However, the consumption channel may aggravate the impacts of monetary tightening given that consumption is found to respond signicantly to negative shocks in share prices. In other words, this channel of inuence needs to be taken into consideration for assessing potential effects of monetary policy. These ndings and policy implications are, however, not without a caveat. As in other studies on consumption, measurement issues remain the core problem and can be more strenuous for emerging markets. While the theory species consumption as the sum of non-durable consumption and ow-of-services from durable purchases, we use the aggregate consumption instead. Moreover, we employ real GDP to proxy labor income.

Aggregate consumption asymmetry 27

SEF 27,1

Finally, we ignore other forms of wealth such as housing wealth in the analysis. Admittedly, data availability is the main factor restricting richer model specication. While some care should be exercised in interpreting our results, we believe that they add to our understanding on the aggregate consumption behavior for Malaysia.
References Ando, A. and Modigliani, F. (1963), The life cycle hypothesis of saving: aggregate implications and tests, American Economic Review, Vol. 53, pp. 55-84. Apergis, N. and Miller, S.M. (2006), Consumption asymmetry and the stock market: empirical evidence, Economics Letters, Vol. 93, pp. 337-42. Barrel, R. and Davis, E.P. (2007), Financial liberalization, consumption and wealth effects in seven OECD countries, Scottish Journal of Political Economy, Vol. 54 No. 2, pp. 254-67. Bonser-Neal, C. and Dewenter, K.L. (1999), Does nancial market development stimulate savings? Evidence from emerging stock markets, Contemporary Economic Policy, Vol. 17 No. 3, pp. 370-80. Boone, L., Giorno, C. and Richardson, P. (1998), Stock market uctuations and consumption behavior: some recent evidence, OECD Economics Department Working Papers, No. 208, OECD, London. Byrne, J.P. and Davis, E.P. (2003), Disaggregate wealth and aggregate consumption: an investigation of empirical relationships for the G7, Oxford Bulletin of Economics and Statistics, Vol. 65 No. 2, pp. 197-220. Cutler, J. (2005), The relationship between consumption, income, and wealth in Hong Kong, Pacic Economic Review, Vol. 10 No. 2, pp. 217-41. Domian, D.L. and Louton, D.A. (1997), A threshold autoregressive analysis of stock returns and real economic activity, International Review of Economics & Finance, Vol. 6 No. 2, pp. 167-79. Enders, W. and Dibooglu, S. (2001), Long-run purchasing power parity with asymmetric adjustment, Southern Economic Journal, Vol. 68 No. 2, pp. 433-45. Enders, W. and Siklos, P.L. (2001), Cointegration and threshold adjustment, Journal of Business and Economic Statistics, Vol. 19, pp. 166-77. Engle, R.F. and Granger, C.W.J. (1987), Cointegration and error correction: representation, estimation, and testing, Econometrica, Vol. 55, pp. 251-76. Habibullah, M.S. and Baharumshah, A.Z. (1996), Money, output and stock prices in Malaysia: an application of the cointegration tests, International Economic Journal, Vol. 10, pp. 121-30. Hall, S.G. and Patterson, K.D. (1992), A system approach to the relationship between consumption and wealth, Applied Economics, Vol. 24, pp. 1165-71. Ibrahim, M.H. (1996), Can stock market movements predict business cycle? An empirical note for Malaysia, Malaysian Journal of Economic Studies, Vol. 33 No. 1, pp. 1-8. Ibrahim, M.H. (1999), Macroeconomic variables and stock prices in Malaysia: an empirical analysis, Asian Economic Journal, Vol. 13 No. 2, pp. 219-31. Ibrahim, M.H. and Aziz, H. (2003), Macroeconomic variables and the Malaysian equity market: a view through rolling subsamples, Journal of Economic Studies, Vol. 30, pp. 6-27. Liu, X. and Shu, C. (2004), Consumption and stock markets in Asian economies, International Review of Applied Economics, Vol. 18 No. 4, pp. 483-96.

28

Maki, D. and Kitasaka, S. (2006), The equilibrium relationship among money, income, prices, and interest rates: evidence from a threshold cointegration test, Applied Economics, Vol. 38, pp. 1585-92. Neftci, S. (1984), Are economic time series asymmetric over the business cycles?, Journal of Political Economy, Vol. 92, pp. 307-28. Patterson, K.D. (1993), The impact of credit constraints, interest rates and housing equity withdrawal on the intertemporal pattern of consumption a diagrammatic analysis, Scottish Journal of Political Economy, Vol. 40, pp. 391-1407. Poterba, J.M. (2000), Stock market wealth and consumption, Journal of Economic Perspectives, Vol. 14 No. 2, pp. 99-118. Rothschild, M. and Stiglitz, J.E. (1971), Increasing risk II: its economic consequences, Journal of Economic Theory, March, pp. 66-84. Shirvani, H. and Wilbratte, B. (2000), Does consumption respond more strongly to stock market declines than to increases?, International Economic Journal, Vol. 14, pp. 41-9. Starr-McCluer, M. (2002), Stock market wealth and consumer spending, Economic Inquiry, Vol. 40 No. 1, pp. 69-79. Stevens, L.K. (2004), Aggregate consumption spending, the stock market and asymmetric error correction, Quantitative Finance, Vol. 4, pp. 191-8. Corresponding author Mansor H. Ibrahim can be contacted at: mansorhi@econ.upm.edu.my

Aggregate consumption asymmetry 29

To purchase reprints of this article please e-mail: reprints@emeraldinsight.com Or visit our web site for further details: www.emeraldinsight.com/reprints