THE BEHAVIOUR OF CONSUMER S EXPENDITURE IN INDIA:

Transcription

1 48 ABSTRACT THE BEHAVIOUR OF CONSUMER S EXPENDITURE IN INDIA: DR.S.LIMBAGOUD* *Professor of Economics, Department of Applied Economics, Telangana University, Nizamabad A.P. The relation between income and consumption has drawn the attention of economists for a long time. However, the study of consumption behavior on theoretical level was first examined by J.M.Keynes. He held the view that current consumption expenditure is a stable function of current income and marginal propensity to consume is less than unity. Since the advent of Keynesian consumption function hypothesis many economists carried out empirical studies and concluded that household consumption was consistent with the proposition of Keynes but the aggregate consumption was not found to be proportional to the level of aggregate income. An attempt is made in the present paper to examine the short and long-run consumption behavior in India during the period to In order to estimate consumption function the distributive lag models have been used and it is found that the short-run MPC was about 0.22 and the long-run MPC was about INTRODUCTION The relation between aggregate consumption and aggregate income is termed as consumption function. It will have a considerable importance in macro economic analysis and policy formulation of the government because household s consumption decisions affect the way the economy as a whole behaves both in the short- run and in the long -run. Household consumption decision is very crucial for short run analysis as consumption expenditure of households largely determines aggregate demand which, in turn, determines the levels of employment and income. At the same time, the long run consumption decision is also crucial as it determines the rate of growth of the economy. The theoretical relationship between consumption and income was first studied in a systematic way by John Maynard Keynes in General Theory of Employment, Interest and Money (1936) and he held that the current consumption expenditure is a highly dependable and stable function of current income that the amount of aggregate consumption mainly depends on the aggregate income. Ever since Keynes explained such kind of relationship between consumption and income, many empirical studies have been taken up to estimate numerical consumption behavior on a short and long term basis. To mention Arthur Smithies, Simon Kuznets, James Duesenberry, Robert Ando, Franco Modigiliani, Brumberg and Milton Friedman have taken up empirical studies on consumption behaviour and suggested alternative ways of interpreting the data on consumption and income.

2 49 LITERATURE REVIEW As mentioned above, the consumption function hypothesis developed by J.M.Keynes stated that current consumption expenditure was highly correlated with current income, the marginal propensity to consume less than average propensity to consume, so the percentage of income saved increased with income. But then a serious conflict of evidence arose. Later empirical studies came to the conclusion that this relationship does not provide a sufficient explanation for the behaviour of aggregate data. The Life Cycle Hypothesis developed by Modigiliani and Brumberg (1954) assumes that consumption makes up a constant part of the present value of the life time income. This theory says that that the propensity to consume is lower with the younger households than with households of older persons who spend their savings after retirement. Aggregate demand thus depends not only on income and wealth, but also on demographic changes. Later Milton Friedman (1957) presented Permanent Income Hypothesis as an extension of the Keynes theory of consumption behavour. His permanent income hypothesis divides consumption and income into permanent and transitory element and includes future income estimations. This theory states that, if an individual considers a change of his income transitory (short run), he has no reason to change his consumption habits. On the other hand, if he finds out that the income change is of permanent nature, he simultaneously adjusts to his consumption. According to Friedman s permanent income hypothesis, the key determinant of consumption is the real wealth of consumer and not his current real disposable income. Friedman s permanent income hypothesis attracted considerable empirical attention. Several experiments such as Modigiliani (1966); Davidson (1978); Ghatak (1998); and Wen-Jen and Hsing, (2005) are famous in literature that focused on the aspect of linearity between income and consumption in different countries. Another theoretical development Duesnburry s relative income hypothesis, which argued that personal consumption is more dependent upon a person s relative income i.e, relative to his/her neighbors and friends. The basic objective of the present paper is to investigate the behavior of consumer s expenditure in India during the period to To examine the same, the Life Cycle Hypothesis developed by Modigiliani and Brumberg is used. Modigiliani and Brumberg related consumption expenditure to demography. According to life cycle hypothesis, the consumption in any period is not the function of current income of that period but of the whole lifetime expected income. Thus, in life cycle hypothesis the individual is assumed to plan a pattern of consumption expenditure based on expected income in his entire lifetime. It is further assumed that individual maintains a more or less constant level of consumption. However, this level of consumption is limited by his expectations of life time income. This theory states that a typical individual in his early years of life spends on consumption either by borrowing from others or spending the assets bequeathed from his parents. And in his main working years of lifetime he consumes less than his income and therefore makes net savings. He invests these savings in assets, that is, accumulates wealth which he consumes in the future years. In this hypothesis wealth is assigned a crucial role in consumption decision, wealth includes not only property such as houses, stocks, bonds and saving accounts etc., but also the value of the future earnings. The individual in his lifetime after retirement again dissaves i.e., consumes more than his income in the later years of his life but is able to maintain his consumption in the lifetime even after retirement. Thus in the life cycle model consumption is not a mere function of current income but on the expected

3 50 lifetime income. Besides, in life cycle theory the wealth presently held by individuals also affects their consumption. DATA AND METHODOLOGY The data used under this study are secondary in nature and have been collected from National Accounts Statistics Back Series ( to ), Central Statistical Organisation, Ministry of Statistics and Programme Implementation, Government of India. As stated above, in order to examine the relationship between consumption and income this study used Ando and Modigiliani s hypothesis. In life-cycle model the determinants of consumption are disposable income and financial wealth. The consumption here is defined as the sum of the expenditure on goods and services of private residents and non-profit organizations. As a result consumption is defined as a dependant variable. The two independent variables are: 1) National Disposable Income, 2) Financial Wealth. In this paper GDP at constant prices has been taken instead of National Disposable Income (NDI) since NDI is not available for the period under study. In the case of lack of NDI in particular at constant prices the best proxy is GDP. As stated above, Albert Ando and Franco Modigiliani provided the life cycle hypothesis. Similar to Friedman s hypothesis, the life cycle hypothesis states that an individual s consumption in any time period does not depend on the current income of the period. They argue that wealth plays a crucial role in consumption decision. According to life cycle hypothesis the consumers visualize themselves as having a stock of initial wealth, a flow of income generated by that wealth over their lifetime. Consumption decisions are made with the whole series of financial flows in mind. After the pioneering work of Ando and Modigiliani the wealth effect on the consumer s expenditure has often been analysed as the life cycle hypothesis. It is exposited by Modigiliani. This model is formulated in the following form: C t = α A t-1 + βy t + ε t (1) Where Y t is labor income, A t-1 is wealth at the end of period t-1. In Ando and Modigiliani estimated model both α and β coefficients appeared with positive values. In order to measure the wealth or assets, Davidson has suggested the following formula: A t = A t-1 + (Y t - C t ).(2) Where (Y t - C t ) is saving at end of the period, t. The equations of (1) and (2) can be transformed to the following equations (3) and (4) respectively, which have included the first and second lag: C t-1 = α A t-2 + βy t-1 + ε t-1 (3) A t-1 - A t-2 = (Y t-1 - C t-1 ).(4) By subtracting equation (4) with equation (1), we get:

4 51 C t - C t-1 = βy t - βy t-1 + α (A t-1 - A t-2 ) + ε t - ε t-1 (5) Replacing equation (4) in (5) and reordering, we obtain the equation (6) as follows: C t = βy t + (α β) Y t-1 + (1- α) C t-1 + U t (6), which is transformed in to a autoregressive distributed lag model of C t and Y t and the error term, ε t - ε t-1, in the equation is assumed as U t. I) THE ADAPTIVE EXPECTATIONS MODEL Suppose we postulate the following model: The consumption, C, is linearly related to permanent income, X* C t = β 0 + β 1 X t * + u t (1) Where C = Consumption X* = Permanent income U = Error term Equation (1), postulates that the consumption is a function of permanent income. Since the permanent income is not directly observable, we need to specify the mechanism that generates permanent income. In order for it, let us propose the following hypothesis about how permanent income is derived: X t * - X t * -1 = γ ( Xt - X t * -1 )..(2) Where γ, such that 0< γ 1, is known as the coefficient of expectation. Hypothesis (Eqn.2) is known as the adaptive expectation and is popularized by Cagan and Friedman. Eqn.(2) implies that economic agents will adapt their expectations in the light of past experience and that in particular they will learn from their mistakes. More specifically this equation states that expectations are revised each period by a fraction γ of the gap between the current value of the variable and its previous expected value. Equation 2 can also be written as : X t * = γ Xt + ( 1- γ ) X t * -1 ( 3 ) Eqn.,(3), shows that the expected value of permanent income at time, t, is a weighted average of the actual value of the previous period with weights γ and ( 1- γ ) respectively. If γ = 1, X t * = Xt, which means expectations are realized immediately and fully in the same period. If on the other hand, γ = 0, X t * = X t * -1 which means expectations are static i.e., conditions prevailing today will be maintained in all subsequent periods. Expected future values then become identified with current values.

5 52 Substituting Eqn. (3), into eqn. (1), we obtain: C t = β 0 + β 1 [γ X t + ( 1- γ ) X t * -1 ] + u t = β 0 + β 1 γ X t + β 1 ( 1- γ ) X t * -1 + u t ( 4 ) Now by introducing lag one in eqn. 1, and multiplying it by 1- γ, and substituting the product from eqn. (4), we obtain: C t = γ β 0 + γ β 1 X t + ( 1- γ ) C t-1 + u t - ( 1- γ ) u t-1 = γ β 0 + γ β 1 X t + ( 1- γ ) C t-1 +V t.. ( 5 ) Where V t = u t - ( 1- γ ) u t-1 In eqn.(1), β 1 measures the average response of C to a unit change in X* or long run value of X. In eqn. (5), on the other hand, γ β 1 measures the average response of C to a unit change in actual or observed value of X. If γ = 1, the current and long run values of X are the same. Once, the estimate of γ is found from the coefficient of lagged X through eqn. (2), we can easily compute β 1 by simply dividing the coefficient of X t (= γ β 1) by γ. II) THE STOCK ADJUSTMENT OR PARTIAL ADJUSTMENT MODEL Marc Nerlove developed this model. To illustrate this model, let us consider the accelerator model, which assumes that there is an equilibrium, desired or long run amount of capital stock needed to produce a given amount under the given state or technology, rate of interest etc,. Let us assume that the desired level of capital Y t * is a linear function of output X: Y t * = β 0 + β 1 X t + u t (6) Since the desired level of capital is not directly observable, Nerlove postulated the following hypothesis known as partial adjustment or stock adjustment hypothesis: Y t - Y t-1 = δ ( Y t * - Y t-1 ) (7) Where δ, such that 0 < δ 1, is known as coefficient of adjustment and where Y t - Y t-1 = actual change and ( Y t * - Y t-1 ) = to desired change. Since Yt - Y t-1 is the change in capital stock between two periods and is nothing but investment and it can alternatively be written as: I t = δ ( Y t * - Y t-1 ) (8) Where I t = investment in time period, t. Equation, 8, postulates that the actual change in capital stock in any given time period, t, is some fraction of δ of the desired change for the period. If δ = 1, it means that the actual stock of capital is equal to the desired stock; that is, actual stock adjusts to the desired stock instantaneously.

6 53 However, if δ = 0, it means that nothing changes since actual stock at time, t, is the same as that observed in the previous time period. The coefficient, δ, is expected to lie between these extremes since adjustment to the desired stock of capital is likely to be incomplete because of rigidity, contractual obligations etc,. hence the name of this model is known as partial adjustment model. Equation ( 7 ) can alternatively be written as : Y t = δ Y t * + ( 1- δ ) Y t-1 (9) The substitution of equation ( 6 ) into equation ( 9 ) gives: Y t = δ (β 0 + β 1 X t + u t ) + ( 1- δ ) Y t-1 = δ β 0 + δ β 1 X t + ( 1- δ ) Y t-1 + δu t..( 10 ) Equation ( 6 ) represents the long run demand for capital stock, on the other hand, equation ( 10 ) represents short run demand for capital stock. Once we estimate short run function and estimate the adjustment coefficient, δ, ( from the coefficient of Y t-1 ), we can easily derive the long run function by simply dividing δ β 0 and δ β 1 by δ and omitting the lagged Y term, which will then give equation ( 6 ). CONSUMPTION FUNCTION ESTIMATE By using data from to at prices, which is published by National Accounts Statistics Back Series ( to ), Central Statistical Organisation, Monistry of Statistics and Programme Implementation, Government of India, the short and long run consumption function for India has been estimated. Suppose consumption, C, is linearly related to permanent income X* then: C t = β 1 + β 2 X t * + u t (11) Since X t * is not directly observable we need to specify the mechanism to generate the permanent income. Suppose we adopt the adaptive expectations hypothesis specified in equation, (1). Using this equation and simplifying we obtain the following estimating equation: C t = α 0 + α 1 X t + α 2 C t-1 + V t..(12) Where α 0 = γ β 0 α 1 = γ β 1 α 2 = ( 1 γ ) V t = [ u t - ( 1- γ ) u t-1 ]

7 54 As explained above, β 1 gives the mean response of consumption to, say, a 1 rupee increase in permanent income, whereas α 1 gives the mean response of consumption to a 1 rupee increase in current income. TABLE-1: PRIVATE FINAL CONSUMPTION EXPENDITURE AND GDP, INDIA YEAR GDP PFCE YEAR GDP PFCE (RS.CRORES) Source: National Accounts Statistics Back Series ( to ), Central Statistical Organisation, Monistry of Statistics and Programme Implementation, Government of India.

8 55 PFCE = Private Final Consumption Expenditure in Domestic Market at prices GDP = Gross Domestic Product at Market Prices at prices From the annual data for India for the period to (presented in Table-1) the following results were obtained: C t = X t C t-1.(13) Se (15937) (0.036) (0.073) t (4.026) (5.951) (8.764) R 2 = 0.99 d = F = n = 32 Where C = Private Final Consumption Expenditure, and X = Gross Domestic Product both at constant prices. The results show that the short run marginal propensity to consume, MPC, is about 0.22, suggesting that a 1 rupee increase in the current real income would increase mean consumption by about 0.22 rupee. But if increase in income is sustained, then eventually the MPC out of the permanent income will be β 1 = γ β 1 / γ = 0.216/ (since γ = 1- α 2 or ) = or about 0.60 rupee. In other words, when consumers have had time to adjust to the 1 rupee change in income, they will increase their consumption ultimately by about 0.60 rupee. Now suppose the consumption function chosen is: C t * = β 0 + β 1 X t + u t (14) In this equation long run consumption C t is a linear function of current income. Since C t * is not directly observable, we applied partial adjustment model. Using this model and simplifying we obtain: C t = δβ 0 +δ β 1 X t + ( 1- δ ) C t-1 + δ u t = α 0 + α 1 X t + α 2 C t-1 + V t...(15) Where α 0 = δ β 0 α 1 = δ β 1 α 2 = (1 δ) V t = δ u t

9 56 In appearance this model is akin to adaptive expectations model. Therefore, the regression results given in equation,(13), are equally applicable here. The model represented through equation (14) is the long run consumption function, whereas equation (15) is the short run consumption function, β 1 measures the long-run MPC, whereas α 1 (=δ β 1 ) gives the short run MPC. The former can be obtained from the latter by dividing it by δ, the coefficient of adjustment. Returning to equation (13), the value of α 1 = can be interpreted as short-run MPC. Since δ = 0.359, the long-run MPC is The value of adjustment coefficient 0.36 suggests that in any given time period consumers only adjust their consumption by 0.36 towards the desired or long-run level. The regression result, reported through equation (13), has shown that all coefficients are statistically significant. The R-squared obtained and F-test has shown a fairly successful regression as well. The equations used in this study to estimate consumption function are autoregressive distributed lag models and hence though the Durbin-Watson test statistic reveals no serial correlation in the residuals, we can not rely on D-W statistic. Because the Durbin- Watson, d, statistic may not be used to detect serial correlation in autoregressive models as the computed d value in autoregressive models generally tends towards 2. Therefore, the best test for checking the autoregressive model is h test. The h test is as follows: h = ρ n/ 1- n[var (α 2 )].(16) where n is the sample size, var (α 2 ) is the variance of the lagged C t = ( = C t-1 ) coefficient in equation (13) and ρ is an estimate of the first-order serial correlation. The value of ρ can be estimated by using ρ (1 d/2). In this study, n = 32, ρ (1 d/2) = , and var (α 2 ) = var(c t-1 ) = (0.073) 2 = Putting these values in equation (16), we obtain: h = /1-32( ) = As h-statistic obtained in the study is -1.74, we significantly accept the null hypothesis of no serial correlation in the residuals at 5% level of significance. (The significance or the critical value of z for two-tailed test at 5% level of significance is and 1.96). CONCLUSION An attempt has been made in this study to estimate the relationship between consumption and income in India during To estimate short and long-run consumption the distributive lag models have been applied. In this study the long-run marginal propensity to consume was found to be 0.60 as against the short-run marginal propensity to consume, REFERENCES Ando.A. Modigiliani.F.(1963) The Life Cycle Hypothesis of Saving: Aggregate implications and tests, American Economic Review, Cagan.P. (1956) The Monetary Dynamics of Hyperinflations, in M.Friedman (ed.), Studies in the Quantity Theoary of Money, University of Chicago Press, Chicago.

10 57 Davidson, James E H, et al. (1978) Econometric Modeling of the Aggregate Time-Series Relationship between Consumers' Expenditure and Income in the United Kingdom" Economic Journal, vol. 88(352), Edwards, Sebastian. (1996) Why are Latin America s Savings Rates So Low? An International Comparative Analysis. Journal of Development Economics, 51(1), Enders, W. (1995) Applied Econometric Time Series. New York: John Wiley & Sons. Friedman, Milton. (1957), A Theory of the Consumption Function, National Bureau of Economic Research, Princeton University Press, Princeton, N.J. Ghatak, Anita. (1998) Aggregate consumption functions for India: A Co-integration Analysis under Structural Changes, Journal of Applied Statistics, Vol. 25, Issue 4, Gujarati, D.N. (2007) Basic Econometrics, Tata Mc-Graw Hill Education Private Ltd,New Delhi, Hall, Robert E. (1978) Stochastic Implications of the Life Cycle-Permanent Income Hypothesis: Theory and Evidence, Journal of Political Economy Vol.86, No.6, Keynes, John Maynard Keynes. (1936) General Theory of Employment, Interest and Money, Macmillan Publishers, Cambridge University Press. Modigiliani,F. Brumberg,R (1954) Utility Analysis and the consumption function : An interpretation of cross section data, Kurihara,K.K.(ed), Post-Keynesian Economics. Modigliani, F. (1966) The Pasinetti Paradox in Neoclassical and More General Models, Review of Economic Studies 33. (4). Nelson, C., and C. Plosser. (1982) Trends and Random Walks in Macroeconomic Time Series: Some Evidence and Implications. Journal of Monetary Economics, Vol. 10(2), Nerlove Marc,(1958), Distributed Lags and Demand Analysis for Agricultural and Other Commodities, Agricultural Hand Book No.141, U.S. Department of Agriculture. Pindyck,R.S. Rubinfield, D.L.(1991) : Econometric Models and Economic Forecasts,Tata Mc- Graw Hill Education Private Ltd,New Delhi. Romer, David. (2006) Advanced Macroeconomics. 3rd edition. New York: McGraw- Hill. Shaw,G.K.,(1984) Rational Expectations : An Elementary Exposition, St.Martin s Press, New York, p.25 Wen-Jen, Hsieh & Yu Hsing. (1994) "Tests of Nonlinear Consumption Functions: The Cases of Korea, Taiwan, Thailand and India," International Economic Journal, Korean International Economic Association, Vol. 8(2),