The demand side block

The demand side block covers private consumption and government‟s final consumption expenditure. This block determines GDP by the GDP expenditure identity condition given inventory levels treated exogenously, and exports and imports which are estimated in the foreign trade block³⁸. The private and government consumption are based on the Keynes‟ absolute income hypothesis and the government budget constraint models respectively.

3.3.2.2.(a) Private consumption

The private consumption function in this model is fundamentally based on Keynes‟

absolute income hypothesis (1967) in which the real consumption is a function of the real disposable income. The linear function of the consumption in the absolute income hypothesis provides three principles: first, the marginal propensity to consume has a value between zero and one. Second, the average propensity to consume decreases as income increases. Last, current income determines consumption, since the interest rate is regarded as an ineffective variable.

While the Keynesian consumption function has been proven to work successfully in the short term periods, empirical estimation of the long term periods contradicted the second principle of the absolute hypothesis by showing that the average propensity to consume is constant over long periods of time (Kuznets, 1942). In other words, the result shows the conflict between the consumer‟s behaviour in the short run and the long run, which is called

38 The use of chained indices in the national account has a non additivity problem, as weights are changed over time. This study uses an error term, defined as the difference between real GDP and the sum of GDP expenditure components such as for calculating the real GDP.

the Kuznets paradox. This empirical result of long time-series has facilitated the development of alternative consumption functions.

Two prominent alternative hypotheses have been proposed in the 1950s. The life-cycle hypothesis proposed by Modigliani and Brumberg (1968) emphasize consumption smoothing by supplementing the current income regarded as the unique determinant of consumption in the Keynesian consumption function with the present value of lifetime earnings which is called wealth. The permanent income hypothesis as proposed by Friedman (1957) also explains a constant ratio of consumption to income in the long term by suggesting that consumption is mainly determined by permanent income.

Both models are based on Irving Fisher‟s intertemporal optimization approach (1930) in which the forward-looking consumers make decisions on their current income and expected income in the future. These theories provide the modern macroeconomist with a foundation.

The intertemporal utility maximization method has been applied to dynamic mathematical methods, which is called the Euler equation approach, in the modern consumption theories and the models have been extended to deal with uncertainty and expectations.

Although it is desirable to include the wealth variable in the consumption function in view of the importance of wealth pointed out in the literature, this model does not include the wealth variable in the behavioural equation of private consumption for the sake of simplicity and unavailable data in this version. Modelling wealth is relatively difficult, since few economic theories explaining the relationship between wealth and other economic variables exist. In addition, creating the wealth series requires modelling the prices of real estate and financial assets variables and the quantity of them consumers hold. Such stock price indices have considerable volatility. Several papers simply use a stock price index as a proxy for

wealth. However, the error from endogenously modelling such a volatile index would worsen the results of the simulation. If the wealth variable is treated exogenously, it requires an additional assumption for simulation. Besides, the current System of National Accounts (SNA) published in Bank of Korea does not provide the „wealth‟ data and so official data is not available.

For those reasons, we stick to the fundamental consumption function, „absolute income‟ hypothesis which is defined as a function of real disposable income and price of the category in this model. The consumption function is given by:

where is the real consumption. The lagged real consumption expenditure, is included to measure the habit persistence which indicates the effect of the previous period‟s consumption on the current consumption as Brown (1952) suggested. is the consumer price index. represents the real disposal income.

Table 3.7. Estimation of the private consumption function (a) Regression model

Dependent Variable : Sample : 1970 ~ 2011

Regressor Coefficient Standard Error T-value P-value

Constant -0.233 0.547 -0.427 0.672

0.654^*** 0.070 9.290 0.000

0.358^*** 0.067 5.300 0.000

-0.051^*** 0.017 -3.007 0.005

Note: ^***,^**, and ^* indicate significance at 1 %, 5 %, and 10 % respectively.

The estimation includes dummy variables equal to unity for the period 1998 and 2001~2011.

(b) Cointegration test : = 8.155^***

(c) Diagnostic tests

̅ = 0.999 = 1.487

̂ = 0.018 = 13193.10 (0.000)

= 2.035 (0.163) = 1.593 (0.451)

= 0.043 (0.837) = 0.934 (0.471)

Note: the right-hand side value in parentheses indicates p-value

The result of a log-linear form for estimation is described in Table 3.7. We confirm that there is a long-run relationship between consumption and disposable income and consumer price index by the ARDL bounds testing; the calculated F statistic is higher than the upper bound critical value at 5% significance level. All of the coefficients on variables are significant and have the expected sign; the coefficient of disposable income is positive and the coefficient of consumer price index is negative. The lagged variable indicates that around 35%

of the adjustment process to the desirable level in long-run takes place in the first year. The result shows that the short-run and long-run elasticity of disposable income are 0.358 and 1.032 respectively, meaning that a 1% increase in disposable income raises consumption by approximately 0.36% and 1% in the short term and long term respectively. The short-run and long-run own-price elasticity are -0.051 and -0.146 correspondingly. The marginal propensity to consume (MPC) is 0.250 in the short-run and 0.721 in the long-run³⁹.

3.3.2.2.(b) Government consumption

The government consumption on goods and service belongs to this component of GDP.

The consumption is chosen by the fiscal policy that determines taxes and spending.

Normally, fiscal and government spending areas in macro-econometrics models generally build up a fiscal reaction function based on the government budget constraint condition as mentioned earlier. It is desirable in that the government is assumed to retain a specific robust fiscal rule, which is regarded as best practice, in order to examine unexpected shocks of the fiscal policy (and monetary policies) on agents‟ behaviour; this should be modelled based on micro-foundations such as preferences and resource constraints rather than

39 We calculate the MPC by multiplying the income elasticity by the average ratio of consumption to disposable income for the period 1970~2011.

simple parameters estimated entirely on the basis of relationships observed in historical data to represent the agent‟s prediction for the change in policy (Lucas, 1976).

When it comes to modelling the government consumption in the system model, it is often modelled by an exogenous variable of government expenditure to analyse the ripple effect of fiscal policy on the economy. However, the disadvantage of the external fiscal policy approach for simulation is that it requires an additional assumption for the fiscal policy decision when the policy experiment is conducted in the model. An arbitrary assumption for the government spending could underrate or exaggerate the effect of the policy experiment on output in the simulation analysis. Therefore, we build up a behavioural equation for government consumption that is able to respond to external conditions properly.

The model follows the assumption from Harrison et al.‟s BEQM model (2005) for the government consumption function; the government has targets for spending, so that it would not generally follow a balanced budget each period. Moreover, spending is allowed to vary temporarily from target levels following a shock. The target level is defined as the relationship between spending and tax revenues in this model. The model specification refers to the Government Budget Constraint by Barro (1974) and Sargent and Wallace (1981) ⁴⁰. The equation is built up in order to evaluate the effect of the change in tax revenues on the government consumption based on these theoretical relationships.

According to the government budget constraint (GBC), the government consumption is equal to the total tax revenue at a point in time and across time. We check the long-run relationship between the government final consumption expenditure ( ) and real tax revenue

40 The intertemporal equation of the GBC was firstly applied by Barro (1974). The modern application of the GBC has been developed from Sargent and Wallace (1981). Overall, the intertemporal GBC contributes to the development of theoretical works in monetary and fiscal policy studies.

( ) by the cointegration test, and then the government consumption expenditure is modelled as follows:

Table 3.8. Estimation of the government consumption function (a) Regression model

Dependent Variable : Sample : 1971 ~ 2011

Regressor Coefficient Standard Error T-value P-value

Constant 1.522^*** 0.473 3.219 0.001

0.891^*** 0.041 21.334 0.000

0.063^** 0.027 2.430 0.025

Note: ^***,^**, and ^* indicate significance at 1 %, 5 %, and 10 % respectively.

The estimation includes a dummy variable equal to unity for the period 1990 (b) Cointegration test : = 5.444^**

(c) Diagnostic tests

̅ = 0.999 = 1.610

̂ = 0.017 = 18005.75 (0.000)

= 1.516 (0.226) = 1.306 (0.757)

= 2.597 (0.116) = 1.119 (0.354)

Note: the right-hand side value in parentheses indicates p-value

Table 3.8 summarises the estimation result of government consumption. We found a long-run equilibrium relationship between government consumption and the tax revenue, meaning that the government‟s plan for consumption has not been made in isolation from the tax revenue. The coefficient on the lagged government consumption variable has a high value which is close to one, implying that the government‟s consumption is subject to substantial inertia meaning the adjustment process to the long run equilibrium relationship is slow. The short-run and long-run elasticity of real tax revenue are 0.063 and 0.578, which means that the government gradually increases consumption as the tax revenue rises for the long-run.