consumption decision on a discounted stream of current and future expected net income
/&,W and on their current stock of financial wealth ):W . The basic reason for doing this is
derived from a concept of inter-temporal utility maximization of households, whereby they
find it optimal to smooth consumption over time. If households expect for example a
temporary decline in their income they will according to this hypothesis mainly react via a
reduction in their savings rate. Alternatively if they expect an increase in their future net
income because of credibly announced tax reductions the current savings rate may also fall,
Lh consumption may already increase in the present period in anticipation of higher future
income. Private consumption &W under this hypothesis can theoretically be represented as
follows
ct =(θ+S )[lci, + f:, ] pt∣pct (1)
where θ is the rate of time preference and S the probability of death, or the inverse of the
forward looking horizon of the household. Life cycle income is defined as
/CI, =∫(1 - WO ) :3L+75
t L
(2)
The permanent income represents the current and discounted future expected net income
stream the household sector is expected to earn. It consists of all non-capital income, LH net
labour income wO:W1W and all other transfers to households, including unemployment
benefits, 75W. The other determinant for private consumption is financial wealth
f:, = O9t + ) + %t (3)
which, on the aggregate level, consists of the market value of firms in the domestic
economy 09W, the net foreign asset position )W and government debt %W. Note, however,
though government debt enters the definition for private wealth, this does not mean that it
has a positive effect on private consumption because households deduct future tax payments
and reductions in transfer payments from their permanent income which are required to
service the debt. This is also known as Ricardian Equivalence. This proposition does,
however, only hold in its extreme form for infinitely lived consumers. Life cycle consumers
will discount the future more heavily and thereby underestimate the tax burden associated
with government debt. Consequently they will regard government debt at least partially as
net wealth of the household sector. As Summers and Poterba have shown, however, this net
wealth effect of government debt is negligible in the life cycle model.
Our empirical specification of the model allows for a deviation to formulation (1) above
that reflects the findings of many empirical studies of consumer behaviour (e.g Campbell
and Mankiw (1989)) that a sizeable fraction of consumption is based on real current
disposable income <'I6W because of liquidity constrained private households. Therefore,
the consumption function used in the model is
C1 = (1 - λ)θ + S )[LCIt + f:, ] 3 ∕3Cf + λ<'I6, (1a)
where the parameter λ denotes the share of liquidity constrained consumption. Table 1 lists
the parameters in our consumption function. A forward looking horizon of 50 years has been
assumed, implying a value for S of 0.02. Estimates of λ based on Euler equations for those
countries for which quarterly data was available ranged from 0.2 to 0.4, while on annual data
they ranged from 0.2 to 0.6. These estimates did not differ significantly between countries