31.01.2013 Views

Empirical life cycle models of labour supply and - Statistisk sentralbyrå

Empirical life cycle models of labour supply and - Statistisk sentralbyrå

Empirical life cycle models of labour supply and - Statistisk sentralbyrå

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

<strong>Empirical</strong> Life Cycle Models Social <strong>and</strong> Economic Studies 91<br />

The first order conditions corresponding to this optimization problem implicitly<br />

define the dem<strong>and</strong> for goods <strong>and</strong> leisure as functions <strong>of</strong> wt, Pt, Yt <strong>and</strong> rtAt-i,<br />

<strong>and</strong> the parameters <strong>of</strong> the tax function <strong>and</strong> the preferences. In order to simplify<br />

the exposition further, ignore income taxation <strong>and</strong> assume that these dem<strong>and</strong><br />

functions can be given closed form solutions. If _C_t f (Ce , Lt) is the vector <strong>of</strong><br />

consumption goods with corresponding price vector pt = (tvt , Pt), the<br />

within-period dem<strong>and</strong>s can be written as<br />

(60) = (Pt , Yt)<br />

where Lt is the vector <strong>of</strong> dem<strong>and</strong> functions.<br />

These dem<strong>and</strong> functions are called y-conditional dem<strong>and</strong>s since consumption is<br />

chosen conditional on full income. They are homogenous <strong>of</strong> degree zero in prices,<br />

pt , <strong>and</strong> full income, Y, <strong>and</strong> the functional form varies across goods. According<br />

-to Blundell <strong>and</strong> Walker (1986), the dem<strong>and</strong> functions for all goods generally<br />

change if there are binding constraints in any market in period t. In contrast<br />

binding constraints in the future only change the value <strong>of</strong> Y. The assumption <strong>of</strong><br />

intertemporal additive separability then allows for decentralization over time.<br />

Although we are assuming perfect certainty, we notice that the introduction <strong>of</strong><br />

uncertainty about future prices <strong>and</strong> preferences leaves the two-stage budgeting<br />

model almost unchanged, since all uncertainty is captured through the<br />

distribution <strong>of</strong> full income across the <strong>life</strong> <strong>cycle</strong>, cf. Blundell <strong>and</strong> Walker (1986).<br />

From the discussion <strong>of</strong> the determination <strong>of</strong> full income it is evident that the<br />

full income variable in the y-conditional dem<strong>and</strong>s summarises the effects <strong>of</strong> past<br />

decisions <strong>and</strong> future prices <strong>and</strong> preferences, on current decisions. In that respect<br />

full income serves the same purpose in the y-conditions dem<strong>and</strong>s as the<br />

marginal utility <strong>of</strong> wealth does in the Frisch dem<strong>and</strong> functions.<br />

Compared with the Frisch dem<strong>and</strong>s, the y-conditional dem<strong>and</strong>s, however, only<br />

include variables that can be observed, <strong>and</strong> they can be observed within the<br />

current period. Estimation <strong>of</strong> the dem<strong>and</strong> function for a particular good<br />

requires observations <strong>of</strong> the dem<strong>and</strong> for the good, the price vector pt <strong>and</strong> full<br />

income. If we take account <strong>of</strong> income taxation, rtAt_i must also be observed in<br />

addition to the parameters <strong>of</strong> the tax system. Since many surveys are either<br />

income or expenditure surveys, it can be hard to find data sets that include all<br />

these variables.<br />

Blundell (1987) points at that empirical specifications based on parametrization<br />

<strong>of</strong> direct utility, can <strong>of</strong>ten be shown to be unnecessarily restrictive compared to<br />

dual representations. In that respect the following results are important:<br />

46<br />

1. If preferences are intertemporally separable, within-period allocation <strong>of</strong>

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!