This study analyses housing and equity wealth effects on
householdsnon-durable consumption using the Survey of Italian Household and
Wealth (SHIW) published by Banca DItalia. This dataset contains detailed
information on Italian households consumption, income and wealth from
1989-2002, and is constructed as a panel. The main contribution of our study is
that, using the variation through time and across households, the idiosyncratic
shocks to households' income and wealth can be identified ed. Furthermore, we
analyse the consumption responses of different age and wealth groups of
households. In addition, we also investigate indirect wealth effects, i.e.how
different types of households react to aggregate equity and housing price
Our main contribution arises from using this single dataset to analyze
household consumption responses to housing and equity wealth shocks.
Furthermore, we analyzed the consumption responses of di¤erent age and wealth
groups of households. In addition, we investigated indirect wealth effects, i.e.
how different types of households react to aggregate equity and housing price
changes. Regarding direct wealth effects (those arising from the self-reported
change in wealth), our results indicate that homeowners.consumption react
statistically and economically significantly to realized housing wealth shocks
(the estimated MPC is over 8 percent). This is slightly larger than many US
studies, where Peek (1983), Skinner (1984, 1986) and Engelhardt (1996) all found
effects between 3 and 5 percent, but is in the range of estimates found by
Disney et. al. for the UK and by Hori and Shimizutani (2003) for Japan. Our
results also indicate that the estimated unrealized equity wealth effects for
stockowners, although statistically signi.cant, are economically quite small,
with an average MPC of 0.4 percent. The estimated effect is lower than estimates
for other countries such as the US where Dynan and Maki (2001) estimated a MPC
of 5-14 percent. We additionally find that the estimated elasticity for old
(45-65 years old) households is larger (the estimated MPC is around 15 percent),
whereas for younger (25-44 years old) house- holds, the estimated elasticity is
smaller (around 5 percent), but not statistically significant. Unexpectedly, the
estimated elasticity for the richest wealth group households is the largest
(around 10 percent), but not statistically significantly di¤erent from the
medium wealth group households. elasticity (around 7 percent). For the lowest
wealth group household, the estimated elasticity is statistically not
signi.cant. One possible explanation is that binding credit constraints are
preventing households in the lowest wealth group households from increasing
their consumption in response to housing wealth gains. We also investigated the
effect of the house-price and stockmarket indices. For these indirect wealth
effects, we find no support for indirect housing wealth effects, whereas
indirect equity wealth effects are found to be statistically signi.cant and
economically large. The indirect equity wealth effects are likely to be related
to expected improvements in income outlook, given that both stockholders and
non-stockholders increase their consumption in response to positive stockmarket
developments, and that the estimated coefficients between these two groups are
found to be similar.
As can be observed from reading the paper the generic application of 'wealth effect analysis is' somewhat impeded by the heterogenous degree of homeownership rate, life course events (in this case purchase of first home), mortgage and credit markets, degree of households' stock ownership ... and I am sure you can come up with many more. In short, institutions and cultural specificities matter; on this area in particular. However, and as many of my regular readers will have expected I am especially intrigued by the implicit findings in terms of the wealth effect over the life cycle. The following is thus an important observation ...
We additionally find that the estimated elasticity for old (45-65 years old)Leaving the issue of statistical significance aside for a moment I do think that this point with respect to the life cycle effect of wealth responsiveness is important to take aboard. When you think of it it is not that unexpected. Older households will thus tend to depend more on their savings and crucially the income they can earn on those savings than their nominal income itself in the form of regular cash flows. At least, this is a pet hypothesis of mine and if you could isolate the effect of income earned on saving it would be interesting to see whether this elasticity divergence holds. However, this point opens up the door to a much more profound discussion as to how we deal with the assumption of dissaving over the life cycle as consumers and by derivative populations age. Specifically, I am talking about how to to incorporate two important stylised facts or, as it were, to investigate the extent to which they exist. These would be; 1) the point that economic agents (consumers) don't dissave to 0 over their life cycle and 2) that increasing life expectancy makes the life cycle pattern uncertain in the latter years. In general the authors make the following point ...
households is larger (the estimated MPC is around 15 percent), whereas for
younger (25-44 years old) households, the estimated elasticity is smaller
(around 5 percent), but not statistically significant.
(...) the standard life-cycle model predicts that older households should have higher MPCs to income and wealth shocks than younger households because of the di¤erences in ex- pected life times. In addition, as shown by Carroll and Kimball (1996), adding income uncer- tainty to the standard life-cycle model induces the consumption function to be concave, in which the marginal propensity to consume out of transitory income shocks, as well as wealth shocks, declines with the level of wealth.
Formally, the authors estimate the following model in first differences except the risk free rate.
Where Cit is non durable consumption for household i, Rt is the risk free rate at time 't', lnYit is income of household i at time 't', Wh is housing wealth at time 't' and We is equity wealth at time 't'. Epsilon (or Ut) as always is an error term. A good place to start for further study I would say.