Page 76 - Invited Paper Session (IPS) - Volume 2

IPS184 Sanvi Avouyi-Dovi et al.
its parameters from the restrictions conditions. We add two exogenous
variables (uncertainty or risk factor and demographic factor). The system is
estimated with three-stage least squares over the sample 1999-2016. The set
of instruments variables contains the lagged values of the endogenous and
exogenous variables. Elasticities corresponding to wealth, interest rates and
additional exogenous variables are derived from the estimates. The Student-t
statistics are heteroscedasticity and auto-correlation consistent. The
conventional validation tests (Wald, ADF, autocorrelation, normality, Arch, etc.)
are also conducted.
Table 1 displays the results of the estimates of the coefficients of the fully
constrained model with two exogenous variables. As regards the effects of the
interest rates, we note that each asset is significantly and positively correlated
with its own return. In the end all things being equal, an increase in the real
return on asset “i” (i= M1, M2M1, ..., Lifebonds) leads to a rise in asset “i”. The
sensitivity of the share to the real return depends on the asset, with the
reactions of M1, M2M1 and Lifebonds to their own real returns being stronger
than those of M3M2, Asset and PEL. However, whatever the asset, the
uncompensated elasticity is significantly different from zero. One can also note
cross-effects between real returns and financial assets. In particular, overall,
the real interest rate of M1 negatively impacts the other shares; in this case, a
substitution effect prevails. A similar pattern exists regarding the effect of the
real interest rate of M2 on the other shares. In fact, overall, substitution effects
prevail.
As regards wealth elasticity, the empirical results bring to light three
situations. In the first case, there is a strong positive effect of wealth on
Lifebonds. Only the richer households seem to react to the variations in
Lifebonds. In the second case, we note a negative link between some financial
assets (M1, M3M2) and wealth. According to Blake (2004), these assets are
wealth-inferior assets with a long run negative elasticity. It is worth noting that
these assets are relatively liquid. Maybe the changes in wealth are offset by
the opposite variations in these liquid assets. In the last case, we note a weak
correlation between two assets (M2M1 and Assets) and wealth as their
coefficients in the corresponding regressions are not significantly different
from 0. In particular, Assets seem to be substantially driven by their own
returns and the substitution effects induced by Lifebonds.
Finally, only M1 is significantly correlated with the business cycle (through
the de-trended unemployment rate) and financial uncertainty (via financial
market volatility). Unexpectedly, the other shares, especially the Assets one,
are not significantly impacted by real or financial risk factors.

63 | I S I W S C 2 0 1 9