The article aimed at assessing whether fully funded pension regimes, based on individual capitalization, produced the distinctive effect of enhancing aggregate private savings and, in turn, helped somehow to strengthen domestic capital stock markets. Likewise, efforts were devoted to analysing the impact upon private savings of a group of economic and demographic variables which the related literature usually link to the performance of both defined benefit and defined contribution pension systems.
In meeting the sought objectives, the traditional life-cycle hypothesis was resorted to, in the first place, in order to explain how individuals´ saving decisions were modified following the introduction of social security taxes within the framework of a PAYG regime. Next, and in line with contributions stemming from Feldstein (1974), Blinder (1982) and Bailliu and Reisen (1997), the theoretical approach was extended in order to include the cases of endogenous retirement age and fully funded regimes.
The impact of individual capitalization systems upon aggregate private savings was next considered within a life-cycle approach in which various hypotheses where successively upheld, such us: homogeneous and heterogeneous individuals, voluntary and compulsory contributions and loose and tight borrowing constraints. The theoretical analysis permitted to prove that only under mandatory contributions and operating liquidity restrictions private savings would unambiguously be increased by pension fund assets.
In ascertaining the validity of the paper´s main hypothesis, the problem of degrees of freedom, stemming from data scarcity caused by the relatively recent implementation of most individual capitalization regimes, had to be dealt with by using a panel data model including statistical series from Argentina, Chile, Colombia, Mexico, Peru and Uruguay for the period 1995-2006.
Also, the recourse to the fixed effect variant whereby intersections were let to vary among countries, permitted to capture countries´ particular features and yet consider similar variables´ coefficients or common slopes for all the cross section units.
In relation to the econometric estimation of coefficients, results gave ample support to the assertion that mandatory pension fund regimes would have a positive impact upon aggregate private savings as the coefficient of pension fund stocks not only held the expected sign but it was also significantly different from 0 in all but one single case.
With regards to the rest of estimations, coefficients´ performance exhibited results of varying econometric soundness, depending on the variable analyzed, but generally falling in line with predictions of the life-cycle model; thus, the interest rate´s coefficient was always positive and statistically significant independent of whether the nominal or the real interest rate were used, the main implications being that substitution effects prevailed over income effect and that the assumed hypothesis of a positive impact of liquidity restrictions upon private savings really held.
The idea of ultrarationality between private and public savings resulted also generally proven as the coefficient held the expected negative sign and resulted significantly different from 0 at 5%, in two cases and at 10% in other two cases. On the other hand, the Keynesian relationship between saving and income (with gross domestic product used as a proxy for the latter) gathered in general econometric support as, apart from the bearing the correct sign, results showed coefficients statistically different from 0.
Poor results were however found for the cases of loans to the private sector and the growth rate of per capita income as, in spite of expected signs generally being achieved, higher significance levels (10% or 15%) were required for discarding equal to 0 coefficients.
Finally, the almost null econometric performance of the dependence index is a worth stressing feature as, contrariwise to what it would have been expected, no relationship could be found between this ratio and the level of aggregate private savings and therefore the idea that demographic variables could somehow influence savings could not at this stage be proven. The point is not however ruled out that the short length of statistical series, as well as the way the ratio was computed, somehow conspired against the variable´s performance at the moment of assessing its real impact upon savings.
