For a survey of the Latin American pension reform attempts and further references, see Queisser (1995); for a review of the current reform approaches in Central and Eastern Europe, see Holzmann (1997).
For a detailed survey and analysis of the Chilean pension reform and further references, see Diamond and Valdes-Prieto (1994).
Merely changing the financing mechanism of an unsustainable, unfunded retirement scheme is not sufficient to put it on a sustainable, funded basis. Any real pension reform essentially has to undertake two changes simultaneously: reduce the commitment of the given pension scheme (given target income replacement rates, essentially by increasing the retirement age) and shift the financing mechanism.
The other four equations of this growth model are traditional. The first specifies the output, Y, via a production function with constant returns to scale on capital, K, and labor, N (man-hours in efficiency units):
The second equation specifies an exogenous growth rate, n, of total employment (in man-hours, L):
dL/dt = nL.
The third specifies a definition equation between N and L via the technicalchange multiplier, T:
N = TL.
The fourth equation defines the capital coefficient:
k = K/N.
d(.)ldt is the time derivative. and δ is the rate of depreciation of capital.
As the model features an external effect, the solution of the social planner’s problem will not necessarily coincide with the competitive equilibrium in a decentralized economy.
In the decentralized solution, the net rate of return with externalities is f’(k*) - δ = λ + n + α (K, . . .) k*, still leaving a growth differential of a (K, . . .) k* to compensate the transition generation.
The data draw from a wide range of different sources, mainly official but also private. Long-term time series in official publications are often not available and sometimes require index-type linking to overcome structural breaks.
No strictly comparable FIRs could be established for the pre- and post-1975 periods. However, the available data for Chile suggest a long-term decrease in FIR-2 from 63 percent (1940) to 32 percent (1950) and 29 percent (1960), with a slight increase to 39 percent (1971) prior to a major shake-up of the economy (Cerda and Zeballos, 1975).
These data are used in Levine and Zervos (1996), and the access granted by Ross Levine is gratefully acknowledged.
For an analysis of the experience and mistakes of financial market liberalization during this early period of economic reform, see De la Cuadra and Valdes-Prieto (1992).
To estimate TFP, a simple growth-accounting exercise is undertaken (with logarithmic approximation to account for discrete time and a constant labor share of 0.65). Admittedly, this estimation ignores its well-known limitations, such as the quality of factor inputs and the time-varying factor shares, because of data constraints.
For definitions of the variables used in the regressions in Tables 3-5, see notes to Table 3. All variables in Tables 3-5 have been subject to a unit root test (augmented Dickey-Fuller test), which has been passed for all endogenous and most exogenous variables at the 5 percent confidence interval. For the unemployment rate, the unit root could not be rejected. This outcome may cast doubt on the estimated parameters and their statistical significance; although the stability of the parameters under alternative measures of financial markets and the Durbin-Watson test statistics of about 2 may give confidence in the results, further investigation is certainly required.
The parameter estimates for the ΔFIR and ΔSMI variables differ slightly from previously presented estimates, owing to data revisions and the normalization of the respective sample average to one. The normalization allows for a direct parameter comparison between different financial market indicators and a straightforward interpretation of the parameter value.
The approach uses the data structure of the Almon lag to calculate a composite variable ΔFMI(1, s)t = 1sΔFMIt + 2sΔFMIt-1 + . . . 1sΔFMIt-1. Thus ΔFMI (2,2)t = ΔFMIt + 4ΔFMIt-1 + 9ΔFMIt-2. This approach results from economic, econometric, and data considerations. One would assume from an economic viewpoint that improvements in the financial market, as measured by changes in the level of FMI, would have little immediate impact but that the impact would grow over time. However, a direct application of the traditional Almon procedure is prevented by the small number of observations, with the increasing s under the full Almon approach reducing the degrees of freedom on a pro rata basis. In addition, AFMI (1,s) for 1 and s = 1, . . . , 3 proved to be highly correlated. Finally, unless very strict conditions are met, the Almon procedure will yield biased and inconsistent estimates.
In view of the weak public sector statistics for Chile, the use of the unemployment rate is preferred to the use of disposable income.
Our reported long-term point estimate of the effects of financial markets on capital accumulation of about 0.015 is close to the estimate in the Levine and Zervos (1996) cross-country study of 0.011 (with a t-value of 2.38).
The net contributions were calculated as premium payments plus other increases minus commissions charged minus benefit payments. The flow returns on pension fund assets were calculated using the average yearly stock and a representative interest rate of the financial market. In principle, the flows ought to be corrected for savings generated in the upstream insurance companies and for dissavings generated by benefit payouts, but these data were not at hand.
The source for this information is Juan Carlos Mèndez, Budget Director until 1980, quoted in Diamond and Valdéz-Prieto (1994).