Aging and Growth in Den, Fra, Norw

This time, we pick up three countries - Denmark (DE), France (FR) and Norway (NO) - because they have the longest time series available on Eurostat.

The basic idea is the same as before, but now we try to investigate longer aging-growth relationships at the cost of concentrating on three country cases.

For DE and NO real per-capita GDP and aging ratios are available since 1975 to 2017; for FR the availability is restricted to the horizon 1991-2017 that means, anyway, some four years more than the generality of statistics for European countries (normally starting from 1995).

We develop a set of log-log regressions with the log of real per-capita Gdp as dependent variable, alternatively regressed on the first lag of the log of three aging explanatory variables:

1) the incidence of people aged 65+ on total residents,

2) the incidence of youngsters aged 14- and elders aged 65+ on active citizens aged 15-64 years,

3) and the incidence of elders aged 65+ on active citizens aged 15-64 years.

Lags help soften problems arising from the characteristic of endogeneity and simultaneity of regressors.

Regressors are completed with observation-unit dummies catching structural long persistent differences across countries, and sets of temporal dummies trying to capture historical and conjuncture events that had repercussions in all countries year by year. Time dummies have also the function to construct the period-specific background against which to detect contributions given by aging processes. Unfortunately, Eurostat series on employment and activity rates start only from 1995 and are not useful for this exercise.They have been already used as control variables in a previous panel, developed bu Reforming, covering 11 European countries.

Given the very long horizon (more than forty years) temporal dummies cover a crucial role in defining estimation results. Even after adopting yearly dummies (one dummy for every single year), coefficients of the three aging variables come out significative and of the expected signs as well: a 1% increase in the incidence of 65+, or in the structural dependency ratio, or in the structural old dependency ratio can be respectively associated with a 0.392%, 0.443%, 0.331% decrease in real per-capita Gdp.

All in all out, using the widest set of dummies, and in particular the most granular set of temporal dummies, does not prevent aging explanatory variables to reveal significative.

In other words, what the exercise does is to first regress real per-capita Gdp on a granular set of dummies, in order to already obtain a good reconstruction of the evolution of the dependent variable over time in all three countries, and then, as second step, to check if the introduction of a new regressor related to aging comes out significative and with the expected sign.

If the exercise is restricted over the horizon 1991-2017 (all three countries have full time series), coefficients of the three aging variables continue to come out significative and of the expected signs as well: a 1% increase in the incidence of 65+, or in the structural dependency ratio, or in the structural old dependency ratio can be respectively associated with a 0.239%, 0.249%, 0.198% decrease in real per-capita Gdp.

Over this shorter horizon (27 year instead of 43), the negative and significative impact of aging on real per-capita GDP dynamics is confirmed but to a lesser scale.

It is not easy to argue about the reduced absolute values of estimated coefficients. This may depends upon the fact that, even if aging (demography per se) is revealing some exponential characteristics, elasticity of productivity and growth can follow a decreasing path from period to period.

But explanations can be also of different nature, for example pointing to some huge breaks that hit World and European economies over a more than 40 years long period (oil shocks in the '70s, technical changes, political and economic integrations, other political and economic turmoils, etc.).

In conclusion, either over the longer or the shorter horizon here investigated for DE, FRA and NO, the negative aging-growth relationship finds another ‘back of the envelope’ confirmation, and also around estimates not so dissimilar from those already presented in previous RNs.