Saving and growth: another look at the cohort evidence

Abstract

In recent years, as longer time series of cross-sectional household surveys have become available, it has become possible to look at the consumption and saving behavior of birth cohorts in a number of developing and developed economies. The cohort evidence is singularly appropriate for the analysis of life-cycle models of consumption because, at least in its simpler forms, the life-cycle hypothesis (LCH) predicts that the cohort average of the logarithm of consumption in any year can be additively decomposed into a time-invariant cohort effect and an age effect, both of which can be readily recovered from the cohort data by linear regression on dummy variables. The results of these cohort level analyses have not been favorable for the LCH interpretation of the international correlation between growth and saving. According to this, higher rates of economic growth drive up rates of national saving by expanding the lifetime resources of younger generations, who are saving, relative to the lifetime resources of older generations, who are dissaving. While it is typically possible to interpret the cohort results in a way that is consistent with the LCH, it is a good deal harder to rescue the prediction that higher growth means higher national saving rates, or at least that the effect is large enough to be consistent with the international relationship in which a one percentage point increase in the rate of per capita growth is associated with a roughly two percentage point increase in the saving rate. In Section 1, we review existing cohort studies from a range of rich and poor countries, and also present new results for Indonesia. This evidence shows no strong negative relationship between saving rates and age, so that when higher growth redistributes lifetime resources towards the young, the effect on savings is modest, and in some cases even negative.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 25 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.