5  Conclusions

In this paper we have focussed on the lifetime saving patterns of different cohorts. We have provided robust evidence that people born at different times, all other things being held constant, demonstrate different lifetime saving rates. Throughout we have used a definition of saving that removes from current consumption items that provide a flow of services over an extended period; ie, we have sought to use an “economic” approach to defining savings. We examined the effect of different definitions and find that the essential cohort differences in saving rates remain. Furthermore, the cohort patterns remain when we included a wide range of conditioning variables to account for within-cell heterogeneity.

There are two reasons why the finding of cohort differences is important. In the first place, it might help to explain the changes in aggregate saving behaviour. Unfortunately, in the case of New Zealand, we cannot carry this too far until we have a better understanding of how to reconcile the divergent saving rates from the national accounts and the HES.[41] Our results here suggest younger cohorts have higher saving rates than their parents, but we are not yet able to link that to the observed decline in aggregate household saving rates. It may hold out some glimmer that saving rates will not continue to decline as sharply in the future.

In the second place, if the different saving behaviour of different cohorts is due in part to the economic and social climate prevailing during their lifetimes, and in particular if different policies were operative for different cohorts, then we can at least start to better understand the impact of social and economic policies on household saving behaviour. This should contribute to better being able to predict the effect of policy changes in the future.

It is then logical to ask why birth cohort appears as such a significant determinant of saving rates. We have posited the hypothesis that an individual’s saving rate is in part a reflection of the economic and policy environment prevailing over their working life, and in particular applying during their peak saving years between ages 50 and 60. Certainly Thompson (1991) has argued that birth-year alone is an important factor in explaining the distribution of the benefits of the welfare state over the past 50 years.

We argue that individuals would tend to have a lower lifetime saving rate if for example, their working life corresponded to a period of low unemployment, greater job security, real earnings growth, dual income households, generous welfare benefits and they held expectations of an assured state pension. By comparing the lifetime environments of different cohorts one could potentially test this hypothesis.

We have made some modest progress with this complex task, by looking at some indicators of the environment facing different cohorts and finding at least a tentative association with the pattern of lifetime saving displayed by those born between 1920 and 1934 and those born before (1910-14) and after 1950. We conclude, perhaps unsurprisingly, that social welfare policies do seem to matter to the amount people are prepared to save. There is the possibility that, in fact, the policies themselves are “endogenous”; they reflect the outcomes of the political process which in turn is function of other “truly” exogenous circumstances. The generation whose working lives commenced during the Great Depression and then faced WW II were already 25-40 years old by 1950. Typically they had delayed marriage and family formation and had little net wealth. Arguably they saw state transfers as one way that would both compensate them and substitute for the loss of “saving time” -and voted for such policies accordingly.

Because of the nature of the HES we have relied on synthetic cohorts. This method has its limitations. We do not observe the behaviour of say the currently middle aged when they were young, nor the currently old when they were middle aged. As Attanasio (1998) points out, when extrapolating the lifecycle profile of each cohort for the ages at which it is not observed, it is necessary to use the information on the behaviour of other cohorts to impose some structure on the data.

Ideally, we would like to be able to use the insights about saving behaviour at the household level to help explain the aggregate saving trends observed in the national income data.[42] Unfortunately, because of the divergent trends in the two sources we clearly cannot use the former to explain the latter. Eventually, we are confident that the insights into individual saving behaviour from the micro data will be useful in explaining aggregate trends. But until we have a reconciliation and can explain the divergent series, then this task remains in the category of unfinished business.

While the results of the cohort saving behaviour (Section 3) seem both significant and robust, our attempts to provide an explanation (Section 4) are partial and tentative. This is a complex area; the saving rates we observe are the resolution of a set of forces encompassing social and cultural norms shaped by the experience of earlier generations, economic conditions over the working life, expectations of future incomes, health status and life expectancy, and myriad state interventions. Arguably, the provision of higher state benefits, or more certainty would be expected to dampen the incentive for private saving.[43] Our preliminary examination of some snippets of evidence is at least consistent with that argument. Further testing of this relationship awaits the development of long-term data series for at least the last one hundred years, together with richer models about how cultural norms and values together with expectations, blend to shape consumption and saving decisions.