6  Conclusions

This paper has projected social expenditures in New Zealand over 50 years, based on a stochastic approach using 14 categories of social spending, decomposed by age and gender. By allowing for uncertainty about fertility, migration, mortality, labour force participation and productivity, and all categories of social spending, it has been possible to generate forecasts with accompanying confidence bands. The results show that there is considerable uncertainty and future projections have wide confidence intervals.

This uncertainty has implications for fiscal policy and the management of the Crown debt. The Fiscal Responsibility Act 1994 states that, as a principle of responsible fiscal management, a New Zealand government should ensure total Crown debt is at a prudent level by ensuring total operating expenses do not exceed total operating revenues. In the face of uncertain levels of future expenditure, the best that can be achieved is to estimate the ex ante cash budget balance needed to achieve an actual cash budget balance, at a given level of probability (Buckle, Kim and Tam 2001). To the extent that the approach adopted in this paper can improve the estimates of the probability distribution surrounding key areas of government expenditure, then it should be possible to make more reliable estimates of the ex ante budget balance needed to sustain a particular level of the Crown’s debt.

It would however be premature to conclude from these results that increased outlays will constitute a burden on society. First place, the well-being of older citizens is presumably enhanced by the receipt of health and pension benefits. Offsetting this is the loss to those who must bear higher rates of taxation to meet the transfers. We have not explored the costs to society of the deadweight losses of extra taxation.[35] Second, as argued by Guest and McDonald (2000) the absolute incomes of future generations will be higher given continued growth in productivity. In fact, even if productivity growth were to remain at its longer term average rate of 1.5%, future workers in 30 years time would be enjoying real incomes more than 50% higher than at present. The implication is that even after allowing for higher relative transfers, they would still have a level of well-being substantially greater than that enjoyed by workers today.

In focussing on the task of explicitly incorporating uncertainty into the projections, we have set aside some other potentially difficult issues. For example, while we have forecast some changes in labour force participation rates, these have relied on projections of past trends. With increased labour scarcity combined with better health and extended life spans, significant increases in the participation rates of older workers may occur. Offsetting that is the increased demand for leisure that comes with higher real incomes. While the demographic projections are based on reasonably disaggregated data, no allowance has been made for other demographic characteristics such as ethnicity or marital status.

The projection model allows for the growth of every category of social spending to differ for every age group. This constitutes a more disaggregated approach than has been applied in other studies. When combined with the extensive stochastic elements this comes at a cost of increased complexity. The empirical work is still at a rudimentary stage. We have often relied on crude assumptions such as the variability of underlying distributions remains constant over time, or across several sub-categories as in the case of health spending. Further we have paid scant attention to the question of covariances across categories of spending.

We have relied, like others, on examining the sensitivity of the projections to different assumptions. More remains to be done to explore different policy settings such as adjustments to the age of eligibility for publicly provided retirement benefits or to the criteria for certain classes of welfare benefits. Above all, even after the extensive disaggregation in the present study together with stochastic simulation of over 500 variables, the approach is vulnerable to the criticism that it fails to capture potential behavioural responses to changes in relative prices and incomes. Those advances remain as challenges.