5.5  Changes in Future Health Costs

The results of Section 4 show that increased health costs represent a significant share of the projected rise in social expenditures. This is particularly so if per capita social expenditures were to continue to grow at their historical rates. This section further explores the growth in health costs, and extends the basic model to allow for differential growth rates of health expenditures by age groups.

It has been stressed that the growth rates in real per capita public health expenditures have been highly variable; see Appendix Figure 4. As shown in Table 9 the selection of the historical period influences the annual average estimated growth rate in health expenditures. Most accounts of long-term changes in medicine and health make the period around World War II a watershed. It therefore probably makes sense to look at average growth rates from then until the present. As is apparent in Table 9, this gives a growth rate of around 3% per year. The corresponding average for the entire OECD for 1960-1995 is around 4.7% per year (Mayhew 2000b).[32]

Source: NZ Treasury; unpublished data from the Long Term Data Series

For present purposes it is necessary to eliminate the contribution of population ageing to this expenditure growth. Strictly speaking, this cannot be done without knowing the age-profile of spending in the past. A Ministry of Health report (Johnston and Teasdale 1999) nevertheless makes an estimate, presumably by assuming a constant age-profile. The report’s estimate is that ageing was responsible for about 0.4 percentage points in expenditure growth per year between 1977 and 1997. Mayhew (2000a: Table 2.3) derives an estimate of 0.35 percentage points per year for the OECD during the years1960 to 1995.

Subtracting an ageing component of 0.4% per year from an average growth rate of 3.0% per annum gives an adjusted rate of increase in per capita public spending on health of 2.6%. This is rounded to 2.5%, a figure in line with international estimates.[33] Mayhew (2000a: Table 2.3) assumes an underlying growth rate of 3.0% for the OECD over the period 1995-2050.

Table 10 presents a comparison of the projected level of social expenditures as a share of GDP under three different growth rates for health costs. In the benchmark estimates all social expenditures grow at 1.5%, equal to the rate of labour productivity growth. In the second case all social expenditures grow at their historical rates, which implies health costs grow at 3.1% (a figure which is not adjusted for the effect of population ageing). In the third case, health costs for all age groups grow at 2.5% while other categories of social expenditure continue to grow at 1.5% per year. Even in the third case, where only health costs are permitted to grow at a higher rate, the projected mean level of social expenditures rises substantially compared with the benchmark case; yet because of the large degree uncertainty, the differences are not statistically significant.

The projections have all assumed that any given rate of growth in health costs applied to all age groups. There is a possibility that the costs for the elderly will rise more steeply than those for younger age groups.[34] Recognising this, the projection model was modified to allow for the annual growth rates for all categories of social expenditure category to vary with age. The mean growth rates for health expenditures in the higher age groups were thereby allowed to be higher than the corresponding rates for lower ages. However, the standard deviation of the growth rate for each age group was set equal to the common value previously used for each social expenditure category (as there was insufficient information on which to base any differences). This increase in the number of distributions from which random draws are made raises a question regarding the correlation between age groups (with the values being jointly normally distributed). Indeed, the earlier projections are equivalent to an equal mean growth rate for each age group, combined with an implicit assumption that the values are perfectly correlated. The allowance for a lower degree of correlation is expected to reduce the standard deviation of the ratio of aggregate social expenditure to GDP, since the selection of higher than average values in one age group in a particular year can be partially offset by the selection of lower than average values for other age groups. For this reason, the two extremes of zero and perfect correlation between the growth rates in different age groups were examined.

Note: ^a Benchmark case

gij refers to the projected growth rate of the i-th category of social expenditure (i=1,...,14) for the j-th age group (j=1,...,19)

ρ refers to the correlation coefficient of growth rates between age groups.

The first two columns of Table 11 show the benchmark case for the two extreme values of the correlation across ages. The mean estimates remain virtually unchanged but the standard deviations are reduced. The final two columns allow growth rates to vary by age; health expenditures were assumed to grow at 2% for those from 0-64 and 3.5% pa for those 65 and over. All remaining categories of social expenditure were held to the benchmark growth rate of 1.5%. The accelerated growth rate of elderly health costs adds some 7 percentage points to the mean share of GDP with the standard deviations again varying according to the assumption about the correlation across age groups. A detailed breakdown of the components of health costs for the case of a unit correlation coefficient (third column in Table 11) is given in Appendix Table 8.