5  The Treasury’s long-term fiscal model and fiscal gap calculations 

The budget identity of equation (1) and the subsequent fiscal gap of equation (8) are stylised. The formulation of the Treasury’s long-term fiscal model (LTFM) captures all the key variables required for the fiscal gap calculation (r, g, b_t, and pb_s). The LTFM can trace out the path of relevant fiscal balance and stock variables under alternative economic, demographic and fiscal assumptions. The model is based on Generally Accepted Accounting Practice (GAAP) and incorporates a fuller set of assumptions around expenses, revenues, assets and liabilities than embodied in equations (1) and (8).

It is important to note the following features embodied in the LTFM when interpreting estimates of the fiscal gap generated using the LTFM. First, although official population and labour force projections are made out to 2101, the LTFM typically does not project fiscal balances and stocks beyond 2051 (year ending June). This is largely because outstanding student loans, which will influence the borrowing requirement, are not currently projected beyond 2051.

Second, the LTFM projections start at the end of a set of short-term economic and fiscal forecasts. The Treasury’s short-term forecasts typically assume that the economy is at or close to trend GDP at the end of the forecast period. As a result, the projected long-term fiscal position is structural and ignores the effects of the business cycle.

Third, the LTFM does not incorporate any feedback between the fiscal position and macroeconomic variables such as productivity or interest rates.

Fourth, the LTFM assumes a positive inflation rate and results are generally expressed as ratios-to-nominal GDP. As discussed in Section 5.3 below, the effects of inflation on the tax take (i.e., fiscal drag) are generally excluded.

Finally, the version of the LTFM used here is deterministic in that each input (e.g., labour productivity growth) is given a single value for each year and the model produces a single outcome. Lee and Edwards (2001) use a stochastic approach to modelling the fiscal effects of ageing in the US. They use time series methods to fit stochastic models for input variables such as labour productivity growth, real interest rates and demographics. In most cases they constrain the central path for each input variable (i.e., its long-run mean) to match the assumptions of agencies like Social Security and the CBO. Historical information provides estimates of the variance of the error term around the imposed mean. The stochastic projection uses random draws to assign values to each input in each year. When combined with other components (e.g., cost drivers) these generate a stochastic outcome. The projections are run repeatedly (1000 times) and the frequency distribution for the outcomes is used to generate a probability distribution of outcomes. Recent CBO analysis of the finances of US Social Security uses a similar approach, attaching ranges of uncertainty for inputs such as mortality, unemployment, inflation and the real interest rate (CBO, 2001). Modelling uncertainty around inputs into the Treasury’s LTFM using stochastic techniques could complement the more typical “what if?” analysis. The “what if?” analysis generally considers changes to the trend in input variables as well as changes to policy parameters such as the indexation of spending to real wages, or the target debt level. Stochastic analysis requires a careful assessment of the interdependence (covariance) of input variables.

5.1  Demographics

Demographic projections are a key input to the LTFM and are summarised in Table 2 (details on the assumptions are in the Appendix).

Note: 1981 is historical and based on a De Facto population definition. From 1999 the definition is Resident population. Projections are Series 4 (see Appendix for details).

Source: Statistics New Zealand

The elderly dependency ratio embodied in the LTFM increases from around 18 (per 100 people in the age group 15-64) in 2001 to 43 in 2041. The increasing proportion of the population aged over 65 is due partly to the “baby boom” generation passing into higher age groups, and also to the effects of increasing longevity and falling fertility. By 2061 the elderly dependency ratio is projected to reach 47 and there is less change after this point.

The effect of the increase in the elderly dependency ratio on the total dependency ratio is partially offset by the decline in the child dependency ratio. The fall in the child dependency ratio dominated during the 1980 to 2000 period. Thereafter, the increase in the elderly dependency ratio dominates. Nevertheless, the total dependency does not reach the 1981 level until around 2021. However, the composition is quite different.

Raw dependency ratios do not necessarily capture “economic dependence” – some people aged 15-64 are not in employment, some aged over 65 are. The LTFM allows for this by incorporating projections of the labour force that are derived using age group specific labour force participation rates. These labour force projections are combined with assumed long-run rates of unemployment and productivity growth to determine real economic output. In addition, the LTFM builds in the change in the composition of the population and the fact that government spending differs across age groups.

5.2  Expense assumptions

Long-term fiscal scenarios generated by the LTFM use a “bottom-up” approach to expense and tax revenue assumptions.[9] Expenses are projected by functional classification (e.g., social security, health, education, law and order). Some of these classifications are influenced by demographic changes and are projected on an age-related basis. For example, health expenses are influenced by numbers in particular age groups and the amount spent on each age group. Per person expenses for health, education, and social welfare transfers are assumed to increase in pre-specified real per capita terms. This increase generally equals the assumed rate of increase in real wages and hence growth of labour productivity. There are several major items that are sensitive to population projections and are expected to have an important effect on the expenditure profile: New Zealand Superannuation (NZS), Health, Education, and Social welfare.

The projected number of retirees and legislation regarding entitlements determines expenses for public pensions (that is, NZS). Payments of NZS are currently linked to nominal wages and so to real wages and labour productivity.

A number of factors make the projection of health expenses uncertain. The LTFM assumes static cost weights for health spending.[10] There is uncertainty as to whether longer life expectancies will see extended periods of health care at higher age brackets or whether costs will be shifted to later years of life (the “proximity to death” issue). Static age-related spending profiles may overestimate the impact of ageing on health spending. On the other hand, the LTFM assumes that the weights increase in line with real wages and labour productivity. This may underestimate future health costs if the long-term elasticity of health spending to per capita income is greater than one. Although technological advances may work to reduce medical costs for specific procedures, this may result in the procedures being applied to a greater proportion of patients and thereby increasing total costs (see Lee and Skinner, 1999).

Education is the other major expenses area influenced by demographics. Social welfare transfers (e.g., unemployment) are influenced by the rate of “take-up” and grow with wages (so as to avoid a significant decline in benefits-to-wages over the long term). Expenses not directly influenced by demographics, such as core government, law and order, and defence grow at a specified real rate.

Compared to NZS, which has a legislated link to wages and an age of eligibility, the assumptions for other areas are more problematic. There is uncertainty surrounding demographic and economic assumptions, technological change, behavioural responses and the role of future governments in providing particular goods, services and transfers. Given their labour content, and in the absence of major changes to input structure (e.g., capital-labour ratios), health, education and other expenses will in the long-term be influenced by wage growth.

5.3  Tax revenue assumptions

In “bottom-up” mode average effective tax rates are assumed to remain constant. Effects on the tax take arising from the interaction of rising incomes and the progressive tax system are therefore not modelled (i.e., there is no “fiscal drag” or “bracket creep”).[11] In the presence of inflation, this assumption is equivalent to assuming labour income tax brackets are inflation-indexed. Possible changes to the tax base are implicitly offset by revenue neutral policy changes (tax scales are implicitly “indexed” because there is inflation).[12]