The economic impact of the New Zealand fiscal stimulus package

Abstract

Unlike many countries affected by the global financial crisis, New Zealand did not announce a formal fiscal stimulus package. However, via a series of policy announcements beginning in October 2008, by March 2009 the government budget balance had moved towards deficit by 1.6% of 2011 GDP. We interpret this discretionary movement towards deficit as New Zealand's fiscal stimulus package. The package largely comprises three policies: cuts to personal income taxes, cuts to business taxes, and infrastructure spending. We investigate the individual and joint effects of these policies using a dynamic CGE model of the New Zealand economy. We find that the package has a small positive effect on short-run employment, but at a cost to long-run real consumption. We examine an alternative package, which generates a larger short-run employment gain, for a similar long-run real consumption cost.

Keywords:

• fiscal stimulus
• dynamic CGE
• New Zealand economy

Acknowledgements

The authors thank John Ballingall, Peter Dixon, John Janssen, John Madden, Jean-Pierre de Raad, Nhi Tran, Louise Roos and two anonymous referees for helpful comments and suggestions. The views expressed in this paper are those of the authors alone.

Notes

 1. While personal income tax cuts were announced prior to October 2008, they were re-affirmed after the global financial crisis. We include them in the fiscal stimulus package since, under more benign macroeconomic conditions, these tax cuts may have been revoked.

 2. The business tax measures are described in English and Dunne (2009). These measures can be broadly classified into two groups: (1) changes in thresholds, limits and tax penalty interest rates, having the effect of permanently lowering the value of tax paid by SMEs. Examples include: increasing the GST payments threshold, increasing the GST registration threshold, increasing the minor FBT limit, and provision for tax-expensing of business-related legal costs. (2) Changes in tax thresholds and tax clauses that have the effect of delaying the payment of tax. Examples include removal of the 5% uplift rate that businesses pay in advance on provisional tax, and increasing the PAYE once-a-month filing and payment threshold from $100,000 to $500,000. These measures do not directly affect legislated ‘tax rates’. They do, however, lower effective tax rates, by lower tax collected off a given tax base. As we discuss in Section 3.1.2, we model the business tax measures as a temporary reduction in production tax rates costing $0.12b of revenue over four years.

 3. Capital and land are specific to each industry. Labour is occupation-specific, but free to move between industries.

 4. The 131 industry-specific production functions are nested constant returns to scale. The top level of the nested production functions are fixed proportions in 210 composite commodities and a primary factor composite. Each of the 210 × 131 composite commodities is a CES (constant elasticity of substitution) composite of source-specific varieties of the commodity. Two commodity sources are identified: New Zealand and foreign. The primary factor composite is a CES composite of land, labour and capital. The model thus has 55,413 (=210 × 131 × 2 + 131 × 3) cost-minimising input demand functions over source-specific commodities and primary factor inputs. See Dixon et al. (1982, p. 76–90).

 5. The representative household maximises a Klein-Rubin or Stone-Geary utility function in current-year consumption of 210 composite commodities subject to current-year income. Each of the 210 composite commodities is a cost-minimising CES (constant elasticity of substitution) composite of source-specific varieties of the commodity. Two commodity sources are identified: New Zealand and foreign. The household optimisation problem thus produces 420 (= 210 × 2) utility maximising demand functions over source-specific commodities. See Dixon et al. (1982, pp. 96–103).

 6. Units of new industry-specific capital are produced via industry-specific nested constant returns to scale production functions. The top levels of the nested production functions are fixed proportions in 210 composite commodities. Each commodity composite is in turn a cost-minimising CES (constant elasticity of substitution) composite of source-specific varieties of the commodity. Two commodity sources are identified: New Zealand and foreign. The model thus has 55,020 (= 131 × 210 × 2) cost-minimising demand equations relating to inputs to capital formation. See Dixon et al. (1982, pp. 94–96).

 7. That is, each of the 210 commodity composites demanded by each of the model's agents is a cost-minimising CES (constant elasticity of substitution) composite of imported and domestic varieties of the commodity. See Dixon et al. (1982, p. 69).

 8. Commodity-specific export demand functions are constant elasticity of demand. See Dixon et al. (1982, pp. 104–105).

 9. With 210 commodities and two sources of supply, the model recognises 420 source-specific government commodity demands. These demands are modelled as either exogenous or indexed to aggregate public consumption spending. See Dixon et al. (1982, p. 105).

10. Direct taxes are levied on labour and capital income. Indirect taxes are potentially payable on every source-specific commodity flow to each agent (=210 × 2 × (131 + 131 + 1 + 1) indirect taxes), production by industry (=131 indirect taxes), imports by commodity (=210 indirect taxes), and exports by commodity (= 210 indirect taxes). However, indirect taxes on most sales are zero. Given New Zealand's VAT system, the bulk of the indirect tax burden falls on commodity-specific flows to household consumption. See Dixon et al. (1982, pp. 108–117).

11. We calculate the average unit cost of production for each of the model's 131 industries. Given the constant returns to scale technology that characterises each industry's production function, the average unit cost is also the marginal cost. The competitive zero pure profits condition that output price is equal to marginal cost is enforced via an assumption that the average cost of each industry's output is equal to the average price received on the commodities sold by each industry. See Dixon et al. (1982, pp. 108–110).

12. In both the short-run and long-run, imports are available in elastic supply at given world prices. Short-run market clearing is imposed on sales of domestic commodities and industry-specific physical capital markets. As discussed in Section 2.1, short-run labour markets are characterised by stickiness of the real consumer wage (see Dixon & Rimmer, 2002, pp. 204–210). Long-run market clearing is imposed on all markets other than commodity-specific import markets, where, again, import supply is assumed to be elastic at given world prices. See Dixon et al. (1982, pp. 122–125).

13. Five types of margin are modelled: wholesale trade, retail trade, road freight, other road transport services, and rail and sea transport services. In the absence of exogenous movements in technical change, demands for margin services are assumed to be a fixed proportion of the physical commodity flows that they facilitate. See Dixon et al. (1982, pp. 106–108).

14. Industry-specific capital accumulation follows the stock-flow accounting rule that capital in year t is equal to depreciated capital in year t-1 plus gross fixed capital formation in year t-1. See Dixon and Rimmer (2002, pp. 4–5, 154–155).

15. See Dixon and Rimmer (2002, pp. 190–195).

16. See Dixon and Rimmer (2002, p. 158).

17. See Dixon and Rimmer (2002, pp. 204–210).

18. NZIER (2009a) forecast annual average growth in real GDP of –1.5% in the year to March 2009, the lowest since 1991. NZIER (2009a) anticipates growth to remain sub-trend at 1.1% in 2010 before stabilising in 2011 and 2012 at rates of 2.8% and 3.0% respectively. Forecasts for the other macro aggregates similarly reflect likely impacts of the global financial crisis. The NZIER (2009a) forecasts are broadly in line with consensus forecasts from other major forecasters around New Zealand (see NZIER, 2009b).

19. We exclude land from BOTE equation (E2) since the stock of land (which is unchanged in both our base-case and policy simulations) exerts little influence on our FSP simulation results.

20. In linking current-year consumption to current-year disposable income via equation (E3) we have: (a) excluded the possibility of Ricardian-equivalence in short-run savings behaviour; and (b) assumed households adjust consumption to current income. We think (a) is justified on two grounds. First, empirical studies have yet to reach a consensus on the existence or size of Richardian-equivalence savings effects. Secondly, we do not find the microeconomic assumptions required for Ricardian-equivalence, such as household adjustment of current savings to anticipated future tax liabilities, to be compelling descriptions of actual behaviour. We justify (b) on the grounds that since our model is annual, rather than quarterly, we expect much of household consumption adjustment to changes in post-tax income to occur within-period.

21. Adding equations (E4) and (E5), the tax terms cancel, giving HINC + GINC = Y × q(TOT) − [NFLG + NFLH] × R = GNP. BOTE does not include autonomous net foreign transfers (MONASH-NZ does). Hence in BOTE, GNP is the same as GNDI. The basis of equations (E4) and (E5) is the nominal GNDI identity NGNDI = NGDP − [NNFLG + NNFLH] × R, where NGNDI is nominal gross national disposable income, NGDP is nominal GDP, R is the rate of interest on net foreign liabilities, and NNFLG and NNFLH are the nominal net foreign liabilities of the public and private sectors respectively. Dividing through by the deflator for consumption, PC, we have: GNDI = Y × PY/PC − [NFLG + NFLH] × R, where PY is the GDP deflator and all other variables are as defined in Table 3. PY/PC is a positive function of the terms of trade, q(TOT). With inclusion of tax terms, this identity can be split into two components: claims on GNDI by the household sector (HINC) and the public sector (GINC), being equations (E4) and (E5) respectively.

22. Equation (E10) is based on the profit maximising first order condition that the value of the marginal product of capital equals the rental price of capital, noting that the production function is constant returns to scale. See Dixon and Rimmer (2002, p. 244).

23. Equation (E11) is based on the profit maximising first-order condition that the value of the marginal product of labour equals the wage, noting that the production function is homogeneous of degree 1.

24. Adding equations (E15) and (E16) provides ΔNFLH + ΔNFLG = I − HINC + APC HINC + G − GINC. Via equation (E3), ΔNFLH + ΔNFLG = I − [HINC + GINC − C − G]. From Note 21 above we know HINC + GINC = GNDI. Hence ΔNFLH + ΔNFLG = I − [GNDI − C − G]. GNDI − C − G is national savings, S. Hence equations (E3), (E4), (E5), (E15) and (E16) imply that the change in national net foreign liabilities equals I – S.

25. Further details on the model's labour market adjustment mechanism are available in Dixon and Rimmer 2002 (pp. 205–210).

26. We base this on the speed of employment adjustment in KITT, the Reserve Bank of New Zealand's quarterly macro-econometric model. This model displays an employment adjustment path in which employment deviations are largely eliminated over 20 quarters. See Benes et al. (2009).

27. In equation (E9), Λ can be interpreted as a normal rate of return. Hence, via equation (E9), investment will be above (below) base-case so long as current rates of return (ROR) are above (below) normal rates of return. The annual capital accumulation process is described by equation (E14). Capital accumulation (depreciation) gradually drives convergence of actual and normal rates of return. In column (2) of Table 3, row (E9), we describe the long-run outcome of this process as effective exogeneity of ROR. Long-run movements in capital require long-run adjustment of investment to maintain the capital. We describe this via long-run exogenous status of Ψ in equation (E10). Equation (E17) ensures that wage rates in the policy-case will rise (fall) relative to the base-case so long as employment in the policy case is above (below) base-case employment. This process continues until policy-case employment returns to its base-case level. While in the MONASH-NZ simulation employment (L) remains formally endogenous at all times, by driving L back to base-case, equation (E17) makes L act like a long-run exogenous variable. Hence, in column (2) of Table 3, we describe L as exogenous.

28. A potential source of gain from the business tax measures, not modelled in our paper, is a lowering of the compliance cost burden related to New Zealand's business taxation. A further study, one focused on the compliance cost consequences of the 2009 business tax measures, might use studies such as Colmar Brunton (2005) and Sandford and Hasseldine (1992) as starting points for the construction of plausible estimates of efficiency gains arising from a lower compliance burden under the new tax arrangements.

29. We model the rise in public infrastructure spending as an increase in public consumption of output of non-dwellings construction services. Hence, in our macroeconomic indicators, we see a positive deviation in public consumption spending (see Figure 6). From a national accounting perspective, the additional infrastructure spending might be assigned to economy-wide gross fixed capital formation rather than public consumption. Beyond this national accounting matter, the assignment of the additional infrastructure spending to one national accounts demand aggregate or another has no implications for our modelling.

30. As Giesecke et al. (2008) note, this is consistent with rates of return on Australian public infrastructure capital found by Demetriades and Mamuneas (2000).

31. The decomposition is calculated by running four simulations: the personal tax measures alone, the business tax measures alone, the infrastructure spending alone and the combined simulation. Since the model is non-linear, the sum of the impacts of the three shocks alone is slightly different from the joint effect of the three shocks considered simultaneously. We report the (small) difference as ‘residual' in Figures 1 – 11.

32. The returns on the new infrastructure contribute approximately 0.14 (= 0.9 × 0.15) percentage points of the long-run real GDP deviation. That is, 0.9 percentage points of GDP (representing the size of the infrastructure spending) generating an annuity of 15 cents in the dollar (see Section 3.1.3).

33. Real consumption is approximately 60% of New Zealand's GDP. By 2011, the two tax measures are worth 1.3% of GDP ( Table 1). Hence the two tax measures directly contribute approximately 2.2 percentage points (= 1.3 / 0.60) to the 2011 real private consumption deviation.

34. =3.5 – 2.2 – 0.3.

35. Fiscal impulse indicator: see Section 1.

36. Compare the year 2017 outcomes for the cumulative FII: final row of Tables 1 and 2.

37. With employment returning to the base-case in the long-run, the long-run incidence of the tax is borne by employees. If we were to leave the employer labour tax cut permanently in place, the real pre-tax wage would rise by the full amount of the tax cut, returning employment to the base-case. The start of this process is apparent in the wage and employment results for the first three years of the simulation period ( Figure 12).