Debt-financed fiscal stimulus in South Africa

Abstract

Levels of government debt-to-GDP have been rising substantially for over a decade in emerging market economies such as South Africa, which has led to much debate around the implementation of large-scale debt-financed fiscal stimulus programs in response to the economic fallout of the global covid-19 pandemic. Debt-financed fiscal policies directly stimulate aggregate demand through government expenditure or tax cuts, but their effectiveness is highly dependent on direct crowding-out of private sector expenditure, spill-over effects to the private sector through a higher risk premium on interest rates, and the interaction between fiscal policy and monetary policy. Using an open-economy dynamic stochastic general equilibrium model for South Africa, we identify the effect of six different fiscal policy instruments on short-term and long-term interest rates. These disaggregated expenditure and revenue shocks raise long-term real yields between 18 and 29 basis points, but there are non-negligible differences in the dynamic responses to each fiscal instrument. Our main findings suggest that, in the context of fiscal sustainability, an investment-driven debt-financed fiscal stimulus programme would reduce the government debt-to-GDP ratio, especially in periods of economic slack when monetary policy would typically be more accommodative. In fact, since the global financial crisis, monetary policy has reduced the burden of fiscal adjustment in response to rising debt and a rising risk premium. But further shocks to the risk premium could offset any gains from the current stance of monetary policy (for example, a credit rating shock raises the long-term government bond rate 155 basis points).

Keywords:

• Crowding out
• fiscal policy
• fiscal sustainability
• government debt
• interest rate
• monetary policy
• risk premium
• credit rating

JEL CLASSIFICATION:

• E43
• E6
• H3
• H6

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 A sustainable fiscal policy means that the public debt-to-GDP ratio remains stable, at or below some benchmark, over the medium to long term. Fiscal sustainability also implies that government should maintain a primary surplus if the real interest rate on government debt exceeds the real economic growth rate (Fourie and Burger, 2003). As this is the case for South Africa, going forward, the government cannot sustain a fiscal position of dissaving.

2 In a comprehensive survey, Ramey (2019) points out the ‘dearth of research’ on the macroeconomic effects of fiscal policy up to the global financial crisis. Much can be said about the effect of DFFS on interest rates—a key transmission mechanism for fiscal multipliers (see, e.g., Ganelli and Rankin, 2020) and fiscal sustainability (Calitz et al., 2014; Fourie and Burger, 2003).

3 The ‘term premium’ in long-term government interest rates serves as a proxy for the economy-wide (or country) risk premium (see, e.g., Erasmus and Steenkamp, 2022a, 2014b; Soobyah and Steenkamp, 2020a, 2014b). One key reason for this is that corporate yields are largely determined by (i.e., benchmarked against) sovereign yields in South Africa (Peter and Grandes, 2005).

4 The South African literature has focused predominantly on the effect of government debt on growth, the effect of interest rates or spreads on the macroeconomy, and the spillover effects of credit ratings or sovereign risk (see, e.g., Fedderke, 2020; Mhlaba and Phiri, 2019; Mothibi, 2019; Peter and Grandes, 2005; Soobyah and Steenkamp, 2020a, b).

5 Peppel-Srebrny (2021) raises a similar point to motivate their econometric approach and contribution to the literature. In fact, the author finds no existing literature that empirically tests the relevance of the composition of government budget deficits for government bond yields.

6 In addition, Ramey (2011) shows that anticipation (‘news’) matters for DFFS. A DSGE model incorporates agents that are forward-looking and it can control for anticipated fiscal policy (‘fiscal news’) shocks (Born et al., 2013).

7 In Section 4 we show that the responses of monetary policy and the risk premium to DFFS are particularly important, whereas direct crowding in/out is relatively less important for the efficacy of fiscal stimulus.

8 Since tax revenue data is used, as opposed to tax rates, a ‘tax cut’ is indistinguishable from a ‘fiscal revenue shortfall’ or even changes to the tax base. Because the results indicate that ‘tax cuts’ are contractionary, the term ‘fiscal revenue shortfalls’ is used to broadly capture all of these intensive margin effects. One way to identify exogenous tax cuts would be to separately identify historically administered changes in tax rates (see, e.g., Kemp, 2020b).

9 For example, in a small open economy like South Africa, a sovereign debt crisis typically leads to capital outflows (and therefore a large exchange rate depreciation), financial sector instability, and tighter credit conditions (through higher borrowing costs and stricter creditworthiness) which severely affects the efficient functioning of the real economy.

10 The term spread is measured by the difference between short-term and long-term interest rates on government debt, typically the three-month and ten-year bond yields.

11 The dynamics (supply) of government debt (both stock and flow) can also influence monetary policy (because government debt provides collateral for liquidity operations and it expands the consolidated government balance sheet, which is inflationary) and financial stability policy (because government debt serves as a high-quality liquid asset, it is typically exempt from risk-weighted capital requirements, and it is a benchmark for market pricing of debt).

12 Notably, with inflation, real GDP per capita (detrended), and the long or short nominal interest rate (depending on the dependent variable).

13 The selection of the seven non-fiscal variables follow from Smets and Wouters (2007). Data for the ten-year government bond yield is excluded from the DSGE estimation due to the limited number of shocks in the model (i.e., eight shocks to match eight variables). The implied long-term interest rate is determined by the risk premium, which has an endogenous component based on the fiscal variable such that higher debt raises sovereign debt risk and an exogenous component that follows an AR(1) stochastic process.

14 Although standard model selection criteria prefer an unrestricted model with four lags over both lower order lags and restricted versions of the model, we show results for a restricted VAR(1) model that matches the the DSGE model equations. One exception being that we allow for lagged effects from the fiscal variable G on all of the other variables in the system. For the fiscal DSGE model analysis in the following section, the individual-level paths for debt and output (GDP) are determined by the model to match the observed debt-to-GDP ratio. For the comparison exercise in this section, we therefore remove the fourth degree polynomial trend from government debt-to-GDP. The more standard alternative transformation to log-difference real government debt provides similar results in the unrestricted VAR(4) model. All of these results are available on request.

15 For a comprehensive analysis of alternative identification approaches for fiscal policy in econometric models, see Kemp (2020a).

16 Historical evidence for South Africa suggests that fiscal policy has responded counter-cyclically to debt accumulation since the 1960s (Burger et al., 2012; Calitz et al., 2014), but over the business cycle (as measured by GDP) the findings reject any consistent counter-cyclical policy (Plessis et al., 2007; Plessis and Boshoff, 2007). Assuming government debt and output grow at constant rates, the log-linear detrended series are correlated –0.84 (–0.74) in nominal (real) terms over the sample period 1994Q1 to 2018Q4. Trend quarterly growth for both government debt and output is 2.46 per cent in nominal terms, and 0.74 per cent and 0.73 per cent, respectively, in real terms (both deflated using the GDP deflator). While fiscal interventions may not systematically respond to output, the data clearly suggests strong counter-cyclical behaviour of debt to output. Kemp and Hollander (2020) show that government expenditure responds counter-cyclically to both real debt and real output, but typically more strongly to debt, which is in line with optimal stabilization policies (Hollander and Havemann, 2022). There is therefore a clear policy trade-off between debt and output stabilization.

17 Figure A.2 plots the estimated implied long-term interest rate with the three-month Treasury bill rate, the ten-year government bond yield, and the effective interest rate derived from the ratio of debt-service costs to outstanding gross debt (annualized and based on deseasonalized monthly data). The implied rate, which is consistent with the government budget balance, tracks the market rates and effective rate derived from debt-service costs very closely. After some minor corrections to the original replication code, the estimated version of this model still follows closely to that of Kemp and Hollander (2020). Since the DSGE model is not part of the contribution of this paper, the full model description, prior selection, estimation diagnostics, and posterior parameter estimates are available upon request.

18 Public sector (government) investment is the sum of fixed capital formation by general government and public corporations.

19 Using panel data for 31 EU and OECD economies from 1990, Peppel-Srebrny (2021) shows that primary budget deficits due to higher government current (non-investment) expenditure (such as social benefits and employee compensation) intensify sovereign risk in the form of significantly higher bond market rates. In contrast, a higher deficit due to investment, on average, leads to lower long-term government bond yields because markets tend to price in a positive return on investment, which is favourable for growth. These results obtain after controlling for the short-term interest rate and under various specifications, which provides strong evidence that this mechanism operates primarily through the sovereign risk premium. We theoretically identify this transmission mechanism in Figure 6a,b.

20 It is again worth noting that, based on how effective tax rates are calculated, it is unclear what the dominant effects of tax policy and tax efficiency and collection are. We take an agnostic and model-consistent view that the identified shocks (to the effective tax rates) are exogenous changes to realized tax revenue. This could incorporate several legislative and non-legislative effects due to, for example, marginal tax rate changes, bracket creep, tax base changes, changes to the elasticity of taxable income, anticipated tax policy, and unanticipated tax revenue windfalls and shortfalls. The effect of negative shocks to tax revenue are therefore contractionary due to lower government expenditure, higher debt servicing costs (through both debt accumulation and a higher risk premium), and relatively weak tax multiplier effects on private consumption and investment.

21 From 2001/02 to 2010/11, Calitz et al. (2016) find that forecast errors for revenue projections account for 71.2 per cent of the errors in the budget balance-to-GDP ratios. Such inaccuracies appear commonplace internationally. In fact, official South African forecasts of budget balances have been relatively accurate by international standards.

22 In the estimated version of the model we account for a time-varying inflation target. Since it is exogenous, it is excluded from the equation.

23 For the estimated model, the parameter values for $\{{\varphi }_R,{\varphi }_{\pi },{\varphi }_{\Delta y}\}$ are $\{0.87,1.57,0.38\}.$

24 See Figure 4 for the inflation response.

25 It is assumed that valuable government consumption enters the households’ utility function in a non-separable way. This feature has two important implications. First, under this specification shocks to government consumption affect optimal private consumption decisions directly. This stands in contrast to the indirect wealth effect associated with separable government consumption. Second, the feature implies that it is theoretically possible to generate co-movements between private and government consumption, conditional on the estimated degree of complementarity.

26 Public capital augments private production at no direct cost for the firm. Therefore, it can be interpreted as an externality to the private productive sector. Financing of public capital is not factored into the cost accounting of firms as it takes place through the general tax system. Private physical capital services (i.e., the usage of private capital in production) includes variable capital utilization. Finally, the stock of public capital is assumed to grow at the same speed as private capital services along the balanced growth path of the model. The log-linear expression for aggregate capital used in production is: ${\tilde{k}}_t={\alpha }_K^{\frac{1}{{\nu }_K}}{(k^s/\tilde{k})}^{\frac{{\nu }_K-1}{{\nu }_K}}{k}_t^s+{(1-{\alpha }_K)}^{\frac{1}{{\nu }_K}}{(k_G/\tilde{k})}^{\frac{{\nu }_K-1}{{\nu }_K}}{k}_{G,t}$. The log-linear expression for aggregate consumption (Equation 11(11) $${\tilde{C}}_{h,t}={\left({\alpha }_G^{\frac{1}{{\nu }_G}}{\left(C_{h,t}\right)}^{\frac{{\nu }_G-1}{{\nu }_G}}+{\left(1-{\alpha }_G\right)}^{\frac{1}{{\nu }_G}}{\left(G_t\right)}^{\frac{{\nu }_G-1}{{\nu }_G}}\right)}^{\frac{{\nu }_G}{{\nu }_G-1}}$$(11) ) follows analogously.

27 The estimated effects of the risk premium on output and inflation are broadly comparable to the impulse responses of term premium shocks estimated by Soobyah and Steenkamp (2020b) (see Figures 14 and 16).

28 The contributions of monetary policy and the risk premium to output, debt, output growth, and debt growth are similar in magnitude for both shocks (see Table A1).

29 We do not account for possible non-linearities with respect to credit ratings or debt levels (threshold effects).

Additional information

Funding

This study was supported by the United Nations University World Institute for Development Economics Research.