WPS8712 Policy Research Working Paper 8712 When the Cycle Becomes the Trend The Emerging Market Experience with Fiscal Policy during the Last Commodity Super Cycle Rashaad Amra Marek Hanusch Charl Jooste Macroeconomics, Trade and Investment Global Practice January 2019 Policy Research Working Paper 8712 Abstract Fiscal buffers have shrunk across the world. This paper When the commodity boom ended, it became apparent that argues that limited fiscal room in emerging market econ- countries had saved less than they should have, and that omies today is partly due to the commodity super cycle of fiscal policy had, perhaps inadvertently, been pro-cyclical. It 2000–15. The super cycle created the mirage that economic left countries with depleted fiscal buffers and large budgets performance had structurally improved, mistaking a long, when the cycle came to an end, limiting room for fiscal commodity-fueled uptick in the business cycle for higher stimulus when needed. The paper illustrates the argument trend growth. This thinking supported fiscal expansions. with reference to the South African experience. This paper is a product of the Macroeconomics, Trade and Investment Global Practice. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://www.worldbank.org/research. The authors may be contacted at mhanusch@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team When the Cycle Becomes the Trend: The Emerging Market Experience with Fiscal Policy during the Last Commodity Super Cycle Rashaad Amra*, Marek Hanusch+ and Charl Jooste+ JEL Classifications: C54, E02, E62, F41 Keywords: Fiscal policy, commodity super cycle, business cycles * Parliamentary Budget Office, South Africa + Global Practice for Macroeconomics, Trade and Investment, The World Bank _____________________________ The views expressed do not necessarily reflect the views of the Parliamentary Budget Office or The World Bank. The authors would like to express their thanks for various contributions and comments to Asli Senkal, Dinar Prihardini,  Ha  Nguyen,  Sudarshan  Gooptu,  Vincent  Dadam,  and  Sebastien  Dessus.  The  paper  draws  on quantitative text analysis conducted by Slava Mikhaylov. 1. Introduction Fiscal stabilization works to effectively maximize the effects of fiscal policy on output while avoiding a sustainability trap (both from a liquidity and solvency perspective). If governments run large deficits and accumulate debt it detracts from future spending as financing deficits requires revenues to service debts and/ or spending cuts. Consequently, the “when and how” to spend are critical for fiscal policy plans. During economic downturns, fiscal policy can be used to cushion some of the adverse economic shocks and help stabilize the business cycle. However, the scope of economic downturn intervention is partly determined by how much fiscal space government created during economic booms. A large saving during booms implies that governments have more room to intervene during recessions. However, fiscal policy can destabilize the business cycle by being procyclical – that is, expenditures increase when there are booms and decreases when there are recessions. This type of policy may prolong business cycles and may create fiscal sustainability concerns, limiting the generation of fiscal buffers. In this paper we analyze the fiscal stance of selected emerging markets and developing economies (EMDEs), and how the commodity super cycle affected it. South Africa will serve as an illustration of our argument. In a nutshell, we argue in this paper that the long commodity super cycle created the mirage among policy makers that the cycle had become the trend, in other words, that growth had structurally—rather than temporarily—increased. This allowed for higher expenditures, commensurate with higher economic potential. This miscalculation inadvertently supported pro-cyclical policy, as expenditures rose with the (long) cycle rather than with potential. When the cycle ended, it became clear that governments had saved less than they should have to create the fiscal buffers needed to stimulate their economies. Commodity super cycles are not infrequent. According to World Bank (2015), four commodity super cycles can be distinguished since the 19th century: from 1894 to 1932, from 1932 to 1971, from 1971 to 1999, and, most recently, from 2000 to 2015. We focus on this most recent commodity super cycle. Figure 1 illustrates the magnitude of the cycle for key commodities. Figure 1: The 2000-2015 commodity super cycle (index, 1990=100, real 2010 US dollars) 400 350 300 250 200 150 100 50 0 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 Energy Agriculture Metals & Minerals Precious Metals Source: World Bank Pink Sheets, 2018. Note: Agriculture includes beverages, food, and raw materials (timber and other). EMDEs managed to significantly reduce public debt levels during the early 2000s, partly with the additional resources associated with the commodities boom. Public debt fell from 52.1% of GDP in 2002 to 33.9% by 2008 (Figure 2). Then the global financial crisis struck, followed by a global economic downturn. Commodity prices dipped briefly in 2009, but returned to elevated levels until 2  about 2015 (led by a plummeting oil price). In response to slow growth, debt levels started rising again, reaching 49.0% by 2017. A large number of EMDEs saw their sovereign credit ratings downgraded since about 2016. South Africa lost its investment grade credit rating in early 2017. Fiscal buffers are depleted in many EMDEs, making them vulnerable to an economic downturn. Rebuilding those buffers is critical (World Bank, 2018), as is ensuring that the mirage of the cycle is the trend does not undermine fiscal policy in the future. Great care is needed in producing economic forecasts, especially during commodity super cycles, paying close attention to distinguishing structural from cyclical factors. Figure 2: Public debt in advanced and emerging economies (gross public debt in percent of GDP) 120 100 80 60 40 20 0 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 Advanced economies Emerging market and developing economies Source: IMF World Economic Outlook, April 2018. This work builds previous research that looks at long cycles and fiscal policy (Strawczynski and Zeira, 2013) as well as commodity prices and fiscal policy (Cespedes and Velasco, 2013). The innovation of this paper is to understand the stance of EMDE countries using the combined the cycle is the trend methodology while controlling for commodity price cycles. In addition, we argue that a fiscal stance need not be fixed over time (something that is commonly assumed). We extract time varying parameters that measure the fiscal stance over different periods. In the next section we use the simple theoretical framework of Strawczynski and Zeira (2013) to illustrate how a structural shift in gross domestic product (GDP) translates into higher expenditures. If a cyclical uptick in growth is mistaken for an increase in trend growth, this increase in expenditure will, unintentionally, constitute pro-cyclical fiscal policy. We then turn to an empirical investigation of fiscal policy in EMDEs during the last 20 years, demonstrating that the latest commodity boom has indeed lengthened business cycles, posing one possible explanation for the pro-cyclical policy stance in our sample of EMDEs. We then zoom into the South African experience, as the commodity boom had a particularly strong impact on lengthening the South African business cycle. We show how this supported overly optimistic growth forecasts, being one of the likely factors explaining an increase in expenditures and a reduction in fiscal buffers that limit South African policy space today. 2. A Simple Framework of the Government’s Fiscal Policy Stance The model of Aguiar and Gopinath (2007) explicitly decomposes shocks into trends and transitory shocks and illustrates that it is optimal for governments to stimulate the economy and smooth out consumption if there is a negative transitory shock. Similarly, it is optimal for government to reduce 3  spending if there is a negative trend shock. The optimal decision here requires knowledge about understanding the structurally permanent vs. transitory aspects of a shock. This has interesting modeling repercussions. Leeper et al. (2013), as an example, show that foresight about economic fundamentals can create rational expectations equilibria with a non-fundamental representation. This implies that it is difficult to recover the structural shocks to which agents react to. The long-commodity cycle and its effect on creating a long economic cycle might have been viewed as a permanent shift in the trend of output and hence led to larger expenditures than what would have been typical if governments thought of movements in output as transitory. Or in the Leeper et al. (2013) example, it would imply that the tools used by government meant that it was unable to correctly pick out the structural vs. transitory shocks – and hence leading to a non-fundamental representation or under-identification. The model illustrated here focuses the optimal decision by government when movements in output are permanent (the model thus abstracts from the cyclical nature of spending). We use a simple model to demonstrate the interaction between GDP and fiscal policy. Barro (1979) and Barro (1990) illustrate the optimal time sequence of choosing taxes, conditional on an expenditure path that is conditional on taxation. The consumer maximizes the following discounted utility function: 1 1 Where measures the relative risk aversion or is the inverse intertemporal elasticity of substitution, is aggregate consumption and is the time preference rate. The economy produces output, with capital ( ), labor ( 1 , exogenous TFP () and government expenditures ( ) using a Cobb-Douglas production function: Where 1 are the income shares of capital and the government. All government expenditures are financed via an output tax: Note that → 1 1 The dynamic budget constraint is given as: 1 To solve for the optimal conditions, we set up the Hamiltonian: 1 1 1 The first order conditions are given as: ′ (1) : 1 1 (2) We can substitute the first FOC into the second: 4    ′ 1 1 (3) Differentiate (1): ′′ ′ (4) Plug (4) into (3): ′ ′ 1 1 We can simplify this expression by dividing by and ′ : 1 1 ′ The condition above (using the functional form of the utility function) determines consumption growth: 1 1 Noting that in steady state 1 1 Using the condition above and the government’s budget constraint, we can solve for the optimal tax rate and then inherently optimal level of expenditures : ⎧ 1 1 ⎫ ⎪ ⎪ max ⎨ ⎬ ⎪ ⎪ ⎩ ⎭ Subject to: This gives us a growth optimizing tax rate equal to when 1. This simply means that government spending is proportional to long-run output. In effect it is equal to a fixed share of permanent output: → . This solution states that a shock to permanent GDP (say by raising () permanently) raises revenues and that the only way to achieve equality is by also increasing expenditures permanently. In practice, this means that expenditures increase with potential GDP, resulting in a steady share of public expenditure in GDP over time. The difficulty lies with estimating potential GDP and separating it from temporary fluctuations. Figure 3 provides a graphical explanation of how the cycle can be mistaken for the trend. To smooth the business cycle, it is economically optimal to temporarily increase expenditures in a downturn of the business cycle, to support the economy, and to temporarily decrease expenditure in an upturn to save for bad times. Yet this is difficult in practice as potential GDP is unobserved. It can easily be seen from the above framework that mistaking a cyclical increase in GDP for a structural increase, any accompanying increase in expenditure will be cyclical too—unintentionally making fiscal policy procyclical. 5    Figure 3: Schematic representation of the cycle is the trend (value added over time, trend GDP, “the cycle is the trend” GDP, and business cycle fluctuations) Source: The authors. 3. An empirical examination of the cycle is the trend in EMDEs We now turn to an empirical assessment of the cycle is the trend in EMDEs, or the mirage of higher trend growth created by the latest commodity super cycle. This analysis requires data on government expenditures, the output gap, commodity prices, real GDP and the GDP deflator for EMDE countries. Commodity price data were collected from the World Bank’s Pink sheets while national accounts and fiscal data were collected from the IMF’s WEO databank. We express variables in real per capita terms using population estimates from the UN. The remaining variables (i.e. the output gap, the cycle is the trend and the structural balance) are calculated with formulas in the following sections. The analysis focuses on 26 EMDEs from 2003 until 2016, which yields 364 observations. The EMDEs are Argentina, Bulgaria, Brazil, Botswana, Colombia, Croatia, the Arab Republic of Egypt, Indonesia, India, Jamaica, Lebanon, Sri Lanka, Morocco, Mexico, Mauritius, Malaysia, Namibia, Nigeria, Pakistan, Peru, Philippines, Poland, the Russian Federation, Turkey, Vietnam and South Africa. Decomposing EMDE Budget Balances Formally one may decompose the budget balance into a cyclical component and a structural one as follows: (5) where is the observed budget balance, is the structural budget balance and is the automatic part of the budget balance that changes with the cycle. The automatic part varies with the economic cycle, which is defined here as the output gap ( , expressed as the percent deviation of GDP from potential GDP. The automatic part of the budget can thus be expressed as which means we can write (5) as: (6) We assume that is an unobserved and a time-varying AR(1) process. The Kalman filter is used to estimate (6). The resultant from equation (6) will be used as a benchmark in determining the fiscal stance along with the growth in real GDP per capita. One particularly sticky issue is the measurement of the output gap ( ). We calculate potential GDP using a Cobb-Douglas technology as in Burns et al. (2014). To clean the output gap from commodity 6    price effects we construct a real commodity export price index as in Keyfitz (2004) that is weighted by trade shares using the World Bank WITS data set and deflating it by the GDP deflator for each country. We take a simple HP-filter on this series to get the commodity price cycle ( ) (see Alberola et al. (2016) for an alternative measure). ∑ ,     (7) where the weights are calculated using the share of each commodity ( ) in total country exports in the base period (the average observed between 2009-2011) deflated using GDP deflator ( ). The output gap minus the fitted line on the real commodity price gap yields an estimate of the non- commodity output gap : (8) Or Cyclicality of Fiscal Policy: The Role of Cyclical Lengths To investigate the possibility that policy makers incorrectly identified long-term growth, and adjusted their fiscal policy accordingly, we follow Strawczynski and Zeira (2013). They show that the extent to which expenditures respond to output depends on the length of the cycle. This is related to the simple illustration in the theory section above. Fiscal policy is prone to become procyclical in cases where shocks lengthen the duration of the cycle. We follow Strawczynski and Zeira (2013) and Aguiar and Gopinath (2007) to create a variable that measures how the cycle becomes the trend: ∆ (9) Where is the log of per-capita real GDP and is the number of lagged differences. We normalize the above to one using ∆ ∆ and express the time sequence as: ∆ (10) ∆ This ratio provides a measure of long-run variability to short-run variability, or the degree to which the cycle is the trend. If the standard deviation of per capita real GDP for any period lagged is the same as the difference of one period lagged, then there is no cyclical trend. However, we interpret a persistent cycle as one where the lagged variance to that of one period ago is larger than one. Note that equation (10) essentially boils down to a variance ratio test (see Cochrane, 1988). In essence, it is testing for a random walk in GDP vs. temporal shifts in GDP. If GDP is trend stationary, then GDP will fluctuate around that trend over time – implying that shocks to GDP are only temporary. If GDP is a random walk with drift, then shocks may push GDP on different paths (i.e. there is no reason to assume trend correction). One way to interpret (10) is that a value of 1 implies a random walk, 1, implies mean reversion and 1 implies some very persistent movements in output. Table A1 in the annex summarizes the statistical significance for each country for various horizons. In our sample, South Africa has the largest persistence (Figure 4): the long-run variability of output to short-run variability is more pronounced compared to its peers, reaching a maximum of about 12 years (this compares well with the average duration of business cycles as estimated by the South African Reserve Bank (according to the 2018 Quarterly Bulletin). The trend is the cycle indicator is above 1 for most countries, indicating that cycles are persistent. The persistent cycle could create the illusion that GDP is on a long-term higher trajectory and could lead to fiscal policy becoming procyclical. 7    Figure 4: Variability in output cycles in selected EMDEs (Long-to-short term variability of GDP per capita at various lags) 4.0 ZAF BRA IND BWA 3.5 RUS MEX PER IDN 3.0 Relative Variances at K TUR 2.5 2.0 1.5 1.0 0.5 0.0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 K Source: Authors’ calculations. Note: ZAF: South Africa; IND: India; RUS: Russia; PER: Peru; TUR: Turkey; BRA: Brazil; BWA: Botswana; MEX: Mexico; IDN: Indonesia. Finally, a measure that captures the cycle is the trend is calculated by taking the above indicator (equation 10) and multiplying it by the smoothed (three-year moving average) growth in per-capita GDP: (11) As with the non-commodity output gap, we also compute the non-commodity CITT variable as: (12) Figure 5 provides a sense of how much the commodity cycles in each country contributed to larger growth rate than what would have been observed without movements in real commodity prices. The left panel highlights CITT before the crisis in 2007, the center panel compares CITT and CITT excluding commodity price movements during 2009, and the right panel provides an estimate of the two CITTs in 2016. A 45-degree line divides the charts in two. Note that the majority of the countries are to the right of the line in 2007 – non-commodity CITT was much lower relative to actual CITT, which supports the hypothesis that commodity prices cycles prolonged the economic cycle. During the financial crisis, countries moved closer to the 45-deggree line, implying that the effects of commodities contributed little to the economic cycle. 8    Figure 5: Commodity vs. non-commodity cycles (impact of commodity prices on the cycle is the trend) Source: Authors’ calculations. The non-commodity CITT and output gap are used as explanatory variables to determine the cyclical stance of the government using both the estimates of the structural budget balance and real per capita expenditure growth rates net of social contributions. We assume that countries or country groups may vary their fiscal stance over time. It is possible that a country followed countercyclical fiscal policy leading up to the financial crisis, but then became procyclical after the crisis. Our model takes the following specification: ∆ , , , , , (13) ∆ log , , , , , (14) Our fixed effects estimates imply that , . Here is the group mean intercept while measures differences that are due to changes between countries. To allow for random variation of the fiscal stance over time, we model , , where is a time index and accounts for random variation over time. Note that , ~ 0, and that time and country variability are 0 independent ~ . 0 The definition of procyclical fiscal policy requires that , 0. This means that the structural deficit widens when the cycle increases. Columns 2 and 3 of Table 1 compare the fixed effects model of equation (14) to one where the slopes vary with time (column 3). The last two columns compare the fixed effects model of equation (13) to one where the fiscal stance varies with time (column 4). In all specifications , 0, implying that EMDEs have been mainly procyclical over time. Column 2 suggests that the model where the fiscal stance varies with time describes the data better (it has a lower log-likelihood function). 9    We plot the fiscal stance parameter as a function of time in Figure 6. The fiscal stance over time has been moderately procyclical with little evidence of becoming more countercyclical irrespective of our measure of the cycle. Table 1: Model comparison1 ∆ log , , ∆ log , , ∆ , , ∆ , , Fixed Estimate S.E Estimate S.E Estimate S.E Estimate S.E effects: , 0.11 0.10 0.14 0.11 0.00 0.00 0.00 0.00 , 1.36** 0.42 1.32 0.68 0.06* 0.03 0.05 0.03 Random effects: 0.19 0.22 0.00 0.00 0.00 0.00 1.11 1.03 0.00 0.00 Fit: AIC 1110.1 1107.3 -1930.2 -1934.4 BIC 1125.7 1134.6 -1914.6 -1907.1 LogLike -551.0 -546.6 969.1 974.18 Source: Authors estimations. Figure 6: EMDE fiscal stance as a function of time 5 , , 5 , , Source: Authors’ calculations. Note: Positive elasticities point to procyclical fiscal policy.                                                              1 The Hausman test that errors are uncorrelated with the regressors cannot be rejected (i.e. we use this as our implicit test of endogeneity – which suggests that regressors and errors are orthogonal). 10    4. Fiscal policy and the cycle is the trend in South Africa Since democracy in South Africa in 1994, three distinct phases characterized fiscal policy. The end of Apartheid heralded a period of debt repayments. South Africa repaid legacy debt to external creditors and pursued consolidated fiscal accounts. In its annual budgets following the transition to democracy in 1994, it identified fiscal consolidation as the country’s main fiscal policy objective – a continuation of the fiscal policy objective from 1992 (National Treasury, Budget Reviews 1994 - 1997). In a context of inherited national debt amounting to 48.7% of GDP, together with weak revenue collection, the government identified the need to reduce the budget deficit and government dissaving (borrowing), moderate government consumption expenditure as a share of GDP, and avoid a permanent increase in the tax burden as its main fiscal policy goals (Budget Reviews 1994 – 1998). The government affirmed this policy stance over the subsequent years until 2001 (Budget Reviews 1999 – 2000). From 1998 onwards a more pronounced emphasis on fiscal sustainability, and an attempt to link expenditure growth to economic growth can be observed. By the 1999 Budget, following poor economic growth in the previous year, the government deferred its fiscal deficit target – an early case of fiscal responsiveness to economic performance. From 2001, the government stated its intention to shift towards an expansionary fiscal policy stance to support the economy, investment and social spending. It noted that the fiscal consolidation of the previous years (1994 – 2000) provided the basis upon which to embrace its expansionary fiscal policy stance. South Africa reduced gross debt from 41% of GDP in 2001 to 26% of GDP in 2009 by running primary surpluses and acquired favorable credit ratings (Figure 7). This period also coincided with real commodity price increases, which benefited commodity trade. GDP growth was on an upward path, which together with an improvement in tax collection methods led to increase in revenues. Sustained strong growth over this period seemed like a permanent shift in GDP. This period also saw a permanent shift in revenues, which allowed for sustained real growth in expenditure. Public expenditure2 increased from 24.4% of GDP in 2001/2002 to 24.9% of GDP in 2007/2008, while revenues3 increased from 23% of GDP in 2001/2002 to 25.9% in 2007/2008 (Budget Reviews). Both revenues and expenditures outpaced nominal GDP growth, reflecting a wider tax base, profitable institutions, higher returns to factor inputs, and most likely spill-overs from commodity prices. It was only in 2007 that the government identified the relationship between its fiscal policy stance and the economic cycle through the introduction of the concept of a cyclically-adjusted budget balance in its 2007 medium-term budget (MTBPS 2007, Budget Review 2007). In 2009, as the country experienced the effects of the global financial crisis, the government retrospectively referred to its fiscal policy stance over the previous period (since 2001) as countercyclical, and explicitly committed itself to a countercyclical fiscal policy. In subsequent budgets, the government affirmed its counter-cyclical fiscal policy stance, and stated its objective of counteracting variations in the business cycle and supporting growth through public spending. It identified countercyclicality as one of the principles underlying fiscal policy in South Africa (Budget Reviews 2008 – 2014). The fiscal policy response during this period seemed standard and appropriately countercyclical with a combination of automatic stabilizers acting together with discretionary fiscal measures to stabilize, or spur on demand during what was likely to be its longest downward phase of the economic cycle. The economic response was muted and consequently left the government with a sizable debt burden with high interest repayments, and possibly lower long-run growth.                                                              2 Main budget expenditure. 3 Main budget revenue.  11    Figure 7: Growth and fiscal policy since democracy in South Africa (budget balance and gross public debt in percent of GDP and percentage growth of real GDP) Source: South African National Treasury and business cycles estimate from Table 2. By 2013, following several years of sustained budget deficits, slow economic growth and gross-debt increasing to 40% of GDP, the government committed itself to a policy of fiscal consolidation combined with a continuation of the countercyclical fiscal policy stance. By 2015, with gross-debt increasing to 44% of GDP, the government noted that its countercyclical approach has “reached its limits” and it was running a structural deficit, warranting significant expenditure-reduction and revenue-raising consolidation measures. The build-up of contingent liabilities, especially from extending guarantees for SOC borrowing, is also likely to have constrained the country’s appetite for continuing to pursue increases in countercyclical expenditure (Bachmair and Bogoev, 2018) Table 2: The South African business cycle and selected fiscal and debt dynamics Average annual change* Gross debt Business cycle period Fiscal years Phase GDP Expenditure Revenue as share of GDP December 1996 - August 1999 1996/97 - 1999/00 downward 1.9% -0.6% 5.4% -1.0 pp September 1999 - November 2007 2000/01 - 2006/07 upward 4.2% 4.7% 5.3% -2.0 pp December 2007 - August 2009 2007/08 - 2009/10 downward 0.6% 9.0% 4.4% 2.5 pp September 2009 - November 2013 2010/11 - 2012/13 upward 2.6% 3.6% 6.9% 4.4 pp December 2013 - 2013/14 - 2016/17 downward 1.1% 1.8% 5.0% 2.3 pp - 1996/97 - 2016/17 - 2.8% 3.3% 4.0% 0.2 pp * Real change in GDP, expenditure and revenue. Based on a fiscal year deflator in constant 2010 prices Source: Statistics South Africa, Budget Reviews (BRs), South African Reserve Bank and authors’ calculations. Table 2 summarizes expenditure, revenue and debt trends across different phases of the business cycle between 1995/96 and 2016/17. Two periods can be observed. Between 1996/97 – 2008/09, as expenditure grew slower than the economy while revenue grew faster, the country was able to realize 12    progressively narrower budget deficits, eventually realizing two years of budget surpluses. This allowed the government to reduce its debt stock (as a share of GDP) – a stated fiscal objective. Between 2009/10 – 2015/16 expenditure growth exceeded economic growth and revenue growth, resulting in sustained large budget deficits and an increased in its debt stock (as a share of GDP). It is possible that the commodity boom period of 2001 to 2007 in South Africa would have seemed like a permanent shift in output. Government spending and revenues during this period increased rapidly. It is difficult to pinpoint whether this increase was due to an upward forecast bias in potential GDP, or whether it was intentional. There is some support that South Africa erred in terms of estimating potential GDP (Anvari et al. (2014)) after the financial crisis. Figure 8 illustrates this point, comparing the three- year (and a few months) forecasts from the Budget Reviews and Medium Term Policy Statements (MTBPS) with actual growth outturn, i.e. it looks at the first outer year estimate/print published by the National Treasury. Figure 8a compares projections and outturn and Figure 8b looks at the cumulative forecast-outturn difference. Figure 8a demonstrates the over-optimism about future growth, following the global financial crisis and 2009 recession. In Figure 8b, a flat line would suggest that there is no discrepancy between forecasts and outturns, a downward trend would suggest that projected growth was lower than observed and an upward trend would suggest that growth projections were larger than observed growth. The lines clearly point upwards both for Budget Reviews and MTBPSs after 2009, indicating an upward growth bias.4 The optimistic bias from 2007/08 until 2016/17 in the graphs provides some evidence that potential GDP growth was perceived as much larger than it actually was. Figure 8: Growth projections vs. growth realization (forecasts vs. actual and cumulated projection growth differentials with actual growth) a) Outer-year forecast vs actual GDP growth 7% Real GDP 3.25yr forecast Real GDP 3.5yr forecast 6% Real GDP growth Actual 5% 4% 3% 2% 1% 0% ‐1% ‐2%                                                              4 Investigating the potential biasness of the National Treasury’s GDP growth forecasts, the Parliamentary Budget Office found some statistical evidence of biasness for several forecast horizons, including the three-and-a-half and three-and-a-quarter year forecasts, with a tendency to overestimate GDP growth (Parliamentary Budget Office, 2016).   13    b) Cumulative projected growth differentials with actual growth 10.0% 8.0% MTBPS BR 6.0% 4.0% % pts 2.0% 0.0% ‐2.0% ‐4.0% Source: Budget Reviews (BRs), Medium Term Budget Policy Statements (MTBPSs) and authors’ calculations. Note: First print of outer-year forecast. An empirical investigation of the cycle is the trend in South Africa We now empirically look at whether the cycle is the trend mattered in South Africa. For the case study of South Africa, we develop a more sophisticated measure of the structural budget balance than in the panel analysis of EMDEs above, which is easier in a simple time-series setting. Details about deriving the structural balance can be found in Annex 2. Figure 9 plots our estimate of South Africa’s structural budget balance. It suggests that although South Africa saved between 2006 and 2008 (a positive overall budget balance), the structural balance was already in deficit by 2005, and deteriorated significantly between 2007 and 2009. Since 2008, both South African overall and structural balances were in deficit and have not recovered since. The sustained structural deficits since 2005, together with corresponding overall budget deficits since 2009, was chiefly responsible for South Africa’s large and sustained increase in debt since 2009, significantly reducing fiscal space. Figure 9: Estimates of South Africa’s budget balances (budget balances in percent of potential GDP) 2.0% 0.0% ‐2.0% % of potential ‐4.0% ‐6.0% ‐8.0% ‐10.0% 2002 2004 2006 2008 2010 2012 2014 2016 Recessions Budet Balance Structural Budget Balance Structural Budget Balance (SS) Source: World Development Indicators and authors’ calculations 14    We augment equations (13) and (14) for South Africa and control for the effects of large policy changes to isolate the cyclical stance of the government – such as national plans that explicitly state what fiscal policy ought to do. These represent outlier cases of fiscal policy and would have an impact if any of the measures are significant: ∆ ∑ (15) ∆ log ∑ (16) We construct using the various policy documents from 1991 onwards. We assign a value of 1 when the document outlines a path of government that supports higher spending. This helps us to approach the narrative of policy and its relation to government outcomes and disentangle policies that were expansionary vs. contractionary (the assumption that contractionary policies do not alter the results). Annex 3 provides further information on the construction of the dummies. Turning to the estimates of equations (15) and (16), we show that:  Each of the policy measures (∑ ) are positive and consequently lead to higher expenditures or a larger structural deficit.  The coefficients on the trend is the cycle variable, however, is positive in all controls, indicating fiscal policy in South Africa has not really been countercyclical. This can be attributed by several factors such as correct intentions by the fiscal authorities, but missing either important turning points of the business cycle, perceiving long-cycles as the trend, poor choices on spending which are difficult to reverse such as wages or weak policy coordination where funds are disbursed and allocated but not spent on time. We discuss some of these factors in the next section. It should be noted that our results are in line with Cespedes and Velasco (2013) who analyze the fiscal stance for commodity exporters. They study whether policy has been less procyclical or countercyclical in the recent commodity boom episode compared to previous episodes. While they find that for most countries the elasticity of overall balance to the cyclical component of commodity prices has increased, for South Africa it has become negative, implying procyclical fiscal policy during the 2006-2009 period as opposed to the 1973-1990 commodity price boom. Table 3: The fiscal stance of the South African government Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Model 8 ∆ ) ∆ ) ∆ ) ∆ ) ∆ ) ∆ ) ∆ ) ∆ ) 0.27** 0.24** 0.01 0.03 (0.11) (0.10) (0.02) (0.02) 0.85 0.69 0.94 1.18 (0.81) (0.73) (13.09) (13.81) 0.13 0.10 1.45 0.13 0.28 0.01 0.44 0.22 -0.03 -0.07 -0.02 -0.16 F-Stat (P-value) 0.02 0.31 0.03 0.16 0.57 0.94 0.48 0.48 Source: Authors’ calculations. Notes: Standard errors in parentheses. 15    Fiscal policy intentions: Accidental or intentional? Did the South African government pursue procyclical policy accidentally or on purpose? Assuming two objective functions, we construct measures of intentions that we call consistency: meaning the government achieves its intended outcomes. We define these objectives as (i) the government attempting to stabilize the economy and (ii) the government attempting to run sustainable fiscal policy. We collect a data set using various Budget Reviews (starting in 1998) and analyze text related to a set of words that are related to these objectives. We build indices of these words (normalized by the total number of words in each document), which we then evaluate with outcomes (correlations to the budget deficit and to the output gap). Our correlations are referred to as consistency measures, i.e. capturing whether a government achieves its stated policy objectives. The words are related to statements regarding stabilizing the economy, smoothing out business cycles, and countercyclical fiscal policy. We create two measures of consistency. 1. One where the government announces that it will aim to smooth the business cycles compared to movements in the actual business cycles. The word frequencies related to business cycle stabilization are correlated with the inverse of the output gap ∗ , ∗ . A positive correlation implies that the government follows through on its statements, i.e. that statements of stabilizing the economy are followed through by a stabilizing event. To this end we have an illustration of consistent demand management policies. 2. The second measure correlates the word frequencies related to the stance of the government to actual changes in the budget balance. This measure is perhaps more important than the first measure given the direct control that the fiscal authorities have on setting budget targets. As an example, words related to stimulus and the deficit are measured as: , ,∆ ,∆ ∆ , or consolidation: ∆ . If the government announces a period of expansion or consolidation and the actual budget balance follows those announcements, then policy is consistent. will naturally fluctuate between -1 and 1. If 0 during a consolidation effort, then the government does what it says – where once again, consistency is the intent of government measured against outcomes. If the government is countercyclical, then it is possible that both objectives can be met and hence they should consequently co-move. A procyclical stance would imply competing objectives where one would for example emphasize output maximization policies over a stable budget. Note here that maximization of output is not stabilization of output – here again one could see how a government can prolong a business cycle. Figure 10 plots the frequency of words related to consistency of the budget and output stabilization. It is interesting to note that budget and output stabilization are almost inversely related to each other during 2011-2015: During the financial crisis (2008/2009) there was an emphasis on using countercyclical fiscal policy to stimulate the economy, however in the years since 2013 the Budget has focused its efforts on terms related to output consistency but at the expense of foregoing countercyclical policy. An obvious caveat is the limited number of words in each Budget Review related to these objectives. It is striking that little space is devoted to discussing budget stabilization policies. 16    Figure 10: Counts of words regarding various consistency measures (word frequencies in Budget Reviews) Word Frequencies 30 25 Consistency: Output Stabilization Consistency: Budget Stabilization 20 15 10 5 0 1998 2000 2002 2004 2006 2008 2010 2012 2014 2016 Source: Budget Reviews and authors’ calculations. In Figure 11 we plot the rolling correlations of these two measures with (i) the output gap and (ii) the realized budgets. Again, these two terms do not always move in a similar manner.  The authorities’ announced budget intentions (“consolidate the deficit”) and realized budget outcomes demonstrate limited consistency. The sudden change in signs from a consistent positive correlation to consistent negative correlation in 2011 is suggestive of either intentional procyclical policy or calculation errors in terms of potential GDP.  Looking at the policy intention to ‘smooth the business cycle’ (i.e. countercyclical fiscal policy), actual policy reflected intentions up to 2006. In 2007, when the overall budget balance was still in surplus but the structural balance deteriorated markedly (as shown in Figure 8), intentions decoupled from the actual stance, suggesting that the authorities had underappreciated that strong commodity prices were still positively underpinning South Africa’s business cycle. The cycle had become the trend. Intentions remains decoupled from the actual dynamics of the economy until 2014/15, the actual end of the commodity super cycle. By 2016, South Africa’s fiscal buffers were depleted and creditworthiness had declined. South Africa lost its investment grade credit rating in early 2017, being downgraded to BB+ by both Fitch and Standard and Poors, although Moody’s maintained an investment grade credit rating. When South Africa slipped into recession in the first half of 2018, room for fiscal stimulus was severely limited. 17    Figure 11: Rolling correlations of various consistency measures (rolling correlation) Rolling Correlations 1.2 0.8 0.4 0.0 -0.4 -0.8 Consistency: Output Stabilization Consistency: Budget Stabilization -1.2 03 04 05 06 07 08 09 10 11 12 13 14 15 16 Source: Various Budget Reviews and authors’ calculations Policy Discussion As many other EMDEs, South Africa seems to have been surprised by the length of the commodity super cycle, and its impact on extending domestic business cycles. Procyclical fiscal policy in South Africa was thus partly unintentional. However, the cycle is the trend is not necessarily the only reason for this procyclicality. This section reviews potential other reasons: 1. Limited access to international credit markets and limited depth of domestic financial markets 2. Political economy and institutional factors 3. Design of countercyclical expenditure response and expenditure composition 4. Underestimated structural dynamics 1. Access to international credit markets and limited depth of domestic financial markets Countries experiencing adverse economic conditions find it difficult to access credit, and if available it is generally very expensive. The absence of affordable credit limits the degree to which countries can employ countercyclical fiscal policy. Adverse economic conditions may force countries to increase taxes and/or reduce government spending, which further entrenches a procyclical fiscal stance (Calderón, Chuhan-Pole, & López-Monti, 2017, Frankel, et al., 2013, and Gavin et al., 1996). Limited access to international and domestic credit markets is highly implausible as a reason for South Africa’s observed procyclical fiscal policy stance. South Africa, after transition to democracy in 1994, was able to access domestic and international credit, despite its relatively high levels of inherited public debt. Its debt issuance increased gradually from 1994, and quite rapidly after 2006. The country’s borrowing costs declined consistently from 1994, and by 2001 credit ratings agencies Standard and Poor’s, Moody’s, and Fitch all rated the country’s foreign and local currency debt at investment grade. By 2012 the country was included in the Citibank World Government Bond Index, allowing it improved access to foreign lenders and lower rates. While the country has experienced downward revisions to its local and foreign currency credit ratings since 2014—and lost its investment grade credit rating on 18    foreign-denominated debt—it still enjoys access to domestic and foreign credit markets, with its bond auctions remaining well subscribed. Limited liquidity for the government is an unlikely contributor to South Africa’s observed procyclical fiscal stance. 2. Political economy and institutional factors In periods of strong economic growth, with high revenue growth and easy access to cheaper finance, ministers of finance find it difficult to resist additional demands for further appropriations, referred to as voracity effects. These voracity effects are stronger when there are genuine and critical social needs (e.g. welfare spending, education, health care, housing, and other unfunded or underfunded mandates), and when there exist significant differences between the demands of different groups exerting demands over public resources (Abbott & Jones, 2013). Under such conditions, ministers of finance are often forced to make additional appropriations, which prove difficult to reverse (Perry, 2003). It is easy to see how voracity effects can interact with overoptimistic growth assumptions when the cycle is the trend: Given higher expectations on growth, more ambitious future spending allocations are budgeted for in medium term fiscal frameworks. The degree to which a minister of finance and treasury can mediate and manage the competition for public spending and pursue countercyclical fiscal policy is also a function of domestic institutional strength. A country with weak institutions—including weak rule of law, lack of enforcement of property rights for investors, and widespread corruption—is less likely to be able to enforce countercyclical fiscal policy. The competition over fiscal allocation, combined with weak political institutions, may result in additional appropriation of public resources during periods of high growth, thereby locking-in a procyclical bias (Abbott & Jones, 2013). World Bank Group (2018) argues that pressures to raise appropriations are high in South Africa, as the country aims to tackle the high levels of inequality it inherited of a history of separation and apartheid. Pressures for redistribution are thus high and can support voracity effects. In addition, post-apartheid South Africa saw an erosion in the quality of institutions since about 2007, including institutionalized corruption (“State Capture”) that not only undermined state-owned enterprises and law enforcement agencies, but also reduced administrative efficiency in the South African Revenue Service and aimed to undermine the National Treasury. These factors may have contributed to undermining countercyclical policy making in South Africa, despite official intentions. Given a propensity to overestimate future growth in South African Budget Reviews, voracity effects may well have interacted with the cycle is the trend dynamics in South Africa. 3. Design of countercyclical expenditure response and expenditure composition The ability of a government to successfully pursue countercyclical fiscal spending is dependent on the responsiveness of public spending to changes in the business cycle. This, in turn, is a function of a) institutional arrangements and design of countercyclical expenditure response, and b) the composition of public spending. In the pursuit of countercyclical fiscal policy, institutional arrangements and design of countercyclical expenditure response must allow for a deliberate, timely, targeted and temporary increase in public spending in response to a downward-phase in the economy, and a winding-down during an upward- phase. Public spending directed towards countering the effects of the business cycle may be automatic (non-discretionary) or discretionary. Given the lags associated with identifying changes in the business cycle, as well as developing, appropriating and implementing spending programs, automatic expenditure stabilizers allow for more agile fiscal responsiveness compared to discretionary interventions (Baunsgaard and Symansky, 2009, Elmendorf and Furman, 2008 and Taylor, 2000). In addition, effective discretionary fiscal spending, effected through legislative appropriation, is contingent on political support for both implementation and winding-down when the business cycle changes. 19    To the extent that South Africa pursued countercyclical fiscal spending, it was predominantly discretionary. Its primary automatic stabilizer on the expenditure side is payments from its Unemployment Insurance Fund (UIF), which comprised a particularly small share of the country’s overall budget, limiting its effect on aggregate demand.5 In its pursuit of discretionary fiscal policy as a stabilization tool, the South African government made only limited explicit reference to which expenditures have been increased specifically in response to the business cycle. The absence of clear identification of discretionary spending, including amounts and time-frames, may have led to fungibility between funding unfunded or insufficiently-funded mandates (baseline expenditure) and discretionary fiscal spending, making the discretionary fiscal policy responses more difficult to reverse upon changes in the business cycle. Interestingly, in the 2015 MTBPS the government proposed a “fiscal guideline” linking expenditure in the outer-year of its three-year budget-planning-framework to potential GDP. While it is not clear if the proposed “fiscal guideline” was adopted and implemented, even if adopted earlier its potential efficacy would have been limited by the accuracy of potential GDP estimates. Moreover, A country is more likely to have a procyclical fiscal policy bias to the extent that its expenditure composition is distributed in favor of expenditure items that display procyclical behavior. Of particular relevance to South Africa is wage spending, given its rapid growth since 2008, contributing 38% to the overall increase in expenditure between 2008 and 2016 (see Table 4). Several studies find the public sector wage bill to behave in a strong procyclical manner across the economic cycle (in booms and busts) (Eckardt and Mills, 2014, Lamo et al. 2013, and Lane, 2003). In addition to the overall wage bill being procyclical, Lamo et al (2013) also find that public employment levels, and remuneration per public sector employee also move in the same direction as the business cycle. With rapid growth in the public wage bill since 2008, due to headcount growth and above-inflation remuneration increases, South Africa’s public wage bill is likely to have contributed to a procyclical fiscal policy bias. As upward-adjustments to the public wage bill are politically difficult to reverse, de facto, wages in South Africa are nondiscretionary. This is likely to have limited the extent to which the government was able to respond to a downward phase in the economic cycle, resulting in entrenched medium-term compositions relative to other expenditure items. Table 4: Consolidated public spending by main category (share of total central government expenditure) Contribution to overall increase in expenditure (%) (share of total) 1997/98 2001/02 2008/09 2014/15 2016/17 2008/09 - 2016/17 Compensation of employees 39% 37% 33% 34% 35% 38% Transfers 20% 29% 34% 40% 33% 32% Goods and services 13% 13% 17% 12% 14% 12% Interest payments 20% 18% 8% 10% 11% 13% Capital 8% 4% 7% 4% 6% 6% Total 100% 100% 100% 100% 100% - Source: Budget Reviews and authors’ calculations.                                                              5 The degree to which Unemployment Insurance Fund disbursements serve as an automatic stabilizer is limited by the portion of the population eligible for UIF unemployment benefits, slow employment growth, and modest rand value of unemployment benefits disbursed. To be eligible for UIF unemployment benefits, a person had to previously contribute to the UIF, generally necessitating formalized employment. On average, about 40% of the employed have historically not been eligible for UIF. 20    The pursuit of countercyclical fiscal policy through discretionary increases to public spending, with the wage bill comprising the largest share of overall expenditure increases, resulted in South Africa’s budget structurally increasing to a higher level of expenditure to GDP. To the extent that the expansion of expenditure was based on a mirage of higher potential GDP, it is difficult to reverse past decisions and return to a budget size more in line with actual potential GDP. In South Africa, in light of high public debt, high expenditures but slowing revenue dynamics, this has only left the option to cut discretionary spending, notably on infrastructure. Yet, cutting investment eventually comes at the expense of raising future potential, to the levels where it was perhaps expected when the cycle became the trend. 4. Underestimated structural dynamics The commodity super cycle is one possible reason why potential GDP growth was overestimated. Another potential reason is that South Africa’s structural constraints—and the ability to resolve them— have been underestimated, resulting in an unanticipated slowdown in potential GDP growth. World Bank Group (2018) shows that total factor productivity growth has been decelerating significantly in South Africa, and declining in recent years. This can be due to multiple factors, including rigid factor markets that keep factors from being allocated more efficiently. It can also be due to South Africa losing global competitiveness due to labor ad factor market rigidities and the weakening of the performance of important state-owned enterprises, in electricity or transport. South Africa has also been attracting relatively limited investment (both physical and in innovation). The World Bank Group partly attributed this to policy uncertainty that is keeping investors wary. An underappreciation of the magnitude of South Africa’s challenges may also have resulted in overestimates of potential growth, in addition and beyond the cycle is the trend. 5. Conclusion Fiscal policy is an important stabilization tool. Adequate countercyclical fiscal policy may stabilize the economic cycle. Unfortunately, policy mis-steps such as the mistiming of fiscal policy can destabilize the cycle, or prolong recessions. This paper estimates the cyclical stance of EMDEs with a focus on South Africa by decomposing the budget into transitory and permanent components. To control for the variability in output, a cycle is the trend variable is constructed, but cleaned of commodity price effects. This accounts for the possibility that the authorities might perceive long cycles as the trend – in which case they run the risk of enacting procyclical fiscal policy. The results show that fiscal policy in EMDEs has been mostly procyclical. This result is robust regardless of different lengths of the cycle is the trend or alternative measures of the cycle such as the output gap. This can be attributed to good intentions by the fiscal authorities, but missing either important turning points of the business cycle, perceiving long-cycles as the trend due to commodity price cycles, poor choices on spending which are difficult to reverse such as wages or weak policy coordination where funds are disbursed and allocated but not spent on time. We study South Africa to understand the extent to which this may have played out in a case study. An interesting outcome of the analysis for South Africa is the relationship between the fiscal stance and a derived intention measure which is calculated using texts from various Budget Reviews. The difference between outcomes and intentions yields a set of consistency measures. They show that meeting budget announcements and stabilizing the business cycle have been at odds at times and the outcomes related to intentions sometimes went in opposite directions. There exists thus a mismatch between what the government aimed to achieve (e.g. countercyclical fiscal policy) and the outcome. Imperfect knowledge regarding the length of the economic cycle (e.g. the commodity boom in the 2000s), and external shocks and other exogenous shocks may have contributed to this mismatch. These results imply that forecasts of GDP and specifically the output gap have a large weight in policy errors. A clearer understanding of the duration of cycles, the impact of expenditure and revenue policies on the cycle and credibility of fiscal policy will go a long way in aligning fiscal policy with macroeconomic objectives of sustainable countercyclical fiscal policy. 21    Annex 1: Variance-Ratio tests Table A1: Variance-Ratio tests  K=2 K=4 K=8 K = 16 K= 32 ZAF 1.5* 2.2* 3.7* 5.7* 0.8 IND 1.3 0.8 1.0 NA NA BRA 1.3* 1.4 1.6 2.5* NA BWA 0.8 0.5 0.2 0.3 NA RUS 1.5* 2.0* 3.0* NA NA CRI 1.1 0.8 0.6 0.6 NA ARG 1.3 1.4 1.2 0.4 NA COL 1.4* 2.0* 2.2* 1.3 2.2 PER 1.4 1.9* 2.4* 2.9* 2.1 MEX 0.9 0.7 0.4 0.5 NA NGA 1.6* 2.7* 4.6* 6.9* NA EGY 1.9* 3.6* 3.3* NA NA MAR 0.7* 1.3 1.6 NA NA MUS 0.7 0.7 0.4 0.2 NA MYS 0.8 0.6 0.0* NA NA IDN 1.4* 1.6* 1.6 1.2 NA POL 1.0 1.1 0.9 2.4* NA PHL 1.3 1.9* 1.7 2.7* NA VNM 1.6* 2.1* 1.6 2.7* NA BGR 1.3 2.1* 3.3* 1.8 NA TUR 1.3* 1.6 0.4 0.4 NA JAM 1.6* 2.3* 2.8* 0.1 NA LBN 1.2 1.4 1.1 2.0* NA PAK 1.4* 1.7* 1.5 0.3 NA LKA 1.4* 2.0* 2.8* 2.3* 0.2 NAM 1.5* 2.3* 3.7* 5.8* 2.3 Notes: *,**,*** indicate significance at 10%,5% and 1% level. 22    Annex 2: Estimating the structural budget balance for South Africa We decompose tax and expenditure lines into a cyclical and structural component. For taxes, this is usually done using a two-step approach whereby one estimates the responsiveness of revenue to its tax base, and then the tax base to output (as an example see Russek and Kowalewksi, 2015). Formally, the structural budget balance ( ) can be written as: ̅ (A1) where and ̅ are structural revenues and expenditures, respectively. If ̅ is discretionary (an assumption we make) then structural revenues are the sum of each revenue response to its trend tax base. Structural revenues are calculated from a VAR. To account for endogeneity in extracting the automatic responses of the revenues to output, we follow a very popular approach by separating the cyclical and structural components out from taxes using the methodology of Blanchard and Perotti (2002) and Caldara and Camps (2008). The reduced form VAR model is specified as: (A2) is a -dimensional vector of endogenous variables, is a constant while is a time trend, is a p-order lag polynomial and is a vector of reduced form errors ( 0 and ∑ ). The list of variables in , , , are the log of real GDP per capita, real government expenditure per capita, real revenues per capita and real debt per capita. All variables are expressed in first differences using annual data. Deflating revenues helps to deal with bracket creep arising due to inflation. The structural form of the model is written as: (A3) Note that describes the relationship between the structural shocks and the reduced form shocks . To identify the model, we need to impose restrictions on both and . If the structural shocks are uncorrelated then the variance-covariance matrix of is diagonal. Furthermore, if we use a Cholesky decomposition of ∑ the system becomes recursive. We impose both restrictions on the A and B matrices. This allows us to explore the annual dimension of the data without explicitly assuming that expenditures do not respond contemporaneously to GDP. GDP then follows expenditure in the ordering, followed by revenues and then debt. We can interpret the restrictions as follows: (i) government spending is purely exogenous in a contemporaneous sense and responds only to GDP; (ii) output is affected by both spending and taxes contemporaneously and not by debt; (iii) revenues react to both government expenditure and GDP contemporaneously – this ensures that we capture the effects of automatic stabilizers while ruling out the effects of discretionary taxes on GDP in the first period after the shock. Debt responds to everything in the system contemporaneously. Blanchard and Perotti (2002), however argue that taxes and spending can also respond to structural shocks, which is captured by the parameters of and . We can relate the reduced form and structural errors as: 1 0 0 1 0 0 0 1 0 1 0 0 1 0 0 0 1 0 1 0 0 0 1 23    Our interest lies with an estimate for , which is the automatic response of revenues to the business cycle. Interestingly there are a quite a few studies in South Africa that estimate using a variety of methodologies (c.f. Du Plessis and Boshoff (2007) and Swanepoel and Schoeman (2003)). In general, it can be shown that this estimate corresponds to the product elasticity of the tax base to revenues and output to the tax base ( ∑ , , ). We calibrate this value to be 1 using an average of the literature in a meta-analysis setting. Table 1: Output elasticities Du Plessis et al. (2007) 1.06 Swanepoel et al. (2003) 0.91 Equation (16) can be used to analyze the historical shock contributions of each variable on revenues (Figure 2). These are labeled baseline (refers to structural revenues), changes in expenditures, real GDP growth, changes in debt, and other factors (noise). Real revenues fell markedly during the recession in 2009. Two major factors contributed to the sharp decline: revenue responses to GDP shocks and noise shocks (which is the residual and accounts for other factors). The feedback effects of expenditure decisions are also clearly seen on revenue responses – especially in 2015 as revenues adjust in a rule- like manner. Growth in structural revenues has been moderating over time and decreasing since 2012 implying a slowdown in trend GDP. Figure A1: Decomposing real revenue growth (annual percentage change) 20.0 15.0 10.0 % change 5.0 0.0 ‐5.0 ‐10.0 ‐15.0 2002 2004 2006 2008 2010 2012 2014 2016 Baseline Expenditures GDP Noise Debt Actual Source: Authors’ estimations. The estimates of equations (5) and (11) allow us to create our fiscal stance measures – which here refers to the discretionary decisions of the government regarding the business cycle. It should be noted that these measures do not apply a statistical filter to obtain structural shares of the budget to output policy as is done typically when smoothing GDP. We argue this is a useful approach – we explicitly model the unobserved potential GDP by identification using a structural model (the SVAR), and by assuming that the long-run part of the government moves slowly in line with the long-run part of the economy around a stationary point. These estimates of the structural budget balance can then be compared to each other. 24    Annex 3: Empirically controlling South African policy shifts Table A2: Dummy variable definition for national policies Year Shifts in plans Relation to fiscal policy  Government 1994 - 1 Reconstruction and establishes extensive Development Program welfare system  Without adequate tax reform, expenditure policies aimed at creating growth were constrained  Effects debt payments 1995 - 1 Abolish dual exchange rate to rest of the world  Reduce fiscal deficits No dummy Growth Employment and Redistribution policy  Increase in efficiency No dummy SARS autonomy granted in collecting taxes  Increased planning No dummy Medium Term Expenditure over budget cycles Framework (MTEF) adopted  Departments plan No dummy Public Finance Management better over budget Act cycle  Targeted spending to 2005 - 1 Accelerated and Shared Growth help reduce poverty Initiative for South Africa and halve unemployment by predefined dates  Spending allocated to 2010 - 1 National Industrial Policy identified economic Framework and subsequent sectors Industrial Policy Action Plans  Eliminate poverty and 2013 - 1 National Development Plan reduce inequality by 2030 25    References Abbott, A., & Jones, P. (2013, March). Procyclical government spending: a public choice analysis. Public Choice, 243-258. Aguiar, M. and G. Gopinath. (2007). Emerging markets business cycles: The cycle is the trend, Journal of Political Economy, 115(1): 11–69. Alberola, E., Gondo, R., Lombardi, M. and Urbina, D. (2016). Output gaps and policy stabilization in Latin America: The effect of commodity and capital flow cycles. BIS Working Paper No. 568. Bank of International Settlement. Anvari, V., Ehlers, N. and Steinbach, R. (2014). A semi-structural approach to estimate South Africa’s potential output. South African Reserve Bank Working Paper Series No. 14/08. Bachmair, F.F, and Bogoev, J. (2018). Assessment of Contingent Liabilities and their Impact on Debt Dynamics in South Africa. World Bank MTI Global Practice Discussion Paper. No.1 May 2018 Barro, R.J. (1979). On the Determination of the Public Debt, Journal of Political Economy, 87(5): 940– 71. Barro, R.J. (1990). Government spending in a model of endogenous growth, Journal of Political Economy, 98(55): 103–125.Baunsgaard, T. and Symansky, S. A. (2009) Automatic Fiscal Stabilizers. IMF Staff Position Note SPN/09/23 Blanchard and Perotti (2002). An empirical characterization of the dynamic effects of changes in government spending and taxes on output, Quarterly Journal of Economics, 117 (4) (2002), 1329-1368. Budget Reviews: http://www.treasury.gov.za/documents/national%20budget/default.aspx Burns, A., Van Rensburg, T.J., Dybczak, K. and Bui, T. (2014). Estimating potential GDP in developing economies, Journal of Policy Modeling, 36 (2014-07), 700-16. Calderón, C., & Schmidt-Hebbel , K. (2008). Business cycles and fiscal policies: The role of institutions and financial markets. Central Bank of Chile Working Paper, 481. Calderón, C., Chuhan-Pole, P., & López-Monti, R. M. (2017). Cyclicality of Fiscal Policy in Sub- Saharan Africa: Magnitude and Evolution. World Bank Policy Research Working Paper, 8108. Caldara, D. and Camps, C. (2008). What are the effects of fiscal policy shocks, European Central Bank Working Paper No. 877/2008, Frankfurt am Main, European Central Bank. Cespedes, L.F. and Velasco A. (2013). Was this time different? Fiscal policy in commodity republics. Journal of Development Economics 106 (2014) 92-106. Cochrane, J.H. (1988). How big is the random walk in GNP, Political Economy, 96, 893-920. del Granado, J. A., Gupta, S., & Hajdenberg, A. (2010). Is Social Spending Procyclical? IMF Working Paper, 10(234). Du Plessis, S. A. and W. H. Boshoff (2007). The potential for counter-cyclical fiscal policy in South Africa. Stellenbosch, University of Stellenbosch, Stellenbosch working paper 13/2007. Du Plesses, S.A., Smit, B. and Steinbach, R. (2014). A medium-sized open economy DSGE model of South Africa. South African Reserve Bank Working Paper WP/14/04. 26    Elmendorf, D.W. and Furman, J. (2008) If, When, How: A Primer on Fiscal Stimulus. Washington DC: The Brookings Institution. Eckardt, S., & Mills, Z. (2014, January). What Goes Up Must Come Down: Cyclicality in Public Wage Bill Spending. World Bank Policy Research Paper(6760). Fölscher, A. (2007). Reforming Public Expenditure Management in Developing Countries: The African Case - Country Case Study: South Africa. In A. Shah, Budgeting and Budgetary Institutions (pp. 501- 534). Frankel, J. A., Végh, C. A., & Vuletín, G. (2013). On graduation from fiscal procyclicality. Journal of Development Economics(100), 32-47. Gali, J. and Monacelli, T. (2005). Monetary policy and exchange rate volatility in a small open economy, The Review of Economic Studies, 72(3), 707-734. Gavin, M., Hausmann, R., Perotti, R., & Talvi., E. (1996, March). Managing Fiscal Policy in Latin America and the Caribbean: Volatility, Procyclicality, and Limited Creditworthiness. Inter-American Development Bank Office and the Chief Economist, Working Paper 326. International Budget Partnership. (2018). Open Budget Survey 2017. Washtington DC: International Budget Partnership. Leeper, E.M., Walker, T.B. & Yang, S‐C.S. (2013). Fiscal foresight and information flows. Econometrica,  81( 3),  Kaminsky, G., Reinhart, C., & Végh, C. A. (n.d.). When It Rains It Pours: Procyclical Capital Flows and Macroeconomic Policies. In M. Gertler, & K. S. Rogoff, NBER Macroeconomics Annual 2014. Cambridge, MA: MIT Press. Keyfitz, R. (2004). Currencies and commodities: Modeling the impact of exchange rates on commodity prices in the World Market. The World Bank, Development Prospects Group. Washingtong D.C. Lamo, A., Pérez, J. J., & Schuknecht, L. (2013). The cyclicality of consumption, wages and employment of the public sector in the euro area. Applied Economics, 45(12), 1551-1569. Lane, P. R. (2003). The Cyclical Behaviour of Fiscal Policy: Evidence from the OECD. Journal of Public Economics, 87. Parliamentary Budget Office of South Africa (2016). Assessing the Performance of National Treasury’s Economic and Fiscal Forecasts. Cape Town: Parliamentary Budget Office. Perry, G. (2003). Can Fiscal Rules Help Reduce Macroeconomic Volatility in the Latin America and Caribbean Region? World Bank Policy Research Working Paper(3080). Russek, F. and Kowalewski, K. (2015). How CBO estimates automatic stabilizers. CBO WP. 2015-07, Congressional Budget Office. Strawczynski, M. and Zeira, J. (2012). Procyclicality of fiscal policy in emerging countries: The cycle is the trend. Published in Fiscal Policy and Macroeconomic Performance, Vol. 17, Edited by Jordi Gali and Luis Cespedes, Central Bank of Chile, 2013. Swanepoel, J. A. and N. J. Schoeman. (2003). Countercyclical fiscal policy in South Africa: Role and impact of automatic fiscal stabilisers, South African Journal of Economic and Management Sciences, 6(4): 802-822 27    Taylor, J.B. (2000) Reassessing Discretionary Fiscal Policy, Journal of Economic Perspectives, 14(3), 21-36. Transparency International. (2016). Corruption Perceptions Index 2015. Berlin: Transparency International. Unemployment Insurance Fund. (2016). Unemployment Insurance Fund – Annual report 2016/17. Pretoria: Department of Labour”. World Economic Forum. (2016). Global Competitiveness Index. Geneva: World Economic Forum. World Bank. (2015). Africa's Pulse (April). Washington, DC: World Bank. World Bank (2018). Global Economic Prospects: The Turning of the Tide? Washington, DC: World Bank. World Bank Group (2018) An Incomplete Transition: Overcoming the Legacy of Exclusion in South Africa. Cape Town: University of Cape Town Press. 28