Identifying Aggregate Supply and Demand Shocks in South Africa¶
Stan du Plessis; Ben Smit; Federico Sturzenegger (2008). Identifying Aggregate Supply and Demand Shocks in South Africa. CID Working Paper No. 164, Center for International Development, Harvard University (South Africa Growth Initiative). Center for International Development, Harvard University.
| Type | working paper |
| Scope | Economics and economic history of apartheid |
| Method | econometric |
| Period studied | 1960–2006 |
| Regions | South Africa |
| Themes | macro fiscal finance debt, methods data sources, transition post apartheid, inequality poverty unemployment, sanctions divestment |
| Citation key | plessis2008 |
Argument¶
The paper applies a three-variable structural VAR (following Clarida and Galí 1994) with long-run restrictions to split South African real GDP into permanent supply shocks and transitory fiscal and monetary demand shocks. Cumulated supply shocks give a measure of potential output, and the demand shocks give a measure of the business cycle. The authors argue the identification is plausible because the impulse responses match theory, the business-cycle measure matches SARB and Du Plessis turning points, and the supply-shock series lines up with events economic historians treat as major shocks. The result is a large cumulative fall in potential GDP from the mid-1970s to the mid-1990s, followed by a rapid recovery after the 1994 transition.
Evidence and method¶
Seasonally adjusted quarterly SARB series, 1960Q2-2006Q4 (Q4 2006 is an estimate): first difference of log real GDP, government consumption/GDP as the fiscal proxy, and the real interest rate built from the discount rate and CPI. The model uses four lags and two samples (1960+ and 1983Q4+). Supporting material includes ADF unit-root tests, variance decompositions, impulse responses, historical decompositions, and comparison with SARB and Du Plessis (2006) turning points and with historical accounts (Moll 1991; Feinstein 2005).
Place in the debates¶
- Economic cost of apartheid / did the apartheid economy 'fail' (Moll vs Feinstein) (see Did the apartheid economy fail? Growth, productivity, investment and the post-1970s crisis). Provides a quantified estimate (about 30% fall in potential GDP relative to trend) that the authors say is consistent with historical accounts by Moll (1991) and Feinstein (2005). It treats the late-apartheid years as a period of cumulative negative supply shocks.
- Methods for estimating potential GDP in South Africa (HP filter and production function vs structural VAR) (see Data reliability and econometric method). Argues for an SVAR long-run restriction approach that jointly yields potential output and a business-cycle measure, rather than the mostly HP-filter and production-function approaches used in earlier South African work.
- Pro-cyclicality and role of fiscal vs monetary shocks in output fluctuations (see Did the apartheid economy fail? Growth, productivity, investment and the post-1970s crisis). Finds that supply shocks dominate GDP over the long run. Fiscal shocks dominate monetary shocks in the short-to-medium run only in the longer sample, and the ranking reverses after 1983, which the authors link to the De Kock-era monetary regime change.
Findings¶
-
Potential GDP fell about 30% relative to trend between the mid-1970s and mid-1990s, which the authors read as a quantification of the cost of maintaining apartheid amid adverse international shocks. (p. 16)
On this measurement potential GDP declined by a cumulative 30% relative to trend between the mid-seventies and the mid-nineties.
-
Positive supply shocks of the 1960s reversed from 1973. The authors attribute the subsequent adverse shocks to international events, hostile border states, domestic unrest, the 1985 debt standstill and the unravelling of National Party power, probably worsened by human and capital flight. (p. 16)
Domestic unrest in 1976 and again in the mid-1980s, the debt standstill in 1985, and the final unravelling of the National Party’s grip on political power were all along the path of declining potential GDP
-
Both models show continued negative supply shocks from the mid-1980s to 1994, followed by positive supply shocks that raised actual and potential GDP. (p. 15)
both recording continued negative supply shocks from the mid-eighties through 1994, after which positive supply shocks moved actual and potential real GDP up
-
Potential GDP rose sharply almost immediately after the political transition, growing at a healthy rate but below government targets. (p. 17)
The new measure of potential GDP also shows a marked rise in potential GDP since 1994, starting almost immediately after the political transition.
-
Post-1994 potential growth is estimated at 2.5% to 4.4% a year, with about 3.5% the likely point estimate. (p. 17)
The estimates in Table 3 suggest a range of 2.5% to 4.4% for the growth rate of potential GDP, while 3.5% would be a likely point estimate
-
Long-run GDP is dominated by supply shocks. In the 1960+ model, fiscal shocks outweigh monetary shocks in the short-to-medium run, but the ranking reverses in the post-1983 sample. (p. 13)
Firstly, the long-run development of real GDP is dominated by the history of supply shocks.
-
The identified demand-shock measure matches SARB and Du Plessis business-cycle turning points well. The one anomaly, the 1984 contraction, is explained as a sharp negative supply shock (falling public investment, depressed terms of trade) alongside expansionary monetary policy. (p. 18)
The measure of aggregate demand identified with the long-run restrictions matches the periods of relative expansion and relative contraction in the South African economy surprisingly well.
-
The monetary policy regime changed around 1980, with monetary shocks becoming more dispersed and fiscal and GDP shocks less so, which the authors tie to the De Kock Commission reforms. (p. 13)
The monetary policy shocks have become more dispersed since the 1960s, while the opposite seems to have happened with the GDP growth and fiscal policy shocks.
-
Appendix tests cannot reject a unit root in the fiscal proxy (government consumption/GDP), raising a stability concern, but all SVAR eigenvalues lie inside the unit circle. (p. 33)
it is not possible to reject the hypothesis of a unit root for the fiscal policy proxy. This raises a concern over the stability of the VAR.
Evidence ledger¶
Sources used¶
| Source | Kind | Period | How used |
|---|---|---|---|
| SARB quarterly real GDP series (RB6006D), seasonally adjusted, 1960Q2-2006Q4 | official statistical series | 1960Q2-2006Q4 | First difference of log real GDP is the first variable in the three-variable structural VAR, from which supply shocks and potential output are derived. |
| SARB government consumption series (RB6008D) divided by GDP (RB6006D) | official statistical series | 1960Q2-2006Q4 | Ratio of government consumption to GDP serves as the proxy for fiscal policy stance, the second VAR variable. |
| SARB discount rate and CPI index (RB 7032N), monthly | official statistical series | 1960-2006, monthly aggregated to quarterly | Real interest rate computed from monthly discount rate and CPI with a within-quarter formula is the monetary policy proxy, the third VAR variable. |
| De Kock Commission (1985), Final report of the commission of inquiry into the monetary system and monetary policy in South Africa (RP 70/1984) | commission or official report | 1985 | Cited for the shift in monetary policy regime in the 1980s from quantitative controls to market-based instruments, which motivates the 1983Q4 sub-sample and explains changes in monetary shock dispersion. |
| SARB business cycle dating committee turning points | official statistical series | 1960s-2006 | Official reference cycle turning points are compared against the identified cumulative demand shock as an independent check on the business cycle measure. |
| Du Plessis (2006) alternative business cycle turning points from a non-parametric dating algorithm applied to real GDP | secondary literature compilation | 1960s-2006 | Alternative turning points serve as a second benchmark against which the new business cycle measure is compared. |
| Du Plessis and Smit (2007) analysis of the 1984 contraction | secondary literature compilation | 1984-85 | Used to explain the apparent anomaly of a positive demand shock during the post-June 1984 contraction (large policy rate cuts). |
| Frankel, Smit and Sturzenegger (2007), CID Working Paper No. 133 | secondary literature compilation | 1984; post-1994 | Supplies the likely point estimate of 3.5% potential growth and the explanation (public investment fall, terms of trade) for the 1984 supply shock. |
| Historical accounts of apartheid's economic costs: Moll (1991) and Feinstein (2005) | secondary literature compilation | apartheid era | Used qualitatively to judge the plausibility of the estimated 30% decline in potential GDP and to match shocks to historical events. |
New data¶
The authors construct new quarterly estimates of potential GDP and a business cycle (output gap) measure for South Africa from structural VAR historical decompositions, for 1960Q2-2006Q4 and 1983Q4-2006Q4, reported in Appendix B. No new primary data were collected.
Inference¶
Quantification: Econometric (causal design).
Three-variable structural VAR in the Clarida-Galí style (real GDP growth, government consumption/GDP, real interest rate) with long-run restrictions identifying supply, fiscal and monetary shocks. Plausibility is judged from impulse responses, variance decompositions, a historical decomposition, a match with SARB and Du Plessis business cycle turning points, and a narrative matching of supply shocks to historical events. Two overlapping samples serve as a robustness check.
Key claims¶
| Claim | Basis | Strength |
|---|---|---|
| Potential GDP fell by about 30% relative to trend between the mid-1970s and mid-1990s, a quantified cost of the late apartheid period and adverse external shocks. (p. 14) | Cumulated supply shocks from the longer-sample SVAR on SARB GDP, government consumption and real interest rate data, matched to events in Figure 10 and judged against Moll and Feinstein. | Partial data |
| Supply shocks dominate long-run GDP variance; fiscal shocks outweigh monetary shocks in the long sample, while the order reverses in the post-1983 sample. (p. 11) | Variance decomposition of real GDP in Table 2 for the two SVAR models. | Systematic data |
| Potential GDP rose sharply after the 1994 transition, growing at roughly 2.5%-4.4% a year, with about 3.5% as the likely point estimate. (p. 15) | Table 3 growth rates of the identified potential GDP over post-1994 sub-samples for both models, with the point estimate taken from Frankel et al. (2007). | Partial data |
| The identified demand shock series matches the South African business cycle, especially the Du Plessis (2006) turning points, supporting the identification scheme. (p. 16) | Visual comparison in Figure 11 of cumulated demand shocks with SARB and Du Plessis turning points. | Partial data |
| Negative supply shocks from 1973 onward were associated with the first oil shock, the collapse of the Portuguese colonies, domestic unrest in 1976 and the mid-1980s, the 1985 debt standstill and the unravelling of National Party rule. (p. 14) | Narrative matching of the cumulated supply shock series to historical events (Figure 10) and economic history literature; no formal test. | Illustrative or anecdotal |
| The impulse responses are consistent with theoretical priors: supply shocks permanently raise GDP, a positive fiscal shock is temporarily expansionary, and a positive real interest rate shock is temporarily contractionary. (p. 10) | Impulse response functions in Figure 4 from the SVAR estimates for both samples. | Systematic data |
| The apparent positive demand shock during the 1984 contraction is an anomaly explained by a sharp negative supply shock, with a fall in public investment and depressed terms of trade named as the likely causes. (p. 16) | Du Plessis and Smit (2007) and Frankel et al. (2007), cited rather than tested within the paper. | Asserted |
The author's caveats¶
The authors note that the Q4 2006 data are an estimate. They also note that the unit root test cannot reject a unit root for the fiscal proxy over the longer sample, though eigenvalues lie inside the unit circle. The identification requires strong assumptions, and Blanchard and Quah's scheme is at best 'nearly correct'. They add that the model is not sensitive enough to identify the precise nature of individual supply shocks.
Gaps in the evidence¶
The supply-shock narrative is matched to events only by eye, with no direct data on capital flight, sanctions, emigration, terms of trade or public investment to test the causal interpretation of the 30% decline. Independent measures of potential output (production function estimates, capital stock and labour series) or counterfactual comparators would test the apartheid cost estimate better; most of these existed and could be compiled. Confidence bands for the historical decompositions and the sensitivity of results to the government consumption proxy and the identifying restrictions are not reported but could be computed.
Replicability: Data described but not public.