Journal Issue
Share
Chapter

3. Estimation Results

Author(s):
Nir Klein
Published Date:
August 2011
Share
  • ShareShare
Show Summary Details

Table 1 presents the estimation results by each methodology divided into several subsamples to identify whether there was a significant shift in potential output growth over the sample’s period. The subsamples are (1) the apartheid regime era in which the economic growth was burdened by economic isolation and the international sanctions (1985:Q1–1994:Q1); (2) the post-apartheid regime until the financial crisis adversely affected South Africa’s economic growth (1994:Q2–2008:Q4); (3) the financial crisis period (2009:Q1–2009:Q4), which was reflected in continuous contraction of real GDP (on a year-on-year basis); and (4) the recent period of economic recovery (2010:Q1–2010:Q4).

Table 1.The Average of Actual Output, Potential Output, and Output Gap
Potential output growth
PeriodActual output growthPotential output growthHP FilterBaxter-KingKalman filter ModelSVARPF
Model 1Model 2Model 3
1985Q1–94Q10.6330.8791.0070.9960.8980.7850.7560.7660.951
1994Q2–2010Q43.1923.2423.1713.1773.3093.3753.2953.1683.189
1994Q2–2008Q43.5533.4313.2473.3033.4583.6283.6173.4453.324
2009Q1–09Q4−1.6961.9232.6411.8352.5822.7301.540−0.5042.638
2010Q1–10Q42.7431.7392.5842.6601.8370.2840.3032.7491.760
PeriodAverage output gapOutput gap as a percent of potential output
1985Q1–94Q1−0.194−0.301−0.292−0.070−0.328−0.4010.201−0.173
1994Q2–2010Q40.2750.142−0.0061.2050.5220.333−0.3840.115
1994Q2–2008Q40.5750.3630.1701.7741.0860.692−0.4300.372
2009Q1–09Q4−2.433−1.547−1.341−3.441−4.872−3.529−0.040−2.266
2010Q1–10Q4−1.435−1.421−1.259−2.536−2.413−1.090−0.046−1.284

By and large, the estimated trajectories of the output gap seem to be highly correlated, although the magnitude of the gaps varied in some periods (Table 1 and Figure 2). The only period in which the estimations differ substantially is the 2000–04 cycle. In this period, which includes the deterioration of external conditions in 2001–02 due to the slowdown of the global economy, the SVAR and the PF approaches estimate a negative output gap due to continued strong pace of potential output, while the Kalman filter models estimate a deceleration in potential output growth and therefore yield a positive output gap. The diverging estimations for this period lead to a relatively weak correlation between these three methodologies (Table 2).

Figure 2.Output gap by different methodologies*

Source: IMF staff calculations.

* The Hodrick-Prescott filter is plotted in all chart for comparison.

Figure 3.Actual and Potential output (in natural log terms)

Source: IMF staff calculations.

Table 2.Output Gap Correlation Matrix
HPBaxter-KingKalman filter, model 1Kalman filter, model 2Kalman filter, model 3SVARPF
HP1.00000.96150.64000.50250.46890.63210.7522
Baxter-King0.96151.00000.60670.48690.45580.62980.7219
Kalman filter, model 10.64000.60671.00000.92860.88270.15270.5253
Kalman filter, model 20.50250.48690.92861.00000.97670.13160.3966
Kalman filter, model 30.46890.45580.88270.97671.00000.17800.3597
SVAR0.63210.62980.15270.13160.17801.00000.5883
PF0.75220.72190.52530.39660.35970.58831.0000

The comparison between the subsamples clearly indicates that the potential GDP growth varied significantly over the years. In particular, the estimations show that in the first period (1985:Q1–1994:Q1), the potential GDP growth was very modest, at slightly below 1 percent, emphasizing the high economic costs associated with isolation of the apartheid regime.8 In the second period (1994:Q2–2008:Q4), the potential output growth accelerated to nearly 3½ percent on average, reflecting also the rapid increase in employment and the steep increase in total factor productivity (see Figure A1 in Appendix 1).

While observations are still limited, the estimations also reveal that potential output growth has significantly decelerated to slightly below 2 percent following the global financial crisis, thus suggesting that the prevailing output gap is smaller than previously thought.9 In particular, the estimations show that the estimated output gap moved from an average level of 2 percent in 2007:Q1–2008:Q2, to an average level of about −2½ in 2009 (Figure 4). Given that the actual growth was higher than the estimated potential growth in 2010, output gap contracted and on average stood at about −1½ percent.

Figure 4.South Africa: The average output gap in percent of potential output by the seven methodologies1

1The shaded areas are defined by the Economic Cycle Research Institute (ECRI) as periods of recessions.

The breakdown of the potential output by the main three sectors reveals that their cyclical pattern is not identical (see Figure A.3 in Appendix 1). The output gaps of the primary and secondary sectors are significantly more volatile than that of the tertiary sector, which accounts for nearly two-thirds of the total added value. The relatively low volatility of the tertiary sector’s output gap, which largely represents the government’s services, may reflect the direct impact of the government’s countercyclical fiscal policy. Additionally, although the output gap of all the three sectors turned to negative levels with the outbreak of the financial period, the potential output of the primary sector seems to have recovered toward the end of 2010, and its output gap reverted to positive levels. The output gaps of the secondary and tertiary sectors remained negative in 2010, though they show signs of improvement.

Figure 5.South Africa: Estimated potential output growth, 1994-2010

The deceleration in potential output growth in 2009–10 reflects the impact of the financial crisis on the multifactor productivity, as well as on the labor and capital factor inputs, More specifically, the decline in innovative activities in advanced economies, which was in part due to the recent tightening in credit conditions and the problems in firms’ balance sheets, was reflected in a considerable decline in the multifactor productivity in most advanced economies during the financial crisis episode (Figure 6). This is likely to have an impact on South Africa’s productivity growth as well, particularly in industries that are fast adopters of new technologies.

On labor, the nontrivial job shedding of nearly 1 million employees may not be fully reversed in the short-term, not only because of the frictions in the labor market and the relatively long adjustment lags, but also due to the dynamics in the labor force.10 And indeed, since 2008, the participation rate has declined by some 4 percentage points and resulted in a growing number of discouraged work-seekers (see Figure A.4 in Appendix 1).11

Figure 6.The average change in multi factor productivity in selected advanced economies, 2008-09*

Source: OECD.

*For Australia, and New Zealand the figures refer to 2008 only.

The deceleration of potential growth also stems from the sharp drop of investment, which affected the pace of the accumulation of capital stock. The overall gross fixed domestic investment declined by 3½ percent of GDP since 2008, largely reflecting the contraction of the private sector’s investment, particularly in the manufacturing and agriculture sectors (Figure 7). Given that labor and capital are often viewed as complementary input factors, the decline in capital stock is likely to affect job creation in the near future.

Figure 7.South Africa: The change in gross fixed domestic investment and capital stock by sectors (constant prices)

External Demand and the Potential Output Growth

A possible explanation for the continued contraction of employment and private investment in 2010 is the substantial and sharp decline in external demand for South African exports in the past two years (Figure 8). During 2009, the South African exports volume fell by 20 percent, and registered only modest recovery in 2010 despite the relatively strong global economic recovery. Consistent with this explanation, the breakdown of the nonagricultural employment to tradable and nontradable sectors indeed confirms that the recent employment loss is exclusively concentrated on the exporting sectors, most notably manufacturing (Figure 9).

Figure 8.South Africa: Real GDP, employment and exports volume (1999=100)

Figure 9.South Africa: Nonagricultural employment (2008:Q2=100)

The fact that South African exports volume failed to recover despite the relatively strong demand of its trading partners may suggest that the decline in external demand is not entirely cyclical and may partly reflect structural factors, including low competitiveness. In this regard, at some 30 percent, the appreciation of South Africa’s real effective exchange rate since the eve of the financial crisis is the highest among South Africa’s peers, is not entirely explained by fundamentals, and significantly reduces South Africa’s ability to compete in the international markets (see Figure A.5 in Appendix 1). Additionally, the sluggish economic recovery in advanced economies and the expectation that it will continue over the medium-term add to South Africa’s weak external demand. The latter was also reflected in the decline in the share of exports of goods to Europe, which until 2008 was South Africa’s largest trading partner, to 28 percent in 2009–10 from a share of 32 percent in 2008 (see Figure A.6 in Appendix 1).

Has the Financial Crisis Led to a Structural Break in Potential Output Growth?

In view of the substantial differences in potential output growth between the examined subsamples, this section evaluates, using Chow Breakpoint test, whether there was a structural break around 2008:Q4 or if it was part of the “normal” volatility of the estimated series. Because the exact timing of the shift in growth is unclear, we also examine the possibility of a structural break two quarters before and after these points. To exclude the impact of the apartheid regime era, the regressions were estimated for the 1994:Q2–2010:Q4 period using the lagged dependent variable as explanatory variables. The probability values of the Chow Breakpoint Test are presented in Table 3.

Table 3.Chow Breakpoint Test1,2
2008Q208Q308Q409Q109Q2
HP (5)0.0130.0090.0090.0310.034
Baxter-King (6)0.6320.2070.1460.1010.947
Kalman filter, model 1 (1)0.0410.0810.0040.0160.103
Kalman filter, model 2 (6)0.1690.1540.0130.0240.032
Kalman filter, model 3 (6)0.2740.2120.1810.1700.069
SVAR (1)0.2760.0340.0080.0040.272
PF (1)0.8050.5930.5690.4020.204
Average potential growth (5)0.1550.0110.0130.0080.013

The results point to a structural break in potential output growth following the outbreak of the global financial crisis as, apart from the potential output that was estimated by the PF approach, the null hypothesis can be rejected in at least one of the examined quarters. The results also point out that the structural break probably took place in 2008Q4 or 2009Q1 given that the null hypothesis can be rejected in four out of the seven estimated series.

But going forward, it is not clear whether the slower pace of potential growth reflects a transitional phase, in which the level of potential output will revert back to the precrisis trajectory as demonstrated in Scenario A in Figure 10, or remain below that level for a protracted period. Scenario A implies a sharp acceleration of potential growth to around 5 to 6 percent over the medium- term, before returning to its precrisis pace. Alternatively, if potential growth resumes its precrisis pace of around 4 percent or continues at the current pace of below 2 percent, this will result in a permanent loss of potential output as illustrated in Scenario B and Scenario C, respectively. This said, based on the pace of recovery and the dynamics in the labor market thus far, it is most likely that Scenario B will prevail.12

Figure 10.South Africa: Alternative scenarios for potential output

The low potential growth rate for this period is consistent with the findings of Arora and Bhundia (2003).

While in the previous two subsamples, the estimations on average differ only marginally from each other, it is less so in the third and fourth subsamples. In the period of the financial crisis, the estimations for potential output growth vary in a relatively wide band ranging from 2¾ percent to a contraction of ½ percentage points. The estimated contraction of the potential output is derived from the SVAR methodology, which excludes the public sector’s impulse in this period.

The loss of a million jobs reflects a shift from a positive output gap of 2 to 3 percent on the eve of the financial crisis to a negative terrain in 2009–10. Therefore, in the process of returning to potential output, fewer than 1 million jobs are expected to be created.

Since 2008, the number of discouraged work-seekers has increased by 1.1 million to 2.2 million. This may lead to a permanent destruction in human capital, provoking further loss in the level of potential output.

Empirical evidence suggests that previous financial/debt crises were associated with large and permanent output loss in other emerging and developing economies (Cerra and Saxena, 2008).

    Other Resources Citing This Publication