Estimating South Africa's Output Gap and Potential Growth Rate

• Date: 01 June 2017
• Author: Johannes W. Fedderke,Daniel K. Mengisteab
• DOI: http://doi.org/10.1111/saje.12153
• Published date: 01 June 2017

ESTIMATING SOUTH AFRICA’S OUTPUT GAP AND

POTENTIAL GROWTH RATE

JOHANNES W.FEDDERKE*

,†,‡

AND DANIEL K.MENGISTEAB

†

Abstract

This paper estimates the potential output of the South African economy using several filters. We

demonstrate that potential output measures are very sensitive to the different methodologies. We

also provide estimates of South Africa’s potential growth rate over the 1960–2015 period. Cur-

rent estimates of the potential growth rate fall in the 1.9%–2.3% range. However, the evidence

suggests that the rate is under considerable downward pressure. South African potential growth

may be headed toward the 1% range. The strongest decline is in the real sectors of the economy

(Manufacturing, Mining), the greatest resilience in the service sectors (financial in particular).

JEL Classification: E23, E32

Keywords: Measurement of potential output, measurement of potential growth, South Africa

1. INTRODUCTION

South Africa’s democratic transition was followed by a recovery in its growth perform-

ance, a growth acceleration that lasted until the 2000s, ending with the financial crisis of

2007/2008. Since the advent of the crisis, South African growth has reverted to much

lower levels, and in trend terms shows little signs of improvement. Given the substantial

welfare challenges that remain for South Africa, this raises the fundamental question of

what the structural growth capacity of the economy is, and whether the down-turn in

growth is temporary or constitutes a sign of a lack of competitiveness. In this paper, we

examine this issue by developing estimates of potential output in the economy, and the

implied output gaps and growth in potential output this implies.

Potential output is widely used in both economic theory, and the formulation of eco-

nomic policy. An immediate and important use of potential output is in determining

whether the economy is subject to a positive or negative output gap, and hence whether

demand-side policy should be expansionary or contractionary. A second application is

that the identification of potential output allows for inferences on the underlying struc-

tural growth performance of the economy.

* Corresponding author: Pennsylvania State University, USA, Economic Research Southern

Africa, South Africa, South African Reserve Bank Research Fellow, South Africa, and Univer-

sity of the Witwatersrand, South Africa. E-mail: jwf15@psu.edu

†

Pennsylvania State University.

‡

Economic Research Southern Africa, South African Reserve Bank Research Fellow, and Uni-

versity of the Witwatersrand.

We thank two anonymous referees for useful comments in improving an earlier version of

this paper. Fedderke acknowledges the research support of Economic Research Southern

Africa.

V

C2017 Economic Society of South Africa. doi: 10.1111/saje.12153

161

South African Journal of Economics Vol. 85:2 June 2017

South African Journal

of Economics

Yet in empirical application measurement of potential output has presented a peren-

nial challenge. Broadly speaking, two approaches to estimating potential output predom-

inate in empirical contexts: statistical filtering and approaches based on structural

models, for instance by the estimation of production functions. In addition, there exist a

range of additional multivariate approaches to output gap derivation.

The benefit of the statistical filtering approach is its simplicity. The disadvantages are

that the potential outputs generated are sensitive to the parameterisation of filters, the

lack of structural economic theoretical foundations to the filter, and hence the failure to

indicate underlying structural, and hence causal, drivers of potential output. Further dis-

advantages of filtering are the potential for spurious cycles being generated, the accuracy

of estimations in real time, particularly with the Hodrick–Prescott (HP) filter whose esti-

mates are less accurate at the end of sample. A useful summary of issues can be found in

Kramer and Farrell (2014).

The production function approach involves the estimation of a production function

to determine the potential output of an economy. The advantage of using a production

function rather than statistical filters, derives from its greater foundation in economic

theory, by linking potential output of an economy to factor inputs. Any changes in the

factor inputs of the economy would be evident in the potential output generated by the

production function. However, production function approaches have their own disadvan-

tages arising from data limitations, particularly as regards the measurement of capital,

questions as to the appropriate functional form to be employed in the production func-

tion, and econometric concerns surrounding identification.

Other approaches used to estimate potential output include dynamic stochastic gen-

eral equilibrium models, and structural vector autoregressive models, amongst others.

The appeal of such models is the greater explicitness with which economic theory is

incorporated. On the other hand, the relative complexity of specifications that result,

makes inference increasingly sensitive to the underlying parameterisation of the model,

which is often incompletely verified through econometric testing. Data limitations can

further constrain successful implementation.

The purpose of this note is to consider the sensitivity of estimated potential out-

put and the implied output gap to alternative methodologies. In doing so, we move

beyond previous contributions to the South African debate, by considering a wider

range of statistical instruments in deriving potential output and associated gap meas-

ures. This allows for a comparison of the sensitivity of inference to the methodology

adopted.

Our focus is primarily on the use of alternative filters. An important motivation for

this is that the Hodrick–Prescott (1997) (HP) filter has become popular in applications

in economics. One important reason for this popularity, is that the filter will identify the

deterministic trend in a series, while removing stochastic trends (be robust to unit roots),

even for series subject to higher order of integration (King and Rebelo, 1993). Doubts

concerning this claim emerge from the finding of Cogley and Nason (1995), who dem-

onstrate that the HP filter when applied to an integrated time-series can generate busi-

ness cycle characteristics even when none are present in the original data. An explanation

for this comes from the Phillips and Jin (2015) consideration of the asymptotic distribu-

tion theory for the HP filter. They show that when the HP filter eliminates deterministic

trends, it does not eliminate stochastic trends (unit roots) as anticipated, thus providing a

formal account of why the HP filter can generate “spurious” cycles in the filtered data.

162 South African Journal of Economics Vol. 85:2 June 2017

V

C2017 Economic Society of South Africa.