The structuralist theory of imported inflation: an application to South Africa

Abstract

This study emphasizes the importance of identifying the origin of inflation in the present context of inflation targeting by many emerging market and transition economies. The analysis shows, based on South African data, how structural (supply) and demand inflation can be distinguished. The results indicate that South Africa's inflation experience between 1973q1 and 1998q4 is characterized by two monetary regimes. During the first regime (1973q1–1984q4) the long-run cause of inflation is demand-pull. The second regime (1987q1–1998q4) represents major changes to structural (‘imported’) and cost-push inflation. The two-year period 1985–1986 signifies structural change from the first to the second regime. Moreover, the results in the second regime remain robust when the inflation model is subjected to ‘new’ out-of-sample data until 2001q2. Evidence of structural (‘imported’) inflation in the second regime suggests that inflation should not entirely be squeezed out of the system nor should it necessarily be kept at the lowest possible level, because some inflation may be regarded as the natural by-product of the growth and development process. South Africa's inflation experience points to several lessons for existing (and potential) emerging market and transition economies with some form of inflation targeting.

Acknowledgements

The author is thankful to Alan Carruth, Andy Dickerson, Basil Moore, Maria de Mello, Tony Thirlwall and an anonymous referee for valuable comments on an earlier draft of this paper. The usual disclaimer applies.

Notes

Although New Zealand and Canada ‘initiated’ such a framework, emerging market and transition economies such as Brazil, Chile, the Czech Republic, Israel, Korea, Poland, Colombia, Thailand and South Africa at the same time adopted many elements of inflation targeting regimes used in more advanced economies (Amato and Gerlach, 2002).

The study identifies unit labour costs and exchange rate changes as the main determinants of inflation. For useful surveys of previous inflation studies in South Africa, see Akinboade et al. (2002) and Hodge (2002).

For an excellent overview of the monetarist-structuralist debate, see Thirlwall (2003).

South Africa's major suppliers include the United States, the United Kingdom, Japan, Germany, France, Italy and the Netherlands. The import share of each country was obtained from International Monetary Fund, Direction of Trade Statistics (various issues).

For example, the idea that inflation rates in emerging market and transition economies should closely approximate those in more advanced economies is advanced by Amato and Gerlach (2002). By contrast, in structuralist writings, inflation up until 10% is regarded as ‘mild’ (see Ghatak, 1995).

Structuralist models assume that developing economies have unemployed resources, thus providing ample scope for demand-side policies to promote faster growth (see Gibson and Van Seventer, 2000).

Unit root tests were based on Phillips and Perron's (1988) semi-parametric correction to the Dickey-Fuller test and Perron's (1989) dummy variable approach when the time series contain structural breaks. The tests showed that all the variables are I(0) in inflation spiral Equation 1 and I(1) in import price Equation 2 during 1973q1–1998q4. All these results are available from the author.

At one extreme, a value of k = 0 implies that potential and actual output are the same, so that a Phillips curve relation disappears. At the other extreme, a very high value of k suggests that most of the movements in the business cycle are associated with actual output and not potential output. For a more detailed discussion of the underlying methodology, see Clark et al. (1996).

The main empirical results in the next sections are invariant irrespective of whether the output gap is defined in levels or growth rates. However, based on a range of non-nested tests described in Pesaran and Pesaran (1997), the unrestricted inflation spiral Equation 8 with the output gap in growth rates outperformed the ‘rival’ model with the output gap in levels.

Unless indicated otherwise, all the data in this paper are obtained from the South African Reserve Bank's historical data set on the internet (http://www.resbank.co.za/Economics/econ.html).

Unlike one-step ahead forecasts, multi-step dynamic forecasts provide a ‘true’ test of the forecasting ability of an empirical model. Multi-step dynamic forecasts are made under the more realistic assumption that the actual inflation rate is unknown during the forecast period (Charemza and Deadman, 1997).

To visualize the decelerating trend more clearly, a large outlier in the third quarter of 1998 is omitted from the inflation series in Fig. 3.

For a more comprehensive discussion of restrictive monetary policy measures during the 1980s and 1990s, see Gibson and Van Seventer (2000); Mohr and Rogers (1995) and Weeks (1999).

The out-of-sample forecast period starts in the fourth quarter of 1999 because the model showed some signs of predictive failure in the first three quarters of 1999. Predictive failure over these three quarters may be attributed to a downward shift in inflation expectations following the sharp exchange rate depreciation in 1998q3 and its subsequent recovery in 1998q4. However, the inflation spiral model quickly recovers its good predictive ability from 1999q4 until 2001q2 (see Fig. 5).