![]()
1 The resource curse, commodity prices and economic growth
David I. Harvey, Neil M. Kellard, Jakob B. Madsen and Mark E. Wohar
1 Introduction
The recent ‘food crisis’ in 2008 starkly demonstrated the importance of global commodities markets. In particular, sharply rising prices for commodities such as wheat, maize, rice and oil pushed many vulnerable groups deeper into poverty (see Ivanic and Martin 2008).1 In any case, the behaviour of primary commodity prices is typically important to many developing countries, wherein a significant proportion of national income is often generated by a small number of primary products (see Harvey et al. 2010).2 The nature and causes of movements in primary commodity prices therefore have significant implications for growth and poverty reduction policies in developing countries.
Analysis of long-run commodity prices is dominated by the Prebisch-Singer (PS) hypothesis, which implies a secular, negative trend in commodity prices relative to manufactures.3 Possible theoretical rationales include low income elasticities of demand for commodities, asymmetric market structures that result from comparatively homogeneous commodity producers generating highly competitive commodity markets whilst facing oligopolistic manufacturing markets, and technological and productivity differentials between core (industrial) and periphery (non-industrial) countries. If a country’s export commodities present long-run downward trends in their relative prices, the policy advice is typically to diversify the export mix to include significant proportions of manufactures and/or services. Additionally, as is noted in Arezki et al. (2014a), understanding the trend and other time series characteristics should enable improved forecasting of commodity price movements.
Empirical evidence examining the PS hypothesis provides an ambiguous picture. The vast majority of recent studies employ the Grilli and Yang (1988) dataset of 24 annual non-fuel primary commodity prices which commences in 1900.4 However, the relatively large variance of commodity prices (see Deaton 1999) and the possibility of trend structural breaks inhibits statistical determination of any trend magnitude and direction with this sample size. A possible approach to address this issue is to provide greater degrees of freedom via a backwards extension of the sample. Recently, Harvey et al. (2010) and Arezki et al. (2014b) employ a unique disaggregated dataset, comprised of 25 separate commodity time series and spanning the 17th to the 21st centuries.
Although there is a large body of literature examining the veracity of the PS hypothesis, there are far fewer attempts to establish a plausible rationale for any estimated trend and associated breaks. In many ways, this is understandable; reliable historical data on possible causal variables such as technology differentials, relative market structures or income elasticities of demand are clearly difficult to source. However, the primary goal of this chapter is to explore the causal underpinnings of long-run commodity prices and hence to provide input to long-range forecasting and policy design. To circumvent the lack of data we suggest a threefold approach. Firstly, aggregate measures of relative commodity prices would be useful over a very long time frame. An aggregate measure of commodity price movements should smooth the idiosyncratic effects of individual commodity prices and provide a summary series shaped primarily by common factors. Secondly, the time series characteristics of these new series can subsequently be cross-referenced with the existing economic history literature to elucidate possible common factors. Finally, we suggest that the trend growth rate of economic activity, data that is available over an historical timeframe, is a useful proxy for these common factors (see, for example, Sachs and McArthur 2002).
In this chapter, our first contribution is to collect a large historical dataset on the export values of 23 individual commodities, which is not a straightforward task. This new data is then used as weights when combined with updated individual commodity series from Harvey et al. (2010) to create aggregate annual series beginning in 1650 and running continuously until 2010. Subsequently, our second contribution is to use the new aggregate series to examine the statistical properties of relative commodity prices over the very long run. Given the well-known problems of identifying the order of integration of price series, and the pervasive influence of any unit root/stationarity pre-tests on subsequent tests of commodity time series characteristics (see Harvey et al. 2010), we apply trend tests and multiple trend break tests which are robust to whether or not the series under consideration contains a unit root. The results show that the trend path of our new aggregate series can be split into four regimes (i.e. 1650 to the early 1820s, the early 1820s to the early 1870s, the early 1870s to the mid-1940s, and the mid-1940s to 2010). Through all but the second regime, a long-run downward trend can be clearly detected, giving new historical support to the PS hypothesis.
Our understanding of the regimes suggested earlier can be deepened...
