Commodity prices played a critical role in the surge and decline of inflation since 2020. This column studies the relationship between commodity prices, consumer prices, and the business cycle for 15 economies in the years prior to World War I. It finds a strong positive relationship between inflation and commodity prices, even when focusing on commodities that are unlikely to have been in the consumer basket. Demand-induced increases in commodity prices, linked to the global business cycle, have a stronger effect on inflation.
Share
Authors
Stefan Gerlach
Rebecca Stuart
Advisor, Monetary Policy Division Central Bank Of Ireland
Fluctuations in oil and other commodity prices played a critical role in the surge and decline of inflation since the start of the Covid pandemic in 2020 (IMF 2024). The impact of supply bottlenecks on commodity prices in the post-Covid period, and of the Russian invasion of Ukraine on energy prices has boosted interest in the relationship between consumer prices and commodity prices. Examples of some recent contributions include Kilian and Zhou (2023) and Coutinho and Licchetta (2024), who study the impact of energy price shocks on inflation in the US and the euro area, respectively. But this is not a new phenomenon: the importance of drastic fluctuations in oil prices for consumer price inflation has been debated since the oil crises of the 1970s.
In a recent paper (Gerlach and Stuart 2024), we look further back and demonstrate that commodity price shocks had a big impact on inflation in a wide range of economies already before WWI. Importantly, we demonstrate that global business cycle swings led to fluctuations in commodity demand and had particularly strong effects on consumer price inflation in this period.
We use annual data on UK commodity prices in sterling and consumer prices in local currency in 15 economies for the period 1851-1913 to study this question. 1 The consumer price data are obtained from various sources and can be thought of as capturing the cost of living. Our commodity prices are sourced from Sauerbeck (1886, 1893, 1908, and 1917) 2 in which data on the prices of 43 commodities are compiled. One issue with historical consumer price series is that often wholesale prices are used as proxies for retail prices. We therefore restrict our analysis to metals and products that would neither enter the consumer basket, nor be used as proxies for items in the consumer basket.
Figure 1 shows the median international inflation rate in the 15 economies alongside the median commodity price inflation. Commodity prices are more volatile than consumer prices, but the two series move together.
Figure 1 Median commodity price (11 commodities) and median inflation rate (15 countries), annual percentage change, 1851-1913
We first use ordinary least squares (OLS) to estimate the relationship between inflation in country i on the lagged inflation rate and the median growth rate of the subset of commodities identified above. Our measure of commodity prices is significant at the 5% level in all cases except for Australia (see Table 5 of our paper for the full set of results). This result is striking when one recalls that these are primarily industrial goods which are unlikely to be included in the consumer basket, and which thus don’t have any direct channel through which they can impact consumer prices. The estimated coefficients range between 0.09 in France and 0.60 in Switzerland, and are 0.26 on average, suggesting that over a quarter of any change in commodity prices passes through to inflation within a year. The proportion of inflation explained by the model ranges from 8% in Australia to 48% in the US and is 24% on average.
An important theme in the literature is that commodity prices co-move strongly with global business cycles and demand factors (Pindyck and Rotemberg 1990, Alquist et al. 2020, Delle et al. 2022). One important question is whether the strength of the transmission of commodity price shocks depends on whether they reflect a global boom and demand for commodities, or supply factors that are likely to be commodity-specific. To explore this issue, we distinguish between demand and supply induced changes in commodity prices.
Figure 2 presents our measure of commodity price inflation and growth in industrial production in the UK from Crafts et al. (1989). The correlation between these two series is 0.43. 3 The shaded areas in Figure 2 show the periods of recession in the UK (based on Klovland 1998). While the relationship is not perfect, commodity prices tend to fall during downturns and rise during expansions.
Figure 2 Median percentage change of commodity price (11 commodities), and industrial production and business cycle chronology in the UK, 1851-1913
While the UK was one of the most important economies of the time, in our case a broader measure that captures the international dimension of the business cycle is preferable. Klovland (2003) shows that shipping freight rates capture movements in the global business cycle well. We measure freight rates using the median of rates from several sources (Klovland 2003, North 1958, Harley 1988) and deflate them using the UK GDP deflator. Figure 3 shows that there was strong co-movement between this measure of the state of the global business cycle and commodity prices.
Figure 3 Median commodity price and freight rates, annual percentage change, 1851-1913
To assess whether supply- and demand-induced movements in commodity prices have the same impact on inflation we compare the OLS estimates above that make no distinction between supply and demand factors with instrumental variable (IV) estimates obtained by using shipping costs as an instrument for commodity prices. If demand-induced increases in commodity prices have greater impact on inflation than supply-induced price changes, then we would expect the IV estimates of γi to be larger than the OLS estimates.
We find that the average parameter on the commodity variable is 0.35 when IV is used in contrast to 0.26 when the equation is estimated by OLS (the full set of IV parameters are available in Table 6 of our paper). The parameter estimates using IV range from 0.01 in the US to 0.84 in Switzerland. Efficiency is always an issue when IV is used. Nevertheless, the parameter is significant at the 5% level in nine of 15 cases and at the 10% level in a further two. These results are compatible with the idea that demand-driven swings of commodity prices are reflected more strongly in inflation in the economies we study.
We go on to use Hausman tests for endogeneity bias to show that IV estimates are generally larger than the OLS estimates, which suggests that fluctuation in the global demand for commodities had particularly large effects on inflation in the 15 economies under study.
Overall, we find that commodity price shocks were important determinants of the international co-movements of inflation before WWI, particularly when they were due to international demand shifts.
source : cepr.org
