Economics Letters 181 (2019) 90–94

Contents lists available at ScienceDirect

Economics Letters
journal homepage: www.elsevier.com/locate/ecolet

Inflation hedging with commodities: A wavelet analysis of seven

centuries worth of data
∗
Adam Zaremba a,c , , Zaghum Umar b , Mateusz Mikutowski a
a
Department of Investment and Capital Markets, Faculty of Management, Poznań University of Economics and Business, al. Niepodległości
10, 61-875 Poznań, Poland
b
College of Business, Zayed University, P.O. Box 144534, Abu Dhabi, United Arab Emirates
c
Dubai Business School, University of Dubai, Academic City, P.O. Box 14143, Dubai, United Arab Emirates

highlights

• We test the inflation hedging properties of commodities on the longest data series ever used.
• We apply wavelet analysis to commodity prices and UK inflation for the years 1265–2017.
• Commodities co-move with inflation in the 4- to 8-year horizon.
• The phenomenon is robust for agricultural, energy, and industrial commodities.
• The correlation was strong from the 13th to the 19th centuries and weakened in the 20th century.

article info a b s t r a c t

Article history: We perform a test of the inflation hedging properties of commodities on the longest data series ever
Received 17 March 2019 used for this purpose. We apply wavelet analysis to commodity prices and inflation data from the
Accepted 1 May 2019 United Kingdom for the years 1265 through 2017, to detect co-movement across different times and
Available online 11 May 2019
frequencies. We demonstrate robust inflation hedging properties of agricultural, energy, and industrial
JEL classification: commodities for the 4- to 8-year horizon through almost the entire seven centuries.
G11 © 2019 Elsevier B.V. All rights reserved.
G14
E44

Keywords:
Inflation hedging
Commodity markets
Early commodity prices
Wavelet analysis

1. Introduction an examination of inflation hedging properties of commodities

over the longest period ever studied, extending the earlier studies
Securing the value of your savings is one of the major goals on this topic by several centuries.
of financial investment. To date, numerous studies have demon- Using a unique dataset, we investigate the co-movements of
strated the inflation hedging properties of different asset classes, aggregate commodity prices and inflation rates in the United
particularly commodities (see Alagidede and Panagiotidis, 2012; Kingdom (UK) for the years 1265–2017. To accomplish this, in-
Arnold and Auer, 2015; Spierdijk and Umar, 2015). Notably, re- stead of relying on classic inflation hedging measures based on
cent development of databases that include early financial data correlations and regression coefficients, we apply state-of-the-art
have allowed examination of the inflation hedging properties of
wavelet analysis as in Tiwari et al. (2018), which allows us to
commodities or equities for periods spanning up to 300 years in
efficiently present the co-movements across different times and
the past (e.g., Bampinas and Panagiotidis, 2015; Kim and Ryoo,
frequencies.
2011; Tiwari et al., 2018). In this study, we increase the period
Our findings confirm strong co-movement of commodity re-
tested to more than seven centuries. In consequence, we conduct
turns, both nominal and real, and UK inflation rates. The effect
∗ Corresponding author. is robust for both an aggregate commodity basket and for dif-
ferent commodity classes: agriculturals, industrials, and energy.
E-mail addresses: [email protected] (A. Zaremba),
[email protected] (Z. Umar), [email protected] The strong inflation hedging properties are visible and significant
(M. Mikutowski). for the 4- to 8-year horizon from the 13th century to the 19th

https://doi.org/10.1016/j.econlet.2019.05.002
0165-1765/© 2019 Elsevier B.V. All rights reserved.
 A. Zaremba, Z. Umar and M. Mikutowski / Economics Letters 181 (2019) 90–94 91

Fig. 1. Wavelet power spectrum and average wavelet power for the composite commodity index. (For interpretation of the references to color in this figure legend,
the reader is referred to the web version of this article.)

century; they have somewhat faded in the 20th century, perhaps earlier periods. The prices are compiled from different sources,
due to the structural changes in the economy and the shift to such as stock exchanges and history books, including Cole (1938)
service-based models. and Clark (2004), and chain-linked to form a continuous time
The remainder of this article is organized as follows. Section 2 series.
presents the data and methods. Section 3 discusses the empirical As the measure of inflation, we use the United Kingdom Retail
findings. Finally, Section 4 concludes the study. Price Index Inflation Rate by GFD. We focus on the UK data
due to its long-run data availability—the inflation rate in the UK
2. Data and methods can be traced back as far as the 13th century. The chain-linked
price index by GFD combines several measures, including the
Our study is based on long-run commodity data sourced from Retail Price Index, the Cost of Living Index, and a compilation of
Global Financial Data.1 The sample runs from 1265 to 2017, and several historical sources, including Brown and Hopkins (1956),
– following Zaremba et al. (2019) we use annual intervals for in particular (details available from GFD). Finally, to be consistent
the analysis because monthly or daily data is generally unavail- with our focus on the UK and its inflation rate, all commodity
able prior to the 19th century. Our base measure of commodity returns are expressed in GBP.
prices is the GFD Commodity Composite Index, which tracks spot Table 1 reports the summary statistics for our research sample.
prices of about 50 commodities. The index is split into three To research the inflation hedging properties of commodities,
sub-indices covering agriculturals—beer, cocoa, coffee, milk, tea, we use wavelet multi-scale analysis (see, e.g., Torrence and
butter, cheese, eggs, potatoes, sugar, barley, corn, hay, oats, rice, Compo, 1998) which allows for a decomposition of time se-
wheat, hogs, cattle, lambs, corn oil, cottonseed oil, cottonseed ries over both time and scale. Hence, we are able to assess the
meal, flaxseed, lard, oatmeal, palm oil, soybean meal, soybean correlation levels at different moments in time and frequen-
oil, soybeans, and tallow; industrials—aluminum, copper, iron, cies. Although computationally complex, the wavelets are finely
lead, nickel, steel, tin, zinc, cotton, rubber, tobacco, wood, wool, detailed, allowing for a good understanding of the spectral char-
gold, palladium, platinum, and silver; and energy—coal, coal gas, acteristics of time series, and dealing well with the non-stationary
firewood, natural gas, lamp oil, whale oil, and petroleum. No exact behavior and non-normality of financial data (Conlon et al., 2018).
production or consumption data is available for the period before In particular, in our investigations we focus on two measures: the
the 1900s, therefore components are weighted on the amount wavelet power spectrum (WPS) and the cross-wavelet transform
of trade and production or using arbitrary weights. The indices (XWT). The first displays the occurrence of extreme events; the
are based on U.S. prices when available (usually only the 20th second measures co-movements of two series. We derive both
century), and on English price data otherwise, particularly for measures using the continuous wavelet transform, and in all our
calculations we closely replicate the implementation of Tiwari
1 https://www.globalfinancialdata.com. et al. (2018).
92 A. Zaremba, Z. Umar and M. Mikutowski / Economics Letters 181 (2019) 90–94

Fig. 2. Cross-wavelet transform and average wavelet power for the composite commodity index. (For interpretation of the references to color in this figure legend,
the reader is referred to the web version of this article.)

Table 1
Descriptive statistics of inflation rates and commodity index returns.
GFD composite GFD energy GFD agricultural GFD industrial United Kingdom
commodity index commodity index commodity index commodity index retail price index
inflation rate
Average 0.0146 0.0221 0.0163 0.0133 0.0145
Standard deviation 0.1066 0.1653 0.1265 0.1209 0.1050
Skewness 1.0981 2.2585 0.7374 1.1998 0.3836
Kurtosis 5.9295 14.1202 3.7605 4.9268 2.9641
Minimum −0.4594 −0.5008 −0.4769 −0.4238 −0.4377
1st quartile −0.0471 −0.0614 −0.0617 −0.0518 −0.0419
Median 0.0069 0.0034 0.0116 0.0019 0.0106
3rd quartile 0.0596 0.0794 0.0766 0.0612 0.0646
Maximum 0.7439 1.5701 0.8032 0.6565 0.6363
No. of observations 752 752 752 752 752

Note. The table reports the descriptive statistics for the yearly changes in commodity and UK retail price indices.

3. Results low (high) frequencies, whereas areas of blue color indicate a lack
of significant scenarios.
Figs. 1–3 summarize our findings, exhibiting the WPS along Let us begin with an overview of the results for the aggregate
with its average power spectrum and the XWT along with its commodity portfolio represented by the GFD Composite Com-
average power spectrum. The outcomes are displayed in con- modity Index. Fig. 1 presents the WPS for nominal commodity
tour plots with three dimensions: time spaces (horizontal axis), returns (Panel A), real commodity returns (Panel B), and inflation
frequency (vertical axis), and the wavelet spectrum values (col- (Panel C), as well as the average wavelet power (Panel D) giving
ored spectrum of rising intensity from dark blue to red). The the average variance contained in all the wavelet coefficients at a
frequency is expressed in yearly units, ranging from the highest given frequency. Although the WPS is not an explicit proof of co-
frequency of two years (bottom of the plot) to the lowest fre- movements or lead–lag relationships between commodity prices
quency of 256 years (top of the plot). A significant red color at and inflation, it helps to identify some similarities in the evolution
the top (bottom) suggests the occurrence of extreme events at of data across the time scale.
 A. Zaremba, Z. Umar and M. Mikutowski / Economics Letters 181 (2019) 90–94 93

Fig. 3. Cross-wavelet transform and average wavelet power for inflation and nominal returns on the agriculture, industrial, and energy indices. (For interpretation
of the references to color in this figure legend, the reader is referred to the web version of this article.)

Importantly, the nominal and real commodity returns, as well run. The vivid red color for the 2- to 8-year frequencies indi-
as inflation, display elevated wavelet power levels for the 2- to cates a reliable relationship between the two series, and the
8-year scale—represented by the red contour. This might be inter- average power plot corroborates its significance at the 5% level
preted as a first sign of interdependencies. The average wavelet for the 4- to 8-year scale. Again, as in Fig. 1, the similarity in
the co-movement is solid from the 13th century through the
power plot also confirms the significant common years scale
19th century, but markedly weaker similarity during the last
of 2–8 between nominal commodity returns and inflation. The
century. This phenomenon could be interpreted as the dimin-
similarities between inflation and commodity behavior are clearly
ishing role of commodity prices in the modern service-based
visible for the 13th through 19th centuries; they are noticeably economy. Notably, the relationship with inflation is similarly
weaker in the 20th century. strong for both nominal and real composite commodity returns.
The findings reported in Fig. 2 formalize our intuition on co- Nonetheless, there is a considerable difference in the direction
movements of inflation and commodity returns over the long of the commodity–inflation relationship when the nominal and
94 A. Zaremba, Z. Umar and M. Mikutowski / Economics Letters 181 (2019) 90–94

real returns are considered. Information in this regard is given 750 years. The co-movement is strong at the aggregate level
by the arrows in the plot. The phase arrows facing west (east) and also for different commodity classes – energy, industrials,
indicate out-of- (in-) phase co-movement, and the south- (north-) or agriculturals – but it visibly weakened in the 20th century.
facing arrows represent the situation where the commodities lead Future studies on inflation hedging using early commodities data
(lag) inflation. For nominal commodity returns, the arrows pre- could extend our research to different countries, currencies, and
dominantly point to the east, suggesting in-phase co-movements: the role of individual commodities.
high commodity returns are accompanied by high inflation rates.
For the real commodity returns, however, the relationship is the References
inverse, with predominantly west-facing arrows indicating an
out-of-phase relationship. Alagidede, P., Panagiotidis, T., 2012. Stock returns and inflation: Evidence from
quantile regression. Econom. Lett. 117 (1), 283–286.
Finally, Fig. 3 extends our aggregate commodity-level investi-
Arnold, S., Auer, B.R., 2015. What do scientists know about inflation
gations to the three sub-indices: agriculture, industrial, and en- hedging? North Am. J. Econ. Financ. 34, 187–214.
ergy. For brevity, we focus solely on the cross-wavelet analysis of Bampinas, G., Panagiotidis, T., 2015. Are gold and silver a hedge against inflation?
the nominal price changes. Remarkably, the relationship between A two-century perspective. Int. Rev. Financ. Anal. 41, 267–276.
inflation and commodity returns is strong for all three commodity Brown, E.H.P., Hopkins, S.V., 1956. Seven centuries of the price of consumables,
compared with builders’ wage-rates. Economica 23 (92), 296–314.
classes. The red color suggests a clear link between commodity Clark, G., 2004. The price history of english agriculture. Res. Econ. Hist. 22,
prices and inflation for the 2- to 8-year frequency for almost the 1209–1913, 41–123.
entire study period, and the east-facing arrows indicate in-phase Cole, A.H., 1938. Wholesale Commodity Prices in the United States. Harvard
co-movements. This relationship, however, weakens in the 20th University Press, Cambridge, pp. 1700–1861.
Conlon, T., Lucey, B.M., Gazi, U.S., 2018. Is gold a hedge against inflation? A
century, similar to the results in Figs. 1 and 2, pointing to the pos-
wavelet time-scale perspective. Rev. Quant. Financ. Account. 51 (2), 317–345.
sible lesser importance of commodity prices in the contemporary Kim, J.H., Ryoo, H.H., 2011. Common stocks as a hedge against inflation: Evidence
service-based economy. The formal analysis of the average cross- from century-long US data. Econom. Lett. 113 (2), 168–171.
wavelet plots (Panels B, D, and E) confirms the statistical validity Spierdijk, L., Umar, Z., 2015. Stocks, bonds, T-bills, and inflation hedging: From
of our findings. Nonetheless, the effect appears to be the most great moderation to great recession. J. Econ. Bus. 79, 1–37.
Tiwari, A.K., Cunado, J., Gupta, R., Wohar, M.E., 2018. Are stock returns and
robust for agricultural commodities, which played a crucial role inflation hedge for the UK? Evidence from a wavelet analysis using over
in the pre-industrial era that constitutes the majority of our time three centuries of data. Stud. Nonlinear Dyn. Econom. Forthcoming.
sample. Torrence, C., Compo, G.P., 1998. A practical guide to wavelet analysis. Bull. Am.
Meteorol. Soc. 79, 605–618.
Zaremba, A., Bianchi, R.J., Mikutowski, M., 2019. Long-run reversal in commodity
4. Concluding remarks
returns: Insights from seven centuries of evidence. Available at SSRN: http:
//dx.doi.org/102139/ssrn.3314834.
In this study, we have demonstrated that the inflation hedging
properties of commodities in the UK have been evident for almost