Content uploaded by William Ginn
Author content
All content in this area was uploaded by William Ginn on Jun 15, 2024
Content may be subject to copyright.
Commodity Prices Exposure to Changing ENSO
Patterns ∗
Gilles Dufr´enot†William Ginn‡Marc Pourroy§
June 15, 2024
Abstract
El Ni˜no–Southern Oscillation (ENSO) is a major climate phenomenon that influences tem-
perature and precipitation across the globe. We study the effect of changing ENSO patterns
on commodity prices using a Global Factor Local Projections (GFALP) model. Firstly, we
demonstrate that unanticipated ENSO movements contribute to commodity price volatility
asymmetrically during El Ni˜no and La Ni˜na periods. Secondly, climate change might disrupt
ENSO patterns. We compare the current situation with potential climate change outcomes to
evaluate its impact on commodity price stability. We compute an index measuring commodity
price exposure to these disruptions. We demonstrate that in most cases, these shifts exacer-
bate commodity price volatility. Finally, we explore several avenues to explain the observed
heterogeneity in the exposure of commodity prices to the evolution of ENSO that could result
from climate change, and we highlight the crucial role of international commodity markets in
adapting to climate change.
JEL Classification: C32, G13, Q54, C50.
Keywords: ENSO, Commodity Price, Climate Change, Agriculture, Energy.
∗This work was supported by the French National Research Agency Grant ANR-17-EURE-0020, and by the
Excellence Initiative of Aix-Marseille University - A*MIDEX. We thank the participants of the CEMA Chicago
conference, IMAC Rennes conference, LEMNA Seminar and L´eP Seminar for fruitful comments and suggestions.
†Aix-Marseille Univ, CNRS, AMSE, Marseille, France, Gilles.Dufrenot@univ-amu.fr
‡Labcorp, Sr. Economist / Data Scientist, USA and Coburg University of Applied Sciences, Germany.
William.Ginn.OBA@said.oxford.edu
§Correspondence author, University of Poitiers, France. Marc.Pourroy@univ- poitiers.fr
1
1 Introduction
Numerous recent studies have highlighted the impact of weather conditions on economic activity.
These studies help us to better understand how economic activity depends on climatic conditions,
and how weather shocks can impact on the economy (e.g., Carleton and Hsiang,2016,Henseler
and Schumacher,2019,Acevedo et al.,2020,De Winne and Peersman,2021,Tol,2024). A central
question in this literature is how ongoing climate change will affect economies, so that society can
adapt institutions and policymakers make informed decisions leading to sound economic policies.
In this paper, we focus on a specific channel that links weather conditions to economic activity:
the commodity price channel. More specifically, we look at how a key global climate phenomenon,
named El Ni˜no–Southern Oscillation (ENSO), affects commodity prices, and how this relation may
evolve due to climate change.
ENSO is a climate phenomenon characterized by periodic fluctuations in sea surface tempera-
tures and atmospheric pressure in the equatorial Pacific Ocean. It involves interactions between
the ocean and the atmosphere and has significant impacts on weather patterns around the world.
ENSO is characterized by the periodic alternation between El Ni˜no conditions (warmer than aver-
age sea surface temperatures) and La Ni˜na conditions (cooler than average sea surface temperatures
in the central and eastern Pacific Ocean). ENSO events typically occur every two to seven years
and can lead to various climate anomalies, including changes in rainfall patterns, temperatures,
and storm activity, affecting ecosystems, agriculture, and economies globally.
Existing literature has identified numerous ways ENSO conditions influence economic activity.
The most obvious is the impact of weather anomalies on the supply of agricultural commodities
(Legler et al.,1999). However, the effects of ENSO extend far beyond agriculture. For instance,
during the El Ni˜no phase, the lack of upwelling of colder, nutrient-rich water near the South
American Pacific coast leads to a decline in phytoplankton populations, resulting in reduced fish
catches (Bertrand et al.,2020). Additionally, metals and minerals can be affected, particularly due
to excessive rainfall causing flooding in mines (Vink and Robbins,2012). This illustrates how all
commodities are impacted by the ENSO phenomenon. Beyond commodities, it should be noted
that this wide range of ENSO conditions’ sectoral effects logically manifests at the macroeconomic
level (Berry and Okulicz-Kozaryn,2008 ,Cashin et al.,2017,Liu et al.,2023).
The examples provided above are just a few of the numerous transmission channels between
ENSO conditions and economic activities explored in the literature. These channels can sometimes
appear contradictory. For instance, the impact of El Ni˜no on coffee varies between arabica and
robusta varieties (Ubilava,2012). Another complexity arises from the asymmetry between El
Ni˜no and La Ni˜na phenomena: meteorologically, La Ni˜na is not simply the opposite of El Ni˜no,
but rather an intensification of “normal” conditions. A brief La Ni˜na period is not necessarily
followed by a similarly brief El Ni˜no period, as there is no predictable cyclic pattern (An and Jin,
2004,An and Jin,2004,Guo et al.,2017). The duration and scale of El Ni˜no events influence
their effects on economic activity, introducing numerous non-linearities (see Smith and Ubilava,
2017 and Generoso et al.,2020 on regime-dependent nonlinearity in the growth response to ENSO
1
shocks). Additionally, many researchers believe that climate change will alter the rate at which
El Ni˜no and La Ni˜na phases reverse and increase their intensity, although these projections are
highly uncertain (Cai et al.,2014,Yeh et al.,2018,Hu et al.,2021,Cai et al.,2021). Ultimately,
the significance of this phenomenon—both meteorologically and economically—combined with the
complexity of its analysis, makes it a critical subject of study.
In this paper, we address two questions: How does climate affect commodity prices through
ENSO conditions? And how might this relationship evolve due to climate change?
First, we introduce a method to measure ENSO unanticipated changes impact on commodities,
while controlling for global factors. As the impact of El Ni˜no or La Ni˜na cannot be reduced to
one country, but rather has global dimensions, this paper employs a global factor augmented local
projections model (GFALP) framework to assess the transmission of weather shocks on commod-
ity, while controlling for global output and financial condition, using monthly data over the period
1986 01 to 2023 06. Using monthly data has the potential benefit of capturing the short-term tem-
poral effect that weather has on the economy (e.g., Barnston,2015), as opposed to using quarterly
data as in previous studies (Brunner,2002,Cashin et al.,2017). To capture the non-linearities
described above and following Auerbach and Gorodnichenko (2012) and Ventosa-Santaul`aria et al.
(2024), we estimate a non-linear local projection model (NLLP, as opposed to Berry and Okulicz-
Kozaryn (2008)orAnttila-Hughes et al.,2021 for example of linear framework). Our analysis
demonstrates that aggregating commodities into broad categories (agriculture, energy, metals)
significantly reduces the observed impact of ENSO on prices compared to examining individual
commodities. This difference is due to the diverse reactions of individual commodities. Addi-
tionally, we show that the impact of shocks varies greatly between El Ni˜no and La Ni˜na periods,
supporting the use of a non-linear model.
Second, we propose an original method to capture the effect of climate change through two
parameters: the frequency and intensity of El Ni˜no and La Ni˜na events. We estimate these two
parameters using historical data and then simulate the effects of shocks by calibrating values to
reflect the assumed impacts of climate change. A key contribution of our paper is to propose three
original statistical criteria to estimate these parameters. This approach allows us to observe how
climate change influences the volatility of commodities through its effects on ENSO. From these
simulations, we develop an index that measures each commodity’s price exposure to changes in
ENSO. This index helps us determine whether climate change will increase the vulnerability of
each of the 67 commodities studied to ENSO events or leave it unaffected.
Our findings reveal considerable variation among commodities in their responses to climate
change. Some commodities show minimal or no impact, with a few expected to have even more
stable prices in the future. In contrast, others are projected to face significantly increased volatil-
ity, as indicated by our index. To explain this disparity, we conducted several tests and found
that factors such as financialization, production concentration, and the limited proportion of a
commodity sold on the global market tend to amplify volatility.
This research overlaps with two broad strands of literature. The first strand relates to a volumi-
2
nous body of literature analyzing the effects of ENSO on activity, and particularly on commodities.
Previous studies generally focus on one commodity, such as Tack and Ubilava (2015) about cot-
ton, or a given country, such as Ubilava (2012) and Melo-Velandia et al. (2022) about Colombia,
Mueller and Osgood (2009) about Brazil, Mainardi (2011) about Burkina Faso and Niger or Li
et al. (2019) about China. We add to the literature by showing that the effects of ENSO operate
also at the global level. Also, our work is close to Ubilava (2018) who finds an effect of ENSO on a
large number of commodities, particularly agricultural commodities. However Ubilava (2018) does
not address the issue of climate change. Conversely, Liu et al. (2023) find a damaging impact from
an El Ni˜no on global production and they show how climate warming will exacerbated economic
damage from changing ENSO. In a close work, Callahan and Mankin (2023) use a cross-country
model with two separated measures to capture El Ni˜no and La Ni˜na (and without controlling for
the global economy). They show that El Ni˜no persistently reduces country-level economic growth.
Hence, Liu et al. (2023) and Callahan and Mankin (2023) take into account the effect of climate
change, but looks at GDP and not commodities. Therefore, our position in the literature is clear:
to our knowledge, this paper is the first to propose a method for projecting the effect of climate
change on the impact of ENSO on a large number of commodities.1
The second strand relates to a growing literature analyzing the impact of climate shocks on
financial stability (see Buhr et al.,2018,Giuzio et al.,2019,Fabris,2020,Stan et al.,2021,Strabel
and Wurgler,2021) such as Flori et al. (2021) who concludes that “climate conditions affect financial
stability by impacting commoditiy comovements”. Among all these papers, our work is connected
to a small body of literature that highlights the financial impact of ENSO. In particular, Damette
et al. (2024) find a significant positive impact of ENSO on sovereign risk in Latin America, and
De Marco et al. (2023) investigate show that ENSO affects the banking system in the US through
lower house prices and mortgage lending during El Ni˜no phase. We focus on the impact of ENSO on
the commodities market, and show in particular that the financialization of this market exacerbates
volatility after ENSO shocks. Because of the importance of climate shocks for financial stability,
there is increased interest from government and central banks to incorporate climatic risk into
adaptation and resilience management (see Pointner and Ritzberger-Gr¨unwald (2019), Battiston
et al. (2021), Svartzman et al. (2021)). In addition, a few papers have pointed out that central banks
1Note that Liu et al. (2023) and Callahan and Mankin (2023) use projections for future climate based on Shared
Socioeconomic Pathways (SSP) projections for 2000–2099. Liu et al. (2023) refers to Ni˜no 3. In addition to Ni˜no 3,
Callahan and Mankin (2023) refers to SST in 5 localisations, aggregated into an E-Index. In this paper, we do not
rely on SSP projections because our ENSO measurement, MEI.V2, incorporates multiple variables, with SST being
only one among them (and, to our knowledge, only SST is projected in the SSP). Our approach is similar to that of
Mourtzinis et al. (2016), who examines the effect of climate change on corn yields via the amplification of the ENSO
phenomenon without using SSP scenarios, instead focusing on temperature changes. We concentrate on the impact
of two parameters: the intensity and frequency of the ENSO phenomenon. Given the uncertainties surrounding
the effects of climate change on all aspects of ENSO, focusing on these two parameters seems prudent. While this
means we cannot compare different CO2 emission scenarios on commodity volatility, that is not our objective. Our
goal is to identify which commodities experience increased volatility due to climate change, reflected by these two
parameters, and to understand why these particular commodities are affected more than others.
3
mandate for price stability is also threatened by weather regimes and climate change (Mukherjee
and Ouattara,2021,Kabundi et al.,2022,Boneva et al.,2022,Cevik and Jalles,2023,Ventosa-
Santaul`aria et al.,2024). By showing that climate change weighs on commodity prices stability,
we contribute to this literature, which defines the channels through which climate change can
jeopardize price stability.2
The rest of the paper is structured as follows: in Section 2describes the data and discusses
the global dimensions of the business cycle considered in the paper. The empirical strategy is
developed in Section 3. Empirical results are summarized in Sections 4and 5. Section 6concludes
the paper.
2 Data
Three types of global variables are considered: a weather indices (MEI.v2); global factors (output
and interest rate); and 67 commodity prices. Each are discussed in turn.
2.1 Global Weather Patterns
ENSO is one of the most important climate indicators, which has a major influence of global
weather conditions (e.g., Ropelewski and Halpert,1987,Rosenzweig et al.,2001,McPhaden et al.,
2006,Dai,2013 and Br¨onnimann et al.,2007).3When a major El Ni˜no (La Ni˜na) occurs, there is an
anomalous loss (increase) of heat from the ocean to atmosphere so that global mean temperatures
rise (fall) (McPhaden et al.,2020). The anomalous atmospheric patterns are known as the Southern
Oscillation, as ENSO relates to cyclical, environmental conditions that occur across the equatorial
Pacific Ocean. Changes to ENSO are due to natural interactions between sea surface temperature,
rainfall, air pressure, atmospheric and oceanic circulation. The effects of ENSO, commonly called
”teleconnections”, emphasize that changing conditions can have a profound effect on global climate,
which can in turn directly affect people’s livelihoods (e.g., Barlow et al.,2001,Diaz et al.,2001,
and Alexander et al.,2002).
Various series have been used in the literature to capture ENSO phenomena.
A first one is the Sea Surface Temperature (SST) anomalies (Hansen et al.,1998,Brunner,
2002,Ubilava,2018,Atems and Sardar,2021). SST indices are measures based on the average
sea-surface temperature over a fixed area in the tropical Pacific. They look particularly relevant
for annual data analysis.
Another measure is the Oceanic Ni˜no Index (ONI) (Sarachik and Cane,2010,Hsiang et al.,
2011,Generoso et al.,2020)). ONI is a 3 month rolling index tracking the ocean part of ENSO.
2Note that some papers find that temperature shocks lead to inflationary pressures, such as Mukherjee and
Ouattara (2021) for developing countries, while other papers seems to indicates the opposite, such as Cevik and
Jalles (2023) who find that following a temperature shock, headline inflation falls. See Kranz et al. (2024) for a
literature review.
3The NOAA considers ENSO as ”one of the most important climatic phenomena on Earth”, see https://www.
weather.gov/mhx/ensowhat.
4
Finally, some indices have been created to integer many dimension of ENSO: SST, winds,
etc. One of the most used in the literature is MEI.v2, published by the National Oceanic and
Atmospheric Administration (NOAA) and available from 1979. MEI.v2 uses 5 variables variables :
sea level pressure (SLP), sea surface temperature (SST), surface zonal winds (U), surface meridional
winds (V), and Outgoing Longwave Radiation (OLR). 4It is a bi-monthly index is calculated for 12
overlapping bi-monthly “seasons” (Jan-Feb, Feb-Mar,...). Therefore, it takes into account ENSO
seasonality and reduce effects of higher frequency intra-seasonal variability (see De Marco et al.
(2023)).
Composite positive MEI events can be read as warm period or El Ni˜no events.5Negative MEI
events have key features of mostly opposite phase and can similarly be read as cold periods or La
Ni˜na evants. NOOA generally applies a +/−0.5 threeshold to define non-overlapping hot and cold
periods, the in-between been neither El Ni˜no nor LN. Figure 1displays the evolution of MEI.v2
since 1979. Top warm El Ni˜no events can be seen in 1983, 1987, 1992, 1998 and 2016. Similarly,
top cold La Ni˜na events can be seen in 1989, 1996, 1999, 2008 and 2011.
-2 -1 0 1 2 3
ENSO (MEI.v2)
1980 1990 2000 2010 2020
Time
El Niño La Niña
Areas shaded blue indicate negative values of the MEI.v2 that represent the cold ENSO
phase, a.k.a. La Ni˜na, while areas shaded red indicate positive MEI.v2 that values represent
the warm ENSO phase, a.k.a. El Ni˜no.
Figure 1: MEI.v2 Evolution over Time
4See https://psl.noaa.gov/enso/mei/
5As defined by NOAA : ”Key features of composite positive MEI events (warm, El Ni˜no) include (1) anomalously
warm SSTs across the east-central equatorial Pacific, (2) anomalously high SLP over Indonesia and the western
tropical Pacific and low SLP over the eastern tropical Pacific, (3) reduction or reversal of tropical Pacific easterly
winds (trade winds), (4) suppressed tropical convection (positive OLR) over Indonesia and Western Pacific and
enhanced convection (negative OLR) over the central Pacific.”
5
2.2 Commodities
A total of 67 international commodities are analyzed in this study, taken from the World Bank
“pinksheet” monthly data. The sample period covers 1986:01 to 2023:06. Figure 2presents the
composition of the data set according to the type of commodities. Apart from a large group of food
and beverage commodities, the data set is well balanced over the 6 dimensions: energy, fertilizers,
metals end minerals, precious metals. To make the results easier to read, we sometimes reason at
an aggregate level, using the World Bank’s 3 categories: agriculture, energy and metals/minerals
prices. All prices are converted to logarithm. The commodity price data, denominated in U.S.
dollars, is deflated by dividing by Producer Prices Index (OECD Total Area).
3
7
5
10
33
9
0 10 20 30 40
Number of commodities
Precious Metals
Metals and Minerals
Fertilizers
Energy
Agriculture Food and Beverage
Agricultural Raw Materials
Figure 2: Composition of the commodities data set
2.3 Control variables
The global dimension of our macro variables is obtained by considering the first principal compo-
nent of nine economies covering output and the interest rate. Following the approach by Ratti and
Vespignani (2016), we construct a global factor using the principal component indices for output
and interest rate using normalized loadings.6The benefit of this approach is that by taking the
first principal component establishes a dimension reduction techniques that can replicate the main
features of a global environment. The global factors represent nine economies which approximate
6Output is proxied by OCED and Fred industrial production data. For India, manufacturing production index
(FRED mnemonic INDPRMNTO01IXOBM) is used as opposed to total production index (FRED mnemonic IND-
PROINDMISMEI), considering data availability (the correlation between the is 0.9918 for Jan 2000 to Dec 2018).
For China, we use total production excluding construction (FRED mnemonic CHNPRINTO01IXPYM). As the pro-
duction index for China includes missing values, the Kalman smoother using an ARIMA state space representation
is used to impute missing values. For India, the interest rate is based on the 90 day Treasury Bill interest rate (e.g.,
Patnaik et al.,2011,Gabriel et al.,2012,Saxegaard et al.,2010,Anand et al.,2014 and Ginn and Pourroy,2022).
6
two-thirds of global output.7
YG
t= [YCAN
t, Y C HN
t, Y E U R
t, Y GB R
t, Y I N D
t, Y J P N
t, Y K OR
t, Y RU S
t, Y U S A
t] (1)
RG
t= [RCAN
t, RCH N
t, REU R
t, RGBR
t, RIN D
t, RJP N
t, RKOR
t, RRU S
t, RU SA
t] (2)
We use one factor (the principal component) for the global variables (Ratti and Vespignani,
2016). The results are provided in Table 1, which shows the top three principal components of each
global variable for the nine economies. The first principle component captures significant share of
the variance relating to output (54.9%) and the interest rate (54.1%).8
Figure 3plots the global factors along with the economy data. The top-pane shows a sizable
decline in output which occured during the global financial crisis.9
Global Output Global Interest Rate
First Principal Component 54.9% 54.1%
Second Principal Component 33.5% 18.3%
Third Principal Component 6.4% 12.3%
Table 1: Variation Explained by First Three Principal Components
The correlation between global variables (output and interest rate) is provided in Table 2.
The correlation between global factor and country output is quite high for CAN, RUS and USA;
and somewhat moderate for EUR and IND. There is lower correlation between the China and
global output. The negative correlation between the UK and global output may be due to higher
uncertainty in the UK related with Brexit. While it remains unclear to the extent that Brexit has
had an impact on the domestic economy, a common thread is uncertainty, which has been linked
with reduced investment, employment and productivity growth (Bloom et al.,2018).
7The nine economies include: Canada (”CAN”), China (”CHN”), Euro zone (19 countries; ”EUR”), United
Kingdom (”GBR”), India (”IND”), Japan (”JPN”), South Korea (”KOR”), Russia (”RUS”) and the United States
(”USA”). Based on IMF data in purchasing power parity terms, the nine economies considered in this paper represent
66.1% of global output, see https://www.imf.org/external/datamapper/PPPSH@WEO/OEMDC/ADVEC/WEOWORLD/EU.
The Euro zone values are based on the 19 member countries (i.e., Austria, Belgium, Cyprus, Estonia, Finland,
France, Germany, Greece, Ireland, Italy, Latvia, Lithuania, Luxembourg, Malta, the Netherlands, Portugal, Slo-
vakia, Slovenia, and Spain).
8The higher dimensions of the principal components are provided in the Appendix, see the Scree plot in Figure
20. These values are similar to Ratti and Vespignani (2016), where the first principle component in their paper
captures for global output and interest rate represents 60.0% and 44.5% of the total variance, respectively.
9According to the NBER, the recession dates for the U.S. is between 2007:DEC to 2009:JUN.
7
Figure 3: Global Factors
Country GLO CAN CHN EUR GBR IND JPN KOR RUS USA
Global Output 1.00 0.63 -0.81 0.51 -0.46 0.92 -0.1 0.91 0.98 0.81
Global Interest Rate 1.00 0.92 0.56 0.95 0.94 0.01 0.61 0.91 -0.25 0.81
Table 2: Correlation by Country and Global Variable
8
3 Empirical strategy
3.1 Baseline Model
The linear local projections model with transition function, developed by Jord`a (2005), is employed
to estimate the dynamic responses that changing ENSO patterns have on commodity prices.10 In
the benchmark specification, we estimate commodity inflation in real terms (πt) as follows:
πt+h= + (1 −F(ζt−1))(αh,EN +ϕh,E N (L)xt−1+βh,EN ut)
+F(ζt−1)(αh,LN +ϕh,LN (L)xt−1+βh,LN ut)
+ϵt+h
(3)
which accounts for an asymmetry, defined as an El Ni˜no (EN ) and La Ni˜na (LN) climate state,
with πt+h=pt+h−pt−1where pt+his 1 commodity price (in log form) among 67 indices which is
projected on the space generated by a set of control variables (xt−1). The vector of control variables
includes lags of the respective commodity price inflation (πt), global output growth, global interest
rate and control variables for the global financial crisis and the Covid-19 lock-down period. In this
specification, we allow the prediction of πt+hto differ according to the state of the climate (i.e., in
an El Ni˜no and La Ni˜na state) when a weather shock (ut) occurs. The coefficient βh,E N (βh,LN )
corresponds with the estimated impact of the weather shock in a El Ni˜no (La Ni˜na) state.
The variable utis the surprise components of MEI.V2 obtained as follows :
yt=α+βyt−h+γTt+ut
with ytour ENSO measure (MEI.v2), considered with h= 1, ..., 6 lags, Ttis a monthly control
variable that captures seasonality, and utis the residual.
F(ζt) is a smooth transition function that represents the state of the climate:
F(ζt) = exp(−γζt)
1 + exp(−γζt)(4)
where γ>0 controls the degree of smoothness of the transition between states and |ζt|<∞is a
standardized transition variable.
The transition variable is taken as MEI.v2. As opposed to Gorodnichenko and Auerbach (2013)
and Ramey and Zubairy (2018), the transition variable is not standardized by taking the cyclical
component using the Hodrick and Prescott filter, as MEI.v2 is already centered and cyclical.
Consistent with Auerbach and Gorodnichenko (2012), the transition function is dated t−1 in
Equation (3) to avoid contemporaneous feedback from policy actions with regard to the state of
the economy (i.e., F(ζt−1)).
In equation (4), the parameter γdenotes the degree of smoothness of the transition between
the two state, the larger γthe faster the regime change from El Ni˜no to La Ni˜na and vice-versa.
10Following Auerbach and Gorodnichenko,2012, we develop here a non-linear version of the local projection
model, sometimes refereed as NLLP (Ventosa-Santaul`aria et al.,2024.
9
In Appendix 8.3 we illustrate how does this function work: we simulate different values for ζ, and
then observe the values taken by F(ζ) assuming alternative values for γ.
We propose a methodology to estimate γvalue in order to obtain a transition function the
closest to the ”true” transition process, which can be proxied by ESOI CDF function.11
Assuming a random evaluation grid: X=−1, ..., x, ...1 , the cumulative distribution function
(CDF) of ζis given by Fζ(x) = P(ζ≤x) which is the probability that ζtakes on a value less than
or equal to x. Empirically, we compute for any xthe proportion:
Fζ(x) = 1
n
n
X
i=1
1[ζi<x](5)
−1.0 −0.5 0.0 0.5 1.0
0.0 0.2 0.4 0.6 0.8 1.0
Evaluation Grid (x)
Enso CDF
Figure 4: ESOI Empirical CDF
The Empirical CDF, Fζ(x) for ESOI, is represented on Figure 4. The calibration of γmust
ensure that the transition function F(x|γ), given in Equation 4, is as close as possible to the em-
pirical CDF, given in equation 5. We suggest three alternative criteria to obtain such an optimal
estimated ˆ
γ, based on Total Sum of Squares, the Kolmogorov–Smirnov statistic and the Dvoret-
zky–Kiefer–Wolfowitz Confidence Interval .
(1) Our first criterion to estimate γis based on the Sum of Squares and consist in minimizing
the mean distance between the Empirical CDF and the transition function. We minimize the
mean square deviation between the Empirical CDF and the transition function. The result is
the transition function that is, on average, closest to the empiric CDF, for all x considered. The
criterion is :
min
γ
Sq =
n
X
1=1
[F(xi|γ)− Fζ(xi)]2
(2) Our second criterion consist in minimizing the Kolmogorov–Smirnov statistic applied to
F() and F, where sup is the supremum function. The focus not on the average deviation but on
11In the literature, this parameter is generally calibrated (see Auerbach and Gorodnichenko (2012) among other).
10
largest one (as we concentrate on the xithat gives the biggest gap between the Empirical CDF
and the transition function). The criterion is :
min
γ
KS = sup
x
|F(ζt|γ)− Fζ(x)|
(3) Our last criteria is based on Dvoretzky–Kiefer–Wolfowitz Confidence Interval with ϵ=
qln 2
α
2nand α∈[0,1] a parameter such that the larger α, the tighter the confidence interval that
contains F&F. In other words, we look for γvalue that gives the tightest confidence interval
around Fζ(x) that contains F(ζt|γ). The criterion is defined as
max
γ,α H(α) = [1]Fζ(x)−ϵ(α)≤F(ζt|γ)≤ Fζ(x) + ϵ(α)
Figure 5displays γestimation outputs. On the left column, we observe the value taken by
the TSS, KS and DKW criteria as a function of γ. The three estimation crieterion give very
similar results: the value of γthat minimizes the Total Sum of Squares is 4.328; similarly the
Kolmogorov–Smirnov statistic is minimal for ˆ
γ= 4.224 ; and the Dvoretzky Kiefer Wolfowitz
Inequality’s alpha is maximized for ˆ
γ= 4.276. Figure 5right column displays ESOI empical CDF
and the transition function obtained with the optimal γ. Whatever the criteria used to estimate
γ, the shape of the transition functions is fairly closed to the empirical CDF.
3.2 Changing ENSO patterns: extreme conditions
Climate change is expected to influence the ENSO. While the precise details are still an area of
ongoing research, there is evidence suggesting that climate change could alter the frequency and
intensity of El Ni˜no and La Ni˜na events (Cai et al.,2014,Yeh et al.,2018,Hu et al.,2021,Cai
et al.,2021). Some models suggest an increase in the frequency of extreme El Ni˜no events, which
could have significant implications for global weather patterns (e.g., Timmermann et al.,1999,
Chen et al.,2001,An and Wang,2000). Accordingly, we investigate anomalous ENSO conditions.
Following NOAA among other, we use two join criteria to define anomalies: amplitude and
duration of the variation. Therefore, we define anomalies as periods where MEI.V2 absolute value
is above .8 and when this threshold is met for a minimum of 5 consecutive overlapping seasons.
Anomalies periods are shown in dark on Figure 6.
To estimate whether an anomalies weather conditions matter in an El Ni˜no state, Equation 3
is extended to include a latent variable and interaction terms:
πt+h= + (1 −IEN AF(ζt−1))(αh,EN A+ϕh,E N A(L)xt−1+βh,EN Aut)
+IEN AF(ζt−1)(αh+ϕh(L)xt−1+βhut)
+ϵt+h
(6)
where IEN Aequals 1 if MEI.v2 > .8 for at least 5 months in a row, 0 otherwise.
11
Evaluation Grid Fζ(x) and F(x|ˆ
γ)
Total Sum of Squares : ˆ
γ= 4.328
0 5 10 15 20 25 30
0 10 20 30 40
Gamma
Total sum of squares
−1.0 −0.5 0.0 0.5 1.0
0.0 0.2 0.4 0.6 0.8 1.0
Evaluation points
Cumulative distribution fonction
ENSO CDF
Logistic
Kolmogorov–Smirnov statistic : ˆ
γ= 4.224
0 5 10 15 20 25 30
0.1 0.2 0.3 0.4 0.5 0.6
Gamma
Kolmogorov–Smirnov statistic
−1.0 −0.5 0.0 0.5 1.0
0.0 0.2 0.4 0.6 0.8 1.0
Evaluation points
Cumulative distribution fonction
ENSO CDF
Logistic
Dvoretzky Kiefer Wolfowitz Inequality : ˆ
γ= 4.276
0 5 10 15
0.0 0.2 0.4 0.6 0.8
Gamma
DKW alpha
−1.0 −0.5 0.0 0.5 1.0
0.0 0.2 0.4 0.6 0.8 1.0
Evaluation points
Cumulative distribution fonction
ENSO CDF
Logistic
Figure 5: γestimation
Similarly we estimate the same equation for La Ni˜na anomalies :
πt+h= + ILNAF(ζt−1)(αh,LN A+ϕh,LN A(L)xt−1+βh,LN Aut)
+ (1 −ILNAF(ζt−1))(αh+ϕh(L)xt−1+βhut)
+ϵt+h
(7)
12
-2 -1 0 1 2 3
ENSO (MEI.v2)
1980 1990 2000 2010 2020
Time
El Niño La Niña
El Niño anomalies La Niña anomalies
Areas shaded dark blue indicate negative values of MEI.v2 below -.8 for more than 5 periods
that represent La Ni˜na anomalies, while areas shaded dark red indicate MEI.v2 values above
.8 that represent El Ni˜no anomalies.
Figure 6: MEI.v2 Anomalies
where ILNAequals 1 if MEI.v2 <−.8 for at least 5 months in a row, 0 otherwise.
3.3 Changing ENSO patterns: faster transition speeds
Climate change could impact ENSO cycle by shortening transition times from one phase to another
(transitions from El Ni˜no to La Ni˜na, and vice versa). How can we account for this change in our
set-up? Faster transitions between El Ni˜no and La Ni˜na, and vice versa, should lead to an increase
in parameter γ. To test this hypothesis, we estimate the value of γby splitting the MEI.v2 series
into two sub-samples: 1979-1999 and 2000-2022. Results are displayed on Figure 7. It appears
that γis already changing over time.
In order to identify whether this evolution of γparameter will have an impact on the influence
of ENSO shocks on commodities, we estimate Equation 3taking into account higher value of γ.
The baseline value is γ= 4. For the robustness section, we use a calibrated parameter, denoted
by ¯
γ, that takes a value ten times larger than the baseline estimated γ.
13
2.8 3.0 3.2 3.4 3.6 3.8 4.0
1.85 1.95 2.05 2.15
Gamma
Total sum of squares
3.5 4.0 4.5 5.0
0.75 0.80 0.85 0.90 0.95 1.00
Gamma
Total sum of squares
ˆ
γestimated over 1970-1990 ˆ
γestimated over 1990-2000
Figure 7: Shortening transition times: historical ˆ
γ
πt+h=trendt
+ (1 −F(ζt−1|¯
γ))(αh,EN |¯
γ+ϕh,EN |¯
γ(L)xt−1+βh,EN |¯
γut)
+F(ζt−1)(αh,LN|¯
γ+ϕh,LN|¯
γ(L)xt−1+βh,LN|¯
γut)
+ϵt+h
(8)
3.4 Changing ENSO patterns: an index to identify commodities under
stress
To summarize the information obtained using the estimates presented above, we construct an index
that captures the exposure of each commodity price to changing ENSO pattern.
To do this, we compare the results of the baseline estimate (equation 3in section 3.1) with
estimates obtained by considering stressed values, i.e. extreme values (equation 6and 7in section
3.2) or a faster transition speed (equation 8in section 3.1).
Each commodity is therefore evaluated under four stress exercises : during El Ni˜no periods
βh,EN is compared with βh,ENAand βh,E N |¯
γand during La Ni˜na periods βh,LN is compared with
βh,LNAand βh,LN |¯
γ.
We classify the result of each stress exercises into three categories : “Less volatility” (e.g
|βh,EN |>|βh,ENA|), “No change” (e.g βh,EN ∼βh,EN A) and “More volatility” (e.g |βh,EN |<
|βh,EN A|) .
Comparisons are made at a significance level of 5%.12
12Note that the value of his set to pick the period with the strongest effect. So we typically compare the strongest
effect of ENSO shocks on the price of cotton in the baseline scheme with the strongest effect of ENSO shocks on
14
Finally, to give an overview of the four stress scenarios in a single measure, we calculate an
index centered on 0, which increases by one unit for any scenario concluding to higher volatility
and decreases for any scenario concluding to lower volatility.
The lowest possible value is therefore -4 (all scenarios lead to more volatility) and the maximum
is +4 (all scenarios lead to more volatility). The central value, 0, corresponds to the case where
there are no more scenarios leading to more volatility than scenarios leading to greater stability
(or simply ENSO shocks are never find to impact significantly the commodity price).
the price of cotton in the case of anomalies, and this effect may arrive after 7 periods in one case and 8 periods in
the other, it doesn’t matter, we just want to measure the capacity of the shock to vary the price of cotton.
15
4 Baseline results
4.1 ENSO impact on aggregated prices
The IRFs are obtained by scaling the estimated coefficient (βh) to a 1 standard deviation shock.
As MEI.V2 is positive (negative) in average during an El Ni˜no (La Ni˜na) phase, we assume a
positive (negative) shock. By doing so, we make sure that a shock can always been interpreted
as strengthening ENSO phenomenon.13 The impulse response functions (IRF) are presented in
Figure 8. The IRF plots include the 68 and 90% confidence band using the Newey West standard
errors.
IRFs provide interesting information. A shock during an La Ni˜na phase tends to raise prices
for all three indices (agriculture, energy, minerals). Conversely, a shock during an El Ni˜no phase
tends to push the Energy and Met/Min price indices down. As the two shocks are opposite in
nature, the asymmetry between El Ni˜no and La Ni˜na seems limited for Energy prices. However,
the magnitude of the results is not the same during El Ni˜no and LN. The three indices have a
more marked reaction during LN. The reaction of the agricultural index is significantly different
from 0 during LN, but not during EN. The same behavior is observed for Energy. Finally, all 3
indices show low reactivity. IRF values are generally not different from 0. This may be due to
heterogeneity of the prices making up these indices.
4.2 ENSO impact on individual commodities
An important limitation of the results presented above is that they are based on aggregated vari-
ables. These three price indices group together very different commodities, produced in different
places and under different climates. They are therefore average results, which may conceal very
different realities. We therefore carried out the same exercise for each of the 67 commodities, one
by one. A quick overview is provided by Figure 9, considering significance at 10%. The effect of
an ENSO shock pushes some prices upwards, but also pushes other prices downwards. This result
explains why the IRF on aggregated indices in the figure above appear limited.
Finally, we plot the reactions, grouped by commodity type on Figure 10. Each point has the
price reaction during El Ni˜no as its ordinate and the reaction during La Ni˜na as its abscissa. For
any group, linear reactions can be found in the upper left-hand box (ENSO shock is positive during
El Ni˜no and negative during LN) or in the lower right-hand box (ENSO shock is positive during La
Ni˜na and negative during EN), along the -45 degrees line . Only a very few commodities are pre-
cisely on the line. For example, among the Agricultural Raw Materials, two commodities, namely
Cotton and Log are on the line, therefore displaying a linear reaction to MEIV2 shocks, whatever
the state (EN or LN). Other commodities reaction to ENSO shocks appear to be dependent on
ENSO state (EN or LN). These is non-linearities confirm our empirical strategy.
13If the response function equals 0.02 at time t+3, it basically means that prices were 2% larger than their mean
value at time t+3, in reaction to a 1 standard deviation increase of MEI.V2 that happened in t0.
16
La Ni˜na El Ni˜no
Agriculture index
-.05 0 .05
Percent
0 5 10 15 20 25
months
Notes: 68 and 90 percent confidence bands
during LN phase
Responses of iAGRICULTURE prices to 1 sd MEIV2 shock
-.02 0 .02 .04
Percent
0 5 10 15 20 25
months
Notes: 68 and 90 percent confidence bands
during EN phase
Responses of iAGRICULTURE prices to 1 sd MEIV2 shock
Energy index
-.1 -.05 0 .05 .1 .15
Percent
0 5 10 15 20 25
months
Notes: 68 and 90 percent confidence bands
during LN phase
Responses of iENERGY prices to 1 sd MEIV2 shock
-.1 -.05 0 .05
Percent
0 5 10 15 20 25
months
Notes: 68 and 90 percent confidence bands
during EN phase
Responses of iENERGY prices to 1 sd MEIV2 shock
Metals and Minerals index
-.15 -.1 -.05 0 .05
Percent
0 5 10 15 20 25
months
Notes: 68 and 90 percent confidence bands
during LN phase
Responses of iMETMIN prices to 1 sd MEIV2 shock
-.06 -.04 -.02 0 .02 .04
Percent
0 5 10 15 20 25
months
Notes: 68 and 90 percent confidence bands
during EN phase
Responses of iMETMIN prices to 1 sd MEIV2 shock
Notes : responses of price indices to 1 sd MEIV2 negative shock during La Ni˜na phase (solid
blue) and positive shock during El Ni˜no phase (dash red), percent, with 68 and 90 percent
confidence bands
Figure 8: Impact of MEI.v2 on commodities
17
La Ni˜na El Ni˜no
22
20
0 5 10 15 20
Number of commodities
Inflation
Deflation
25
15
0 5 10 15 20 25
Number of commodities
Inflation
Deflation
Notes: responses of price indices to 1 sd MEIV2 negative shock during La Ni˜na phase and
positive shock during El Ni˜no phase percent, considering significance at 10%, over a total of 67
commodities
Figure 9: Inflationary or deflationary impact of MEI.v2 on commodities
18
Agricultural Raw Materials Agriculture Food and Beverage
LOGSMYS
PLYWOOD
COTTONAINDX
RUBBER1MYSG
SAWNWDCMR
TOBACUS
LOGSCMR
RUBBERTSR20
SAWNWDMYS
-.05 0 .05 .1 .15 .2
ENSO shock impact on commodities during EN
-.15 -.1 -.05 0 .05
ENSO shock impact on commodities during LN
COFFEEARABIC
COFFEEROBUS
FISHMEAL
GRNUTOIL
RICEA1
SHRIMPMEX
SORGHUM
TEAAVG
BANANAEU
BARLEY MAIZE
PLMKRNLOIL
SUGARWLD
WHEATUSHRW
ORANGE
COCONUTOIL
PALMOIL
TEAKOLKATA
WHEATUSSRW
BANANAUS
SOYBEANMEAL
COCOA
SUGAREU
TEACOLOMBO
TEAMOMBASA LAMB
SOYBEANS
BEEF
CHICKEN
GRNUT
RICE05
SOYBEANOIL
SUGARUS
-.1 -.05 0 .05 .1 .15
ENSO shock impact on commodities during EN
-.15 -.1 -.05 0 .05 .1
ENSO shock impact on commodities during LN
Energy Fertilizers
COALSAFRICA
iNATGAS
CRUDEWTI
COALAUS
CRUDEBRENT
CRUDEDUBAI
CRUDEPETRO
NGASEUR
NGASJP
NGASUS
-.2 -.1 0 .1 .2
ENSO shock impact on commodities during EN
-.2 -.1 0 .1 .2
ENSO shock impact on commodities during LN
TSP
UREAEEBULK
DAP PHOSROCK
POTASH
-.2 -.1 0 .1
ENSO shock impact on commodities during EN
-.1 0 .1 .2 .3
ENSO shock impact on commodities during LN
Metals and Minerals Precious Metal
COPPER
ZINC
NICKEL
ALUMINUM
TIN
IRONORE
LEAD
-.1 -.05 0 .05 .1 .15
ENSO shock impact on commodities during EN
-.15 -.1 -.05 0
ENSO shock impact on commodities during LN
GOLD
SILVER
PLATINUM
-.05 0 .05
ENSO shock impact on commodities during EN
-.05 0 .05
ENSO shock impact on commodities during LN
Notes : Maximum response over 24 months, of each individual commodities to 1 sd MEIV2
negative shock during La Ni˜na phase and positive shock during El Ni˜no phase.
Figure 10: Reactions to ENSO shocks grouped by commodity categories
19
5 ENSO under stress: an exposure index
5.1 Index overview
For the 67 commodities, we calculate the price response to an ENSO shock in the baseline frame-
work, and then repeat the process according to our two scenarios for the evolution of ENSO:
assuming extreme conditions and a faster transition speed (see Figure 13). Then, we compute,
for the 67 commodities, our index of commodity price exposure to the evolution of ENSO. The
index, represented on Figure 11, is centered on 0, in which case the evolution of ENSO patterns
does not, on average, change the effect of ENSO shocks on the price of these commodities. This is
the case for 31% of the commodities. The index takes a negative value for 13% of the commodi-
ties (9+1+3). For these commodities, the effect of ENSO shocks on commodity prices should be
less significant in the future, due to the evolution of ENSO patterns. More specifically, for 3%
of the commodities, we observe a highly stabilizing effect from the evolution of ENSO patterns.
However, this stabilizing effect remains rare. A majority of commodities are expected to be more
exposed to ENSO shocks in the future, due to the evolution of ENSO patterns. Indeed, in 63% of
cases (31+28+3+1), our index takes a positive value. In 4% of cases (3+1), we even find a highly
destabilizing effect, with our index taking values greater than or equal to three.
Thus, we can draw two conclusions from our index of commodity exposure to the evolution of
ENSO. Firstly, the evolution of exposure is highly heterogeneous across commodities, with some
becoming less exposed while others are much more exposed. Secondly, the dominant effect clearly
points towards an accentuation of commodity price volatility due to ENSO shocks, as the index
takes a positive value in a ma jority of cases.
0
3
1
9
22
31
28
3
1
0 10 20 30
Percentage of commodities
-4 -3 -2 -1 0 1 2 3 4
less volatility ←---→ more volatility
Exposure Index
Figure 11: Commodity price exposure to the evolution of ENSO: index overview
20
La Ni˜na El Ni˜no
ALUMINUM
BANANA EU
BANANA US
BARLEY
BEEF
CHICKEN
COAL AUS
COAL SAFRICA
COCOA
COCONUT OIL
COFFEE ARABIC
COFFEE ROBUS
COPPER
COTTON A INDX
CRUDE BRENT
CRUDE DUBAI
CRUDE PETRO
CRUDE WTI
DAP
FISH MEAL
GOLD
GRNUT
GRNUT OIL
IRON ORE
LAMB
LEAD
LOGS CMR
LOGS MYS
MAIZE
NGAS EUR
NGAS JP
NGAS US
NICKEL
ORANGE
PALM OIL
PHOSROCK
PLATINUM
PLMKRNL OIL
PLYWOOD
POTASH
RICE 05
RICE A1
RUBBER1 MYSG
RUBBER TSR20
SAWNWD CMR
SAWNWD MYS
SHRIMP MEX
SILVER
SORGHUM
SOYBEAN MEAL
SOYBEAN OIL
SOYBEANS
SUGAR EU
SUGAR US
SUGAR WLD
TEA AVG
TEA COLOMBO
TEA KOLKATA
TEA MOMBASA
TIN
TOBAC US
TSP
UREA EE BULK
WHEAT US HRW
WHEAT US SRW
ZINC
iNATGAS
-.6 -.4 -.2 0 .2 .4
Percent
Baseline Large gamma Anomalies
ALUMINUM
BANANA EU
BANANA US
BARLEY
BEEF
CHICKEN
COAL AUS
COAL SAFRICA
COCOA
COCONUT OIL
COFFEE ARABIC
COFFEE ROBUS
COPPER
COTTON A INDX
CRUDE BRENT
CRUDE DUBAI
CRUDE PETRO
CRUDE WTI
DAP
FISH MEAL
GOLD
GRNUT
GRNUT OIL
IRON ORE
LAMB
LEAD
LOGS CMR
LOGS MYS
MAIZE
NGAS EUR
NGAS JP
NGAS US
NICKEL
ORANGE
PALM OIL
PHOSROCK
PLATINUM
PLMKRNL OIL
PLYWOOD
POTASH
RICE 05
RICE A1
RUBBER1 MYSG
RUBBER TSR20
SAWNWD CMR
SAWNWD MYS
SHRIMP MEX
SILVER
SORGHUM
SOYBEAN MEAL
SOYBEAN OIL
SOYBEANS
SUGAR EU
SUGAR US
SUGAR WLD
TEA AVG
TEA COLOMBO
TEA KOLKATA
TEA MOMBASA
TIN
TOBAC US
TSP
UREA EE BULK
WHEAT US HRW
WHEAT US SRW
ZINC
iNATGAS
-.4 -.2 0 .2 .4 .6
Percent
Baseline Large gamma Anomalies
Figure 12: Commodities under stress
La Ni˜na El Ni˜no22
ALUMINUM
BANANA EU
BANANA US
BARLEY
BEEF
CHICKEN
COAL AUS
COAL SAFRICA
COCOA
COCONUT OIL
COFFEE ARABIC
COFFEE ROBUS
COPPER
COTTON A INDX
CRUDE BRENT
CRUDE DUBAI
CRUDE PETRO
CRUDE WTI
DAP
FISH MEAL
GOLD
GRNUT
GRNUT OIL
IRON ORE
LAMB
LEAD
LOGS CMR
LOGS MYS
MAIZE
NGAS EUR
NGAS JP
NGAS US
NICKEL
ORANGE
PALM OIL
PHOSROCK
PLATINUM
PLMKRNL OIL
PLYWOOD
POTASH
RICE 05
RICE A1
RUBBER1 MYSG
RUBBER TSR20
SAWNWD CMR
SAWNWD MYS
SHRIMP MEX
SILVER
SORGHUM
SOYBEAN MEAL
SOYBEAN OIL
SOYBEANS
SUGAR EU
SUGAR US
SUGAR WLD
TEA AVG
TEA COLOMBO
TEA KOLKATA
TEA MOMBASA
TIN
TOBAC US
TSP
UREA EE BULK
WHEAT US HRW
WHEAT US SRW
ZINC
iNATGAS
-.6 -.4 -.2 0 .2 .4
Percent
Baseline Large gamma Anomalies
ALUMINUM
BANANA EU
BANANA US
BARLEY
BEEF
CHICKEN
COAL AUS
COAL SAFRICA
COCOA
COCONUT OIL
COFFEE ARABIC
COFFEE ROBUS
COPPER
COTTON A INDX
CRUDE BRENT
CRUDE DUBAI
CRUDE PETRO
CRUDE WTI
DAP
FISH MEAL
GOLD
GRNUT
GRNUT OIL
IRON ORE
LAMB
LEAD
LOGS CMR
LOGS MYS
MAIZE
NGAS EUR
NGAS JP
NGAS US
NICKEL
ORANGE
PALM OIL
PHOSROCK
PLATINUM
PLMKRNL OIL
PLYWOOD
POTASH
RICE 05
RICE A1
RUBBER1 MYSG
RUBBER TSR20
SAWNWD CMR
SAWNWD MYS
SHRIMP MEX
SILVER
SORGHUM
SOYBEAN MEAL
SOYBEAN OIL
SOYBEANS
SUGAR EU
SUGAR US
SUGAR WLD
TEA AVG
TEA COLOMBO
TEA KOLKATA
TEA MOMBASA
TIN
TOBAC US
TSP
UREA EE BULK
WHEAT US HRW
WHEAT US SRW
ZINC
iNATGAS
-.4 -.2 0 .2 .4 .6
Percent
Baseline Large gamma Anomalies
Figure 13: Commodities under stress (cont.)
22
5.2 Index construction
Before we address the question of which commodities are most exposed and why, it is important
to revisit the values taken by the index. The index reflects, for a given commodity, whether, in
the event of a change in ENSO patterns, the effect of a shock is stronger or weaker than currently
observed (identified in the baseline). However, as shown in the Figure 14, the two scenarios have
vastly different implications. In the case of a faster transition speed (represented by an increase in
the value of parameter γ), the effect of an ENSO shock on the commodity price remains unchanged
for a majority of commodities. There is no significant difference. Conversely, in the scenario with
stronger ENSO events, anomalies increase volatility for approximately one commodity out of two.
This is an important finding: an increase in the transition speed of ENSO cycles is less concerning
than an increase in the intensity of ENSO phases.
5.3 Index by commodity categories
In order to identify which commodities are most exposed to the evolution of ENSO patterns,
we represent the value of the index for each commodity, and group commodities by categories
(Agriculture, Energy, etc.). The result is depicted in Figure 15.
No category stands out from the others. In all cases, significant heterogeneity is observed. One
exception to note is that for oil prices, the index systematically takes the value of 0. There are
two possible explanation. Firstly, climatic conditions only have a minimal impact on oil prices.
Secondly, oil prices appear to be linearly related to MEI.V2, meaning that positive shocks (El
Ni˜no) are offset by negative shocks (La Ni˜na), and thus these prices do not particularly stand out
when one looks at the overall trend.
5.4 Determining factors
Our index reveals a significant heterogeneity in commodity price exposure to changing ENSO
patterns. Several explanations can be put forward to account for this heterogeneity. We consider
three possibilities. Firstly, volatility may be higher for commodities produced in geographic regions
most influenced by ENSO. Secondly, volatility may be higher for commodities whose production
is concentrated in a small number of areas. Finally, price volatility may be influenced by the
financialization of certain commodities, which tends to correlate prices.
5.4.1 Financialization
As pointed by Tang and Xiong (2012), index investment in commodity markets increases the
correlation between non-energy and energy commodity prices. For these authors, the financial-
ization of the commodity markets explains part of the price volatility of non-energy commodities
around 2008. More recently, Kang et al. (2023) update this result and confirm that financializa-
tion increase pairwise return correlation within commodity futures markets. Figure 16 display our
Exposure Index for commodities exposed to financialization (top panel) and for commodities not
23
8
10
49
52
10
5
0 10 20 30 40 50
Number of commodities
Less volatility No change More volatility
Assuming shorter ENSO cycles, how do commodities evolve in response to ENSO shocks?
EN LN
11
3
31
17
25
47
0 10 20 30 40 50
Number of commodities
Less volatility No change More volatility
Assuming stronger ENSO events, how do commodities evolve in response to ENSO shocks?
EN LN
Notes : Based on the maximum response over 24 months, for each individual commodities, to
1 sd MEIV2 negative shock during La Ni˜na phase and positiv shock during El Ni˜no phase,
considering on a 90 percent confidence bands, comparing local projection with γ= 4 and
γ= 40 on the upper graph and comparing local projection with baseline EN/LN definition and
anomalies on the lower graph.
Figure 14: Changing ENSO paterns impact on commodities
exposed (bottom panel). We assume commodities to be exposed to financialization if they appear
in the Bloomberg Commodity Index (BCOM Index) or the Thomson Reuters CoreCommodity
(TR CRB) Index . As the Exposure Index is more right skew on the top panel, financialization
seems to contribute to volatility.
Indeed, financialization involves greater participation of financial investors, such as hedge funds
24
Agricultural Raw Materials Agriculture Food and Beverage
-3 -2 -1 0 1 2 3 4
Exposure Index
COTTON A INDX
LOGS CMR
LOGS MYS
PLYWOOD
RUBBER TSR20
RUBBER1 MYSG
SAWNWD CMR
SAWNWD MYS
TOBAC US
-3 -2 -1 0 1 2 3 4
Exposure Index
BANANA EU
BANANA US
BARLEY
BEEF
CHICKEN
COCOA
COCONUT OIL
COFFEE ARABIC
COFFEE ROBUS
FISH MEAL
GRNUT
GRNUT OIL
LAMB
MAIZE
ORANGE
PALM OIL
PLMKRNL OIL
RICE 05
RICE A1
SHRIMP MEX
SORGHUM
SOYBEAN MEAL
SOYBEAN OIL
SOYBEANS
SUGAR EU
SUGAR US
SUGAR WLD
TEA AVG
TEA COLOMBO
TEA KOLKATA
TEA MOMBASA
WHEAT US HRW
WHEAT US SRW
Energy Fertilizers
-3 -2 -1 0 1 2 3 4
Exposure Index
COAL AUS
COAL SAFRICA
CRUDE BRENT
CRUDE DUBAI
CRUDE PETRO
CRUDE WTI
NGAS EUR
NGAS JP
NGAS US
iNATGAS
-3 -2 -1 0 1 2 3 4
Exposure Index
DAP
PHOSROCK
POTASH
TSP
UREA EE BULK
Metals and Minerals Precious Metal
-3 -2 -1 0 1 2 3 4
Exposure Index
ALUMINUM
COPPER
IRON ORE
LEAD
NICKEL
TIN
ZINC
-3 -2 -1 0 1 2 3 4
Exposure Index
GOLD
PLATINUM
SILVER
Figure 15: Index value by categories
or institutional investors in commodity markets. These investors often engage in trading aiming
to profit from short-term price movements rather than physical delivery or consumption of com-
modities. Their trading activities can amplify price fluctuations and contribute to increased market
volatility. Financialization can also lead to herding behavior among investors, where large numbers
of market participants follow similar investment strategies based on trends or market sentiment.
While the liquidity induced by financialization can enhance market efficiency and price discovery,
it can also lead to rapid price changes as large volumes of capital flow in and out of markets,
25
particularly during periods of market stress or uncertainty.
BCOM Index TR CRB Index
Commodities included in the index
0
6 6 6
17
33
22
6 6
0 10 20 30 40
Percentage of commodities
-4 0 4
less volatility ←---→ more volatility
Exposure Index: commodities included in BCOM Index
0
5 5
10
19
24 24
10
5
0 5 10 15 20 25
Percentage of commodities
-4 0 4
less volatility ←---→ more volatility
Exposure Index: commodities included in TR CRB Index
Commodities not included in the index
0
2
0
12
22
29
32
2
0
0 10 20 30
Percentage of commodities
-4 0 4
less volatility ←---→ more volatility
Exposure Index: commodities excluded from BCOM Index
0
2
0
10
22
34
32
0 0
0 10 20 30 40
Percentage of commodities
-4 0 4
less volatility ←---→ more volatility
Exposure Index: commodities excluded from TR CRB Index
Figure 16: Commodity price exposure and financialization
5.4.2 Production concentration
Commodity production frequently exhibits high levels of concentration due to natural resource
distributions or natural endowments. Consequently commodities elasticity of supply may be low:
the responsiveness of quantity supplied to changes in price is limited, commodities are difficult
to substitute in the short term, leading to more pronounced price movements. When commodity
production is concentrated in a few countries, any disruptions to production in these areas can have
a significant impact on overall supply. We therefore expect our Exposure Index to be positively
correlated with concentration.
Figure 17 display our Exposure Index for commodities whose production is concentrated on
a limited number of countries (top panel) and for commodities produced more broadly (bottom
panel). We capture this feature using IMF (2023) data about the share of countries that import
26
a given commodity from three suppliers only (left column) and data about the share of top three
countries in total commodity world production (right column). We assume production to be
concentrated if the two variables take values larger than 8% and 15% respectively. As the Exposure
Index take greater values (+3, +4) on the top panel, production concentration seems to contribute
to volatility. This is less clear-cut when considering the share of top three producers (right column).
Share of Countries that Import Share of Top three Countries
from Three suppliers in Total Commodity Production
Exposure Index: commodities with production concentration
2 2
9
26 26
30
4
2
0 10 20 30
Percentage of commodities
-4 0 4
less volatility ←---→ more volatility
Exposure Index: commodities with Share of Countries that Import from Three suppliers > 8%
2
8
22
29
35
2 2
0 10 20 30 40
Percentage of commodities
-4 0 4
less volatility ←---→ more volatility
Exposure Index: commodities with Share of Top three Countries in Total Commodity Production > 15 %
Exposure Index: commodities without production concentration
0
8
0
8 8
54
23
0 0
0 20 40 60
Percentage of commodities
-4 0 4
less volatility ←---→ more volatility
Exposure Index: commodities with Share of Countries that Import from Three suppliers < 8 %
0
13
0
13
25
38
6 6
0
0 10 20 30 40
Percentage of commodities
-4 0 4
less volatility ←---→ more volatility
Exposure Index: commodities with Share of Top three Countries in Total Commodity Production < 15 %
Figure 17: Commodity price exposure and production concentration
5.4.3 Trade
As a result of commodity production concentration (mentioned earlier) access to global commodity
markets is essential for many countries. If the global market for a commodity is large, it promotes
27
market liquidity, facilitates access to diversified sources, facilitates arbitrage activities, improves
the flow of information and supports risk management strategies - all of which help to reduce
price volatility. In the opposite, according to Campos et al. (2023) and IMF (2023), market
fragmentation (typically due to geopolitical events) can lead to more volatility on commodity
markets. Figure 18 display our Exposure Index for commodities whose world production is largely
available on the market (top plot), considering commodities whose share of traded world production
is above 1/3, based on IMF (2023) data. The lower plot represent commodities with limited share
of traded world production. Our Exposure Index is clearly lower for largely traded commodities.
This is consistent with Gouel and Laborde (2021) among other, who shows the crucial role of
international markets for agricultural products in adapting to climate change.
Share of Traded World Production
Commodities with large Share of Traded World Production
4
2
6
15
38
30
4
0 10 20 30 40
Percentage of commodities
-4 0 4
less volatility ←---→ more volatility
Exposure Index: commodities with Share of Traded World Production > 33%
Commodities with limited Share of Traded World Production
0 0 0
15
40
15
25
0
5
0 10 20 30 40
Percentage of commodities
-4 0 4
less volatility ←---→ more volatility
Exposure Index: commodities with Share of Traded World Production < 33%
Figure 18: Commodity price exposure and trade
28
5.4.4 South America
Finally, as its name suggests, ENSO could affect South American countries more than others. This
idea is not correct, ENSO being a global phenomena. For example, El Ni˜no brings drier conditions
to southern Africa and parts of the Sahel, while eastern equatorial Africa experiences wetter con-
ditions during the short rainy season, and rainfall in South and Southeast Asia is decreasing. To
illustrate this point, we plot on Figure 19 our Exposure Index for commodities with a least one of
the top three countries in total commodity production located in South America (top panel, based
on IMF (2023) data ). Our exposure index is no higher than for commodities whose main producer
is in any other region (bottom panel), confirming that ENSO is a global phenomenon and that all
regions are affected by its evolution.
Top Three Countries in Total Commodity Production
At least 1 of the Top Three Countries in Total Commodity Production is located in South America
8
17
33
8
25
8
0 10 20 30 40
Percentage of commodities
-4 0 4
less volatility ←---→ more volatility
Exposure Index: at least 1 of the Top is located in South America
None of the Top Three Countries in Total Commodity Production is located in South America
0
4 4
7
25
36
21
0
4
0 10 20 30 40
Percentage of commodities
-4 0 4
less volatility ←---→ more volatility
Exposure Index: none of the Top is located in South America
Figure 19: Commodity price exposure and production concentration in South America
29
6 Conclusion
This paper analyzes the global transmission of weather on commodity prices and proposes an
assessment of the potential effects of climate change on price stability.
We estimate a global factor augmented non-linear local pro jections model using a rich and
extensive monthly data set from 1986 01 to 2023 06 relating to circa two-thirds of global output
and 67 international commodity price sets. We contribute to a growing literature on the ”new
climate economy” (Dell et al.,2014) in two ways. First, this paper exploits the global factor
structure to investigate the global dimensions of weather and commodity shocks. Second, we
exploit the multivariate dimension of data using a nonlinear framework to account for possible
changes in climate regimes.
We first demonstrate that El Ni˜no and La Ni˜na climatic events have an impact on commodity
prices. At the aggregate level, a non-expected evolution of the ENSO cycle has a particularly
significant impact during La Ni˜na phases, especially pronounced for energy and agricultural goods.
This effect is observed with a lag of 6 to 12 months for agricultural goods, which could reflect both
the time lag between weather events and crop outcomes and the importance of futures markets in
price determination.
These relatively modest results at the aggregate level contrast with much more significant
impacts at the disaggregated level when estimating the effect of a non-expected evolution of the
ENSO cycle on commodity prices individually. We show that about two third of commodity prices
are impacted by a non-expected evolution of the ENSO cycle (or ENSO shock), generally with
non-linearity associating climatic conditions with commodity prices.
We conduct two exercises to simulate the impact of climate change on the ENSO cycle and
thereby measure the repercussions on commodity prices. One exercise focuses on intensity, and
the other on the transition speed of the cycles. Thus, we obtain an index that captures commodity
price exposure to the evolution of ENSO, due to climate change. We then show that in most cases,
climate change is likely to result in greater commodity prices volatility. This result is particularly
explained by the assumption of increased extreme events, which seem to have a greater impact
on commodity prices than the evolution of the frequency of EN/LN cycles. Our results indicate
significant heterogeneity among commodities, with some being minimally or not impacted by cli-
mate change (and a handful of commodity prices expected to me even more stable than now) while
others are expected to experience significantly increased volatility, as represented by our index.
We carry out several tests to explain this heterogeneity, and show that financialization, production
concentration and the fact that only a small proportion of a commodity is sold on the world market
tend to increase volatility. Conversely, the origin of production of the commodity plays little role:
ENSO does not only affect South American countries; ENSO is a global phenomenon, therefore all
part of the world may somehow be impacted.
The results from this research have both immediate and long-term policy implications regarding
the adaptation to climate changes. A starting point for short-term policy is to establish sources of
vulnerability that could create economic risks. The findings from this paper serve to do just that
30
in a global factor environment by analyzing the propagation mechanisms through which weather
shocks influence commodity inflation. Our work highlights the ways in which climate change can
create new challenges for financial and price stability, and underscores the importance of inter-
national trade. First, our work is important for policymaker with regards to financial stability:
we show how climate change can create new challenges, particularly when commodities are inte-
grated into many financial products (see Adams and Gl¨uck,2015,Basak and Pavlova,2016 on the
spillovers between commodities and the stock market). Second, these insights are of paramount
importance for for central banks whose mandate is to insure price stability. We show that climate
change will contribute to commodity price volatility, which is a key determinant of headline in-
flation. Although the primary objective of central banks is generally core inflation, central banks
need to monitor headline inflation to ensure that second-round effects are limited. In this respect,
increased commodity volatility can complicate their mission of price stability. Finally, our research
underscores the significance of integrated global commodity markets in managing supply shocks
and mitigating rising price volatility. Consequently, international commodity markets should be
harnessed to meet the challenges posed by climate change.
7 Bibliography
Acevedo, S., Mrkaic, M., Novta, N., Pugacheva, E., and Topalova, P. (2020). The effects of weather
shocks on economic activity: What are the channels of impact? Journal of Macroeconomics,
65:103207.
Adams, Z. and Gl¨uck, T. (2015). Financialization in commodity markets: A passing trend or the
new normal? Journal of Banking & Finance, 60:93–111.
Alexander, M., Blad´e, I., Newman, M., Lanzante, J., Lau, N.-C., and Scott, J. (2002). The
atmospheric bridge: The influence of ENSO teleconnections on air–sea interaction over the
global oceans. Journal of Climate, 15(16):2205–2231.
An, S.-I. and Jin, F.-F. (2004). Nonlinearity and asymmetry of ENSO. Journal of Climate,
17(12):2399–2412.
An, S.-I. and Wang, B. (2000). Interdecadal change of the structure of the ENSO mode and its
impact on the ENSO frequency. Journal of Climate, 13(12):2044–2055.
Anand, R., Ding, D., and Tulin, V. (2014). Food inflation in India: the role for monetary policy.
Number 14-178. International Monetary Fund.
Anttila-Hughes, J. K., Jina, A. S., and McCord, G. C. (2021). ENSO drives child undernutrition
in the global tropics.
31
Atems, B. and Sardar, N. (2021). Exploring asymmetries in the effects of El Ni˜no-Southern
Oscillation on US food and agricultural stock prices. The Quarterly Review of Economics and
Finance, 81:1–14.
Auerbach, A. J. and Gorodnichenko, Y. (2012). Measuring the output responses to fiscal policy.
American Economic Journal: Economic Policy, 4(2):1–27.
Barlow, M., Nigam, S., and Berbery, E. (2001). ENSO, pacific decadal variability, and US sum-
mertime precipitation, drought, and stream flow. Journal of Climate, 14(9):2105–2128.
Barnston, A. (2015). Why are there so many ENSO indexes, instead of just one. NOAA Climate.
gov.
Basak, S. and Pavlova, A. (2016). A model of financialization of commodities. The Journal of
Finance, 71(4):1511–1556.
Battiston, S., Dafermos, Y., and Monasterolo, I. (2021). Climate risks and financial stability.
Berry, B. J. and Okulicz-Kozaryn, A. (2008). Are there ENSO signals in the macroeconomy.
Ecological Economics, 64(3):625–633.
Bertrand, A., Lengaigne, M., Takahashi, K., Avadi, A., Poulain, F., and Harrod, C. (2020). El
Ni˜no Southern Oscillation (ENSO) effects on fisheries and aquaculture, volume 660. Food and
Agriculture Organization.
Bloom, N., Chen, J., and Mizen, P. (2018). Rising Brexit uncertainty has reduced investment and
employment. VOX CEPR policy portal, 16.
Boneva, L., Ferrucci, G., and Mongelli, F. P. (2022). Climate change and central banks: what role
for monetary policy? Climate Policy, 22(6):770–787.
Br¨onnimann, S., Xoplaki, E., Casty, C., Pauling, A., and Luterbacher, J. (2007). ENSO influence
on Europe during the last centuries. Climate Dynamics, 28(2-3):181–197.
Brunner, A. D. (2002). El Nino and world primary commodity prices: warm water or hot air?
Review of Economics and Statistics, 84(1):176–183.
Buhr, B., Volz, U., Donovan, C., Kling, G., Lo, Y. C., Murinde, V., and Pullin, N. (2018). Climate
change and the cost of capital in developing countries.
Cai, W., Borlace, S., Lengaigne, M., Van Rensch, P., Collins, M., Vecchi, G., Timmermann, A.,
Santoso, A., McPhaden, M. J., Wu, L., et al. (2014). Increasing frequency of extreme El Ni˜no
events due to greenhouse warming. Nature climate change, 4(2):111–116.
Cai, W., Santoso, A., Collins, M., Dewitte, B., Karamperidou, C., Kug, J.-S., Lengaigne, M.,
McPhaden, M., Stuecker, M., and Taschetto, A. (2021). Changing El Ni˜no–Southern oscillation
in a warming climate. Nature Reviews Earth & Environment, 2(9):628–644.
32
Callahan, C. W. and Mankin, J. S. (2023). Persistent effect of El Ni˜no on global economic growth.
Science, 380(6649):1064–1069.
Campos, R. G., Estefania-Flores, J., Furceri, D., and Timini, J. (2023). Geopolitical fragmentation
and trade. Journal of Comparative Economics, 51(4):1289–1315.
Carleton, T. A. and Hsiang, S. M. (2016). Social and economic impacts of climate. Science,
353(6304):aad9837.
Cashin, P., Mohaddes, K., and Raissi, M. (2017). Fair weather or foul? the macroeconomic effects
of el ni˜no. Journal of International Economics, 106:37–54.
Cevik, M. S. and Jalles, J. T. (2023). Eye of the storm: The impact of climate shocks on inflation
and growth. Technical report, International Monetary Fund.
Chen, C., McCarl, B., and Adams, R. (2001). Economic implications of potential ENSO frequency
and strength shifts. Climatic Change, 49(1):147–159.
Dai, A. (2013). Increasing drought under global warming in observations and models. Nature
Climate Change, 3(1):52–58.
Damette, O., Mathonnat, C., and Thavard, J. (2024). Climate and sovereign risk: The Latin
American experience with strong ENSO events. World Development, 178:106590.
De Marco, F., Limodio, N., Eufinger, C., Gorostiaga, J. P., Richter, B., Martini, F., Steffen, Z.,
Theunisz, C., and Lattanzio, G. (2023). The financial transmission of a climate shock: El Nino
and US banks. Available at SSRN, 4279194.
De Winne, J. and Peersman, G. (2021). The adverse consequences of global harvest and weather
disruptions on economic activity. Nature Climate Change, 11(8):665.672.
Dell, M., Jones, B. F., and Olken, B. A. (2014). What do we learn from the weather? The new
climate-economy literature. Journal of Economic Literature, 52(3):740–98.
Diaz, H. F., Hoerling, M. P., and Eischeid, J. K. (2001). ENSO variability, teleconnections and
climate change. International Journal of Climatology: A Journal of the Royal Meteorological
Society, 21(15):1845–1862.
Fabris, N. (2020). Financial stability and climate change. Journal of Central Banking Theory and
Practice, 9(3):27–43.
Flori, A., Pammolli, F., and Spelta, A. (2021). Commodity prices co-movements and financial
stability: A multidimensional visibility nexus with climate conditions. Journal of Financial
Stability, 54(C).
33
Gabriel, V., Levine, P., Pearlman, J., and Yang, B. (2012). An Estimated DSGE Model of the
Indian Economy, The Oxford Handbook of the Indian Economy, (Ed). Chetan Ghate. Oxford
University Press.
Generoso, R., Couharde, C., Damette, O., and Mohaddes, K. (2020). The growth effects of El Nino
and La Nina: Local weather conditions matter. Annals of Economics and Statistics, (140):83–
126.
Ginn, W. and Pourroy, M. (2022). The contribution of food subsidy policy to monetary policy in
India. Economic Modelling, 113:105904.
Giuzio, M., Kruˇsec, D., Levels, A., Melo, A. S., Mikkonen, K., and Radulova, P. (2019). Climate
change and financial stability. Financial Stability Review, 1.
Gorodnichenko, Y. and Auerbach, A. J. (2013). Fiscal Multipliers in Recession and Expansion.
University of Chicago Press.
Gouel, C. and Laborde, D. (2021). The crucial role of domestic and international market-
mediated adaptation to climate change. Journal of Environmental Economics and Management,
106:102408.
Guo, Z., Zhou, T., and Wu, B. (2017). The asymmetric effects of el ni˜no and la ni˜na on the east asian
winter monsoon and their simulation by cmip5 atmospheric models. Journal of Meteorological
Research, 31(1):82–93.
Hansen, J., Hodges, A., and Jones, J. (1998). ENSO influences on agriculture in the southeastern
United States. Journal of Climate, 11(3):404–411.
Henseler, M. and Schumacher, I. (2019). The impact of weather on economic growth and its
production factors. Climatic change, 154(3-4):417–433.
Hsiang, S., Meng, K., and Cane, M. (2011). Civil conflicts are associated with the global climate.
Nature, 476(7361):438–441.
Hu, K., Huang, G., Huang, P., Kosaka, Y., and Xie, S.-P. (2021). Intensification of El Ni˜no-induced
atmospheric anomalies under greenhouse warming. Nature Geoscience, 14(6):377–382.
IMF (2023). Divergences, navigating global. World Economic Outlook.
Jord`a, `
O. (2005). Estimation and inference of impulse responses by local projections. American
Economic Review, 95(1):161–182.
Kabundi, A. N., Mlachila, M. M., and Yao, J. (2022). How persistent are climate-related price
shocks? implications for monetary policy. Technical report, International Monetary Fund.
Kang, W., Tang, K., and Wang, N. (2023). Financialization of commodity markets ten years later.
Journal of Commodity Markets, 30:100313.
34
Kranz, T., Bennani, H., and Neuenkirch, M. (2024). Monetary policy and climate change: Chal-
lenges and the role of major central banks. Technical report, University of Trier, Department of
Economics.
Legler, D. M., Bryant, K. J., and O’brien, J. J. (1999). Impact of ENSO-related climate anomalies
on crop yields in the us. Climatic Change, 42:351–375.
Li, Y., Yi, F., Wang, Y., and Gudaj, R. (2019). The Value of El Ni˜no-Southern Oscillation
Forecasts to China’s Agriculture. Sustainability, 11(15):1–23.
Liu, Y., Cai, W., Lin, X., Li, Z., and Zhang, Y. (2023). Nonlinear El Ni˜no impacts on the global
economy under climate change. Nature Communications, 14(1):5887.
Mainardi, S. (2011). Cropland use, yields, and droughts: spatial data modeling for Burkina Faso
and Niger. Agricultural Economics, 42(1):17–33.
McPhaden, M., Santoso, A., and Cai, W. (2020). Introduction to El Ni˜no southern oscillation in
a changing climate. El Ni˜no Southern Oscillation in a Changing Climate, pages 1–19.
McPhaden, M., Zebiak, S., and Glantz, M. (2006). ENSO as an integrating concept in earth
science. Science, 314(5806):1740–1745.
Melo-Velandia, L. F., Orozco-Vanegas, C. A., and Parra-Amado, D. (2022). Extreme weather
events and high Colombian food prices: A non-stationary extreme value approach. Agricultural
Economics, 53(S1):21–40.
Mourtzinis, S., Ortiz, B. V., and Damianidis, D. (2016). Climate change and ENSO effects on
southeastern us climate patterns and maize yield. Scientific reports, 6(1):29777.
Mueller, V. and Osgood, D. (2009). Long-term consequences of short-term precipitation shocks:
evidence from Brazilian migrant households. Agricultural Economics, 40(5):573–586.
Mukherjee, K. and Ouattara, B. (2021). Climate and monetary policy: do temperature shocks
lead to inflationary pressures? Climatic change, 167(3-4):32.
Patnaik, I., Shah, A., and Bhattacharya, R. (2011). Monetary policy transmission in an emerging
market setting. Number 11-15. International Monetary Fund.
Pointner, W. and Ritzberger-Gr¨unwald, D. (2019). Climate change as a risk to financial stability.
Financial Stability Report, 38:30–45.
Ramey, V. and Zubairy, S. (2018). Government spending multipliers in good times and in bad:
evidence from US historical data. Journal of Political Economy, 126(2):850–901.
Ratti, R. A. and Vespignani, J. L. (2016). Oil prices and global factor macroeconomic variables.
Energy Economics, 59:198–212.
35
Ropelewski, C. and Halpert, M. (1987). Global and regional scale precipitation patterns associated
with the El Ni˜no/Southern Oscillation. Monthly Weather Review, 115(8):1606–1626.
Rosenzweig, C., Iglesias, A., Yang, X., Epstein, P., and Chivian, E. (2001). Climate change and
extreme weather events. Global Change and Human Health, 2(2):90.
Sarachik, E. and Cane, M. (2010). The El Nino-southern oscillation phenomenon. Cambridge
University Press.
Saxegaard, M., Anand, R., and Peiris, S. (2010). An estimated model with macrofinancial linkages
for India. Number 10-21. International Monetary Fund.
Smith, S. and Ubilava, D. (2017). The El Ni˜no Southern Oscillation and economic growth in the
developing world. Global Environmental Change, 45:151–164.
Stan, K., Watt, G. A., and Sanchez-Azofeifa, A. (2021). Financial stability in response to climate
change in a northern temperate economy. Nature Communications, 12(1):7161.
Strabel, J. and Wurgler, J. (2021). What do you think about climate finance? CEPR Discussion
Papers 16622, C.E.P.R. Discussion Papers.
Svartzman, R., Bolton, P., Despres, M., Pereira Da Silva, L. A., and Samama, F. (2021). Central
banks, financial stability and policy coordination in the age of climate uncertainty: A three-
layered analytical and operational framework. Climate Policy, 21(4):563–580.
Tack, J. and Ubilava, D. (2015). Climate and agricultural risk: measuring the effect of ENSO on
U.S. crop insurance. Agricultural Economics, 46(2):245–257.
Tang, K. and Xiong, W. (2012). Index investment and the financialization of commodities.
Financial Analysts Journal, 68(6):54–74.
Timmermann, A., Oberhuber, J., Bacher, A., Esch, M., Latif, M., and Roeckner, E. (1999).
Increased El Ni˜no frequency in a climate model forced by future greenhouse warming. Nature,
398(6729):694–697.
Tol, R. S. (2024). A meta-analysis of the total economic impact of climate change. Energy Policy,
185:113922.
Ubilava, D. (2012). El Ni˜no, La Ni˜na, and world coffee price dynamics. Agricultural Economics,
43(1):17–26.
Ubilava, D. (2018). The role of El Nino Southern Oscillation in commodity price movement and
predictability. American Journal of Agricultural Economics, 100(1):239–263.
Ventosa-Santaul`aria, D., Tapia, E., and P´erez-Pe˜na, A. (2024). Inflation dynamics under different
weather regimes: Evidence from Mexico. Ecological Economics, 220:108179.
36
Vink, S. and Robbins, S. (2012). Mine water management in variable climate regimes. International
Mine Water Association, Annual Conference.
Yeh, S.-W., Cai, W., Min, S.-K., McPhaden, M., Dommenget, D., Dewitte, B., Collins, M., Ashok,
K., An, S.-I., and Yim, B.-Y. (2018). ENSO atmospheric teleconnections and their response to
greenhouse gas forcing. Reviews of Geophysics, 56(1):185–206.
37
8 Appendix
8.1 Data sources
Table 3provides a description a the data used in the paper.
Item Symbol Source Description
ESOI Weather WG,ESOI
tNOAA Equatorial Southern Oscillation Index
SST Weather WG,SST
tNOAA Sea Surface Temperature
Output Growth ∆ ln YtOECD, FRED Industrial Production
Commodity Inflation ∆ ln PG
tWorld Bank (see [a]) 34 International Commodity Prices Considered
Interest Rate RtOECD, FRED, CEIC Short-term interest rate
[a]: https://www.worldbank.org/en/research/commodity-markets.
Table 3: Variable Selection
All variables are converted to logarithm with the exception of the interest rate and weather in-
dices. The commodity price data, denominated in U.S. dollars, is deflated by dividing by Producer
Prices Index (OECD Total Area) . The ESOI and SST indices are used as a global measure of
weather, which is collected from the National Oceanic and Atmospheric Administration (NOAA).
38
8.2 Scree Plots
The scree plots are provided for the global factor analysis of output and interest rate. The scree
plot is a plot of the eigenvalues of principal components.
Figure 20: Scree Plots for Global Variables
39
ζ F (ζ)
γ= 5 γ= 1 γ= 0.1
-1 0.99 0.73 0.52
-0.1 0.62 0.52 0.50
00.50 0.50 0.50
0.1 0.38 0.48 0.50
0.2 0.27 0.45 0.50
0.3 0.18 0.43 0.49
10.01 0.27 0.48
Table 4: Simulation for illustration
purposes
0 .2 .4 .6 .8 1
f(ζ)
-1 -.5 0 .5 1
ζ
γ=5 γ=1 γ=.1
γ= 5 ; γ= 1 ; γ= 0.1
Figure 21: Representation of F(ζ) for alternative
γvalues
8.3 Exploring the transition function
To illustrate how does this function works, we simulate different values for ζ, and then observe
the values taken by F(ζ) assuming alternative values for γ. Table 4shows the numerical values
(the first column contains simulated values for ζwhile the second column shows the value taken
by F(ζ), the transition function, for each of the proposed values of ζ, assuming γ= 5. Similarly,
columns 3 and 4 show the values of F(ζ) for the same values of ζ, but assuming γ= 1 and γ= 0.1.
As it can be seen in Figure 21, for very small values of γ, the transition function F(ζ) varies
only slightly. Conversely, the higher the γ, the more abrupt the transition from a high to a low
value of F(ζ) .
40
