This analysis is an abbreviated version of a paper written for Penn State in early 2020. For a full version, contact INNER//JOIN

As this new decade has begun, it is incumbent upon all of us to recognize that we now exist in a world that is at risk of delivering a future far different from the present we preside in now; year over year, findings detail how our behaviors have caused irreversible environmental damage. These effects range from the more segmented to the far-reaching; while all these ramifications should be seen as distressing, it’s important that we use the plant life around us, the natural world’s KPI, as a litmus test for the current state of the climate. One crop in particular serves as both the backbone for our global food supply while also exposing itself to the consequences of our environmental malfeasance: soy. Soybean production is an industrial, logistical and agricultural behemoth with an increase of 200 million tons in global consumption since the seventies. The importance and reliance of the soybean lies within its multiple applications-- soybeans can be used in food products as edible vegetable oil, biofuel and most importantly its meal can be used as a protein source in livestock feeds. In recognizing this, we now have the tools at our disposal to properly measure and assess the effects of an ever-changing climate on such a crucial part of the global economy. Environmental analytics, agricultural biology, and even financial markets provide invaluable insights to growers, producers, and consumers on the extent of environmental effects. To focus on the financial sector, we must ask ourselves: How can we use the futures market to show the effect weather and climate factors have on soybeans? How can we use tools provided by the markets to analyze and benchmark climate variables, measure their effects on agriculture, and even improve the markets themselves? One of the greatest assets for this cause is utilizing the commodities market -- specifically in this case, soybean futures (CME ticker symbol: ZS). Historical data on futures give us insights on the directional move of the underlying asset in terms of market value, and gives us the price of a specific amount at a specific point in time. Cross-referencing these futures with historical climate and weather data gives us a crystal-clear window into how such atmospheric variables affect the pricing of given amounts of soy at a determined date. 

Such climate disruption brings unprecedented risk that will continue to put financial strain on growers and producers. While long-term solutions are seemingly becoming more perpetually long-term, there is an increasing need for growers to understand the value of their product, as it will help make decisions and project future efforts for their business. Enter: futures and the commodities market. A “futures contract” is an indicator of market value that is derived from the cash value of the underlying asset. The problem itself is that changes in climate patterns continue to threaten the routine of our global commodity system; to help assist with this problem, we need to see how much weather has an effect on commodity prices a la the commodity futures market. 

Knowing this problem lies ahead, the question becomes how can we use weather patterns to predict the future value of commodities? The major stakeholders affected by this continued risk include growers, the agricultural industrial complex, importers/shipping, traders, and investors. With respect to regionality, our analysis will be assessing global trends as they pertain to soy value and climate effects. While it’s easy to focus solely on the United States (due in part to their market share of soy), we will remain privy to major global weather or geopolitical events, as these can greatly impact the futures market. Now, what’s the hook? How can relevant stakeholders be convinced to consider weather data in futures projections, and therefore how can weather data be applicable in calculating the value of a crop or harvest? The cost here is whatever it takes to get all necessary data and to provide analysis in a manner where it can evenly mesh with the current data stream surrounding futures projections and grower profits. Would we be able to say that increased temperature and decreased precipitation affects not only crop yield, but the overall future value of the market? How can that be proven? These are all important questions to keep in mind as we go forward. 

In order to give us a high-level overview of global weather patterns during the given timeframe, the data collected is from the Global Climate Model HadGEM2-AO. From this model, monthly global average temperature and precipitation was pulled. While it might seem overly general to get global monthly averages, remember that the data we’re measuring against these extracts is affected by global weather patterns, not just domestic or location-specific. This is true not just with soybean futures, but across the entire commodity market -- per MarketWatch, extreme weather events may have as much impact on world oil supply, demand and prices as geopolitical risks in 2020, according to S&P Global Platts annual outlook. Our problem revolves around using historical weather data and the futures market to determine the volatility of a specific commodity, given weather effect “X”. Here, the “X” can mean an fluctuation in temperature or precipitation in such a manner that it disrupts the markets. How can weather change the overall value? To understand that, one must understand both the meteorology around basic weather variables as well as the mechanics behind futures.

Since we’re looking at global changes in climate over large swaths of time, one thing to keep our eyes out for is the ebbing and flowing of the ENSO effect. NOAA defines this effect with it’s two distinct phases-- El Niño and La Niña. These are opposite phases of a natural climate pattern across the tropical Pacific Ocean that swings back and forth every 3-7 years on average. Together, they are called ENSO, short for El Niño-Southern Oscillation. The ENSO pattern in the tropical Pacific can be in one of three states: El Niño, Neutral, or La Niña. El Niño (the warm phase) and La Niña (the cool phase) lead to significant differences from the average ocean temperatures, winds, surface pressure, and rainfall across parts of the tropical Pacific. Neutral indicates that conditions are near their long-term average. Measuring global temperature and precipitation levels can give us a general idea as to which phase we are in the midst of. During El Niño, the surface winds across the entire tropical Pacific are weaker than usual. Ocean temperatures in the central and eastern tropical Pacific Ocean are warmer than average, and rainfall is below average over Indonesia and above average over the central or eastern Pacific. Rising air motion (which is linked to storms and rainfall) increases over the central or eastern Pacific, and surface pressure there tends to be lower than average. Meanwhile, an increase in sinking air motion over Indonesia leads to higher surface pressure and dryness. El Nino has been attributed to causing larger storm systems and affecting global food production, as the FAO outlines that ag is one of the main sectors of the economy that could be severely affected by the El Niño phenomenon. The figure below gives an outline on the general effects of both climate systems over the past few decades in the United States:

Here, we can see increased-rainfall and lower-temperature anomalies across the continental south for El Nino, with the opposite effect across the Midwest and Pacific region. La Nina causes slightly different anomalistic changes, with around 1.0-1.5 degree increases in temperature differences from the average, and a massive dip from the average in precipitation levels in the same region-- some areas in the Gulf saw decreases as low as -5.0 inches. However, it needs to be said that a main underlying factor in this assessment is climate change. With our data, we’ll be able to measure deviations from the average, and we’ll be able to test if there in fact is a relationship between climate deviations and changes in (in our case, soy) commodity market valuation. El Nino and La Nina are recurring climate systems that change climate variables across different parts of the world to different degrees, and while it makes sense to keep these in mind, what we really want to keep in the forefront of our minds is the overall effects of climate change. Below, we can see the overall increase in temperature anomaly from 1950 - 2020 as generated by NASA:

Here, we can visualize our slow climb towards a 1 degree Celsius increase over the 1951-1980 global temperature baseline. While these effects have been creeping into our lives over the past few decades, we’ve seen rapid acceleration in the past few years that have the potential to create far-reaching externalities across the globe; affecting everything from sea ice levels to precipitation, from fauna migration patterns to (potentially) interrupting financial markets for necessary commodities. Now the second half of this is to quantify these changes in specific commodities per the futures market. In this particular instance, the focus is on the commodities market and it’s measure of valuation for different agricultural products. This market is run by the basic economic principles of supply and demand; low supply theoretically causes an increase in demand, which would lead to price increases until the market finds an “equilibrium”-- the continually-sought-yet-rarely-found middle point where an ideal price-point results in balanced supply and demand. This equilibrium can be disrupted by internal and external factors-- think how geopolitical events can rattle oil prices. With commodity futures, the mechanism is similar to trading stocks or options. Futures are contractual obligations to buy or sell a commodity at a given date in the future at a particular price-point. The contract is for a set amount, and it specifies when the seller will deliver the predetermined amount of the commodity in question. If the value of the established commodity goes up, the buyer of the futures contract makes money. The product was initially bought at the lower, agreed-upon price and can now sell it at today's higher market price, thus “flipping it”. If the price goes down, the futures seller makes money-- the commodity can be bought at today's lower market price and flipped to the futures buyer at the higher, agreed-upon price. Think of it like a company’s stock-- if the company becomes more “valuable”, it’s stock price rises and traders can make money selling it at $X, given that the stock was purchased for <$X. Traders can also lose money when the inverse occurs-- with this, the futures market serves as a baseline for the value of a commodity across the globe. There’s a lot more intricacies that are involved with such a process, but for the purpose of this analysis, the focus is going to be on the futures market itself and the fluctuations in price for a commodity (in this case, soybeans).

To build an analysis like this there needs to be careful consideration with the “ideal” temporal data resolution in order to get an apt comparison across both weather and commodities. Using GCM for climate data allowed for flexibility around the observation frequency-- offering anywhere from hourly to yearly observations. Futures data is collected on a smaller scale (typically with daily summaries of price lows-highs, as well as closing price and change over time). The issue with this is that a lot of “static” can be introduced from comparing such small increments of time. Another consideration is that the (futures, in particular) data source will require some imputation if there are missing/NA values, as it should be kept in mind that there will inevitably be some gaps in data that spans such a large amount of time. Since the focus is on changes in market trends versus changes in climate trends due to the large period of time being analyzed (10-20 years), I determined an ideal resolution would be monthly averages in both closing price and change over time (for futures) and monthly averages in max temperature, max precipitation, and changes in both variables over time (for climate). This way, daily spikes in futures, which can usually be written off as an outlier or caused from an external market effect, can be smoothed out into an average. The same can be said for each weather variable, and while there might be something to be said for daily changes in weather versus daily changes in futures, remember that the goal here is to link trends in order to give growers insights about the value of their outputs. 

The data used to collect precipitation and temperature data comes from ESGF, particularly the Lawrence Livermore National Laboratory portal within the Department of Energy. The climate data in question does not have a particular site as the aim was to measure global averages; however, there are some data specifics that are pertinent to understand for context. As previously mentioned, the model this data was taken from is the HadGEM2-AO (or colloquially, “Hadley”), and the European Geosciences Met Office describes this model as such: The HadGEM2 family of configurations includes atmosphere and ocean components, with and without a vertical extension to include a well-resolved stratosphere, and an Earth-System (ES) component. The HadGEM2 physical model includes improvements designed to address specific systematic errors encountered in the previous climate configuration, HadGEM1, namely Northern Hemisphere continental temperature biases and tropical sea surface temperature biases and poor variability. The mechanics behind a Global Climate Model prop up the credibility of this data source, as predicted data points are altered based on real-world observations, thus improving future calculations based off of past records and interpretations. This gives us a model that is always running, calculating, editing, and improving itself as real-world data catches up with it’s predictions. These models are typically run very far out into the future (the Hadley model had temperature data going to 2100), and since the data for this analysis is “historical”, there is confidence that the model’s forecasts for these dates has been validated by real-world inputs. 

Again, given the nature of these robust GCMs, there is limited risk of missing values or errors in particular site observations. GCMs use the equations for the conservation of momentum, energy, and mass to predict outcomes of the future state of the atmosphere, rather than a given locale. Global Climate models are a gridded data source -- GCMs are designed on comparatively coarser grids than comparable systems, however, spatial resolution has been increased in an ongoing effort to downscale GCM output. Also, the data from the model is calculated and then validated by observational data, instead of relying solely on remote or in situ observations that would be more inclined to give gaps and errors. Missteps in calculation are corrected by validated observations, thus minimizing historical errors or biases.

In order to make our findings relevant to our problem, the null hypothesis for this analysis is: Changes in climate do not have a significant impact on crop valuation as determined by futures data. Now, off the bat it should be noted that a lot can affect crop valuation; soil chemistry changes, blight, geopolitical events, labor market changes, demand changes, and logistics changes are all external economic factors that will shift our market away from its “equilibrium”. However, the aim here is to assess the trends within historical temperature and precipitation data, and compare it against soybean futures data on a smaller scale. Due to the many factors that can influence commodity valuation, there will not be an overarching correlation between changes in weather patterns by temperature and precipitation and soybean value--expecting such a relationship would be fruitless. The main data sources used for our analysis will be 20 years of historical Soybean Futures data extracted from the massive repository of CME historical market data. With this futures data, we’ll be able to assess the historical value of the soy market, and be able to see what the climate variables correspond to specific peaks and valleys in soy futures value. While it is important to note that correlation does not necessarily equal causation, the aim of this analysis is to find a statistically significant relationship (≥ 0.05) and, from there, be able to extrapolate more findings about the relationship between trends that exist within the two data sources. Remember - we need to prove the relevance of climate data as it pertains to growers and relevant stakeholders via the futures market’s valuation. It’s important to understand how the historical data for futures behaves before moving forward.

Here, we can see the gamut of soybean price futures across the past 20 years. Looking from around 2000-2010, this general upward trend is indicative of a growing global soybean market. According to a 2019 WWF report, the area of land in South America (who hosts 3 of the top 6 global soy producers) devoted to soy grew from 17 million ha in 1990 to 46 million ha in 2010, and between 2000 and 2010, 24 million ha were brought into cultivation in South America. Soybean production expanded by 20 million ha in the same period. Globally, in the last 50 years the production of soy has grown tenfold, from 27 to 269 million tons. However, by 2013 or so we can see that soy value peaked, and has been on an inconsistent decline ever since (reverting 2020 futures prices to that of mid-2006). But hasn’t global demand increased? Will it not continue into the future? Recent FAO projections suggest a production increase to 515 million tonnes by 2050; others project a 2.2 per cent increase per year until 2030. How has soy not only plateaued, but decreased in projected future value? The data from 2010-2020 points to something intriguing; let’s assess Figure 2F below to see if there have been similar changes across yearly soy value:

Not surprisingly, we see the same overall pattern: a peak in 2012-2013, a decline steepening around 2014, and settling at 2006-2007 prices in 2019-2020. While it’s important to keep the 20-year span of data in mind, it would behoove our analysis to focus particularly on 2010-2020, as this decline that bucks traditional economic principle may be caused by more than typical external factors. To dig into the declining trend, let’s try to assess where exactly it began. With a peak in 2013 and a decline bottoming-out at around 2015, let’s compare prices between 2014 and 2015 below:

Here we can see that, as expected, the decline begins in 2014 with an overall drop in valuation by $1317.8 in January to $910.63 in mid-September, followed by a decrease in slope in 2015. While the data concerning soy futures is quite granular, it may make more sense to measure monthly averages as opposed to daily prices, since we want to measure changes in overall trend. Below shows the average monthly closing price of soy futures in the timespan in question (2010-2020): 

With this, we can see that the drop-off in 2014 has had quite an impact on averages as well, and the markets have not quite recovered to its previous glory. Given this, we need to analyze our values against relevant temperature and precipitation data from the same timeframe to truly test our null hypothesis of: changes in climate do not have a significant impact on crop valuation as determined by futures data. Below we can see the global average daily maximum temperature from April of 2010 - April of 2020:

Riveting, I know. But there could be a lot of intricacy with this data, and while it makes sense that average temperatures over the course of a year will fluctuate, I think there’s more that can be uncovered if we dig deeper. How does the global average temperature trend compare to the trends we’re seeing with soy futures? What are some other areas we can dig into further? Below outlines the crossover:

While this might not look like much off the bat, let’s assess some specific crossovers-- looking after 2012, we can see peaks occurring for both data sources at the same time. This could be indicative of a strong growing season; we can also see both beginning to decline around the same timeframe (remember--futures are man-made, and their reaction to market events is not instantaneous). We also see a potential relationship between declining average temperatures for the year and the value-drop in 2014, and again appears a potential relationship between rising temperatures and increased value into 2016. While peaks-do-not-equal-peaks and valleys-do-not-equal-valleys, we can see remnants of a relationship with our data that will require more digging. If we’re going to look at temperature changes compared with market changes, wouldn’t it make sense to compare differences from the average on both metrics, as opposed to just raw values?

Now here is where we see something truly interesting -- the average temperature difference from 2010-2014 has an absolute value of 0.2656434, and the average temperature difference from 2014-2020 has an absolute value of 0.3191719 -- an increase of 20.1505%. Right after the decline in 2014, we can see that the temperature deviations from the average increased, while the value of the market was not able to recover. Looking at the graph above, we can see that the differences in monthly average loosely resembles the same trend we were seeing with regards to overall futures price. Luckily, our data source gave different metrics to measure futures, one of which being Change (%). With securities, Change represents the percent difference between the closing value on day N compared to day N-1. Calculating “Monthly Average Change” should give us a better picture of overall changes in trends, compared to the temperature deviation from monthly average (the same temperature metric measured above). The figure below gives us the following:

Notice the decreases in change during the 2014 drop-- this further pinpoints the trend we’ve been chasing from the beginning. Even with an apparent bounce-back around 2016, we know that the market has not recovered all the value that was lost after 2014. But statistically speaking, how much of that loss can be contributed to changes in weather? Do changes in temperature have a significant impact on crop valuation, as determined by futures data? To piggy-back off of a previous point, a lot can affect the futures market-- factors include underlying price, interest (dividend) income, storage costs, and convenience yield. But how much can we contribute this to changes in climate, and how can we imply such a relationship using statistics? In order to mitigate the effects of externalities, we’ll be focusing on the time period during the initial decline in 2014. To measure the relationship between temperature changes and soy valuation, we will employ the use of a Two Sample t-test. A t-test analyzes the differences between two groups, and it lets you know if those differences could have happened by chance. We will be using a statistical significance point of 0.05; if the analysis generates a  p-value less than 0.05, we can determine that, for the given timeframe, the relationship between soy futures value and temperature is statistically significant, thus rejecting the null hypothesis. Figure 9F below: 

While the figure above may not seem like it represents much, our t-test tells us a different story: in comparing the differences from monthly averages across both soybean futures and global temperature, we are given a p-value of 0.02891, which is less than our point of statistical significance and thus rejects our null hypothesis of: changes in climate do not have a significant impact on crop valuation as determined by futures data. Understanding this, the aim is to assess changes in climate, not solely temperature. To help us paint a more vivid picture, let’s pull in precipitation data from the same timeframe.

The figure above shows a rough yet similar to what we saw earlier, the loosely correlated inclines and declines give us something to dig into further. Will we see statistical significance for precipitation as well as temperature? Below may provide some insight:

Again, we are met with a similar conclusion-- while the graph above may appear nebulous, our t-test gives us a p-value of 0.0289, which again is statistically significant. This solidifies the fact that we can reject our null hypothesis, and declare that there is some correlation between changes in climate and the value of an agricultural commodity. 

While we were able to reject the null hypothesis due to statistical significance, it would be inappropriate to assume that this relationship explains every peak and valley of the soybean values that were initially assessed. However, weather data is excluded from calculations with regards to futures valuation, and to prove that there was a statistically significant relationship at a critical market point may suggest that such relationships exist elsewhere within the data, or within other commodities. Does sea surface temperature affect oil futures, since the market is so reliant on overseas shipping? Are corn futures affected by humidity, since higher humidity levels hinder a plant’s ability to absorb moisture from the soil? These are intriguing rabbit holes, considering that there’s now credibility in the fact that we can attribute climate effects to at least some changes in market value for a global commodity. What we set out to prove is that climate effects (mainly climate change, El Nino, and La Nina ) have a noticeable impact on the value of global commodity markets, and what we saw is that increased deviation from both average global temperature and precipitation levels led to a shift in the market valuation for soybeans-- one from which the market has yet to recover from.

The findings from this study fail to indicate some massive, unreported externality that has been dragging down global markets and siphoning billions of dollars of worth. What it did accomplish, however, is prove that there does in fact exist a relationship between changes in market valuation for specific commodities and changes in global averages for temperature and precipitation. In The Best American Science and Nature Writing - 2019, Sy Montgomery simply states “science is important because this is how we seek to discover the truth about the world.” The truth that we have potentially uncovered isn’t some interesting linkage between two seemingly-dichromatic data sources, but instead a dire warning about what’s to come. Let’s run with the idea that climate changes can affect commodity valuation-- where does this put us 10 years from now? 20? 30? What happens when basics like cereal grains or crude oil have been so disrupted that they either plummet in value or skyrocket due to scarcity? 

This analysis has led to the deduction recommendations to both growers and to the market as a whole-- growers need to pay attention to the world around them and understand that a changing planet could mean a shift in their place in the market. There needs to be an understanding that, even if you do not believe in the concept of climate change, there will in fact be repercussions. Similarly, the second recommendation is to those active in the futures market-- changes are coming and they can impact your wallet. The soy market has still not recovered from a massive downshift in 2014, and while that may be a specific example, we have seen in the NYSE that even hints at certain changes (in that case, coronavirus circa 03/01/2020) can scare a market into shedding billions of dollars of valuation. Traders need to be wary of this ever-changing world, and it would be wise for both growers and traders to make preparations to insulate themselves and their wallets from what can be potentially far-reaching effects due to climate change. 

Sources

Amadeo, Kimberly. “Why Do Prices of the Things You Need the Most Change Every Day?” The Balance, The Balance, 24 Apr. 2020, www.thebalance.com/commodities-futures-and-how-they-work-3305647.

Challinor, Iizumi. “Impacts of El Niño Southern Oscillation on the Global Yields of Major Crops.” Impacts of El Niño Southern Oscillation on the Global Yields of Major Crops, 17 Feb. 2020, ccafs.cgiar.org/publications/impacts-el-ni%C3%B1o-southern-oscillation-global-yields-major-crops#.XqYeRJNKjBJ.

“El Nino, ENSO, and Climate Change.” Food and Agriculture Organization of the United Nations, www.fao.org/el-nino-enso-and-climate-change/hyxsdtaDilY_vDO3c/en

“El Niño, La Niña and Climate Impacts on Agriculture: Southeastern U.S.” Southeast Climate Extension, agroclimate.org/wp-content/uploads/2016/03/ENSO-Impacts-southeast.pdf.

“The Growth of Soy - Impacts and Solutions.” Wwf.org, World Wildlife Federation, The Growth of Soy Impacts and Solutions.

Karuga, James. “10 Countries With Largest Soybean Production.” WorldAtlas, WorldAtlas, 10 Mar. 2016, www.worldatlas.com/articles/world-leaders-in-soya-soybean-production-by-country.html.

Marini, Annalisa. “The Impact of Weather on Commodity Prices: A Warning for the Future, by Annalisa Marini.” Discussion Papers, University of Exeter, Department of Economics, 1 Jan. 1970, ideas.repec.org/p/exe/wpaper/1902.html.

Martin, G.M., et al. “The HadGEM2 Family of Met Office Unified Model Climate Configurations.” Https://Www.geosci-Model-Dev.net/, Met Office, 7 Sept. 2011, www.geosci-model-dev.net/4/723/2011/gmd-4-723-2011.pdf.

McKeef, Clive. “Extreme Weather Seen Impacting Oil Prices in 2020 as Well as Geopolitics.” MarketWatch, MarketWatch, 11 Dec. 2019, www.marketwatch.com/story/extreme-weather-seen-impacting-oil-prices-in-2020-as-well-as-geopolitics-2019-12-11.

“Overview: Weather, Global Warming and Climate Change.” NASA, NASA, 28 Aug. 2019, climate.nasa.gov/resources/global-warming-vs-climate-change/.

“US Soybeans Futures Historical Prices.” Investing.com, www.investing.com/commodities/us-soybeans-historical-data.

Zuba, Gerhard. “A Look Back at the 2014 Price Drop and What's Ahead for the 2015 Crop Season.” AIRWorldwide, www.air-worldwide.com/publications/air-currents/2015/a-look-back-at-the-2014-price-drop-and-whats-ahead-for-the-2015-crop-season/.