the 2017 watson analytics global competition investigating ... · the 2017 watson analytics global...
TRANSCRIPT
The 2017 Watson Analytics Global Competition
Investigating the Relationship between Financial Indices and Ecological Factors
Joseph Adamski
Suhail Pathan
Rudy O’Neil
Dr. Michael Gendron
Faculty Sponsor
Central Connecticut State University
New Britain, Connecticut
March 15, 2017
Financial Indices & Ecological Factors 2
Contents Table of Figures ............................................................................................................................................ 2
Introduction ................................................................................................................................................... 3
Literature Review .......................................................................................................................................... 3
Methodology ................................................................................................................................................. 4
Hypotheses............................................................................................................................... 4
Data Cleanup and Acquisition .................................................................................................... 4
Research ........................................................................................................................................................ 5
Descriptive Analytics ................................................................................................................ 6
Temperature ........................................................................................................................11
Carbon Dioxide ....................................................................................................................13
Correlative Analytics ................................................................................................................14
Limitations .................................................................................................................................................. 19
Discussion ................................................................................................................................................... 20
References ................................................................................................................................................... 20
Data Sources ............................................................................................................................................... 21
Table of Figures
Figure 1: Deforested Land over Year ........................................................................................................... 6
Figure 2: Sea Level over Year ...................................................................................................................... 6
Figure 3: Temperature over Year .................................................................................................................. 7
Figure 4: Mexican IPC and Deforested Land over Year ............................................................................... 8
Figure 5: American Dow Jones and Deforested Land over Year ................................................................. 8
Figure 6: Brazilian iBovespa and Sea Level over Year ................................................................................ 9
Figure 7: American Dow Jones at Sea Level over Year ............................................................................... 9
Figure 8: Australian S&P/ASX 200 and Sea Level over Year ..................................................................... 9
Figure 9: Canadian S&P/TSX Toronto and Sea Level over Year .............................................................. 10
Figure 10: German DAX and Sea Level over Year .................................................................................... 10
Figure 11: Mexican IPC and Sea Level over Year ..................................................................................... 11
Figure 12: Mexican IPC and Temperature over Year ................................................................................. 12
Figure 13: Canadian S&P/TSX Toronto and Temperature over Year ........................................................ 12
Figure 14: Hong Kongese Hang Seng and Temperature over Year ............................................................ 12
Figure 15: American Dow Jones and CO2 over Year ................................................................................. 13
Figure 16: German DAX and CO2 over Year ............................................................................................ 13
Figure 17: American Dow Jones Predictor Dashboard Panel ..................................................................... 14
Figure 18: Chinese Shanghai Predictor Dashboard Panel........................................................................... 15
Figure 19: Canadian S&P/TSX Toronto Predictor Dashboard Panel ......................................................... 16
Figure 20: European Stoxx 50 Predictor Dashboard Panel ......................................................................... 17
Figure 21: Brazilian iBovespa Predictor Dashboard Panel ......................................................................... 18
Figure 22: Mexican IPC Predictor Dashboard Panel .................................................................................. 19
Financial Indices & Ecological Factors 3
Introduction
4,621 square miles of Brazilian rain forest were cut down in 2008 with hopes of bringing
great local economic benefits to the population (Walsh, 2009). However, according to Rodrigues
et al. (2009), this is only a short term view and deforestation brings a short term boom with long
term detriment to the local economy. Building upon this academic research, our group sought to
expand existing research by applying similar methodology to global economies and
incorporating environmental indicators. The environmental indicators chosen were CO2 levels,
sea levels, temperature and deforested land as they could be applied at the global level. The aim
of this study is to determine if there is a relationship between environmental indicators and
global financial indices.
Financial indices were chosen based on the index being representative of a country’s
market. A domestic example might be the performance of the Dow Jones Industrial Average
which tracks 30 large corporations from a wide variety of sectors. However, Walsh (2009) points
out that the effects of deforestation are delayed upon the local economy. Hence, our group will
have to keep in mind that effects on these financial indices may not be immediate. Furthermore,
due to the global scale of indicators utilized, results may be skewed by the whole host of other
factors that influence the market and environmental health.
Starting with 1993, the relationship between the chosen environmental indicators and the
health of various global exchanges will be determined. IBM Watson will be utilized in order to
perform descriptive and correlative analytics. Descriptive tools of Watson will be used in order
to visualize the data and identify trends. Based on descriptive results, we can employ Watson’s
correlative tools to understand these indicators’ correlative strength. By creating correlative
models, this research could be useful to a variety of policy makers in many different countries
and organizations.
Literature Review
In a study by Geist and Lambin (2002), the driving factors of tropical deforestation were
explored. Their analysis revealed that the local decisions to deforest were driven by national and
global level economic opportunities and policies. No consistent link was shown between
deforestation and local issues such as impoverishment and political ecology. Additionally,
political corruptions or poor implementation of forestry rules can allow non-locals to come in to
deforest and commercialize the area. If decisions to deforest are driven by global factors, then
how does deforestation actually affect the global economy after it occurs? This is the substance
of the analysis of our study being done for the 2017 Watson competition.
In recent years, starting from 2004 to 2013, the world has seen a significant decline in the
area of rainforest that has been deforested. The decline from 2004 is due to the Brazilian
government putting regulations on the deforesting large properties (Godar, 2014). However,
Financial Indices & Ecological Factors 4
Godar believes that these measures have done as much as they could and deforestation will not
decline further without additional legislation.
Campari (2002) took a look at what happened in areas of deforestation both in local
economic impact and socially. Campari found that if areas are deforested and turn out to be more
useful for farming then deforestation accelerates quickly in that area. Natives also choose to stay
based on productivity of the land, as they have incentive if the new value of the land is less than
how much they can make by farming the land. However, Campari also found that those who
stayed had poor health care, education and poor housing. In reducing deforestation, Campari
suggested creating better infrastructure such as better transportation for those already in living in
deforested areas in order to create alternative methods of improving quality of life other than
deforestation.
According to Pohkrel (2012), deforestation contributes to a rising sea level by increasing
the amount of inland water flow to the oceans. Previously, streams may have been blocked by
vegetation. This not only affects the oceans but has the potential to flood areas that were
previously inhabited by wildlife. With wildlife and usable land reduced, local economies can be
negatively impacted especially those that rely on farming as an income. This is one way
deforestation may indirectly damage economic growth.
There is a lack of literature involving deforestation and how it is related to global economic
indicators. This project will help fill the gap in literature and assist future studies in identifying
relationships.
Methodology
This project focuses on the utilization of IBM Watson in order to analyze and forecast
financial indices based on environmental indicators. Data cleansing was done using Microsoft
Excel before being uploaded to Watson. IBM Watson allows for analysis, visualization and
prediction utilizing multiple data sets.
Hypothesis
If environmental indicators are negatively affected (i.e., global temperature, carbon
levels, deforestation rates and sea level will change in a negative manner), then financial indices
will also decline.
Data Cleanup and Acquisition
Deforestation data was obtained through The World Bank (2015), an organization
dedicated to providing assistance to developing nations via research and analysis. Financial
Financial Indices & Ecological Factors 5
index data was obtained through Yahoo Finance’s historical data tab. Carbon dioxide was
obtained through readings taken by NASA (NASA, 2016) and adjusted for seasonal corrections.
Temperature data was also from NASA (2017, Global surface temperature) which involved
gathering temperature readings around the globe and calculating a global surface temperature
change. Lastly, Sea level data was also obtained through NASA (2017, Global sea level change)
which was measured through instruments stationed throughout the world.
The financial indicator data was transformed from a monthly dataset into a yearly dataset
by taking the market’s opening value on the first day of the following year in which the
appropriate market was open. For example, the opening value of the first trading day in 2002 is
2001’s final. Deforestation, carbon dioxide and sea level datasets were refined from raw values
into a dataset which describes the cumulative yearly change since 1993. All data sets had to be
cleaned of erroneous values and blank data cells. Upon completion of cleaning, data was
uploaded to IBM Watson and received a quality score of 96%.
Research
In order to understand and analyze the data, our group conducted descriptive analysis and
attempted to create meaningful correlative models using Watson’s analytical tools. Watson has a
powerful and dynamic engine which assists in selecting the proper visualizations to help describe
the data. Additionally, Watson allows for quick generation of in-depth correlative models.
Although many prominent indices were tested, only the ones with the most significance are
mentioned.
Financial Indices & Ecological Factors 6
Descriptive Analytics
Figure 1: Deforested Land over Year
Visualized above is the square kilometers that have been deforested in the world by year.
Although the rate of deforestation per year has seen a significant decrease since 2000, the total
area deforested continues to rise.
Figure 2: Sea Level over Year
The graph above shows the yearly change in sea level by millimeter from the year 1993.
In terms of trends, it can be seen that sea level has been rising. However, there are a few years,
1996 and 2011, where the sea level receded a few millimeters.
Financial Indices & Ecological Factors 7
Figure 3: Temperature over Year
Similar to sea level, temperature is also trending upwards as time passes. Out of 20 years,
16 of them saw an increase from the previous year. As seen in the figure above, the rate of
change between years is accelerating as time goes on. For example, the change in temperature
during 2012 was the highest change seen in the dataset.
Deforestation
Financial Indices & Ecological Factors 8
Figure 4: Mexican IPC and Deforested Land over Year
Figure 5: American Dow Jones and Deforested Land over Year
Figures 4 and 5 compare deforestation levels and its potential impact on financial
indicators. Interestingly, in the year 2000, when the market indices contract, deforestation levels
decrease in yearly change for the first time. However, for the most part, excluding 2008,
deforestation levels grow alongside market indicators. Based on these descriptive graphs, it
would seem that a trend somewhat exists between these financial indices and deforestation
levels.
Sea Level
Financial Indices & Ecological Factors 9
Figure 6: Brazilian iBovespa and Sea Level over Year
Figure 7: American Dow Jones at Sea Level over Year
Figure 8: Australian S&P/ASX 200 and Sea Level over Year
Financial Indices & Ecological Factors 10
Figure 9: Canadian S&P/TSX Toronto and Sea Level over Year
Figure 10: German DAX and Sea Level over Year
Financial Indices & Ecological Factors 11
Figure 11: Mexican IPC and Sea Level over Year
The above figures display sea level against a selection of financial indices. Interestingly,
all the economies aside from Brazil and Mexico, have sea level generally trending with financial
indices in a positively correlated fashion from 2002 to 2013. The trend becomes noticeably
stronger as the years go on, most notably from 2008 to 2013. As sea levels grow, the financial
indices tend to rise with it and vice versa. In the cases of Brazil and Mexico, they follow the
same pattern as the other economies except in the year 2013. In light of these trends, we have
decided to investigate whether the correlative relationship between market indicators and sea
level exist.
Temperature
Financial Indices & Ecological Factors 12
Figure 12: Mexican IPC and Temperature over Year
Figure 13: Canadian S&P/TSX Toronto and Temperature over Year
Figure 14: Hong Kongese Hang Seng and Temperature over Year
Figures 12 through 14 layer temperature levels over financial indices. The graphs exhibit
an almost erratic trend where some years consists of an inverse relationship while other years
contain a positive relationship. In the years from 1998 to 2004, the Canadian and Hong Kongese
indices have a contradicting experience in relation to temperature. As temperature readings go
down, the Canadian and Hong Kongese indices go up and vice versa. However, when observing
2006 to 2013, a positive relationship is displayed in which as temperature rises, financial indices
also grow and vice versa. In Mexico’s case, the same positive trend that exists for Canada and
Financial Indices & Ecological Factors 13
Hong Kong also appears except for the year 2013. In years before 2002, Mexico’s IPC did not
change much on a yearly basis and thus could not be used to draw conclusions for temperature.
For other financial indices, temperature did not seem to follow any solid trend.
Carbon Dioxide
Figure 15: American Dow Jones and CO2 over Year
Figure 16: German DAX and CO2 over Year
Figures 15 and 16 serve to describe carbon dioxide’s relationship with financial
indicators. Carbon dioxide does not decrease in total quantity. Despite fluctuations within
Financial Indices & Ecological Factors 14
various financial indices, carbon dioxide grows in an almost linear fashion nonetheless. Similar
to the deforestation figures, these graphs indicate a lack of trend between carbon and the market.
Correlative Analytics
Figure 17: American Dow Jones Predictor Dashboard Panel
Using IBM Watson’s correlative analytical features, we were able to understand levels of
correlative accuracy of most of the market indices that we examined. The American Dow Jones
is positively correlated with deforestation, CO2, sea level and temperature. The index has a
correlative strength of 77% for deforestation, CO2 and sea level which is fairly strong. As the
levels of deforestation, CO2 and sea level increase, so does the index. However, temperature is
regarded as below average, having a correlative strength of only 44%. Low strength for
temperature’s correlation is most likely due to its fluctuations being more drastic than the other
variables.
Financial Indices & Ecological Factors 15
Figure 18: Chinese Shanghai Predictor Dashboard Panel
We decided to include the Chinese Shanghai Composite analysis to serve as a
counterexample to the American Dow Jones index. Despite being the world’s second largest
economy, it does not have as high of a correlative strength since it sits at a value of 68%. This
score is in stark contrast to the smaller economies in the coming figures with their correlative
strength being the strongest out of all the indices tested.
Financial Indices & Ecological Factors 16
Figure 19: Canadian S&P/TSX Toronto Predictor Dashboard Panel
Our group had hypothesized that the Canadian S&P/TSX might not correlate well with
the ecological factors since it did not recover as well from the 2008 crisis in comparison to other
economies. However, at a correlative strength of 83%, deforestation, CO2 and sea level
correlations are stronger than most other indices. Temperature only has a slightly above average
correlative score of 59%. This relationship reflects that deforestation, CO2 and sea levels rise
with Canada’s S&P/TSX Toronto.
Financial Indices & Ecological Factors 17
Figure 20: European Stoxx 50 Predictor Dashboard Panel
IBM Watson’s correlative modeling for Europe’s Stoxx 50 is quite similar to the
American Dow Jones. Compared to the United States, the correlative strength for sea level, CO2
and deforestation only varied by 1%, for a strength of 74%. However, temperature did not show
any correlation with the valuation of the European Stoxx 50. As with the other indices, the
relationship is a positive correlation with both independent and dependent variables increasing
with one another.
Financial Indices & Ecological Factors 18
Figure 21: Brazilian iBovespa Predictor Dashboard Panel
The Brazilian iBovespa is of particular interest due to the vast amounts of deforestation
affecting the Brazilian rainforests. This index shows a very strong correlation with deforested
land, CO2 and sea level with a correlative strength of 89%. Similar to other index analyses,
temperature proved to have an average correlative strength with a value of 52%. Hence, it seems
that as deforestation levels, carbon levels, and sea levels rise, iBovespa’s valuation does the
same.
Financial Indices & Ecological Factors 19
Figure 22: Mexican IPC Predictor Dashboard Panel
Deforestation, CO2 and sea level is more correlated with Mexico’s IPC than any other
index with a 96% correlative strength. This relationship was of great surprise as we had
hypothesized that Mexico would have an above average correlative strength. Nevertheless,
deforestation, CO2 and sea levels rise alongside the IPC’s value, albeit much stronger when
compared to other indices.
Limitations
Time span was the biggest limitation by far in validity of our results. Historical data for
indices could only be retrieved from 1993 due to some only being recently created. Without
additional data, there may be time lagged effects not shown in the data due to our 20 year time
span simply not being large enough. Furthermore, the data had to be transformed into yearly time
spans due to deforestation data only being recorded on a yearly basis.
Another limitation is that this research could not develop a method to reconcile the
differences between developed and developing economies. For instance, it could be entirely
possible that these results are skewed by the fact that Mexico is a developing nation with an
economy that is still coming into its own. More research should be done to ensure the correlation
is valid. Yearly changes in economic value are most likely more volatile in developing
economies as compared to developed.
Financial Indices & Ecological Factors 20
Upon testing, deforestation, sea level and carbon dioxide had similar correlative strengths
in relation to the indices. Temperature was the only variable that produced differing results.
Hence, in future studies, better variables could be selected for ecological factors in order to better
represent ecological changes.
Discussion
This research sought to identify correlations between ecological factors and global
economic performance. In order to accomplish this, factors were identified such as deforestation
and carbon dioxide to be compared against financial indices. Based on the conducted research, a
trend appears where economic performance increases along with deforestation, carbon dioxide
and sea level. Temperature showed weak to no correlation with the financial indices. However,
based on the research’s time-span limitation, there may be missed long term correlations between
ecological factors and indices. Further research should be done about the long term effects of
environmental change. Future research might consider only utilizing one of these such as
deforestation.
Potential beneficiaries of this research, such as non-profits, policy makers and investors,
can utilize this information to justify and make environmentally-sound decisions. Businesses can
also employ this research to make decisions to sustainably attain their profits. Hence, the
research can assist these decision makers in understanding the economic impact of their
decisions, at least in the short term.
References
Geist, H. J., & Lambin, E. F. (2002). Proximate Causes and Underlying Driving Forces of
Tropical Deforestation Tropical forests are disappearing as the result of many pressures, both
local and regional, acting in various combinations in different geographical locations.
BioScience, 52(2), 143-150.
Rodrigues, A. S., Ewers, R. M., Parry, L., Souza, C., Veríssimo, A., & Balmford, A.
(2009). Boom-and-bust development patterns across the Amazon deforestation frontier. science,
324(5933), 1435-1437.
Walsh, B. (2009). Study: Economic Boost of Deforestation Is Short-Lived. Retrieved
December 27, 2016, from http://content.time.com/time/health/article/0,8599,1904174,00.html
GSFC. 2015. Global Mean Sea Level Trend from Integrated Multi-Mission Ocean
Altimeters TOPEX/Poseidon Jason-1 and OSTM/Jason-2 Version 3. Ver. 3. PO.DAAC, CA,
USA. Dataset accessed [2017-02-26] at http://dx.doi.org/10.5067/GMSLM-TJ123.
Pokhrel, Y. N., Hanasaki, N., Yeh, P. J., Yamada, T. J., Kanae, S., & Oki, T. (2012).
Model estimates of sea-level change due to anthropogenic impacts on terrestrial water storage.
Nature Geoscience, 5(6), 389-392.
Financial Indices & Ecological Factors 21
Godar, J., Gardner, T. A., Tizado, E. J., & Pacheco, P. (2014). Actor-specific
contributions to the deforestation slowdown in the Brazilian Amazon. Proceedings of the
National Academy of Sciences, 111(43), 15591-15596.
Data Sources
World Bank. (2015). Forest area (sq. km). Retrieved March 17, 2017, from
http://data.worldbank.org/indicator/AG.LND.FRST.K2?end=2013&name_desc=false&start=199
3
NASA. (2017). Global surface temperature | NASA Global Climate Change. Retrieved
March 17, 2017, from https://climate.nasa.gov/vital-signs/global-temperature/
NASA. (2017). Global sea level change Climate
Change.https://sealevel.nasa.gov/Retrieved March 17, 2015, from https://sealevel.nasa.gov/
N. (2016, December). Vital Signs. Retrieved February 04, 2017, from
http://climate.nasa.gov/vital-signs/carbon-dioxide/
Prodes. (2016). Retrieved February 04, 2017, from
http://data.globalforestwatch.org/datasets/4160f715e12d46a98c989bdbe7e5f4d6_1
Yahoo Finance (n.d.). Retrieved January 30, 2017, from http://finance.yahoo.com/