Death to the Data: A Data Visualization of Unique Death Indicators

Alex Hugon ([email protected])

Jason Shah ([email protected])
Sergius Wat ([email protected])
Introduction
Death rates are often compared with obvious indicators such as violence, crime, and disease
outbreaks. We wanted to visualize death rates compared to unique indicators: urban proportion
of population (referred to as ‘urban proportion’), GDP, and Internet usage. Perhaps as more
people have moved into the cities, death rates of spiked, or as GDP sank, death rates rose.
Indeed, perhaps Internet usage is related to an increase in death rates. We do not claim causation
or even correlation; however, observing the relationship between death rates and each of these
three factors could be informative in a manner that previous attempts at visualizing death rate
data has failed and surely entertaining. And after all, as we have learned, keeping the audience
engaged is fundamental to the success of a visualization.

Users can perform a variety of tasks with our visualization. The most central task is observing
the relationship over time between death rates and our three indicators: urban proportion, GDP,
and Internet usage. Our users can also compare death rates, urban proportion, GDP, and Internet
usage between countries across different years. Moreover, for greater detail on a given country, a
user can browse the ten-year context data for the selected data pairing (e.g. death rates and urban
proportion) to observe a given year’s data relative to five years prior and five years after.

Our users include those interested in global demographics. We believe that students of
international development, members of the World Bank and World Health Organization, and
political leaders of various countries would find this visualization thought provoking. Given our
focus on this audience, we have used high quality and standardized data from the World Bank.
Moreover, our inclusion of ten-year context is informed by our awareness of how important
context is to international development work.

The data includes death rates, urban proportion of population, gross domestic product (GDP),
and Internet usage for more than 50 countries since 1965. We obtained the death rate data
through Java scraping, and the latter three data sets through browsing the World Bank online
database.

Related Work
Within the domain of death, GE developed a visualization to highlight the actual causes of death
(unlike our three indicators) ranging from suicide to cancer, by age and race
(http://visualization.geblogs.com/visualization/causes/).

From a different angle, this Time Magazine visualization shows the volume of cigarette users in
different countries and the correlating medical conditions
(http://www.time.com/time/interactive/0,31813,1911060,00.html). Unlike our visualization, this
one chose not to choose a map for a canvas and focused on a very specific contributor (i.e.
smoking cigarettes) to medical reasons for death.

In 2010, AllFacebook.com developed a simple visualization that compared global birth and death
rates to Facebook usage and deaths. For example, as of September 2010, 300 million people have
died since Facebook’s launch, and 200,000 of Facebook’s members die each year
(http://www.allfacebook.com/death-on-facebook-2010-09).
With respect to our three unique indicators, there are many separate visualizations of urban
proportion, GDP, and Internet usage. For example, in 2009 Google began publishing simple
standalone visualizations of various World Bank indicators, including GDP and Internet usage
(http://googleblog.blogspot.com/2009/11/world-bank-public-data-now-in-search.html).

Also this visualization shows a more detailed look at Internet usage, breaking out ‘Bitnet’ and
‘Email only’ (http://www.cs.ubc.ca/~tmm/courses/cpsc533c-04-spr/a1/mtan/eg.htm).

Approach
Our visualization design uses a map canvas to display death rate data by country over time as
compared to different indicators, such as urban proportion, GDP, and Internet usage. We decided
to use a map in order to allow the audience to most easily browse between countries based on
known locations and proximity. Moreover, it enables an integrated view of regions and not just a
view of individual countries. On the map, there is size and color encoding for each country and
the bubble that represents its death data. The size of the bubble represents the proportion of that
indicator (i.e. urban proportion, GDP, Internet usage) relative to other countries. For example, a
roughly medium sized bubble when a user has selected ‘urban proportion’ represents about 50%
of the population living in the cities, whereas a much smaller bubble represents a much smaller,
say 22% for the American Samoa, urban proportion. On a related note, the color of the bubble,
and its relative hue (e.g. very dark) represent the death rate. A very light, perhaps nearly white,
color represents a low death rate, close to 0 per 1000 people, whereas a very dark bubble
represents a higher death rate, closer to 32 deaths per 1000 people. We overlay data on the map
as users interact with the bubbles that represent different countries. For example, clicking on the
bubble that represents the American Samoa when on the urban proportion/death rate
visualization will display a semi-transparent information box in the upper right corner of the
visualization providing the following information: “Samoa, 22% of population is urban, and 6
deaths per 1000”, which is dynamically updated each time a user clicks on a different bubble
representing a different country.

Naturally we allow users to scroll through and zoom in on the map either through direct
manipulation using their touchpad/mouse, or through directional arrows overlaid on our own
map.

We decided to accompany our primary map with a control panel that enables the user to
manipulate the data visualization and view more detailed information about a given country
during a given year. We allow uses to slide a simple bar to change the year; users can also click
on a section of the bar to jump to that year instead of forcibly sliding the year-control bar.
Furthermore, users can toggle between playing and pausing the loop in order to either passively
view trends over time or more actively stop the year, view data, and control the next year view.
We then provide a legend explaining how the darkness of the bubbles on the map represent a
higher death rate. Below this, the options for toggling between the three different visualizations
are providing. We made sure to match the hover-over color for these three options to the actual
color scheme of the bubbles on the maps for these visualizations.

As mentioned, users can interact with the visualization with hovering and clicking. In our control
panel, we display drill-down information about countries by showing 10-year context for a given
country when clicked. This data provides 10-year context to the death rate and presently selected
indicator of the three indicators. The line graph here complements the global map that represents
the primary component of the visualization.

Data
Our data is entirely sourced from the official World Bank data repository called the ‘World
Development Indicators’ (WDI). Therefore, we downloaded original datasets on death rates,
urban proportion, gross domestic product (GDP), and Internet usage. This data, and data for
many other indicators, is available here: http://data.worldbank.org/indicator. Indeed, we selected
our indicators from the 298 featured indicators, although there are more than 1000 indicators
with data available through this source.

Death rate represents how many people from this country died for every 1000 people. We chose
this metric instead of total deaths. By doing so, we avoided the inherent skew that would come
with different population sizes. This essentially normalized the data. Also, we chose to use
overall death rate as opposed to the very common indicator of mortality rates for children under
five. The latter indicator is often used in development research because the survival of children
and ability to make it past the age of five is an indicator of progress and basic ability to survive
in a country. However, given that we were comparing death against indicators such as Internet
usage, it was unlikely the under-five mortality rates would be especially relevant. Therefore, our
death rate indicator helps control for variation in population size and is more relevant for our
other three indicators.

Implementation
We used Ben Fry’s Processing language for our full implementation. In order to import the tab-
delimited data into Processing we used Ben Fry’s table.pde file. Furthermore, we used the
Modest Maps API to enable functionality on our map.

Results
We experienced a couple of important pivots during our visualization development. Originally
we anticipated showing the relationship between democracy and mortality. However, this idea
relied on using a relatively unchanging democracy number that also had little meaning to those
not subscribing to this single method of ranking democracy. As a result, one could hardly
observe any trends over time. Moreover, for those who did not agree with the 1-10 democracy
ranking that we were building on top of, the visualization had no credibility.

Once we changed to focus on death rates as compared to these three key indicators, we found
much more success. Here we describe three different scenarios in which our visualization is
useful.

SCENARIO #1: Passively view an indicator to see its change over time and relationship
with death rate
Firstly, imagine one wants to view both the change in Internet usage over time and particularly
the relationship between Internet usage and death rates from a macro or global perspective.
Below are two screenshots from simply loading the visualization and allowing it to play the loop.
One can notice two key findings related to the stated objectives. Firstly, comparing 1996 to 2005,
one can readily observe the growth in Internet usage. Admittedly this should be obvious even
without the visualization; however, this visualization helps show in which countries and regions
usage was most population. Apparently North America, Western Europe, and key parts of East
Asia showed strong Internet usage by 2005 as opposed to a region such as Africa.

However, it is also especially interesting to use the size and color encoding that the bubbles
provide for each country. Specifically, one may observe how some bubbles are especially dark,
representing a relatively higher death rate. Each of these countries, which seem to be in Western
Europe, are about medium sized bubbles, meaning relatively moderate Internet usage relative to
other countries displayed on this map. Indeed, both smaller and larger bubbles, or countries in
which there was less and more Internet usage, are lighter hues, which means that it is hard to
form some hard conclusion about the relationship between Internet usage and death rates.
However, this visualization makes that conclusion readily available through a simple, passive
view of the data.
Internet Usage in 1996 by country

Internet Usage in 2005 by country

SCENARIO #2: Comparing Two Countries’ GDP and Death Rates

Another common task would be the desire to compare two different countries’ GDP values and
death rates. In this instance we loaded the ‘GDP/Death Rate’ visualization and hovered over
different countries to make a sense of comparison beyond viewing the relative size and color
differences on the main overview. Here we observe virtually identical visualizations except that
the first provides greater detail for the U.S., whereas in the second visualization, the mouse is
then hovering over the bubble that represented Germany, so the additional detail provided in the
upper right corner and in the right hand panel differ to allow one to see what changes by
browsing different countries.

GDP and Death Rates in 2000: Detailed View for US

GDP and Death Rates in 2000: Detailed View for Germany

Upon hovering for further detail, the user can instantly and easily observe how the US GDP is
9900 billion USD, whereas the German GDP is just 1900 billion USD. Yet the death rate is
identical. While this is just one data point, it suggests that even a lower GDP, perhaps above a
certain threshold, doesn’t necessarily relate to a higher death rate. One could browse to more
impoverished countries, browsing their data to see if lower GDP does in fact share a relationship
with death rate.

SCENARIO #3: Getting a Detailed View of One Country’s Urban Proportion and Death Rates
with 10 Year Context

Consider that a given country may experience significant urbanization, and one may want to
know what happened to the death rate in the same period of time. Take Bangladesh for example.
In our visualization, one need only to browse to the urban proportion visualization and view the
right hand panel with details to better understand Bangladesh, its urbanization, and the
simultaneous changes in death rates.

Urban Proportion and Death Rates, Focused on Bangladesh in 1984

As we see in the detailed view below, Bangladesh’s proportion of its population living in the
urban parts of the country steadily rose from 1979 through 1989, the five years before and after
1984. Simultaneously the death rate steadily declined as urbanization increased.
Detailed View of Bangladesh, Its Urban Proportion, and Death Rates

If one wanted to zoom back out to the broader visualization, countries with lighter red hues were
often experiencing lower death rates, and often with higher urbanization figures. The case of
Bangladesh just illustrates how one may drill down on a given country when suspecting a trend.

Discussion
We learned a lot throughout this process. Some strengths of our work include a map that reflects
the countries that were included in our data set. Being able to navigate the map rather than an
overly simplistic and relatively hard to use drop down menu offers key advantages for
performance speed and ease of use. Further, the ability to hover to receive additional information
makes it extremely easy for a user to get just the info that they want, on-demand, and without
much significant effort beyond what the user would otherwise do normally. Additionally, the
ability to simply toggle between any of the three visualizations and view an instant update
without needing to swap to a different visualization entirely allows for more flexible review of
the data.

There are also some key weaknesses. In the US, the GDP is so large it overshadows other North
American countries, and it is hard to browse to these other countries. When tabbing between the
three different visualizations, it may have helped if the selected box remained checked rather
than the view toggling back to unchecked. For the 10-year context, we could have considered
allowing a user to scroll through the different years context rather than showing a relatively static
view. On a more fundamental level, our peer evaluation suggested not using a map for our
canvas as it may be inferior to a more abstract canvas and does not appear to deliver any
immediately unique geographic information. It is possible that a different canvas would have
helped the visualization. We made this design decision in order to view regional differences and
find given countries more easily than one could do in alternative visualization canvases. This
decision appears to be useful for these reasons, but conceivably some other canvas may have
provided better information somehow.

Perhaps the visualization should not play by default; for some users this is distracting, but for
others, it is helpful in seeing trends immediately and knowing how to use the control panel.
We could have also made the canvas one and the same; one should not be able to move off of the
map, whereas with some bizarre navigation, a user can make it to our blue background that is off
the map.

References
“Causes of Death”. GE. <http://visualization.geblogs.com/visualization/causes/>.

“Example of a GOOD Visualization”. University of British Columbia.

<http://www.cs.ubc.ca/~tmm/courses/cpsc533c-04-spr/a1/mtan/eg.htm>.

“Leading Causes of Death”. Time Magazine.

<http://www.time.com/time/interactive/0,31813,1911060,00.html>.

“Surprising Facts About Death on Facebook [Infographic]”. AllFacebook.com.

<http://www.allfacebook.com/death-on-facebook-2010-09>.

“World Bank public data, now in search”. Google.

<http://googleblog.blogspot.com/2009/11/world-bank-public-data-now-in-search.html>.