Visual thinking and Visualization Report

Robert Passmore & Adam Sundström

 

Topic: The effective global mapping of the correlation between income inequality and crime rate.

The problem 

Crime and forms of inequality are two of the biggest issues in South Africa. From the research we conducted, we found that poverty and income inequality are powerful predictors of homicide and violent crime. The problem is that there do not exist visualisations that display this relationship effectively for use by the layman. Through creating a visualization which shows this correlation between violent crime and income inequality, we hope to allow for a new way for people to perceive and address these social issues. 

 

Data we used in our visualization are measures of income inequality, crime, population and Gross Domestic Product (GDP). Income inequality will be measured using the Gini coefficient, which is a measure of statistical dispersion intended to represent the income distribution of a nation’s residents, and is the most commonly used measure of inequality (the term ‘Gini’ comes from the name of the developer of the measure, an Italian statistician named Corrado Gini). Crime rate will be measured using homicides per 100000 people. Population is measured by the total number of people in the given country. GDP per capita data will be in current $USD.

 

Our visualization acts to illustrate a social issue present around the world, and is aimed at users who show an interest in such social issues but who do not normally put in effort into their search for data. This visualisation makes the data easy to access, interact with and query.

 

 

Literature Review:

Fajnzlber, Pablo, Daniel Lederman, and Norman Loayza. "Inequality and violent crime." JL & Econ. 45 (2002): 4.

Fajnzlber, Pablo, Daniel Lederman, and Norman Loayza. "Inequality and violent crime." JL & Econ. 45 (2002): 5.

 

These visualizations make use of scatterplots to show the Gini coefficient versus robbery rates and homicide rates for different areas. It succeeds in showing a correlation between increasing inequality and increasing crime, and makes use of various different symbols for each area, allowing the user to look for clusters of certain symbols. However, one cannot see the relationship between income distribution and crime for each geographic location easily, as all we get is a symbol corresponding to a large area, rather than a specific country. Since the symbols used are also only black or grey, they are not as easily distinguishable as they would be if they differed on two channels like shape and colour.

It is worth noting that the graphs were also each labelled incorrectly, which brings in the question of whether the title of each graph is incorrect, or if the Y-axes of the graphs were incorrectly labelled.

 

Segall, Jordan.. "Inequality in the United States." (2010): 3.

 

This visualization is another scatterplot graph, which shows the income inequality versus homicide rates in American States. As with the other scatterplot, it succeeds in showing the visualization, while making it difficult to identify and compare individual locations. This visualization has labelled each point with the name of the corresponding state, making users able to compare individual states unlike the previous example, but still requires users to perform a search through all of the points to find the names of locations which they wish to compare. The visualization could be improved with interactivity like a search feature.

 

Daly M, Wilson M, Vasdev S. “Income Inequality and Homicide Rates in Canada and the United States.” Canadian Journal of Criminology 2001: 43: 219-36

 

 

This scatterplot comparing homicide rates vs income equality for USA states and Canadian provinces doesn’t allow for querying individual states, but does allow for easy comparisons between two countries very effectively. USA and Canada’s symbols differ on two channels, being a different shape as well as a different colour, allowing for very significant contrast and ability to notice a pattern between the two.

Visual comparisons between income inequality and crime are being shown through scatterplot graphs in an attempt to show that there is a correlation between crime and inequality. Our visualization aims to display this same correlation with a focus on geographic location, allowing users to clearly see the occurrence of this correlation intuitively.

 

Important visual queries:

• Do countries with high crime rates have high income inequality?

• Does this correlation between crime rate and income inequality occur consistently between different countries/continents?

• How does one country’s inequality and crime rate compare to another?

• Does a high population effect crime/inequality?

• Does GDP effect income inequality?

Initial Design:

Analysis of initial design:

 

This design used a mercator projection map in order to represent the geographic location of the data in an intuitive way, as it is a familiar pattern for the representation of the world. We made use of different colour schemes for each dimension of data, each contrasting with one another so that they stand out from one another. There is the use of depth of each visual element by making use of a ‘layering’ approach with our visualization. A pattern was used to display the population category of a given country.

 

Critical Feedback

This design failed in that there was just too much information on a single diagram, making it a noisy representation of the data. This noise resulted in an inability to effectively make the visual queries which the visualization was made for. By trying to show too many dimensions of data using too many channels at the same time, we weren't able to make the dimensions of data distinct enough from each other. Some helpful feedback was to use 2 channels to display each dimension of data so that it would be be easier to distinguish from the other channels. There was also criticism of the use of colours, some being too similar when looking at the diagram from afar, such as the darker blue, the black, and the darker gray.

 

At the time we also opted to not use real data for our display, which compromised the effectiveness of the visualization. Smaller countries were extremely difficult to see, especially with this large amount of noise. There was no mention of interaction with the visualization, which would have potentially fixed some of these issues which we had.

We partially implemented our design, and chose a framework called amcharts. Looking through their example maps, we found this example of a heatmap, using a scale from light to dark blue to show intensity. We chose to use a similar approach as a part of our design, since it is an effective way of showing contrast between each area. People are familiar with light to dark showing an increase in intensity, making it an intuitive way to make easy visual queries for each area.

Final design:

This is an overview for our visualization, showing homicate rates per 100000 vs Gini coefficient.

This shows how you can use the slider to filter the results to display only a selected bracket of data.

Zoom functionality allows one to take a closer look at the map, making it easy to see where each symbol belongs.

Mousing over a country will provide a white box which will display the country information in the above format.

Clicking on countries will select them, highlighting them in purple. The map will shrink to make space for a graphical representation on the right, comparing each of the four datasets. Mousing over any of the bars will give the name of the country and exact number represented by that bar.  Clicking on the countries again will deselect them.

Final design overview



 

Reaction to criticism

Our initial design of the visualisation fell victim to noise given the high degree of dimensions of data we want to show. After exploring the idea and considering the criticisms and feedback from our initial prototype and presentation, we decided to adjust what we are displaying on the overview to rather just support our most important visual queries. As a result, we decided to move GDP and population statistics to an interactive element of our design.

 

Data

We used real data this time, using data from the World Bank data portal for each of our datasets (Gini coefficient for income inequality, murders per 100,000, population and GDP per capita in current $USD).

 

Background colour

We changed the colour of the background to a lighter colour, white, with the fill colour of the countries changing from a grey gradient, to a light blue to dark blue scale. This fixed the issue of the darker grey fill colour and the black background sometimes looking like they are the same. The use of the gradient light to dark blue still has easy contrast between each shade, with the familiar pattern of light to dark representing low to high intensity.

 

Symbol shape and colour

To represent crime, we chose to use both symbol shapes and colour as a 2 dimensional way to show each bracket of crime (high, medium and low). This allows for significantly easier identification and comparisons between the crime rate of each country. The colours we chose for high, medium and low are red, orange and green respectively. This choice of colour stemmed from the intuitive heat scale of yellow to red which had great contrast with the blue fill colour of the countries, but yellow was often on the white background which had extremely low contrast, resulting in us replacing yellow with green. Symbols we chose for crime were circles, squares and triangles. These are very simple but distinct shapes. The size of the symbols varies very slightly, with circles being smallest and triangles being biggest, reinforcing the scale of low to high. The use of triangles for the maximum was due to the similarity of the triangle and an upwards arrow, being a familiar symbol for a high value.

 

By assigning categories for crime rate, we do hide the data on the overview, but it allows for easier ability to see trends in crime occurrence and to make easy comparisons between two countries. Exact figures can still be seen when clicking on or mousing over a country.

 

By having the recognizable symbols overlap any part of the map below them, there is a sense of depth given, with the symbols appearing to be on top.

 

Map navigation

When looking at the overview, there is still some difficulty in seeing smaller countries, so we implemented capability to zoom in and out on the map, making the placement of the symbols as well as the fill colour easy to see for even small countries. One can also drag the map in any direction to navigate through it, should they wish to view a different part of it.

 

Adaptive interaction

There is also a slider for the threshold that the symbols appear at (minimum and maximum value), allowing users to filter data being visualized to make specific queries, such as ‘which countries have the highest homicides per 100000?’.

 

Hovering your mouse over a country will provide you with all data for it in a white popup box, including country name, Gini coefficient, murders per 100,000, GDP per capita and population. The choice of white for the box was so that the text would be perfectly clear even when clicking on a country in the middle of a continent with a darker fill colour.

 

Filtering

There is a set of dropdown boxes allowing for one to choose what is represented by both the blue fill gradient and the symbols. This will allow users to compare any two measures, such as GDP per capita versus Population and Population versus Inequality etc. While this is not the primary aim of our visualization, we wanted users to have access to a visualization of any data which we have.

 

Adaptive graphing

Clicking on a country will select it, adding it to the bar graphs on the right. This allows for easy comparison between any dataset for selected countries. Selected countries are highlighted in purple, contrasting enough to any shade of blue to be noticeable, while still allowing the symbols to be visible.

 

Strengths and Weaknesses

 

Strengths

 

• Easily see correlation between crime and income inequality by location.

• The visualisation shows data dimensions very clearly and does not obstruct any data by being too noisy.

• Severity/intensity of dimension easy to see with colour scale.

• Flexibility in data being displayed.

• Quick satisfaction of location based visual query due to use of world map.

• Use of colour contrast for effective display.

• Quick visual grouping of countries by metric based on symbols of different shapes and colours.

• Side by side comparison of data sets in column graph.

• Ability to filter data to see the desired bracket of data.

 

Weaknesses

 

• The use of the slider is not accurate in that it does not display the values to which the user is setting it, only a relative position to the min and max of the range.

• Use of categorization ‘high to low’ for symbols results in less accurate comparisons made by visual queries.

• Sliders allow for one to select chosen brackets of data to be represented, between a selected minimum and maximum, but does not allow for accurate minimum and maximums to be chosen.

• The visualisation relies on some sort of user interaction to show additional dimensions of data beyond the two initial data dimensions shown.

• Graph axis do not start at zero making comparison between data inaccurate

• Selecting large amounts of countries will result in difficulty reading the graphs. Also, while bars for each country are kept parallel between graphs, the larger axis labels resulted in shifts of the graphs.

• The data used for the visualisations was from large datasets, making checking the data for integrity very difficult. The visualisation is expected to show shortcomings in the quality of the data.

 

Difficulties

 

• Effective implementation/demonstration of design ideas were restrained by the implementation tools, not because of the limit of the tools (amchart javascript libraries) but because of our lack of expertise with javascript. We dealt with this by dedicating time to becoming familiar with the tools although this came as a tradeoff since we had less time to spend in the design process, while still only partially implementing our design.

• Transforming the data from its raw source files (from the World Bank) into formats that we could use with the Amcharts javascript library. We used Python scripts to manipulate the data but we could not double check the integrity of the data from before and after the transformation.

• We had to make compromises in our design in order to best satisfy our primary visual queries, for example the categorization of the data represented by the symbols allows for very easy visual queries, while forcing users to interact with the visualization to get data of smaller granularity.

 

Future work

A concern for us is how the data might show correlation between the the different dimensions of data but it might not be a false implication of causality. New research is currently trying to show that measures like Consumer Spending in a country are more causally linked to violent crime. Researchers believe this is because an abundance of luxury goods (cars, jewelry, etc.) in society alienates the impoverished, making them feel like they are not part of society. This in turn leads to an emotional disconnect linked to violent crime. Future work could include visualising this data as an additional dimension.

 

We also realise that the design of this visualisation is not perfect and we could improve upon the weaknesses we have identified and upon any additional weaknesses highlighted through further rounds of presenation and critical feedback.

 

Team workload balance

 

Robert - Implementation of design and proofreading

Adam - Design and report

The project overall was a joint effort with both group members contributing equally in the completion of deliverables. For the purpose of efficiency, Robert placed focus on the implementation of the design, while Adam placed focus on designing the visualization as well as working on the report.