Blog Archive

Tuesday, June 11, 2013

When Maps Fail.

Maps are, after all, data visualizations.  They help us visualize geographical features of immense scale relative to our human-sized perspective and surroundings.  Over time, as our technology has become more sophisticated, we’ve been able to embed more and more information into our maps, such as ocean currents, land altitudes and distances.  This has made our way-finding more precise and manageable.     

Maps have helped us find our way through wilderness and unfamiliar territory, but they have also fueled the imagination of the homebound for millenniaAs maps have evolved from a navigation tool to a tableau for story-telling, we’ve started to confuse the purposeful geographical features required for navigation with those map-features needed for storytelling.  In fact, we may not need rivers, roads, mountains, or even distance as a scale to help tell stories of location and place. 

Maps often reflect artifacts of a time past, rather than the pressing issues of the present time.  Take, for example, the boundaries of the 50 United States.  The first 16 states of the United States were established by 1800, and 45 of its 50 states were established by 1900.   During that time, states declared their boundaries mainly by geographic features much more relevant to the economies of the time, including:

  • Rivers and bodies of water
  • Mountain ranges
  • Natural resource claims

As population and demographics have evolved in each of the states, their area and geographical boundaries are not proportional to their population density.  Furthermore, size either in relation to population or to area alone does not drive political attitudes.  Rather, it’s things like:

  • The emergence of cities and their economic importance
  • Ports and other sources of international traffic;
  • Centers of industry, such as auto manufacturing, or mining

Hopefully no-one interprets this as a call to re-examine state boundaries, state's rights, and the sovereignty of the local over the federal.  However, these kinds of artifacts continue to create problems for data visualization.  And we need to be aware of them.  Below are three examples where map-based representation can distort the story that we tell.  

Elections and Electoral Votes

The United States’ 2012 presidential election gives a great example of map-based representation.   Below are two different approaches to conveying the importance of electoral votes.  The Huffington post portrayed the election results in the more traditional fashion, by state, with a complementary analysis up top of the electoral vote breakdown:

Compare this to the New York Times, who took a wonderful, alternate approach to story-telling of the election by separating an outcomes-based depiction from a geographic depiction: 

The Huffington Post’s depiction is more recognizable and familiar to most people.  It's therefore a safer, conventional approach.  The New York Times, on the other hand, did a great job of challenging our conventions in order to convey a more meaningful and interesting story about the electoral system, the importance of swing states, and the many different potential outcomes.  

Postal Codes

Postal codes in the United States evolved into the ZIP code during the 1960’s.  ZIP codes have since been influenced by many different factors, primarily population density.  The U.S. postal service must practically deliver correspondence in the most efficient way possible, and ZIP codes in part reflect the location of distribution centers that best serve this population.  

ZIP codes provide us with a tempting subdivision of city and state boundaries for placing locational data.  But ZIP codes become irrelevant in the face of the ethnic and economic diversity of those within those geographical zones.    

The Federal Reserve attempted some visualizations of regional mortgage delinquency and foreclosure conditions.  The visualization allowed the viewer to focus by ZIP code. 

Compare the limited insight we get from this visualization to the wonderful visualization concept created by the USDA on food deserts - geographical zones where the population has limited access to healthy, fresh foods.   

In the Federal Reserve example, the ZIP code artifact limits our ability to group by more relevant issues such as income, ethnicity, and occupation.  The USDA example, on the other hand, gives us great insight into the geographic issue of access and distribution, which is a more relevant cause of poor diet and economic inequality.

Our Own Context and Attachments

This is a softer, more subjective issue, less associated with data.  But it is still a powerful distortion.  Because we have emotional attachments to our own locale, or ones that we have visited in the past, we naturally tend to emphasize those in how we think about the world.  If I think of Colorado, I think about the Rocky Mountains.  However, I've overlooked the fact that a large part of the state is plains.  Another example: countries that have, historically, wielded very centralized power such as China and Russia, tend to be associated with their capitals.  But Russia has a huge Asian expanse that is far from Moscow, and China has a huge western expanse beyond Bejjing that covers far more cultures, religions and ethnicities than we are often aware.      

A great project that brings this issue to life is Benjamin Pollach’s World Map Archive.  It’s a wonderful crowdsourcing of world maps drawn by people around the world.  Just one rule: no peeking at a map.  And the results are fascinating.  

Of course, people’s unfamiliarity with geography influences all of the results.   But for those that do get most of the generalities right, there’s plenty of additional insight that comes out of the size, shape and placement of important features. 

So, in general, when we think of a tableau for our location related data, we must always ask:
  • Just because it is locational data, should it be shown by location? 
  • If by location, is there causality?  What are the defining features of that geography that have truly influenced this data? 
  • Are the influencing factors on the data ultimately geographic, social, or something else? 

No comments:

Post a Comment