LEARN NC was a program of the University of North Carolina at Chapel Hill School of Education from 1997 – 2013. It provided lesson plans, professional development, and innovative web resources to support teachers, build community, and improve K-12 education in North Carolina. Learn NC is no longer supported by the School of Education – this is a historical archive of their website.

Maps get really interesting when we start adding human data to them — population, economic production, social behavior, and so on. Mapping is a powerful way to summarize and communicate those kinds of data. Unfortunately, mapping is also an excellent way to confuse them.

Presented with the same data in a chart or a table, we might take the initiative to analyze it — in fact, we’d be forced to analyze raw data in order to make sense of it. But because the impact of a visual map is so immediate and powerful, we rarely question it. One possible explanation is the way the brain processes information: visual processing occurs on the right side of the brain, while analysis with words and numbers happens on the left. Another is that maps just look finished and don’t invite analysis — they’re pre-chewed, you might say.

Whatever the explanation, maps can easily deceive. Let the reader beware.

Choosing the right data

An easy way to see the problems that can arise is to map the number of crimes committed in each state.

Figure 15-1. Crimes committed by U.S. state, 2004. From U.S. Department of Justice data via Wikipedia.

Map: Crimes committed, by state, 2004

Glancing at that map, we’d think California, Texas, Florida, and New York were horribly dangerous places to live, and that there was hardly any crime at all committed in New Mexico. The problem is that we’re basically looking at a map of population. Of course, the more people live in a place, the more crimes (all else being equal) will be committed there. Here’s a population map for comparison:

Figure 15-2. Population of U.S. states, 2000. From U.S. Census data via Wikipedia.

Population, by state, 2000

That’s why when you see data on crime, it’s typically crime rates — crimes committed per 100,000 people, for example. This map shows crime rate by state.

Figure 15-3. Crimes committed per 100,000 residents, by U.S. state, 2004. From U.S. Department of Justice data via Wikipedia.

Crimes per 100,000 residents, by state, 2004

Now, a distinctly regional pattern emerges — crime rates are actually lowest in the heavily populated Northeast. This is data worth mapping, but it shows the importance of choosing the right data to map — and of asking, when we’re presented with a map, whether the data behind it were well-chosen.

Using color

You might also note that I used red — blood-red, in fact — to indicate a high crime rate. It’s always worth noting the colors used on a map. You’re going to respond to those colors viscerally whether you mean to or not! Once you’ve recognized what the colors are and what they imply, you can, if necessary, dismiss them.

Scale

Another way maps can distort data is by choosing the wrong scale. In the maps above, I used darker shades of a single color to indicate more of something. That’s a common convention, and it’s easy to read. The question is, how much more does a darker shade represent? It’s important to read the key, and it’s important to keep telling students always to read the key, because that’s where the answers are. Here again, though, a map can make a powerful visual impression that we may have to work hard to correct.

In drawing my map of crimes committed, I had a choice of scales. I could choose a linear scale, in which each darker shade of color represented an equal number more crimes (say, 1,000 more). Or, I could choose a logarithmic scale, in which each shade darker represented an equal number times as many crimes (e.g. twice as many).

Choosing the right scale

It’s usually best to choose a linear scale, because logarithmic scales confuse most people. For example, the Richter Scale, which measures the intensity of earthquakes, is a logarithmic scale, so that 6 on the Richter Scale means a quake 10 times as powerful as one measuring a 5, and so on. Logarithmic scales are great for distinguishing among very small things and among very big things. Richter 3 isn’t noticeable, while Richter 9 — just six steps up the scale, but a million times more powerful — will destroy a city. But logarithmic scales are not as intuitive as linear ones.

Try this example. The January 2010 earthquake in Haiti measured a 7.0; the quake six weeks later in Chile measured an 8.8. How much more powerful was Chile’s earthquake? See the footnote for the answer.1

In mapping crime rates, I used a linear scale with a break in the middle. Because state-by-state crime rates fell into a fairly narrow range, I needed only six shades of red for a linear scale. But absolute numbers of crimes ranged from 1,000 to more than 200,000. I wanted to distinguish between states having 5,000 crimes and states having 10,000 crimes — twice as many is a big difference — but a linear scale with such fine distinctions would have required more than forty shades of red. So I used a logarithmic scale, roughly doubling the number of crimes at each shade darker.

Figure 15-4. The scale from Figure 14-1.

Map: Crimes committed, by state, 2004

Scales are rarely chosen with an intent to confuse; in fact the mapmaker often has little practical choice. But they can, nevertheless, cause confusion if the reader isn’t paying close attention.

Trouble arises quickly when you pair maps with different scales. I’ll pick on myself this time.

In compiling these maps of North Carolina’s changing demographics, I and LEARN NC’s graphic designer used shades of blue to represent population characteristics of North Carolina counties over time. Most of what we were representing was percentages, so we were nearly always able to use a linear scale. Here’s a map showing percent urban population of each county in 2000:

Figure 15-5. Percent urban population, 2000. From Mapping a Changing North Carolina.

county map

When it came to mapping Latino population, though, we faced a problem. In 1980, there were very few Latinos living in North Carolina, and so to show geographical variation, we needed a scale that would show the difference between, say, 1 percent and 2 percent. But by 2000, Latino population had grown tremendously, and so our scale needed to go as high as 20 percent. The only way to show big numbers without losing fine distinctions among small numbers was to use a logarithmic scale.

Figure 15-6. Percent Latino population, 1980. From Mapping a Changing North Carolina.

county map

Figure 15-7. Percent Latino population, 2000. From Mapping a Changing North Carolina.

county map

The drawback is that the logarithmic scale makes the pace of Latino immigration look even higher than it is — especially since we’d used linear scales for everything else. Drawn with a linear scale, Figure 14-7 would be a lot lighter, but Figure 14-6 would be practically useless. We’re counting on teachers and students to read the materials accompanying the maps, but we’re also aware that this sort of map can give false impressions and be used for political purposes.

Area vs. population

If you watched more than five minutes of television in late 2009 and early 2010, you’ve seen this:

Figure 15-8. Maps from Verizon ad, 2009.

cell phone coverage map from Verizon ad

These maps, from ads for Verizon Wireless, show 3G network coverage by Verizon and AT&T. The impression they give is that most of the United States isn’t covered by AT&T at all — a misrepresentation over which AT&T sued Verizon. (A judge refused to issue a temporary restraining order forcing Verizon to pull the ads, and AT&T decided to drop the suit and retaliate the old-fashioned way — with ads of their own.)

Let’s suppose, though, that you noticed that the maps refer only to a particular type of wireless coverage. AT&T claimed in their lawsuit that “our 3G service is available in over 9,600 cities and towns.” In fact, if you look closely, the blue splotches on the map above cover the country’s major population centers. Here’s a map from the U.S. Census Bureau showing population density:

Figure 15-9. U.S. population density, 2000.

map of United States population distribution, 2000

The similarity between the Census Bureau’s map and the blue AT&T map would suggest that there’s some truth to AT&T’s claim. On television, of course, most people will see all the empty space and think “most of the country isn’t covered.” But territory covered may not be the best reason to choose a cell phone plan, unless you spend a lot of time just driving back and forth across the prairie in a beater truck that’s liable to break down. Otherwise, you should probably be more interested in the number of people covered.

By drawing the viewer’s attention to area instead of population, the maps create potential for confusion. This is a frequent error — and sometimes a deliberate deception — in mapmaking. Even when the data refers to population, what we see is area, and that initial visual impression can overpower

Population and politics

This error is a variation on the last one. Here, again, what we see is area, but what matters is population.

We’ve all seen political maps showing “red states” and “blue states,” with Republican-majority states shaded red and Democratic-majority states shaded blue. Here’s a map of 2008 presidential election results by county, using that convention:

Figure 15-10. 2008 presidential election results by county. By Mark Newman, Department of Physics and Center for the Study of Complex Systems, University of Michigan.

map of presidential election results by county, 2008

If you knew nothing about American politics or current events, you would probably look at that map and assume that surely a Republican was in the White House. It’s nearly all red!

The problem here is that in U.S. politics, there is a strong correlation between high population density and Democratic voting. In other words, cities tend to go Democratic, while rural areas are more likely to go Republican. All those big red counties in the middle of the map have very low populations. Again, when we look at this map we see area — but we need to be thinking population.

Mark Newman, a physicist at the University of Michigan, came up with a solution to that problem after the 2004 elections. He used a computer algorithm to generate what he called a cartogram — a map in which each region’s size is proportional not to its land area but to its importance in the data being communicated. Newman’s cartogram for 2008 county-by-county election results shows each county in a size relative to its population — and thus its voting power — not its area.

Figure 15-11. “Cartogram” showing 2008 presidential election results by county. By Mark Newman, Department of Physics and Center for the Study of Complex Systems, University of Michigan.

Presidential election results by county, 2008: Cartogram

You can see more cartograms on Newman’s website.

We know, of course, that Barack Obama won the election. But that first map nevertheless overstates the importance of a lot of big western counties and makes the U.S. look far more conservative than it is. If we didn’t already know that what it implied wasn’t true, how would we figure it out?

Knowing what we don’t know

If it’s hard for Americans to spot the errors in a map of the United States, how much harder is it when we’re presented with a map of another country? Suppose you were presented with a map of election results from Norway, or of economic production in China? Would you be able to interpret it?

On the process guide for reading maps that follows, one of the questions is “What don’t I know?” That’s a deceptively hard question, and it has implications beyond maps and visual literacy — how do you know what you don’t know?

Here are a few map-related suggestions for helping students figure out what they don’t know.

  1. Use lots and lots of maps. Use as many maps, and as many different kinds of maps, as you possibly can. Maps are all over the internet; you can find a map showing just about anything. (The links throughout this edition will give you some ideas.)
  2. Use bad maps. If you teach a subject where this is remotely appropriate, look for examples of bad maps (and other bad visuals) in the news and take a few minutes to analyze them as a class. (Try to avoid complete cynicism, though. It’s easy to start thinking that everybody is lying to you, when most often, the mapmakers and graphic designers just didn’t have the time or resources to think the problem through and execute it carefully.)
  3. Use maps up front. Let students wrestle with maps on their own, before they read a related text. That way, they’ll be forced to interpret the map for themselves.
  4. Role-playing. Try to think of a real-world use for any given map and ask students to put themselves in the position of someone trying to use it. I mentioned, above, a map of Chinese economic production. In that case, suppose you were the CEO of an American company looking to build a factory in China. Where should you put it? Does this map give you what you need to know? What else would you need to know, and where could you find it?