Become a Creator today!Start creating today - Share your story with the world!
Start for free
00:00:00
00:00:01
Episode #98: Mark Monmonier
181 Plays7 years ago
Mark Monmonier is a Distinguished Professor of Geography at the Maxwell School at Syracuse University. In the data visualization field, he is probably best known for his fantastic book, How to Lie with Maps. Professor Monmonier’s research focuses on the history...
The post Episode #98: Mark Monmonier appeared first on PolicyViz.
Recommended
Transcript
Introduction to Maps and Guest Speaker
00:00:11
Speaker
Welcome back to the Policy This Podcast. I'm your host, John Schwabisch. If you are interested in data, if you're interested in data visualization, you have assuredly come across a time where you are reading a map or you are making a map and you have questions and you may not be exactly sure how to plot geographic data. You may not be sure the map that you're looking at is providing or illustrating the data to you in a responsible, objective way.
Meet Dr. Mark Monmonier
00:00:37
Speaker
So I'm very excited to have on this week's episode, Mark Monmonier, a distinguished professor of geography at the Maxwell School, where I also did my graduate work. So there's a little bit of a camaraderie there. Dr. Monmonier has written a great book, How to Lie with Maps, soon to be in the third edition, along with several other books on maps and geography. Professor, thanks so much for coming on the show.
00:01:00
Speaker
And thank you for inviting me. I'm really excited to chat with you about maps. I have the second edition in front of me. It is all marked up with notes and highlights everywhere. And now I'm excited about the third edition coming out. I'd like to ask you just to introduce yourself for folks who may not know about you and your background. And then I'm going to ask you very simply to tell us a little bit about how people lie with maps. OK, I'm a family member of Maxwell.
00:01:26
Speaker
I did graduate work at Penn State, spent a year at the University of Rhode Island, three years at SUNY Albany, and I've been at Syracuse since 1973, and I like it here. I teach courses in map design, and I also write about maps.
Exploring Monmonier's Works and Concepts
00:01:42
Speaker
I've written a number of books concerned with such things as meteorological charts, specifically the history of weather maps, the use of maps in disaster management,
00:01:56
Speaker
Coastal Mapping, a book called Spying with Maps. It's Concerned with Surveillance Cartography. I have a book called No Dig, No Fly, No Go, which, well, that's the title that the publisher chose. I would have preferred to use a title, maps that say no, but publishers can be very persuasive in this thing. They hold all the cards.
00:02:22
Speaker
I edited volume six of the history of cartography concerned with the 20th century. And for Syracuse fans, I wrote a book called Lake Effect, which started out basically to be a history of cartography concerned with lake effect snow and found out some interesting things about, well, it took a long while for people at the weather bureau to actually start to map snow.
The Art of Lying with Maps
00:02:48
Speaker
Snow was treated as something which was actually beneficial.
00:02:52
Speaker
If you had sleighs in winter, it would provide traction. And when it snowed, they would basically take the snow and melt it and measure the rain equivalent of it. As born in Buffalo and gone to school in Syracuse, I'm going to put that second on my list here after the third edition. So I'm making a special, special note of the lake effect book. Okay, I want to know about lying with maps. Well, this comes from the notion that there are different kinds of lies.
00:03:22
Speaker
and black lies and white lies, and white lies being small lies. And they're also the kinds of lies that children might tell where they don't tell the whole truth. They leave some things out. And I think the most cartographic lines are the result of a map leaving things out. And map readers are used to this because you simply can't show everything and show everything and have someone that's hopelessly cluttered, and then it wouldn't work.
00:03:50
Speaker
So because users, map readers are used to sort of seeing maps and maps seem to be working, they've developed, I think, a reasonable degree of credibility as data objects, as facts. And a problem is a lot of people who make maps don't really realize that there are a variety of different kinds of maps that you could create from the same data.
00:04:19
Speaker
And this is especially true for a type of map that we call the quarter
Mapping Challenges and Solutions
00:04:24
Speaker
-life map. But I mean, it can also be other kinds of maps. If you have land use maps, you just can't show all different kinds of land uses. There are also issues of scale. You can use a relatively large scale, large scale being referring to relatively detailed scale, where you can show a lot of information for a relatively small geographic scope. You could use a small scale
00:04:48
Speaker
for a very, very broad area. But when you do that, obviously, the information that you're able to show for neighborhoods and even for counties is relatively small. But going back, though, to the issue of the Corplath map, which is one of my favorites, and it's a type of map that we oftentimes see in discussions of census data.
00:05:15
Speaker
and other kinds of information collected to inform a discussion of policy. What we have here is data collected for aerial units. And the aerial units are oftentimes political units, states and counties, or in some cases even nations. When you get into the issue of, let's say, worldwide maps, things become really complicated because different countries
00:05:44
Speaker
have different ways of defining things, or they take their censuses at different times. And their censuses would vary in accuracy. When you're, let's say, working with data from the United States, we have a situation where we generally have maps for states, or we have maps for counties. And some states have lots of counties, others have relatively few. We have a situation where, if you look at a national map,
00:06:13
Speaker
And if you look at a distribution that's mapped at the state level, we have some huge states that have relatively small populations. And that population might be concentrated in a small number of relatively large cities. But what the eye sees when you use one symbol to represent an entire state
00:06:38
Speaker
It looks as if it's uniform, and in many cases, it really isn't. Also, too, where you have, let's say, small states like Rhode Island and relatively large states like Wyoming, which has more people than Rhode Island, but Rhode Island, a very, very tiny piece of territory that's going to get a relatively small symbol.
00:07:02
Speaker
and Wyoming is going to get a relatively large symbol, which is going to create a much greater visual impact. And this is especially true, let's say, when we look at maps showing the results of a national election, where very commonly, I mean, if you look, let's say, at 2016, Donald Trump not only won the Electoral College, but he basically won the US landmass, you know, where you basically get
00:07:30
Speaker
these large red states, which might have populations and also numbers of electoral votes, much smaller than some relatively small states. Right. So do you have a favorite alternative to the choropleth to adjust for the geographic scaling or the geographic distortion that occurs? Yes. It's something that they call a visibility base map.
00:07:59
Speaker
And I used it in a book on population geography that I wrote. I even went and registered a copyright for it because I was just concerned with how the copyright process worked. And I got some blowback from the copyright office, but ultimately it went through. And basically I've told various people if you want to use it, go ahead. So I mean, but what it is, is each state is presented by relatively small polygon.
00:08:28
Speaker
I don't think I have any state that has a polygon with more than 10 sides. And you can identify, if you know anything about the geography of the United States, you can tell what Florida is, you can tell what Mississippi is. Florida has a distinctive shape. Mississippi is next to Louisiana. So the geographic position of a state is going to be one clue. Its intrinsic shape is going to be another one.
00:08:56
Speaker
Most people know anything about the geography, know that the state we have in the upper right is Maine.
Data Representation Techniques
00:09:04
Speaker
But these states are drawn in such a way so that they're in proportion to their population. Now, it's not a perfect fix.
00:09:16
Speaker
because, well, things might have changed recently. But, you know, if you look at a map of Georgia and Florida, and you might think that Florida is bigger or smaller than Georgia, you know, and actually Georgia is bigger. More square miles. Yeah. Right. Right. Hand on them. They has that large part that drops toward Cuba.
00:09:41
Speaker
And so, you know, people are not necessarily going to register relative size in a very, in a highly accurate way. But I think, you know, this is certainly far more reliable if you're concerned with large areas with small populations not having an overwhelming impact. So, yeah, that's one thing I tried and it's occasionally used and there's some other
00:10:09
Speaker
things that are done. If you have count data, generally with a core path map, what you have would be data which would be either a percentage or a ratio. An example of the count variable would be number of people, number of inhabitants, number of people who voted in an election. We distinguish between count data and intensity data. Intensity data would be the percentage of the population that voted.
00:10:38
Speaker
the percentage of the populations that 65 went over. Generally, choropleth maps, where you're using shaded values, basically, the conventional metaphor is that darker means more, lighter means less. So if you have a large mortality rate, you'd have a symbol that would be relatively dark. If you have a relatively low mortality rate, the symbol would be relatively light.
00:11:07
Speaker
In any event, if you are concerned with setting up a chain of mortuaries, and if you're trying to identify areas that, let's say, maybe have a lot of people dying, you might want to have count data where you're looking at a number of deaths. If you use a choropleth map on this, you're going to find out with something where basically states that have lots of people
00:11:36
Speaker
irrespective of, let's say, health situation. I mean, relatively dark symbols states with relatively few people are going to have light symbols, but it would probably make more sense if you use symbols that would be graduated in size. So that rather than darker means more, lighter means less, which would be useful for an intensity measure, you have something like bigger means more.
00:12:03
Speaker
smaller means less. And a typical approach here is what we call graduated circles, where you have one circle for each state. And the circle, if it's relatively large, it means you have a lot of people or a lot of deaths. If it's relatively small, you have fewer. But another way around that is to have what I call a dot matrix array map.
00:12:31
Speaker
where you basically have, you can start out in making this map with sort of a grid. And you have this grid where you have dots at each grid intersection. And these dots would all be the same size. And each dot represents, if you have deaths, I don't know, it's going to be like 10,000 deaths. And you can basically allocate these dots to the states based upon the number of deaths that you have.
00:12:58
Speaker
This is an alternative to the graduated circle. A problem with the graduated circles are that you have one here, you have one over here, and let's say this one has a diameter which is twice the diameter of this one here. Well, what does that mean? Well, we know that the area of a circle is proportional to the square of the radius or the square of the diameter, the one that has a diameter
00:13:26
Speaker
twice as big or that it's twice as tall, is representing four times as many as something. And a problem is that the human librarians system, when you look at those symbols, what you're seeing is not just area, you're also seeing height. So there is a tendency to underestimate the size of the larger symbols. Now, a way around that, if you use the dot matrix symbol, whereas
00:13:53
Speaker
An example would be, let's say if you have a state that has a lot of something, it basically might have a little grid. We have these little circles and maybe five circles this way, four circles this way, altogether 20 circles. You can count, you can look at them and you say four times five, okay, that's 20. And it would be a more accurate way of actually estimating the amount. And if you keep these,
00:14:20
Speaker
so that they're in something like a rectangle, or it could be something like this. If you have relatively tall states, such as Illinois or Indiana, you have a symbol where it is proportional to some magnitude that you're trying to show. But the symbol also has something which is at least vaguely countable. And so there's something nicely intuitive about it.
00:14:50
Speaker
I just want to distinguish between two principal kinds of data. We have count data, we actually have a count, and we have intensity data where we're taking this count and we're adjusting it to sort of make it more
Creating Effective Map Legends
00:15:05
Speaker
relevant. So if we have number of people, we can divide population by land area, we can get population density. If we have
00:15:15
Speaker
population that's 65 and older, you can divide it by total population and you can get the percentage of the population in the elderly cohort. And I think that when you have things like this, darker means more, lighter means less is a very good way for representing intensity measures.
00:15:37
Speaker
You talked about setting up this legend or this key with proportional symbols. I also want to get your take on using a legend or a key with a choropleth map. So I guess I have a few questions here, and I'll just blurt them out, and you can take one at a time, I guess. But the questions are as follows. How does one pick...
00:15:58
Speaker
The breaks in the legend so if i'm going from let's say zero two hundred percent i think most tools like tableau and google maps they'll just take zero hundred percent of divided into five and i'll get these these bins and the second question is related to that.
00:16:14
Speaker
My chapter 10 in your book is all is all bookmarked and underlined. So I have an idea of what you're gonna say. But the other question is, if I have bins where I have no observations, so let's say I have a group that goes from one to two, and a group from, you know, two to three, then three to four, but if I have no one in two to three,
00:16:33
Speaker
Do I need to show that bin as well? Or can I just show the bins even though they're not adjacent? So two questions there. One is, what's the, is there a best way, is there a strategy to pick the breaks? And then the second question is, do I need to show all of the bins or all of the breaks, even if there are no observations in some of them? Oh boy. Okay. Okay. You have a situation where if you're basically using five bins,
00:17:00
Speaker
and you have the range between the lowest value and the highest value, you could very well wind up with three empty bins. And if you're putting symbols in there, it's obviously a waste of symbolization. It would probably make more sense to, well, it might make more sense to
00:17:25
Speaker
divide the data if you have some really extreme values. Now these extreme values, they're called outliers, and you can maybe identify the outliers by pointing them out. Maybe for the outliers you could actually write in, if we're dealing with state data, write the numbers in the particular polygons,
00:17:49
Speaker
and say maybe, you know, give them a relatively dark symbol, but put them all in this outlier category. And if you put the actual numbers in, you know, people can get the sense that, okay, these are sort of, you know, head and shoulders, eyeballs, and crotch above everything else. And they are special, and we're not letting them interfere with their ability to show variation for the other more moderate values. Now, that's one strategy.
00:18:20
Speaker
Probably the worst case would be what some software does or at least used to do is to take the lowest value and the highest value and chop it up into five categories. And somewhere along the line somebody figured out that, well, five categories is maybe an ideal number. I mean, seven probably in some cases would work if you have seven symbols that are relatively distinct from each other.
Color Scales and Data Breaks
00:18:49
Speaker
If you're able to use color, you have some other possibilities now. One can really go down that rabbit hole because you can use a variety of different colors and get something that, you know, some would be red, some would be green, some would be yellow. And if the data have nothing to do with the electromagnetic spectrum, I say this is going to be very confusing.
00:19:16
Speaker
If you have, let's say, temperature data, if you're concerned, let's say, with the average January temperature, you know, typical scale that might be useful here would be some sort of a warm-cold contrast where you run from a strong blue to a strong red, strong red representing relatively warm, strong blue representing relatively cold, and you would have what's called a divergent scale.
00:19:45
Speaker
where maybe you could put something in the center, which could be sort of a neutral color, something like white, and you go out in one direction, and you're heading into light red, medium red, dark red, light blue, medium blue, dark blue. And I think most readers, if you make this clear in the key, would probably grasp the fact that you're using this red-blue metaphor
00:20:12
Speaker
to explain color. And you can do this, too, for some other kinds of distributions where you have some sort of maybe meaningful center. But if you're concerned, let's say, with polling data, and if you're concerned, let's say, with divergent views, whether it's some places where people are really extremely opposed to something
00:20:40
Speaker
and some other areas where they're extremely in favor of something. But if there's sort of a situation where a lot of people either couldn't care less or they have very mixed opinions. And by the way, this is raising another issue here, too, because we're concerned with ecological data. And it's not just, let's say, one state, which is uniformly viewing something. You have a lot of variation going on within that state.
00:21:11
Speaker
Probably one of the best things, by the way, if you can think of a situation where you have a group of governors and they're having some sort of national governors conference and they're concerned with comparing states, you know, they are a figure that you have for the whole state probably has maybe a bit more meaning. Maybe, maybe not, you know. And if you really want to have a good view of something more reliable map, hey,
00:21:38
Speaker
don't go to states, actually go to counties and you get the goods, the eternal vacation there. Yeah. Okay. I realize, you know, I was sort of jumping off topic there, but going back to the issue, let's say, of how we partition data. Why not simply show the viewer the distribution of numbers. You can get what I refer to as a number line.
00:22:04
Speaker
And basically, just trying to think, I mean, it can also be referred to as a univariate scatterplot. The standard scatterplot uses x and y axes, and you have a two-dimensional space with lots of little dots in it. But if you're concerned with mapping one variable, you just put all of these little dots on a line. Now, if you have problems with overlap, you can have some of them sort of jump up, so it looks like a little histogram.
00:22:30
Speaker
But this is the friggin dirty way of giving the viewer a sense of, OK, here we have a lot of values that are clustered around here. Here we have an area that's relatively empty. Here we have a few outliers. Now, sometimes people make maps say they like to look for natural breaks. And sometimes you have them and sometimes you don't. You have an actual break.
00:22:55
Speaker
if you actually have a big gap in here, and you could say, OK, there's not very many values in here. So it makes sense to have a break somewhere in here and treat these in one category and these in another category. Oftentimes, you don't have natural breaks. Now, there's another kind of break that very few people who make maps seem to recognize. Software generally doesn't recognize it. And this is what I call the inherently meaningful break. And what makes a break inherently meaningful?
00:23:23
Speaker
You have a map that you're making for the 50 states plus only DC. One thing that is an inherently meaningful break would be the national average. Because if you use the national average as a break, you can look at the state, oh, it's above the US average. This one is below the US average. Now, if you do that, you have to make certain somewhere that the map key indicates that this break that you're using is, in fact, the national average.
00:23:53
Speaker
If you don't do that, you know, what doesn't make much sense? There's another kind of natural, of inherently meaningful break you might have. If you're concerned with the rate of population increase, and an inherently meaningful break there is zero, because some states gained and other states lost. And, you know, you wouldn't want to have a category where you had
00:24:20
Speaker
some some modest winners in with some modest losers. I mean, so and software, unless you tell it that unless you use an override, it maybe is not going to show up. Now we talk about, let's say, you know, smart programs and smart laptops and smartphones. You can also have a smart database. A smart database would be one that would actually indicate any inherently meaningful breaks that you have
00:24:50
Speaker
in a numerical distribution, which is a part of that database. And so the software could take this, recognize it, and use it. And if you make a map and if you're all shrewd, if you're all concerned with helping your viewer, you basically look at your data and have you heard of John Tukey?
Interactive and Advanced Mapping
00:25:11
Speaker
Of course, yep. Okay, the famous statistician.
00:25:14
Speaker
You know, Tukey, like I said, probably has a number of famous sayings, but the one that I like is that to understand your data, you have to look at them. And it's amazing sometimes how many people go ahead and make maps without looking at the data. Yeah. Look at them with a number line, you are looking at them. And, you know, that thing you could do is put them into an Excel spreadsheet,
00:25:39
Speaker
and rank order them, and actually look at them. Don't think other than mapping, even though you're using software which is quick and dirty, and it is inherently instantaneous. Well, don't let that exclude you're actually taking a look at these data and understanding them. Now, having said that, how about the viewer? You can have a situation nowadays with interactive maps or dynamic maps where you could give the viewer
00:26:08
Speaker
Corbileth map with only two categories. And you have a break point, otherwise known as a cut point. And you have a slider. And you can move that slider back and forth, back and forth, and actually work with the data, engage with the data, and get maybe a better sense of where the extreme outliers are, where the extremely high values are, where the extreme low values are.
00:26:35
Speaker
You could have some way in which you could sort of notch it at the US average. You could also, however, even have something where you could have that slider move across the number line at a uniform rate. And you could then note some shifts in your categories. And this makes the map interesting.
00:27:02
Speaker
And it can sort of underscore the fact that, well, if you're moving from low to high, and if there are a lot of relatively low values, all of a sudden these low values light up. And not a whole lot happens until much later on. And you can look maybe at the last place that sort of is turned on. And that basically sends a signal, ah, this is the highest one.
00:27:25
Speaker
Right. I love this idea of plotting the data in an additional way to the map. So you have the map and we know people like maps, but you also actually show, you know, at its very base, just the distribution of the data so that you can see, oh, you know, this data is skewed to the left or skewed to the right.
00:27:41
Speaker
Before we go, I want to ask you what can we expect in the third edition of How to Lie with Maps? So now I'm all excited about what we're going to have. So what have you added or subtracted or changed in the edition that's about to come out?
Future of Mapping and Closing Thoughts
00:27:59
Speaker
Okay. I'm talking a lot more about web maps. I talk about also known as Slippy Maps.
00:28:08
Speaker
I still talk about the Mercator projection, which has been used in a lot of situations inappropriately, because it basically magnifies areas as you get closer to the poles. Well, they also talk a little bit about the Web Mercator, which is used by Google Maps and some other mapping applications, which are zoomable, because basically you have little tiles.
00:28:34
Speaker
And those applications draw upon a storehouse of lots of little tiles. Some tiles are really small and you have a lot of them. Other tiles are considerably larger. And in order to get this interactive zooming, what the software is doing, it's basically serving up contiguous tiles.
00:29:01
Speaker
And the Mercator map, the web Mercator makes it easy to sort of keep track of everything. But if you go into some of these applications and you zoom all the way out to a world map, some of them don't let you do this, but if you do, you're going to get a Mercator map. If you zoom all the way in,
00:29:24
Speaker
The projection is not going to be noticeable. The Mercator projection works out very nicely if you're dealing with a small area. It provides you with roughly the same scale going this way and that way. And there is no distortion of angles and very, very little distortion of area. So it's an extremely versatile thing. I don't spend a lot of time going off on that. But I talk about web mapping. I have a chapter on image maps.
00:29:54
Speaker
I have a chapter on prohibitive cartography, which I've identified as a completely new genre, but it's something which is much more prominent in the 20th century. And prohibitive cartography basically is concerned with maps that tell you where you can't go or what you can't do. And we find this, let's say, in an urban context,
00:30:20
Speaker
in our zoning maps. Or let's say if you have a historic district, basically a place where if you want to paint your door some other color, you can't. So if you do, you're going to get a fine or something or sort of a nasty letter. Anyway. Right. So one more question before we go. So I want to ask, do you have a single favorite map, either a cartographic map or a data map? But do you have a favorite map?
00:30:49
Speaker
I have a favorite type of map. Okay. Yeah, I'll settle for that. Yeah. Okay. Interactive web maps, basically radar maps from Noah's next rad radar system. And I get these on my iPhone and what's neat about them is that they show the intensity of precipitable moisture. They also show you weather systems that are moving. They can show weather systems as they're getting closer to where you are.
00:31:16
Speaker
They can give you a good hint of, you know, if you really zoom in, they can give you a good hint as to how long it's going to be before it starts raining. They're dynamic. They're certainly timely and very engaging.
00:31:33
Speaker
Terrific. This has been great, Professor. There is a ton of information in How to Lie with Maps, and I'm sure in many of your other books, and I'm certainly going to take a look in the book on Lake Effect Snow. So thank you so much for coming on the show. It's been a real pleasure. And thank you, John. Appreciate it.
00:31:51
Speaker
And thanks to everybody for tuning into this week's episode. Please leave comments or questions or suggestions on the website or on Twitter. And please do rate the show on iTunes or your favorite podcast provider. So until next time, this has been the Policy This Podcast. Thanks so much for listening.