A Thousand Words for a Picture

Is the overvaluation of GIS
disrupting a critical balance in crime analysis?

  Christopher W. Bruce
Cambridge, MA Police Department

“As geographers crowd into the edges of their maps parts of the world which they do not know about, adding notes in the margin to the effect that beyond this lies nothing but sandy deserts full of wild beasts, dragons, and unapproachable bogs.”

                                            --Plutarch, The Life of Theseus

Saito’s purchase was the zenith of the art overvaluation that permeated auction houses in the late 1980s and early 1990s, cheating a number of less affluent individuals out of artwork that they should have otherwise reasonably been able to afford. In 1997, the United States Justice Department began an investigation into both Christie’s and Sotheby’s, eventually charging that the two auction houses had conspired to fix the commissions that they charged both buyers and sellers, dating back to the early 1990s.

It’s a new millennium, a different industry, and a different kind of graphic, but pictures are similarly being overvalued in the crime analysis world, sometimes hampering the efforts of skilled analysts everywhere. A Google® search for the terms “picture is worth a thousand words” and “GIS” returns no less than 240 web pages. It’s been said so often that not only is it a cliché, but it’s a cliché to mention that it’s a cliché. (“It’s a tired cliché, but a picture is worth a thousand words,” says Deena Bowman-Jamieson, information systems analyst for the San Diego Police Department, describing the benefits of GIS in the January/February 2001 edition of Mercator’s World.)

The problem with clichés is that no one ever stops to question whether or not they are true, and in the case of crime mapping, it isn’t. A picture of crime isn’t worth a thousand words of qualitative analysis any more than a picture of Dr. Gachet is worth $82.5 million, and thinking so is hurting the profession. Further, crime mapping has the insidious trait of masking issues of data quality and interpretation—issues that were so insistent when viewing tabular data—with the result that it is increasingly becoming a cover sheet; an opaque plotter-printed piece of paper that a Crime Analysis Unit hangs over its front window to hide the chaotic mess inside.

When the Tool Supercedes the Product

Crime analysis, in contrast, has existed as a profession for a quarter century, and as a concept for almost 40 years. Its origins are centuries old. Yet the whole of crime analysis has never had a national center for its study and application. Similarly, the National Law Enforcement & Corrections Technology Center has offered a Crime Mapping & Analysis Program (CMAP) since 1998, but no national nonprofit agency offers a straight crime analysis program.

 Crime analysis does have an international association — appropriately called the International Association of Crime Analysts (IACA) — which receives no government funding other than the dues of its members. The IACA holds a conference in the fall each year; in 1997, the Crime Mapping Research Center began doing the same thing. In 2000, the IACA drew some 157 attendees to its conference; the CMRC conference, held a month later, drew 493 — 172 of whom were employees of police agencies. It’s a telling moment in our profession when a conference dedicated to the use of a particular tool of crime analysis draws so much more attendance than a conference dedicated to the whole of crime analysis; it is comparable to an American Medical Association symposium on electrocardiograms drawing more registrants than the general AMA conference.

Other signs and portents: a list of current job openings maintained by the International Association of Crime Analysts shows 100% more “GIS Analyst” positions available than “Crime Analyst” positions, and the GIS Analysts have an average top salary 37.5 percent higher than the crime analysts. [Author's note: this was true during the Fall of 2000 and Winter of 2001; it has not been true since then, which admittedly weakens this argument.] Analysts are frequently dazzled by current developments in GIS technology, while other technologies of importance to crime analysis, including Records Management Systems, remain in the Paleolithic era.

These signs point to a growing culture in which crime mapping—a tool of crime analysis—is superceding the analysis itself. One can easily see how this might happen: crime mapping is flashy, fancy, and fun. It looks good on a computer screen and even better printed in 36x44. Training, software, and equipment can cost a lot of money, so a heavy concentration of the analyst’s time on crime mapping justifies the expense. And it’s hard to do—impenetrably hard for the layperson. Anyone can read a stack of crime reports and write a bulletin, but crime mapping—like forecasting—is one of those areas of crime analysis that borders on wizardry.

Data and Analysis

 The distinction between data and analysis is crucial in this profession, particularly in crime mapping. Data is raw information—a collection of facts in a database, spreadsheet, or on a piece of paper. Analysis is a product of data. It places data in context, signifies it, explains it, and—in our profession—suggests a course of action. Data answers the questions of who, what, when, where, and how. Analysis answers the question of why, and of what it all means.

A dangerous thing happened to the profession of crime analysis—as well as to the field of problem-oriented policing at large—when people began thinking of maps as analysis rather than data. Usually, a map is simply data—data that has been rearranged, plotted geographically, and printed in color, but data nonetheless. (Not all crime mapping, of course, is meant to result in a visible product: much of the field involves tools and techniques geared towards manipulating data to answer questions, but the result is still only data until the analyst applies his or her own mind to the data and generates analysis.)

(The same thing is true of statistics. Crime in the United States is an enormous collection of raw data, with various chapters and sections mislabeled “analysis.” Many crime “analysis” units produce little more than raw data on a regular basis. This led us, in 1999, to create the Fifth Commandment of Crime Analysis: “Thou Shalt Never Present Statistics (or Maps) by Themselves.”)

To illustrate these principles—and to prove that a picture certainly isn’t worth a thousand words—we’ve presented two items, side by side, below. One is a map; the other is a bit of text. Both concern an increase in car burglaries over the past two months around a shopping mall. Look them over and ask yourself which is more valuable to its intended purpose—the eradication of a crime pattern. (Assume that both items are completely correct.)

Exhibit 1

Item 1

Item 2 "Car burglaries are up around the mall for the holidays. We should put more police officers there."

 The counter-intuitive, though correct, answer is that Item 2 is more valuable than item 1. Item 1 is a display of data, in thematic form, on a map. It is raw data, showing only the numbers of crimes by grid block. It conveys no information that is not contained, or at least suggested, in Item 2. And it took 10 minutes to create.

 Item 2 is analysis. We’re not claiming that it’s good analysis, but it is analysis, and it was written in 15 seconds. It conveys more information than the map, namely: 

1. By saying that car burglaries are “up,” it suggests that not only are the raw numbers high, but that they are higher than average for this particular area.
2. The words “for the holidays” are an attempt to answer the “why” question—to explain what the data is showing. A layperson would understand the implication: the cars burglarized belong to shoppers who park near or at the mall, perhaps leaving holiday presents in the cars while they shop.
3. “We should put more police officers there” satisfies the analyst’s duty to recommend—either implicitly or overtly—a course of action. It is the second-to-last step in the crime analysis process.

  The map requires interpretation; the text does not. No chief or patrol commander will have to furrow his brow for even a second to comprehend what the text is saying. The map needs to be printed and presented to its audience; the text can be fired off in an e-mail or spoken in the hallway. The text is important; the map is superfluous.

If the analysis presented in item 2 were good analysis, it would include further information—a description of the most common cars targeted, means of entry, and property stolen; the most likely time frame for an incident to occur; suspect descriptions, if any exist; a list of individuals suspected or arrested for car burglary incidents in this area in the past, particularly during the holiday season; perhaps a report of current patrol operations in the area. All of these pieces of information are incredibly valuable to a police officer or executive trying to suppress or eliminate a crime pattern, and if the analyst tried to represent them geographically, it would take at least half a dozen maps, none of which would be more efficient than plain descriptive text.

The choice of whether to present the map or the text does not have to be an either/or proposition for the analyst. If he has a chance, he should present both—the map accompanied by the analysis. (Few analysts have such a high level of trust from their agencies that they could present a recommendation to increase police patrols with no supporting data!) If the analyst used the map in Item 1 to arrive at the conclusions in Item 2, then he has made good use of crime mapping. Our point is not that crime mapping is entirely worthless, nor that it doesn’t have it’s place as either a tool or product of analysis—only that the map never eliminates the need for qualitative analysis. Qualitative analysis, if need be, can stand alone and still convey the essence of a crime pattern with astonishing power and clarity; maps almost never can.

Legitimate Uses for Crime Mapping in Crime Analysis

  Crime mapping assists crime analysis in two areas:  

1. Analyzing information
2. Presenting information

  As a tool for analyzing information, crime mapping is both overemphasized and underemphasized. Underemphasized because, since its object is not to create a visible product, it’s often “down and dirty” — no pretty colors, no north arrow, no scale bar, no meticulously-crafted layout window. Books and articles meant to entice analysts into using crime mapping rarely show any screen shots of maps used to analyze information because they’re often bare and crude and thus counter-productive to the intent of the author.

Exhibit 2

A down-and-dirty crime map. The analyst has thrown five auto thefts onto his streets layer to check geographic proximity, and has numbered them in order of occurrence. There are only five incidents, so he has plotted them all with the GIS program’s “find” feature rather than importing them from his records management system. He has labeled only two streets because he knows what the other ones are. He knows where north is, and approximately how far it is from one point to the next. He is familiar with the area, so he has no need for supporting data layers such as parks, buildings, and railroad tracks. After he analyzes the pattern more closely, he may simply discard the work or clean it up so that it can be presented along with his analysis. Needless to say, this type of unadorned map will be unlikely to find its way into a crime mapping book.

Crime mapping as an analysis tool is overemphasized because much of the literature assumes that mapping is the best way to analyze particular types of data, when often it is more efficient for the analyst to simply look at numbers. By over-relying on GIS methods, the analyst retards his own ability to scan a series of data and quickly comprehend the information shown. Most thematic maps are a poor replacement for a trained analyst’s own mind and judgment.

Exhibit 3

Sector Day Shift Eve. Shift Night Shift Total 1 30 22 5 57 2 25 27 10 62 3 12 32 8 52 4 10 10 10 30 5 8 17 21 46 Total 85 108 54 247

In this example, the analyst has taken the trouble to create a thematic map that shows pie charts for each sector of the city, indicating the number of burglaries on each shift in that sector. It makes a nice looking map, but ask yourself whether anything is gained by the mapping process. A good analyst should be able to scan the data in tabular form (on the right) and pick out the information he needs much more quickly. Note that both items, again, are merely data, and require further analysis for the crime analyst to justify his title. (Why does Sector 2 have the most burglaries, and what are their characteristics? Why is Sector 5 the only sector with a high percentage of burglaries on the last half?)

Patterns can be quickly identified through the use of crime mapping, but only those that are geographically based. Many crime patterns exhibit no geographic similarities and must be found through a careful analysis of the modus operandus. Crime mapping greatly facilitates the identification of geographically based patterns, but it will never be a substitute for a thoughtful, regular review of the agency’s crime reports by a sharp analytical mind. (A review of crime patterns in Cambridge over the summer of 2000 revealed that 34% could not be found through crime mapping — they exhibited no geographic clustering.)

  When presenting information, crime maps are best used as a visual aid to accompany qualitative analysis. A map clarifies information in ways that text cannot. Officers may have trouble digesting a bit of text such as, “The pattern is most active in the area bordered by First Street, Main Street, Fifth Street, and Elm Street,” while a clear map with the border carefully outlined and the individual crimes plotted within brings this description to life.

  A second use for maps as a product is to support the conclusions the analyst has reached. A police executive would not be acting unreasonable if he asked for some material to support an analyst’s recommendation to increase patrol in Sector 2 by 50% during the month of June. Various maps can provide this supporting material and show how the analyst arrived at his analysis of the situation.

Again, I emphasize that maps must never be presented alone. Maps are data only, and the analyst is not living up to his name if he does not endeavor to interpret and explain the data.

 
Exhibit 4 shows a good example of a crime map as a product. The analyst has identified robbery hot spots in his jurisdiction and has included textual analysis of each of the hot spots.

Exhibit 4

Conclusions 

Responsible use of crime mapping in crime analysis requires the adherence to three basic principles: 

1. Remember that crime analysis is the profession; crime mapping is the tool. Watch out for an inordinate amount of time spent answering the question of “where” to the exclusion of the questions of “who,” “what,” “when,” “how,” and “why.”

2. Before using crime mapping—either as a tool of analysis or as a product—ask yourself whether the same goal could be achieved more simply and more clearly without the use of mapping tools.

3. Make sure that you do your job as a crime analyst by analyzing and interpreting geographic information. A map of crime is not a product of crime analysis; a map accompanied by qualitative analysis is.

Crime mapping is a powerful, complicated tool, but power and complexity do not necessarily equate with importance, and many worthless products have issued from powerful, complicated tools. Try to keep some balance in your training, in your work focus, and in your product.