All 4 books by Edward Tufte now in
paperback editions, $100 for all 4
Visual Display of Quantitative Information
Envisioning Information
Visual Explanations
Beautiful Evidence
Paper/printing = original clothbound books.
Only available through ET's Graphics Press:
catalog + shopping cart
All 4 clothbound books, autographed by the author $150
catalog + shopping cart
Edward Tufte e-books
Immediate download to any computer:
Visual and Statistical Thinking $2
The Cognitive Style of Powerpoint $2
Seeing Around + Feynman Diagrams $2
Data Analysis for Politics and Policy $2
catalog + shopping cart
Edward Tufte one-day course,
Presenting Data and Information
Bethesda, November 8
Arlington, November 9, 10
San Francisco, December 12, 13, 14
San Jose, December 16
News graphics by Amanda Cox (NY Times)

Here is an excellent news graphic that provides enormous historical context,
describes rich variation not just a recent average, combines words and graphics,
uses annotation to call out important points, and contextualizes recent changes
in market volatility. It integrates traditional news reporting with high-resolution
(sparkline-like) graphics, and makes no distinction among words, numbers,
graphics--the idea is whatever it takes to explain something.

This superb work was done by Amanda Cox, The New York Times,
and published January 6, 2008. here

Some design suggestions: The color could be more evocative, with greater decreases
indicated by greater saturation in the red. The title type is way too big and clunky
given the refinement of the report's graphics. And is the blob of type "GRAPHIC ESSAY"
for readers from out of town? Why attract attention to the mode of production and
bureaucratic status of the report? A conventional news story doesn't have a blob of
type "WORDY REPORT WITH A PHOTOGRAPH" eaten out of the surrounding text.

Some editing suggestions: The title for the report, "The Pulse of Uncertainty," is
imprecise and off-key: pulse suggests a regular repeated rhythm, and uncertainty
differs substantially from volatility (which what the story is about). A better title is
"Market Volatility: Panics, Wars, Recessions," a title which replaces the inaccurate
descriptive words of the original title with some causal ideas. Then, in the opening
sentence of the story, the pulse metaphor is abandoned for "rockiness," which is not
rocket science. In the second sentence, it's now "tumultuous." Block those metaphors.
My outsider guess is that an (unnamed) editor did the title and first 2 sentences since they
all lack the tone, elegance, respect for the reader, and precision of thought of the rest
of the report.

-- Edward Tufte

Response to Brilliant news graphic by Amanda Cox (NY Times)

Thought some might also be interested in the following graphic about casualties in Iraq- also from NYTIMES (1/6/08):

-- DClark (email)

Response to Brilliant news graphic by Amanda Cox (NY Times)

Uncertainty, in this context, does not differ substantially from volatility. When stock returns have volatility, this means they are different from one period to another. Meaning, a priori, you do not know what the returns will be: returns are uncertain. Zero uncertainty in returns means there is (mathematically) zero volatility. Of course, this graphic charts returns; it does not chart volatility directly (variation in returns).

-- Stephanie Lee (email)

A random walk down Wall Street (revisiting Burt Malkiel)

Stephanie Lee has a lot better professional credentials on these matters than I do, but here's my notion.

If asset prices behave like a random walk, at least from the point of view of most investors, then any price event--a large swing or a small swing--is random, which is a lot like uncertain. Thus the size of the swing in prices differs from uncertainty; any market behavior (up, down, same) is uncertain. On the other hand, it is usually the case empirically that errors in estimation increase when estimating more extreme values.

Alas there's a little problem with volatility as well. Volatility has a technical meaning as an individual asset's beta, and this meaning differs from the NYT graphic's use of volatility. But probably that technical use should not deny or trump the everyday meaning of volatility as used by the Times. Here is another try at a new headline that reduces the confounding with uncertainty and with beta:

"Market Swings: Panics, Wars, Recessions"

This is not, however, very poetic with all those s sounds at the end of the last 4 words.

Surely a Kindly Contributor can do better than my latest attempt.

-- Edward Tufte

Response to Brilliant news graphic by Amanda Cox (NY Times)

What strikes me as remarkable in this graphic is the relative stability of the markets during wartime. I see no noticeable change in volatility in WWII, the Korean War, or the Vietnam War.

Where I do see the largest variations are immediately after the market reopened - in WWI and 9/11 - which gives a hint to a possible article title.

Add in the tremendous chaos surrounding the market and bank collapse in '29 and the early 30s, and a proper title becomes clear:

Government (mis)Intervention

-- Michael Round (email)

Another interesting NYT graphic

This past Saturday's (2008-02-23) online version of the New York Times includes a timely, interactive timeline, The Ebb and Flow of Movies: Box Office Receipts 1986 - 2007, recounting 20+ years of film revenue.

The artists who designed this timeline made a number of interesting choices regarding placement and color, and I am curious about this community's impressions of the graphic and the accessibility of the story being presented.

While the subject matter is, perhaps, not as relevant to many readers as the market volatility graphic, the techniques used in this timeline may be applicable to a variety of data sets that can be viewed in a time series.

Bloch, Mathew; Carter, Shan; Cox, Amanda. "The Ebb and Flow of Movies: Box Office Receipts 1986 - 2007", New York Times, 23 February 2008. Available:

-- Ari (email)

NYT Movie Revenue Graph

The Ebb and Flow of Movies is an interesting graphic with a lot of detailed information, but it plots the same thing, revenue, in two directions, up and down. This leaves artifacts in the two different profiles which together describe the total revenue.

Should the lower profile at section A stand out? If the Pirates of the Caribbean were plotted up it would form a higher peak when added to Shrek. Would we draw a different conclusion if Shrek were plotted down and Pirates up?

We see a dip in the top profile at section B. Is it really there or is it filled in by the Harry Potter plotted down?

Does the smoothing gives the appearance of more data points between the weekly revenue totals?

In the Dow Jones Industrial average graph, the daily gains are graphed up and the daily losses are graphed down. The DJIA graph resembles a seismogram, and there the vertical axis records the actual direction of the Earth's movement under the seismometer as well as its magnitude.

-- Dave Nash (email)

Another interesting graphic, "All of Inflation's Little Parts"

Ms. Cox and her team have broken "the average" consumer's spending down into proportionally-sized chunks of a circle. Interesting approach.

Would a bar chart be easier to understand? Would comparisons become easier or more difficult with a bar chart? What does the circle do for us that a bar chart cannot; alternately, what do we miss by not using a bar chart?

Reference: Cox, Amanda. "All of Inflation's Little Parts", New York Times, 2008-05-03.

-- Ari (email)

Response to Another interesting graphic, "All of Inflation's Little Parts"

That this circular chart is a mosaic rather than a pie chart makes it possible to show more meaningful detail at the center. That region isn't sacrificed to the convergence of the wedges. I hadn't seen this approach before. It's like a stained-glass window: The whole image conveys a coherent theme, but each detail offers its own story, as well. (That story emerges as one zooms in on this graphic.)

Similar treatment could be applied to a bar chart -- grouping small categories into larger ones -- and conceivably vertical distance could translate directly into relative size of the category. But I don't see the value of being able to make exact comparisons here -- it's enough to know that "Cleaning Supplies" is a relatively small part of the relatively large category of "Cost of Housing."

When being able to make exact comparisons between categories is important, this particular mosaic window on the data (using a circular window with irregularly shaped panes) would be the wrong approach. In this case, though, I think it's an elegant solution. I'm curious to hear what others have to say.

-- Cliff Tyllick (email)

Response to Excellent news graphics by Amanda Cox (NY Times)

That beautiful chart used in the NYT article is, in fact, a Voronoi Treemap - a variant of Ben Scheiderman's original concept.

Treemaps are excellent options for displaying multi-level hierarchical data, and the Voronoi variant allows for very compact and natural looking shapes. The use of a circular 'frame' for the chart is not very common, but I think it works well for their purposes.

In the end the chart uses area to encode information on spend for each budget category and subcategory as well as cell color to encode change in price. An elegant solution to a dense data set.

Michael Balzer of the University of Konstanz has been working on these for a few years now. See, for example the following paper for other beautiful examples of 'arbitrary shaped' Voronoi Treemaps:

-- Luiz Pires (email)

Response to Excellent news graphics by Amanda Cox (NY Times)

I showed this particular Voronoi treemap to a colleague, and he and I were puzzled at first by the color coding, because values near zero have cool, not neutral, hues. Then the method became clear:

  • A neutral hue means prices in that category increased at or near the overall rate of inflation.
  • A warmer hue means prices in that category increased more.
  • A cooler hue means prices in that category increased less.

So, thanks to a designer who realized that zero is an arbitrary point within the range of these data, it’s easy to identify the hottest and coolest subcategories and infer their shared traits — for example, that the hottest seem likely to be the most sensitive to fuel costs within their respective parent categories.

Economical and elegant!

-- Cliff Tyllick (email)

Response to Voronoi Treemap by Amanda Cox (NY Times)

Interesting and visually appealing?—an enthusiastic YES to both.

This graphic condenses a lot of information and uses relatively unsaturated colors in a rational color gradient, paralleling the familiar ROYGBV rainbow sequence. Of course, to see individual numbers, the reader must move the mouse, hover and read a small pop-up or zoom in and pan. This data compression tradeoff is somewhat inconvenient; however, the purpose here is not to display a table of numbers for reference. The irregular two-dimensional areas present quantification and comparison difficulties, although Cliff Tyllick makes sense in discounting this issue.

I would love to see the Voronoi Treemap as a clockwise spiral of increasingly larger random polygons as the major categories. Within the major categories, smaller spiraling polygons would depict the subcategories. For uniformity in positioning, each spiral would start in the same position, say 8 o'clock. The spiral concept would allow ordering by magnitude and give meaning to position. The mathematics necessary to generate the spirals might be a bit daunting, though.

Ari asked about the possibility of using bar charts. Consider a table, instead. The graphic contains many descriptions and numbers: there are 8 major categories (title, description, percent) and approximately 180 subcategories (description, percent and percent change). If we consider only the subcategories for simplicity, using 10 point Times Roman font, it is possible to squeeze 180 subcategories (60 rows by 3 sets of 3 columns) onto one 8½×11 page of paper, but with room for little else. It might also be necessary to shorten some item descriptions. To present these data in a typographically reasonable document would require two 8½×11 pages or one 11×17 page.

From the University of Maryland Computer Science Department, Here is an article by Ben Shneiderman treemap innovator.

One word about color gradients and coding: they are usually counter-intuitive and annoying, as the reader inevitably returns to the legend for decoding. Envisioning Information has the wonderful bathymetric chart on page 91, which shows part of the Japan Sea’s topography. The range of color to depict elevation goes from dark brown (high land) to deep blue (low sea). This color gradient is instantly understandable and extremely appealing in its intuitive simplicity. Few color gradient schemes offer such clarity—even the ROYGBV color sequence of the rainbow.

-- Jon Gross (email)

Response to Excellent news graphics by Amanda Cox (NY Times)

Amanda Cox's NYT graphics certainly have the power to astonish. See:

Three variables (oil price and consumption over time) are plotted in an x-y line chart, resulting in a figure which has no obvious corresponding mathematical equation (the line loops back across itself in a most jarring fashion!)

I thought at first that there must be a more elegant and intuitive representation of this dataset, but have begun to suspect that this format was intentionally chosen for its shock value. There appears to be a metameaning to this graphic - that oil consumption and price are not correlated in an ordinary supply-demand relationship, but are instead deeply affected by political and other factors.

Or am I reading too much into this startling figure? Your thoughts?

-- PB Turgeon (email)

Response to Excellent news graphics by Amanda Cox (NY Times)

In response to the criticism of the graphic entitled, "The Ebb and Flow of Movies: Box Office Receipts 1986 - 2007", may I direct you to what appears to be the creator of this graph type, Lee Bryon, along with a detailed whitepaper explaining the rationale behind plotting data points in both up and down directions:

-- Todd Spencer (email)

Response to Excellent news graphics by Amanda Cox (NY Times)

A great interactive chart from the New York Times (based on an idea from the OECD) showing the behaviour of the US business cycle and where we are now. The problem with static plots of these types of signals is that, as the first chart shows, they look like a random set of squiggles, but the last animated plot in particular performs the best job I've seen of demonstrating an argument for why the US is already past the turning point (whether one believes this is, of course, a different matter).

-- Will Oswald (email)

The challenge of a 4-quadrant "function" -Response to Owald & Turgeon on news graphics by Amanda Cox (NY Times)

When looking at the composite print version (NYT 7/5/09 Business p. 5) of the chart noted by Will Oswald, I had a reaction similar to the one PB Turgeon had to a similar chart on June 9, 2008: "three variables [here: amount and trend of industrial production over time] are plotted in an x-y line chart, resulting in a figure which had no obvious corresponding mathematical equation (the line loops back across itself...)"

- I think that Turgeon's comment is worth reconsidering. If X and Y variables have no mathematical relationship, then why use X-Y coordinates, with their typical use in showing relationships of variables, as the template for demonstrating this data? This 4-quadrant presentation seems to be cognitively and visually confused and confusing. If I'm mistaken in that view, can you direct me to a text or article that explains how to plot and understand data that is displayed as cycling in differing patterns through four quadrants?

-- JS Gellman (email)

Response to Excellent news graphics by Amanda Cox (NY Times)

"No obvious corresponding mathematical equation" is a much weaker statement than "no mathematical relationship between X and Y variables".

In the case of the NYT business cycle graph, the X and Y variables are most certainly mathematically related: X shows the rate at which Y is changing , and it's actually quite a simple relationship: the X-variable shows how quickly Y is changing over time (i.e., X = dY/dt, or nearly so - there are a few practical details that mean this isn't exact).

However, while we know how X and Y relate *to one another*, that doesn't here give us a mathematical equation that lets us predict them as functions of time.

This sort of "phase space" representation is often used when examining the behaviour of dynamic systems subject to feedback. A simple example is a pendulum or a vibrating spring: if you use the X axis to plot position and the Y axis to plot velocity, the plot will describe a clockwise circle. In that particular case, the relationship between X and Y *does* allow us to predict them as a function of time: it can be shown that X=sin(t), Y=cos(t) (leaving out constant terms).

Phase space plots are popular because they often expose underlying characteristics of the system that are difficult to see when you look at just one variable at a time. They're particularly useful in the study of chaotic systems (as economics is) - in these systems you can't predict exactly what will be happening a year from now, but you might at least be able to set some bounds on it. (Imagine a slow child wandering around a carousel that has an erratic motor - we can't tell exactly where the child will be at any given time, but we know they won't be outside the bounds of the carousel, and there's a limit on how quickly they can move.)

I don't have a good intro reference handy, but and might be of some help.

-- Geoffrey Brent (email)

Response to Excellent news graphics by Amanda Cox (NY Times)

Dear ET,

here is an interesting collection of news and science graphics created by the 13pt design agency in New York ( for the New York Times.

Here is an example - A Deep Calm (July 21, 2009)for the NYT article Is the Sun Missing Its Spots? (

Best wishes


-- Matt R (email)

Response to Excellent news graphics by Amanda Cox (NY Times)

Another informative visual from the NY Times ...

-- Michael Round (email)