Data Visualization in Sociology

SO40CH05-MoodyHealy ARI 4 July 2014 13:29

Data Visualization

in Sociology

Kieran Healy and James Moody

Sociology Department, Duke University, Durham, North Carolina 27708;

email: [email protected], [email protected]

Annu. Rev. Sociol. 2014. 40:105–28

First published online as a Review in Advance on

June 6, 2014

The Annual Review of Sociology is online at

soc.annualreviews.org

This article’s doi:

10.1146/annurev-soc-071312-145551

⃝ 2014 by Annual Reviews.

Keywords

visualization, statistics, methods, exploratory data analysis

Abstract

Visualizing data is central to social scientiﬁc work. Despite a promising

early beginning, sociology has lagged in the use of visual tools. We

review the history and current state of visualization in sociology. Using

examples throughout, we discuss recent developments in ways of seeing

raw data and presenting the results of statistical modeling. We make a

general distinction between those methods and tools designed to help

explore data sets and those designed to help present results to others.

We argue that recent advances should be seen as part of a broader shift

toward easier sharing of code and data both between researchers and

with wider publics, and we encourage practitioners and publishers to

work toward a higher and more consistent standard for the graphical

display of sociological insights.

105

Annu. Rev. Sociol. 2014.40:105-128. Downloaded from www.annualreviews.org

Access provided by Duke University on 08/09/17. For personal use only.

Click here for quick links to

Annual Reviews content online,

including:

• Other articles in this volume

• Top cited articles

• Top downloaded articles

• Our comprehensive search

Further

ANNUAL

REVIEWS

SO40CH05-MoodyHealy ARI 4 July 2014 13:29

INTRODUCTION

From the mind’s eye to the Hubble telescope,

visualization is a central feature of discovery,

understanding, and communication in science.

There are many different ways to see. Visual

tools range from false-color photographs of

telescopic images in astronomy to reconstruc-

tions of prehistoric creatures in paleontology.

In the statistical sciences, images are often more

abstract than models of ﬁghting dinosaurs—

depending asthey must onconventions that link

size, value, texture, color, orientation, or shape

to quantities (Bertin 1967 [2010]). But statisti-

cal visualizations are nonetheless critical to pro-

moting science. One need only think of the now

iconic hockey-stick diagram of earth tempera-

ture for a clear case (Mann et al. 1999). De-

spite its ubiquity in most of the natural sciences,

visualization often remains an afterthought in

sociology.

In this article, we review the history and

current state of data visualization in sociology.

Our aim is to encourage sociologists to use

these methods effectively across the research

and publication process. We begin with a brief

history, then present an overview of the the-

ory of graphical presentation. The bulk of our

review is organized around the uses of visualiza-

tion in ﬁrst the exploration and then the presen-

tation of data, with exemplars of good practice.

We also disc uss workﬂow and software issues

and the question of whether better visualization

can make sociological research more accessible.

SOCIOLOGY LAGS

First, why are statistical visualizations so com-

mon in other ﬁelds and rare in sociology? Al-

though model summaries offer exacting preci-

sion in expressing particular quantities—such

as the slope of a line through data points—

getting a sense of multiple patterns simulta-

neously is typically easier visually. The point is

made forcefully by Anscombe’s (1973) famous

quartet, reproduced in Figure 1a.Eachdataset

contains 11 observations on two variables. The

basic statistical properties of each data set are

almost identical, up to and including their bi-

variate regression lines. But when visualized as a

scatterplot, the differences are readily apparent

(see also Chatterjee & Firat 2007). Lest we think

such features are conﬁned to carefully con-

structed examples, consider Jackman’s (1980)

intervention in a debate between Hewitt (1977)

and Stack (1979) over a critical test of Lenski’s

(1966) theory of inequality and politics, repro-

duced in Figure 1b. The argument is won at

aglance,astheﬁgureshowsthattheseem-

ingly strong negative association between voter

turnout and income inequality depends entirely

on the inclusion of South Africa in the sample.

Given the power of statistical visualization,

then, it is puzzling that quantitative sociology is

so often practiced without visual referents. One

need only compare a recent issue of the Ameri-

can Sociological Review or the American Journal of

Sociology to Science, Nature, or the Proceedings of

the National Academy of Science to see the radical

difference in visual acuity. It is common for the

premier journals in sociology to publish articles

with many tables, but no ﬁgures. The opposite

is true in the premier natural science journals.

There, a key ﬁgure is often the heart of the ar-

ticle. In Nature,forexample,theonlinetable

of contents includes a thumbnail of the central

ﬁgure to serve as the link to the rest of the paper.

It has not always been so. Early in the history

of the discipline, data visualizations were com-

mon and not appreciably out of step with the

wider scientiﬁc community. Exemplars of bar

charts (Hart 1896), line graphs (Marro 1899),

parametric density plots and dot plots with

standard errors (Chapin 1924), scatterplots

(Sletto 1936), and social network diagrams

(Lundberg & Steele 1938) are easy to ﬁnd in

early sociological journal articles. Du Bois’s

(1898 [1967]) The Philadelphia Negro is ﬁlled

with innovative visualizations, including choro-

pleth maps, table-and-histogram combinations,

time series, and others. But somewhere along

the line sociology became a ﬁeld where sophis-

ticated statistical models were almost invari-

ably represented by dense tables of variables

along rows and model numbers along columns.

Though they may signal scientiﬁc rigor, such

106 Healy

Moody

Annu. Rev. Sociol. 2014.40:105-128. Downloaded from www.annualreviews.org

Access provided by Duke University on 08/09/17. For personal use only.

SO40CH05-MoodyHealy ARI 4 July 2014 13:29

5.0

7.5

10.0

12.5

5.0

7.5

10.0

12.5

5 10 15 5 10 15

x values

y values

For all panels, N = 11; mean = 7.5; regression: Y = 3 + 0.5(X); r = 0.82.

SE of slope estimate: 0.118, t = 4.24; sum of squares (X − X): 100

South Africa

Bivariate slope including South Africa (N = 18)

Bivariate slope excluding South Africa (N = 17)

a Anscombe’s quartet (1973) b Jackman (1980)

Figure 1

Visualizations reveal model summary failures: (a) Anscombe’s quartet shows how statistically identical data sets can look very different;

(b) visualization from Jackman (1980) decisively demonstrates the inﬂuence of outlying data points in an analysis.

tables can easily be substantively indecipherable

to most readers and perhaps at times even to

authors. The reasons for this are beyond the

scope of this review, although several possi-

bly complementary hypotheses suggest them-

selves. First, to the extent that graphical im-

agery was thought of as descriptive, statistical

images may have been collateral damage in the

war between causal-inferential modeling and

descriptive reportage. Second, ﬁgures may have

seemed unsophisticated. The very clarity of a

(good) ﬁgure made the work seem too sim-

ple. Third, and more charitably, visualization

in sociology might have been a victim of the

ﬁeld’s relatively rapid embrace of quantitative

methods. American sociology adopted sophis-

ticated modeling techniques quite early com-

pared with other social sciences. The range

and variety of its research questions and data

sources meant that the statistical tool kit in so-

ciology in the late 1960s and into the 1970s

was more varied than in economics or psy-

chology at the time and much more developed

than what was then current in political science.

But this was also a period when the visual-

ization tools of statistical software lagged well

behind their strictly computational abilities.

Conventions of data presentation may have

standardized at a time when the possibilities for

visualization were narrower. Finally, some of

the resistance to ﬁgures may have come from

the fact that the tables in early journal articles

and monographs often contained actual data

rather than summaries or model results. In a re-

view of a history of graphical methods in statis-

tics written in 1938, John Maynard Keynes re-

marked that he wished the author

could have added a warning, supported by

horrid examples, of the evils of the graphical

method unsupported by tables of ﬁgures. Both

for accurate understanding, and particularly to

facilitate the use of the same material by other

people, it is essential that graphs should not be

www.annualreviews.org

•

Data Visualization in Sociology 107

Annu. Rev. Sociol. 2014.40:105-128. Downloaded from www.annualreviews.org

Access provided by Duke University on 08/09/17. For personal use only.

SO40CH05-MoodyHealy ARI 4 July 2014 13:29

published by themselves, but only when sup-

ported by the tables which lead up to them. It

would be an exceedingly good rule to forbid

in any scientiﬁc periodical the publication of

graphs unsupported by tables. (Keynes 1938,

p. 282, emphasis added)

To speak anachronistically, here Keynes is

arguing that economists need the underlying

data along with the visual summary for the sake

of reproducibility. We are now at a point when

the volume of data used in a typical quantita-

tive article far exceeds what can be presented in

a series of tables. But Keynes’s point is worth

bearing in mind. The utility of visualization

methods—in particular their ability to effec-

tively summarize large quantities of data or so-

phisticated modeling techniques—is partly de-

pendent on related advances in our ability to

easily share data and reproduce analyses. If data

are accessible as needed, using ﬁgures instead

of tables becomes much easier. Not coinci-

dentally, this is another area where sociology

has lagged behind other social sciences (Freese

2007).

Whatever their relative importance, the net

result of these processes for sociology has been

a training and publication standard that rarely

includes graphical treatments of statistics. New

students are typically not taught to think about

graphics and statistics in a consistent, coherent

way.

Our argument is not that sociologists should

be producing more visualizations just because

everyone else is doing it. Indeed, as we discuss

below, there is considerable debate about what

sort of visual work is most effective, when it can

be superﬂuous, and how it can at times be mis-

leading to researchers and audiences alike. Just

like sober and authoritative tables, data visual-

izations have their own rhetoric of plausibility.

Anscombe’s quartet notwithstanding, summary

statistics and modeling can be thought of as

tools that deliberately simplify data to let us see

past the cloud of data points. We do not think

visualization will give us the right answer sim-

ply by looking. Rather, we should think about

how visualization might be more effectively in-

tegrated into all stages of our work. Software

now makes routinely generating ﬁgures easier

than ever. Even if many disciplinary journals

still lag in their editorial desire or ability to

present good data visualizations, we argue that

it is time for these methods to be fully integrated

into sociology’s research process.

VISUALIZATION IN PRINCIPLE

Book-length treatments of good statistical visu-

alization practice abound. Their content ranges

from the more theoretical—emphasizing, for

instance, the nature and origins of visual

conventions—to more pragmatic collections

of current best practices meant to serve as

an inspiration to practitioners. In between

are efforts to codify practice and develop

taste, and guides to working implementations.

The most inﬂuential general treatments are

probably Bertin’s (1967 [2010]) Semiology of

Graphics,Cleveland’sThe Elements of Graphing

Data (1994) and Visualizing Data (1993), and

Wilkinson’s (1995 [2005]) The Grammar of

Graphics. Overviews of contemporary practice

can be had in Few (2009, 2012) and Yau (2012).

There are also several books based speciﬁcally

on visualization techniques within a particular

software program, such as Friendly (2000) for

SAS, Mitchell (2012) for Stata, Murrell (2011)

for R, and Kleimean & Horton (2013) for

comparisons of multiple programs. Sometimes

the graphical capabilities of particular software

applications are loosely related to the more

theoretical work, taking from them a concern

with aesthetic principles and possibly speciﬁc

sorts of plots. In other cases , the linkage is

closer. Sarkar (2008) describes a data visu-

alization package for R that closely follows

Cleveland’s ideas (and some earlier associated

software), and Wickham (2009, 2010) describes

a s oftware package for R that implements and

extends principles worked out in Wilkinson’s

(1995 [2005]) The Grammar of Graphics.

The conceptual literature is deep and com-

prehensive, although its representatives do not

always speak in one voice. This is to be expected

in an area where theoretical development

108 Healy

Moody

Annu. Rev. Sociol. 2014.40:105-128. Downloaded from www.annualreviews.org

Access provided by Duke University on 08/09/17. For personal use only.

SO40CH05-MoodyHealy ARI 4 July 2014 13:29

involves judgments of taste. The best-known

critic and tastemaker by far in the ﬁeld is

Edward R. Tufte. It is fair to say that The Visual

Display of Quantitative Information (Tufte 1983)

is a classic in the ﬁeld, and its three follow-up

texts are also widely read (Tufte 1990, 1997,

2006). Described as “self-exemplifying” (Tufte

2006, p. 10), the bulk of the work is a series

of negative and positive examples with more

general principles (or rules of thumb) extracted

from them rather than a direct guide to practice,

akin more to a reference book on ingredients

than to a cookbook for daily use in the kitchen.

At the same time, Tufte’s early work in politi-

cal science shows that he applied his ideas well

before codifying them in this way. His Political

Control of the Economy (Tufte 1978) combines

data tables, ﬁgures, and text in a manner that

remains remarkably fresh almost 40 years later.

Across his work, Tufte preaches a consistent

set of principles, though they vary in their de-

gree of speciﬁcity. Thus,

Graphical excellence is the well-designed pre-

sentation of interesting data—a matter of sub-

stance,ofstatistics,andofdesign. ... [It] consists

of complex ideas communicated with clarity,

precision, and efﬁciency. ... [It] is that which

gives to the viewer the greatest number of

ideas in the shortest time with the least ink

in the smallest space. ... [It] is nearly always

multivariate. ... And graphical excellence re-

quires telling the truth about the d ata. (Tufte

1983, p. 51)

Tufte illustrates the point with Charles Joseph

Minard’s famous visualization of Napoleon’s

march on Moscow, reproduced in Figure 2.

He remarks that this image “may well be the

best statistical graphic ever drawn,” and argues

that it “tells a rich, coherent story with its mul-

tivariate data, far more enlightening than just

a single number bouncing along over time. Six

variables are plotted: the size of the army, its lo-

cation on a two-dimensional surface, direction

of the army’s movement, and temperature on

various dates during the retreat from Moscow”

(Tufte 1983, p. 40). It is worth noting how dif-

ferent Minard’s image is from most contem-

porary statistical graphics. Until recently, these

Figure 2

Minard’s visualization of Napoleon’s advance on and retreat from Moscow is a classic of visualization, but its design is in many ways

atypical.

www.annualreviews.org

•

Data Visualization in Sociology 109

Annu. Rev. Sociol. 2014.40:105-128. Downloaded from www.annualreviews.org

Access provided by Duke University on 08/09/17. For personal use only.

SO40CH05-MoodyHealy ARI 4 July 2014 13:29

have tended to be generalizations of the scatter-

plot or barplot, either in the direction of seeing

more data or seeing the output of models. The

former looks for ways to increase the volume of

data visible, the number of variables displayed

within a panel, or the number of panels dis-

played within a plot. The latter looks for ways

to see results of models—point estimates, con-

ﬁdence ranges, predicted probabilities, and so

on. Tufte (1983, p. 177) acknowledges that a

tour de force such as Minard’s “can be described

and admired, but there are no compositional

principles on how to create that one wonderful

graphic in a million.” The best one can do for

“more routine, workaday designs” is to suggest

some guidelines such as “have a properly cho-

sen format and design,” “use words, numbers,

and drawing together,” “display an accessible

complexity of detail,” and “avoid content-free

decoration, including chartjunk” (p. 177).

Among this set of general goals are some

speciﬁc details that can be employed to good

use across applications. This includes extensive

use of layering and separation, for example,

building on the insights of good cartography.

Judicious use of stroke weight and color allows

one to layer multiple meanings on a single

visual plane. The ability to successfully pull

off such effects depends on use of the smallest

effective difference—lighter lines, smaller color

variations, and simpler textures. It has long

been a complaint of chart designers that accom-

plishing this often means working very much

against the (highly detailed, drop-shadowed,

rich, Corinthian leather) grain of t he default

settings in spreadsheet or other chart-making

applications. Comparison and evaluation are

often enhanced by the use of many small

multiples—plots that repeatedly display some

reference variable or relationship (e.g., gross

domestic product versus health care costs over

time) and iterate across some other variable of

interest (e.g., country) in an ordered fashion

(see also Bertin 1967 [2010], pp. 217–45). The

use of such multiples highlights the notion of

parallelism that allows a reader to carefully

compare across instances of similar-but-

crucially-different items. Combined, these fea-

tures facilitate a simultaneous micro and macro

reading where key points are clearly communi-

cated at the surface, but deeper meaning is ob-

tained through careful review and exploration.

A common complaint about Tufte’s work

is that there are so few direct instructions.

Busy cooks want a cookbook, not a picture

of a fantastic meal. The tendency for the

codiﬁcation of data visualization to vacillate

between overly abstract maxims and overly

speciﬁc examples is characteristic of any craft

where a practical sense of how to proceed—a

taste or feeling for the right choice—matters

for successful execution. A long-standing and

plausible response to the problem is to have the

designer make many of the judicious choices in

advance and then embed them for users in the

default settings of graphics applications. Given

that graphical software aimed at regular users

has been around for several decades now, how-

ever, these efforts have proven less successful

than initially hoped. In the foreword to the

new edition of Semiology of Graphics, Howard

Wainer (2010, p. xi) reﬂects on the hope he

and others once felt that easy-to-use graphical

tools and software would lead to better general

practice by way of smarter defaults. But, he

argues, this has not happened. In the end, high-

quality graphical presentation requires crafting

a deliberately designed message rather than

accepting the pre-established setting. Recent

theoretical work explicitly recognizes the limits

of relying on defaults. Following Wilkinson in

implementing ggplot’s “grammar of graphics”

for R, Wickham (2010, p. 3) notes that the

analogy to grammar is useful because although

“[a] good grammar will allow us to gain insight

into the composition of complicated graphics,

and reveal unexpected connections between

seemingly different graphics[,] ...there will

still be many grammatically correct but non-

sensical graphics. ... [G]ood grammar is just

the ﬁrst step in creating a good sentence.”

If software defaults cannot enforce the ele-

ments of good taste, the next best—or maybe

better—thing is a means to easily expose the

mechanics of good practice. One of the most

positive developments in statistical software

110 Healy

Moody

Annu. Rev. Sociol. 2014.40:105-128. Downloaded from www.annualreviews.org

Access provided by Duke University on 08/09/17. For personal use only.

SO40CH05-MoodyHealy ARI 4 July 2014 13:29

over the past 15 years has been its integration

with a much broader set of tools built to

facilitate the sharing of both data and code.

The ﬁrst wave of modern statistical graphics

and information design could convey, in print,

the general principles and the quality products.

But the crucial piece in between—the design

process and practical assembly—remained

opaque. Subsequently, communities of users

began to share not just output but code much

more widely, whether under the auspices of

a for-proﬁt developer (as in the case of Stata)

or actively backed by free or open-source

licensed platforms (as with R) or expert user

blogs (http://sas-and-r.blogspot.com, http://

ﬂowingdata.com, http://www.r-statistics.

com/tag/visualization). Some of these have

developed into comprehensive references

aimed at the practicing researcher (Chang

2013). Most recently, pastebins and software

development platforms backed by distributed

version control systems—most notably

Github—have made sharing code both techni-

cally much easier and normatively expected.

As with the move toward replication data

sets, everyday sharing of code allows novices

to look behind the curtain much more easily

than before. And perhaps unlike the earlier

emphasis on accepting sensible defaults, it

encourages new users to tinker with various

methods and learn by doing. In many cases,

software now allows users to control very

detailed layout elements in their program

scripts, which (with a little extra language

work) allows one to override defaults with

principled graphical choices. This ongoing

integration of guidebooks, how-to websites,

code repositories, and fully reproducible

examples is a major step forward for improving

visualization practice. As one particularly well-

developed example among many, UCLA’s

Institute for Digital Research and Education

has a large library of worked graphical examples

implemented across several statistics packages

(http://www.ats.ucla.edu/stat/dae). Finally,

because most statistical packages can now pro-

duce graphics as editable vector graphics ﬁles,

one can use any graphical editor to ﬁne-tune

elements (such as line thickness, greater

subtlety in color selection, etc.) for production.

These developments do not make questions

of judgment and good practice go away. Sta-

tistical visualization needs to be thought of as

part and parcel of analysis and presentation. We

should be crafting visualizations thoughtfully in

the same way we craft arguments or build mod-

els. Resources of this sort cannot by themselves

guarantee that code snippets will not simply be

mechanically copied or inappropriately applied

by users looking for a shortcut to a good out-

come. But, to paraphrase Keynes from a dif-

ferent context, they do seem to promise if not

civilized visualization, at least the possibility of

civilized visualization.

VISUALIZATION IN PRACTICE

We have argued that there are several promis-

ing ways that general principles of visualization

can become more tangible in everyday use. We

now turn to the question of current practice

in a little more detail. Here we follow the

common distinction between visualization

for exploration versus presentation of a ﬁnal

ﬁnding. The former is meant for internal

consumption, as the researcher examines the

data to ﬁgure out what is going on; the latter

is designed to convince a wider audience. Nat-

urally, these processes overlap to some degree.

The general principles covered in the previous

section—regarding clarity, honesty, showing

the data, and so on—apply equally to both the

backstage and frontstage of visualization work.

But what is needed in each case does differ.

Some recent developments on each side are

worth highlighting.

Exploring the Data

Graphical methods are now well integrated into

the process of checking assumptions and ro-

bustness in most statistical packages and are

often generated by default. Figure 3 shows a

typical example of some diagnostic plots of an

ordinary least squares regression. They were

produced on demand and by default, with no

www.annualreviews.org

•

Data Visualization in Sociology 111

Annu. Rev. Sociol. 2014.40:105-128. Downloaded from www.annualreviews.org

Access provided by Duke University on 08/09/17. For personal use only.

SO40CH05-MoodyHealy ARI 4 July 2014 13:29

Fitted values

Residuals

Residuals vs Fitted

Theoretical Quantiles

Normal Q−Q

Scale−Location

Leverage

Residuals vs L

Standardized Residuals

Fitted values

Figure 3

Default diagnostic

plots for a linear

model: (a)R,(b)SAS.

Though automatically

produced, both panels

present information

clearly and with

judicious use of

labeling and color.

112 Healy

Moody

Annu. Rev. Sociol. 2014.40:105-128. Downloaded from www.annualreviews.org

Access provided by Duke University on 08/09/17. For personal use only.

SO40CH05-MoodyHealy ARI 4 July 2014 13:29

further tweaking or polishing. Note that al-

though we voiced some skepticism above about

the ability of defaults to shape practice, these

plots are models of clarity. They could be

called into service for presentation purposes in

a pinch. Their real utility, however, is the ease

with which they can be produced and viewed

as part of one’s everyday workﬂow as a social

scientist: With tools like these, comments on

outliers such as Jackman’s (1980) should never

again be necessary.

Diagnostic plots of this kind are—in

principle—what you look at after a model has

been chosen. They are conﬁrmatory rather than

strictly exploratory. Advocacy of exploratory

data analysis (EDA), of looking carefully and

creatively before modeling, is most closely as-

sociated with John Tukey (1972, 1977). His-

torically, EDA has been closely tied to the rise

of graphical capabilities in statistical comput-

ing, particularly tools that allow rapid interac-

tive visualization. A mild sense of unease with

EDA is a feature of the statistical literature. The

approach is explicitly inductive and concerned

with exploring data in a relatively freewheeling

fashion as an aid to discovery, which at times can

seem uncomfortably opportunistic or unstruc-

tured. To working social scientists these are of-

ten virtues, but statistics is also th e discipline

where the avoidance of spurious associations is

a major focus of technical work.

As data sets have continued to increase in

both size and dimensionality, and as computing

power and graphical methods have tried to

keep up, there has been a rapprochement

between the strictly exploratory and strictly

conﬁrmatory approaches. Working social

scientists routinely explore their data as part

of the process of cleaning and checking it. It

would be naive to think researchers were not on

the lookout—literally—for interesting patterns

in complex data sets. Recent developments in

EDA have focused on extending established

methods of easily looking at a lot of data at

once, and on developing new ways for visually

checking the validity of apparent relationships.

The idea is to make the exploratory a little

more conﬁrmatory.

A ﬁrst useful tool for this sort of exploration

is a generalized scatterplot matrix. I n a standard

pairs plot, the goal is to see all the bivariate rela-

tionships in the data at once, presented in a grid

so that quick comparisons can easily be made.

An unfortunate limitation, particularly for the

social sciences, is that these plots do a poor job

with categorical variables. Ideally we would like

to see the panels of the matrix display the data

in a form appropriate to the underlying vari-

able. A generalized pairs plot (Emerson et al.

2013) accomplishes this, using barcode plots,

boxplots, mosaic plots, and other methods.

Figure 4 shows an example. The speciﬁc soft-

ware implementation adds additional function-

ality, including the ability to display different

plots—such as barcode and mosaic plots—in

the upper and lower triangles of the plot ma-

trix, histograms along the main diagonal, and

the option of adding smoothed or linear regres-

sion lines to panels.

Generalized pairs plots can be extended even

further, depending on the software, by allow-

ing further partitioning within panels. For in-

stance, we can show separate histograms of a

continuous variable broken out by the values

of a categorical variable. Multipanel plots are

intrinsically rich in information. When com-

bined with several within-panel types of repre-

sentation and a large number of variables, they

can become quite complex. But, again, the main

utility of this approach is less in the presenta-

tion of ﬁnished work—although it can certainly

be useful for that—and more in the way it en-

ables the working researcher to quickly inves-

tigate aspects of her own data. The goal is not

to pithily summarize a single point one already

knows, but to open things up for further ex-

ploration. Harrell (2001) remains an exemplary

book-length demonstration of the virtues of in-

tegrating graphical methods with the process

of data exploration (including exploring pat-

terns of missingness in the data) right across

the process of model building, diagnostics, and

presentation.

With many variables and large amounts

of data, a square matrix of plots can become

unwieldy even to the trained eye. Seeing more

www.annualreviews.org

•

Data Visualization in Sociology 113

Annu. Rev. Sociol. 2014.40:105-128. Downloaded from www.annualreviews.org

Access provided by Duke University on 08/09/17. For personal use only.

SO40CH05-MoodyHealy ARI 4 July 2014 13:29

Figure 4

Ageneralizedpairsplothandlescategoricaldataeasily,andindifferentways.

data more quickly, and in particular exploring

high-dimensional data in a controlled way,

has been a focus of recent visualization re-

search. Early work—going back to Tukey, and

others—allowed for the exploration of data

in three dimensions, for instance by way of

rotating a cloud of points on a screen. This

sort of approach “demoed well,” as spinning

around a cloud of colored points looks quite

impressive to the casual observer. But in-

terpreting these displays is another matter.

Thus, methods for interactively exploring

114 Healy

Moody

Annu. Rev. Sociol. 2014.40:105-128. Downloaded from www.annualreviews.org

Access provided by Duke University on 08/09/17. For personal use only.

SO40CH05-MoodyHealy ARI 4 July 2014 13:29

−1 −0.8 −0.6 −0.4 −0.2 0

Correlation

0.2 0.4 0.6 0.8 1

cerebvas

pubhealth

pubtopriv

proglib.trim

assault

donors

roads

external

gdp

health

pop

pop.dens

tradcon.trim

cerebvas

pubhealth

pubtopriv

proglib.trim

assault

donors

roads

external

gdp

health

pop

pop.dens

tradcon.trim

−0.04

0.59

0.45

−0.13

0.12

−0.27

0.07

−0.08

−0.12

0.25

0.28

−0.01

0.01

0.11

0.48

−0.04

−0.27

−0.01

−0.37

0.27

0.02

0.83

0.44

0.15

−0.2

−0.09

−0.41

−0.35

−0.19

−0.33

−0.28

0.39

0.14

−0.01

0.07

−0.44

−0.26

−0.42

−0.3

−0.19

0.27

0.06

−0.03

−0.06

−0.04

−0.33

−0.76

−0.64

0.35

0.14

0.33

−0.34

0.14

−0.11

−0.16

0.51

0.47

0.03

−0.23

−0.02

0.06

0.58

0.38

−0.21

0.04

−0.07

0.26

0.04

−0.16

0.11

−0.03

−0.12

0.33

0.07 0.85

Figure 5

Acorrelationmatrixrepresentedasatiledheatmap(upper triangle) with color-keyed correlation coefﬁcients

(lower triangle).

data sets advanced on two fronts. The ﬁrst

moved toward further development of mul-

tiple panels, notably with innovative ways of

visually conditioning on additional variables

or highlighting interactively selected cases

across panels. Co-plots, shingles, and contour

or surface plots are all examples of this kind

of development (Cleveland 1993, pp. 186–271;

Sarkar 2008, pp. 67–115). Increasingly, these

methods take advantage of color for presenting

data, as with heatmaps or tiled representa-

tions of a correlation matrix (see Figure 5).

Tools for permuting correlation matrices,

either in the order produced by factor-analytic

techniques or other direct optimization, allow

one to identify higher-order patterns in such

ﬁgures (Breiger & Melamed 2014).

A second direction has been the develop-

ment of parallel coordinate plots, which show

multiple variables side by side in a way that

allows for the visualization of both speciﬁc

outliers and clusters of association across

many variables at once (Moustafa & Wegman

2006, Inselberg 2009). Figure 6 gives a simple

www.annualreviews.org

•

Data Visualization in Sociology 115

Annu. Rev. Sociol. 2014.40:105-128. Downloaded from www.annualreviews.org

Access provided by Duke University on 08/09/17. For personal use only.

SO40CH05-MoodyHealy ARI 4 July 2014 13:29

−2.5

0.0

2.5

roads

consent.law

txp.pop

opt

pubtopiv

external

pubhealth variable

pop.dens

donors

health

assault

gdp

cerebvas

Value

Corporatist

Liberal

SocDem

World

Figure 6

A parallel coordinates plot highlighting a possibly relevant grouping variable.

example, although the approach is best suited

to much larger numbers of variables and obser-

vations than shown here. This sort of plot also

beneﬁts from being used interactively, as the

ordering of the variables (and the highlighting

of possible grouping variables) can change

the interpretability of the graph quickly. The

GGobi system, for example, is designed to

provide interactive, semiautomated facilities

for “touring” large, high-dimensional data in

real time using parallel plots and a variety of

other methods (Cook & Swaine 2007).

This broad EDA tradition has recently be-

gun to reconnect with the model-checking

or diagnostic approach, with convergence

happening from both directions. The long-

standing concern here is that a striking visual-

ization might not correspond to any robust un-

derlying phenomenon. Early advocates of data

visualization typically presented a “parade of

horribles” (e.g., Wainer 1984) showing how bad

visual presentation can distort or misrepresent

the data. But even properly presented visual-

izations can be vulnerable to spurious pattern

attribution on the part of researchers and ob-

servers. From the EDA side, Wickham et al.

(2010) and Buja et al. (2009) provide some prin-

cipled ways for assessing, in a broadly graphi-

cal manner, whether or not the patterns one is

seeing are likely to be spurious. For example,

a permutation lineup presents observed data

in a small-multiple context surrounded by null

plots of generated data. “Which plot shows the

real data?” Buja et al. (2009, p. 4372) ask. If

observers cannot reliably pick it out, then we

should doubt both the utility of the plot and

the soundness of any inferences (or arguments)

based on it. From the modeling side, Gelman

(2004, pp. 773–74) argues that a Bayesian ap-

proach provides a principled framework for as-

sessing “the implicit model checking involved

in virtually any data display.”

116 Healy

Moody

Annu. Rev. Sociol. 2014.40:105-128. Downloaded from www.annualreviews.org

Access provided by Duke University on 08/09/17. For personal use only.

SO40CH05-MoodyHealy ARI 4 July 2014 13:29

Although we have argued that sociologists

have been relatively slow to adopt data visual-

ization, several of the issues we have discussed

have independently appeared within the socio-

logical literature. Sociologists routinely deal

with data where almost all the variables of inter-

est are categorical, for example. And, as noted

above, the routine and effective display of cate-

gorical data (especially cross-classiﬁed categor-

ical data) has not been a trivial problem to solve.

Furthermore, sociology has a long tradition of

using methods that reduce high-dimensional

data in some way—especially via factor analysis,

principal components, correspondence analy-

sis, or other related methods. In Distinction, for

example, Bourdieu (1984, pp. 128–29, 262, 266,

343) presents his analysis of the space of French

social class and taste in a way that is both highly

visual but also—for some critics—decidedly dif-

ﬁcult to interpret. This family of methods lends

itself to suggestive visualization in what might

be called a conﬁgurational mode. This is some-

what inimical to the Anglo-American tradition

of seeking causal relations in statistical models.

Breiger (2000) provides a useful discussion of

some of the issues here, emphasizing points of

convergence.

Dimensional reduction of this sort typically

characterizes the problem of interest in terms

of space or distance, which naturally encourages

the mapping of social systems. Sociologists have

been among t he earliest users of these visualiza-

tion tools, particularly with network analysis.

The earliest interactive network tools were lit-

erally peg boards and rubber bands (Freeman

2004) or pins-and-strings.

Interactive explo-

ration of social network data has obviously been

made much easier with the advent of efﬁcient

computer programs. Released in 1996, PAJEK

was one of the earliest completely interactive

visualization tools that was also optimized for

large networks. Earlier software typically sepa-

rated the visualization and analysis steps. There

has since been rapid growth in the development

See http://www.soc.duke.edu/∼jmoody77/VizARS/sna_

peg.jpg.

of interactive network exploration tools, in-

cluding on the web (http://www.theyrule.net,

http://dirtyenergymoney.com). The chal-

lenge for such work is excess reduction in the

inherent complexity of the data, which has led

methodologists to propose ﬁt statistics for net-

work layouts (Moody et al. 2005, Brandes et al.

2012).

The rapid availability of fully dynamic net-

work data has created opportunities and chal-

lenges for visualization. Network movies, for

example, allow one to capture the relational dy-

namics as they unfold in space and time (Moody

et al. 2005, Bender-deMoll et al. 2008, Morris

et al. 2009). The clear advantage of a net-

work movie is that one can reserve the two

dimensions of the visual plane for mapping

the topography of the social system and watch

the shape of the system change as the anima-

tion runs. This is particularly useful for explo-

ration, as it makes visible dynamic features that

are otherwise difﬁcult to capture in summary

statistics. But there are also costs. People tend

to have poor visual memories, so comparing

nonadjacent moments in time is challenging,

and the analyst must make strong assumptions

about how to aggregate the network events

over time. Similar visualization challenges are

becoming common in dynamic statistical dis-

plays, such as the GapMinder data set, which

allows one to explore associations over time

(http://www.gapminder.org).

Presenting the Results

These considerations lead naturally to the ques-

tion of presenting data. Most of t he principles

discussed above regarding the construction of

ﬁgures for exploring data also apply to present-

ing it, if only because the audiences are often

the same—that is, experts in a particular ﬁeld.

But effective statistical graphics have a rhetor-

ical aspect, too (Kostelnick 2008). In general,

the goal is to look for ways of presenting the

data that are both effective with respect to one’s

argument and honest with respect to the data.

Though conceptually simple and among the

earliest examples of statistical visualizations,

www.annualreviews.org

•

Data Visualization in Sociology 117

Annu. Rev. Sociol. 2014.40:105-128. Downloaded from www.annualreviews.org

Access provided by Duke University on 08/09/17. For personal use only.

SO40CH05-MoodyHealy ARI 4 July 2014 13:29

0 20 40 60 80 100

Authors (%)

Total number of lifetime publications Total number of lifetime publications

Authors (%)

10 100

0.001

0.01

0.1

100

Figure 7

The distribution of authors’ lifetime number of publications in three very selective sociology journals is

highly skewed. In comparison to a standard histogram (a), a log-log histogram (b) is much better at revealing

details in the “long tail” of the distribution.

variable distributions remain of keen substan-

tive interest. Many of the distributions typically

studied in sociology are extremely skewed and

difﬁcult to display as simple histograms. Con-

sider, for example, some data on the number of

times authors publish in a select set of journals

(here the American Sociological Review, Ameri-

can Journal of Sociology,andSocial Forces) over

the course of their career. Figure 7a presents

a standard histogram, whereas Figure 7b fol-

lows the convention now common in the phys-

ical sciences of presenting the distribution on a

log-log scale.

When comparing distributions across cat-

egorical variables, comparative boxplots allow

one to examine multiple moments of a distri-

bution across multiple categories or over time

(with some loss of resolution). The presentation

of joint distributions of multiple categorical

variables has similarly been improved with

area-accurate Venn diagrams (see for example,

http://www.eulerdiagrams.org/eulerAPE).

An important contribution to this literature

is the work of Handcock & Morris (1999) on

relative distribution methods. By comparing

the ratio of two distributions at each point

along the x-axis, one is quickly able to identify

differences in both shape and central tendency.

Figure 8 reproduces the relative distribution

in permanent wage growth for two cohorts of

the N ational Longitudinal Survey. If the wage

distributions were identical, the density would

be a simple horizontal line at 1.0; instead we

see much greater inequality (heavier tails at

both ends) in the recent cohort.

A related problem involves effectively

displaying trends over time, particularly when

attempting to demonstrate strong variability

across units. The convention of reserving the

x-axis for time and the y-axis for magnitude

becomes tricky if many series are given equal

weight. An effective solution involves carefully

choosing colors, line weights, and labels to

highlight a particular strand among many (see

Figure 10 below). Moody et al. (2011) are

able to demonstrate the wild variability in

adolescent popularity sequences by generating

a scatterplot of trajectory summaries with

exemplar labels.

Because each position in the

See http://www.soc.duke.edu/∼jmoody77/VizARS/

Figure5.jpg for trendspace; http://www.soc.duke.edu/

∼jmoody77/VizARS/Figure%206.pdf for application of

this space to model prediction outcomes.

118 Healy

Moody

Annu. Rev. Sociol. 2014.40:105-128. Downloaded from www.annualreviews.org

Access provided by Duke University on 08/09/17. For personal use only.

SO40CH05-MoodyHealy ARI 4 July 2014 13:29

Proportion of the original cohort

Permanent dierences in log wages

Relative density

0.0

0.5

1.0

1.5

2.0

2.5

3.0

0.0 0.2 0.4 0.6 0.8 1.0

−1 0.5 1 1.5 2

Figure 8

The relative probability density function distribution of permanent wage growth in the original and recent

National Longitudinal Survey cohorts. A decile bar chart is superimposed on the density estimate. The

upper axis is labeled in permanent differences in log wages (adapted from Handcock & Morris 1999).

ﬁeld captures a unique trend, the distributional

coverage of the space suggests there is no

typical sequence.

Moving beyond simple variable compari-

son displays, the bulk of statistical work in so-

ciology involves complex multivariate models.

Even with good statistical training, tables of

coefﬁcients are hard to decipher quickly and

tend to foreground statistical signiﬁcance over

substantive magnitudes. Straightforwardly in-

terpreting the effects of independent variables is

rarely intuitive, especially for models with com-

plex link functions, categorical components,

or interaction terms. Although odds ratios are

margin free and thus nominally interpretable,

knowing whether an effect is substantively large

is often difﬁcult without comparative context

and may be impossible to discern directly from

the table without intimate knowledge of the un-

derlying distribution of control variables. The

simplest solution to this problem is to use the

model to predict outcome variables at differ-

ent levels or combinations of the independent

variables of interest. Figure 9a shows a pow-

erful example from Mirowsky & Ross (2007).

They use a new style of vector graphs for la-

tent growth models by age (see Mirowsky &

Kim 2007) to display predicted values from in-

teraction terms. This enables them to take re-

sults from a complex structural equation model

of people’s perceived sense of control and si-

multaneously illustrate both within-cohort and

between-cohort changes at varying levels of ed-

ucation in a way that would be otherwise very

difﬁcult to represent.

The ﬁgure allows one to identify changes

within cohorts (change within vector) and over

time (sequence of arrows by group). Here we

see that high s chool dropouts have a lower

sense of control overall but a dramatic drop in

sense of control during youth that levels out

as they age. College-educated respondents, in

contrast, have a generally high sense of control

that is continuously optimistic through adult-

hood, turning negative only after about age 60.

Recent advances in the use of statistical graphics

www.annualreviews.org

•

Data Visualization in Sociology 119

Annu. Rev. Sociol. 2014.40:105-128. Downloaded from www.annualreviews.org

Access provided by Duke University on 08/09/17. For personal use only.

SO40CH05-MoodyHealy ARI 4 July 2014 13:29

1.2

1.0

0.8

0.4

0.2

0.0

0.6

Predicted sense of control

18 24 30 36 42 48 54 60 66 72 78 84 90

Age (years)

College

degree

High school

degree

No high school

degree

Y Percentile

LabourLiberal DemocratConservative

0.0

0.2

0.4

0.6

0.8

369 369 369

Attitude toward Europe

Probability

Knowledge

Figure 9

(a) Vector diagram for latent trajectory model of perceived control by age, cohort, and education (adapted from Mirowsky & Ross

2007, with permission from the University of Chicago Press). (b) Predicted probabilities and standard errors plotted from a multinomial

model (adapted from Fox & Hong 2009).

for model interpretation include estimates of

the uncertainty of the mod el predictions. Most

software now provides easy access to model

predictions from the data, and this allows one

to provide results under varying scenarios (see,

for example, Alkema et al. 2011). In this case,

the hard work is done before the plot is made.

Figure 9b shows a series of predicted proba-

bilities from a multinomial model at different

levels of various predictors and outcomes,

with appropriate standard errors shown. Here

no conceptual advances are needed on the

120 Healy

Moody

Annu. Rev. Sociol. 2014.40:105-128. Downloaded from www.annualreviews.org

Access provided by Duke University on 08/09/17. For personal use only.

SO40CH05-MoodyHealy ARI 4 July 2014 13:29

Colorado Springs

HIV risk network

a Default PAJEK view b Edited for presentation

High

Low

Closeness

centrality

Figure 10

Network exemplar of moving between software default and presentation results. Subtle adjustments to line widths and color palettes

and the addition of a centrality scale greatly aid interpretability in (b).

graphical side, just the ability to get informa-

tion out of the model in a readily interpretable

form (Fox 2003, Fox & Hong 2009).

The distance between exploratory and pre-

sentation graphics is most pronounced as the

density of information necessary to display in-

creases. Network images are particularly inter-

esting in this case. A little effort with layering

and coloring makes a real difference. Consider

also Figure 10, which shows a before and after

of the same data. The basic layout is retained

(with the addition of a little jittering to allevi-

ate algorithmically induced stacking), but the

result is much more interpretable.

Recent work on constructing visually inter-

pretable social networks has focused on care-

ful data reduction, either by suppressing nodes

entirely in favor of contour-style diagrams

(Moody 2004, Moody & Light 2006) or by

deleting or bundling edges to highlight struc-

ture (Crnovrsanin et al. 2014). Other work

has focused explicitly on quantifying the layout

model using stress or multidimensional scaling–

related techniques (Frank & Yasumoto 1998,

Brandes & Pich 2006, Brandes et al. 2012; see

Lima 2011 for exemplars).

Our focus so far has been on presenting re-

sults to professional peers. But in recent years

the clear presentation of data to broader publics

has become increasingly important. It has never

been easier to circulate full-color graphics of

original data analysis to large groups of peo-

ple. Social sharing of data through the Inter-

net generally, but especially through services

such as Facebook and Twitter, has accelerated

the rise of infographics or info-visualization.

To many working statisticians, infographics

are the descendants of Tufte’s Ducks—those

“self-promoting graphics” where “the over-

all design purveys Graphical Style rather than

quantitative information” (Tufte 1983, p. 116).

The contemporary infographic in its pure

form is a supercharged megaduck incorporat-

ing not only the bells and whistles derided by

www.annualreviews.org

•

Data Visualization in Sociology 121

Annu. Rev. Sociol. 2014.40:105-128. Downloaded from www.annualreviews.org

Access provided by Duke University on 08/09/17. For personal use only.

SO40CH05-MoodyHealy ARI 4 July 2014 13:29

Tufte but far more besides, such as a spurious

quasi-narrative structure, pictographic se-

quencing, or excessive dynamic elements.

Gelman & Unwin (2013) discuss Infovis-style

work from a statistical point of view. They argue

that most infographics do not meet the stan-

dards normally demanded of statistical visual-

izations, but they concede that sometimes the

goals of the latter are not those of the former.

It seems clear, though, that information

visualization tools will become ever more

widespread. In keeping with our general argu-

ment that good visualization is a component of

broader good practice around data analysis, a

key issue is the openness of standards and tools

for data analysis on the web. Social scientists

have typically worked within dedicated statisti-

cal applications to produce static graphics in a

format geared primarily for print publication.

But there has been tremendous development

over the past decade, and even just within the

past ﬁve years, in tools designed to present data

interactively on the web. The development of

powerful libraries written in JavaScript has al-

lowed developers to present statistical graphics

in a way that is quite open with respect to both

code and data. Mike Bostock’s D3 library, for

instance, is increasingly used by statisticians and

media analysts alike and provides a powerful set

of dynamic visual methods (Murray 2013). It is

always difﬁcult to know ex ante which particu-

lar software tool kits have staying power in the

long run—functionally similar platforms and

libraries have come and gone before—which

is why static formats such as Postscript and

portable document format, or PDF, are so long-

lived. But even so, the leading edge of develop-

ment in this area seems to be moving to fur-

ther integrate speciﬁc statistical tools such as

R with data formats (notably JavaScript Object

Notation, or JSON) that can be presented effec-

tively and interactively in the browser. For some

kinds of data, notably the generation of dynamic

choropleth maps and cartograms, the standard

of presentation in some media outlets is now

very high. It can be difﬁcult to interpret com-

plex and colorful maps with data chunked into

units that vary radically by size (e.g., US coun-

ties). Nevertheless, a map such as the one shown

in Figure 11,whichappearedintheNew York

Times (Bloch & Gebeloff 2009), makes for a very

engaging way to explore patterns both spatially

and over time. Presenting data of this sort in

an effective, interactive package is difﬁcult for

small teams of researchers to accomplish. But

it is not impossible. Katz’s (2013) dialect survey

maps are a compelling recent example of what is

now within reach. Developers seem interested

in building the production of web-enabled con-

tent into the software sociologists are used to

using, and thus these tools are likely to continue

to become more powerful and easier to use.

For sociologists thinking about the public

impact of their work, it is worth bearing in

mind that, the sins of Infovis notwithstanding,

a well-crafted statistical graphic is the fastest

way to propagate one’s ﬁndings. Moreover, it is

easy to forget how revelatory the general public

can ﬁnd even a relatively ordinary descriptive

image if it is properly constructed. The panels

in Figure 12 show two examples. Figure 12a

shows the rate of deaths due to assault in 24

OECD countries between 1960 and 2011.

The point of the image is to emphasize the

exceptionally high death rate in the United

States compared with other countries (as

well as the large changes in the US number

that are visible over the timeframe), and so

the US series is colored separately from the

rest, with every other country getting their

own smoothed line and data points, but not

individual colors. The unique trajectory of the

United States is immediately apparent. The use

of color probably helped the image circulate

more widely in social media and traditional

outlets than it otherwise might have. Color is

not strictly necessary, however, as the superb

image in Figure 12b makes clear. Taken from

Kenworthy (2014), Figure 12b shows trends

in life expectancy plotted against a measure

of health expenditures for 20 countries. The

United States is singled out with a bolder line

than the others. Individu al data points are not

plotted. There are only seven numbers labeled

on the graph (including the one in “19 other

rich countries”), yet a strong argument based

122 Healy

Moody

Annu. Rev. Sociol. 2014.40:105-128. Downloaded from www.annualreviews.org

Access provided by Duke University on 08/09/17. For personal use only.

SO40CH05-MoodyHealy ARI 4 July 2014 13:29

Figure 11

A New York Times interactive choropleth map allows users to explore historical and geographical patterns of migration to the United

States (Bloch & Gebeloff 2009, adapted with permission from the New York Times;theinteractivemapisavailableathttp://www.

nytimes.com/interactive/2009/03/10/us/20090310-immigration-explorer.html).

on rich data is beautifully made about what has

happened to the returns to health spending in

the OECD generally, and in the United States

in particular. In the original presentation,

Kenworthy characterizes the data and mea-

sures with a compact note in the caption,

specifying the methods and measures. There is

nothing about this ﬁgure that is conceptually or

technically new. And yet a clearly conceived and

cleanly executed image like this is still relatively

uncommon in the sociological literature.

Visualizations of categorical data remain

more difﬁcult to convey effectively, partly be-

cause the general public is not always familiar

with conventional ways to present it. Mosaic

plots, for instance, can be effective representa-

tions of contingency tables, but people are not

taught to read them in the same way they can

read bar charts or scatterplots. The effective

visualization of network data presents similar

issues. The dual problems of dimen sionality

and scale require creative ways to layer and

aggregate information in a manner that high-

lights the key features of interest. In an attempt

to characterize trends in political polarization

in the US Senate, Moody & Mucha (2013)

relied on a combination of multiple aggrega-

tion strategies and visual “identity arcs” linking

individuals over time that effectively pushed

“party loyalists” to the background while

www.annualreviews.org

•

Data Visualization in Sociology 123

Annu. Rev. Sociol. 2014.40:105-128. Downloaded from www.annualreviews.org

Access provided by Duke University on 08/09/17. For personal use only.

SO40CH05-MoodyHealy ARI 4 July 2014 13:29

1960 1970 1980 1990 2000 2010

Year

Assault Deaths per 100,000 population

United States 23 other OECD Countries

19 other rich

countries

Life expectancy

5 12 18%

Health expenditures

a Assault deaths by country b Life expectancy by country

Figure 12

(a) Assault deaths in the United States and 23 other OECD countries (Healy 2012). (b) Health expenditure (as a percentage of GDP)

and life expectancy in the United States and 19 other rich countries (see Kenworthy 2014; image courtesy of L. Kenworthy).

highlighting those (increasingly rare) senators

who reach across the aisle (Figure 13).

CONCLUSION

We have argued that quantitative visualization

is a core feature of social-scientiﬁc practice from

start to ﬁnish. All aspects of th e research pro-

cess from the initial exploration of data to the

effective presentation of a polished argument

can beneﬁt from good graphical habits. Good

graphics are not, of course, the only thing—see

Godfrey (2013) for a discussion of the situation

of blind and visually impaired users of current

statistical software. But the dominant trend is

toward a world where the visualization of data

and results is a routine part of what it means to

do social science.

Getting general audiences comfortable with

different kinds of data visualization is a long-

term project, and not one that any particular

researcher or journal editor has any meaningful

control over. But given that the interpretability

of statistical graphics rests on both their inter-

nal coherence as objects and the shared rep-

resentational conventions they embody, a ﬁrst

step is to insist on good standards in the peer

review process. A glance at recent issues of,

say, the American Sociological Review shows that

the standards for publishable graphical material

vary wildly between and even within articles—

far more than the standards for data analysis,

prose, and argument. Variation is to be ex-

pected, but the absence of consistency in ele-

ments as simple as axis labeling, gridlines, or

legends is striking. Just as training in elemen-

tary visualization methods should be a standard

component of graduate education, our ﬂag-

ship journals should encourage their authors to

think about the most effective ways to encour-

age visual clarity. This should not take the form

of overly strict style guides but instead aim for

an ideal of consistent, considered good judg-

ment in the presentation of data and results in

the service of sociological argument.

Effective data visualization is part of a

broader shift in the social sciences where data

are more easily available, code and coding tools

are more widely accessible, and high-quality

graphical work is easy to produce and share.

We hope for professional audiences who ex-

pect to see effective graphics as a routine as-

pect of presented work, and we look forward

to wider publics who are able to comfortably

read and interpret good graphical work. Sociol-

ogists should take advantage of the remarkable

124 Healy

Moody

Annu. Rev. Sociol. 2014.40:105-128. Downloaded from www.annualreviews.org

Access provided by Duke University on 08/09/17. For personal use only.

SO40CH05-MoodyHealy ARI 4 July 2014 13:29

Polarization modularity

0.27

0.13

‘07–’08‘03–’04‘99–’00‘95–’96‘91–’92‘87–’88‘83–’84‘79–’80‘75–76

Democrats

US Senate voting similarity networks, 1975–2012

Timeline: president, Senate party balance, and date (through June 7, 2012)

Republicans

CarterFord Reagan G.H.W. Bush Clinton G.W. Bush

DRRDRRRDDDDRRRDDD

0.1

0.2

0.3

1910 1930 1950 1970

Year

1990 2010

Detail

Modularity

Obama

‘11–’12

Group size Senators

crossing time

5062 6256 56

Senate balance

Within-group vote similarity

0.72

0.89

0.78

0.83

Vote similarity (≥0.6)

Figure 13

Aggregation and a known dimension (a polarization scale) simplify a complex network layout. (Adapted from Moody & Mucha 2013 with permission from Cambridge

University Press.)

www.annualreviews.org

•

Data Visualization in Sociology 125

Annu. Rev. Sociol. 2014.40:105-128. Downloaded from www.annualreviews.org

Access provided by Duke University on 08/09/17. For personal use only.

SO40CH05-MoodyHealy ARI 4 July 2014 13:29

progress in methods, tools, and means to

share—from statistics to computational social

science to web development—the better to see

the social world, and help others see it, too.

DISCLOSURE STATEMENT

The authors are not aware of any afﬁliations, memberships, funding, or ﬁnancial holdings that

might be perceived as affecting the objectivity of this review.

ACKNOWLEDGMENTS

We thank Jaemin Lee, Achim Edelmann, and Richard Benton for comments on earlier drafts.

Permission to use copyrighted material was granted by the American Sociological Association

(Figure 1b), the University of Chicago Press (Figure 9a), the New York Times (Figure 11), and

Cambridge University Press (Figure 13). All other ﬁgures are taken from the public domain and/or

signiﬁcantly redrawn and adapted by the authors. Partial support for this work was provided by

NIH grants 1R21HD068317-01 and 1 R01 HD075712-01.

LITERATURE CITED

Alkema L, Raftery AE, Gerland P, Clark SJ, Pelletier F, et al. 2011. Probabilistic projections of the total

fertility rate for all countries. Demography 48:815–39

Anscombe FJ. 1973. Graphs in statistical analysis. Am. Stat. 27:17–21

Bender-deMoll S, Morris M, Moody J. 2008. Prototype packages for managing and animating longitudinal

network data: dynamicnetwork and rSoNIA. J. Stat. Softw. 24(7). http://www.jstatsoft.org/v24/i07

Bertin J. 1967 (2010). Semiology of Graphics: Diagrams, Networks, Maps. Redlands, CA: ESRI Press

Bloch M, Gebeloff R. 2009. Immigration explorer. New York Times,March10.http://www.nytimes.com/

interactive/2009/03/10/us/20090310-immigration-explorer.html.

Bourdieu P. 1984. Distinction: A Social Critique of the Judgment of Taste. Cambridge, MA: Harvard Univ. Press

Brandes U, Indlekofer N, Mader M. 2012. Visualization methods for longitudinal social networks and stochas-

tic actor-oriented modeling. Soc. Netw. 43:291–308

Brandes U, Pich C. 2006. Eigensolver methods for progressive multidimensional scaling of large data. Int.

Symp. Graph Drawing (GD), Lect. Notes Comput. Sci. (LNCS)4372:42–53

Breiger RL. 2000. A toolkit for practice theory. Poetics 27:91–115

Breiger RL, Melamed D. 2014. The duality of organizations and their attributes: turning regression modeling

‘inside out.’ Res. Sociol. Organ. 40:261–74

Buja A, Cook D, Hofmann H, Lawrence M, Lee EK, et al. 2009. Statistical inference for exploratory data

analysis and model diagnostics. Phil. Trans. R. Soc. A 367:4361–83

Chang W. 2013. The R Graphics Cookbook.Sebastopol,CA:O’Reilly

Chapin FS. 1924. The statistical deﬁnition of a societal variable. Am. J. Sociol. 30:154–71

Chatterjee S, Firat A. 2007. Generating data with identical statistics but dissimilar graphics: a follow up to the

Anscombe Dataset. Am. Stat. 61:248–54

Cleveland WS. 1993. Visualizing Data.Summit,NJ:Hobart

Cleveland WS. 1994. The Elements of Graphing Data.Summit,NJ:Hobart

Cook D, Swaine DF. 2007. Interactive and Dynamic Graphics for Data Analysis. New York: Springer

Crnovrsanin T, Muelder CW, Faris R, Felmlee D, Ma K-L. 2014. Visualization techniques for categorical

analysis of social networks with multiple edge sets. Soc. Netw. 37:56–64

Du Bois WEB. 1898 (1967). The Philadelphia Negro.NewYork:ShockenBooks

Emerson JW, Green W, Schloerke B, Crowley B, Cook D, et al. 2013. The generalized pairs plot. J. Comp.

Graph. Stat. 22:79–91

126 Healy

Moody

Annu. Rev. Sociol. 2014.40:105-128. Downloaded from www.annualreviews.org

Access provided by Duke University on 08/09/17. For personal use only.

SO40CH05-MoodyHealy ARI 4 July 2014 13:29

Few S. 2009. Now You See It: Simple Visualization Techniques for Quantitative Analysis. Oakland, CA: Analytics

Few S. 2012. Show Me the Numbers: Designing Tables and Graphs to Enlighten. Burlingame, CA: Analytics.

2nd ed.

Fox J. 2003. Effect displays in R for generalised linear models. J. Stat. Softw. 8(15). http://www.

jstatsoft.org/v08/i15/paper

Fox J, Hong J. 2009. Effect displays in R for multinomial and proportional-odds logit models: extensions to

the effects package. J. Stat. Softw. 32(1). http://www.jstatsoft.org/v32/i01/paper

Frank KA, Yasumoto J. 1998. Linking action to social structure within a system: social capital within and

between subgroups. Am. J. Sociol. 104:642–86

Freeman LC. 2004. The Development of Social Network Analysis: A Study in the Sociology of Science.Vancouver,

Can.: Empirical

Freese J. 2007. Reproducibility standards in quantitative social science: why not sociology? Soc. Methods Res.

36:153–72

Friendly M. 2000. Visualizing Categorical Data. Cary, NC: SAS Inst.

Gelman A. 2004. Exploratory data analysis for complex models. J. Comput. Graph. Stat. 13:755–79

Gelman A, Unwin A. 2013. Infovis and statistical graphics: different goals, different looks. J. Comp. Graph.

Stat. 22:2–28

Godfrey AJR. 2013. Statistical software from a blind person’s perspective. RJ.5:73–79

Handcock MS, Morris M. 1999. Relative Distribution Methods in the Social Sciences.NewYork:Springer-Verlag

Harrell F. 2001. Regression Modeling Strategies.NewYork:Springer

Hart HH. 1896. Immigration and crime. Am. J. Sociol. 2:369–77

Healy K. 2012. America is a violent country. Kieran Healy Blog,July20.http://kieranhealy.org/blog/

archives/2012/07/20/america-is-a-violent-country

Hewitt C. 1977. The effect of political democracy and social democracy on equality in industrial societies: a

cross-national comparison. Am. Sociol. Rev. 42:450–64

Inselberg A. 2009. Parallel Coordinates: Visual Multidimensional Geometry and its Applications.NewYork:

Springer

Jackman RM. 1980. The impact of outliers on income inequality. Am. Sociol. Rev. 45:344–47

Katz J. 2013. Regional dialect variation in the continental US.Work.Pap.,Proj.Beyond“Soda,Pop,orCoke,”

Dep. Stat., N.C. State Univ., Raleigh. http://www4.ncsu.edu/∼jakatz2/project-dialect.html

Kenworthy L. 2014. Social Democratic America. New York: Oxford Univ. Press

Keynes JM. 1938. Review of HG Funkhouser, Historical Development of the Graphical Representation of Statistical

Data. Econ. J. 48:281–82

Kleimean K, Horton NJ. 2013. SAS and R: Data Management, Statistical Analysis, and Graphics.BocaRaton,

FL: Chapman & Hall/CRC. 2nd ed.

Kostelnick C. 2008. The visual rhetoric of data displays: the conundrum of clarity. IEEE Trans. Prof. Commun.

51:116–29

Lenski G. 1966. Power and Privilege.NewYork:McGraw-Hill

Lima M. 2011. Visual Complexity: Mapping Patterns of Information. New York: Princeton Archit. Press

Lundberg GA, Steele M. 1938. Social attraction-patterns in a village. Sociometry 1:375–419

Mann ME, Bradley RS, Hughes MK. 1999. Northern hemisphere temperatures during the past millennium:

inferences, uncertainties, and limitations. Geophys. Res. Lett.

26:759–62

Marro A. 1899. Inﬂuence of the puberal development upon the moral character of children of both sexes. Am.

J. Sociol. 5:193–219

Mirowsky J, Kim J. 2007. Graphing age trajectories: vector graphs, synthetic and virtual cohort projections,

and virtual cohort projections, and cross-sectional proﬁles of depression. Sociol. Methods Res. 35:497–541

Mirowsky J, Ross C. 2007. Life course trajectories of perceived control and their relationship to education.

Am. J. Sociol. 112:1339–82

Mitchell M. 2012. A Visual Guide to Stata Graphics. College Station, TX: Stata. 3rd ed.

Moody J. 2004. The structure of a social science collaboration network: disciplinary cohesion from 1963 to

1999. Am. Sociol. Rev. 69:213–38

Moody J, Brynildsen WD, Osgood DW, Feinberg ME, Gest S. 2011. Popularity trajectories and substance

use in early adolescence. Soc. Netw. 33:101–12

www.annualreviews.org

•

Data Visualization in Sociology 127

Annu. Rev. Sociol. 2014.40:105-128. Downloaded from www.annualreviews.org

Access provided by Duke University on 08/09/17. For personal use only.

SO40CH05-MoodyHealy ARI 4 July 2014 13:29

Moody J, Light R. 2006. A view from above: the evolving sociological landscape. Am. Sociol. 38:67–86

Moody J, McFarland DA, Bender-deMoll S. 2005. Dynamic network visualization: methods for meaning with

longitudinal network movies. Am. J. Sociol. 110:1206–41

Moody J, Mucha PJ. 2013. Portrait of political party polarization. Netw. Sci. 1:119–21

Morris M, Kurth AE, Hamilton DT, Moody J, Wakeﬁeld S. 2009. Concurrent partnerships and HIV preva-

lence disparities by race: linking science and public health. Am. J. Public Health 99:1023–31

Moustafa R, Wegman E. 2006. Multivariate continuous data—parallel coordinates. In Graphics of Large

Datasets, ed. A Unwin, C Theus, H Hofmann, pp. 143–56. New York: Springer

Murray S. 2013. Interactive Data Visualization for the Web. Sebastopol: O’Reilly

Murrell P. 2011. RGraphics.Boca Raton, FL: Chapman & Hall. 2nd ed.

Sarkar D. 2008. Lattice: Multivariate Data Visualization with R.NewYork:Springer

Sletto RF. 1936. A critical study of the criterion of internal consistency in personality scale construction. Am.

Sociol. Rev. 1:61–68

Stack S. 1979. The effects of political participation and social party strength on the degree of income inequality.

Am. Sociol. Rev. 44:168–71

Tufte ER. 1978. Political Control of the Economy.Princeton,NJ:PrincetonUniv.Press

Tufte ER. 1983. The Visual Display of Quantitative Information.Cheshire,CT:Graphics

Tufte ER. 1990. Envisioning I nformation.Cheshire,CT:Graphics

Tufte ER. 1997. Visual Explanations: Images and Quantities, Evidence and Narrative.Cheshire,CT:Graphics

Tufte ER. 2006. Beautiful Evidence.Cheshire,CT:Graphics

Tukey JW. 1972. Some graphic and semigraphic displays. In Statistical Papers in Honor of George W. Snedecor,

ed. TA Bancroft, pp. 293–316. Ames: Iowa State Univ. Press

Tukey JW. 1977. Exploratory Data Analysis.NewYork:AddisonWesley

Wainer H. 1984. How to display data badly. Am. Stat. 38:137–47

Wainer H. 2010. Foreword. See Bertin 1967 (2010), pp. ix–x

Wickham H. 2009. ggplot2: Elegant Graphics for Data Analysis.NewYork:Springer

Wickham H. 2010. A layered grammar of graphics. J. Comput. Graph. Stat. 19:3–28

Wickham H, Cook D, Hofmann H, Buja A. 2010. Graphical inference for Infovis. IEEE Trans. Vis. Comput.

Graph. 6:973–79

Wilkinson L. 1995 (2005). The Grammar of Graphics. New York: Springer. 2nd ed.

Yau N. 2012. Visualize This: The FlowingData Guide to Design, Visualization, and Statistics.Indianapolis,IN:

Wiley

128 Healy

Moody

Annu. Rev. Sociol. 2014.40:105-128. Downloaded from www.annualreviews.org

Access provided by Duke University on 08/09/17. For personal use only.

SO40-FrontMatter ARI 8 July 2014 6:42

Annual Review

of Sociology

Volume 40, 2014

Contents

Prefatory Chapter

Making Sense of Culture

Orlando Patterson ♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣1

Theory and Methods

Endogenous Selection Bias: The Problem of Conditioning on a

Collider Variable

Felix Elwert and Christopher Winship ♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣31

Measurement Equivalence in Cross-National Research

Eldad Davidov, Bart Meuleman, Jan Cieciuch, Peter Schmidt, and Jaak Billiet ♣♣♣♣♣♣♣♣♣55

The Sociology of Empires, Colonies, and Postcolonialism

George Steinmetz ♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣77

Data Visualization in Sociology

Kieran Healy and James Moody ♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣105

Digital Footprints: Opportunities and Challenges for Online Social

Research

Scott A. Golder and Michael W. Macy ♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣129

Social Processes

Social Isolation in America

Paolo Parigi and Warner Henson II ♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣153

War

Andreas Wimmer ♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣173

60 Years After Brown:TrendsandConsequencesofSchoolSegregation

Sean F. Reardon and Ann Owens ♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣199

Panethnicity

Dina Okamoto and G. Cristina Mora ♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣219

Institutions and Culture

A Comparative View of Ethnicity and Political Engagement

Riva Kastoryano and Miriam Schader ♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣241

Annu. Rev. Sociol. 2014.40:105-128. Downloaded from www.annualreviews.org

Access provided by Duke University on 08/09/17. For personal use only.

SO40-FrontMatter ARI 8 July 2014 6:42

Formal Organizations

(When) Do Organizations Have Social Capital?

Olav Sorenson and Michelle Rogan ♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣261

The Political Mobilization of Firms and Industries

Edward T. Walker and Christopher M. Rea ♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣281

Political and Economic Sociology

Political Parties and the Sociological Imagination:

Past, Present, and Future Directions

Stephanie L. Mudge and Anthony S. Chen ♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣305

Taxes and Fiscal Sociology

Isaac William Martin and Monica Prasad ♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣331

Differentiation and Stratiﬁcation

The One Percent

Lisa A. Keister ♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣347

Immigrants and African Americans

Mary C. Waters, Philip Kasinitz, and Asad L. Asad ♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣369

Caste in Contemporary India: Flexibility and Persistence

Divya Vaid ♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣391

Incarceration, Prisoner Reentry, and Communities

Jeffrey D. Morenoff and David J. Harding ♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣411

Intersectionality and the Sociology of HIV/AIDS: Past, Present,

and Future Research Directions

Celeste Watkins-Hayes ♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣431

Individual and Society

Ethnic Diversity and Its Effects on Social Cohesion

Tom van der Meer and Jochem Tolsma ♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣459

Demography

Warmth of the Welcome: Attitudes Toward Immigrants

and Immigration Policy in the United States

Elizabeth Fussell ♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣479

Hispanics in Metropolitan America: New Realities and Old Debates

Marta Tienda and Norma Fuentes ♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣499

Transitions to Adulthood in Developing Countries

Fatima Ju´arez and Cecilia Gayet ♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣521

vi Contents

Annu. Rev. Sociol. 2014.40:105-128. Downloaded from www.annualreviews.org

Access provided by Duke University on 08/09/17. For personal use only.

SO40-FrontMatter ARI 8 July 2014 6:42

Race, Ethnicity, and the Changing Context of Childbearing

in the United States

Megan M. Sweeney and R. Kelly Raley ♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣539

Urban and Rural Community Sociology

Where, When, Why, and For Whom Do Residential Contexts

Matter? Moving Away from the Dichotomous Understanding of

Neighborhood Effects

Patrick Sharkey an d Jacob W. Faber ♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣559

Gender and Urban Space

Daphne Spain ♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣581

Policy

Somebody’s Children or Nobody’s Children? How the Sociological

Perspective Could Enliven Research on Foster Care

Christopher Wildeman and Jane Waldfogel ♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣599

Sociology and World Regions

Intergenerational Mobility and Inequality: The Latin American Case

Florencia Torche ♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣619

A Critical Overview of Migration and Development:

The Latin American Challenge

Ra´ul Delgado-Wise ♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣643

Indexes

Cumulative Index of Contributing Authors, Volumes 31–40 ♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣665

Cumulative Index of Article Titles, Volumes 31–40 ♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣♣669

Errata

An online log of corrections to Annual Review of Sociology articles may be found at

http://www.annualreviews.org/errata/soc

Contents vii

Annu. Rev. Sociol. 2014.40:105-128. Downloaded from www.annualreviews.org

Access provided by Duke University on 08/09/17. For personal use only.