Sie sind auf Seite 1von 16

Information Introduction Vivien Beattie is a Senior Lecturer

in Accounting at the University of


design and Stirling, and Michael John Jones is
a Lecturer in Accounting at Cardiff
Corporate information design is big business.
manipulation: One particularly important corporate commu-
Business School. Both authors are
pleased to receive correspondence.
financial graphs nications medium is the corporate annual
in corporate report. Recognition of the need for professional Authors' addresses
Vivien Beattie
annual reports and effective corporate design reflects a funda-
Department of Accountancy
mental sea-change in management's attitude to, and Finance
and appreciation of, the annual report's func- University of Stirling
Vivien Beattie and Stirling FK9 4LA
Michael John Jones tion. Until relatively recently the corporate Scotland
annual report was perceived as an annual regu- Michael John Jones
Cardiff Business School
We discuss the results of an latory chore, a primarily statutory document University of Wales, College of
investigation into the graphic whose sole purpose was to provide shareholders Cardiff
reporting practices used by Aberconway Building
with stewardship information so that they could Colum Drive
240 leading UK companies in
assess managerial performance. T h e increasing Cardiff CF1 3EU
their 1989 corporate annual Wales
reports. Our main findings are
sophistication of the corporate communication
that 79 per cent of companies process has, however, encouraged management
used graphs and that 64 per to exploit the annual report's potential as a
cent of all graphs were bar/ major public relations and promotional oppor-
column graphs. Many of these tunity (Hanson, 1989). Over the last 20 years
were poorly designed and con­
structed. There was evidence
companies have, therefore, increasingly provid-
of biasing in graphic choices, ed voluntary information and adopted more
with the use of graphic presen­ imaginative and creative presentational tech-
tation being contingent upon niques. A particularly important and interesting
'good' rather than 'bad' finan­ trend has been the growing use of graphic forms
cial performance. Companies
of presentation. The majority of UK and USA
were three times more likely to
include graphs in their annual companies now include graphs of selected
report which exaggerated, financial variables in their annual reports. This
rather than understated, fav­ article focuses upon this particular practice.
ourable time series trends in
Financial graphs offer companies an alterna-
key performance variables.
There was also evidence of the tive presentational format to the traditional
use of certain design and con­ alphanumeric table and continuous narrative
struction techniques intended text formats. They have several key advantages.
to create a favourable visual Graphs, particularly those in colour, draw the
impression. There is a need
eye. This attention-attracting quality is re-
for more studies of graphic
practices in other domains,
inforced by memorial potency. The brain's
and for guidelines to raise the memory for pictorial and graphical representa-
standards and fidelity of tion is, in general, superior to that for numbers
financial graphs. (Leivian, 1980). Graphs can therefore be
impressive communicators of financial, numeri- 1. Macdonald-Ross
(1977) provides an authorita-
cal information. 1 As Korol (1986) notes, graphs tive review of the use of graphs
possess 'the ability to better make comparisons to convey quantitative data in a
and identify trends, the ability to uncover previ- wide variety of contexts.

211

Information Design Journal 7:3 (1994), 211-226. DOI 10.1075/idj.7.3.03bea


ISSN 0142-5471 / E-ISSN 1569-979X © John Benjamins Publishing Company
Vivien Beattie & Michael John Jones ■ Information design and manipulation: financial graphs in corporate annual reports

ously hidden facts; and, the avoidance of variables were excluded solely on the grounds
lengthy analyses to determine the message' (pp. that they would present an unflattering picture
148-9). A company's results are often sum- of the company's financial performance.
marised using a few key financial indicators, Second, certain graphical forms are particu-
such as sales, profit, earnings and dividends. larly suited to specific tasks. T h e information
Graphs, therefore, enable a concise financial processing of graphs initially involves perceptu-
picture to be presented which can be readily al, rather than highly cognitive, tasks. These
assimilated. In a survey, more than 75 per cent perceptual tasks include initial comparative
of shareholders in the UK expressed a desire for judgements of visual dimensions and the per-
more graphs and charts in annual reports to ceptual accuracy of such judgements has been
help explain financial performance {The shown to vary. For example, position along a
Accountant's Magazine, 1992). This was sup- common axis (the visual dimension used in
ported by a further recent survey of the annual bar/column graphs) is a more accurately per-
reports of UK and American companies which ceived dimension than angle/slope (the visual
concluded that graphs were valuable for the dis- dimension used in pie and line graphs)
play of performance data and ratios (Peter (Cleveland and McGill, 1987). Most financial
Prowse Associates 1993). graphs in annual reports portray time series
The increasing incidence of corporate finan- data which can be presented using column
cial graphics has been paralleled by largely inde- graphs or line graphs. Column graphs are there-
pendent developments in the fields of statistical fore the preferred presentational format for
graphics and computer graphics. Statistical such trend data, since this graph type utilises
graphics researchers (such as Schmid and position along a common axis. Furthermore,
Schmid, 1979; Schmid, 1983) have been partic- financial graphs in annual reports are common-
ularly interested in further distilling, from exist- ly used to present a quick overview of corporate
ing practice, a set of design and construction performance, rather than forming the basis of
principles. Technological developments in an in-depth analysis. Simple information acqui-
computer graphics have recently provided a cat- sition tasks of this type rely on perceptual
alyst for experimental studies designed to test impression rather than cognitive evaluation.
psychophysical principles of graphical percep- Consequently, the suitability of the graphical
tion. format, which directly affects the impression
Five important premises underpin current conveyed, is of critical importance.
thinking about the graphical presentation of Third, there is a consensus that the represen-
quantitative material. These relate to: selectivi- tation of numbers, as physically measured on
ty/biasing in the use of graphs, suitability of the graph's surface, should be directly propor-
graph type, measurement distortion, presenta- tionate to the numerical values being represent-
tional enhancement, and context-dependency. ed (Tufte, 1983). In column graphs, therefore,
We now briefly describe each of these premises. a doubling in the financial variable portrayed
First, the initial decision about the use of should be represented by a doubling of the col-
graphs must avoid biasing. It is a fundamental umn's height. Violations of this fundamental
principle in the communication of financial principle can be measured using Tufte's 'lie fac-
information that information should be present- tor' (Tufte, 1983). Distortions can arise simply
ed in a neutral and unbiased manner. Biasing because of inaccurate draughtsmanship or
would, for example, occur if commonly graphed because basic graph design principles have been

Information design journal 7/3 (1994) 211-226

212
Vivien Beattie & Michael John Jones ■ Information design and manipulation: financial graphs in corporate annual reports

transgressed. Possible transgressions include values. For column graphs the specifiers
the use of a non-zero axis, a broken axis, or a (columns) should be of uniform width, uni-
non-arithmetic scale. formly spaced, and not cross-hatched since this
T h e fourth underlying premise is that a well- can cause a distracting vibratory visual impres-
designed graph has a number of key graphic sion. Labelling should be clear, precise and
components (illustrated in figure 1) which must meaningful. The interaction of these graphic
be correctly displayed for effective communica- components should result in an impression of
tion to occur. For example, by convention, a balance and harmony. Furthermore, the graph-
column graph has four structural components: ic components should be arranged so that a
background, framework, graphic specifier and neutral and fair representation of the underlying
labels. The background has no essential func- financial information is portrayed. In particular,
tion and must be unobtrusive. The framework presentational enhancement designed to give a
for column graphs normally consists of two more flattering view of the results than is war-
intersecting perpendicular lines, which should ranted by the underlying data should be
each be scaled. The specifier is the symbol avoided.
which represents the underlying numerical

Figure 1
Key graphic Title
components
Width of specifier
of a graph
XYZ plc Group Turnover
Interspace

40...

Specifier/
graphical
Scale division/ 30 element
tick mark

Turnover
(£m) 20

Gridline
Numeric label
10

o
1990 1991 1992 1993 1994
Axis/scale line
Year

Alphabetic label

Information design journal 7/3 (1994) 211-226

213
Vivien Beattie & Michael John Jones ■ Information design and manipulation: financial graphs in corporate annual reports

Finally, the context-dependency of graphs Our main findings were that 79 per cent of
has been emphasised. T h e effectiveness of leading UK companies used graphs and that
graphical presentation is contingent upon many of these were poorly designed and con-
decision task complexity (see., for example, De structed. We found evidence of systematic
Sanctis and Jarvenpaa, 1989) and upon user selectivity, measurement distortion and presen-
characteristics (Garceau et al., 1988). For exam- tational enhancement. These findings are dis-
ple, it has been found that graphs are generally cussed in the next section in more detail and in
superior to tables for simple information acqui- table 1 we provide a summary. 3 We have, where
sition tasks, with the degree of superiority possible, supported our analysis by including
depending on the user's graphical literacy and black-and-white reproductions of real-world
learning style preferences. examples.
Surprisingly, given the ongoing debate con-
cerning the theoretical and practical aspects of
graphical design, we are unaware of any sub- Findings
stantive, systematic investigations of empirical
practice in any discipline. 2 Consequently, there In the first part of this section we summarise
is a lack of knowledge about graphical usage. the frequency of graph usage, categorise the
This article addresses this area of relative types of graphs employed and the topics cov-
neglect by discussing the results of a domain- ered, and comment on both the use of graphic
specific analysis of particular graphical practices designers and the prominence of the graphs.
in financial reporting. T h e main aims of this T h e second part considers the important issue
analysis were: first, to establish the extent and of measurement distortion. Finally, in the third
nature of graph use by UK companies; and sec- part, we discuss our findings with regard to
ond, to document and analyse cases where the aspects of the financial graph's design and con-
principles of representational neutrality had struction, covering the four main graphical
been violated as regards selectivity, measure- components and associated issues, such as
ment distortion, and presentational colour and illusory effects.
enhancement.
We summarise the findings of a systematic Nature and extent of graphical usage
study of 240 annual reports, drawn by random Frequency of graph usage
sampling from the largest (by market capitalisa- T h e corporate annual reports of 189 out of the
tion) 500 listed UK companies. Using a check- 240 companies (79 per cent) contained at least
list, we collected for each company general one graph (of either a financial or non-financial
graphical information (for example, use and
non-use of graphs), and details about those 3. Readers wishing a more in-depth coverage of our
specific aspects of graph design and construc- findings are referred to Beattie and Jones, 1992 a, b. The
first of these articles is a Research Report, published by the
tion which are discussed above. This included Chartered Association of Certified Accountants, which
detailed information on the degree and poten- reviews the accounting graphics literature, gives a basic
tial causes of measurement distortion and on description of graphs and graphical understanding based
presentational enhancement involving the on the statistical graphics literature, and includes 21 colour
graph's four structural components. examples of corporate financial graphs. The second article
presents a detailed statistical analysis of the relationship
2. There are, however, a large number of empirical between the incidence of graph use, measurement distor-
studies about map design and use. tion, and measures of corporate performance.

Information design journal 7/3 (1994) 211-226

214
Vivien Beattie & Michael John Jones ■ Information design and manipulation: financial graphs in corporate annual reports

variable). More particularly, 65 per cent of all


companies graphed one of four financial vari-
ables (turnover, profit before tax, earnings per
share (EPS) and dividends per share ( D P S ) )
which are of key importance in assessing a com-
pany's performance. The average number of
graphs used by graph-using companies was 7.5.
Interestingly, the decision to use graphs of the
four key financial graphs is contingent upon
good financial performance. In other words,
companies with good financial performance, as
measured by an increase in E P S , were more like-
Figure 2
ly to include graphs of at least one key variable
Scottish
than those companies where EPS had declined. Metropolitan
For example, of the 183 companies where EPS Property plc
increased over the current reporting year, 73 per 1989: earn­
cent included graphs while of the remaining ings per share
graph. This
companies, where EPS declined, only 41 per
fan-shaped
cent did so. Using a chi-square test this was sta- pictorial graph
tistically significant at the 1 per cent level. And, has no conven­
except in the case of turnover, significant asso- Earnings Per Share tional axes and
ciations were discovered between an increase in the specifiers
an individual key financial variable and the like- are of inconsis­
tent width. The
lihood of that specific variable being graphed. Graphic designers
remote key
T h e use of professional external designers was impedes fast
Types of graph very much the norm: 69 per cent of the annual information
In all there were 1,426 graphs, of which 31 per reports which contained graphs credited an processing. In
external design company.4 However, there was the colour orig­
cent showed key financial variables. T h e most
no evidence of oligopolistic concentration - 130 inal, the speci­
common graph type overall was the column/bar fiers darkened
graph (64 per cent). T h e column graph is par- companies used 93 different external designers. progressively
ticularly useful for displaying time series data, Addison Design and Michael Peters Literature from light blue
since it portrays time sequentially, from left to were the only two designers to be acknowledged to dark purple.
right. It was used for 84 per cent of the key by more than 5 per cent of the 130 companies.
financial variables, which were typically shown Only five companies acknowledged in-house
as five year time series (72 per cent). A small designs. A chi-square test supported the hypo-
minority of companies used graphical types thesis (at the 5 per cent level) that those compa-
considered to be totally unsuitable for time nies employing graphic designers would be
trends. For example, Scottish Metropolitan more likely to use graphs. In an interesting
Property (see figure 2) has no conventional testament to the standards of British graphic
axes, the columns form a fan-shape, emerge
4. We have no way of telling whether those companies
from an uncertain origin, and also progressively which did not give explicit credits used in-house designers
widen. Consequently, problems are posed when or left the design in the hands of the annual reports'
inter-column comparisons are attempted. printers.

Information design journal 7/3 (1994) 211-226

215
Vivien Beattie & Michael John Jones ■ Information design and manipulation: financial graphs in corporate annual reports

designers, we found no evidence to suggest that Figure 3


the use of graphic designers resulted in a higher Tomkins plc,
1989: divi­
incidence of material measurement distortions.
dends per
share graph.
Prominence and location of graphs The up-pointing
Companies typically presented key financial arrows serve
graphs prominently as a group at the front of to draw the eye
upwards, as do
the annual report. We termed prominent pre-
the rather dis­
sentation of this nature 'highlighting'. One tracting back­
company, Tomkins plc, presented its graphs on ground vertical
the front cover (figure 3). Two-fifths of the key stripes. The
financial graphs were included in a financial lack of both
gridlines and
highlights section, with on average 0.78 pages
numbers
being dedicated to key financial graphs. attached to
the specifiers
Measurement distortion make it diffi­
As previously discussed, one of the fundamental cult to read
premises underpinning graphical presentation is this graph
that a graph's physical measures should propor- accurately.
tionately represent the underlying numerical
values. Transgressions of this nature are termed The major cause (82 per cent) of materially
measurement distortion. We measured all the discrepant graphs was non-specific errors in
key financial graphs using Tufte's 'lie index' draughtsmanship. In the case of Dowty Group
(Tufte, 1983; Beattie and Jones, 1992b), class- (figure 4) the measurement discrepancies for
ing distortions of over 5 per cent as material D P S and EPS are +140 per cent and + 22 per
measurement distortion. We chose 5 per cent cent respectively. In both cases, there is a conse-
for two reasons: first, Tufte (1983) judged dis- quent exaggeration in the trend - financial per-
tortions in excess of 5 per cent to be 'substan- formance is presented more favourably than is
tial' (p.57); and, second, Pany and Wheeler's warranted by the data.
(1989) review of the auditing literature deter-
A significant minority of material discrepan-
mined 5 per cent to be an important threshold.
cies had specific causes: for example, a non-zero
Of the 465 key financial variables, 142 graphs
axis (17 graphs), a non-arithmetic scale (6
(30 per cent) showed material measurement
graphs), a broken vertical axis (2 graphs) and
distortion. Even more interestingly, in 129
truncation of a negative scale (1 graph). Figure
cases, where both the final year change and the
5, Glynwed plc, illustrates a non-zero axis. Such
trend were upwards, the upward trend was
graphical practices have been much criticised
exaggerated in 98 cases and understated in only
for their violation of the proportionality premise
31 cases. In other words, we found that mea-
(see, for example, Schmid, 1983; Tufte, 1983).
surement distortion was three times more likely
In cases where graphs are casually scanned by
to give a more, rather than a less, favourable
readers, rather than thoroughly processed, the
portrayal of a company's performance. This
risks of the transmission of inaccurate informa-
result was statistically very significant, using the
tion are obviously increased. In the example
binomial test.
which we have chosen to illustrate these risks

Information design journal 7/3 (1994) 211-226

216
Vivien Beattie & Michael John Jones ■ Information design and manipulation: financial graphs in corporate annual reports

DIVIDENDS AGAIN

I NCREASE

S U BSTANTI ALLY,

BY 20 PER CENT

Figure 5
Glynwed plc, 1989: dividend per
share graph. This graph illus­
trates the use of a non-zero axis,
resulting in measurement dis­
crepancy. The wavy lines in the
background distract from the
specifiers, while the absence of
either gridlines or numbers
Figure 4 attached to the specifiers makes
Dowty Group plc, 1989: dividends per share, earnings accurate graph reading relatively
per share, profit before tax, and turnover graphs. The difficult.
judicious bunching of these four graphs and careful scale
selection results in an almost continuous rising trend
across the page. The lack of a scaled y-axis helps to dis­
guise the scale selection. The dividends per share and
earnings per share graphs show graph measurement dis­
crepancies o f + 140% and +22% respectively. In the origi­
nal colour graphs the final year's specifier was a darker
colour than the previous years' specifiers.

Information design journal 7/3 (1994) 211-226

217
Vivien Beattie & Michael John Jones ■ Information design and manipulation: financial graphs in corporate annual reports

(see figure 6, T.I.P. Europe plc), the cause of the


discrepancy was not immediately obvious. The
context in which T.I.P. Europe's EPS graph was
presented held a major pitfall for the unwary
reader. T h e other three graphs presented by this
company on the same page shared a common
format, but with zero axes. They also contained
four, rather than three, specifiers. We suggest
that the unsophisticated graph user may easily
misread this graph.

Construction and design


T h e findings discussed in this section are
summarised in tables 1 and 2.

Background
Graphical theory dictates that the background
should be non-distracting and generally unob-
trusive (e.g. Kosslyn, 1989).5 In 22 cases, how-
ever, we considered backgrounds to be obtrus- Table 1, however, demonstrates that the Figure 6
ive. Backgrounds, such as supermarket shelves majority of companies provided neither a scaled T.I.P. Europe
plc, 1989
and a cafe scene, submerged the specifiers and time (69 per cent) nor a scaled financial variable
Earnings per
often drew attention to themselves rather than axis (73 per cent). Neither were gridlines preva- share graph.
to the graphs. In figure 5, we show Glynwed lent which, in the absence of scaled axes, would The 'three-
plc's D P S graph. T h e wavy lines in the back- have helped graphical perception. In addition, dimensional'
ground draw attention away from the specifiers. five companies ordered their time series uncon- nature of the
specifier caus­
ventionally.
es some ambi­
Framework T.I.P. Europe (figure 6) illustrates both a guity: it is hard
Column graphs, like other Cartesian rectilinear scaled vertical axis and gridlines. However, to tell whether
coordinate graphs, should be drawn with clearly these helpful graphical aids are offset by the we are looking
specified, scaled axes and with time series data dysfunctional non-zero axis. Figure 4 (Dowty at the front of
a wall or the
ordered from left to right. Gridlines may also plc) illustrates the potential problems created
inside of a box.
facilitate effective visual interpretation. Korol by the absence of an effective framework. All The non-zero
(1986), in particular, has emphasised the four graphs are set within a common frame- axis causes a
importance of scaling: 'errors in perception work, which disguises the different scales used graphical dis­
drop almost to zero if adequate scales are pro- to draw the D P S and EPS graphs (in pence) and crepancy of
+720 percent.
vided' (p. 374). the profit before tax and turnover graphs (in
Finally, in the
£m). A potentially misleading impression of a colour original,
5. We acknowledge that in certain cases the graph continuous rising trend is created by judicious the later years
background can be effectively used to gain attention. In 22
ordering of the four graphs, since a false conti- gradually dark­
cases, however, our experience of viewing the graphs sug-
nuity is established. ened in colour.
gested that the background design was dysfunctional,
attracting attention to itself and impeding the underlying
message.

Information design journal 7/3 (1994) 211-226

218
Vivien Beattie & Michael John Jones ■ Information design and manipulation: financial graphs in corporate annual reports

Table 1 Number of key Number


Summary of Proportionality of physical graph financial displaying
graphic report­ measure and numerical values variables feature percentage
ing practices in
the 240 corpo­ Material measurement distortion 465 142 30
rate annual
reports.sur­ Number of
veyed here. companies Number
to which this displaying
Graph type issue applies feature percentage

Unsuitable graph type (bar or pictorial) 156 16 10

Structural components
Background
Obtrusive background 156 22 14
Framework
No scaled time axis 154 106 69
No scaled financial variable axis 154 113 73
No gridlines 154 121 79
Unconventional ordering of time series
(right to left or bottom to top) 156 5 3
Specifiers
No space between graphical elements 148 19 13
Inconsistent spaces between graphical elements 129 9 7
Three dimensional specifier 156 29 19
Labels
No numeric label on financial variable axis 154 105 68
No number attached to individual specifier 156 41 26
Numerically labelled, scaled financial variable axis,
but neither attached numbers, nor gridlines 154 16 10
Financial variable axis located at left hand side 154 9 6
Non-existent or unconventional location of numeric
label on time axis 154 12 8
Unconventional location of number attached
to individual specifier 112 26 23

Typeface
Number attached to specifier not horizontal 115 17 15
Typeface changes within series of numbers
attached to specifier 115 20 17
Numeric year labels not horizontal 156 2 1

Colour
Colour progressively darkened with time 129 8 6
Colour of last year's individual specifier was darker
than all other years which were of uniform colour 129 16 12
Colour of last year's individual specifier was a
different hue to all other years 129 18 14

Information design journal 7/3 (1994) 211-226

219
Vivien Beattie & Michael John Jones ■ Information design and manipulation: financial graphs in corporate annual reports

Specifiers label commonly used for the time axis. On


T h e specifier represents the underlying num- occasion, this general sparsity of labelling led
bers. Many original and innovative specifiers to confusion.
were discovered, including those which reflect- The numeric labelling on the time and finan-
ed the 'image' of the corporate report. For cial variable axes is conventionally located out-
example, Capital Radio (figure 7) adopted side the graph's inner framework, close to the
aerials as a specifier, while Fisons (principally axes. Furthermore, in the case of the financial
a scientific, research-based company) used variable axis, the values at each scale point
pipettes to convey a scientific image. Devenish should be indicated (Schmid, 1979). These
(J. A.) and Marston, Thompson and Evershed, principles were not universally followed. On the
who are both brewers, use lager bottles (figure time axis, two companies failed to provide year
8) and barrels respectively to transmit their cor- details, while in 10 cases the numerical labels
porate image. In the case of Devenish (J. A.) not were not positioned conventionally. For exam-
only the height, but the width, of the bottles ple, in figure 2, Scottish Metropolitan Property
changes. This gives the impression of a conse- plc, a remote key was used. Remote keys are
quent increase in volume creating ambiguity as considered inappropriate for graphical presen-
to which visual dimension is associated with tation (Lewandowsky and Spence, 1989).
the numerical values. On the financial variable axis 105 (68 per
Specifiers and spaces between specifiers cent) out of 154 companies using graphs with
(interspaces) should be of equal width and Cartesian axes gave no numerical labels on the
evenly spaced (Jarett, 1983). This principle was axes. T h e more common practice adopted by
not universally practised. Figure 9, Lovell (Y.J.) the majority of companies (115) was to attach
Holdings plc, for example, portrays a series of numbers directly to the top of the specifiers.
steadily widening columns and steadily narrow- There is a division of opinion in the statistical
ing interspaces. Attention is thereby drawn to graphics literature on such a practice. Without
the columns which represent more recent years. gridlines it can make inter-specifier compar-
Twenty-nine companies employed three- isons difficult. Of those companies attaching
dimensional specifiers (see, for example, T.I.P. numbers, three recorded only the first and last
Europe in figure 6). T h e three-dimensional years, while 26 companies positioned them
nature of the specifiers may cause ambiguity unconventionally (for example, in a horizontal
(see Wainer, 1984, p. 145). For example, when row above all the columns). Figure 2 (Scottish
combined with inadequate scaling, or the lack Metropolitan Property plc) and figure 4 (Dowty
of a common axis for all years, accurate plc) provide examples of specific numbers
perception may be inhibited. attached to specifiers. Figure 3 (Tomkins plc)
illustrates the use of numeric labelling of the
Labelling financial variable axis. These labels are located,
A clear title and effective labelling obviously unconventially, on the right hand side of the
facilitates graphical perception. All the graphs graph. It is also noteworthy that Tomkins'
we encountered, except for two, had titles. In alphabetic financial variable label serves as a
most cases firms considered that the use of such title and that the lack of gridlines and numbers
titles made alphabetical labelling for the finan- on the specifiers makes inter-year comparative
cial variable axis superfluous (for example judgements difficult.
Tomkins plc (figure 3). Nor was an alphabetic

Information design journal 7/3 (1994) 211-226

220
Vivien Beattie & Michael John Jones ■ Information design and manipulation: financial graphs in corporate annual reports

Alignment of type in labels TURNOVER Figure 7


An important element of labelling is the way in £000 Capital Radio pic, 1989: turnover
graph (£000). This graph uses an
which the type is used. Type should be present-
original design - aerials - as a spec­
ed horizontally to facilitate reading and should ifier. The numerical labelling for the
be internally consistent in style and size to avoid final year in the time series, 1989,
confusion. Of the 115 companies that directly is bolder than that used for the earli­
attached numbers to specifiers 13 aligned type er years. There is extensive use of
type aligned at an angle.
vertically, and 4 placed type at an angle. The
angled type was used in a particularly unusual
way by Capital Radio, figure 7. The numbers on
top of the specifiers slope from top left to bot-
tom right, whereas the numeric labels on the
time axis sloped from bottom left to top right.
Another interesting and frequently employed
Figure 8 (below)
method of presentational enhancement was the
Devenish (J.A.) plc, 1989: ordinary
differential emphasis placed upon the final share dividend graph. In this graph
year's number. One of the 20 companies adopt- both the height and width of the
ing this practice was Lovell (Y. J.) Holdings plc lager bottle specifiers change, cre­
(figure 9). Here bold type is used to stress the ating the impression of a greater
importance of the final, and most impressive, increase in the dividend than is war­
ranted. Lack of axes compound the
year's performance.
difficulties for the graph reader.

Colour
A particularly interesting feature of many
graphs was the association of darker colours
with the later years of a time series graph. In the
case of figure 2, Scottish Metropolitan Property Ordinary share dividend
plc, there is, for example, a progressive darken- £000's
ing of colour from light blue in 1985 to dark
purple in 1989. In this case, and 7 others, the
numerical quantities portrayed increased con-
tinuously over time, while the intensity of the
colour progressively darkened. There is, thus, a
prima facie ordered ordinal relationship
between colour intensity and numerical quanti-
ty. There appears to be little consensus in the
literature about such a relationship. Tufte
(1983, p.154) suggests that intensity and quanti-
ty are directly proportionate. Kosslyn (1989,
p.204), on the other hand, believes that if
colour is used, then lighter colours should cor-
respond to higher values.
In addition to this progressive use of colour,

Information design journal 7/3 (1994) 211-226

221
Vivien Beattie & Michael John Jones ■ Information design and manipulation: financial graphs in corporate annual reports

we commonly found the colour of the final Turnover


year's specifier to be either darker than the uni- £ millions
formly coloured preceding years (12 per cent)
or of a completely different colour to the single
colour of earlier years (14 per cent). This pre-
sentational enhancement of the final year coin-
cides with Kosslyn's (1989) observation that
abrupt changes draw the eye. The eye is thus
drawn to the last year of a time series. In figure
9, Lovell (Y.J.) Holdings plc's graph illustrates
this point: 1989, the final year's turnover graph,
is yellow; by contrast 1985-88 are blue. The
change in colour naturally draws the eye to the
1989 column.

Special and illusory effects


We identified 75 examples of various special
effects (see table 2). Of these we judged that 44
were forms of presentational enhancement
while the remainder had variable effects on our
85 86 87 88 89 85 86 87 88
perception. Four types of presentational
enhancement can be distinguished: sloping pre- Figure 9
sentation, emphasis on top of specifiers, non- Lovell (Y.J.) Holdings plc, 1989: turnover graph.
consecutive time series and embedded graphs. This graph has specifiers which gradually widen through­
out the time series and interspaces which gradually nar­
Generally, all the examples of presentational
row. The combination of bold typeface used for the final
enhancement depend upon two fundamental year's labels, different colour (in original graph) used for
characteristics of graphical perception. First, the final year's specifier, and sloping specifiers draws the
graphs are read from left to right, in the same eye upwards to the right, enhancing the perception of a ris­
way that we read print. Second, the top of the ing trend. The position of the baseline with respect the
platform is unclear.
graphs, which forms the trend line, is the focal
aspect of a graph. T h e consequence of this rea-
soning is that, where presentational enhance- themselves sloped from left to right. Lovell
ment occurs, the graphical techniques focus on (Y.J.) (figure 9) utilises this technique. Third,
the top right. the labelling sloped (see discussion above and
Sloping presentation, therefore, typically figure 6, Capital Radio). Fourth, in two cases
slopes from left to right. We identified five dif- the individual graphs themselves contained no
ferent ways in which angled presentation sloping elements; however, the series of graphs
enhanced a company's graph. First, the were arranged across the page, in stepwise fash-
specifiers within a graph sloped back in a step- ion, from bottom left to top right. Finally,
wise manner. The consequence of this may well Dowty's graphs (figure 4) were presented in a
be that, as in Laura Ashley Holdings plc (figure carefully arranged, ordered sequence: D P S ,
10), the later years may appear more impressive EPS, profit before taxation and turnover. The
than the facts warrant. Second, the specifiers almost subliminal impression is that we have a

Information design journal 7/3 (1994) 211-226

222
Vivien Beattie & Michael John Jones ■ Information design and manipulation: financial graphs in corporate annual reports

Table 2 Feature Number of


Special and companies
illusory effects displaying
feature
used in the
240 reports Presentational enhancement
surveyed.
Use of slope
stepwise specifiers 6
specifiers 5
labels 4
stepwise graphs 2
combined series of graphs 1

Top of specifier emphasised


multishading 9
background shading 4
marking 4
upward-pointing arrows 3
top faces of 3D columns 2

Two year nonconsecutive time series 2


Embedded graphs 2

total 44

Variable effects

Crosshatching 11

Shadows 9

Individual specifiers 7
Figure 10
Obscured baseline 4 Laura Ashley Holdings plc, 1989: sales, operating profit,
and profit before taxation graphs. These graphs slope
total 31 towards the top right, creating the possibility that later
years are perceived to be progressively larger than they
actually are. The baselines of the sales and operating
graphs are obscured by the profit before taxation graph,
making it difficult for readers to make accurate judge­
ments.

Information design journal 7/3 (1994) 211-226

223
Vivien Beattie & MichaelJohnJones ■ Information design and manipulation: financial graphs in corporate annual reports

continuous rising trend of 20 years. In fact, we vibratory effect called Moire vibration and is
have four distinct five-year time series. therefore best avoided in colour graphs
We identified five broad design aspects which (Schmid, 1983). Second, nine specifiers had
emphasised the top of individual specifiers. shadows attached to them. These shadows had
First, multishading occurred, where each col- no obvious effect but, in certain types of projec-
u m n was shaded from bottom to top - with the tion, enhanced the perception of column length
top being shaded darkest. Second, differential (Schmid, 1983). Third, there were seven
shading in the background adjacent to the top instances where individual specifiers were dis-
right of the graph drew attention to this area of tinguished in some (previously undiscussed)
the graph. Third, the tops of the columns were way. For example, in one case they formed part
emphasised, for example, by drawing them in of an underlying picture. This proved distract-
black. Fourth, three companies (including ing. Finally, in four cases the baseline of the
Tomkins plc, figure 3) used upward pointing graph was obscured. Thus, as in the cases of
arrows. This is an interesting psychological Lovell (Y.J.) Holdings plc (figure 9) and Laura
ploy, since subtle connotations of directionality Ashley Holdings plc (figure 10), uncertainty
are imputed to the time series. Finally, the top reigned as to where the baseline began. Clear
face of three-dimensional columns was empha- judgement of the length of individual specifiers
sised, the effect of which is once again to draw was, therefore, impeded.
one's attention to the top of the graph.
We identified two further forms of presenta-
tional enhancement. First, one company's Conclusion
graphs showed specifiers for only 1980 and
1989. The immediate, almost unconscious,
This study discusses the graph design issues
inference is that this represents two consecutive
which arise from a domain-specific empirical
years. By omitting the years 1981-88, the initial
investigation into the particular graphical
impression is of a substantial and impressive
reporting practices of 240 leading UK firms in
leap in turnover and profit. Second, we identi-
their 1989 corporate annual reports. The main
fied the use of embedding, where graphs of a
findings are that 79 per cent of companies used
different type and nature are 'embedded'
graphs, 64 per cent of all graphs were bar/col-
among other very different graphs. For exam-
umn graphs and the most popular variables
ple, one company included a graph entitled
graphed were turnover, profit before taxation,
'1989 Turnover by Section (continuing busi-
earnings per share and dividends per share.
nesses)' together with profit before tax, EPS and
90 per cent of all key financial graphs were
D P S graphs. However, this is a non-time series,
in colour.
sectoral turnover graph, which is arranged and
Many of the graphs were poorly designed and
presented, superficially at least, as if it were a
constructed. 10 per cent of time-series graphs
time series of the fourth key financial variable.
were drawn using a bar or pictorial graph
We isolated four categories of special effects
instead of the more appropriate column graph.
which did not necessarily enhance presentation.
In addition, many graphs suffered from obtru-
First, cross-hatching was used in 11 cases.
sive backgrounds, lack of a clear framework,
Cross-hatching is where the interior of the spec-
non-uniform specifiers, poor labelling, inappro-
ifier consists of parallel or criss-crossed lines.
priate typeface and cross-hatching.
Cross-hatching may cause a disturbing optical,

Information design journal 7/3 (1994) 211-226

224
Vivien Beattie & Michael John Jones ■ Information design and manipulation: financial graphs in corporate annual reports

There was also evidence of systematic selec- would argue that the development of graphical
tivity, measurement distortion and presenta- guidelines is long overdue. We believe that
tional enhancement. The decision to choose a comprehensive guidelines would improve the
graphical presentation was contingent upon representational faithfulness of graphs.
financial performance. In short, the annual
reports of companies were more likely to
include graphs when corporate financial per- Acknowledgements
We thank the 240 companies who sent us their annual
formance was favourable rather than unfavour- reports, and particularly the 9 who gave us permission to
able. Turning to measurement distortion, we reproduce the graphs illustrated in this article. The
found that companies were three times more financial support of the Chartered Association of Certi-
likely to include graphs in their annual report fied Accountants is gratefully acknowledged. On the
which exaggerated, rather than understated, a individual level, we are extremely appreciative of the help
and encouragement of the editor of IDJ, Paul Stiff, and
rising trend. Finally, we commonly found
for the valuable comments of the referees.
instances where firms appeared to use colour,
angled presentation and emphasis on the top of
specifiers to enhance their graphs.
These findings have implications both for the References
development of knowledge about graphs and
for the practical role of graphs in the corporate Accountant's Magazine, The , 1992. Annual Reports 'Not
True and Fair', Say Shareholders, September 1992.
financial reporting process. Systematic,
Beattie, V.A. and Jones, M.J. 1992a. The communication of
domain-specific, empirical studies, which have information using graphs in corporate annual reports.
been relatively neglected to date, uniquely offer Certified Research Report 3 1 , Chartered Association of
the opportunity to study the practical applica- Certified Accountants. London: Certified Accountants
Educational Trust
tion of graphical theory. They provoke new
Beattie, V.A. and Jones, M.J. 1992b. The use and abuse of
insights into issues such as the role of colour in graphs in annual reports: a theoretical framework and
graphical displays as well as complementing empirical study, Accounting and Business Research,
theoretical studies. There is thus a need for Autumn, vol. 22(88), pp. 291-303
more such substantive studies of actual prac- Cleveland, W.S. and McGill, R. 1987. Graphical perception:
the visual decoding of quantitative information on graphi­
tice. In addition, there is also a need for comple-
cal displays of data, Journal of the Royal Statistical
mentary studies which explore the way in which Society, vol. 150(3), pp. 192-229
people read, interpret and make sense of graphs DeSanctis, G. and Jarvenpaa, S.L. 1989. Graphical investi­
in specific contexts. gation of accounting data for financial forecasting: an
T h e main practical implication of this study is experimental investigation, Accounting, Organizations and
Society, vol. 14(5/6), pp. 509-525
that it demonstrates the need for financial Garceau, S., Oral, M. and Rahn, R.J. 1988. The influence of
reporting graphical guidelines to be developed. data presentation mode on strategic decision-making per­
At present, the form and content of financial formance, Computers and Operations Research, vol.
statements contained in the annual reports of 15(5), pp. 479-488
Hanson, J.D. 1989. Developments in Financial Reporting
the UK and other developed countries, such as
Over the Last 20 Years, Financial Reporting 1988-89. A
the USA, is regulated and audited. Graphs, on Survey of UK Published Accounts, (20th anniversary edi­
the other hand, being voluntary, are subject to tion), pp. 3-13
no external validation. Given our findings con- Jarett, I.M. 1983. Computer Graphics and Reporting
cerning systematic selectivity, measurement Financial Data. New York: John Wiley.
distortion and presentation enhancement, we

Information design journal 7/3 (1994) 211-226

225
Vivien Beattie & Michael John Jones ■ Information design and manipulation: financial graphs in corporate annual reports

Korol, J.K. 1986. Graphical perception and the representa­


tion of financial information, Georgia Journal of
Accounting, vol. 7, pp. 147-157
Kosslyn, S.M. 1989. Understanding charts and graphs,
Applied Cognitive Psychology, vol. 3, pp. 185-226
Leivian, G.M. 1980. How to Communicate Financial Data
More Efficiently, Management Accounting (USA),
vol. 61(1), pp. 3 1 - 3 4
Lewandowsky, S. and Spence, I. 1989. The perception of
statistical graphs, Sociological Methods and Research,
vol. 18(2/3), pp. 200-242
Macdonald-Ross, M. 1977. How numbers are shown:
a review of research on thepresentation of quantitative
data in texts, Audio-Visual Communication Review,
vol. 25(4), pp. 359-409
Pany, K. and Wheeler, S. 1989. Materiality: an inter-industry
comparison of the magnitude and stabilities of various
quantitative measures, Accounting Horizons, vol. 3(4),
pp. 71-78
Peter Prowse Associates 1993. The company report report -
1993. Leatherhead, UK: Peter Prowse Associates
Schmid, C.F. and Schmid, S.E. 1979. Handbook of graphic
presentation. 2nd edn, New York: John Wiley
Schmid, C.F. 1983. Statistical graphics: design principles
and practices. New York: John Wiley/lnterscience
Tufte, E.R. 1983. The visual display of quantitative
information. Cheshire, CT: Graphic Press
Wainer, H. 1984. How to display data badly, The American
Statistician, vol. 38(2), pp. 137-147

Information design journal 7/3 (1994) 211-226

226

Das könnte Ihnen auch gefallen