Environmental Impact Assessment Review 73 (2018) 114–121

Contents lists available at ScienceDirect

Environmental Impact Assessment Review

journal homepage: www.elsevier.com/locate/eiar

Does size matter? An evaluation of length and proportion of information in T

environmental impact statements
⁎
Germán Marino Rivera Fernández, Ludmila Ladeira Alves de Brito, Alberto Fonseca
Graduate Program in Environmental Engineering, Federal University of Ouro Preto, Minas Gerais, Brazil

A R T I C LE I N FO A B S T R A C T

Keywords: For decades, authors and institutions have argued that the quality of Environmental Impact Statements (EISs) is
Environmental impact statement (EIS) somehow affected by the volume of information they contain. Both too little and too much information can be a
EIS length problem. However, very few academic studies have addressed the issue of EIS length in detail. The objective of
EIS quality this article was to systematically analyze the volume of information presented in EISs, using Brazil as the em-
Information management
pirical context. More specifically, this study evaluated the volume and proportion of information disclosed in 49
Environmental impact assessment
Decision-making
Brazilian EISs. This study also tried to identify sectorial variations and whether variables such as project size and
number of pages in Terms of References are likely determinants of information volume. > 146 thousand pages of
EIS information were scrutinized in two rounds of content analysis. Data were organized in spreadsheets and
then coded and analyzed through various descriptive and inferential statistical techniques. Overall, findings
corroborate the fact that EISs are now significantly longer than the early ones, and still heavily loaded with
baseline information. The average number of pages in EISs and in Non-technical Summaries was found to be
2993 and 94, respectively. Kruskal-Wallis and linear regression tests indicated that EIS length is likely affected
by a combination of variables, including project size, territorial and sectorial characteristics. Such findings
suggest that the historical approach of setting page limits to EISs through regulations and Terms of References is
no longer appropriate for EIA practice in connection with large enterprises in Brazil, and arguably elsewhere.
The article discusses its practical and academic implications, and highlights the need to further investigate the
actual impacts of EIS length on decision-making.

1. Introduction decisions (Glasson et al., 2005). The realization of this scenario, how-
ever, is complicated by various problems in the generation and review
Environmental Impact Assessment (EIA) has become one of the of EISs (Morrison-Saunders et al., 2014).
world's most influential environmental policy tools. Virtually every Among the most frequently debated issues in connection with EISs is
country on Earth use some form of EIA in the decision-making processes ‘lack of quality’, which can significantly affect the overall effectiveness
of projects and strategic undertakings (Morgan, 2012). While the of the EIA process (Ross et al., 2006). While the concept of quality has
practice of EIA has gone through significant changes since it was first been loosely defined in the EIA literature, as Bond et al. (2018, p. 50)
regulated in the United States in 1969, many of its early challenges still recently noted, in the context of positivist/rationalist theory, where
occupy scholars and analysts all over the world (Lawrence, 2013). EIA, better information means better decision-making, “(…) the quality of
as Sadler (1996, p.13) put it, is “a systematic process of evaluating and the information underpinning and presented in the environmental im-
documenting information on the potentials, capacities, and functions of pact report, has been assessed as the key quality measure in a number of
natural systems and resources in order to facilitate sustainable devel- studies (…)”. In the early 1990s, Wathern (1990) had already pointed
opment planning and decision making in general”. Such information, out that the quality of information in EISs had been investigated since
which is usually presented in a document entitled Environmental Im- the mid-1970s. Since then, numerous studies have revisited this issue,
pact Statement1 (EIS), is expected to inform authorities to make better often based on the Environmental Statement Review Package

⁎
Corresponding author at: Morro do Cruzeiro, s/n, Ouro Preto, MG 35400-000, Brazil.
E-mail address: [email protected] (A. Fonseca).
1
The documents that carry information related to the impact assessments studies (e.g. project description, baseline information, impact prediction, mitigation
measures, etc.) can be described in a variety of terms and acronyms. This paper adopts the term Environmental Impact Statements (EIS), as it is often used in grey and
academic literature. Other common terminology includes Environmental Impact Report, Environmental Statement and Environmental Impact Study.

https://doi.org/10.1016/j.eiar.2018.08.002
Received 9 May 2018; Received in revised form 7 August 2018; Accepted 8 August 2018
0195-9255/ © 2018 Elsevier Inc. All rights reserved.
G.M.R. Fernández et al. Environmental Impact Assessment Review 73 (2018) 114–121

developed by Lee and Colley (1991). The Lee and Colley Review 2. EIS length: a puzzling quality dimension
Package is basically a checklist of quality criteria that were originally
developed to assist in assessing the quality of EISs submitted to the EIS quality has a variety of dimensions, such as completeness, re-
United Kingdom (UK) authorities. It has been inspiring many studies in liability, comprehensibility, clarity, accessibility, and length. The latter,
various jurisdictions worldwide. A study by Anifowose et al. (2016) while not always realized by practitioners and regulators, affects not
found that the Lee and Colley Review Package underpinned the meth- only the shape/form of EISs; it can affect several aspects of the whole
odology of 26 out of the 47 articles (55%) on EIS quality that they had EIA process, such as cost, time, decisions and participation. One of the
identified in the literature. first peer-reviewed studies to shed light on the link between EIS quality
The Lee and Colley Review Package and many other checklist-based and EIS length was the one by Lee and Brown (1992). This study, while
frameworks for assessing quality in EIS, as Põder and Lukki (2011, p. reviewing the content quality of a sample of 83 UK EISs, realized that
35) argue, “(…) do not reflect the quality of EIS at large, but only one the high-quality statements tended to be the long ones (with > 100
quality component, namely completeness of relevant information”, that pages), whereas the low-quality statements were found to be the
is, the degree to which a particular EIS addresses a list of topics. Various shortest (< 25 pages). A similar relationship was found the next year in
other factors may affect EIS quality. For example, a study by João a study that compared that UK sample with another sample of 40 Irish
(2002) found that scale choices (both in terms of spatial extension and EISs (Lee and Dancey, 1993).
geographical detail) can significantly affect the quality of EISs. Another The relationship between EIS quality and EIS length had already
factor that may affect quality is the length or amount of information been addressed in early EIA regulations. For example, in 1978 the
presented in EISs. This factor was particularly emphasized by the In- United States Council of Environmental Quality (CEC) changed the
stitute of Environmental Management & Assessment (IEMA) in its re- NEPA regulations, so agencies were “(…) directed to write concise EISs,
view of EIA practice in the UK: which normally shall be less than 150 pages, or, for proposals of unu-
sual scope or complexity, 300 pages” (CEC. National Environmental
IEMA's research has found that EIA practitioners recognise the fact that
Policy Act, 1978, p. 55978). The CEC regulation encouraged agencies to
generally the length of a typical UK ES [terminology for impact as-
prepare analytic rather than encyclopedic EISs, and set page limits not
sessment statement in UK] acts to reduce the value of the EIA's findings
only for the whole statement but also for its cover (< 1 page) and its
it contains to the majority of audiences. It is clear that action is needed by
summary (< 15 pages). According to Wathern (1990, p. 26), this reg-
EIA practitioners and all those involved in practice to begin to produce
ulatory change was driven by fear of litigation: “approximately a 1 in
more proportionate ES that provide clear and focussed information.
10 chance of a suit being filed against an EIS encouraged agencies to
(IEMA, 2011, p. 87).
adopt an ‘encyclopedic’ approach to EIA, particularly in the early years
While practitioners and analysts usually agree that the volume of of EIA”. That 150-page limit was later endorsed as a guideline by UK's
information can somehow affect the quality in EISs as well as the Department of the Environment, Transport, and the Regions, which
overall effectiveness of the EIA process (Lyles, 2017), very few aca- recommended the following:
demic studies have evaluated this issue in detail. In fact, quite often,
For projects which involve a single site and relatively few areas of sig-
studies on EIS quality completely overlook the length or volume of
nificant impact, it should be possible to produce a robust ES of around 50
information in EISs (e.g. Kabir and Momtaz, 2014; Veronez and
pages. Where more complex issues arise, the main body of the statement
Montaño, 2017). This is to a great extent a consequence of the meth-
may extend to 100 pages or so. If it exceeds 150 pages it is likely to
odologies of these studies, which tend to target quality mainly through
become cumbersome and difficult to assimilate and this should generally
the aforementioned lenses of ‘completeness of information’. The
be regarded as a maximum. Any additional information should be in-
shortage of studies on volume of information in EIS is worrisome, as the
corporated in appendices (Department of the Environment Transport
increasing availability of data and growing societal expectations of
and the Regions, 1995, p. 35).
content are creating incentives for lengthy EISs. The effects of this trend
on decision-making need further and continuing investigation, parti- Interestingly, the 150-page limit, as a maximum threshold for EIS
cularly in contemporary Brazil, where pressures are mounting for length, has been endorsed in other peer-reviewed studies. Glasson et al.
streamlining its EIA system (Fonseca and Rodrigues, 2017; Fonseca (1997), for example, in their review of EIS quality in the UK, concurred
et al., 2017). that length could be used as a proxy of EIS quality, as statements longer
The objective of this article was to systematically analyze the than 150 pages tended to have undesirable effects on the decision-
amount of information that is being presented in EISs, using Brazil as making process. The World Bank, in the early 1990s, also encouraged
the empirical context. More specifically, this study evaluated the vo- submissions no longer than 100 pages (World Bank, 1991, p. 8).
lume and proportion of information that were presented in 49 Brazilian However, burgeoning empirical data in the mid-1990s exposed a
EISs, and discussed implications to decision-making. This study also more complicated picture of the relationship ‘length versus quality’.
tried to identify sectorial variations and whether variables such as One of the most cited EIA studies of that period (Sadler, 1996) included
Project Size and number of pages in Terms of Reference are likely de- a survey of EIS reports that classified them into three length categories:
terminants of information volume in EISs. Although addressing the small (< 50 pages); medium (51–400 pages); and long (> 400 pages).
Brazilian empirical context, this study is likely to be relevant to a broad The survey found that many EISs were longer than 150 pages, but
international audience concerned with EIS quality and EIA effective- without any discernible pattern. Interestingly, in that survey, a ‘long
ness. This study is arguably the most thorough empirical analysis of statement’ was perceived to be those documents longer than 400 pages,
length and volume of information in EISs to date. More than 146 rather than 150 pages.
thousand pages of EIS information were systematically evaluated by the In the 2000s, studies continued to argue that EIS length is somehow
authors through descriptive and inferential statistics. Findings may in- related to EIS quality (e.g. Cashmore et al., 2002; Lawrence, 2003;
spire future research and underpin policy-making in a variety of con- Morrison-Saunders et al., 2006). In the US, a NEPA Task Force (2003)
texts. found that EISs were becoming significantly longer than the early ones,
This article is organized in five sections, including this introduction. reaching up to 2000 pages-long documents. Some jurisdictions in
The next section reviews previous studies and regulations that ad- Europe (e.g. IEMA, 2011; and The Scottish Government, 2013) started
dressed the issue of EIS length. The third section explains the metho- to emphasize the need for ‘proportionality of information’, rather than a
dology used to collect and analyze data in connection with Brazilian fixed number of pages, as EISs tended to carry too much baseline in-
EISs. Findings are presented and discussed in section four, and con- formation without relevant implications to decision-making (Anifowose
cluding remarks are finally drawn in section five. et al., 2016). More recently, Landim and Sánchez (2012) found that, in

115
G.M.R. Fernández et al. Environmental Impact Assessment Review 73 (2018) 114–121

Brazil, EISs were also becoming lengthier, a fact that was corroborated length.
by Paiva et al. (2015) and IBAMA (2016). In light os such trends, both One should note, however, that 49 EISs do not represent a statisti-
Borioni et al. (2017) and Lyles (2017) argued that improving scoping is cally ‘representative’ sample of EISs produced in Brazil between 2005
a key way to reduce document size. and 2015. The population of EISs filed in IBAMA in this period, while
Altogether, these studies suggest that the issue of EIS length is not not clear in its website, is likely in the order of multiple thousands of
an easy one: on the one hand, developers, when preparing EISs, are statements related to > 20 different sectors. The purposive sample se-
encouraged to be as complete as possible, thus avoiding delays with lected here, however, was sufficient for the purpose of exploring sec-
requests for further information and eventual litigation; on the other torial variations and likely drivers of length. The sample included 7 EISs
hand, they are also encouraged to be focused, including only relevant of 7 different sectors. To add further context to the analysis, this study
information to decision-making. However, guidelines and regulations, counted the number of pages of the EISs' respective Terms of Reference
most of the time, are not specific about what ‘focused’ means in prac- and also characterized declared investments, impacted municipalities,
tice, i.e. how many pages or volume of information developers should and affected areas.
use for each section, chapter and volume of the EIS document. Terms of This study adopted a checklist of categories and sub-categories to
Reference are also getting longer, imposing the need for more and more evaluate the proportion of information in each page of the 49 EISs. The
information. Scoping is, of course, a means to address EIS length, but terminology used to describe the sub-categories was inspired by the Lee
how to balance completeness and length - or as Snell and Cowell (2006) and Colley Review Package. The idea was to analyze how much in-
say, how to balance precaution and efficiency - is still a rather sub- formation was used to address the typical EIS topics, using a termi-
jective issue. nology that is commonly used worldwide.
Moreover, the likely negative effects of lengthy EISs have been Like in previous studies, the approach adopted here to measure the
mostly anecdotal; very few studies have tried to understand the effects volume of information was to count pages. The counting was carried
of length on the review and decision-making process. Most publications out by each co-author separately. Findings from this first round of page-
analyzed here address EIS length as a marginal issue, and often based counting were later contrasted in a co-authors meeting that revealed a
on descriptive statistics; the only exception was found to be the study number of divergent interpretations of EIS content. This led to a review
by Drayson et al. (2015), which undertook a few inferential statistical of the content analysis criteria and to another round of page-counting.
tests. In any case, until the time of writing this article, very few studies The findings presented in the next section are, therefore, the result of
(if any) had addressed EIS length as the main topic of research, as this 7–8 months of two-rounds of content analysis. The raw data generated
study set out to do. in this process were organized in spreadsheet software and is fully
available for download and future use by different authors and in-
3. Methodology stitutions. The complete spreadsheet presents the information in var-
ious sub-categories and brings additional details related to the number
This study aimed at analyzing the amount of information presented volumes, section, and chapters of each EIS, as well as to the exact
in EISs, using Brazil as the empirical context. This was carried out number of pages covered by each sub-category across the documents'
through comparative (qualitative and quantitative) content analyses of many sections.
a sample of 49 Brazilian EISs submitted to the Brazilian Institute of the Findings were finally evaluated in SPSS through descriptive and
Environment and Renewable Natural Resources (whose acronym in graphical statistics. A number of inferential tests (linear regressions and
Portuguese is IBAMA), which is the government agency responsible for Kruskal-Wallis) were carried out to evaluate the potential associations
project permitting and approval at the federal level. Content analysis of EIS length with Project Size and Terms of Reference. It is important
has long been a methodological choice in many previous studies on EIS to emphasize that these tests did not include blank pages in the cal-
quality (Anifowose et al., 2016). However, as Julien (2008) notes, the culations. It is also important to emphasize that, during the content
level at which “(…) content analysis occurs varies widely, from obvious analysis, the authors realized that there was a vast amount of in-
surface-level groups of similar responses to a particular interview formation in appendices and annexes that could have relevant im-
question to deeper inductive insights inferred from more sustained, plications to decision-making. As opposed to simply counting the
iterative, and recursive interaction with textual data” (p. 120). Here, overall number of pages in those annexes, the authors evaluated their
content analysis was carried out in a superficial manner, as the objec- contents and respective proportion of information.
tive was simply to classify the broad themes disclosed by developers in The documents analyzed here were all digital documents officially
each page of the EIS. disclosed by the Brazilian agency. However, they were merely digital
The sampling approach used here was purposive, i.e. “type of versions of print documents; the agency review process is still based on
nonprobability sampling in which the units to be observed are selected the traditional rationale of reviewing contents organized in a ‘linear’
on the basis of the researcher's judgment about which ones will be the format. While the technology that is used in the preparation of EISs,
most useful or representative” (Babbie, 2010, p. 193). The sample of 49 such as computers, software, scanners, image processing equipment,
Brazilian federal-level EISs, published between 2005 and 2015, was etc., has improved substantially in the past decades, it has not yet af-
arbitrarily selected because they are arguably what Gerring (2007, p. fected the overall structure of documents filed in Brazil's federal-level
101) calls ‘extreme case’, i.e. “corresponds to a case that is considered EIA. Nonetheless, it is clearly affecting the use of different page formats
to be prototypical or paradigmatic of some phenomena of interest”. and layouts. While obvious, this trend raises not-so-obvious questions
Federal-level EIAs, in Brazil, tend to be related to projects that affect about how to measure the length of an EIS. Many of the reviewed pages
large areas and populations, which, in turn, are more likely to result in were A3 or A2-sized pages with graphical information, which demand
lengthy EISs. A recent study by IBAMA (2016) found that, at the federal different review approaches from A4 or letter-sized pages with textual
level, Brazilian EISs were found to have, on average, > 1000 pages. In information. While realizing that such pages can demand more time
other words, the sample of federal-level Brazilian EISs addressed here and attention to review, for the purpose of this study, they were
are arguably clear cases of lengthy EISs (at least based on the average counted as one page.
number of pages mentioned in the reviewed literature). Moreover, the
selected EISs were filed in licensing applications that were all approved 4. Results
by the federal agency, which suggests that, although lengthy, they met
minimum quality requirements by the agencies. The analysis of their 4.1. Proportion and average number of pages
structure, number of pages and proportion of information gives an
opportunity to further explore the causes and implications of EIS As shown in Table 1, this study found that the average number of

116
G.M.R. Fernández et al. Environmental Impact Assessment Review 73 (2018) 114–121

Table 1
– Number of pages and additional characteristics of the 49 Brazilian EISs analyzed in the study.
Sector Project EIS publication Pages Pages with Non-numbered Project declared Pages of non- Pages of the Number of
EIS code year without annexes Blank Pages investments (in US technical term of impacted
annexes dollars)b summary reference municipalities

Railways RW1 2013 652 1248 0 81,158,077 123 10 2

RW2 2010 1914 2046 0 515,164,694 119 n/a 29
RW3 2009 2210 2986 0 1,060,125,173 28 51 37
RW4 2008 846 1262 0 571,141,445 100 13 21
RW5 2008 1057 1391 0 808,236,542 101 13 20
RW6 2010 2018 4663 0 54,950,901 41 23 3
RW7 2011 1378 1812 0 380,928,935 112 23 1
Highways HW1 2014 845 4464 0 800,981,021 68 19 19
HW2 2013 883 1095 28 18,130,268 78 29 2
HW3 2010 1574 1634 0 141,040,645 85 33 2
HW4 2010 540 873 6 128,218,768 48 22 2
HW5 2009 579 793 0 16,585,069 98 7 6
HW6 2008 1134 1147 0 n/aa 53 n/a 1
HW7 2007 1172 1172 0 71,965,561 76 46 1
Transmission Lines TL 1 2015 2469 5495 0 1,498,578,670 68 41 12
TL 2 2015 1936 3755 270 963,784,848 88 46 29
TL 3 2014 2362 3051 0 306,836,443 230 25 19
TL 4 2010 3046 11,824 315 200,711,041 116 32 83
TL 5 2014 2175 3965 150 n/a 122 36 9
TL 6 2013 2755 4935 0 507,335,640 136 25 45
TL 7 2009 3198 4410 349 369,200,432 106 57 83
Hydroelectric H1 2011 2021 3279 201 496,863,828 80 74 8
Power Plants H2 2010 1906 2632 302 1,133,340,142 68 68 2
H3 2009 2268 2289 0 227,034,522 96 51 5
H4 2011 1876 3062 337 949,404,082 110 67 2
H5 2005 493 513 0 1,216,091,208 54 16 2
H6 2012 2172 2267 84 192,671,361 88 97 2
H7 2009 1677 2142 284 143,019,866 68 79 3
Mining Plants M1 2012 277 303 0 519,297,309 121 17 2
M2 2012 2046 2493 0 319,567,575 142 25 3
M3 2009 1425 1493 0 21,675,130 105 29 1
M4 2014 1465 1501 2 275,808,039 37 21 1
M5 2011 2498 5403 14 207,026,595 113 n/a 1
M6 2007 745 1151 2 1,655,208 210 n/a 1
M7 2010 3260 4913 37 22,896,209 137 26 1
Ports P1 2009 868 2687 4 89,191,311 40 20 1
P2 2009 2364 3441 1 1,041,567,142 118 18 1
P3 2009 1543 2281 0 1,487,660,348 60 17 1
P4 2011 1438 3371 4 683,187,764 94 23 1
P5 2009 1594 5998 0 706,750,115 69 33 1
P6 2008 1138 2266 3 446,872,595 56 17 1
P7 2010 1334 1972 0 276,712,129 290 9 1
Oil & Gas OG 1 2011 1741 3036 128 3,921,083,710 64 93 n/a
OG 2 2010 1506 2301 50 8,912,349,204 60 57 n/a
OG 3 2014 3544 5466 42 41,371,205,821 69 65 n/a
OG 4 2012 1003 2660 14 119,837,841 89 39 n/a
OG 5 2011 6620 8698 50 4,417,506,187 74 n/a 2
OG 6 2007 1397 1988 2 1,526,542,197 42 39 n/a
OG 7 2012 1479 3037 48 2,164,471,132 74 57 n/a
Total 86,441 146,664 2727 73,767,824,642 4624 1608 469
Average 1764 2993 56 1,731,624,739 94 37 11

a
n/a = not applicable / - means that data were not available in reviewed documents.
b
Investment values were declared by developers in the EIA process at the time of EIA initiation. The values in the table, originally declared in Brazilian currency
(Reais), were adjusted for inflation (using IGP-M rates) and converted to US dollars using a conversion rate of 3.5.

pages in the sample of EISs was found to be 1764 (without annex) or The same argument is valid for non-technical summaries (NTS),
2993 (with annex). It also found that the total number of pages in an- whose earliest page limits mentioned in the literature (15 pages) are 6
nexes can be almost as large or in many cases larger than the total times smaller than the average number of pages found here (94 pages),
number of pages in the main EIS document. Content analysis revealed as shown in Table 1. Interestingly, this average (94 pages) corresponds
that the annexes may carry not only baseline information, but also to about 3% (with annex) or 5% (without annexes) of the average
impact prediction content that is clearly relevant in decision-making. number of pages in the complete EIA reports. In the United States,
Thus, it seemed appropriate to include the annexes' contents in the regulations expect NTSs to be limited to 15 pages, that is, 10% of the
statistical tests presented further below. maximum report length (150 pages) (CEC. National Environmental
The fact that the EISs analyzed are on average almost 3000-page Policy Act, 1978). The meaning of these contrasting figures is open for
long strongly suggest that the historical 150-page limit endorsed by debate, as the desirable proportion of size between the main EIS
many institutions and scholars is no longer appropriate for EIA practice document and its NTS has been largely overlooked by scholars.
in connection with large enterprises in Brazil, and arguably elsewhere.
Even in the United Stated, where the 150-page limit was regulated, EISs 4.1.1. Environmental impact statements or environmental baseline reports?
have been found to have on average 700 pages (Lyles, 2017). In order to further explore the potential reasons why EISs are

117
G.M.R. Fernández et al. Environmental Impact Assessment Review 73 (2018) 114–121

Fig. 1. Proportions of the Main Categories of Contents in Each EIS.

becoming lengthy documents, this study classified the types and pro- with the size of projects. This potential association has not been clearly
portion of information in each page of the 49 s EISs. As previous re- signaled in previous studies, but is a logical assumption. After all, the
search suggested (e.g. Landim and Sánchez, 2012), EISs have been in- larger the project, the more area and municipalities it is likely to affect,
cluding large proportions of baseline information in comparison with and the more baseline information one can expect to find. To explore
other common EIS contents like impact prediction and alternatives this potential association, this study arbitrarily adopted three proxies
evaluation. This study tried to quantify such proportions. To guide the for project size available in the reviewed EISs: a) project construction
content analysis, this study adopted a list of categories and sub-cate- area (in hectares); b) the number of municipalities directly impacted by
gories. The complete results of the content analysis are presented in the project; and c) total declared investments (in US dollars).
spreadsheet format and available for download. Summarized findings Scatterplots were created to correlate the number of pages in each EIS
of the main categories (e.g. pretext contents, project description, with their respective project size proxies. Moreover, simple linear re-
baseline information, impact prediction, etc.) are presented in Fig. 1. gressions were carried out to further explore the potential roles that
Source: Based on the study's data. The contents of all annexes were each proxy plays in the variations of EIS length. Results are presented in
analyzed and distributed according to their respective contents. Post- Fig. 2.
text pages include non-textual information, such as bibliography, ta- The scatterplots a, b and c presented in Fig. 2 reveal that the tested
bles, graphs, which were presented after the main textual sections/ proxies of project size tended to be positively correlated to EIS pages,
volumes. Sector codes: R = Railways, HW=Highways, that is, as project construction area, impacted municipalities and de-
TL = Transmission lines, H=Hydroelectric power plants; M = Mining clared investments increase, the more pages one can expect in their
projects; P=Ports; OG = Oil & Gas. respective EISs. However, the amount of variance in the outcome ex-
As expected, the great majority of EIS content, in the sample, was plained by the various linear models (R2) was found to range from
found to be baseline information (49% of the total pages), followed by 0.023 to 0.272 (Fig. 2 a, b, and c). The highest value corresponds to the
post-text pages (19%), identification and prediction of impacts (11%), regression model for EIS pages and Impacted Municipalities, which
project description (9%), mitigation measures and alternatives (6%), means that 27.2% of the variation of EIS pages could be explained by
pretext pages (5%), blank pages (2%) and communication or results the number of municipalities impacted by the proposed project. The
(0.1%). Interestingly, on average 26% of the total number of pages other two proxies were found explain only 2.3% and 2.0% of the var-
reviewed in the sample accounted for contents that are not directly iation in EIS pages. That is, these proxies, alone, are weak predictors of
relevant to decision-making (pretext, post-text and blank pages). EIS length. These linear regressions, however, were calculated for the
However, in some EISs this proportion was > 40% (e.g. R6, TL4, OG1). overall sample, which included 49 EISs from seven different sectors.
As shown in the stacked column bars of Fig. 1, such contents seem to Sector-specific regressions will result in very different R2 values.
vary across statements and across sectors. For example, the EISs from Therefore, one needs to careful consider these results when analyzing
the Oil & Gas sector contrast with the remainder of the sample, as the sector-specific trends.
number of pages dedicated to ‘identification and prediction of impacts’ EIA practice has long suggested that the length of an EIS is expected
in a number of cases outweigh those dedicated to describing the af- to be influenced by scoping and its respective Term of Reference (ToR)
fected area. This emphasis on impact prediction was found to be a di- (FAA, 2002, Wende, 2002). This document is, by definition, a pre-
rect result of that sector specific Terms of Reference, which asked for scription of content for impact assessment studies. Some ToRs, for ex-
various types of risk analysis, which were categorized here as ‘identi- ample in Brazil (e.g. Paiva et al., 2015), explicitly set page limits to
fication and prediction of impacts’. However, some sectorial variations EISs. However, even when ToRs do not set those limits, they affect the
of proportion of information were difficult to explain. For no clear length of EISs simply by requiring more or less information. The longer
reason, for example, the EISs from the Hydroelectric sector were found the list of issues required by ToRs, the more content one can expect in
to have many blank pages. EIS. To test this assumption, a scatterplot was created to correlate the
number of pages in each EIS with the number of pages in their re-
4.2. Exploring potential determinants of EIS length spective ToRs (Fig. 2 d). Number of pages in ToRs was arbitrarily used
here as a proxy for the volume of information required in ToRs. Find-
Findings corroborate the fact that EISs tend to carry large propor- ings indicate that the amount of variance in the outcome explained by
tions of baseline information, a fact that could be somehow associated the resulting linear models (R2) was also found to be 0.020. That is,

118
G.M.R. Fernández et al. Environmental Impact Assessment Review 73 (2018) 114–121

Fig. 2. Scatterplots of EIS Pages related to: a) Project Area, b) Impacted Municipalities, c) Total Investments, and d) Terms of Reference (ToR) Number of Pages.

while positively correlated, the number of pages in ToRs explained only

2% of the variation of pages in EISs. This regression was also calculated
for the overall sample of 49 EISs from seven different sectors. Sector-
specific regressions will result in different R2 values.

4.3. Sectorial variations in the number of pages in EISs and respective non-
technical summaries (NTS)

The data and descriptive statistics previously presented suggested

that the lengths of EISs could be somehow related to their respective
sectors, as the page numbers seemed to change significantly across
sectors. This apparent sectorial-dependence is particularly visible when
boxplots of the total number of pages in EISs are compared across
sectors. As shown in Fig. 3, in the Highway sector, the mean length of
EISs was of 1147 pages with quartiles clearly smaller than the ones from
the Transmission Lines (mean of 4410) and Oil & Gas (mean of 3036)
sectors. Similar boxplots created to show the distribution of number of
pages in Non-Technical Summaries across sectors (see Fig. 4) also
suggest that there are sectorial differences in those shorter documents.
Fig. 3. Boxplots of the number of pages in EISs across different sectors.
NTSs from the Oil & Gas sector (the mean number of pages was 69) Sector codes: RW = Railways, HW = Highways, TL = Transmission lines,
seemed significantly shorter than the ones from the Mining sector H = Hydroelectric power plants; M = Mining projects; P = Ports; OG = Oil and
(mean of 121). Gas.
To confirm if such visual differences are statistically significant
across the seven categories of sectors addressed in this study, Kruskal-
might be associated with EIS length, the ‘causes’ of such differences are
Wallis tests were carried out to test the null hypothesis that the dis-
still unclear. A variety of variables, not analyzed here, are likely to
tribution of pages in both EISs and Non-technical Summaries are the
influence the generation of EISs and their respective lengths. EIS design
same across categories of sectors. Kruskal-Wallis tests were adopted, as
is just partly regulated; consultancies have lots of room to maneuver its
opposed to ANOVA or other common statistical models used to analyze
contents according to multiple variables. The analysis presented here
differences among groups, because the data were not normally dis-
explored just a few of them.
tributed (based on both Kolmogorov-Smirnov and Shapiro-Wilk tests).
Results of the Kruskal-Wallis tests are presented in Table 2.
The Kruskal-Wallis tests confirmed what the boxplots suggested 4.4. Discussion: EISs are longer than the early ones, and still heavily loaded
visually, i.e. that the distributions of pages in EISs and Non-technical with baseline information; so what?
Summaries are not the same across sectors. However, one needs to
careful interpret these results, because, while some specific sectors The various analyses presented here are important in various ways.

119
G.M.R. Fernández et al. Environmental Impact Assessment Review 73 (2018) 114–121

sample (almost 50%) also suggest that the UK calls for better pro-
portionality of information in EISs (e.g. IEMA, 2011 and The Scottish
Government, 2013) are likely valid in Brazil.
The technology involved in preparing and reviewing EISs has gone
through profound changes in the past decades. Consultants, practi-
tioners and developers have more data available, as well as more ca-
pacity to process and communicate data in a variety of formats (Lyles,
2017). Likewise, environmental agencies today have more information
systems, software and hardware to review EIS content. This technolo-
gical change raises questions about what would be the appropriate
page-limit thresholds under and above which EIS quality could be ne-
gatively affected. The first studies that tried to identify or define such
thresholds (e.g. Lee and Brown, 1992; Sadler, 1996), did so in a sim-
plistic way, with little attention to the many variables that might affect
EIS quality and length. Moreover, those studies did not provide detailed
information about the criteria they used to measure EIS length, which,
as shown here, is not so obvious, particularly today, when EIS are using
more alternative formats and graphical contents. A recent EIA hand-
book from Scotland (HES and SNH, 2018) is arguably aware of the risks
Fig. 4. Boxplots of the number of pages in Non-technical Summaries (NTSs) involved in setting simplistic page limits to EISs. That handbook asks
across different sectors. practitioners to avoid lengthy EISs, as “too much detail can distract and
Sector codes: RW = Railways, HW = Highways, TL = Transmission lines, deter readers and make key issues difficult to appreciate” (p. 81);
H = Hydroelectric power plants; M = Mining projects; P = Ports; OG = Oil and however, it also recognizes that “(…) the size of an EIA Report will
Gas. depend on the range and complexity of issues and no standard size can
be given” (p. 81).
Table 2 Concerns as to the effects of EIS length has not ceased. Recently, an
Results of the Kruskal-Wallis tests. UK EIA specialist, in reaction to the phenomenon of lengthy EISs argued
that there “is a risk local authorities are just reading the non-technical
Null Hypothesis Test Sig. Decision
summaries because of the size of the statements” (Tomos, 2016, p. 4).
The distribution of pages in EISs Independent- 0,004 Reject the null But how likely is this risk? As mentioned in the second section of this
is the same across Samples Kruskal- hypothesis article, very few studies have tried to understand the actual effects of
categories of sectors Wallis Test
EIS length in the review and decision-making process. A recent survey
The distribution of pages in Independent- 0,042 Reject the null
Non-technical Summaries is Samples Kruskal- hypothesis in Brazil found a “a clear predominance of negative appraisal both
the same across categories Wallis Test about the quality of the process and the quality of EIS” (Duarte et al.,
of sectors 2017, p. 1). The findings of this survey coupled with the findings pre-
sented here suggest that longer EISs are not necessarily contributing to
Asymptotic significances are displayed. The significance level is 0,05.
higher quality in EISs. But there is also no evidence that longer EISs are
contributing to lower quality. In this context, one needs to carefully
First, the empirical data corroborate the perception held by many au- consider any proposal of regulation or policy change targeting EIS
thors and institutions that EISs are getting longer. The sample of 49 length.
Brazilian EISs analyzed here ranged from several hundred to a few
thousand-long documents; even their respective non-technical sum- 5. Concluding remarks
maries were found to have on average almost 100 pages: these are
numbers significantly greater than the numbers that preoccupied early This article tried to fill an empirical gap in connection with the
EIA specialists and regulators. However, the average length of the marginal and puzzling issue of EIS length. It did so by systematically
documents reviewed here are not exactly what Ross, Morrison-Saunders analyzing the amount and proportion of information presented in 49
et al. (2006) humorously called “everything under the sun” documents, Brazilian EIS. Sectorial variations and a few likely determinants of in-
that is “Multiple volume EISs that are measured in metres of bookshelf formation volume were explored. Findings corroborate previous claims
space they occupy” (p. 8). The literature often cites cases of lengthy that EISs are now longer than the early ones, and still heavily loaded
EISs. For example, Borioni et al. (2017) recently mentioned a 14,000- with baseline information. While worrisome, the implications of such a
page EIS produced in the approval process of a hydroelectric power trend are still unclear.
plant in Brazil. But such documents are unlikely to reflect the average The exploratory statistics presented here corroborate the fact that
EIS filed in environmental agencies. The findings presented here give a EIS length is context-dependent. The volume of contents is likely af-
better picture of how long federal-level EISs have been in Brazil. fected by a combination of variables related to project size, territorial
Moreover, this study has shown that, without annexes, pretext, post- and sectorial characteristics. Setting page-limits through regulations or
text and blank pages, EISs are about a fourth shorter; which means that Terms of References without an understanding of those variables could
decision-makers should worry more with the proportion of content that be detrimental to the quality of EISs. However, the amount of variation
is likely to be relevant for them, rather than with the total number of explained by the variables addressed here is relatively small; many
pages. other issues are likely to affect document size. Therefore, future studies
In any case, setting the often-cited 150 page limit to current EISs or should analyze a longer list of variables, including, for example, sta-
15-page limits to NTSs would mean a profound change to current EIA keholder pressure, organizational culture, data availability, and time
practice; and arguably a wrong one in the Brazilian context, as recent constraints.
literature has been emphasizing the need for better scoping through Likewise, setting targets of information proportion without a clear
public participation and technical guidance, rather than by fixing tar- understanding of what would be the needed proportion of information
gets of information volume (Barretto and Montaño, 2012; Borioni et al., in a particular decision-making context is risky. Findings presented here
2017). The proportion of pages dedicated to baseline information in the confirmed that about half of the pages in the samples of EISs were used

120
G.M.R. Fernández et al. Environmental Impact Assessment Review 73 (2018) 114–121

to describe the biophysical and socioeconomic characteristics of af- IBAMA, 2016. Avaliação de Impacto Ambiental: Caminhos para o fortalecimento do li-
fected areas, with the exception of the Oil & Gas sector. But is that cenciamento ambiental federal: resumo executivo. Instituto Brasileiro do Meio
Ambiente e dos Recursos Naturais Renováveis (IBAMA), Brasília.
necessarily a problem? What would be the ideal proportion of baseline IEMA, 2011. Special Report - The State of Environmental Impact Assessment Practice in
or any other type of information? Such questions are open for debate. the UK. Lincoln, Institute of Environmental Management & Assessment (IEMA).
Scholars should also note that the actual impacts of EIS length in João, E., 2002. How scale affects environmental impact assessment. Environ. Impact
Assess. Rev. 22, 289–310.
decision-making are still obscure, and thus need to be further in- Julien, H., 2008. Content analysis. In: Given, L.M. (Ed.), The SAGE Encyclopedia of
vestigated. For example, there is an obvious opportunity to advance this Qualitative Research Methods: Volumes 1 & 2. SAGE Publications, Thousand Oaks,
knowledge gap through interviews and surveys with those who actually pp. 120–122.
Kabir, S.M.Z., Momtaz, S., 2014. Sectorial variation in the quality of environmental im-
review and make decisions about EIS's many contents. Their perception pact statements and factors influencing the quality. J. Environ. Plan. Manag. 57,
can clarify, for example, whether a lengthy EIS requires longer review 1595–1611.
times, whether it distract readers or encourages the reading of NTSs as Landim, S.N.T., Sánchez, L.E., 2012. The contents and scope of environmental impact
statements: how do they evolve over time? Impact Assess. Proj. Apprais. 30, 217–228.
opposed to the main document.
Lawrence, D.P., 2003. Environmental Impact Assessment: Practical Solutions to Recurrent
Future studies should keep exploring the increasingly rich data Problems. John Wiley & Sons, Hoboken.
available in the public administrations in charge of the EIA process. Lawrence, D.P., 2013. Environmental Impact Assessment: Practical Solutions to Recurrent
Without a better understanding of how government agencies and ana- Problems and Contemporary Challenges. John Wiley & Sons, Hoboken.
Lee, N., Brown, D., 1992. Quality control in environmental assessment. Proj. Apprais. 7,
lysts process information, the effects of EIS length and proportion of 41–45.
information are likely to remain unclear. More important than under- Lee, N., Colley, R., 1991. Reviewing the quality of environmental statements: review
standing how long an EIS should be is to understand how much in- methods and findings. Town Plan. Rev. 62, 239–248.
Lee, N., Dancey, R., 1993. The quality of environmental impact statements in Ireland and
formation stakeholders need and can process in a particular EIA setting. the United Kingdom: a comparative analysis. Proj. Apprais. 8, 31–36.
Lyles, K.C., 2017. Expediting the NEPA process via a document management system and
Acknowledgments virtual GIS-based NEPA platform. Environ. Pract. 19, 139–147.
Morgan, R.K., 2012. Environmental impact assessment: the state of the art. Impact Assess.
Proj. Apprais. 30, 5–14.
This paper was possible thanks to the financial support of CNPq Morrison-Saunders, A., Annandale, D., Cappelluti, J., 2006. Practitioner perspectives on
[grant number 310758/2015-7]. The authors are sincerely grateful to what influences EIA quality. Impact Assess. Proj. Apprais. 19, 321–325.
Morrison-Saunders, A., Pope, J., Gunn, J.A.E., Bond, A., Retief, F., 2014. Strengthening
IBAMA (the Brazilian Environmental Agency), who quickly responded impact assessment: a call for integration and focus. Impact Assess. Proj. Apprais.
to our information requests, as well as to the three anonymous re- 32, 2–8.
viewers, for their excellent suggestions. NEPA Task Force, 2003. The NEPA Task Force Report to the Council on Environmental
Quality: Modernizing NEPA Implementation. Washington, Executive Office of the
President.
References Paiva, I.C.P.S.B.R., Oliveira, A.K.M., Bononi, V.L.R., 2015. Análise da abordagem
socioeconômica no contexto do licenciamento ambiental de empreendimentos
Anifowose, B., Lawler, D.M., van der Horst, D., Chapman, L., 2016. A systematic quality sucroenergéticos no estado de Mato Grosso do Sul. Soc. Natureza 21, 97–110.
assessment of environmental impact statements in the oil and gas industry. Sci. Total Põder, T., Lukki, T., 2011. A critical review of checklist-based evaluation of environ-
Environ. 572, 570–585. mental impact statements. Impact Assess. Proj. Apprais. 29, 27–36.
Babbie, E., 2010. The Practice of Social Research, 10th ed. Belmont, CA, Wadsworth/ Ross, W.A., Morrison-Saunders, A., Marshall, R., Sánchez, L.E., Weston, J., Au, E., et al.,
Thomson Learning. 2006. Improving quality. Impact Assess. Proj. Apprais. 24, 3–22.
Barretto, FRdM, Montaño, M., 2012. Avaliação da etapa de delimintação do escopo de Sadler, B., 1996. Environmental Assessment in a Changing World: Evaluating Practice to
estudos de impacto ambiental no estado de São Paulo. 1Congresso Brasileiro de Improve Performance. Canadian Environmental Assessment Agency (CEAA) and
Avaliação de Impacto São Paulo. International Association for Impact Assessment (IAIA), Ottawa.
Bond, A., Retief, F., Cave, B., Fundingsland, M., Duinker, P.N., Verheem, R., et al., 2018. A Snell, T., Cowell, R., 2006. Scoping in environmental impact assessment: balancing pre-
contribution to the conceptualisation of quality in impact assessment. Environ. caution and efficiency? Environ. Impact Assess. Rev. 26, 359–376.
Impact Assess. Rev. 68, 49–58. The Scottish Government, 2013. Planning Advice Note 1/2013: Environmental Impact
Borioni, R., Gallardo, A.L.C.F., Sánchez, L.E., 2017. Advancing scoping practice in en- Assessment. The Scottish Government, Edinburgh.
vironmental impact assessment: an examination of the Brazilian federal system. Tomos, K., 2016. The Importance of Proportional Scoping in EIA. IEMA Wales Network
Impact Assess. Proj. Apprais. 35, 200–213. EIA Event: ARCADIS Design & Consultancy for Natural and Built Assets.
Cashmore, M., Christophilopoulos, E., Cobb, D., 2002. An evaluation of the quality of Veronez, F., Montaño, M., 2017. Análise da qualidade dos estudos de impacto ambiental
environmental impact statements in Thessaloniki, Greece. J. Environ. Assess. Policy no estado do Espírito Santo (2007–2013) Desenvolvimento e Meio Ambiente. vol. 43.
Manag. 4, 371–395. pp. 6–21.
CEC. National Environmental Policy Act, 1978. Implementation of Procedural Provisions: Wathern, P., 1990. Environmental Impact Assessment: Theory and Practice. London and
Final Regulations. Council on Environmental Quality (CEC), Washington. New York, Routledge/Taylor & Francis Group.
Department of the Environment Transport and the Regions, 1995. Guide on Wende, W., 2002. Evaluation of the effectiveness and quality of environmental impact
Environmental Statements for Planning Projects that Require Environmental assessment in the Federal Republic of Germany. Impact Assess. Proj. Apprais. 20,
Assessment. Department of the Environment Transport and the Regions, London. 93–99.
Drayson, K., Wood, G., Thompson, S., 2015. Assessing the quality of the ecological World Bank, 1991. Environmental Assessment Sourcebook. Volume I - Policies,
component of english environmental statements. J. Environ. Manag. 160, 241–253. Procedures, and Cross-Sectoral Issues. Washington, The World Bank.
Duarte, C.G., Dibo, A.P.A., Siqueira-Gay, J., Sánchez, L.E., 2017. Practitioners' percep-
tions of the Brazilian environmental impact assessment system: results from a survey. German Marino Rivera Fernández, MSc, is an agroforestry engineer with a Master's
Impact Assess. Proj. Apprais. 1–17. degree in Socioeconomics and Environment. He is currently a PhD candidate in
FAA, 2002. Best practices for Environmental Impact Statement (EIS) Management. Environmental Engineering at the Federal University of Ouro Preto (UFOP), Brazil. His
Federal Aviation Administration, Washington. research addresses the challenge of managing information volume in impact assessment.
Fonseca, A., Rodrigues, S.E., 2017. The attractice concept of simplicity in environmental
impact assessment: perceptions of outcomes in southeastern Brazil. Environ. Impact Ludmila Ladeira Alves de Brito, MSc, is a biologist with a Master's degree in Sanitation,
Assess. Rev. 67, 101–108. Environment and Hydric Resources. She is currently a PhD candidate in Environmental
Fonseca, A., Sánchez, L.E., Ribeiro, J.C.J., 2017. Reforming EIA systems: a critical review Engineering at the Federal University of Ouro Preto (UFOP), Brazil. Her research ad-
of proposals in Brazil. Environ. Impact Assess. Rev. 62, 90–97. dresses the means to increase transparency in EIA decision-making.
Gerring, J., 2007. Case Study Research: Principles and Practices. Cambridge University
Press, Cambridge.
Alberto Fonseca is an Assistant Professor in the Department of Environmental
Glasson, J., Therivel, R., Weston, J., Wilson, E., Frost, R., 1997. EIA-learning from ex-
Engineering at the Federal University of Ouro Preto (UFOP), Brazil. Alberto has a PhD
perience: changes in the quality of environmental impact statements for UK planning
degree in geography and environmental management (University of Waterloo, Canada),
projects. J. Environ. Plan. Manag. 40, 451–464.
an MSc degree in environmental engineering (UFOP, Brazil), and a BSc degree in civil and
Glasson, J., Therivel, R., Chadwick, A., 2005. Introduction to Environmental Impact
sanitation engineering (UFMG, Brazil). Alberto has worked in several public and private
Assessment. Routledge, London and New York.
organizations, including Brazil's Ministry of the Environment and South America's largest
HES and SNH, 2018. Environmental Impact Assessment Handbook: Guidance for
mining company. He is the former president and current scientific director of the
Competent Authorities, Consultation Bodies, and Others Involved in the
Brazilian Association for Impact Assessment.
Environmental Impact Assessment Process in Scotland. Historic Environment
Scotland (HES) and Scottish Natural Heritage (SNH), Edinburgh.

121