Coding Instructions for Manifestos

ÖSTERREICHISCHE NATIONALE WAHLSTUDIE
AUSTRIAN NATIONAL ELECTION STUDY
Department of Government
AUTNES | Supply Side
Coding Instructions for Manifestos
English Language Version
Wolfgang C. Müller, Martin Dolezal, Laurenz Ennser-Jedenastik, Anna Katharina Winkler
July 2012
Content
1.
Introduction ..................................................................................................................................... 3
2.
Theoretical Framework ................................................................................................................... 3
3.
Training of Coders........................................................................................................................... 8
4.
The Coding Process ......................................................................................................................... 9
Step 1: Document Selection ................................................................................................................ 9
Step 2: Converting Text to Electronic Format .................................................................................. 13
Step 3: Unitizing................................................................................................................................ 13
Step 4: Coding ................................................................................................................................... 14
Coding Actors................................................................................................................................ 14
Coding Issues ................................................................................................................................ 15
Linking Subjects to Objects .......................................................................................................... 17
Additional Variables...................................................................................................................... 22
Appendix 1: Variable List ..................................................................................................................... 25
References ............................................................................................................................................. 28
2
1. Introduction
The present document provides a short overview and introduces the AUTNES manifesto
coding scheme along several steps, summarizing the essence of far more extensive German
language documents. Following two brief sections concerning the theoretical basis of the
coding and the training of coders, this document provides insight into the criteria of document
selection, the preparation of the documents to be coded, the unitizing procedure, and the
actual coding of the individual variables. In line with the press releases and leader statement
coding, the AUTNES coding of manifestos follows a relational approach that was developed
by Kleinnijenhuis and his collaborators (Kleinnijenhuis & Pennings 2001) and has been
adapted to the specific purposes of an election study.
2. Theoretical Framework
In line with Kleinnijenhuis and his collaborators (Kleinnijenhuis & Pennings 2001), the
AUTNES approach rests on the idea of a relational, kernel-sentence based content analysis.
First, the manifestos’ grammatical sentences are broken down into the units of analysis which
are subsequently coded.
In the Kleinnijenhuis and Pennings approach, the units of analysis are ‘nuclear sentences’
which consist of a subject and an object, linked by a predicate that indicates the direction of
the relationship between the two (Kleinnijenhuis & Pennings 2001, 168). In AUTNES, these
units of analysis are generated based on Noam Chomsky’s (1957, 1965) idea of a phrase
structure model: In his theory the smallest possible but nevertheless complete grammatical
sentences – the so-called ‘kernel sentences’ – are quite similar to what Kleinnijenhuis and
Pennings call ‘nuclear sentences’.
Chomsky distinguishes between noun phrases and verbal phrases: A noun phrase consists of a
noun (or pronoun) and of possible additions related to that noun, e.g. articles, adjectives, or
relative clauses. A verbal phrase includes a verb (and possible proverbs) and a second noun
phrase. Following Chomsky’s model, each grammatical sentence consists of at least one noun
phrase and one verbal phrase, which again contains a verb and usually a second noun phrase.
Thus one ‘kernel sentence’ would consist of one noun phrase and one verbal phrase – e.g.
‘Party X demands a tax cut’ (Figure 1).
3
Figure 1
Noam Chomsky’s Phrase Structure Model
Grammatical Sentence
verbal phrase
noun phrase I
Party X
verb
noun phrase II
demands
a tax cut
However, not all grammatical sentences are as short as a kernel sentence. To a large extent,
they include more components and therefore contain more than one kernel sentence.
Chomsky’s theory offers a set of rules that help to deconstruct even longer sentences into
kernel sentences. Relative clauses, for instance, are inseparably connected with the noun
phrase they refer to and therefore do not make up a distinct kernel sentence, whereas a second
object would do so. Thus a grammatical sentence like ‘Party X that won an overall majority
of the vote in the last elections now wants to cut taxes’ still consists of only one kernel
sentence (as above: ‘Party X demands a tax cut’) whereas ‘Party X demands a tax cut and a
rise in pensions’, consisting of three noun phrases (Party X, tax cut, rise in pensions), can be
split into two kernel sentences:
Party X demands a tax cut.
Party X demands a rise in pensions.
These ‘kernel sentences’ – called ‘statements’ in AUTNES terminology – are our unit of
analysis. Following Kriesi et al. (2006, 2008) the AUTNES coding scheme distinguishes
between three different types of statements, depending on their content (Table 1): (1) the
actors’ position towards an issue (actor-issue statements), (2) the relationship between actors
(actor-actor statements), and (3) statements describing facts in a rather unbiased way (reality
statements). The second type of statement can either occur with respect to a specific issue
(actor-actor-issue statement) or without (actor-actor statement). In the course of the unitizing
procedure coders extract these statements from the grammatical sentences in the texts. Then
they record them in a standardized format that clearly indicates the subject, the object and the
relation between the two.
4
Table 1
Standardized Coding of the Different Statement Types
Statement Type
Direction of the
Predicate
Ficticious Example
Standardized Statement
Actor-Issue
positive
We want to reduce VAT.
Party X for reduction of VAT
negative
We reject EU membership.
Party X against EU
membership
neutral
GMOs are an important
issue.
Party X talks about GMOs.
positive (with
caveat)
If necessary we will raise
care allowance.
Party X for* raising care
allowance
negative (with
caveat)
We are against raising taxes
if the budget permits it.
Party X against* raising
taxes.
positive
Chancellor Jones did a good
job.
Party X praises Paul Jones.
negative
Chancellor Jones failed
completely.
Party X criticizes Paul Jones.
neutral
We were in negotiations with
Chancellor Jones.
Party X talks about Paul
Jones.
positive
The government’s tax policy
is brilliant.
Party X praises the
government (Party X / Party
Y) because of its tax policy.
negative
Chancellor Jones‘ European
policy is a complete failure.
Party X criticizes Paul Jones
because of his European
Policy.
neutral
We negotiated a budget deal
with Chancellor Jones.
Party X talks about
Chancellor Jones because of a
budget deal.
(always) neutral
The inflation rate increased
by 3 per cent.
Party X: Inflation rate
increased.
Party X is the best party.
Party X praises Party X.
Actor-Actor (no
issue)
Actor-ActorIssue
Reality
Statement
Special Case:
Self-Praise (no
issue, no second
actor)
5
The ‘actor’ – what Kleinnijenhuis and Pennings and Kriesi et al. call the ‘subject’ – typically
positions her- or himself with respect to an ‘object’, which can either be an issue or a second
actor. However, in both cases subject and object are linked with a numerical value – the socalled ‘predicate’ – that indicates the direction of the relationship. This numerical value can
range from 1 to -1. A positive value (1) represents approval of an issue position or a positive
relationship towards another actor, while -1 stands for the subject's rejection of an issue or
criticism of another actor (Dolezal 2008, 68; Kleinnijenhuis & Pennings 2001, 169).
Sometimes parties do not commit to a definite policy position but add a caveat, e.g. conditions
that need to be fulfilled before a complete approval of an issue in the future. In these cases,
the numerical value is reduced to 0.5 / -0.5.
It is also possible that the actor might express her- or himself in a rather neutral way or simply
describes ‘what is’. Here in both cases the predicate obtains a value of 0. To distinguish
between neutral actor-issue relations (that are included in the later analysis of the parties’
policy position) and the reality statements (that in our analysis only contribute to the issues’
saliency) a second dichotomous variable is coded, specifying the type for all statements with a
predicate value of zero.
The differentiation between the types of statements offers a number of advantages for the data
analysis: The first type of statement indicates a party's issue position, which is estimated with
a numerical value. Thus one can clearly distinguish between the parties’ policy positions.
However, scholars interested in the overall issue saliency still can revert to all types of
statements and thereby get an exact measure of an issue’s relative importance.
As the actor-actor statements are recorded whenever an actor refers to another, these provide
additional value when it comes to an analysis of party competition. On the basis of these
statements, researchers may draw conclusions from the relative frequencies of a party
mentioning its competitors or its own candidates and from the positive or negative
relationships between pairs of actors. Moreover, the extra coded issue provides information
about the topics on which one party attacks another.
With the issue recorded, the AUTNES manifesto coding provides the basis to estimate the
issue saliencies. Here the AUTNES manifesto coding comes with an additional benefit: By
applying a relational method of content analysis, parties and issues are linked by a numerical
value from which policy positions can later be calculated.
In the context of a national election study, the data derived from manifestos should be
comparable over time and compatible with other data. Thus it is necessary to generate a
6
flexible coding scheme that is open to new issues that may arise in the future and that can be
linked with other data, e.g. voter surveys or media coding data. With over 650 individual
issues, the AUTNES system of categories allows for easy adaption to individual research
purposes and linkage with other datasets.
7
3. Training of Coders
Automated methods for coding political texts have rapidly developed in the past decades.
When it comes to generating issue saliency, these provide excellent results. However, for the
purposes of an election study such as AUTNES, a far deeper analysis is required. This is
especially relevant when it comes to representing the relations between the different political
actors in the dataset. Thus despite the current notable achievements and benefits of
computerized text coding, it is still inevitable to bear the higher costs of manual coding. This
is because human coders are not only able to comprehend the content of the manifestos’
sentences, but also grasp the relations between the different actors and issues appearing in the
text.
On the other hand, human coders are more vulnerable to produce unreliable results (e.g.
Mikhaylov et al. 2012). Here computerized methods clearly outperform manual text coding.
However, a strict set of rules and an intense process of training help to reduce error and thus
enhance reliability. Hence all AUTNES coders receive a 100-odd page manual with detailed
coding instructions and plenty of examples. All coders read these instructions carefully before
running through an intense process of training in which, under the supervision of AUTNES
researchers, they learn to employ these rules on actual manifestos. After several weeks of
training and extensive tests concerning inter-coder reliability, the coders have internalized the
rules and are ready to start the actual coding.
In order to ensure data quality, the coding procedure is organized in several steps, each of
which is carried out by two independent coders. After both coders have finished unitizing,
inter-coder reliability is measured. In cases of disagreement, all AUTNES researchers decide
by judgement call based on stated coding instructions. This ‘approved’ version then provides
the basis for the next step of coding, which is then carried out by two other coders. Thus one
manifesto is eventually coded by several coders. Note that the pairings of coders are not fixed,
but rather rotate to minimize systematic error.
8
4. The Coding Process
A valid and reliable set of data is supported by a clearly-arranged coding procedure. Thus our
process of coding is organized in several steps: In the first step, the actual manifestos are
identified. As parties publish a variety of different programmatic documents, this step
sometimes requires thorough archival research, especially for earlier years. In the second step,
documents that are not already available in electronic form are converted into electronic files.
In order to eliminate mistakes in the data files resulting from incorrect transformation, these
electronic files are cross-checked sentence by sentence with the original documents.
These files then provide the basis for the unitizing procedure. This third step is carried out
independently from the eventual coding of variables. Each manifesto is unitized by two
trained human coders working autonomously. After both coders have finished the unitizing
process, inter-coder reliability is measured. The two datasets are then merged into one version
that provides the basis for further coding. In case of inter-coder disagreement, all AUTNES
researchers decide by judgement call based on the stated grammatical rules.
The fourth step of coding is again carried out by two independent human coders. Once they
have finished, inter-coder reliability is again checked and both versions are merged into one
final dataset. As with the unitizing, if the two coders’ decisions deviate from each other, all
AUTNES researchers decide by judgement call based on coding instructions.
The following sections outline these individual steps of coding in further detail. Examples of
coding decisions and inter-coder reliability test results are also provided.
Step 1: Document Selection
In our dataset we include all manifestos of Austrian political parties that are represented in the
Austrian Nationalrat at the time of the election or those that win enough votes to enter
parliament. As we assume that parties who miss the threshold of four per cent of the vote may
still influence party competition, we include the manifestos of all parties that won more than
two per cent of the vote.
9
On the basis of thorough archival research (‘historical’ cases) and in some cases interviews
with politicians, we identified the actual election manifestos (Table 2; see also Dolezal et al.
2012). The selection of the appropriate documents is not trivial, as parties do not always
designate a particular document as their manifesto. In some cases they may not have produced
one, while in others they might have several documents that fulfil some of the criteria.
Therefore we faced some challenges especially with respect to newly founded parties:
Sometimes these parties issued documents that deviate from what more established parties
publish as their ‘manifesto’ in form and specificity. However, it was finally possible to
identify a document fulfilling the criteria of a manifesto for all parties. The only exception
with this regard is the Green party: In 1990 the young party issued their first basic
programme, but beyond that did not publish a manifesto. Thus instead of a genuine
‘manifesto’ for the 1990 election, we resorted to a document that shares characteristics of a
basic programme, especially with respect to the party explaining the Greens’ idea of an ideal
society. As some party officials identified this document as the party's election manifesto, we
also treated that document as such despite its unusual characteristics.
10
Table 2
Coded Manifestos, 1986-2008 (respectively their functional equivalents)
SPÖ
FPÖ
Grüne
Unser Österreich.
Jörg Haider. Ein
Politiker der neuen
Art
Offenes
Kurzprogramm.
Es wird Zeit,
etwas zu tun!
1990
Das Österreich von
morgen. Wie die SPÖ Den Aufschwung
in den nächsten vier
wählen! Mit uns ist
Jahren Österreich
er sicher
erneuern will
Blaue
Markierungen.
Schwerpunkte
freiheitlicher
Erneuerungspolitik für
Österreich 1990
Leitlinien Grüner
Politik zu den
Themen:
Umwelt,
Demokratie,
Soziale
Gerechtigkeit
(= basic
program)
1994
Es geht um viel. Es
geht um Österreich.
Die Erhard-BusekSozialdemokratische Pläne für
Arbeit für die Zukunft Österreich
Österreichs
ÖsterreichErklärung zur
Nationalratswahl
1994. Dr. Jörg
Haider
Sie bestimmen.
Die Qualität der
Politik.
1986
1995
1999
Das VranitzkyProgramm. Vor uns
liegt das neue
Österreich
Für Österreich! Das
Kanzlerprogramm
Das Wahlprogramm
der SPÖ für die
Nationalratswahlen.
Der richtige Weg für
Österreich
ÖVP
Österreich zuerst.
Das MockProgramm für eine
Wende zum
Besseren
Der Schüssel-Ditz- Vertrag mit
Kurs
Österreich
Der bessere Weg
Der HaiderPrinzhorn Plan.
Zukunft für unser
Land
Jetzt Farbe
bekennen. Die
Grünen
LIF
VGÖ
Vereinte GRÜNE
Das UmweltBürgerforum
ÜBERLEBEN
Die Politik des
Liberalen
Forums
(= Folder)
Die Offensive
Mitte
Persönlichkeiten
und Programm
für Österreich
Kompetent.
Engagiert. Grüne Liberale
Positionen für
Antworten
eine neue Politik
11
BZÖ
MARTIN
Table 2
Coded Manifestos, 1986-2008 (respectively their functional equivalents) (continued)
SPÖ
ÖVP
FPÖ
Grüne
2002
Faire Chancen für
alle! 26 Projekte für
die Zukunft
Österreichs
Das ÖsterreichProgramm der
Volkspartei.
Unsere Ziele für
Österreich
Programm. Wir
gestalten
Österreich mit
Sicherheit
Österreich
braucht jetzt die
Grünen
2006
Den Wohlstand
gerecht verteilen. 20
Projekte für mehr
Fairness in
Österreich
Wahlprogramm
der Freiheitlichen
Kursbuch Zukunft.
Zeit für Grün.
Partei Österreichs
Modern. Sicher.
Das grüne
FPÖ.
Menschlich
Programm
Nationalratswahl
2006
2008
Wahlmanifest der
Sozialdemokratischen
Neustart für
Partei Österreichs.
Österreich
Nationalratswahl
2008
Österreich im
Wort. Auswahl und
Zusammenfassung
inhaltlicher Ziele
der Freiheitlichen
Partei Österreichs
für die neue
Legislaturperiode
LIF
Für ein
modernes
Österreich. Das
liberale
Zukunftsprogramm 2002
Neu beginnen!
Das grüne
Fairness
Programm für
einen Neubeginn
BZÖ
Wahlprogramm
VGÖ
MARTIN
Demokratie
Kontrolle
Gerechtigkeit.
Grundsatzprogramm
und
Kandidaten
Deinetwegen.
Österreich. Das
Wahlprogramm
des BZÖ
Note: When running in 2008, the newly founded Liste Fritz did not have a genuine manifesto. As to date (June 2012) despite intense research we could not elicit a functional
equivalent, this party is not included in the analysis.
12
Step 2: Converting Text to Electronic Format
The documents identified as the parties’ manifestos are available in different formats: While
contemporary documents are usually available in electronic format (pdf), manifestos from
earlier election years are most likely to be found in printed format. As the steps of unitizing
and coding are carried out in MS-Excel, it is necessary to transform these documents first into
MS Word and then into MS Excel format.
Each grammatical sentence – with headlines or enumerations treated as distinct sentences –
occupies one row in the Excel sheet. These sheets then provide the basis for the unitizing
procedure, in which coders add rows in the file according to the number of statements they
derive from one grammatical sentence.
Step 3: Unitizing
In the third step, the AUTNES coders deconstruct the grammatical sentences in the
manifestos into standardized statements, that later constitute the units of analysis. The strict
rules derived from Chomsky’s syntax grammar – which we adapted to the specific structure
of the German language – guarantee that the process of unitizing is independent from the text
content. Furthermore, the fixed rules of unitizing should lead to high inter-coder reliability.
To test this, each manifesto is unitized twice. Table 3 shows the reliability scores of the
unitizing procedure for the Austrian manifestos from 2002 to 2008:
Table 3
Inter-coder reliability of the unitizing procedure
Party
Social Democratic Party (SPÖ)
People’s Party (ÖVP)
Freedom Party (FPÖ)
Greens
Alliance Future of Austria
(BZÖ)
Liberal Forum (LIF)
2002
0.84
0.84
0.84
0.89
2006*
0.89
0.92
0.91
0.85
2008
0.91
0.88
0.90
0.90
-
0.93
0.86
0.77
-
0.91
Note: Figures are Krippendorff’s Alphas for the number of statements extracted from each sentence by two
independent coders; * data for 2006 based on random samples of 10 % taken from the manifesto text rather than
the entire manifestos.
13
The results display excellent overall agreement, with only one case closely missing the
conventional benchmark of 0.8. In order to proceed with the issue coding, inter-coder
disagreements need to be resolved authoritatively by the AUTNES researchers to produce a
single set of statements that represent the starting point for further coding.
Step 4: Coding
In the last step, these standardized statements are coded. Here the framework follows the
statement structure: We record a subject actor as well as the object which in case of actorissue statements consists of an issue and in case of actor-actor statements of a second actor.
The subject is linked to the object through a numerical value that indicates a positive,
negative, or neutral relationship. In the case of actor-actor-issue statements, an issue is coded
in addition to the object actor.
This basic structure is complemented by three additional variables that provide the basis for a
wide range of further analyses: The record variable distinguishes between retrospective and
prospective statements; the characteristics variable captures references to the actors’
characteristics such as competence or leadership qualities, while the justification variable
records the actors’ justification of their issue positions.
Coding Actors
The subject and object actors are recorded with two variables: organization and name. With
regard to manifestos, in most cases the organization variable records the parties’ names, but
in general this variable is open to all kinds of organizations such as political institutions,
governments, NGOs, or research institutions. For references to a coalition government, we do
not multiply the statements, but rather include all parties that are part of the cabinet when
recording the actor’s organization (‘Government (Party X/Party Y)’).
For individuals, typically the party’s top candidate, a second variable additionally records the
name of the actor.
14
Coding Issues
When it comes to coding, establishing a list of issues is certainly the most challenging part: It
is required to be comprehensive and needs to capture both the parties’ very specific as well as
general statements. Moreover, in a longitudinal perspective, the coding scheme has to be open
to new issues coming up in the future, such as future EU treaties. We therefore created a list
consisting of over 650 different issues that includes broad labels such as ‘economic growth’
as well as very specific ones, e.g. ‘VAT’ or ‘mandatory labelling of GMOs’. Thus it is
possible to capture both the parties’ positions on the same level of specificity at which they
are expressed by the parties themselves, and to keep track of the precise contents of the issue
position what is operationally important for the building of policy scales later on.
With regard to the powers of the European Union, we include several issues concerning EU
treaties and policies such as ‘Common Foreign and Security Policy’ in our list. For issues that
are not specific to the EU, we created a dummy variable by which coders can specify the
required level of regulation (European vs. national) that a party prefers.
In a relational method of content analysis, the parties’ policy positions are recorded by the
predicate and the issue variable. Thus most issues are non-directional – e.g. ‘immigration’
instead of ‘pro-immigration’ and ‘anti-immigration’.
Our list of issues is structured into three hierarchically nested levels, which we name issue
categories (level 1), issue sub-categories (level 2), and issues (level 3). Coders are trained to
allocate each statement into one of the more than 650 issues on level 3 (e.g. ‘early
retirement’). As human coders cannot be expected to memorize all 650 categories, these fall
into one of the 99 issue sub-categories (e.g. ‘pensions’) which again belong to one of the 15
level-1 categories (e.g. ‘welfare’) (Table 4). This hierarchically structured system of issue
categories enables coders to first locate the statement in one of the 15 issue categories, and
then search for the most appropriate sub-category.
15
Table 4
Hierarchical Issue Coding Scheme, Example
Level 1: Issue
Level 2: Issue Sub-
Categories
Categories
Level 3: Issues
12000
Budget
12200 Taxes (general)
12201
Tax policy (general)
12000
Budget
12200 Taxes (general)
12202
Tax reform (general)
12000
Budget
12200 Taxes (general)
12203
Simplification of the tax code
12000
Budget
12200 Taxes (general)
12204
12000
Budget
12200 Taxes (general)
12205
12000
Budget
12200 Taxes (general)
10508
12000
Budget
12300 Taxes (Individuals)
12301
VAT (general)
12000
Budget
12300 Taxes (Individuals)
12302
VAT (specific products)
12000
Budget
12300 Taxes (Individuals)
12303
Income tax
Reduction of tax and contribution
ratio (general)
Tax increase (general)
Taxes on speculative products /
stocks
Coders are allowed to come up with new issues by simply adding them to the list when
coding the manifestos.1 When creating a new issue, coders should also suggest the appropriate
issue category and sub-category. When merging the two coding files into the final dataset, the
AUTNES researchers check each new issue and decide on a definite inclusion of the new
issue in the existing list. If the new issue is accepted, it is adopted for all cases in which
coders opted for that issue; beyond that, all cases with inter-coder disagreement are checked
on whether the new issue would fit in there as well. As the new issue is to be included in the
general list of issues, it is available for future coding. Thus our system of issue categories is
flexible enough for longitudinal analyses of party competition.
For the manifesto coding we rely on a number of human coders who, after an intense training
process, are able to accomplish this task independently (see above). To provide not only valid
but also reliable results, our coding is conducted by two independent coders. After both
coders have finished their work, inter-coder reliability is checked. So far, this reliability test
has provided excellent results (Table 5).
1
However, this is only a last resort in case a statement could not be validly coded into one of the existing
categories. Empirically this is done in less than 1 per cent of the cases.
16
Table 5
Inter-Coder Reliability of the Issue Coding (before final check to resolve coder
disagreements)
Party manifesto
Alliance Future of Austria
(BZÖ)
Freedom Party (FPÖ)
Liberal Forum (LIF)
Greens
People’s Party (ÖVP)
Social Democratic Party (SPÖ)
Level 3
Level 2
Level 1
.74
.80
.86
.65
.68
.69
.66
.65
.80
.82
.77
.80
.81
.85
.86
.85
.89
.86
Note: Agreement measures are Krippendorff’s Alphas for two independent coders.
The levels refer to our system of categories. The actual coding is done at level 3 (see the main text).
After reliability is checked, both coders’ works are merged into one final dataset. This final
step is carried out by the AUTNES researchers who, in the deviant cases, collectively decide
on which coding fits the party’s intention best.
Linking Subjects to Objects
The subject actor is linked to the object (issue or second actor) with a numerical value that
indicates a positive, negative, or neutral relationship. This relationship is captured by the
predicate variable that consequently can adopt the values of 1, 0, and -1. For rare cases in
which an actor adds a caveat to her/his issue position, these predicate values may be reduced
to 0.5 / -0.5.
Accordingly, the predicate can adopt the following values:
1 = approval of issue or praise of object actor
0.5 = approval of issue with caveat / conditions for future approval
0 = neutral position, neutral relation to object actor, or reality statement (see below)
-0.5 = rejection of issue with caveat / conditions for future rejection
-1 = rejection of issue or criticism of object actor
The numerical value of zero is recorded in three different cases: A neutral issue position, a
neutral relationship between the subject and the object actor, and for cases in which the
17
subject describes ‘what is’ (reality statement). While an actor-actor statement can be clearly
identified, it is not possible to distinguish between neutral actor-issue relations and reality
statements, as both types include one subject actor and one issue that are both linked with a
numerical value of zero. Therefore, in this case, a second dichotomous variable is coded that
distinguishes between neutral issue positions and reality statements (Table 6).
Note that for actor-actor-issue statements the predicate captures the relationship between the
two actors, but does not indicate the subject’s issue position. Consequently, the actor-actorissue statements are included in the issue saliency measure, but excluded from the analysis of
the issue position in the subsequent analysis.
18
Table 6
Type of Statement
Actor-Issue
Coding of Statements
Direction
of the
Predicate
Ficticious Example
Standardized
Statement
Coding:
Subject
Actor
Coding:
Predicate
Coding:
Predicate
(Neutral /
Reality)
Coding:
Coding:
Object Actor
Object
(Organization) Actor
(Name)
Coding: Issue
Party X for
reduction of
VAT
Party X
+1
--
--
--
VAT
positive
We want to reduce
VAT.
negative
We reject EU
membership.
Party X against
EU membership
Party X
-1
--
--
--
EU membership
neutral
GMOs are an
important issue.
Party X talks
about GMOs.
Party X
0
Neutral
Position
--
--
GMOs
positive
(with
caveat)
If necessary we will
raise care allowance.
Party X for*
raising care
allowance
Party X
0.5
--
--
--
Care allowance
negative
(with
caveat)
We are against tax
increase if the budget
permits it.
Party X against*
tax increase.
Party X
-0.5
--
--
--
Tax increase
19
Table 6
Type of Statement
Actor-Actor (no
issue)
Actor-Actor-Issue
Coding of Statements (continued)
Direction
of the
Predicate
Ficticious Example
Standardized
Statement
Coding:
Subject
Actor
Coding:
Predicate
Coding:
Predicate
(Neutral /
Reality)
Coding:
Coding:
Object Actor
Object
(Organization) Actor
(Name)
Coding: Issue
positive
Chancellor Jones did
a good job.
Party X praises
Paul Jones.
Party X
1
--
Party Y
Jones, Paul
--
Chancellor Jones
failed completely.
Party X
criticizes Paul
Jones.
Party X
-1
--
Party Y
Jones, Paul
--
negative
We were in
negotiations with
Chancellor Jones.
Party X talks
about Paul
Jones.
Party X
0
--
Party Y
Jones, Paul
--
neutral
Party X
1
--
Government
(Party X / Party
Y)
--
Tax policy
The government’s tax
policy is brilliant.
Party X praises
the government
(Party X / Party
Y) because of
its tax policy.
Party X
-1
--
Party Y
Jones, Paul
Chancellor Jones‘
European policy is a
complete failure.
Party X
criticizes Paul
Jones because
of his European
policy.
European policy
(general)
positive
negative
20
Table 6
Type of Statement
Coding of Statements (continued)
Direction
of the
Predicate
Ficticious Example
Standardized
Statement
Coding:
Predicate
Coding:
Predicate
(Neutral /
Reality)
Coding:
Coding:
Object Actor
Object
(Organization) Actor
(Name)
Coding: Issue
Party X talks
Party X
about
Chancellor
Jones because
of a budget deal.
0
--
Party Y
Jones, Paul
Budget deal
Party X
0
Reality
--
--
Inflation rate
Party X
1
--
Party X
--
--
Actor-Actor-Issue
(continued)
neutral
We negotiated a
budget deal with
Chancellor Jones.
Reality Statement
(always)
neutral
The inflation rate
increased by 3 per
cent.
Party X:
Inflation rate
increased.
Party X is the best
party.
Party X praises
Party X.
Special Case: SelfPraise (no issue, no
second actor)
Coding:
Subject
Actor
21
Additional Variables
Apart from coding actors and issues, the AUTNES manifesto data includes three additional
variables that provide the basis for deeper analysis of party competition: the parties’
references to records, the justifications for their policy positions, and the actors’
characteristics.
Record
The record variable distinguishes between retrospective and prospective statements. It records
references to the actors’ past behaviour and performance in office. Moreover, the coded value
indicates the level on which the parties’ record has taken place: The national level, the
international level (e.g. the European Union), or on the Land level (e.g. a record established as
part of the Land government). A fourth category indicates the ‘historic’ record, e.g. a parties’
reference to its achievements decades ago.
0 = no record
1 = record national level
2 = record Land level
3 = record international level
4 = historic record
As it is coded for each observation – actor-issue statements as well as actor-actor statements
– the record variable makes it possible to draw conclusions on the parties’ strategic
employment of their past performance in political competition.
22
Characteristics
Voters base their decisions not only on pure policy positions but also draw on their
perceptions of the competitors’ public image (e.g. Funk 1996; Lewis-Beck et al. 2008; Miller
& Shanks 1996). This is why our coding scheme includes the so-called characteristics
variable: It is coded for each actor – both subject and object actors – and indicates the specific
characteristics ascribed to the respective individual or organisation, distinguishing between
‘competence’, ‘leadership qualities’, ‘character’, and ‘appearance’. To account for negative
and positive attributions, there is a positive as well as a negative category for each
characteristic. Moreover, a residual category is included to capture references that do not fit in
any of the given categories.
Justifications
Parties not only differ with regard to their policy positions and the emphasis they put on an
issue, but also with respect to how they argue the issue positions they adopt. Thus it is
possible that parties adopt the same policy position, but use different arguments to justify it.
Two parties could both be in favour of building a new railroad track. However, one of them
may cite economic arguments such as faster delivery of goods and an anticipated boost for the
local economy (category: economy), while the other party may stress the benefits for the
environment that come with the expected reduction in car emissions (category: environment).
Hence for each issue the justification variable records the arguments parties use to motivate
their issue positions. Here we distinguish between 10 different categories (Table 7).
For each statement, all arguments used to justify the party positions are coded. These are
identified on the level of the grammatical sentence, including both the sentence from which
the coded statement originates as well as the precedent and the subsequent grammatical
sentence. Each argument is treated as one individual justification, regardless of the category it
belongs to. For this reason a position justified by three economic arguments results in coding
three justifications. Note that there is no limitation in the number of justifications coded.
23
Table 7
Categories of the Justification Variable
Category
Examples for justifications belonging in category
Economy
economic growth, unemployment (reduction), protection of
corporate profits, maintaining income levels
Welfare (Expansion)
expansion of the welfare state, redistribution, needs of individual
social groups, public welfare, public health, public responsibility,
social justice
Welfare (Retrenchment)
hedge of the welfare state, increasing the efficiency of the
welfare state, possible savings, avoiding malpractice, individual
responsibility, individualization
Protection of the Environment
protection of the environment (general), climate protection,
animal welfare
Security
internal security, crime, terrorism, external security, war and
peace
Education
education, science, art, (high) culture
Governance
government / bureaucratic efficiency, all references to the
budget, financial restriction (excluding restrictions concerning
the welfare state  category ‘welfare (expansion)’), government
/ bureaucratic authority
Ethnic-Nationalist
Austrian traditions, customs, traditional way of life, Austrian
sovereignty / independence
Religious
Christian-religious aspects, secularism
Universalistic
justifications on a normative-abstract level, general principles
(e.g. fairness, democracy), universal values (e.g. human rights)
Residual Category
24
Appendix 1: Variable List
v01 sentence_id
– Description: All grammatical sentences in the manifesto get an individual ID number.
v02 sentence
– Description: This variable entails the complete wording of each grammatical sentence.
v03 statement
–
Description: Statement derived from the grammatical sentence.
v04 page
– Description: Page number of the manifesto where the grammatical sentence can be found.
v05 subjectactor_organization
– Description: Organization of the subject actor (usually the party).
v06 subjectactor_name
– Description: Name of the subject actor if she / he is an individual; notation: last name, first
name.
v07 subjectactor _characteristics
– Description: Characteristics of the subject actor (i.e. her / his self-perception).
– Categorization: Four dimensions, each with positive and negative values (1) competence (+/-),
(2) leadership qualities (+/-), (3) character (+/-), (4) appearance (+/-), residual category.
v08 predicate
– Description: This variable measures the relation between subject and object (i.e. either between
the subject and an issue or an object actor).
– Categorization: (-1) rejection, (-0.5) rejection with caveat, (0) neutral, (0.5) approval with
caveat, (+1) approval.
v09 predicate_specialcase
– Description: If the predicate is 0 the coder has to decide whether it is a neutral position or a
statement about reality.
– Categorization: (0) neutral position, (1) reality statement.
25
v10 issue
– Description: Here we code the specific policy / issue mentioned.
– Categorization: 15 level I categories (e.g. ‘budget’), 90 level II categories (e.g. ‘taxes for
individuals’), 650+ specific level III issues (e.g. ‘inheritance tax’).
v11 eu_reference
–
–
Description: Level on which the issue coded in v10 should be regulated.
Categorization: (0) national level, (1) European level.
v12 justification
– Description: All justifications mentioned by the subject actor in actor-issue statements.
– Categorization: (1) economy, (2) welfare (expansion), (3) welfare (retrenchment),
(4) protection of the environment, (5) security, (6) education, (7) governance, (8) ethnicnationalist, (9) religious, (10) universalistic, (11) residual.
v13 record
– Description: Retrospective statements about a party’s own or a political opponent’s behaviour /
achievement.
– Categorization: (0) no record, (1) national level, (2) Land level, (3) international level,
(4) historical.
v14 objectactor_organization
– Description: Organization of the object actor (usually the party).
v15 objectactor_name
–
Description: Name of the object actor if she / he is an individual; notation: last name, first
name.
v16 objectactor_characteristics
– Description: Characteristics of the object actor.
– Categorization: Four dimensions, each with positive and negative values (1) competence (+/-),
(2) leadership qualities (+/-), (3) character (+/-), (4) appearance (+/-), residual category.
26
v17 manifesto_year
– Description: Year of manifesto.
v18 manifesto_party
– Description: Party authoring the manifesto.
v19 manifesto_title
– Description: Title of the manifesto.
_________
Note: All values of variables 16, 17, and 18 are identical for all observations of one manifesto.
27
References
Chomsky, Noam (1957). Syntactic Structures. The Hague: Mouton & Co.
Chomsky, Noam (1965). Aspects of the Theory of Syntax. Cambridge, Mass.: The Massachusetts
Institute of Technology Press.
Dolezal, Martin (2008). ‘The design of the study: the distinguishing characteristics of our approach’,
in Hanspeter Kriesi, Edgar Grande, Romain Lachat, Martin Dolezal, Simon Bornschier &
Timotheos Frey, West European Politics in the Age of Globalization, Cambridge: Cambridge
University Press, 53-74.
Dolezal, Martin, Laurenz Ennser, Wolfgang C. Müller & Anna Katharina Winkler (2012). ‘The Life
Cycle of Party Manifestos: Exploring the Austrian Case.’ West European Politics, 35(4): 869895.
Funk, Carolyn L. (1996). ‘Understanding Trait Inferences in Candidate Images’, in Michael X. Delli
Carpini, Leonie Huddy & Robert Y. Shapiro (eds.), Rethinking Rationality. Greenwich, Conn.:
JAI Press, 97-123.
Kleinnijenhuis, Jan & Paul Pennings (2001). ‘Measurement of party positions on the basis of party
programmes, media coverage and voter perceptions’, in Michael Laver (ed.), Estimating the
Policy Positions of Political Actors. London: Routledge, 162-182.
Kriesi, Hanspeter, Edgar Grande, Romain Lachat, Martin Dolezal, Simon Bornschier & Timotheos
Frey (2006). ‘Globalization and the transformation of the national political space: Six
European countries compared.’ European Journal of Political Research. 45(6): 921-956.
Kriesi, Hanspeter, Edgar Grande, Romain Lachat, Martin Dolezal, Simon Bornschier & Timotheos
Frey (2008). West European Politics in the Age of Globalization. Cambridge: Cambridge
University Press.
Lewis-Beck, Michael S., William G. Jacoby, Helmut Norpoth & Herbert F. Weisberg (2008). The
American Voter Revisited. Ann Arbor: The University of Michigan Press.
Mikhaylov, Slava, Michael Laver & Kenneth Benoit (2012). ‘Coder Reliability and Misclassification
in the Human Coding of Party Manifestos.’ Political Analysis, 20(1): 78-91.
Miller, Warren E. & J. Merrill Shanks (1996). The New American Voter. Cambridge, Mass.: Harvard
University Press.
28