Creswell (2003, p. 3) claims that the researcher needs to consider three elements while designing the research framework: philosophical assumptions about creating knowledge; general procedures of research, (called strategies of inquiry or methodologies); and detailed procedures of data collection and analysis, (called methods). In section 4.3 the philosophical assumptions for the current study have been highlighted as was the methodological approach adopted. This section concentrates on research methods, the data collection process as well as the procedures for data analysis (Creswell, 2003).
Previous studies had adopted different approaches to examining assurance practices. Several of these studies have adopted a content analysis approach, using some disclosure instrument to capture data on aspects of assurance statements (see Ball et al., 2000; CPA Australia, 2004; O’Dwyer and Owen, 2005; Deegan et al., 2006). Other studies have involved interviewing assurance providers and stand-alone reporters (Park and Brorson, 2005) while others have used a mix of research methods (Kamp-Roelands, 2002; Wilson, 2003).65
In the light of the research objectives and questions of this study a quantitative approach has been taken through designing a research instrument. The development of this instrument relied upon the previous literature in the field (Ball et al., 2000; Kamp-Roelands, 2002; Wilson, 2003; O’Dwyer and Owen, 2005) as well as a review of assurance statements. Thus, both a deductive and inductive approach to designing the research instrument was used. Data on assurance statements’ characteristics was gathered using content analysis and it is to this approach that attention now turns. Content analysis has been used to gather details of the content of the assurance statements. This element of the study has identified the presence of certain pre- defined characteristics. In addition, content analysis was also employed within the research instrument to investigate the amount of space devoted to each element of the assurance statement. This approach was undertaken to allow the evolution of the shape and contents of the assurance statement to be captured. This section describes
65 Kamp-Roelands (2002) had employed two research methods in her research project: verbal protocol content analysis and questionnaire, whereas Wilson (2003) employed: the participatory qualitative research approach (case study) in addition to content analysis.
the nature of content analysis as well as providing a description of the way in which this method is used in the study.
Content analysis has been used frequently in accounting research (see for example, Zéghal and Ahmed, 1990; Dirsmith and Haskins, 1991; Gray et al., 1995a, 1995b; Deegan and Rankin, 1996; Hackston and Milne, 1996; Kolk, 2003; Fisher et al., 2004; Freedman and Patten, 2004; Unerman and Bennett, 2004; Murray et al., 2006). In the context of investigating the assurance practice, content analysis has been used to analyse the content of the assurance statements by Ball et al., (2000), Kamp-Roelands (2002), Wilson (2003), and O’Dwyer and Owen (2005).
Various definitions have been proposed by different authors for the content analysis. Stone et al., (1966, p. 5) state that content analysis “is any research technique for making inferences by systematically and objectively identifying specified characteristics within text”. Krippendorff (1980, p. 21) defines the content analysis as a “research technique for making replicative and valid inferences from data to their context”. Krippendorff (1980) also emphasised the relationship between the content of texts and their institutional, societal, or cultural contexts. Another definition was offered by Abbott and Monsen (1979 in O’Dwyer, 1999, p. 215) who define content analysis as a “technique for gathering data that consists of codifying qualitative information in anecdotal and literary form into categories in order to drive quantitative scales at varying levels of complexity”. Weber (1990) emphasised that content analysis provides the opportunity for inferences to be drawn about the issuer of the text (sender of the message), the text for itself (the message) and/or the audience (receiver of the message).
Content analysis can be employed as a research method for many purposes. Weber (1990, p. 9)66 outlined many examples of these purposes, such as: comparing the media or ‘level’ of communication, identifying the intentions and any other characteristics of the communicator, reflecting the cultural patterns of group of actors, institutions, or societies, revealing the focus of individual, group, institutional, or societal attention, and describing the trends in communication content. In comparison
66 Weber (1990) relied on the earlier work of Berelson (1952) who outlined nine purposes for content analysis, most of these purposes concerned the communication process (communicator, text, and audience).
with other research techniques, content analysis usually generates ‘unobtrusive’ measures in which, neither the sender nor the receiver of the text is aware is being analysed (Weber, 1990, p. 10). Hence, there is less chance that the measurement actions itself will act as a force for change that confounds the data (Weber, 1990). There are several research techniques can be used to perform the content analysis. These techniques include (Krippendorff, 2004, pp. 44-45):
(i) Pragmatical content analysis: this method focuses in the procedures which classify events and signs according to their probable causes or effects (for example, counting the number of times that something is said which is likely to have the effect of producing favourable attitudes toward sustainability issues in an identified audience).
(ii)Semantical content analysis: this technique focuses in those procedures which classify event and signs according to their meanings (for example, counting the number of times that sustainability is referred to, irrespective words that used to make this reference). Semantical procedures may include: (a) designations analysis (this analysis provides the frequency with which certain objects (persons, things, groups, or concepts) are referred to; (b) attribution analysis (this analysis provides the frequency with which certain objects are characterisations are referred to), and (c) assertions analysis (this analysis provides the frequency with which certain objects are characterised in a particular way, namely, thematic analysis).
(iii) Sign-vehicle analysis: this technique relies on procedures which classify content according to the psychophysical properties of the signs (for example, counting the number of times the word ‘sustainability’ appears).
To adopt any of the above techniques a researcher needs to decide the basic unit of text to be classified (Weber, 1990, p. 21). In this context, there are six frequently used coding units: word; word sense; sentence; theme; paragraph; and the whole text (Weber, 1990, pp. 22-23). 67
67 In order to make a valid inference from a text, the words, phrases, or any other units of the text classified into a category are supposed to have similar meaning (Weber, 1990, p. 12). This similarity may be based on the precise meaning of the word (for example, grouping synonyms together) or may be based on words sharing similar connotations (Weber, 1990, p. 12).
Making valid inferences from the text depends on the reliability of the classification procedures (Krippendorff, 2004). This simply means conducting a consistent procedure on the classification process and that “different people should code the same text in the same way” (Weber, 1990, p. 12). At the same time, classification procedures should also generate valid variables. This form of validity concerns the extent to which the generated variable measures or represents what the researcher intends to measure (Weber, 1990, p. 12).
As with any research method, content analysis has limitations. The key issue is how to deal with the subjectivity inherent in the method. Subjectivity arises because the same text can mean different things to different researchers (Carney, 1972 in O’Dwyer, 1999, p. 220). Thus, the central problem of content analysis is how to ensure that the data reduction process is robust (Weber, 1990, p. 15). The second issue concerns the validity of inferences that were sought from the data. In particular, the variables used in content classification have to have some defined relationship to the research questions. To overcome these problems and to minimise their effects, researchers usually develop and pre – test their research instruments (Weber, 1990). Establishing the reliability of the instruments as well as specified decision categories would enhance confidence in the results (Milne and Adler, 1999, p. 239 and see section 4.6.3 that addresses reliability of the instrument used in this study).