Once all hypotheses have been brought forward, the design of the research will be set. The research design is the blueprint for conducting the research and answering the research ques-tions. Both main research A and main research B are sample-based and involve primary data which is collected via surveys. As part of the research design, data collection and sampling represent decisive elements (Blumberg, Cooper, Schindler, 2005, pp. 64-65, 132-137). There-fore, arrangements from both will be discussed in more detail below.
Data collection design
The choice of an appropriate data collection design is decisive for the strength of the research results. Only if the design matches the objectives of the research and clearly addresses the hy-potheses, will the results be meaningful (Blumberg, Cooper, Schindler, 2005, pp. 36-39).
In research theory, a large number of data collection methods are discussed. An overview of these methods is shown in the table below:
Table 1: Data collection methods
Source: Own resource, based on Lang, 2014, pp. 6-11
Both main research A and main research B are being performed by means of quantitative sur-veys which are distributed among internal auditors. This form of data collection method bears the advantage that it is comparably easy to administer, includes a large number of questions, and allows to cover large samples (Lang, 2014, p. 14). The surveys take the form of structured questionnaires with mostly closed-ended questions. Consequently, each respondent is con-fronted with the same survey setup. This standardisation delivers a consistent picture and allows for broad coverage of a specific topic. They also deliver a strong representation of results and thereby provide a sound basis for comparisons. The surveys are therefore suitable for gathering information (in contrast to opinions) in a most neutral manner. Due to the high numerical focus of the surveys, their results can be analysed with statistical methods.
There is a risk that quantitative surveys do not deliver differentiated results as they do not ac-count for individual response options (Oak Rich Institute for Science and Education, 2017). To overcome this shortage, the questionnaire used in main research B allows respondents to phrase individual responses with an additional open-ended question. Also, there is a risk that questions are misunderstood or that the survey structure (e.g. order of questions, type of answer alterna-tives) systematically influences the answering behaviour, if the survey is not conducted person-ally (Oak Rich Institute for Science and Education, 2017). Therefore, the questionnaire of main research A is accompanied by a guideline to provide further details to respondents. For main research B, further details and instructions are orally provided to the respondents.
The data collection method chosen for the preliminary research differs from the approach cho-sen for the main research A and main research B. The focus of the preliminary research is about understanding rather than describing how practitioners deal with CA. Nor does it aim to estab-lish any causal connections. Therefore, a qualitative approach is chosen. This not only does do justice to the high complexity of CA, but also allows a deeper analysis than a quantitative anal-ysis would do. It enables the researcher to manage the train of discussion in a more flexible way while respondents can answer more generously. At the same time, results from this in-depth research complement and validate findings from the literature review.
Sampling design
When doing research, the researcher needs to determine whether he wishes to make use of a sample to conduct his research or whether he wants to conduct a census study. A census is appropriate when the population is small, and elements of the population are different from each other. Both preconditions are not true for either main research A or main research B. In-stead, the use of samples (as a subset of the population) is more advantageous due to the fol-lowing reasons: Firstly, making use of samples comes at a lower cost. Secondly, data collection is faster than under a census. Thirdly, research quality is increased as there is a higher chance of outliers not being considered in the samples (Blumberg, Cooper, Schindler, 2005, p. 202).
Sampling can occur by using one of several techniques. The most prominent ones are shown in the following table:
Table 2: Overview of sampling techniques
Source: Own resource, based on Blumberg, Cooper, Schindler, 2005, pp. 206-209
The preliminary research features a convenience sampling approach during which amount and sort of elements (i.e. internal auditors) are assembled at the researcher’s discretion to become the sample. This approach is chosen as it is considered the easiest and most cost-effective sam-pling technique. On the downside, precision of this techniques is low. However, this disad-vantage is neglected, as the preliminary research “only” is the starting point for further research activities (Blumberg, Cooper, Schindler, 2005, pp. 206-209).
Main research A and main research B feature judgemental sampling techniques. In both cases, the sample is composed of elements selected by chance based on predefined criteria (i.e. inter-nal auditors of German companies).
Alongside the sample technique, the sample size is of central importance. Finding the right sample size (to ensure representativeness) when doing research is not an easy task. Blumberg, Cooper, and Schindler (2005, pp. 212-213) mention specific rules which need to be true in order to achieve a sound representativeness. These rules are listed below:
• A minimum sample size needs to be achieved, whereby absolute size is not the ultimate goal. Instead, focus lies on appropriate fit.
• The more homogeneous the population, the easier it is to achieve representativeness and the fewer elements are needed.
• The more complex and detailed the evaluation of data is planned to be, the more ele-ments are needed.
• The more reliable the conclusions need to be and the lower the error tolerance is, the more elements are needed.
Given these general rules, a total of eight elements are chosen to be sufficient for the prelimi-nary research. For main research A and main research B, the exact sample size will depend on the level of responsiveness of addressees.