RESEARCH METHODOLOGY AND APPROACH
4.4 Study Design and Units of Analysis
The study design logically connects the empirical data to the research questions, and ultimately to the stud\¶V conclusions. It guides the researchers to collect, analyse and interpret the data, and consequently draw inferences concerning the causal relationship among variables (Yin, 2009). It is suggested that there are five important steps in a research design: 1) the research questions; 2) its propositions; 3) its units of analysis; 4) the logic linkage between the data and the propositions; 5) the criteria for interpreting the data.
During the literature review, the focus of this study was narrowed down to this research question ³+RZ DQG ZK\ GR organizational changes lead to different consequences for SHUIRUPDQFH"´6ix propositions have been proposed. Some studies with similar topics in other research settings provide support to the research questions and suggest the way to investigate them. The unit of analysis in this study will be discussed next. The fourth and fifth steps of research design as suggested by Yin (2009) as above-mentioned will be discussed in the data analysis section of this chapter.
The unit of analysis can be identified by the research question and propositions of this study. The main unit of analysis should be organizational change phenomena and connected with organizational characteristics and performance. In this study, the case organisation is in the safety and filtration industry in China. Analysing change events and solving the problems in the process of change for the case organisation is also the motivation for this work; the research was funded by the case organisation. Hence the change events in three branches of the case organisation will be the major units for exploring various change phenomena since the time of the organisational founding. The rationale of using a single case will be discussed in the next section. Although change events are at the sub-organizational level, and the organizational characteristics and performance are analysed at the organizational level, it is still true that the definition of the XQLW RI DQDO\VLV LV WKH VDPH DV WKH GHILQLWLRQ RI WKH µFDVH¶ <LQ 7KH embedded units ± change events ± in this study have parallel functions to organizational
characteristics and so implicate an organisational level. For instance, ecological theories of organisational change examine the subunit levels of the organisation, and implicate the organizational level (Hannan et al., 2007). This kind of research design is named by Yin (2009) as a single case embedded design. It aims to understand organisational change by collecting in depth information from few or a single case organisations but with embedded units/cases. Using multiple subunits makes it possible for replication logic in this study. The embedded design is at two levels: 1) change events; 2) organizational characteristics and performance. The environmental variation surrounding the units forms the contextual elements.
4.4.1 The Rationale of Single Case Embedded Design
This study employs single case with embedded analysis cases. Yin (2009) argues that a single case is appropriate if the conditions apply to five rationales. The first rationale is when a single case represents a critical case to test a well formulated theory, can confirm, challenge or extend the theory, and demonstrate whether the proposition is correct or an alternative explanation is more relevant. The second rationale is that the case is unique or extreme case which is worth documenting and analyzing. Thirdly, the case is a representative or typical one. Fourth and fifth rationales are the case can be either revelatory or longitudinal respectively.
In this study, a single case embedded design is reasonable because of several reasons. First, it complies with the first rationale to be as single case design, because that it considers the major arguments in organisational change studies wherever applicable relating with the focused concepts in this study. It examines the rival theoretical discussion and two models from both organisational adaptation and selection theories. This study tests under what condition, the propositions can be better explained and supported by one theory than the others. It does not assume each one alone can explain all the change situations. Thus, for instance, both proposition 5 and proposition 6 examine the effect of change processes but from different angles and each from one school. Although the organisational change study is well formulated before in Western
countries, we see only few empirical studies of issues with which this study is concerned. Hence, this study takes the model by considering both, to put in a unique under studied context to empirically examine whether it can be extended to this situation.
Second, the case organisation is in the safety and filtration13 industry in China. This industry only emerged in the last 25 years or so in China and has not been studied by previous organisational researchers. The case organisation was founded almost at the same time the industry emerged, and it is also in the first generation of private enterprises after the Open Door Policy was introduced in 1978 in China. Since 1978, government macroeconomic intervention and control have been implemented eight times, leading to a large number of bankruptcies each time. The first law to protect and support private enterprises was only set up in 2002 (Center, 2007). The business environment and policies are very unstable and the average mortality rate of SME (small and medium enterprises) is 45%-65% within five years after their founding, 85% within 10 years, and the average age of SMEs is only 3 years (Chen, 2008). The case company has made dramatic internal changes in reaction to the changing external political regulations, economic recession and disease explosion14. This provides the researcher with a number of change events as good empirical resources from both prior history and current conditions, to make a critical test of change theory. Taking account of the above reasons, this single case organisation can be said to be both unique and typical for investigating organisational change and its consequences for performance. It thus accords with the second and third rationales.
Thirdly, as the events in this study cover fifteen years, and the financial data of case organization are traced back to 1994, the event-history and time continuous techniques are adopted in this study15. This study could be therefore regarded as a longitudinal one and comply with the fifth rationale. Another, the author of this study worked in this case
13 The definition of safety and filtration industry is explained in this chapter later. Why to choose safety and filtration
industry is explained in introduction chapter and in the section later of this chapter.
14 Why the disease explosion is relevant will be discussed later in this chapter. 15 The event-history and time continuous techniques are discussed in next section.
organization for eight years; previous working experience is very helpful for data collection and for improving reliability.
Moreover, the embedded analysis can avoid attending to data on an unduly abstract level, focusing instead on each change event covering a long period of time. This makes it possible to supply a clear insight into comparable change processes between change events. The logic of replication can be attained by comparing the results from similar types of change pairs. If the findings from change events turn out as predicted, it can provide strong support to the initial propositions in this study.
Additionally, a single researcher is unlikely to adopt multiple cases with embedded designs because the time and extensive resources required go beyond the means of an independent student (Yin, 2009). The result of a pilot study shows that direct replication of paired case organizations is unlikely to be achieved16 in this work. Due to all of above reasons, overall, the single case embedded design is appropriate and is adopted in this study.
4.4.2 Historical Events and Time Continuous Technique
Organizational change studies require historical information. Historical information about organizational change is strongly suggested by previous researchers (Amburgey et al., 1993). If information about previous organizational changes has not been considered, or if the only focus is on the external environment or contemporary information, the research would supply a misleading and incomplete picture. An event-history approach records the data containing information on the timing, numbers and consequences of significant actions and changes in the organisation from the founding to the end of observation period (Tuma and Hannan, 1979, Kelly and Amburgey, 1991). Event-history methods are appropriate for the study of discrete state, time-continuous change processes, and it is helpful to distinguish the consequences of change content and change process by a fine grain approach (Kelly & Amburgey, 1991;
Tuma & Hannan, 1984). Since exploration of the change process is one of the aims of the study, as we mentioned earlier, and since the case design is a fine grain approach, event history is appropriate to be used here. Furthermore, a multivariate point process PRGHO LV SURSHU IRU µVLWXDWLRQV ZKHUH GLVFUHWH SRLQW HYHQWV RFFXU LQ RQH GLPHQVLRQDO FRQWLQXXP XVXDOO\ WLPH¶ (Amburgey et al., 1993: P62). If multiple, repeatable events need to be examined, like organizational change, a multivariate point process model defines each type of event separately as a µPDUJLQDO SURFHVV¶ FRPSDUHG WR ZKROH processes.
Moreover, it is strongly suggested that since the processes of changes and performance are unlikely in equilibrium, a dynamic approach to examine the impact of the timing and type of changes on performance in an organisation is more proper than a cross sectional approach, which implicitly assumes temporal equilibrium (Tuma and Hannan, 1984). Following this advice, the goal of this study is to model the impact of multiple changes at different points of time, instead of considering a single spells and the author breaks the event history into some major change events which each involve some multi-level changes and represent that historical period17. Each marginal process of
major change case is to be defined18, however the change process during the cascading
change might not be clear cut. Eventually it is possible to combine the whole time series of DQ RUJDQLVDWLRQ¶V FKDQJH history. The company information of change events and financial data of the organisation are complete and accessible, which benefits this study.
Previous theoretical work and methodological discussions claim that tracing changes over time is the major strength of a case study (Yin, 2009). The time series technique is the match between theoretical significant trends and the empirical trends. This means one should examine the financial data before and after the change events compared to the market trends or industry average data, especially the performance might have some decline followed by some rise in the same change, the time series can strengthen the analysis and is of value by generating a rich explanation for the complex outcome of
17 The change events selection will be discussed in a later section of this chapter.
change events (Greve, 1998). This study measures performance of each change event at four points: pre-change performance is assessed twice, and after change condition are approached twice. The proposition for performance measurement of this study are tested continuously (Ocasio, 1994).