Competence barriers to innovation in SMEs

Get Complete Project Material File(s) Now! »

Reliability

In order to implement a research method that yields high reliability, i.e. the extent to which the measured results are consistent and without errors (Saunders et al., 2007; Zikmund, 2000), the authors of this thesis have applied methods previously used by other authors.
This implies a high degree of replication both in terms of data collection and analysis. The use of formal and structured tests of the data, in this case through SPSS, can be argued to increase the reliability (Robson, 2002). Further, the telephone interviews were kept structured to reduce observer error that could arise by all three authors conducting the interviews.
As the questions were closed in the survey, observer bias were overcome. The interviews were also kept anonymous to further ensure reliability, due to the sensitive subject of the thesis (Saunders et al., 2007). Therefore the CEOs did not need to worry when admitting weaknesses within organization. The reliability might be reduced due to misunderstanding or any other circumstances which resulted in the managers giving an answer, which would not exactly represent what was intended by the respondent at the point the answer was given. The questions were kept as short and clear as possible, and as the data were collected through telephone interviews, the respondents had the possibility to ask if they still had problems understanding the questions. This is argued by the authors to be important, as high reliability would result in the same answers if the research was conducted again.
This could be tested through a retest, which in this case was not possible due to the limited time for the research project (Saunders et al., 2007; Zikmund, 2000).
As this study is highly influenced by previous research, the reliability of the previous studies is important. By using secondary data, mainly derived from government statistics, these issues are reduced. Finally, by adapting method of analysis and questionnaire from earlier research, reliability and validity can be further increased (Saunders et al., 2007).
Furthermore, the data collected from the main and complementary study have been estimated for reliability by applying the Cronbach’s alpha test. The reliability is found by testing the scale variance, where an alpha value of at least 0.7 is desirable (Cortina, 1993). This test was applied to all numerical scale variables used in the regression analyses; perceived level of competition, level of training and the three selected competence barriers to innovation.
For the main study the Cronbach’s alpha was 0.787 and for the complementary study 0.715, both above the reliability level (Appendix 9). These results are further strengthened by only including four and five variables respectively, since adding another variable to the test increases the alpha value, even though it may add to the error variance (Cortina, 1993).

Internal validity

The validity of this thesis will firstly be discussed through the concept of internal validity, i.e. the extent to which the findings of this thesis can be attributed through interventions rather than any flaws in the research design (Saunders et al., 2007). If external factors have had an influence, validity has been affected (Zikmund, 2000). A threat to validity within this research project is that the study was conducted at two different occasions. This might have caused a maturation threat, in the sense that the managers for some reason have changed their view of the problem between the studies (Zikmund, 2000). What can be argued to increase validity is that there were no pre-test or contact with the CEOs before the interviews were conducted. A pre-test of this sort could have caused a testing effect, as the managers would then have had the possibility to prepare for it (Robson, 2002). In order to overcome an instrumentation effect that could arise during the interviews, such as the interviewer changing words or other circumstances that cause an effect in the measuring of the dependent variable (Zikmund, 2000), the telephone interviews were kept highly structured, with short and clear questions. Further, during the pilot study, the interview were
tested and slightly modified for the final survey. Sample attrition, which implies that some managers choose to end their participation before the study was fully conducted or split the interview on more occasions (Zikmund, 2000), was not experienced. However, as only approximately half of the respondents replied to the second survey, it still poses a threat to the validity of this thesis. Measurement, or construct validity in this study is high, with the main argument that most questions posed to the CEOs have been adopted from earlier studies (Bryman & Bell, 2003). It must however be considered, that there is always a possibility that the respondents might misunderstand the questions thus affecting the validity (Saunders et al., 2007).

READ Financial management implementation in the public service

External validity

External validity, closely connected and often referred to as generalizability, concerns to what extent the results can be applicable to other situations than the one studied (Robson, 2002). It is argued by the authors that the external validity is high within this sample, but one must be cautious when making generalizations due to the relatively small number of responses in relation to the population size (Saunders et al., 2007). External validity can also be seen as a question of sampling (Zikmund, 2000). Probability sampling was utilized when selecting individual cases for this thesis, as this technique removes bias during the random selection process. This sampling technique ensures that a sample is representative for the population (Bryman & Bell, 2003). However, in order to collect a sample that would have been representative on the 5% significance level, more firms would have needed to be included in the sample (Saunders et al., 2007). Therefore external validity of this thesis decreases and generalizations should be limited to the sample. If generalizations stretch
beyond the sample, these should only be considered as indications.

1 Introduction .
1.1 Background
1.2 Problem
1.3 Research questions .
1.5 Delimitations
2 Frame of Reference.
2.1 Barriers to innovation in SMEs
2.2 Competence barriers to innovation in SMEs
2.3 HRM and training.
2.4 The theoretical framework tested in this study
3 Metho
3.1 Research approach
3.2 Research design.
3.3 Getting access and research ethics
3.4 Selecting sample
3.6 Statistical method .
3.7 Credibility of the research
4 Empirical Findings and Analysis
4.1 Research questions 1 & 2.
4.2 Research questions 3 & 4 .
5 Conclusion .
6 Discussion
References