Before diving into the data visualization and construction of the study’s variables of interest, it is crucial to test the reliability and validity of the questionnaire (Erkan, 2016). There is a total of 4 control variables and two independent variables that we want to investigate to identify the impact they have on Greek and Finnish participants' online consumer buying behaviour. Thus, it is necessary to measure each variable's internal reliability with its different number of items (Kwakye, 2016).
Cronbach’s alpha test was implemented to test the items' internal reliability in the variables of interest. Cronbach’s alpha is the most commonly used test for reliability when one wants to assess the internal consistency and is expressed as a number between 0 and 1. Cronbach’s alpha test shows the consistency between measurements which form the scales (Erkan, 2016).
Internal consistency describes the extent to which all the items in a test measure the same perception, and hence it is linked to the inter-relatedness of the items within the test (Tavakol et al., 2011). According to Tavakol et al. (2011), Cronbach alpha's value increases if the test items are correlated to each other. They point out that when experiencing a high Cronbach alpha value, that does not always mean a high degree of internal consistency. What is important is the length of the test conducted. If the test length is too short, the alpha value is reduced (Nunnally et al., 1994).
When it comes to the acceptable range of the Cronbach’s alpha values, Hair et al., (2006) distinguish that the measures are consistent only in values with a threshold of 0.7 and above (Hair et al., 2006). On the other hand, Field (2009) mentions that Cronbach’s alpha values' acceptable range is between 0.5 and 1 (Field, 2009). Therefore, taking into account both these options, in this study, we conclude that a Cronbach alpha value more than 0.5, is acceptable.
On the other hand, the validity of the items in the questionnaire was tested using the Kaiser–
Meyer–Olkin (KMO) test (Kaiser et al., 1974). The KMO test is used to determine the sampling adequacy of each variable of interest. The statistic that is computed is a measure of 0 to 1. The sampling is adequate or sufficient if the value of Kaiser Meyer Olkin (KMO) is more extensive than 0.5. On the other hand, according to Pollant (2013), the value of KMO is 0.6 and above.
Hutchenson and Sofroniou (1999), and Kaiser (1974) recommend:
1. A bare minimum of 0.5
2. A value between 0.5 and 0.7 is mediocre.
3. A value between 0.7 and 0.8 is good.
4. A value between 0.8 and 0.9 is excellent.
5. A value between 0.9 and above is superb.
In this section, the Cronbach alpha test and validity test will be conducted using SPSS. The Cronbach alpha test output includes the reliability statistics, the item statistics that represent the item’s mean and the standard deviation, the inter-item correlations and item-total statistics
that indicate what would the Cronbach alpha value be if the item was deleted. All these outputs, for each variable of interest, are presented in the Appendix.
9.1 Reliability and Validity Testing (Greek Sample)
Using SPSS, the Cronbach’s alpha test and Kaiser–Meyer–Olkin (KMO) test was conducted to gain the reliability and validity results respectively of the different items in the Greek questionnaire. These tests were first applied to the Greek sample, and the results are presented at the tables below:
Table 12 Initial Reliability and Validity Testing (Greek sample) Variables Variable Name Question
no.
Table 13 Final Reliability and Validity Testing (Greek sample)
As one could indicate from the Table 12, the initial reliability and validity results suggest that all the variable values scales reach acceptable reliability, accept the Greek consumer attitude values scale comprising four items indicating an alpha lower than the 0.5 thresholds (0.328<0.5). On the other hand, in terms of validity testing, we can observe that the Online Consumer Buying Behaviour, Innovation, Consumer Attitude, Subjective Norms and Online Advertisement show mediocre sampling adequacy since their values are between 0.5 and 0.7 (Kaiser,1974). Besides, the e-WOM independent variable’s values scale comprising six items demonstrates non only the highest Cronbach’s alpha test (0.845). It displays the highest
sampling adequacy value (0.775) compared to all the other variable items. On the contrary, the perceived risk scale values encompass two questions and seem to have a bare minimum (Kaiser, 1974).
As mentioned previously, the Cronbach’s alpha test's output includes the reliability statistics, which gives us a hint of the mean and standard deviation of the different items of interest. The output also provides a correlation matrix and a table that displays the “new” alpha values if a particular item would be deleted from the questionnaire. The output results displayed in the Appendix, of the Cronbach’s alpha test for each of the variable items, the e-WOM’s, Online Advertisement’s and Perceived Risk’s items were not deleted. They stayed as they were because if an item got deleted in these variables, the Cronbach’s alpha test score would go down. Therefore, all the questions that consisted of these variables were kept as they were.
On the contrary, changes were implemented on the Online Consumer Buying Behaviour, Innovation, Subjective Norms and Consumer Attitude variable items because many of their items demonstrated week correlations and higher (or lower) mean scores than the other variables. This aspect suggests that they need to be removed from the questionnaire. Indeed, as we can observe from Table 13, the number of items of the variables mentioned previously have been reduced, and the outcome is a higher Cronbach’s alpha test. An important aspect worth mentioning is that although the Greek consumer attitude at the initial reliability test showed a Cronbach’s alpha test less than 0.5 now that we have deleted question 24 the alpha test score has risen to 0.752 (Appendix). In terms of the Greek sample, the results in Table 13 will be taken into account for the construction of the Greek dependent, control, and independent variables.
9.2 Reliability and Validity Testing (Finnish Sample)
Using SPSS, the Cronbach’s alpha test and Kaiser–Meyer–Olkin (KMO) test was conducted to gain the reliability and validity results respectively of the different items in the Finnish questionnaire. The results are displayed at the tables below:
Table 14 Initial Reliability and Validity Testing (Finnish sample) Variables Variable Name Question
no.
Table 15 Final Reliability and Validity Testing (Finnish sample)
Variables Variable Name Question no. Number of Items The initial reliability and validity test results in the Finnish questionnaire sample, suggest that the Innovation control variable and e-WOM and Online advertisement independent variables exhibit high Cronbach’s alpha test scores, above 0.8, and “great” sampling adequacy values (values between 0.8 and 0.9) (Kaiser, 1974). However, the Subjective Norms variable items demonstrate a very low Cronbach’s alpha test score (0.115) and a sampling adequacy score below 0.5. Therefore, one could consider deleting all of the variable’s items if the output result suggests that this action should be taken.
The output of the Cronbach’s alpha reliability test, in the Finnish sample, suggest that ten items from the Online Consumer Buying Behaviour variable should be deleted (Appendix). After deleting these items, the alpha test and sampling adequacy score have increased from 0.650 and 0.588 to 0.885 and 0.780, respectively. Also, the Consumer Attitude and e-WOM variable items will remain the same because the results do not suggest otherwise. However, the Subjective Norms variable scales comprising of 2 questions will be deleted. Even after deleting question 26, the results still lead to a low Cronbach’s alpha test (below 0.5suggesting that the items are not internally consistent.
Moreover, looking at the Innovation and Online Advertisement item’s reliability and validity output results, even if the obliteration outcomes suggest deleting some of their questions, it reduces their sampling adequacy scores from 0.806 to 0.500 for Innovation and from 0.814 to 0.751 for Online advertisement. Because before deletion the Cronbach’s alpha test for both Innovation and Online Advertisement was way above the 0.5 thresholds (0.868 and 0.886 respectively). Their sampling adequacy was within the range of 0.7 and 0.9, and the final results will keep the items as they were before their deletion. In terms of the Finnish sample, the results in Table 15 will be considered to construct the Finnish dependent, control and independent variables.