CHAPTER 5 THE IMPACT OF PUNISHMENT ON THE RATE OF EFD USE
5.3 Fear of the punishment and rate of EFD uses
5.3.2 Perceived level of punishment
Low Reference value Source: (Author’s Design, 2020)
Based on the information presented in Table 5-13, the analysis suggests a significant impact of the perceived level of awareness of punishment on the rate of EFD use. First, the relationship between the awareness of punishment due to non-compliance and the rate of EFD use fits the ordinal equation, because the model fitting p-value is 0.000. Furthermore, a significant difference is identified across parameters of the input variable, when the referenced parameter is set as “low”; all p-values are less than 0.05. In this case, the awareness of respondents on the perceived level of punishment (exercised by the revenue authority) determines the rate of using EFD. When the perception is low, the use is equally low, and vice versa. This element of findings is supported by authors such as Armborst (2017) and Nkwe (2013), where both suggested the size of the punishment to determine compliance with tax laws.
High 45 52% 6 7% 87
Moderate 19 56% 10 29% 34
Low 5 28% 4 22% 18
Very low 4 50% 1 13% 8
Total 157 56% 24 9% 279
Source: (Author’s Design, 2020)
According to Table 5-14, those who fear punishment perceive the current setting of punishment as high or severe (58%), compared to those who do not fear (5%), on average.
This indicates that if penalties and other associated punishments are properly enforced on those who are non-compliant, the magnitude of the problem will decrease because fear leads to compliance. The observation of the impact of punishment on taxpayers who are not compliant is equally supported by researchers such as Devos (2013), and Mohdali, Isa, and Yusoff (2014), and Yunus, Ramli, and Hassan (2017), who collectively suggested that when there is no fear, the chance for compliance is low. Further, the study used the One Way ANOVA to test the significance of the categorical relationship between the perceived level of punishment and the fear of punishment expressed by taxpayers. Table 5-15 presents the results.
Table 5-15: Perceived level of punishment and fear of punishment – ANOVA test Sum of Squares df Mean
Square
F Sig.
Between Groups 16.431 4 4.108 4.669 .001
Within Groups 241.060 274 .880
Total 257.491 278
Source: (Author’s Design, 2020)
Based on Table 5-15, the results suggest a significant categorical relationship between the perceived level of punishment and the fear of punishment due to non-compliance. The observed p-value is 0.001. This value is significant because it is less than the threshold, which is 0.05. Apart from the descriptive information provided in Table 5-13, additional descriptive information in Table 5-16shows a steady increase in the mean value across the measurements of the “perceived level of punishment” – that is, very high, high, moderate,
and low. The “very low” response showed the decrease in the mean to 2.31. A general explanation for this observation is that respondents who perceive a high/very high level of punishment are more inclined to show a high fear of punishment when compared to the rest of the categories. The current study ignores respondents with a very low level of punishment because their frequency is too low to show statistical relevance (Levers, 2013).
Table 5-16: The mean perception on the fear of punishment based on the level of punishment
Level of punishment N Mean
Very high 132 2.13
High 87 2.32
Moderate 34 2.68
Low 18 2.94
Very low 8 2.13
Total 279 2.31
Source: (Author’s Design, 2020)
A further analysis was conducted to determine the impact of the perceived level of punishment to the fear of punishment. In this part of analysis, the study engaged the Ordinal Regression analysis for testing the relationship. Based on the results presented in Table 5-17, the model fits the analysis because the model fitting information is represented by the p-value equal to 0.011 (p<0.05). Accordingly, the study relied on Nagelkerke Pseudo R-Square because it provides a close equivalency to traditional r-square value, compared to other pseudo values (Maroun, 2012). Based on the current study, the Nagelkerke r-square value is 0.042, which is 4.2% of impact. Although, the impact is viewed as low, it should still be considered given the fact that the study is based on respondents‟ perceptions. Furthermore, a simple observation of parameter estimates shows that only respondents with a very high perception of the level of punishment showed a significant response towards the fear of punishment, when the low level perception of punishment is set as a reference value. The observed p-value is 0.012, which is less than the threshold value. Therefore, it suggests that respondents who perceive a very high level of punishment are more inclined to fear punishment associated with non-compliance. This
observation is supported by Devos (2013) and Fellner, Sausgruber, and Traxler (2013) who suggested that punishment increased the desire to comply due to fear.
Table 5-17: Level of punishment and fear of punishment ordinal regression extracts
Element of measurement Value
Model fitting information P-value = 0.011
Nagelkerke Pseudo r-square r-square = 0.042 or 4.2%
Parameter estimates for the input variable (level of punishment)
Very high p-value = 0.012
High p-value = 0.108
Moderate p-value = 0.945
Low Reference value
Source: (Author’s Design, 2020)
Further to this, the study tested whether the perceived level of punishment determined the rate of EFD use. In this analysis, data from the perceived level of punishment were transformed to a three-level Likert scale, to have closer-related inputs grouped together. In addition, this eliminated input elements with fewer responses to fit to Ordinal Regression (Al-Khulaifi, 2012). The results of the Ordinal regression model are presented in Table 5- 18.
Table 5-18: perceived level of punishment and the rate of EFD use – ordinal regression extracts
Element of measurement Value
Model fitting information p-value = 0.449
Nagelkerke Pseudo r-square R-square = 0.006
Parameter estimates for the input variable (level of punishment)
High 0.193
Moderate 0.471
Low Reference value
Source: (Author’s Design, 2020)
According to the results in Table 5-18, the information used in the analysis do not fit the Ordinal Regression model. This is because the model-fitting information value (p=0.449) is greater than the threshold (p=0.05). Furthermore, in all cases the impact of the perceived
level of punishment on the rate of EFD use is insignificant because all elements of the perceived level of punishment have a p-value greater than the threshold. Therefore, there is no significant difference on the impact, across all values of the independent variable.