Before evaluating the effect of each explanatory variable in the model, it was very important to determine whether the model would improve its ability to predict outcomes. A statistically significant chi-square statistic (p < 0.05) indicates that the final model significantly improves the baseline intercept model.
As the model summary above shows, the likelihood ratio for the chi-square statistic (116.72) is high and statistically significant at the 5% significance level. This indicates that the included parameters provide better predictions compared to the null model with no predictors.
According to the (22) criteria for the best-fitting model (pseudo R2 values between 0.2 and 0.4 are considered to be a good fit), we found the overall fit of the logit model to be worthwhile. Because the regression results model shows that the pseudo R2 (0.26) satisfies the above assertion.
One of the important steps in a logistic regression model is figuring out whether or not there exists multicollinearity amongst unbiased variables.
Multicollinearity happens while or greater explanatory variables are notably correlated to every other. Therefore, the variance inflation factor (VIF) can be customized to detect the multicollinearity diagnostic for independent variables.
VIF (variance inflation factor) is a hallmark of how much of the inflation of the standard error could be caused by collinearity. As stated by (23), if the value of VIF is less than 10, then no multicollinearity problem exists.
As reflected in the table above, the result confirms that the VIF for all factors is less than 10. Since the variance inflation factor is less than the cutoff point, this indicates that all variables are relevant and multicollinearity is not there since the variance inflation factor (VIF) for each variable is less than the cutoff point 10.
4.3. Regression Results
To test the hypothesized relationships between the independent variables (age, gender, marital status, education, income, business type, duration in the business, tax knowledge, tax fairness, complexity of tax system, tax rate, financial constraint, service delivery, penalties, audit) and the dependent variable (taxpayers' tax evasion perception), the binary logistic regression analysis was conducted.
The output from this analysis, a beta coefficient, provides an assessment of the significance, the impact of the explanatory variables on the dependent variable, level of likelihood, and the pseudo R squared which indicates the model fitness. In the regression result, the independent variables may have a positive or negative coefficient, which describes the nature of the effect that they exerted on the dependent variable.
The independent variable with negative coefficients implies that it hurts the dependent variable and vice versa.
The first iteration (called iteration 0) is the log-likelihood of the "null" or "empty" model; that is, a model with no predictors. At the next iteration, the predictor(s) are included in the model. At each iteration, the log-likelihood increases because the goal is to maximize the log-likelihood.
Prob > chi2 is the probability of obtaining the chi-square statistic given that the null hypothesis is true. This is, of course, the p-value, which is compared to a critical value to determine if the overall model is statistically significant. The likelihood ratio chi-square of 116.72 with a p-value of 0.0000 tells us that the model as a whole is statistically significant.
Table 2
TE | Coef. | Marginal effect | Std. Err. | z | P>|z| |
SEX | − .3807322 | − .0899275 | .3152445 | -1.21 | 0.227 |
AGE | − .2095392 | − .0494923 | .1479765 | -1.42 | 0.157 |
MART | .2343165 | .0553446 | .1812703 | 1.29 | 0.196 |
EDUC | .2566447 | .0606185 | .1336001 | 1.92 | 0.055 |
FAIR | .427364 | .1009417 | .1521802 | 2.81 | 0.005 |
DUR | − .2260923 | − .0534021 | .1544366 | -1.46 | 0.143 |
SQ | − .5091038 | − .1202483 | .1665873 | -3.06 | 0.002 |
Teduc | − .4738096 | − .111912 | .2023156 | -2.34 | 0.019 |
FC | − .6958697 | − .1643617 | .1424733 | -4.88 | 0.000 |
CTS | .0047239 | .0011158 | .1654089 | 0.03 | 0.977 |
TR | .6524214 | .1540994 | .1627031 | 4.01 | 0.000 |
IL | .4139863 | .0977819 | .1270285 | 3.26 | 0.001 |
TS | − .0218883 | − .0051699 | .1472003 | -0.15 | 0.882 |
TAU | .500048 | .1181094 | .15956 | 3.13 | 0.002 |
PEN | − .280243 | − .0661923 | .1440523 | -1.95 | 0.048 |
The regression result shows that there are eight independent variables including the fairness of the tax system, income level, tax rate, tax education, financial constraint, audit, penalty, and service quality in the models that have a significant influence on tax evasion at a significance level of 5%.
The results show that sex, age, marital, education, complexity, business sector, and duration are not important factors in determining tax evasion at a 5% significance level. Accordingly, financial Constraint, tax education, service quality, and penalty are negatively associated with tax evasion, while the fairness of the tax system, income level, audit, and the tax rate are positively associated with tax evasion. That means, that tax knowledge, improved service delivery and absence of liquidity (financial constraint) will reduce tax evasion; meanwhile, better income levels, high tax rates, unfair tax, and poor audit regime will increase tax evasion.
The variable that influences tax evasion perception of taxpayers are education level of taxpayers, which is positive and (p = 0.05) significant at a 1% level of significance, this indicates that a one-year increase in education will lead to 0.25 units increase in tax evasive holding all other variables constant. The other variable that influences tax evasion is the taxpayer’s knowledge of tax rules and regulations (β = − .47380, p = 0.019). This means that; when the taxpayers’ know-how and understandings of tax rules are relatively high then the tax evasion perception of the taxpayer decreases by 0.47 units, other factors held constant.
Financial constraint is the other significant factor (β = − .69, p = 0.000, and marginal effect = − .1643617), for every financial constraint that happens, the probability of tax evasion perception of taxpayers decreases by 16.4%, other factors being at their margin; when every financial constraint will go on, the taxpayer flight from evading tax. Tax fairness was found to have a significant impact on tax evasion. The result indicates that when the tax system is not fair, the probability of tax evasion perception of taxpayers increases by about 10.1%.
Another significant factor is quality service; when quality service is delivered to taxpayers, the Probability of the taxpayers being evasive is decreased by 12.0%, holding all other factors constant. The tax rate is one of the factors that determine tax evasion. A regression table shows that a one percent increase in marginal tax rate will encourage tax evasion by 15.4%, other factors being constant.
Income level was found to be significantly determining tax evasion at a 5% level of significance with a marginal effect of.0977819. The result of the regression table tells us when an increase in the income level of taxpayers, the probability of taxpayers' tax evasion perception also increases by about 9.8%. The result of the tax audit was found to be positive and significantly determined tax evasion. Other factors remain the same when the probability of being audited is high the probability of tax evasion perception of taxpayers increases by 11.8%. The penalty of a taxpayer was also found that hurts the tax evasion behavior of taxpayers at a 5% level of significance.
Fairness of tax system
The logistic regression result shows that the relationships of tax evasion regarding attitudes toward the fairness of the tax system (β4 = 0.42) with marginal effect (0.10094) are positive and significant, holding other explanatory variables constant when an increase in the perception of the tax system is unfair, the probability of tax evasion perception increase by about 10.1%. This means that when taxpayers feel the tax system is fair their willingness to pay taxes is also increased. The result is consistent with the study (9), (17), and (19).
Tax rate
Tax rate (ß =.65242) and marginal effect (.1540994) have a positive and significant association with tax evasion perception. This result suggests that a one percent increase in marginal tax rate will encourage tax evasion by 15.4%. This implies that if the government increases the tax rate, the probability of being at a higher tax evasion level is realized. This result is consistent with (24).
Income level
The coefficient of income level is.41398 and its marginal effect is (.0977819). This means that there is a positive significant relationship between the income level of taxpayers and tax evasion. This implies that holding other explanatory variables constant when the income level of taxpayers increases the probability of taxpayers' tax evasion perception also increases by about 9.8%. The result is contrary to the work of (15).
Tax office quality service delivery
The tax office's good quality service delivery has a statistically negative (ß= -0.51) with a marginal effect of (-.1202483), the P-value is 0.002, which is less than 0.05 significant levels. This implies that improved service delivery decreases the probability of tax evasive behavior of individual taxpayers by 12.0%. This result is consistent with the finding of (19).
Tax knowledge (Tax Education)
The binary logistic result shown in Table (above) revealed that tax evasion was influenced by taxpayers' knowledge (β =-.47380, P < 0.05) with a marginal effect of (-.111912). It was found to have a negative and significant effect on the tax evasion behavior of taxpayers at a 5% level of significance.
This implies that as the individual's tax awareness improves, the tax evasive behavior of the individual decrease by 11.2%, other factors being constant. This result is also consistent with (17) and (19).
Personal financial constraints
As shown in the regression table, the explanatory variable financial constraint was found to be a negative and statistically significant relationship with tax evasion at a 5% significance level. The ordered regression result (β= − .6958697) with marginal effect (-.1643617) indicates that an increase in financial constraint causes tax evasion behavior of the taxpayer to decrease by 16.4%, other factors being constant. The result is at odds with the study of (25) (26) and (27), people who face personal financial problems are likely to be more prone to evade tax.
Audit coverage
The regression analysis states that lower audit regimes have a positive (ß =.5000) and very strong significant relationship with tax evasion perception of taxpayers at a 5% significance level. The marginal effect (.11810) indicates that other factors remain the same when the probability of being not audited increases by one unit the probability of tax evasion perception of taxpayer increases by 11.8%. This result is consistent with that of (13), (28), and (29) the probability of being audited, the more positive compliance attitude of taxpayers, and a higher audit regime reduce tax evasion significantly.
Penalties
Regarding penalty (ß =-.2802), it has a negative and significant association with tax evasion at 5% significance. This implies that an increase in the penalty of tax non-compliant taxpayers and the likelihood of taxpayers’ tax evasion perception decreases by about 6.6%. This result is consistent with that of (13), (9), and (17).