Methodology And Analytical Choices

Published Date: 02 Nov 2017

This section of the thesis will be discussing the theoretical framework, the linkages with theoretical justifications of the independent variables with the dependent variable. The tax effort is represented by the tax gap between actual and budgeted tax revenue as mentioned before. Furthermore, the statements of hypothesis have been stated covering the major components of foreign aid; grants and loans, along with other economic and institutional variables affecting the tax gap. Keeping the relevant determinants of tax effort in view as well as keeping in line with the purpose of the thesis, three competing models have been constructed largely depicting the impact of the major foreign aid components affecting the tax gap of Pakistan. Weighted Least Square has been used on models 2 and 3 for the incidence of heteroskadasticity.

3.1. Framework of Analysis

To build a regression model we have to choose an explained variable that reflects the effort in tax collection and tax administration in the period under view, and explanatory variables that reflect the dependency and requirement of foreign aid in terms of its impact on the efficiency and quality of collecting tax revenue.

Theoretical Framework

The flowchart above shows this relationship. In literature, tax effort has been calculated by using the tax to GDP ratio of the country or countries, in case of panel data. But in the case of Pakistan, the tax to GDP ratio data is not used as a proxy to tax effort. This is because of its lack of volatility in the data. The results using tax to GDP ratios were coming out ambiguous. Therefore, instead of using tax to GDP, the dependent variable used is the difference in actual and budgeted tax revenue.

The data for the actual and budgeted tax revenue was taken from the Economic Survey of Pakistan statistical supplements, issues ranging from 1989-90 to 2010-11.

Among many explanatory variables identified from literature, only nine relevant variables to Pakistan were selected in order to carry out three regressions, assessing the relationship of components of foreign aid, individually and as a whole. The first regression focuses on the relationship of grants received by the Pakistan Government with the tax effort. The second regression focuses on the relationship of loans financed and credited to Pakistan Government with the tax effort. The third regression focuses on the relationship of net official development assistance received by the Government along with the estimates of corruption in the Federal Board of Revenue by using number of tax payers as a proxy to tax administration, against tax effort.

In many studies, the effect of grants on the tax effort, specifically tax revenue has been ambiguous, depending on the scope and the context of the study. Despite their nonpayment nature and less fiscal burden, they still tend to affect tax revenue and tax effort adversely in most cases. According to Khan and Hoshino (1992) grants, mostly indicating project aid, reduce the tax effort in their study on India and Kenya. Another study by Gupta et al. (2003) showed that grants have a negative effect on tax revenue. He said the impact of aid on tax revenue depends on the composition of aid. They said that their results reflect the fact that grants have a modest affect on revenue on average but it is high in countries with weak institutions. Furthermore an OECD report, (Hansjörg Blöchliger and Oliver Petzold, 2007). It assessed the revenue composition of sub central government of OECD countries. It revealed that SCG’s revenue is not increasing as much as the increase in government expenditure. This is largely due to the dependence on intergovernmental grants. This resulted in the conclusion ‘that grants reduce SCG tax effort, inflate SCG spending and increase SCG deficits and debt.’

Similar to grants, the effect of loan on tax revenue or tax effort is ambiguous. In some studies it shows a positive relationship with tax effort or tax revenue while others shows a negative trend. Khan and Hoshino (1992) study shows that loans have a positive relationship with tax effort. Moreover, Gupta et al (2004) study showed that loans have no significant or positive effect on government revenues. Furthermore, a study by Benedek et al (2012) has its conclusions in line with Gupta et al (2004) showing a positive relationship of loans and tax revenue in a cross sectional data of 118 countries. In another incidence he examined that the affect of concessional loans is positive on domestic revenue mobilization and negative for grants (2003).

If we examine the effect of Foreign Aid as a whole on Tax effort or tax revenue, it shows a different picture as well. Studies based on panel data show a negative effect of aid on tax but individual countries show both negative and positive. It depends on the composition of aid as various studies suggest. According to a study by Feeny (2006) on Melanesian countries, it was found that foreign aid causes increase in development expenditure and fall in tax revenues and borrowing. It was also examined that aid grants have more favorable public sector response relative to loans. Moreover, it is also examined that aid has a statistically insignificant affect on revenue and does not reduce revenue collection efforts in developing recipient countries (Ouattara 2006).

Level of Economic Development is one of the important variables impacting the tax revenue collection of a country. In most studies GDP per capita is incorporated as a proxy for level of economic development. It is seen that increase in GDP per capita affects the tax collection efficiency to increase significantly. According to Clist and Morrissey (2009) that the relationship between GDP per capita and tax revenue is non linear across 82 developing countries over the period from 1970–2005. It also examined that GDP per capita coefficient positively affected ‘tax collection efficiency’. Moreover, Le et al (2012) examined a cross sectional analysis of taxable capacity in various countries around the world. It was seen that there is a positive correlation of tax revenue with GDP per capita.

Industry value added (%age of GDP) is also seen to be a positive contribution towards increasing the tax collection efforts of the country. This is because the industrial sector is easier to tax, unlike the agricultural sector, especially in the developing countries. According to Wang et al (2009), while assessing the provincial tax performance in China, it was examined that a large share of the industry value added generated China’s revenue. Industrial share of GDP was found one of the most positively significant determinants for generating tax revenue and increasing taxable capacity. In most studies the agricultural share of GDP has been incorporated in the models assessing the affect on tax revenue. Le et al. (2012), Bird (1976), Bahl. (2003), Ahmad and Stern (1991).

For corruption, Tax Evasion to GDP is incorporated in the model. The Tanzi’s methodology was used to estimate the size of the underground economy by taking the log-ratios of currency in circulation and foreign currency accounts to M2 as a dependent variable, while log of total tax revenues as a percentage of GDP, log of interest rate on time deposits and log of dummy variable were taken as independent variables in the paper by the Pakistan Institute of Development Economics paper (2004). In the previous literature, the relationship of tax evasion with tax revenue has been seen as significantly negative. While assessing the tax structure of Uganda, Tessa (2002) identified that tax evasion was one of the factors that negatively affected tax revenue. According to Cobham (2002), in a set of 142 countries, the cost to the developing countries in terms of tax evasion and avoidance is around $385 million annually. Moreover, Lacke (1999) studied the effect of underground economy that it will increase the tax gap. This is because an increase in tax burden will cause more tax evasion thus widening tax gap.

POLCONV is the Political Constraint Index taken from the POLCON Database. It is the measurement of the level of Political constraint to the Executive in the policy making. Its scale ranges from 0 to 1. POLCON=0 is interpreted as having the minimum political constraint on the executive leading to most hazardous political environment, while POLCON=1 stands for having maximum political constraint leading to most stable political environment. (Henisz, W. J. (2000). "The Institutional Environment for Economic Growth." Economics and Politics 12(1): 1-31.)

Total tax collection is the combination of federal as well as provincially collected tax revenue. These are the indirect and the direct tax revenues from the formal economy, assigned to the respective tiers of the Government. The relationship of the total tax revenue collected with the tax gap is expected to be significantly negative. This is because an increase in tax revenue indicates better tax collection efficiency, better tax administration and tax compliance, pointing towards high actual tax revenue over budgeted tax revenue hence reducing the gap. This may also indicates towards higher tax rates imposed on the economy which would result in more tax collection.

Fiscal discipline is defined as the responsible taxing and spending pattern of the government and the institutional limits placed on political decisions in settling a budget for the country. This definition is according to the Advisory Commission of Intergovernmental Relations in the United States of America in their report regarding fiscal discipline in the federal system (1987). In another incidence it is defined as ‘the capacity of a government to maintain smooth daily financial operation and long-term fiscal health.’ (Hou 2003) It is a very broad term and has various definitions in the previous literature. But in this thesis, it is used as to assess the ratio of federal expenditures over revenues and how much these expenditures are in excess or short from the revenues during the period 1990 to 2010. It is proposed that the expenditures of the Federal Government of Pakistan are in excess of the tax revenues they receive. This gap between expenditure and revenue is expected to be largely due to fiscal Indiscipline and maladministration at the part of the tax authorities.

3.2. Statement of Hypothesis

The aim of this thesis is to assess whether foreign aid inflow in Pakistan affects the quality of tax administration and improves the effort put in by the Government and the Federal Board of Revenue, in the collection of tax revenue from the formal economy of Pakistan, or not, for the time period 1990 to 2010.

Following are the proposed hypothesis of this thesis:

H0 = Amount of Foreign Aid inflow received by Pakistan, in the form of Grants, does not have a significantly negative relationship with tax effort of Pakistan.

H1 = Amount of Foreign Aid inflow received by Pakistan, in the form of Grants, has a significantly negative relationship with tax effort of Pakistan.

To test the proposition whether the component of Foreign aid Inflow in the form of Grants negatively impacts the quality and effort of Tax collection in Pakistan, or not, in the time series setting from 1990 to 2010.

H0 = Amount of Foreign Aid inflow received by Pakistan, in the form of Loans, does not have a significantly positive relationship with the Tax Effort of Pakistan.

H1 = Amount of Foreign Aid inflow received by Pakistan, in the form of Loans, has a significantly positive relationship with the Tax Effort of Pakistan.

To test the proposition whether the component of Foreign aid Inflow in the form of Grants positively impacts the quality and effort of Tax collection in Pakistan, or not, in the time series setting from 1990 to 2010.

H0 = Amount of Total Foreign Aid inflow in Pakistan does not have a significantly negative impact on the Tax Effort of Pakistan

H1 = Amount of Total Foreign Aid inflow in Pakistan has a significantly negative impact on the Tax Effort of Pakistan

To test the proposition whether the total Foreign Aid inflow in Pakistan negatively affects the quality and effort in tax collection of Pakistan, or not, in the time series setting from 1990 to 2010. Here the variables used for total Foreign Aid are Net Official Development Assistance (ODA) ($ millions) and Non Development Aid in the form of Military Assistance ($ millions), predominantly from the USA.

H0 = Level of Economic development of Pakistan does not have a significantly positive effect on the Tax effort of Pakistan.

H1 = Level of Economic Development of Pakistan has a significantly positive effect on its tax effort.

To test the proposition whether the level of Economic Development of Pakistan have a significantly positive relationship with the quality and effort in the collection of tax, or not, in a time series setting from 1990 to 2010. Here, a proxy variable to represent the level of Development has been taken as "GDP per capita."

H0 = Industry value Added (%age of GDP) of Pakistan does not have a significantly positive impact on its Tax Effort.

H1 = Industry value Added (%age of GDP) of Pakistan have a significantly positive impact on its Tax Effort.

To test the proposition whether the value of the industrial sector of Pakistan (%age of GDP) have a positive impact on its quality and effort of Tax collection, or not, in the time series setting from 1990 to 2010.

H0 = Incidence of fiscal indiscipline of Pakistan does not have a significantly negative impact on the Tax Effort of the country.

H1 = Incidence of Fiscal indiscipline in Pakistan has a significantly negative impact on the Tax Effort of the country.

To test the proposition whether the incidence of Fiscal Indiscipline in Pakistan has a significantly negative impact on the quality and effort of Tax collection in the country, or not, in a time series setting from 1990 to 2010. Here, the ratio of Federal Expenditure to Federal Tax revenue has been incorporated as a proxy for Fiscal indiscipline for the analysis of the respective time period.

H0 = Incidence of Corruption in Pakistan Economy does not have a significantly negative impact on the tax effort of Pakistan

H1 = Incidence of Corruption in Pakistan Economy has a significantly negative impact on the Tax Effort of Pakistan

To test the proposition whether the incidence of corruption impacts the Tax Effort of Pakistan in a significantly negative manner, or not, in a time series setting from 1990 to 2010. Here, a proxy variable to represent corruption has been taken as ‘Tax evasion to GDP ratio’

H0 = Incidence of Minimum Political Constraint on the Executive level of Government of Pakistan does not have a significantly negative impact on the Tax Effort of the Country.

H1 = Incidence of Minimum Political Constraint on the Executive level of Government of Pakistan has a significantly negative impact on the Tax Effort of the Country.

To test the proposition whether the incidence of Minimum Political Constraint on the Executive Government of Pakistan, in policy decision making has a significantly negative impact on the quality and effort of the collection of tax revenue of the country, or not, in the time series setting from 1990 to 2010. Here, the variable incorporated into the analysis is the POLCONV from the POLCON Database.

3.3. Elements of Research Design

The dissertation is a co-relational study- an inquiry to know the important variables associated with the problem. The research engages in hypothesis testing which would explain the nature of relationships between the variables. Hypothesis testing is undertaken to explain the variance in the dependent variable and to predict organizational outcomes.

Econometric modeling, Weighted Least Squares (WLS) in particular, will be used in the analysis. This is because to avoid the incidence of heteroskadasticity in the variables of the models 2 and 3. Three competing models were run to depict the model building stage of the procedure. The study settings will be purely artificial and it will be a time series analysis ranging from 1990 to 2010. Therefore the data hopes to illustrate the trend till the very recent time period. There are proxies for some included independent variables.

Data Sources

The data for these variables has been taken from the handbook of the Economic Survey of Pakistan, The World Development Indicators and the Statistical Bulletin, State Bank of Pakistan.

The definitions of the keywords including Loans, Grants, Tax Revenue and Tax Effort were taken mostly from the World Development Indicator.

Data Reliability and Consistency

As most of the data of the independent variables have been taken from the World Development Indicator, (WDI), they do not guarantee full reliability of the published data and do not accept responsibility of any consequences. Therefore, the reliability of data is not very high.

Interpolated Data

There was an incidence of missing data for two variables; Tax evasion to GDP and the POLCONV. As the data for Tax Evasion to GDP was taken from the PIDE publication by Ali Kemal (2007), it had estimated data from the years 1974 to 2005. Therefore, by using the statistical software MINITAB, the missing data for the five year period from 2006 to 2010 was interpolated and incorporated in the original data.

Similarly, from POLCON database, there was missing data of POLCONV from 2008 to 2010. Therefore, by using MINITAB, the interpolated data for the respective three years was incorporated.

3.4. Model Specification

Competing Model # 1:

Difference in Actual and Budgeted Tax revenue = f ( Grants, Tax evasion to GDP, total tax collection, industry value added (%age of GDP))

Competing Model # 2:

Difference in Actual and Budgeted Tax revenue = f (Loans, total tax collection, industry value added (%age of GDP), Tax Evasion to GDP, POLCONV)

Competing Model # 3:

Difference in Actual and Budgeted Tax revenue = f (ODA, Military Assistance, total tax collection, GDP per capita, industry value added (%age of GDP), POLCONV, ratio of Federal expenditure to Federal tax revenue)

Econometric model, Ordinary Least Squares in particular, will be used to gauge the correlation between the dependent and the independent variables. Three competing models have been taken into analysis. OLS Regression was run in model 1, but due to the incidence of heteroskadasticity in the independent variables including ratio of expenditure to revenue (Refer Appendix for Tables) in models 2 and 3, Weighted Least Square was run in the regression in order to avoid heteroskadasticity. Some of the variables in the equations carry transformations in form, otherwise linear forms are incorporated.

Chapter 4: Estimation, Analysis and Conclusion

The forth chapter focuses on the econometric model in greater detail demonstrating the estimation and the analysis of the three econometric models and lastly ending the chapter with the basic conclusion of the topic and the econometric models. This matches the literature with empirical evidence of Pakistan.

4.1. Model Estimation

Model 1: Impact of Grants on Tax Effort (Difference in Actual and Budgeted tax Revenue)

Tax Revenue Gap = 3.20e+11 + 1.71e+10 [log of Grants] + 1.62e+10 [Tax Evasion to GDP] – 0.1549814 [Total Tax Revenue Collected] – 1.33e+13 [1/ Industry value added %age of GDP]

Dependent variable: Difference in Actual and Budgeted Tax Revenue

Independent variables:

Grants Assistance signed

Tax Evasion to GDP

Total Tax Revenue Collected

Industry valued added (%age of GDP)

Analysis of Variance

Source

Sum of Squares

DF

Mean Square

F-Ratio

P value

Regression

1.9074e+22

4

4.7684e+21

23.26

0.000

Residual Error

3.2806e+21

16

2.0504e+20

Total (Corr.)

2.2354e+22

20

1.1177e+21

R-Sq = 85.32% R-Sq(adj) = 81.66%

In the Model above the independent variables, especially Log of grants are regressed against the proxy for tax effort which is the Tax Gap. Other independent variables include tax evasion to GDP, total tax revenue collected and reciprocal of industry value added %age of GDP. Tax evasion is taken as a proxy for corruption. It can be observed that all the variables are appearing as significant at 95% confidence interval or 5% level of significance. The p-values all stand at being less than 0.05, which means they are all accepted statistically. The p-value in the ANOVA table also shows that the probability of rejecting the model is zero percent.

Taking the model in totality, the R- squared of the model shows that the 85.32% of tax gap is explained by all the independent variables. The adjusted R-squared shows that after adjusting for the degrees of freedom 81.66% of the dependent variable is explained by the independent variables. The coefficient of the log of grants shows that 1% increase in grants will cause the tax gap to increase by 0.0171%, ceteris paribus. The coefficient is shown as statically significant. The coefficient of tax evasion to GDP shows that Re. 1 increase in tax evasion to GDP will cause the tax gap to increase by Rs. 1.62e+10, ceteris paribus. The coefficient is statistically significant. The coefficient of total tax revenue collection shows that Re. 1 increase in total tax revenue will cause the tax gap to decrease by Rs. 0.1549; ceteris paribus. The coefficient is shown to be significant. The last coefficient of reciprocal of industry value added %age of GDP shows that 1% increase in industry value added shows decrease in the tax gap by 1/1.33e+13=Rs.0.75 million, ceteris paribus. Coefficient and signs of all variables in this model are consistent with the hypothesis.

Model 2: Impact of Loans on Tax Effort (Difference in Actual and Budgeted Tax Revenue)

Tax Revenue Gap = 7.39e+11 -8.56e+09 [log of Loans] – 0.1067935 [Total Tax collected] – 1.56e+13 [1/industry value added %age of GDP] + 2.04e+10 [Tax Evasion to GDP] + 8.61e+10 [POLCONV] – 1.30e+11 [POLCONV2 ]

Dependent variable: Difference in Actual and Budgeted Tax Revenue

Independent variables:

Loans and Credit Contracted

Total Tax Revenue Collected

Tax Evasion to GDP

Industry valued added (%age of GDP)

Political Constraint – POLCONV

Analysis of Variance

Source

Sum of Squares

DF

Mean Square

F-Ratio

P value

Regression

1.8365e+22

6

3.0609e+21

10.74

0.0001

Residual Error

3.9891e+21

14

2.8494e+20

Total (Corr.)

2.2354e+22

20

1.1177e+21

R-Sq = 82.16% R-Sq(adj) = 74.51%

The above model focuses on the impact of loans on tax effort. Among the five variables incorporated, three of them are appearing significant, including total tax collected, industry value added as a %age of GDP and tax evasion to GDP. The insignificant variables are log of loan and the POLCONV. In totality, 82.16% of the dependent variable is explained by the independent variables, while after adjusting for degrees of freedom, 74.51% of the tax gap is explained by them.

The coefficient of log of loan shows that 1% increases in loan received causes the tax gap to decrease by 0.0856%, ceteris paribus. This is appearing as insignificant at 95% confidence interval. The coefficient of tax collected shows that Re. 1 increases in total tax collected causes a decrease in tax gap by Rs. 0.106, ceteris paribus. The coefficient is appearing significant. The coefficient of tax evasion to GDP shows that increase in tax evasion to GDP will cause an increase in tax gap by Rs. 2.04 million. The coefficient is seen as significant. The coefficient of the reciprocal of industry is shows that 1 unit increase in industry value added will cause decrease in tax gap by Rs. 0.64 million, ceteris paribus. The coefficient is coming out to be significant. The coefficient of POLCONV shows that it will increase tax gap by Rs.8.61e+10. As the relationship of POLCONV with tax gap has been seen as increasing at a decreasing trend thus the square of this variable has been incorporated in the model in order to represent the relationship.

Model 3: Impact of Net Official Development Assistance on Tax Effort (Difference in Actual and Budgeted Tax Revenue)

Tax Revenue Gap = 6.38e+11 + 2.53e+09 [Log of ODA] – 9.46e+15 [1/GDP per capita] – 0.2385731 [Total Tax Revenue Collected] + 366710.9 [Military Assistance] – 7.35e+12 [1/Industry value added %age of GDP] + 1.01e+10 [Federal Expenditure/ Federal Tax Revenue]

Dependent variable: Difference in Actual and Budgeted Tax Revenue

Independent variables:

Net Official Development Assistance

Total Tax Revenue Collected

GDP per capita constant

Industry valued added (%age of GDP)

Military Assistance

Fiscal Indiscipline ( Federal Expenditure/ Federal Tax Revenue)

Source

Sum of Squares

DF

Mean Square

F-Ratio

P value

Regression

1.8233e+22

6

3.0388e+21

10.32

0.0002

Residual Error

4.1215e+21

14

2.9439e+20

Total (Corr.)

2.2354e+22

20

1.1177e+21

R-Sq = 81.56% R-Sq(adj) = 73.66%

The model above focuses on the impact of total foreign aid, development and non development, on tax effort. In the regression, only two variables are appearing insignificant, namely the log of net official development assistance and the proxy of fiscal indiscipline that is ratio of federal expenditure over federal revenue. In totality, the model is shows as significant with a p-value as 0.0002, less than 5%. While the f ratio shows that there is 10.32 probability of accepting the model. The R-squared, referred to as coefficient of determination, shows that 81.56% of the tax gap is explained by the independent variables, while after adjusting for the degree of freedom, tax gap is 73.66% explained.

The coefficient of log of ODA shows that 1 % increase in ODA received to Pakistan causes a 0.0253% increase in tax gap, ceteris paribus. The coefficient is appearing as insignificant. The coefficient of reciprocal of GDP per capita shows that 1 unit increase in GDP per capita causes the tax gap to decrease by 1/9.46=Rs0.1057 million, ceteris paribus. It is appearing as significant. The coefficient of the total tax revenue collected shows that Re.1 increase in tax revenue collection will cause the tax gap to decrease by Rs. 0.2385791, ceteris paribus. It is appearing significant. The coefficient of military assistance shows that increase in military assistance by $ 1 causes the tax gap to increase by Rs. 366710.9, ceteris paribus. The coefficient of reciprocal of the industry value added %age GDP shows that 1% increase in industry value added will cause a decrease in tax gap by 1/7.353+12=Rs. 0.1359, ceteris paribus. The coefficient of ratio of federal expenditure over federal revenue shows that it is greater than 1 thus federal expenditures are more than the revenues, resultantly, causing the tax gap to increase, ceteris paribus.

4.2. Findings and Analysis of the Model

This section explains the research hypothesis statement in the above sections, keeping in view the study objectives of the thesis. Each significant independent variable will be explained about its linkage with the Tax Gap.

4.2.1. Nexus between Loans and Tax Effort

Model 2 shows the impact of loans received, as a component of foreign aid, on the tax effort of Pakistan. The data of loans is taken from the Statistical Bulletin of the State Bank of Pakistan for the years 1990 to 2010. The amount was in million dollars thus the amounts were converted to rupees by multiplying it with the official exchange rate (LCU per US$ per average) (WDI). It is expected that the loans received will have a positive impact on tax effort. Therefore, as loans increase, it will cause the gap of tax revenue to decrease. On the contrary, the relationship of loan with the tax gap has been seen as insignificantly negative in the model. This can be said that the burden of debt from loans increases due to which tax rates also increases, there is more fiscal management. Loans to Pakistan has always been attached with conditionalities, thus increases tax rates and tax effort. Furthermore, loans are taken in order to fulfill recurrent expenditures of filling the tax gap this showing an increase in tax effort. Therefore, the second null hypothesis is rejected but it shows insignificance. This is because of high incidence of corruption and tax evasion present in the economy.

4.2.2. Linkage between Grants and Tax Effort

Model 1 focuses on the relationship between grants received, as a component of foreign aid, by Pakistan and its tax effort. Tax effort is depicted by the difference in actual tax revenue and the budgeted tax revenue. The data on grants is taken from the State Bank of Pakistan website, Statistical Bulletin, the Public Finance section. The amount of grants was in dollars, thus it was converted in rupees by multiplying the amounts with the official exchange rate (LCU per US$ per average) (WDI). It is expected that grants negatively relates with tax effort. Therefore, an increase in grants causes the gap of tax revenue to increase, thus causing a decrease in the tax effort. This is due to tax rebates, tax exemptions and less tax burden on the Federal Government. Grants are showing a significant relationship to difference in tax revenue.

4.2.3. Nexus between Total Foreign Aid and Tax Effort

Model 3 focuses on the impact of Net ODA received by Pakistan on its tax effort. This data contains inflow of loans and grants alike. It is expected that ODA will decrease tax effort. Therefore, in the model it shows that increase in ODA received will cause increase in the gap of tax revenue leading to a decrease in the tax effort of Pakistan. The data of ODA is taken from WDI.

US Military assistance received by Pakistan through the years is taken as a non-developmental component of the foreign aid. It is expected to be negatively related to tax effort. Therefore, in this model it is shown that an increase in military assistance inflow to Pakistan by the US will cause the difference in tax revenue to increase. It shows high significance with difference in tax revenue.

4.2.4. Linkage between Corruption and Tax Effort

Tax evasion to GDP is taken as a proxy of corruption affecting Tax Network. It ranges from 3.02 to 7.18. It is expected to decrease the tax effort as it increases. Therefore, as tax evasion to GDP increases, it increases the gap between actual and budgeted tax revenue. It shows a high statistical significance to difference in tax revenue at 95% confidence interval in model one and model 2. This might be because of the high relevance and incidence of tax evasion in Pakistan. It is one of the major reasons for the increase in tax gap. There is high incidence of tax evasion in Pakistan because according to Bukhari and Haq (2011), people have lost their faith in any kind of government prevalent in the country and the departments functioning under it. There is a perception of lack of fairness in the institutions thus people avoid paying taxes.

4.2.5. Nexus between Value of Industrial Sector and Tax Effort

Industrial sector is one of largest sector being taxed in Pakistan. It is expected that the share of industry value added in percentage of GDP will have a positive effect on the tax effort of Pakistan. It is shown high significance in the models.

4.2.6. Linkage between Level Economic Development and Tax Effort

Level of economic growth or GDP per capita is expected to have a positive and a significant impact on tax effort. On graph, there is an increasing and decreasing relationship with difference in tax revenue. The model shows that as GDP per capita increases; there will be more tax collection which will reduce the difference between actual tax revenue and the budgeted. It is a statistically significant variable in the models. GDP per capita is taken from WDI.

4.2.7. Nexus between Fiscal Indiscipline and Tax Effort

Fiscal indiscipline in the federal government of Pakistan is quantified by taking the ratio of federal expenditure to federal revenues. It can be observed that federal expenditures has been increased by 12.57% from 1990 to 2010 while federal revenues have increased by only 11.55%, less than the expenditures. Therefore, there has been incidence of fiscal indiscipline in the country due to the gap between expenditures incurred and revenue collected. This is largely because of the incompetence of the tax administration and corruption in the economy. On the other hand, we can also say that the there has been many unexpected expenditures incurred by the government during the period under view, including the war against terrorism, earthquakes, floods etc. Thus, not all the gap between expenditure and revenue can be attributed to corruption and maladministration.

4.2.8. Linkage between Political Constraint and Tax Effort

POLCONV is expected to have a negative impact on the tax effort in case of Pakistan. The data of POLCONV for Pakistan ranges from 0 to 0.5, only 0.75 during 1996-97. Therefore, indicating minimum constraint on executive powers through the years. Low constraint on the executive powers means high corruption levels and an unstable government. This would lead to low tax collection by the authorities. This relationship is depicted by the model as POLCONV causes a Rs. 7.95 million increase in the gap of tax revenue thus decreasing tax effort. Nevertheless, it shows a high insignificance to tax effort. This might be because the data observations were inadequate as regards getting the full impact of corruption in the tax network and administration.

4.2.9. Other Determinants of Tax Effort of Pakistan

Other determinants of tax effort includes fiscal deficit, regulatory quality of the government, the bureaucratic index, trade openness, age dependency, Total tax collection, Political Stability and agriculture sector value added %age of GDP. Total Tax collection in the model is taken as a proxy of tax rates. The relationship between tax collection and tax effort is positive. Therefore, as tax collection increases, there is a decrease in the gap of tax revenue. It is a highly significant variable. It is a highly significant variable. Data is taken from various issues of Economic Survey of Pakistan.

4.3. Model Consideration, Tests and Overall Goodness of Fit

Due to the incidence of heteroskadacity in the dependent variable as well as few independent variables, all the models have been adjusted for this using Weighted Least Square using STATA statistical software. The test of serial correlation was found using MINITAB for all three models. Following table shows post estimation tests on the three models and their statistics:

Model 1 (Grants)

Model 2 (Loans)

Model 3 (ODA)

R-squared

85.32 percent

82.16%

81.56%

(adj.) R-squared

81.66 percent

74.51%

73.66%

F-Ratio

23.26

10.74

10.32

Breusch-Pagan / Cook-Weisberg test (heteroskadasticity)

chi2(1) = 0.03

Prob > chi2=0.8623

chi2(1) = 0.21

Prob > chi2=0.6486

chi2(1)=0.08

Prob>chi2=0.7756

Durbin-Watson statistic

(Serial Correlation)

2.21935

2.03916

1.84206

Mean VIF (Multicollinearity)

2.77

7.26

7.56

Overall, the models did not collapse. The coefficient of determination, R-squared of the three models, which is showing their overall goodness of fit, is respectable above 80%. The R-Squared adjusted for degrees of freedom (a better measure of the model specification) also shows a respectable percentage. The F ratios of the three models show that there is low probability of accepting the hypothesis. Moreover, the chi-squares of the Breusch-Pagan / Cook-Weisberg test for heteroskadasticity shows that they are small thus no incidence of heteroskadasticity. Furthermore, the Durbin Watson Statistics to test Serial Correlation is almost equal to 2 for all the models indicating no incidence of serial correlation. Lastly, in model 1, the Mean Variance inflation Factor (VIF) testing for multicollinearity, is shown to be less than 5 thus no incidence of multicollinearity. Nonetheless, the Mean VIF of the econometric models 2 and 3 are shown to be more than 5 showing incidence on multicollinearity. This is because of the presence of two important explanatory variables namely total tax revenue collection and GDP per capita constant in the same models. There is high correlation between these variables.

4.4. Conclusions

The study included an analysis of the tax system, administration and the efficiency of tax collection or tax effort of Pakistan and how it is affected by the inflow of aid in the country. It basically addresses the question whether and how one of the largest receipt countries of foreign aid is using it to improve and progress the efficiency of tax collection and tax administration of Pakistan. in previous literature there have been many international tax effort models discussed including Bahl Model, Lorz and Morss model, fiscal decision maker model; but in this research three econometric models were run focusing on the impact of the major components of foreign aid on tax effort as well as Foreign aid inflow as a whole.

Various explanatory variables were incorporated that determine the tax gap between actual and budgeted tax revenues of the country. This dependent variable was used as a proxy for assessing Tax Effort of Pakistan. Foreign aid was decomposed to two major components received by Pakistan; grants and loans. Resultantly, it was observed that grants had a significantly positive relationship with tax gap resulting in decrease in tax effort. While on the other hand, loans were seen to have an insignificantly negative relation with tax gap resulting in increase in tax effort. On the other hand, Net ODA was used as Development Aid and Military Assistance was used as Non Development Aid. Both the variables were indicating an increase in tax gap suggesting decrease in tax effort. Nevertheless, the ODA was seen to be appearing statistically insignificant as opposed to military assistance.

Other relevant variables included were GDP per capita, Tax Evasion to GDP, Industry value added as a percentage of GDP and total tax revenue collection. These were statistically significant variables at 95% confidence interval, and were crucial in determining Tax gap. Tax Evasion to GDP was taken as a proxy for Corruption. The institutional variables including Political Constraint on the Executive and Fiscal Indiscipline showed statistical insignificance with the dependent variable in question.

4.6. Limitations of Research

Limitation of the study include data collection period which was relatively short for an in-depth econometric analysis. The years totaled 21; hence the degree of freedom was 20, quite small for adequate analysis. Some of the variables were missing some data, which had to be interpolated using the ‘MINITAB’ software. Hence, the most accurate values might not have been achieved as actual validity of data was lost; nevertheless they were reasonably adequate for consideration. Also, the strategy for analysis included usage of econometric modeling, which has its own merits and loopholes. Furthermore, the competing models each did not have more than five variables in the regression equations and so may be including some specification bias.