Potential tax reforms and Kuwait economic growth

Размещено на http://www.allbest.ru/

Kuwait University, College of Business Administration

Potential tax reforms and Kuwait economic growth

K. Lawler, Prof. Farah Ali Al-Sayegh, Graduate Stud

Kuwait

Abstract

The objective of this study is to identify whether tax reforms are viable in Kuwait in order to create more government income from sources other than oil. The study examines the relationship between the changes in tax revenues, changes in oil revenue and changes in GDP in Kuwait using time series data from 1998 to 2015. The Augmented Dickey-Fuller (ADF) is used to check for the existence of a unit root. The cointegration test is applied to test for long term relationships between variables using the General Least Square (GLS) method of estimation. The results of the tests find that the impact of changes in tax revenues on changes in the GDP of Kuwait is insignificant. Therefore, Kuwait's government could rationally implement tax reforms to have incremental sources of income other than oil revenue. Moreover, it is argued that the government might consider implementing broad based consumption taxes and value added taxes into the tax structure Kuwait, and to invest the revenues from those taxes in productive policies, to induce long term economic growth.

Keywords: GLS Methods, ADF Tests Stationarity, Granger Causality, Elasticity of Tax Base.

Мета цього дослідження визначити, чи життєздатні податкові реформи в Кувейті, щоб отримати більший дохід уряду з інших джерел, окрім нафти. Розглянуто взаємозв'язок між змінами в податкових надходженнях, змінами в нафтових доходах і змінами ВВП у Кувейті на основі даних часових рядів з 1998 по 2015 рік. Розширений тест Дікі-Фуллера (ADF) використано для перевірки існування оди-ничного кореня. Тест коінтеграції застосовано для перевірки довгострокових співвідношень між змінними з використанням методу найменших квадратів (GLS). Результати дослідження показали, що вплив змін податкових надходжень на зміну ВВП Кувейту незначне. Таким чином, уряд Кувейту може виважено проводити податкові реформи, щоб мати додаткові джерела доходу, окрім доходів від на-фти. Стверджується, що уряд міг би розглянути питання про введення широкомасштабних податків на споживання і додану вартість до податкової структури Кувейту та інвестувати доходи від цих податків у виробничу політику, щоб стимулювати довгострокове економічне зростання.

Ключові слова: метод найменших квадратів GLS, розширений тест Дікі-Фуллера, причинність Грейнджера, еластичність податко-вої бази.

Цель этого исследования в том, чтобы определить, являются ли налоговые реформы жизнеспособными в Кувейте, чтобы полу-чить больший доход правительства из других источников, кроме нефти. Рассматрена взаимосвязь между изменениями в налоговых поступлениях, изменениями в нефтяных доходах и изменениями ВВП в Кувейте на основе данных временных рядов с 1998 по 2015 год. Расширенный тест Дики-Фуллера (ADF) использован для проверки существования единичного корня. Тест коинтеграции применен для проверки долгосрочных соотношений между переменными с использованием метода наименьших квадратов (GLS). Результаты иссле-дования показали, что влияние изменений налоговых поступлений на изменение ВВП Кувейта незначительно. Таким образом, правите-льство Кувейта может взвешенно проводить налоговые реформы, чтобы иметь дополнительные источники дохода, помимо доходов от нефти. Утверждается, что правительство могло бы рассмотреть вопрос о введении широкомасштабных налогов на потребление и добавленную стоимость в налоговую структуру Кувейта и инвестировать доходы от этих налогов в производственную политику, чтобы стимулировать долгосрочный экономический рост.

Ключевые слова: метод наименьших квадратов GLS, расширенный тест Дики-Фуллера, причинность Грейнджера, эластичность налоговой базы.

Introduction

This research study takes Kuwait, as an oil dependent GCC member, as a surrogate for studying the need for structural reforms in the country itself, and in the GCC by choosing specifically tax reforms as a focus. The objective of the study is to assess the viability of reforming the tax structure in Kuwait, and the introduction of new taxes to diversify sources of income and induce long term economic growth.

Kuwait has a limited scope for tax revenues. The government implements taxes on foreign owned corporations only. However, most of the private sector is held by Kuwaiti individuals exclusively. Moreover, since only the foreign corporations are within the purview of Corporate Income tax law, is makes the generated revenues very limited. Additionally, the corporate income tax is not levied on the income of companies owned by GCC nationals in the same way it does foreigners, further limiting the tax base. However, Kuwait neither implements Value Added Taxes (VAT), income tax nor sales taxes. This tax friendly environment is a common among the members of the G.C.C. This type of governance derives its legitimacy through tradition, religion and more importantly the mobilization of oil wealth to increase the citizens' living standards. Consequently, it is very difficult to broach the subject of taxation, and the issue usually faces major political opposition [4].

Methodology

The empirical model used in the study is a multiple regression model based on the Generalized Least Square method of estimation (GLS). It utilizes ANOVA to assess the significance of the changes in tax revenues that are generated to the changes in GDP. Several other tests are conducted. Hence the Augmented Dickey and Fuller test is used to test if the variables are stationary at the same level. The Johansen test is used to isolate long run relationships between variables over time and finally the Granger causality test determines the causal relationships between the variables. Moreover, problems in the data of multicollinearity, heteroscedasticity [18] and serial correlation are tested and corrected. The software used to conduct the analysis of this econometric l model is the STATA-IC, where all tabulated results use this platform. All the macro economic data utilized in this GLS estimation is retrievable at Kuwait Ministry of Finance, which can be accessed on the World Data Atlas, Central Statistical Bureau Kuwait. However, if taxes are used for productive purposes, such as a public goods that is an input in production or for public welfare (e.g. public infrastructure, cleaner environment ...etc.) there may be an optimal level of taxation. It is feasible therefore, that a country with higher tax environment could have higher GDP per worker than a country with a lower tax incidence [5].

Eltony N. [14], implements a regression analysis using OLS methods using cross country, time series data for 16 Arab countries that included the GCC countries to determine the significance of taxation towards revenues yield in these countries. He argued that in the GCC, there was a significant gap between actual and potential taxes especially in the UAE, Kuwait, Qatar and Saudi Arabia. He also found, that the political system, attitudes toward government and the quality of the tax regimes were also very important determinants of the tax share in GDP, and therefore these aspects should be considered by governments. He concluded that in order to reach maximum tax revenue generating capacity these countries should implement relevant tax reforms to current systems.

The G.C.C consists of Saudi Arabia, Bahrain, United Arab Emirates, Qatar, Oman and Kuwait. The political and economic issues are similar in nature in these countries and may cause positive or negative spillover effects from one country to another, making it easier to implement reforms if neighboring countries do the same, or more difficult if they fail. The Kuwaiti economy is similar to that of other GCC members in its dependence on petroleum exports, as 95 % of its export revenue comes from oil [36]. Kuwait considered a high-Income country, having a GDP per capita of about 70,500 USD as of 2015, falling behind only Qatar, which has the highest GDP per capita in the world (IMF, 2016). Like other members of the GCC, the government is considered obligated to provide education, healthcare and most basic services for the citizens, as a result of the social contract between them. This has caused the governments to subsidize most public goods. The government currently owns the water and electricity services, while food, education and healthcare are available as public and private entities. Most citizens rely on public services. The current subsidy system ensures that education, health services, basic food items, electricity and water are all sold below market prices. [13].

In Kuwait, most taxes that apply are for services to be received, such as stamp duties, fees for documentation renewal and fees for licenses and property transfers. Currently Kuwait does not implement personal income taxes on citizens. corporate profit tax on Kuwaiti nationals and GCC national owned companies or, Value Added Tax (VAT). Zakat on Kuwaiti companies is set at 1 %, but this is not considered part of government tax revenues. Corporate taxes are implemented at 15 % on foreign owned companies [28]. Other revenue sources under tax revenues are the fees government charges on the provision of the public goods and services [22].

Arnold J. [2] argues that corporate tax is the least growth friendly tax, in comparison to consumption and property taxes. Therefore, this is considered as one of the least attractive taxes. [Arnold ibid] also, finds that a broadly based consumption taxes would be superior and would cause the least distortions in comparison with taxes that have many exemptions [2]. Moreover, this study found that only when corporate tax rates are lower than income tax rates, is there encouragement to owning a business rather than being an employee. Hence, in this scenario, agents would be encouraged to take risks and invest in private businesses, increasing corporate tax revenues as result.

Kuwait does not levy individual citizen income taxes and therefore corporate taxes are expected to discourage investment in this case, as agents prefer to be employed rather than risking owning a business [29]. A detailed inspection of Table 1 summarizes the impact of ten studies which have aimed at estimating the impact of taxation on economic growth. As may be seen for some studies: for example, Hassan B. [17] indicates a strong positive impact of VAT revenue on economic growth. Moreover, Easterly W. [12] find no significant correlation exists between taxes and economic growth in a large sample. Furthermore, most studies summarized in Table 1 used econometric methods/ OLS/ GLS/panels to estimate the significance of various tax regimes and we adopt similar procedures except we that deliver cointegration and other enhancements/procedures to encompass a similar path of research.

Table 1. Empirical Studies on the effect of Taxes on Growth

Source: Authorial Compilation and Tabulation.

Econometric Model: ADF Tests and Stationarity Profiles

We applied the ADF test to all the 7 variables and difference the non-stationary variables until we found their order of stationarity [8, 9, 10]. T able 2 shows the results. Our inferences were on the critical values under a 95 % confidence interval, aka value of a=5 % for all the tests conducted in the full model in order to maintain consistency. We applied these tests to the Gross Domestic Product (GDP) variable. After the second difference, or at I (2) order of integration the data becomes stationary with a calculated value of the test statistic -4.765 and p-value of 0.0001 as the shown in table. We proceeded to implement this test using the same method on Oil Revenue (OR) data. We found that it was also stationary after second differences when p-value is at 0.0060 and the test statistic is -3.586. For the variable of Taxes on Corporate Income and Profits (TCIP), we found stationarity after the second differences with a p-value of

0002 and the test statistic at -4.531. This made both of those variables of I (2) order of integration. The fourth variable, Taxes on Property Transfers (TPT), was found to be non-stationary therefore we took the first difference and found that it was stationary at p-value 0.0017, which made its order of integration I (1). This causes the variable (TPT) to be at a lower order of integration than the GDP, OR and TCIP variables [25, 21] Therefore, we dropped this from the next step of the econometric model since it meant that it did not exhibit a long term trend with the fundamental variables in the model GDP and OR. The variable Taxes on International T rade (TIT) was tested and found to be stationary after taking the first difference, where the calculated test statistic is more negative than the critical value at -3.740 and with a p-value of 0.0036 whereas the variable Stamp Duties (SD) was tested and found to be stationary after the first difference as well at p-value of 0.000. tax revenue cointegration

Results

Therefore, these two variables were omitted from (the empirical model along with the (TPT) variable. Finally, (AT) was found stationary after taking second differences. So, we kept the 4 variables integrated at order I (2). The reason behind this is that GDP is the dependent variable and hence, we include only the variables integrated at the same level in order to avoid spurious regression results [3]. The multiple regression equation therefore includes the A Gross Domestic Product (GDP), A Oil Revenue (OR), A Taxes on Corporate Income and Profits (TCIP) and A Activities Taxes (AT) as follows:

Д GDP (t) = Я0 + ЯiAOR (t) + Я2ATCIP (t) + Я3AAT (t) + s(t).

Table 2. The Augmented Dickey and Fuller Unit Root Test Results

*The Calculated value of the test statistic on the raw data set without taking any differences.

** Variables GDP, OR, TCIP and AT are stationary at the same level and will be included in the model. The rejection of the null hypothesis of unit root is compared at 5 % significant level for all variables.

Source: Authorial Compilation and STAT-IC Output.

Table 3 shows the results of the application of the test. We analyzed the results starting with the Trace statistic against the Critical Value at the maximum rank of 0. The Trace Statistic of 118.3966 is bigger than the Critical value of 47.21. Therefore, we rejected the null hypothesis of no co-integration and concluded that there was at least one co-integrating relationship between variables. We proceeded to compare the Trace Statistic to the Critical value at rank 1, and we concluded the same since 55.5348 is bigger than 29.86. At rank 2 and 3 we find that the Trace Statistic of 25.9286 and 10.3100 are bigger than the 15.41 and 3.76 Critical Values respectively. Thus, AGDP, AOR, ATCIP and AAT are cointegrated, and there is a long-run relationship among them [19, 20].

Table 3. The Johansen co-integration test results

Source: Authorial Compilation and STATA-IC Output.

Table 4 shows the results of the Granger-causality test. The number of lags used in the causality test is 2 lags. The appropriate number of lags for the causality test was determined by the Akaike Information Criteria (AIC) in order to select the maximum lag length. The first variable to be tested was the AGDP. We observed each variable against

AGDP and found if each of them Granger-causes it then AGDP would be chosen as the Y (dependent) variable, and all others would be considered the X (independent) variables. For the first row, the null hypothesis is that AOil Revenue (OR) does not Granger-cause AGDP, and the p-value of 0.000 is less than 0.05 therefore we rejected this hypothesis and concluded that AOR does Granger- causeAGDP, therefore there is a causal relationship between them. The results also show a bilateral relationship between the two variables, which means AOR Granger- causes AGDP and AGDP also Granger-causes AOR. The next variable to be tested against AGDP was the ATCIP. From the p-value we find that ATCIP Granger-causes AGDP at 0.002, but the AGDP does not Granger-cause ATCIP since the p-value of 0.161 is bigger than 0.05. Testing AGDP against ATCIP we found that it Granger-causes it while the inverse is incorrect. At p-value 0.186, the AAT does not Granger-cause AGDP, while at p-value of 0.033 AGDP causes AAT. Finally, we observed if all variables together Granger-cause AGDP and concluded that they do from the

000 p-value. Therefore; the choice of AGDP as the dependent variable is consistent with the results [15].

Table 4. Granger causality test results

*Based on lag 2 which is determined by the Akaike Information Criteria (AIC).

Source: Authorial Compilation and STATA-IC Output.

Assessment using the Variance Inflation Factor (VIF) which, which quantified the severity of multicollinearity. As a rule of thumb, VIF less than 10 is acceptable for the model to conclude that there is no multicollinearity problem [31]. Table 5 contains the results of the VIF test. From the results we see that all the three independent variables do not have a VIF equal to or, above 10. At 1.08, 1.08 and 1.01 we concluded that there is no multicollinearity between the variables [6].

Source: Authorial Compilation STATA-IC Output

By observing the F-value of 21.76 and comparing it to the tabulated F-value of 3.49 at 5 % significance level, we find that it is higher. Therefore, we reject the F-test's null hypothesis that the model is insignificant. We also examined the P-value. At 0.0000 which was less than 0.05 and we rejected the null of an insignificant model. Therefore, we concluded that full model was statistically significant and based conclusions on this finding.

Table 5. Multicollinearity test (Variance Inflation Factor)

Conclusion and Discussion

The [Prais-Winsten, 1954] regression was used to overcome serial correlation that appeared in regression models. It uses the Generalized Least Square (GLS) to estimate the parameters of the model that has serially correlated errors. (Table 6) The GLS method gives more efficient estimations than OLS in this case [16]. [Prais- Winsten Ibid] also prevents the loss of the first observation in the model. Therefore, it increases the efficiency of the test and is more appropriate for smaller a sample model [Maeshiro A. [27], Park, 1980]. The test produced a transformed Durbin Watson value of 2.01 which when applied on our data, and the serial correlation was solved. From the observed value of R-Squared which is 8447, we inferred that the independent variables in the model (AOR, ATCIP and AAT) explain 84.47 % of the variability in the AGDP. Therefore, a significant amount of the changes that appear in the GDP of Kuwait is explained by our independent variables and only 15.53 % is left and is unexplained. Therefore, this is considered part of the error term and is unexplained by the independent variables chosen. Thus, from these results, we conclude that the general model results are statistically significant.

We observed the effect of AOR , which had a t-value of and a P-value of 0.000. This effect is significant on AGDP at our 5 % significance level. The value of its beta coefficient of 1.25 indicates a positive relation. The confidence interval also indicated that the rate of change of the conditional mean of AGDP with respect to AOR was estimated to be between 0.863 and 1.639. This confirms the dependence on oil for the GDP of Kuwait [16].

To attain the core objective, we studied the effects of the changes in tax revenues in Kuwait on changes in Gross Domestic Product (GDP) and measured their relative statistical significance. The econometric model used was a multiple regression model based on the Generalized Least Square (GLS) method of estimation. The data utilized was time series data, collected for 18 years starting from 1998 to 2015, on Gross Domestic Product (GDP), Oil Revenues (OR), Taxes on Corporate Income and Profits (TCIP) and Activities Tax (AT). The data sources were from Kuwait's statistical bureau and the World Bank.

Table 6. Regression Model (Prais-Winsten)

The results indicate that the impact of changes in tax revenues do not affect changes in the GDP of Kuwait significantly. The fundamental issue was that the only significant factor affecting changes in long run GDP growth were changes in oil revenues. This was due to the country's dependence on oil for the generation of GDP. Therefore, the government could viably utilize tax reform policies and introduce new taxes to generate income from sources other than oil. VAT, foreign labor taxes and broad-based consumption taxes are recommended for being growth friendly in comparison with corporate income and personal income tax as per the literature reviewed.

References

1. Adekunle A.R. Evaluating the Effects of VAT on Tax Revenue in Nigeria / A.R. Adekunle, A.R. Remi, O. Ayorride // Pakistan Jour. of Social Sciences, 2013. - P. 22-26.

2. Arnold J. Do Tax Structures Affect Aggregate Economic Growth? // OECD Economics Department Working Papers, 2008. P. 28-30.

3. Asteriou D. Applied Econometrics: A Modern Approach Using Eviews and Microfit Revisited Edition ? / D. Asteriou, S.G. Hall // Palgrave Macmillan, 2007. - New York. - P. 220-256.

4. Auty, R. M. Political economy of resource-abundant states / R.M. Auty, A.H. Gelb, 2001. - P. 126-144.

5. Barro R.J. Government spending in a simple model of endogeneous growth / R.J. Barro // Jour. of political economy, 1990. - Vol. 98. (Section 5, Part 2). - P. 103-125.

6. Breusch T.S. A Simple Test for Heteroscedasticity and Random Coefficient Variation / T.S. Breusch & A.R. Pagan // Econometrica, 1979. - Vol. 47, No.5. - P. 1287-1294.

7. De Mooij R. Fiscal Devaluation" and Fiscal Consolidation: The VAT in Troubled Times / de Mooij R. and M. Keen // Internat. Monetary Fund, 2012. - No. WP/12/85.

8. Dickey D. A. Distribution of the Estimators for Autoregressive Time Series with a Unit Root / D.A. Dickey & W.A. Fuller // Jour. of the American Statistical Association, 1979. - P. 427-431.

9. Dickey D. A. Likelihood Ratio statistics for Autoregressive Time Series with a Unit Root, Econometrica / D.A. Dickey & W.A. Fuller // Jour. of the Econometric Society, 1981. - P. 1057-1072.

10. Dickey D. A. Determining the Order of Differencing in Autoregressive processes / D.A. Dickey & S.G. Pantula // Jour. of Business & Economic Statistics, 1987. - Vol. 5, No.4. - P. 455-461.

11. Durbin J. Testing for Serial Correlation in Least Squares Regression II / J. Durbin & G.S. Watson // Biometrika, 1951. - Vol. 38. - P. 159-177.

12. Easterly W. Marginal Income Tax Rates and Economic Growth in Developing Countries / W. Easterly & S. Rebelo // European Economic Review, 1993. - Vol. 37, issue 2-3. - P. 409-417.

13. El-Katiri L. Anatomy of an oil-based welfare state: Rent distribution in Kuwait. Kuwait Programme on Development. Governance and Globalization in the Gulf States research / L. El-Katiri, B. Fattouh, P. Segal // The London School of Economics and Political Science, 2011. - P. 13.

14. Eltony N. The Determinants of Tax effort in Arab countries / N. Eltony. - Arab Planning Institute : Kuwait Information Center, 2002.

15. Granger C.W. Investigating Causal Relations by Econometric Models and Cross-Spectral Methods / C.W. Granger // Econometrica: Jour. of the Econometric Society, 1969. - P. 424-438. DOI: 10.2307/1912791.

16. Hansen C.B. Generalized Least Squares Inference in Panel and Multilevel Models with Serial Correlation and Fixed Effects? C.B. Hansen // Jour. of econometrics, 2007. - Vol. 140(2). - P. 670-694.

17. Hassan B. The Role of Value Added Tax (VAT) in the Economic Growth of Pakistan / B. Hassan // Jour. of Economics and Sustainable Development, 2015. - Vol. 6, No 13. - P. 174-183.

18. Hayes A.F. Using heteroskedasticity-consistent standard error estimators in OLS regression: An introduction and software implementation / A.F. Hayes, L. Cai // Behavior research methods, 2007. - Vol. 39(4). - P.709-722.

19. Johansen S. Statistical analysis of cointegration vectors / S. Johansen // Jour. of economic dynamics and control, 1988. - Vol. 12(2-3). P.231-254.

20. Johansen S. Estimation and hypothesis testing of cointegration vectors in Gaussian vector autoregressive models / S. Johansen // Econometrica: Jour. of the Econometric Society, 1991. - P. 1551-1580.

21. Kumari S.S. Multicollinearity: Estimation and elimination ? / S.S. Kumari // Jour. of Contemporary research in Management, 2008. - Vol. 3(1). - P. 87-95.

22. Kuwait Al-Youm Magazine, Law No. 2 of 2008 on amending some provisions of Kuwait Income Tax Decree No. 3 of 1955, 2008. Kuwait Al-Youm Magazine, No. 856.

23. Kuwait's Ministry of Finance (2016). The country's financial managements' final account for the year 2015/2016.

24. Kuwait Ministry of Finance. General Accounting Affairs. Guidance and System Department, 2016. Government Budget Report 2015/2016. - Kuwait. - P. 3-40.

25. Testing the null hypothesis of stationarity against the alternative of a unit root: How sure are we that economic time series have a unit root? /

26. Kwiatkowski, P.C. Phillips, P. Schmidt, Y. Shin // Jour. of econometrics, 1992. - Vol. 54(1-3). - P. 159-178.

27. Looney R.E. Oil Revenues and Dutch Disease in Saudi Arabia: Differential Impacts on Sectoral Growth / R.E. Looney // Canadian Jour. of Development Studies/Revue Canadienne d'йtudes du Dйvelopement, 1990.

28. Maeshiro A. On the retention of the first observations in serial correlation adjustment of regression models / A. Maeshiro // International Economic Review, 1979. - P. 259-265.

29. Mehra Y.P. Oil Prices and Consumer Spending / Y.P. Mehra, J.D. Peterson // Federal Reserve Bank of Richmond Economic Quarterly, 2005. - Vol. 91. - P. 53-72.

30. Metwally M. Determinants of Aggregate Import s in The GCC Countries // Applied Econometrics and International Development. AEDI, 2004. - Vol. 4-3.

31. Njogu L. K. The effect of value added tax on economic growth in Kenya / L.K. Njogu // Internat. Academic Jour. of Economic and Finance, 2015. - Vol. 1(5). - P. 10-30.

32. O'Brien. A caution reading rules of thumb for variance inflation factors / O'Brien // Quality and Quantity, 2007. - Vol. 4(19). - P. 49-56.

33. Okafor R.G. Tax revenue generation and Nigerian economic development / R.G. Okafor // European Jour. of business and management, 2012. - Vol. 4(19). - P. 49-56.

34. Onwuchekwa J.C. Value added tax and economic growth in Nigeria / J.C. Onwuchekwa, S.A. Aruwa // European Jour. of Accounting Auditing and Finance Research, 2014. - Vol. 2(8). - P. 62-69.

35. Prais S.J. Trend estimators and serial correlation / S.J. Prais, C.B. Winsten. - Chicago : Cowles Commission discussion paper, 1954. - Vol. 383. - P. 1-26.

36. Shinohara M. Tax Structure and Economic Growth / M. Shinohara, 2014. Retrieved

37. Squalli J. Electricity consumption and economic growth: Bounds and causality analyses of OPEC members / J. Squalli // Energy Economics, 2007.

38. Wildman J. The impact of income inequality on individual and societal health: absolute income, relative income and statistical artefacts / J. Wildman // Health Economics, 2001. - Vol. 10(4). - P. 357-361.

Размещено на Allbest.ru