DATABASE AND RESEARCH METHODOLOGY

CHAPTER III DATABASE AND RESEARCH METHODOLOGY The nature of the present study Direct Tax Reforms in India: A Comparative Study of Pre and Post-liberalization periods is such that it requires secondary data. In this chapter, database and research methodology of the study has been discussed. First section of this chapter presents the database of the study and second section explains the various statistical and econometric techniques used in the study. Section-I 3.1: DATABASE OF THE STUDY The study is mainly based on secondary data covering a time period of forty-two years from 1970-71 to 2011-12. However, different sources of data have been consulted to prepare a continuous set of data as per the requirements of the study. The followings are the main sources of data consulted to conduct this study: Reserve Bank of India Publications Two main secondary data sources published by Reserve Bank of India were heavily consulted and relied upon for the present study. These sources are, namely: A Hand Book of Statistics on Indian Economy, different issues; and Report on Currency and Finance, different issues. Data regarding Gross Domestic Product (GDP), Expenditure, Deficits, Governement Liabities etc. Has been taken from these reports. Government of India Publications Various publications by Ministry of Finance, Government of India, and other government agencies were also used for the present study. These sources include: India Public Finance Statistics, Ministry of Finance, various issues; Economic Surveys, various Issues; Annual Union Budget Speeches of Finance Ministers of different years; All India Income Tax Statistics, Directorate of Income Tax, various issues; Income Tax 64

Department Performance Statistics, Directorate of Income Tax, various issues; Compliance report on Direct Taxes, Comptroller and Auditor General of India, various issues; Income Tax Act, 1961 as amended from time to time; and various reports of different Finance Commissions constituted from time to time. Data regarding various components of taxes were collected and used from these sources. Reports of various Committees appointed by Government of India The following reports of various committees appointed by Government of India from time to time have also been considered to analyse the policy changes on direct taxes and administrative reforms introduced during the post-independence period. These reports include: The India Taxation Enquiry Committee, 1924-25; Income Tax Enquiry Committee, 1935-36; Income Tax Investigation Commission, 1948; Taxation Enquiry Commission, 1953-54; Indian Tax Reforms- A Survey, 1956; Rationalization and Simplification of the Tax Structure, 1968; Direct Tax Enquiry Committee Report, 1971; Taxation of Agricultural Income and Wealth, 1972; Direct Tax Laws Committee, 1978; Tax Measures to Promote Employment, 1980; Long Term Fiscal Policy, 1985; Interim Report of Tax Reforms Committee, 1991; Final Report (Part I & II) of Tax Reforms Committee, 1992, 1993; Advisory Group on Tax Policy and Tax Administration for the Tenth Plan; Report of the Task Forces on Direct Taxes, 2002; The Fiscal Responsibility and Budget Management (FRBM) Act, 2003; The Fiscal Responsibility and Budget Management Rules, 2004; and The Direct Tax Code Bill, 2010. These policy documents have been consulted to achieve the objective of analyzing policy reforms in India. World Bank Publications To compare the performance of India in regard to generation of revenue from taxes, a comparison of Indian tax performance with some developed and developing countries has also been made. For this, data have been collected from international publications like World Development Report, various issues; World Development Indicators, various issues; Tax Policy in Developing Countries; Tax Reforms in Developing Countries; the publications of OECD countries; and European Commission Report on Tax Trends, 2012. However, due to different methodologies used in collection 65

of data, some differences were found in various figures and those are duly accounted for at appropriate places. Other Publications Following data sources were also referred for the compilation of the data for the present study. Such as: Basic Statistics relating to Indian Economy, various issues, Center for Monitoring of Indian Economy (CMIE); Public Finance in India, various issues; publications of India Tax Foundation; Taxmann s Direct Tax manual; and Taxmann s Direct Tax Ready Reckner. Keeping in view the data availability, efforts have been made to collect data for the period from 1970-71 to 2011-12. However, due to gaps in the data, some analysis has been restricted for short period as per the availability of data. In case of variations in the figures, the Comptroller and Auditor General of India s publication- Compliance Reports on Direct Taxes have been relied upon for figures relating to the direct taxes. Furthermore, Ministry of Finance Publication- Indian Public Finance Statistics was relied upon for indirect taxes; and, Reserve Bank of India Publication-A Handbook of Database on Indian Economy was relied upon for data relating to other variables. The direct tax reforms undertaken since 1991 and their effects have been examined in detail. The study has focused mainly on two major direct taxes, namely, personal income tax and corporation tax. Reference of other taxes like wealth tax, gift tax etc. has also been made wherever it was required. Section-II 3.2: RESEARCH METHODOLOGY The following variables have been identified to achieve the objectives of the study. It includes: Gross Domestic Product (GDP); Gross fiscal Deficit (GFD); Revenue Receipts (RR); 66

Direct & Indirect Taxes; Corporate & Personal Income Taxes; Capital & Revenue Expenditure; and Liabilities (both external and internal) of the Governments (the Centre & the States). The above variables have been chosen on the basis of review of vast literature available on the similar studies conducted by other researchers. However, due care has been given to consider the relevant variables for the present study. The following aspects have been studied in the present study, which include: Tax-GDP ratio; Buoyancy of various taxes; Annual Growth rates of different variables; Performance of Indian Tax system in regard to revenue generation and its effects on different macro-economic parameters under study; Profile of corporate and personal income tax assessees; and Arrears of taxes. Hypothesis Tested To achieve the objectives of the study, the following hypothesis has been framed and tested: 1. H 0 : The direct tax reforms have not led to higher Tax-GDP ratio and tax buoyancy during the post-reforms period. H 1 : The direct tax reforms have led to higher Tax-GDP ratio and tax buoyancy during the post-reforms period. 2. H 0 : The direct tax reform measures have not succeeded to reduce fiscal deficit during the post-reforms period. 67

H 1: The direct tax reform measures have succeeded to reduce fiscal deficit during the post-reforms period. 3. H 0 : The direct tax reforms have not affected uniformly to all the stakeholders during the post-reforms period. H 1 : The direct tax reforms have affected uniformly to all the stakeholders during the post-reforms period. 4. H 0 : The direct tax reforms have not led to increase in tax base vis-à-vis tax revenues proportionately during the post-reforms period. H 1 : The direct tax reforms have led to increase in tax base vis-à-vis tax revenues proportionately during the post-reforms period. 5. H 0 : The direct tax reforms have not succeeded to generate more revenue from the upper section of the society during the post-reforms period. H 1 : The direct tax reforms have succeeded to generate more revenue from the upper section of the society during the post-reforms period. 6. H 0 : The direct tax reforms have not led to higher tax compliance during the postreforms period. H 1 : The direct tax reforms have led to higher tax compliance during the postreforms period. 7. H 0 : The direct tax reforms have not led to higher level of revenue generation during the post-reforms period. H 1 : The direct tax reforms have led to higher level of revenue generation during the post-reforms period. 68

8. H 0 : The direct tax reforms have not led to higher level of direct tax revenue generation during the post-reforms period. H 1 : The direct tax reforms have led to higher level of direct tax revenue generation during the post-reforms period. Statistical and Econometric Tools used The data collected for the purpose of the study have been analysed using different statistical and econometric tools, which are as follows: Tabular analysis Ratios and percentages Averages Standard Deviation Coefficient of Variation Compound Annual Growth Rates (CAGR) Simple regression Multiple Regression Analysis [Distributed Lag AR(p) Model] Test of stationarity using Augmented Dickey Fuller (ADF) Unit Root Test Histogram-Normality test for residuals using Jarque-Bera Test Statistic Breusch-Godfrey Serial Correlation LM Test for residuals Test for Multicolinearity using Tolerance Level and Variation Index Factor (VIF) scores. Compound Annual Growth Rates (CAGR) To evaluate the performance of tax reforms, compound annual growth rates have been computed. To estimate the compound growth rates of various vaiables, the following model has been estimated: Y t = ab t e u t -----(1) 69

Taking log on both sides Where Y = Value of the variable log Y = α + t log β + t ----(2) t = α = β = Time variable Intercept term Slope coefficient = Stochastic disturbance term The compound growth rate r is worked out with the following formula: r = (antilog -1) * 100 ----(3) where is the ordinary least square estimator of β in model (2). Simple Regression Analysis to calculate Buoyancy of Taxes To calculate buoyancy of tax revenue during the pre and post-reforms period, the following regression function has been used by regressing log of tax revenue on the log of the base (GDP at market price): Log T t = α + β Log Y t Where T t = tax revenue (including discretionary changes) of each type of tax (individually) for the period t α = Constant β = Buoyancy coefficient Y t = GDP at current price for period t 70

If b < 1, tax revenue is considered to be less buoyant If b > 1, tax revenue is considered to be more buoyant If b = 1, tax revenue is considered to be equally or proportionately buoyant Multiple Regression Analysis To test the relative effectiveness of direct tax policies in India over the study period, the following reduced form equation has been estimated: d(log Y t ) = βo + β 1 d(log X 1t ) + β 2 d(log X 2t ) + β 3 d(log X 3t )+ β 4 d(log Y t-1 ) + β 5 d(log(x 1t- 1) + β 6 d(log X 2t-1 ) + β 7 d(log(x 3t-1 ) + β 8 d(log(y t-2 ) + β 9 d(log(x 1t-2 ) + β 10 d(log(x 2t-2 ) +β 11 d(log(x 3t-2 ) + D + Where: d(log Y t ) : Estimated value of the dependent variable at time t d(log X 1t X 2t, Xn t ) : Values of indipendent variables at time t α : Intercept β 1, β 2, βn : Values of coefficients d(log Y t-p ) : Dependent variables at (t-p) years lag (differenced) d(log X nt-p ) : Independent variables at (t-p) years lag (differenced) D : Dummy variable : Residuals It has been noted that all time-series data of the above variables have strong trend (increasing), therefore the equation has been estimated in the first difference form. In first difference case, yearly change from period (t-1) to period (t) is labeled as the change at period (t). To test the structural shifts, the period of the study has been divided into two 71

parts representing the pre-reform (ie, 1970-71 to 1990-91) and post-reform (i.e., 1991-92 to 2011-12) periods and a dummy variable has been introduced for the same. Test of Stationarity Before knowing causality between variables, it requires that all the variables to be stationary. A stochastic process Y t is said to be stationary if i. E (Y t ) = μ ii. E(Y t μ) 2 = σ 2 iii. E[(Y t μ)(y t+k μ)] = Y k The first two conditions require the process to have constant mean and variance and third condition requires the value of covariance between two time periods depends only on the distance between two time periods and not on the actual time at which the covariance is computed. Thus, first of all the time series data were analyzed for various time series properties, for instance, unit roots. Augmented Dickey-Fuller (ADF) Unit Root Test A test of stationarity (or nonstationarity) that has been popular over the past several years is the unit root test. For testing the unit roots in the data Augmented Dickey-Fuller (ADF) unit-root test has been applied. In the ADF test, lag order was determined using Schwarz criterion (BIG). This unit root test has been applied with intercept. This procedure allows for the error term to be correlated. To take care of that, it includes longer lags of dependent variable into the regression. This test estimates following three regressions, I. ΔY t = ΔY t-1 + α i t-1 +Et ------- random walk II. ΔY t = β 1 + ΔY t-1 + α i t-1 +Et ------- random walk with drift III. ΔY t = β 1 + β 2 t + ΔY t-1 + α i t-1 +Et ------- random walk with drift and trend 72

If = 0 : Series is non stationarity or unit roots are present If < 0 : Series is stationarity or unit roots are not present In all the three equations, the null hypothesis, i.e. =0, is same. But, the alternative hypothesis for equation I is, Y t is zero mean stationarity. For II, Y t is stationarity. For third equation, Y t is trend stationarity. However, the signifinance of can be tested by employing t-test. t = where is the coefficient of Y t-1 and SE() is the standard error of. Jarque-Bera Test Statistic The JB test of normality is a large-sample test. It is based on the residuals of ordinary least square (OLS). This test first computes the skewness and kurtosis measures of the OLS residuals and then uses the following test statistic to test the normality of residuals: where n is the number of observations and k is the number of regressors when examining residuals to an equation, S = skewness coefficient, and K = kurtosis coefficient. For a normally distributed variable, the vale of S = 0 and that of K = 3. Therefore, the Jarque-Bera (JB) test of normality is a test of the joint hypothesis that S and K are 0 and 3, respectively. In that case the value of the Jarque-Bera (JB) statistic is expected to be 0. 73

Breusch-Godfrey Serial Correlation LM Test Presence of autocorrelation is another very common problem which is generally faced in time-series data. The word auto means self. It shows the correlation of a variable and its lags. This property of time-series is quite important to regression modelling as it is closely related with the problem of multicolinearity. A test of autocorrelation (or serial correlation) was developed by the statisticians Breusch and Godfrey that is general in the sense that it allows for (1) non-stochastic regressors, such as the lagged values of the regressand; (2) higher-order autoregressive schemes, such as AR(1), AR(2), etc.; and (3) simple or higher-order moving averages of white noise error terms, such as t. Without going into the mathematical details, the BG test, which is also known as the LM test, can be used by adding many regressors to the model along with the lagged values of the regressand. The BG test involves the following steps: Consider a linear regression of any form, for example where the residuals might follow an AR(p) autoregressive scheme, as follows: The simple regression model is first fitted by ordinary least squares to obtain a set of sample residuals. Breusch and Godfrey proved that, if the following auxiliary regression model is fitted 74

and if the usual statistic is calculated for this model, then the following asymptotic approximation can be used for the distribution of the test statistic when the null hypothesis holds (that is, there is no serial correlation of any order up to p). Here n is the number of data-points available for the second regression, that for, where T is the number of observations in the basic series. Note that the value of n depends on the number of lags of the error term (p). That is, asymptotically, n p times the R 2 value obtained from the auxiliary regression follows the chi-square distribution with p d.f. If in an application, (n p)r 2 exceeds the critical chi-square value at the chosen level of significance, we reject the null hypothesis, that there is no serial correlation among the residuals of variables. The serial correlation can also be tested by looking at the significance of F-statistic. If F-statistic is found insignificant, then null hypothesis that there is no serial correlation among the variables. Moreover, if all the coefficients of explanatory variables and their lags are insignificant in BG test, the null hypothesis on no serial correlation is accepted again. Tolerance and Variance Inflation Factor for Collinearity Diagnosis Collinearity is a linear relationship between two explanatory variables. Two variables are perfectly collinear if there is an exact linear relationship between the two. Whereas Multicollinearity refers to a situation in which two or more explanatory variables in a multiple regression model are highly linearly related. The problem of multicollinearity is very severe in time-series data. Presence of multicollinearity in the data may create problem like high R 2 but few or no significant coefficients of independent variables etc. To overcome this problem, two statistical measures have been used, as suggested by researchers (Gujrati, 2008). These include: 75

Variation Inflation Factor The Variation Inflation Factor (VIF) is one of the widely used score to test the level of multicollinearity among the different variables in an ordinary least square regression analysis. It provides a value that can be compared with set standards or rule of thumbs to check the level of multicollinearity among the variables. According to Gujrati (2008), if the VIF scores is 10 or more, then it can be said that the data is suffering from the problem of multicollinearity. To check the presence of multicollinearity in data, VIF scores have been used for the regression analysis. Tolerance Tolerance is a measure of to check the multicollinearity problem in a data-series. If tolerance value is approaching towards zero, it implies that the variable under consideration is forming a linear relationship with other independent variables in the equation. So, such variable should not be included in the regression equation. All variables having linear relationship will have small tolerance value. The measure of tolerance is given as: where is the coefficient of determination of a regression of explanator j on all the other explanators. A tolerance of less than 0.20 or 0.10 and/or a VIF of 5 or 10 and above indicates the presence of multicollinearity in the data (O'Brien 2007). Whereas Gujrati (2008) has given the rule of thumb that if the VIF of a variable exceeds 10, which will happen if R 2 j exceeds 0.90, that variable is said to be highly collinear, and the closer is Tolerance (TOLj) to zero, the greater the degree of collinearity of that variable with the other regressors. On the other hand, the closer TOLj is to 1, the greater the evidence that Xj is not collinear with the other regressors. 76

3.3: LIMITATONS OF THE STUDY Limitations have always been a part of any research work. The present study is also not an exception of this. Some of the limitations of the study have been listed as under: The study is totally based on secondary data only. Therefore, it suffers from all the limitations suffered by a research based on secondary data. There are many governmental and non-governmental agencies in India which collect data relating to public finance and taxes. All of these agencies follow different methodology to collect the data. So the problem of data mis-match was faced many a times during the study. Due to non-availability/non-publication of data, the study period had to be adjusted accordingly for part of analysis. There were huge gaps in data in regard to some of the variables, so the necessary adjustments were required to be done in the data to make it compatible. The topic of the research is too vast to cover all the macroeconomic variables in the study. So, the study remained confined to the direct taxes and its various components. So, the findings of the study may not be generalised to indirect taxes and other sources of revenue of the government. A separate research can be conducted for these variables. Due to the limitations of time and resources, the study has excluded separate analysis of inter-state differences in regard to revenue generation. As the state governments also collect majority of the taxes, a separate research in this area can also be pursued. Due to paucity of time, the possibility of collection of primary data to know the tax-payers perception in regard to the various aspects of direct tax reforms has not been carried out. 77