CHAPTER 3 METHODS AND RESEARCH DATAMETHODS AND RESEARCH DATA
3.3 Research data and its source
3.3.1. Determining appropriate variables and its sources
Before choosing data, this study seeks to select the appropriate variables for designing the research models. This study measures the public finance variable by two subcomponents. First, the ratio between total tax revenue and GDP represents the “TAXgdp” variable. Second, “GEXgdp”
stands for a variable of the rate of total government expenditure to GDP. These two variables denote the public finance in this research and we collect the data of these two variables from the IMF’s database - Government finance statistic (GFS). Following d'Agostino et al. (2016), Acemoglu et al. (2003), Acemoglu et al. (2008), Johansson et al., 2008, and Kneller et al. (1999), as well as a few other previous researchers, we evaluate economic growth using
GDP per capita (measured in constant US dollars using 2010 as the base year).
GDP per capita helps us get more robust results than GDP or average growth rate of GDP per capita. GDP per capita is gross domestic product divided by midyear population. The World Bank considered that GDP is the sum of gross value added by all resident producers in the economy plus any product taxes and minus any subsidies not included in the value of the products. This calculation of GDP is without making deductions for depreciation of fabricated assets or for depletion and degradation of natural resources. Data are in constant 2010 U.S. dollars.
To get the third research objective, a “control of corruption” score obtained from Kaufman et al. (2011) measures the “governance” variable (CCI). This variable measures perceptions of corruption, conventionally defined as the exercise of public power for private gain. The scores are oriented so that higher values correspond to better outcomes, on a scale from – 2.5 to 2.5. A higher index indicates lower corruption or lack of corruption and higher control of corruption. This study collected this data from the World Bank’s database - World Governance Indicators (WGI). Since 2002, this examination has taken place annually; therefore, the data from 1997, 1999, and 2001 in this study were added up and divided to get the average value (See Torgler and Schneider, 2009; Law et al., 2013.) This variable may support taxes system as well as public spending. For robustness check, we continue to extract the corruption perception index (CPI) of business, which was evaluated by Transparency International (TI) from 1996 to 2011, they computed the maximum index is ten, however, from 2012, this CPI was changed the computation method, and the highest index is 100, which represents the area, where corruption is free. Almost developing countries lacked the index in 1996 and 1997. This study assumes that the beginning score of this index is the same score in 1998, so this research chooses the nearest index to fill in this missing value for these two years.
Furthermore, we extract the annual data for whole sample, which includes 38 developed and 44 developing countries over a 21-year period (1996-2016.) (See Appendix A1- List of studied countries).
Due to the reason of that, instability of economies affects economic activities so that we choose inflation annual index (INFL) for saying about economic status. In this research, FDI’s rate to GDP denotes the investment climate (FDI). This study collects this data from the World Bank’s database – World Development Indicators (WDI).
The human development index is a variable that indicates the quality of human capital in a society. We collect the human development index (HDI) from the United Nations Development Program (UNDP) (see table 3.1).
Table 3.1 Variables
Variables ‘meaning
Gross domestic per capita (US. dollars) Total tax revenue (% of GDP)
Total government expenditure (% of GDP) Inflation(Consumer annual Price index)
Inflow of foreign direct investment value (% of GDP) Human development index (index)
Control of corruption indicator
Corruption perception index (CPI) of business
Source: World Bank’s database – WDI and WGI, IMF’s database –
Since the positivism philosophy is suitable for large amounts of numeric data, this study collects the secondary data from the World Bank’s database, International Monetary Fund (IMF)’s database and UNDP’s database. Secondary data is more convenient to use, and is credible because it has been checked by previous researchers, or by carefully using the factors for analysis.
3.3.2 Collecting secondary data
The topic of public finance, governance, and economic growth is a social topic that debates and conducts an experiment to examine the linkage between tax revenue and expenditure, as well as the role of public finance and governance in both developed and developing economies.
Commonly, to resolve a research topic, the researcher should review the theories that provide the core analytical framework or arguments supporting the author’s hypotheses. In addition, a researcher should offer his or her viewpoint to evaluate the change in tax revenue or expenditure when governments change their current policies, or issues any new laws or regulations related to these variables, such as tax revenue and government expenditure.
Second, to convince readers to be interested in the topic with readers’
strong beliefs as well as to avoid bias from subjective ideas, the researcher should collect a large data set from developed and developing countries; then he or she should conduct an experimental study to provide proof of his or her above-mentioned theories.
Creswell (2009) indicated that researchers always begin the scientific method with theory, then collect data that supports or refutes their theory, and make any necessary revisions before adding a test. The data are collected on
an instrument that measures attitudes, and the information is analyzed using statistical procedures and hypothesis testing. The quantitative methos is the best approach to performe the test a theory or explanation, therefore, this research collects the secondary data from three major sources mentioned in 3.3.1.
3. 4 Check balance and essential test
Based on characteristic of large, longitudinal, and cross-sectional panel data, this study has to check balance of data via statistical software due to that imbalanced data is more harmful to general performance. It also poses a big challenge for statistics to ensure the robustness (Wang and Yao, 2012). In addition, to answer the research question 1, and 2, this study needs to verify the stationary of research variables for justification of that Granger test, long- run test and any regression those are appropriate to the dynamic panel data.
Regarding the panel data, this study makes aware of that non-stationary variables may lead to the spurious correlations and presenting the unreliable and invalid empirical results.
This thesis employs the unit root test following Harris-Tzavalis (HT) (1999) test and Im-Pesaran-Shin (IPS) (2003), owing to the reason that Im- Pesaran-Shin (IPS) (2003) relaxes the assumption of a common rho and does not require a strong balanced panel for the less bias from spurious regression as well as co-integration test running. While the Harris-Tzavalis (HT) (1999) test hypothesizes, which assumes that all panels have the same autoregressive parameter and rho is smaller than 1 and the periods are fixed. However, the IPS test does not necessitate balanced data but requires that T must be at least 5 if the dataset is strongly balanced for the asymptotic normal distribution of Z - t-tilde-bar to hold.
3. 5 Choose appropriate analytical methods
Holley (2011) confirmed that short-term tax changes can be different from long-run effects because of the great elasticity of the demand curve. In past decades, some previous researchers have performed a comprehensive analysis of this difference to help policymakers design appropriate fiscal policies. Since it helps avoid the bias given the case of regressions from non- stationary variables, multiple studies employed co-integration test to clear up the problem of spurious regression (e.g. McCoskey and Kao, 1999; Bai and Ng., 2004; Pedroni, 2004; Breitung and Pesaran, 2005; Westerlund and Edgerton, 2008; Persyn and Westerlund, 2008).
First, to answer the first research question, this study conduct co- integration test due to following reasons that the error-correction (EC) model is often applied to investigate the long-run relationship between stationary as well as cointegrated variables (Ojede and Yamarik, 2012). The Westerlund error-correction based panel cointegration tests are very flexible and allow for an almost completely heterogeneous specification of both the long- and short- run parts of the error-correction model, where the latter can be determined from the data. The series are allowed to be of unequal length. If the cross- sectional units are suspected of being correlated, robust critical values can be obtained through bootstrapping.
Assuming that i represents a country and t is time period, the long-run relationship can be represented as below:
, = 0, + , , + , ,
where , is logarithm of GDP per capita (dependent variable), 0, is a country-specific intercept term, , ′ denotes country- characteristic slope coefficients, , indicates the vector of public finance, and , is an error term of country i at time t.
By using public finance, which was measured by the rate of total tax revenue to GDP and the ration between general government expenditure and GDP in the equation (1), this study provides the better evidence of the relationship between public finance and economic growth than previous studies. First, the government collects total tax revenue following the tax base and tax rate. If the government change any tax rate or tax base this decision may change any behavior of the private investors who may lead the economic growth. Second, the government expenditure supports the input of production function such as labor force, infrastructure or safety society. Using public finance helps policymakers better to understand the whole situation of an economy than explore each kind of taxes or each share of the public spending.
In case a co-integration linkage exists between , , and , variable, and error term
, is an I(0) process for all countries i, we can re-write the growth equation in terms of an autoregressive distributed lag (ARDL) of order (p, q) as below:
, = 1, , −1 +2, , −2 + ⋯ + , , − + , , +
1, , −1 + ⋯ + , , − + + , ,
where p is number of lag of dependent variable, and q is number of lag of independent variables.
Then, we re-design the error-correction model as follows:
∆ , = ∑ =1−1 , ∆ , − + ∑ =0−1 , ∆ , − + [ , −1 − 0, −
1, , ] + ,
(1.2) where , and , are short-run coefficients, 0, , and 1,
stand for long-run
coefficients, and and , represent an adjustment-speed (error-correction term) to the long-run equilibrium.
For co-integration test, this study follows Persyn & Westerlund’s (2008) proposed technique, developed by Westerlund (2007). This allows for a complete check of heterogeneous characteristics of long-run parts of error correction model. The null hypothesis is H0: ai = 0 for all i, (i= 1,…N) and H1:
: ai < 0 for all I, (i= 1,…N). The error correction model is as follows:
. = + 1 ∗ . −1+ 2∗ . −2+ ⋯ + ∗ . −
+ 0 ∗ . + 1 ∗ . −1 + ⋯ + ∗ . −
+ ( −1 − ∗ −1) +
This test uses the and Gt test statistics to check the null hypothesis for at least one i. These statistics start from a weighted average of the individually estimated Ga to measure ′s and Gt measuring their t-ratio's, respectively. The test also requires the null hypothesis (H0) be rejected for accumulating evidence of co-integration of at least one of the cross-sectional units. The Pa
and Pt test statistics pool information over all the cross-sectional units to test H0: ai = 0 for all i, (i= 1,…N) and H1: : ai < 0 for all I, (i= 1,…N). Rejection of H0 is thus substantial to validate an existence of co-integration given the entire the panel. There are two other types of panel cointegration test that follow the residual approach such as Kao and Pedroni's test for identification of a long- run link between public finance and economic growth. Kao (1999) investigated the spurious regression for the panel data and presented four different DF type test statistics. Following the serial limitation theory of Phillips & Moon (1999) he attained the asymptotic distributions of his statistics. Pedroni (1999) expanded his panel cointegration testing procedure for the models, where researchers try to reject the null hypothesis of no cointegration panel. The panel-ρ statistic is an extension of the non-parametric Phillips-Perron ρ-statistic, and the parametric panel-t the statistic is an extension of the ADF t-statistic. Gutierrez (2003) demonstrated that the
group-ρ the statistic is the best power among the Pedroni (1999)'s test statistics that we can choose this statistic. The Westerlund (2007)’s is one of the tests that based on the structural rather than residual one. He followed Banerjee, Dolado, and Mestre (1998) test that proposed the panel extensions in the time- series context. These researchers designed the null hypothesis by inferring whether the error correction term in a conditional error correction model is equal to zero. If this test provides the evidence to reject the null hypothesis of no error correction, then the null hypothesis of no cointegration is also rejected. For checking the robustness of co-integration test, this dissertation also conducts the other test that follows residual base such as Kao, Pedroni and Westerlund's test for identification of a long-run link between public finance and economic growth.
By showing the persistence of the relationship between public finance and economic growth, this study becomes a little research, which verified the relationship between public finance and economic growth in the long run.
Almost previous studies checked the relationship between each type of tax rate or tax base or share of expenditure with corruption or economic growth only.
As we may know that, the co-integration test does not affirm the direction of causal linkage; therefore, this study runs the Granger (1969) test to identify the direction of linkage between tax revenue and government expenditure and sets out to identify whether there exists the one way directional linkage or bi-directional linkage between economic growth and public finance. The results of the Granger test could help this study to answer the second research question. Furthermore, Maziarz (2015) confirmed that scholar should apply Granger test when theory on connecting mechanism of two-time series does not exist or it is insufficient. He also conducted this test when he does not consider the prior theoretical knowledge. The results could suggest a way to control deficit in both developed and developing countries
based on the relationship between taxation and expenditure arguments.
Additionally, identifying the direction of the relationship between tax revenue and expenditure also helps this study to enrich the hypothesis of fiscal synchronization.
The null hypothesis can be formulated as follows:
: ( )
0
: ( )
1
The corresponding F test is:
(SRR − 1)/ ( − 1)
=SRR1/[ − (1+ )− ]
The empirical research equation for Granger test is computed as:
, = 0 + ∑ =0 1 , −1 + ∑ =1 1 , −1 + + , ,
(2.1)
, = 0 + ∑ =0 1 , −1 + ∑ =1 1 , −1 + + , ,
(2.2)
where , is the proportion of total tax revenue to gross domestic products (GDP) of country i (i=1,…N) at time t (t=1,…T), , denotes the proportion of total government expenditure to GDP, k and p are latencies. stands for country -characteristic effects and , represents the observation error with E( , ) = 0.
In the last decades, a few studies investigated the linkage between economic growth and public spending or tax revenue that are two representative factors of public policy. Furthermore, confusing based on Barro (1990)’s argument that tax revenue is equal to public expenditure, nevertheless, tax revenue also depends on public expenses. The question is that “how does tax revenue correlate to government expenditure?” In the past,
the answers have been mixed and confusing. Applying Granger causality test not only helps this study identify the direction of this relationship, but also to enrich the hypothesis of fiscal synchronization that supports the research to capture solution for the second research question.
However, for the less bias from non-stop variables, we should run unit root test first.
After identifying the co-integration and directions of the linkages between dependent and independent variables, to answer the third question, this paper conducts a regression for SUR model (Zellner, 1962, and Yanev and Kontoghiorghes, 2007). This model also verifies the role of governance in modifying the effects between public finance and economic growth in long- run. The SUR model can ensure the efficient computation with orthogonal regression (See equation (4) in chapter 2) and it can help this study to reduce bias from cross-countries data extracted from two financial crises.
In this research, M stands for 3 equations, and ’th dependent variables are 3 factors such as “tax revenue”, “government spending” and “economic growth”. The independent variables are “governance, inflation, FDI, and the human development index”
The empirical model and equation for performing SUR model should be designed as seen as below:
12 ..
[ . 1
2
.
. +
.
][ ]
(3.1)
, = , , + , , (3.2)
where , are dependent variables, which stand for economic growth (lrgdp), tax revenue (TAXgdp), and government expenditure (GEXgdp) of country i at time t, while , represent the independent variable “Governance
- Gov” and other control variables such as inflation rate (infl), the ratio of foreign direct investment value per GDP (FDI) that was computed by logarithm for reducing the bias, and human development indext (hdi).
Conducting SUR and SGMM models helps this study to answer the research question 3 and to fix the endogeneity issue. Blundell and Bond (1998) showed that when the series are closed to a random walk, the system GMM estimation is more robust. In addition, the outcome of economies could be affected by dependent variables with first lag, that indicating the endogenous phenomena. Moreover, the auto-correlation with error term can be existed. In each equation, can re-write as below: = + and transformed lagged dependent variable that correlates with transformed error
term ( − ̅ ), the ∆ also correlates error term U (Baltagi 2005).
So to solve the endogenous phenomena and auto-correlation, the study has to apply two-step system generalized method of moments estimation (Hsiao 2003 and Baltagi 2005). Arrellano and Bond (1991), Baltagi (2005), D’Agostino, Dunne, and Pieroni (2012), and Sasaki (2015) indicated that a dynamic panel data technique with SGMM can help the endogenous growth model be more consistent than the fixed effect model. Furthermore, Barro (1990), Acemoglu and Robinson (2001) revealed that endogenous variables always appear in growth models that make OLS regression biased, and using an exogenous instrument could help regressors fix this issue. In addition, Windmeijer (2005) noted that the two-step GMM procedure obtains consistent and efficient parameters of estimation.
Before designing the empirical research models, this work defines the measurement of economic growth and using other control variables. There is a
i,t-1
large debate of literature on measuring economic growth. In this work, we follow the definition of economic growth at countries’ level only. Represents for neoclassical economists, Solow (1956) took the ratio of production (Yt) stands for growth with the equation: Y(t) = A(t) F(K,L). Stiglitz (1989) indicated the economy growth if rising rate of income. Barro (1991), Cooray (2009) calculated the economic growth is a change of GDP per capita at time t to compare with the previous time, this parameter can explain the real of changing economy. The equation representing the growth rate per capita was expressed as:
( ) − (0) = (1 − −λ )[ ( ∗) − ln (0)]
where y(0) is initial level of output per worker, and y* is stead state level of income per capita and λ is speed of convergence and λ = (1 - α - β - γ)(η+ ϖ +δ). Mankiw et al. (1992), Barro & Sala-i-Martin (1992), Acemoglu et al.
(2005), Acemoglu et al. (2008) used GDP per capita to measure the economic growth. While Siddiqui & Ahmed (2013) considered the rate of growth of real GDP per capita is economic growth. Almost above previous reaserchers tried to control the endogenous phenomena and reduce the error term of specific characteristic of each countries by adding some control variables to their models such as: total investment, government expenditure, tax revenue, trade value, inflation, total population and human development index beacause of that these variables are close relationship with growth of economy.
In accordance with Barro (1990) and Barro and Sala-i-Martin (1992), the empirical model for estimating degrees of tax revenue and government expenditure on economic growth are expanded as seen as below:
, = 0 + 1 , −1 + 2 , + 3 , + 4 , +
5 , + 6 , + , + ,
(4.1)