Chapter Iii Research Methodology

CHAPTER III RESEARCH METHODOLOGY

3.1 RESEARCH DESIGN In this research, the writer is using quantitative research method. Quantitative research can be described as entailing the collection of numerical data and exhibiting the view of relationship between theory and research as deductive, a predilection for natural science approach, and as having and objectivist conception of social reality. In other words, quantitative research primarily examines relationships between numerically measured variables with the application of statistical techniques (Bryman & Bell, 2015). The writer uses this method because the purpose of this research is to understand and know the level of influence of Regional Taxes and Retributions towards the amount of Local Own Source Revenue with the methods that have been specified in quantitative research.

3.2 POPULATION AND SAMPLE In quantitative research, population is the research subject. The population can be defined as a generalization which consists of an object or subject that has a certain quantity and characteristic set by the researcher to be studied and then drawn its conclusions (Firdaus & Zamzam, 2018). From the definition, the population used in this research is all the cities or regencies in the province of North Sumatra. According to Firdaus & Zamzam (2018), a sample is a smaller collection of units from a population or representative that is used to determine truths about that population. The sample data used in this research is taken by using census. Census or saturation sampling technique is a sample determination technique when all members of the population are used as a sample. In other words, census is a

method of data collection in which the entire population is investigated. The sample used in this research consists of 25 regencies and 8 cities. This research is using the realization data of APBD take over four-year period, from year 2013 to 2016, with Tanjung Balai city, Padangsidempuan city and Nias Utara regency do not upload the data of APBD realization in 2014. Also, in 2013, Batu Bara regency, Nias Selatan Regency and Gunung Sitoli city do not upload the data of APBD realization. So, the total sample used in this research is 126 samples.

3.3 DATA COLLECTION METHOD The data that is required in this research are internal data which is the realization report of APBD in the period of 2013 to 2016 which show the data of Local Own Source Revenue, Regional Taxes and Retributions revenue in all cities and regencies of North Sumatra province. This data is in the form of quantitative data and time series. The data collection methods that are used to collect the data for this research are as follows: 1. Documentation Documentation is the data collection methods in the form of written sources of books, directories, and other data related to the research. Documentation in this research means collecting all secondary data in which the writer obtains from Directorate General of Fiscal Balance of Indonesia Republic that is www.djpk.kemenkeu.go.id. The required data is the report of realization of regional budget (APBD) including the Local Own Source Revenue, the Regional Taxes and Retributions revenue for 2013-2016 and other data related to this research.

2. Literature Studies To obtain the basic of study and strong concept about the topic in order to solve the problem, therefore the writer conducts literature studies by studying and collecting data through books, tax journals, literature, articles, taxation laws, data from the internet, and others which is related to the research that is being conducted. 3. Time Series Analysis This analysis is essentially looking at the measurement from time to time. The measurements are in various ways and usually by means of frequency, percentage or by looking at the central tendency of an event. The data to be analyzed in this time series method are the data of PAD North Sumatra province, consists of Regional Taxes and Retributions of each regency or city in North Sumatra province.

3.4 OPERATIONAL VARIABLE DEFINITION AND VARIABLE MEASUREMENT Operational variable definition is the definition given to the variable by giving meaning or specifying activity. The variables used in this research are: 1. Dependent Variable (Y) Dependent Variable is the variable that is being measured in the experiment and what is affected during the experiment. Dependent variable in this research is Local Own Source Revenue (PAD). According to Law Number 33 Year 2004 on Financial Balance between Central and Regional Governments, Local Own Source Revenue (PAD) is part of the regional revenue that is derived from the potential of the region itself which is levied in

accordance with the regional regulations and applicable law. 2. Independent Variable (X) Independent Variable is a variable that affects or causes the changes or the occurrence of dependent variable. In this research, the independent variable is divided into two groups, such as:

a. Regional Taxes (X1) According to Law Number 28 Year 2009, Regional Taxes shall mean obligatory contribution to the region owed by private individuals or entities of enforced nature based on the law, without receiving direct compensation and used for the needs of the region mostly for the welfare of the people.

b. Regional Retributions (X2) In article 1 paragraph 26 of Law Number 34 year 2000 states that regional retributions are the payment for the services or granting certain permits that are specifically provided and/or managed by the regional government for the benefit of an individual (Fitriana, 2014).

3.5 DATA ANALYSIS METHOD Data analysis method used in this research is descriptive analysis, classic assumption test and hypothesis testing which is done through multiple regression analysis using SPSS (Statistical Package Social Science) 25.0 software.

3.5.1 DESCRIPTIVE ANALYSIS Santoso (2017) reveals that descriptive statistics provide an overview or description of data such as mean, standard deviation, variance, maximum, minimum, sum, rage, kurtosis and skewness.

3.5.2 CLASSIC ASSUMPTION TEST The use of regression analysis in statistics should be free of classical assumptions with the purpose of avoiding or reducing the bias of the research results. The classical assumption used is as follows:

3.5.2.1 NORMALITY TEST Normality test aims to test whether in a regression model, confounding variable has normal distribution. This test is necessary because in order to conduct t test and F test assumes that the residual values follow the normal distribution (Arifin, 2017). According to Ghozali (2013), the good regression model should have a normal or near to normal distribution. There are two ways to detect whether the residual is normally distributed or not, such as: a. Graphical Analysis Normality test can be done by looking at the histogram graph that compares the observed data with the distribution near to normal distribution or by looking at the normal probability plot that compares the cumulative distribution of normal distribution. Normal distribution will form a straight diagonal line and the plots of residual data will be compared to the diagonal line. If the residual is normally distributed, the line that describes the actual data will follow the diagonal line. b. Statically Analysis Normality test can be done by using Kolmogrov-Smirnov (K-S). In Kolmogrov-Smirnov, the guidelines used in decision making are as follows: 1. If the values is significantly greater than 0.05 then the data is normally distributed.

2. If the values is significantly smaller than 0.05 then the data is not normally distributed (Ghozali, 2013).

Some ways to overcome the abnormal distribution are: 1. By transforming the data to another form, e.g. in the form of logarithms. 2. Trimming is deleting the outlier data. Outlier data is the data that has a very deviating value from the value of other data. 3. Winsorizing, it changes the value of outlier data into maximum or minimum value in order to the data is normally distributed (Ghozali, 2013).

3.5.2.2 MULTICOLLINEARITY TEST Multicollinearity test aims to test whether the regression model has a correlation between independent variables. A good model should not have a correlation between the independent variables. The detection of multicollinearity problem can be seen from: a. The R2 value generated by estimation of empirical regression model is very high but individually many of the independent variables do not significantly affect the dependent variable. b. Analyze the correlation matrix of the independent variables. c. The Tolerance Value and Variance Inflation Factor (VIF) values in which Tolerance Value measures the variability of the selected independent variables that are not explained by other independent variables while the VIF is an estimation of how much multicollinearity increases the variance of a coefficient estimation of an independent variable. Generally, the cutoff values used to indicate the

presence of multicollinearity are Tolerance ≤ 0.1 and VIF value > 10 (Gani & Amalia, 2015).

3.5.2.3 HETEROSCEDASTICITY TEST Heteroscedasticity is a condition in which the variant of residual value is unequal between one observer (observation) with another observer. If the variance and residual value are equal between one observer with another observer, then this condition is called homoscedasticity condition. A good regression is a regression that is in a position of homoscedasticity and not a condition of heteroscedasticity. The variable is expressed in the heteroscedasticity position if the observer points spread above and/or below the zero on the Y axis and lead to clear pattern. On the contrary, the observer points spread above and/or below the zero on the Y axis and lead to an unclear pattern, there has been homoscedasticity (Gani & Amalia, 2015).

3.5.2.4 AUTOCORRELATION TEST The autocorrelation test aims to test whether in a linear regression model there is a correlation between the confounding error in period t with error in period t-1 or the previous. Autocorrelation arises due to the consecutive observations over time are related to each other. This problem arises because residual or confounding error is not free from one observation to another. A good regression model is a regression which is free from autocorrelation (Ghozali, 2013). In this research, autocorrelation test is done by using Runs Test. The guidelines used in decision making are: a If significant value is greater than 0.05, there is no correlation relationship between residual. b If significant value is less than 0.05 then there is correlation between residuals.

3.5.3 HYPOTHESIS TESTING

3.5.3.1 MULTIPLE REGRESSION ANALYSIS According to Gani & Amalia (2015), regression analysis is the study of the reliant between dependent variables and one or more independent variables, with the purpose of estimating and / or predicting the average population or the mean value of the dependent variable based on the value of the independent variable that is known. Hypothesis testing in this study is using multiple regression analysis because this study is using more than one independent variable. In accordance with the formulation of the problem, the equation model of regression analysis of this study is as follows:

Y= a + b1X1 + b2X2 + et

Where: Y = Local Own Source Revenue (PAD) (Dependent Variable) X1 = Regional Taxes (Independent Variable) X2 = Regional Retributions (Independent Variable) A = Constants B = Regression coefficients that show the numbers increase or decrease of dependent variable based on independent variable e = Error

3.5.3.2 COEFFICIENT OF DETERMINATION (R2) Coefficient of Determination (R2) is used to measure how far the model’s ability to explain variations of dependent variables. The coefficient of determination is between zero to one (0

variables gives almost all the information needed to predict the dependent variable. The closer the value of R2 to 1, it means the influence of dependent variables that can be explained is stronger (Ghozali, 2013).

3.5.3.3 SIMULTANEOUS SIGNIFICANT TEST (F TEST) According to Gani & Amalia (2015), F test is used to determine the effect of all independent variables that are intended in the regression model together to the dependent variable in the test. The number F can be found by using the formula as follows: R!/(k − 1) F = !"#$%&'$#( (1 − R!)/(n − k) The model feasibility test is carried out with the following criteria:

If F arithmetic > F table (a, k-1, n-k), then H0 is rejected

If F arithmetic < F table (a, k-1, n-k), then H0 is accepted Where:

H0 = The model is not feasible therefore it cannot be used to estimate population.

H1 = The model is feasible so that it can be used to estimate the population.

3.5.3.4 PARTIAL SIGNIFICANT TEST (T-TEST) The t test is used to determine whether or not the influence of each independent variable individually to the dependent variable is tested at a significant level of 0.05 (5%) and to test whether the proposed hypothesis is accepted or rejected. For t test, this research compares t arithmetic with t table, it can be concluded that each independent variable individually affects the dependent variable, with the test criteria as follows:

a. If minus t table ≤ t arithmetic ≤ positive t table, then Ho is accepted and Ha is rejected.

b. If t arithmetic ≤ minus t table or t arithmetic > positive t table, then Ho is rejected and Ha is accepted. In t test, this is done on the degrees of freedom (n-k- 1), where n us the number of respondents and k is the number of variables for the confidence level used is 95% or α=5% (Gani & Amalia, 2015).