INFLUENCE COMPENSATION, CAREER DEVELOPMENT AND JOB SATISFACTION TO TURNOVER INTENTION

DOI:10.38035/DIJEFA Abstract: The purpose of this study aims to test and analyze the effect of compensation, career development and job satisfaction on employee turnover intentions at PT. XYZ Research data is data collected from interviews, questionnaires and supporting data from PT. XYZ for 3 consecutive years from 2016 to 2018. The sampling method used is the saturation sampling method. From a population of 156 people, the sample used was 113 people as well. The analytical method used in this study is multiple linear regression analysis. The results showed compensation, career development and job satisfaction together (simultaneously) significantly influence turnover intention. Partially shows that compensation has a negative and significant effect on turnover intention. Career development has a negative and significant effect on turnover intentions. Job satisfaction has a negative and significant effect on turnover intentions.


INTRODUCTION
The development of telecommunications technology has transformed the telecommunications industry in the world. The shift of voice and SMS telecommunications to data-based telecommunications is supported by the development of a massive internet. The availability of infrastructure networks and telecommunications access, in Indonesia has driven the growth of internet usage through wireless data access increasing rapidly. Based on secondary data, the volume of PT XYZ employee resignations has continued to increase in the last 3 (three) years. Based on Figure 1.1, it can be calculated the percentage of employee resignations in 2016 of 8.6%, in 2017 it increased to 10.6%, and in 2018 it increased to 11.5%. This is thought to be the problem of turnover intention in PT XYZ's work environment. each company, employees can leave from time to time. Some researchers have suggested that having the intention to quit is the best predictor of turnover. Conceptual models and empirical models of turnover intentions provide strong support for propositions which state that behavioral intention forms the most important determinant of actual behavior (Putrianti et al., 2014).

RESEARCH METHODS
Based on the title and background of the problem previously discussed, this type of research falls into the category of causality research. According to Ferdinand (2014: 7) causality research is a form of research with the aim of seeking understanding of the causal relationship between more than one variable that has been developed in management.
According to Ferdinand (2014: 171) population is a combination of all the elements that can form events, things or individuals who have similar characteristics or characteristics and become the focus of a researcher.
Sampling in this study uses the category of nonprobability sampling in the form of saturated sampling. According to Indra and Cahyaningrum (2019: 54) nonprobability sampling is a way of determining samples where members of the population do not get the same opportunity during the sample selection process. According to Sugiyono (2014: 126) saturated sample is a sampling technique if all members in the population are sampled and this technique is often used in populations with small numbers or in studies that aim to generalize with relatively small error rates.
According to Ferdinand (2014: 229) descriptive statistics are used to get an empirical picture through data obtained through research where the types can be frequency distribution, average statistics or index numbers.
Validity test is carried out using the Statistical Package Service Solution (SPSS) number 25 program using correlations from Pearson Product Moment. The principle of validity testing on Pearson Correlation Product Moment is by connecting the score of each item with the total score obtained. The process of knowing whether an item is valid or not can be seen in the magnitude of significance where if the significance is less than 0.05 then the item is considered valid or valid but conversely if the significance is greater than 0.05 then an item is considered invalid or invalid. Another method is to compare the r count (the result of data processing) with the r table score (obtained from r table), if the value is positive and r the calculated result is greater than r in the table then the item is considered valid, applies vice versa if r count results are more smaller than r in the table then the item is considered invalid (Purnomo, 2017: 70).
According to Ferdinand (2014: 218) a scale or measurement instrument is considered to be reliable if the instrument consistently or consistently produces relatively similar values each time a measurement is made. According to Herlina (2019: 70) the reliability test using the Cronbach alpha method is generally divided into categories:  Cronbach's alpha with a value <0.06 = reliability is not good  Cronbach's alpha with a value of 0.6 -0.79 = reliability is accepted  Cronbach's alpha with a value> 0.08 = good reliability Normality test is used to see the value of residuals having a normal distribution or even having an abnormal distribution. The model in regression is said to be good if the value of the residuals is fairly or normally distributed. In the normality test that is tested is the residual value only and not each variable (Ansofino et al., 2016: 94). In this study the normality test used by the name Kolmogorof-Smirnov. According to Ismail (2018: 193) the Kolmogorof-Smirnov test was carried out by comparing the distribution of research data with the standard normal distribution in which the standard normal distribution was distributed in the form of z-scores.
Multicollinearity test is used to determine the presence and absence of a high relationship between each independent variable in the multiple linear regression model. If there is a high relationship between independent variables, it can interfere with the relationship that occurs between the independent variable and the dependent variable (Ansofino et al., 2016: 94). This multicollinearity test uses Pearson correlation on each independent variable.
Heteroscedasticity testing is used to determine the presence or absence of dissimilarities from one observation residual to another. A regression model can qualify if the residuals of observations from one to another have similar variance (Ansofino et al., 2016: 94). The method used in this heteroscedasticity test is Glejser. The Glejser method is done through the process of regressing independent variables on absolute residual values. If the significance value of the independent variable over the absolute residual value is greater than 0.05, it is stated that there are no symptoms of heteroscedasticity.
According to Ghozali (2013: 96) multiple regression analysis is an analysis used to measure the effect of two or more independent variables on the dependent variable and also to determine the direction of the relationship that occurs between the independent variable and the dependent variable.
Multiple regression analysis is required to meet the requirements of the classic assumption test which has a normal distribution of data, non-multicollinearity and nonheteroscedasticity.
This study uses a hypothesis test in the form of a determination test (R2), F-test, t-test, and correlation test between dimensions.
F statistical test is used to determine whether the independent variables contained in the regression model have a joint or simultaneous effect on the dependent variable. The results of the F test calculation are then compared with the F table values. If the value of F from the calculation results is greater than F in the table, together or simultaneously the independent variables affect the dependent variable. F-test decision making can be seen from the large significance. If the significance is below 5% or 0.05, together or simultaneously the independent variable has an influence on the dependent variable, and vice versa if the significance is above 5% or 0.05, then simultaneously or simultaneously the independent variable has no effect on dependent variable (Ma'arif in Wati and Primyastanto, 2018: 196).
Partial or t-test is used to find out individually or partial influence of the independent variable on the dependent variable. T test is carried out by making comparisons of the calculated t value with the t value in the table or seeing the significance value. For example the value of t arithmetic> table t and the value of the significance <α 0.05 can be said to be the independent variable has a significant effect on the dependent variable. The reverse also applies if the value of t arithmetic <from t table and the value of the significance> α 0.05, it can be said that the independent variable does not have a significant effect on the dependent variable (Rusmana et al., 2019: 200) The use of the coefficient of determination (R2) can get a picture of how closely the regression relationship between the independent variables with the dependent variable. The higher the value of R2 (0 ≤ R2 ≤ 1) the higher the estimate of a regression model (Nugraha, 2014: 191). The degree of determination coefficient can be calculated using the formula: Correlation analysis in statistical analysis is used to measure the level of relationship that occurs between the independent variables with the dependent variable. Correlation analysis between dimensions is used to determine the relationship of the dimensions of the independent variable to the dependent variable. Data processing in correlation analysis between dimensions uses SPSS (Statistical Product for Service Solution) version 25.

FINDINGS AND DISCUSSION Validity and Reliability Test
The validity test results obtained valid items as many as 6 items of compensation variable, 6 items of career development variable, 6 items of job satisfaction variable, 9 items of turnover intention variable.
Reliability is related to the consistency and predictability test of a measuring instrument. The test is carried out by comparing the Cronbach Alpha number where the Cronbach Alpha value is at least 0.6 or ≥ 0.6. If the value generated from the SPSS calculation results is greater than 0.6 then the questionnaire is reliable, if on the contrary it is not reliable.
Here are the results of the reliability test to the variable Compensation ( 1), Career development ( 2), Job satisfaction ( 3), and Variable Turnover Intention (Y):

Normality Test
The normality assumption test is carried out to test the data of the independent variable ( ) and the dependent variable (Y) in the resulting regression equation, whether it is normally distributed or not normally distributed. If the data distribution is normal, then data analysis and hypothesis testing are used parametric statistics. The normality test aims to test whether in the regression model the dependent variable and the independent variable both have normal distributions or not. A good regression model is having a normal or near normal distribution. The trick is to look at the Normal Probability Plot image. It can be said if the data points spread around the diagonal line and follow the direction of the diagonal line (Ghazali, 2016) Based on Table 2 the results of the Kolmogorov-Smirnov normality test, it can be seen that the Sig. = 0.102> α = 0.05. Thus the Kolomogorov-Smirnov test results from the four variables above have met the normality requirements. This, it can be concluded that the data tested has a normal data distribution

Multicollinearity Test
The multicollinearity assumption test is used to measure the degree of association, closeness of relationships or linear relationships between independent variables. One of the multicollinearity tests commonly used is the Variance Inflation Factor (VIF) test, if the VIF value of the variable <10 then multicollinearity does not occur. Referring to table 3, the tolerance score for each variable is greater than 0.1 and the VIF value for each variable is smaller than 10 so that it can be concluded that the regression model in this study is free from multicollinearity between independent variables.

Heteroscedasticity Test
Multiple regression equations need to be tested whether or not the variant of one observation residual with another observation. If the residuals have the same variant, then it is called homoscedasticity, if if the variance is not the same, it is called heteroscedasticity. A good regression equation is if there is no heteroscedasticity. By using the Scatter Plot, the following results are obtained:

Figure 2. Heteroscedasticity Data Results
(Source: data processing, 2020) Based on Figure 2, it appears that there are no clear patterns such as dots spreading above and below the number 0 on the Y a is, so the points spread randomly, and do not form certain patterns. It can be concluded that the data tested is free from heteroscedasticity

Multiple Linear Regression Analysis
Linear regression is needed to measure the magnitude of the relationship that occurs between two variables or it can be more. Linear regression is also needed in determining the direction of the strength of the relationship that occurs between the independent variable with the dependent variable, the direction can be positive or negative. Regression analysis in this Available Online: https://dinastipub.org/DIJEFA Page 650 study is used to analyze whether there is an influence of the independent variable Compensation (X1), Career Development (X2) and Job Satisfaction (X3) on the dependent variable, Turnover Intention (Y), and how much influence the independent variable has on the independent variable bound together. The following table will be presented based on the results of the multiple linear analysis test:

Tabel 4 Results of Multiple Linear Regression Analysis
(Source: data processing, 2020) Based on Table 4 in the Unstandardized Coefficients B column shows a constant value (α) of 55.527 and in the next row the coefficient of the independent variable (b) is Compensation of -0.494; Career Development is -0.699 and Job Satisfaction is -0.352 so the regression equation is arranged as follows: Y = α + β1X1 + β2X2 + β3X3 + e Y = 55.527 -0. T-test was conducted to find out the magnitude of the effect of the independent variables individually on the dependent variable. This t-test is also intended to estimate the extent of the contribution of changes that occur in each of the independent variables influencing the magnitude of changes in the dependent variable. The criteria for drawing conclusions from the t-test are explained in the sentence below:  If the value of t is calculated> t in the table and the value of sig. <0.05 then Ha is accepted  If the value of t is calculated <t in the table and the value of sig. > 0.05 then Ha is declared rejected T test results based on Table 4 can be concluded as follows: 1) There is a negative and significant influence from Compensation (X1) to Turnover Intention (Y) partially. Table 4 shows that the relationship between Compensation (X1) and Turnover Intention (Y) is significant with a t-count of 2.352 (t-count> t table = 1.98) and a Sig. = 0.020. The coefficient value is negative that is equal to -0.494 which indicates that the direction of the relationship between Compensation (X1) and Turnover Intention is negative at 49.4%. Thus the H1 hypothesis in this study which states that "Compensation (X1) has a significant effect on Employee Turnover Intention (Y)" is accepted. 2) There is a negative and significant influence from Career development (X2) to Turnover Intention (Y) partially. Table 4 shows that the relationship between Career development (X2) and Turnover Intention (Y) is significant with a t-count of 3.694 (t-count> t table = 1.98) and Sig. = 0,000. The coefficient value is negative that is equal to -0.699 which indicates that the direction of the relationship between Career development (X2) with Turnover Intention is negative at 69.9%. Thus the H2 hypothesis in this study which states that "Career development (X2) has a significant effect on employee turnover intentions (Y)" is accepted. 3) There is a negative and significant influence from Job satisfaction (X3) to Turnover Intention (Y) partially.

F-Test
The F test is also called the simultaneous test or ANOVA. The F test is used to determine whether all independent variables in the regression series have a joint influence on the dependent variable. F-test results using SPSS 25 will be presented in the table below:

ANOVA a
Model F Sig.

R 2 Test
R 2 test is used to obtain information on the amount of contributions or contributions that the independent variable gives to the dependent variable. The results of the determination test can be seen in the following table: (X2), and Job satisfaction (X3). The remaining contribution of 36.4% can be explained by other variables.

Matrix Correlation
Correlation test between dimensions is performed to determine the magnitude of the relationship between each dimension of all independent variables with each dimension on the dependent variable. Correlation test results using the Pearson's Correlation two-tailed method and will be described in the following table: The results of the study indicate that Compensation has a significant negative effect on Turnover Intention. The results of this study are supported by the results of previous studies conducted by Johannes et.al (2014) and Khadir & Tinik Sugianti (2016).
The results of the study indicate that Career Development has a significant negative effect on Turnover Intention. The results of this study are supported by the results of previous studies carried out by Rijanti et.al (2016).
The results of the study indicate that Job Satisfaction has a significant negative effect on Turnover Intention. The results of this study are supported by the results of previous studies carried out by Johartono and Retnaningtyas Widuri (2013).
The results of research based on the F test show that Compensation, Career Development and Job Satisfaction simultaneously or simultaneously have a significant influence on Turnover Intention.

CONCLUSION AND SUGGESTION
Based on the results of data processing that has been shown in the previous chapter on "The Effect of Compensation, Career Development, Job Satisfaction on Turnover Intention of XYZ Company" then can draw conclusions as below: 1) The compensation variable has a negative and significant effect on Turnover Intention. This means that the higher the compensation given by the company, the lower Turnover Intention. Satisfaction compensation received by employees, especially the financial dimension consisting of salary, bonuses, benefits has a strong relationship to variable turnover intention, so that the desire of employees to find work with higher pay will be reduced if the company provides better compensation.
2) Career development variables have a negative and significant effect on Turnover Intention. This means that the better career development provided by the company, it will reduce turnover intention. Career development that has a strong relationship to variable turnover intention. So the desire of employees to find other jobs will be reduced if the career given by the company is better.
3) Job satisfaction variable has a negative and significant effect on Turnover Intention. This means that the better job satisfaction of an employee in a company, it will reduce turnover intention. Job satisfaction that has a strong relationship to variable turnover intention. So that the desire of employees to find other jobs will be reduced if the job satisfaction of an employee in a company is better. 4) Variable Compensation, Career development and Job satisfaction simultaneously have a significant effect on Turnover Intention. This means that the variable Compensation, Career development and Job satisfaction are getting worse, it increases Turnover Intention.
Based on the results of the study and the conclusions previously presented, the researcher provides input for school leaders and researchers who will further research as follows: 1) Researchers can then enter the variables or other factors that are suspected to affect Turnover Intention in addition to the variables or factors that exist in this study.
2) The foundation or the principal should increase simultaneously the Compensation, Career Development and Job Satisfaction of the employee to be more effective in reduce Turnover Intention.
3) The company is expected to be able to increase compensation staff such as bonus, extra leave day, as well as other funds.
4) The company is expected to be able to manage & track career development staff so that they feel comfortable and happy to work.
5) The company is expected to be able to increase job satisfaction so staff feel comfortable and happy to work.