Analysis of Determinants of Unemployment Rate in Indonesia

Solving the unemployment issue is one of the best ways to reduce poverty. Through the provision of job opportunities, the poverty rate can be reduced. Therefore, this research explores the factors that influence the unemployment rate across 34 provinces over the 2015-2018 period using the panel regression technique. The variable used consisting of economic growth, the percentage of people with IT competence, and the average school duration. This study indicates that the unemployment rate can be reduced by increasing the average school duration. Meanwhile, the level of economic growth and the proportion of people with IT competence have an insignificant influence on Indonesia's unemployment rate. Based on these findings, the government needs to ensure that every resident in its territory can receive an adequate education.


INTRODUCTION
The Labour issue is still an important thing in the development agenda, especially in welfare discussion. Poverty appears because people do not have the purchasing power to meet their daily needs of living standards (Sartika et al., 2016). The employment provision will provide income for them to be used as capital in meeting their daily needs. Hence, it may conclude that job opportunity is the most fundamental effort to reduce the poverty rate (Karnani, 2011).
Meanwhile, to create jobs, a stimulus is needed to increase demand, which is reflected in increased public consumption (Mardalena et al., 2019). Theoretically, the aggregate demand will stimulate the producers to increase the output produced, which needs more production factors such as the labours. To get a picture of the increase in demand can be reflected through economic growth (constant price).
Besides, the development of information and technology (IT) also influences the labour market (Pianta, 2017). Sultanuzzaman et al. (2019) also argue that technology is quite enough for the economy because of the capacity to boost the economy on a broad scale. Some other studies show a contradiction where technology gets rid of a lot of labour (Rotman, 2013). However, other opinions state that technology can prevent a job and create alternative work that did not exist before (Caliskan, 2015). Therefore, mastering IT becomes essential to survive in the current labour market (Deloitte, 2017). Technological developments that have penetrated every line of life, including business and industry, make workers must be able to adjust to improve their ability on IT. Even though technology has mostly replaced human workers' position, with adequate technological literacy, their position will not be replaced (ILO, 2010).
In Indonesia, the studies that link IT to employment are still limited to the management sector, specifically discuss the employees' IT mastery to work performance, for example, as the researches were done by Handayani et al. (2018), Muzakki et al. (2016), and Siregar (2019). From an economic perspective, the technological approach is more famous in endogenous growth theory, which mentions that technology will affect total production (Budiono, 2011;Fazri et al., 2017). Nevertheless, IT development is not limited only to employee performance and increased production. It also massively Analysis of the Determinants ...

Muhammad Fathul Muin
influences a region's socio-economic patterns due to its digitalization (Alemie, 1998;Tisdell, 2014). Therefore, it is quite important to include elements of a technological approach in labour studies in Indonesia.
As for education, labour absorption cannot be separated from the role of education obtained by the community (Nugroho & Moonti, 2019). The diverse employment opportunities and the availability of a quality workforce are also inseparable from education's role in it. Educated people are considered capable of seeing various opportunities (Jimenez et al., 2015) and have adequate competence (Abdulrahamon et al., 2018). Therefore, education is needed to support the production process to boost output optimally and following the expectations expected by demand.
Besides, the average level of people education can influence the community's socioeconomic development (Brennan, 2008;Sharma & Monteiro, 2016). The education level will also escalate the variety and level of community consumption, which is increasingly diverse (Ioncica et al., 2012;Worsley et al., 2004). This condition causes the demand for various types of goods and services to increase, and in turn, it will increase the number of labour demand and increase the absorption of existing labour (Keynes, 2008;Michaillat & Saez, 2015).
Most of the previous studies only linked education to workers' aspects and ignored the entire population's educational conditions in general. Even though education influences labour absorption, it is not only on the assumption of increasing workers' capability but also on changing the social environment that allows the creation of many job vacancies.
Therefore, an educational approach with a different perspective needs to be done in this study.
Regarding this research's urgency, Indonesia is still facing an extensive workforce of 131.01 million people or 49.34 per cent of the total population as of August 2018 (BPS, 2018). Also, the potential for economic growth is quite good, with an average economic growth of 5 per cent, making it momentum for income distribution and poverty alleviation through job creation. Therefore, it becomes essential to evaluate the effect of economic growth, the mastery of technology, and education level on the employment rate in

Indonesia.
Analysis of the Determinants ...

Muhammad Fathul Muin
Therefore, this study aims to determine the extent of economic growth, IT mastery by the people, and the average education on influence labour absorption. Based on the findings and conclusions in this study, it hopes that it can serve as an evaluation material for the government in determining macro steps to increase labour absorption so that the unemployment rate can be reduced.

The Scope of Research
The scope of this research includes all 34 provinces in Indonesia. The secondary data used from 2015 to 2018, covering four variables obtained from the BPS Statistics Indonesia, viz: 1. Percentage of the unemployment rate.
It is the ratio of the number of unemployed people to the total workforce.

Economic growth with a constant price.
It is an increase in the regional economic comparison between one period to the previous period using constant prices in 2010.
3. Percentage of the people with IT competence.
It is the proportion of adolescents and adults aged 15-59 years who have skills in Information and Computer Technology (ICT).

The average duration schools.
It is the average length of school that has been taken by every resident aged 15 years and over.

Research Model
From the consumption side, economic growth reflects an increase in aggregate demand for goods and services compared to the previous period. Likewise, when viewed from the production aspect, economic growth can be interpreted as an increase in output produced by economic units. The increase in consumption and production can occur when production inputs are met, one of which is labour. Therefore, the effect of economic growth on the number of unemployed is negative.
Analysis of the Determinants ...

Muhammad Fathul Muin
Meanwhile, people's IT ability is also an added value for someone related to the workforce's modern era skills. Their good ability towards IT reflects that the workforce can adapt to various production process changes, include changes in production technology. A good adaptation is necessary to self stay the current status as an employee in a corporation.
Also, it gives the current opportunities with various entrepreneurial ideas. Thus, mastery of IT should have a positive effect on employment. It means that mastery of IT will negatively affect the number of unemployed people.
As for education, as expressed by experts, it is the primary key in producing quality human beings, starts in technical skills, thinking abilities, and adaptability. Therefore, education has a central role in producing a competent workforce that follows the industrial need. Thus, education will have a negative effect on the number of unemployed people.
Based on this description, in general, the relationship between economic growth, mastery of IT, and education level on the unemployment rate are diagnosed as having a negative and significant relationship. Therefore, visually, the relationship between the four variables can be described as follows:

Analysis Method
The inference analysis used in this research is panel data regression analysis. The number of cross-section data (i) are 34 provinces (i = 1, 2, …, 34) and time-series data (t) are 4 years (t = 2015, 2016, 2017, 2018). The type of panel data used is the balanced panel so that the total observations in this study are 136 observations. Based on Hill et al., (2011), there are four stages of the analysis are as follows:

School Duration
Analysis of the Determinants ...

Muhammad Fathul Muin
1. Determine the best model estimate using three types of statistical tests. In general, the possible models that can be obtained consist of 3 types, namely: a.) The common-effect model (CEM) is a model that assumes that there are no differences between individuals or the time. In other words, these effects are constant. The decision to determine this model is based on two test statistics, namely the Chow test and the LM test. When the model probability in the two statistical test results is not significant, then CEM is the best model estimate. When the Hausman test results are not significant, and the LM test is significant, REM is the best estimation model.

Checking for classical assumptions.
The classical panel regression assumption test consists of testing the assumptions for normality, homoscedasticity, and non-multicollinearity. The use of these assumption test is intended to obtain unbiased and consistent estimator values. In detail, each of these assumptions are described as follows: a.) Normality The normal error distribution is required for regression modelling. If this assumption is violated, the various statistical analysis cannot be performed. The statistical test used follows the chi-square distribution. As for an alternative, when the error distribution is not normal, the data need to be transformed.
Analysis of the Determinants ...

b.) Homoscedasticity
In regression, it assumed that residual variance between observations is constant. If there is heteroscedasticity in the data, and there is no cross-sectional correlation, then the estimation method is Generalized Least Square/Weighted Least Square. Meanwhile, when data heteroscedasticity occurs, and there is a cross-sectional correlation, the estimation method is Estimated Generalized Least Square/Feasible Generalized Least Square with Cross Section SUR/Panel Corrected Standard Errors.
c.) Non-multicollinearity Ideally, each independent variable in the regression is independent and correlate with each other. To detect the presence of multicollinearity, the variance inflation factor (VIF) can be used, which measures the increase in variance in parameters. When the VIF calculation results show more than 5 or 10, there is multicollinearity between exogenous variables. The next step to be taken when multicollinearity occurs is reviewing the independent variables used.
3. Test the significance of the model with F test, t-test, and coefficient of determination.
These three tests are used to determine the model's ability to explain the effect of the independent variable on the variation of the dependent variable. A good model that does not require many independent variables but can explain the phenomenon in the variation of the dependent variable accurately.

Interpreting the chosen model
The model design for this research developed as follows:

Selection of the Best Model
The first stage to determine the best model is using the Chow Test. It has a purpose to compare pooled and fixed-effect models. Based on the test results in Table 1, the probability is 0.000 and smaller than 0.01. Using one per cent significance level, there is at Analysis of the Determinants ...

Muhammad Fathul Muin
least one unequal interception from the 34 provinces studied. Therefore, the fixed effects model is better than the pooled model. The next stage is to determine the best model between fixed effect and random effect models. We used the Hausman test to find the best model between them. Showed in table 2, the calculation of probability in the amount of 0.0001 and smaller than 0.01. This result concludes that using a one per cent significance level, the fixed-effect model is better than the random effect model. Thus, from the two tests that have been carried out, it can be concluded that the fixed-effect is the best model. Cross-section random 20.580 3 0.0001

Classical Assumption Testing
For the panel regression, three assumptions need to be fulfil required. They are normality, non-multicollinearity, and homoscedasticity. Based on the test results like figure   2, the probability value is 0.0367 and greater than the significance level of 0.01. It shows that the error from the data set is normally distributed.

Figure 2. Normality Test
Analysis of the Determinants ...

Muhammad Fathul Muin
The second step is the non-multicollinearity test. Based on the test results obtained by the VIF as table 3, the score of all variables are below the specified threshold of 5.
Thus, it can be concluded that there is no linear relationship (non-multicollinearity) between the independent variables used.   After going through several stages of testing, the best estimation model selected is the fixed-effect model. The formed regression equation is as follows:

Discussion
Based on the regression result, economic growth has an insignificant effect on the unemployment rate. This result indicates that the increase in economic growth does not affect decreasing the unemployment rate in Indonesia. This finding opposite of the hypothesis in this research that assumes economic growth will decrease the unemployment rate. However, this finding is supported by previous research conducted by Funlayo (2013) and Safatillah (2014).
In several cases, economic growth ordinarily able to create additional job opportunities. Growth in Regional Gross Domestic Product (RGDP) indicates that consumption has increased. This condition will push the production to increase with additional production factors, such as the capital (Limam & Miller, 2004) and labour (Winanto, 2019). However, in this case, there is an anomaly phenomenon.
There is an explanation of why economic growth does not significantly reduce the unemployment rate. Economic growth will not significantly affect additional employment when the drivers of economic growth come from non-labour-intensive sectors. The increase in production output, which is only influenced by the increase in the use of production technology or the increase in the number of employee work shifts, will Analysis of the Determinants ...

Muhammad Fathul Muin
certainly not affect the job opening. Several sectors that are not effective in creating employment are the mining and quarrying sector, the information and communication sector, the financial services and insurance sector, and the real estate sector. The four sectors have a relatively large share of RGDP, but the contribution to labour absorption is low.
Meanwhile, economic growth that will be positive in absorbing labour will occur in the agriculture, forestry, fisheries sectors, the manufacturing sector, the wholesale trade sector, the transportation, warehousing sector, and accommodation food and beverage provision sector. The five sectors have a relatively high share of labour, some of which even exceed their RGDP share. Therefore, it becomes logical that an increase in economic growth will not have a positive impact.
Next, the regression result shows that IT competence has an insignificant effect on the unemployment rate decline. It can say that an increase in IT skills on the people does not guarantee that they would get jobs. This finding opposite the hypothesis assumed that IT competence has a negative effect on the unemployment rate. This result also contradicts the previous studies that documented that IT competency plays a role in increasing job opportunities, and IT penetration triggers employment termination and makes a transition for new business types (Dachs, 2017).
From a workforce aspect, IT skills make people able to compete for new jobs.
However, the existence of IT also eliminates some of the pre-existing jobs. It is a cancelout effect, where there is a replacement of one another between work types. It means that IT's existence only changes the work structure without changing the number of existing jobs. This argument is also supported by Michaillat & Saez (2015) in their research, which reveals that shocks to labour demand tend to be caused by shocks to demand, not IT developments.
Besides, if viewed from a broader perspective, the insignificance of IT mastery by the people on labour absorption is also due to the low technology penetration in Indonesia.
As a result, the creation of a multiplier effect is too small. Although in several big cities, the use of IT has penetrated until the transportation sector, the share of this sector to the workforce is still relatively small. As a result, the technological literacy possessed by some of these people is unable to create a massive labour effect. It is different if the penetration  Riddell and Song (2011), who found that education, especially in 12 and 16 years of schooling, significantly increases the employment rates.
Education affects life satisfaction achieved through various mediums, including income and work (Powdthavee et al., 2015). It confirms that education improves a person's ability to adapt to the work environment or solve various problems. In a broader scope, people's education can accelerate the community's social life to develop continuously, which will create agglomeration of demand. This condition stimulates producers to take advantage of this momentum by increasing the scale of production and product variation.
So that in the end, it will increase labour demand, and many workers will be absorbed.
From this point of view, it can be understood that the average population's education will have a positive impact on employment in Indonesia.
Currently, the average duration of school in Indonesia is eight years or equal to the junior high school level. It makes sense that increasing school duration will increase job opportunities and decrease the unemployment rate significantly. However, it does not mean that increasing education will create more jobs and optimal labour absorption. The statistics showed that people in higher education have more unemployment (Maryati, 2015). It may happen because more people educated need a better job with high income (Ishchenko-Padukova et al., 2017). Regardless of the pros and cons, education is essential for labour. Education will have the primary role to create a new quality of economy and society in common (Lavrinovicha et al., 2015). Nevertheless, it necessary be noted that the curriculum must be set based on market-needed oriented (Martinez, 2018), as like entrepreneurship skills and other specific skills.
Analysis of the Determinants ...

Implication
Based on this study's findings, several follow-ups need to be done to increase labour absorption so that the unemployment rate in Indonesia can be reduced. These follow-ups are a concrete effort to intervene based on the variables used in this study.
First, this study finds that aggregate economic growth does not affect labour absorption since the economic growth not in the labour-supporting sector. Based on this, the government needs to reobserve the fiscal policy related to government spending. It needs to spend out on the leading sectors of each region by still taking into labour absorption. Government spending can increase economic growth by providing a multiplier effect if spending is made in the right sector. In the future, the paradigm of budget absorption needs to pay attention to each region's economic posture. This effort is made so that the role of the Regional Budget as one of the economic supports and social welfare can be realized.
Second, the insignificance of IT mastery by the population on labour absorption is due to the cancel-out effect and the lack of multiplier effect of IT on labour-intensive job creation. As a follow-up, the efforts that can be made by the government to increase public literacy towards technology are the introduction of massive IT-based services. So far, several government steps such as smart-city, e-government, internet entering villages, and e-money through QRIS are strategic steps to increase public literacy towards IT in its role in everyday life. This step needs to be promoted more massively so that the achievement of IT penetration is step up.
Third, the significant reduction in education level in the unemployment rate gives a signal to the government that the level of education is essential to get attention. Education is the key to human resource development, both in the mastery of industrial skills, possession of insight and information, problem-solving, and creativity. Therefore, the government needs to encourage people to easily access education, both in terms of physical affordability, appropriate quality, and cost. Equality in the people's ability to access education will increase the participation rate of education, which will stimulate the economy as a whole and prepare a reliable workforce that can absorb this stimulus with production activities.
Analysis of the Determinants ...

CONCLUSION
This research has the objective to test the effect of economic growth, IT mastery, and education level on Indonesia's unemployment rate. The method to determine the relationship among three variables uses panel regression with 34 provinces in range 2015-2018. The finding of this research is that the high economic growth and a sufficient percentage of IT mastery do not guarantee a decrease in most provinces' unemployment rates in Indonesia. This study also concluded that the factor that could reduce the unemployment rate is the average school duration.
Based on these findings, it becomes work for governments to increase the average length of school years while still paying attention to the curriculum which relevant to the job necessity. It also needs further research to investigate the comparison between vocational and non-vocational education and rediscuss the postgraduate level's existence than undergraduate and diploma levels.