Poverty remains one of Indonesia's most persistent development challenges. Beyond low income, it reflects limited access to education, healthcare, employment, and economic opportunities. Because these factors often influence one another, conventional statistical models frequently struggle to distinguish their individual effects.
Simalungun Regency has experienced encouraging progress in reducing poverty over the past decade. The poverty rate declined from approximately 10.81 percent in 2016 to 7.72 percent in 2024, although the COVID-19 pandemic temporarily pushed the figure up to 8.81 percent in 2021. These fluctuations highlight the need for more robust analytical approaches capable of handling complex relationships among socioeconomic variables.
To investigate the issue, the researchers analyzed annual socioeconomic data from 2017 to 2025 using official statistics published by Indonesia's Central Statistics Agency (BPS). The study examined poverty as the dependent variable and evaluated five potential influencing factors:
- Human Development Index (HDI)
- Open Unemployment Rate
- Gross Regional Domestic Product (GRDP)
- Population Density
- Total Population
The data were processed using the R statistical software, with particular attention given to multicollinearity, a statistical problem that occurs when independent variables are highly correlated. The analysis revealed that Population Density and Total Population produced Variance Inflation Factor (VIF) values greater than 10, confirming severe multicollinearity and indicating that traditional regression techniques would not produce reliable estimates.
The research then compared Ridge Regression and Partial Least Squares based on their predictive accuracy and ability to explain variations in poverty levels.
The results clearly favored Ridge Regression.
The Ridge Regression model achieved:
- Adjusted R-Squared: 88.93%
- Mean Squared Error (MSE): 0.0369
Meanwhile, the Partial Least Squares model produced:
- Adjusted R-Squared: 57.20%
- Mean Squared Error (MSE): 0.1426
A higher Adjusted R-Squared indicates that a model explains more variation in poverty levels, while a lower MSE reflects greater predictive accuracy. Based on both indicators, Ridge Regression delivered a significantly stronger statistical model than PLS.
The study also identified the socioeconomic variables that significantly influence poverty in Simalungun. According to the best-performing Ridge Regression model, only Human Development Index (HDI) and Total Population showed statistically significant effects.
The findings indicate that improvements in HDI—which measures achievements in education, health, and living standards—are associated with lower poverty levels. Total Population also demonstrated a significant negative relationship with poverty within the study's model. In contrast, the Open Unemployment Rate, GRDP, and Population Density were not statistically significant predictors after multicollinearity had been addressed.
The researchers from HKBP Nommensen University emphasize that improving human development should remain a central strategy for reducing poverty. Investments in education, healthcare, and living standards not only enhance people's quality of life but also contribute directly to lowering poverty rates.
The study further recommends that the Simalungun Regency Government prioritize policies aimed at strengthening the components of the Human Development Index while also reviewing population management strategies to ensure poverty alleviation programs are more effectively targeted.
For future research, the authors suggest comparing Ridge Regression with other statistical approaches such as Principal Component Regression (PCR) and Least Absolute Shrinkage and Selection Operator (LASSO), as well as incorporating additional socioeconomic indicators and broader geographical coverage to improve predictive performance.
Beyond its statistical contribution, the study offers practical implications for policymakers. More accurate analytical models enable governments to identify the factors that genuinely influence poverty, allocate public resources more efficiently, and design evidence-based development programs. For researchers, the findings demonstrate that selecting the appropriate statistical method is essential when analyzing highly correlated socioeconomic variables. Ultimately, the research reinforces the importance of investing in human development as a sustainable pathway toward long-term poverty reduction.
Author Profile
Leony Purba is a researcher at HKBP Nommensen University whose work focuses on applied statistics, regression modeling, and socioeconomic data analysis.
Juli Antasari Sinaga is an academic at HKBP Nommensen University specializing in statistics, quantitative research methods, and data analytics.
Yoel Octobe Purba is a researcher at HKBP Nommensen University with expertise in statistical modeling and quantitative approaches to socioeconomic research.
Research Source
Article Title: A Comparison of Ridge Regression and Partial Least Squares in Addressing Multicollinearity Among Factors Affecting Poverty in Simalungun Regency in 2025
Authors: Leony Purba, Juli Antasari Sinaga, and Yoel Octobe Purba
Journal: Indonesian Journal of Advanced Research (IJAR), Vol. 5, No. 6, 2026
0 Komentar