Dewi, Lailli Arifia (2026) Analisis Risiko Sistemik Saham Subsektor Transportasi dan Logistik Menggunakan Conditional Value-at-Risk Berbasis AR-GARCH dan QRNN dengan Stochastic Search Variable Selection. Other thesis, Institut Teknologi Sepuluh Nopember.
![[thumbnail of 5003221085-Undergraduate_Thesis.pdf]](http://repository.its.ac.id/style/images/fileicons/text.png) |
Text
5003221085-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (20MB) | Request a copy |
Abstract
Sektor transportasi dan logistik merupakan salah satu sektor yang memiliki peran penting dalam mendukung aktivitas perekonomian, namun juga menghadapi berbagai risiko akibat ketidakpastian kondisi ekonomi dan pasar. Risiko tersebut tercermin pada fluktuasi return saham yang dapat menimbulkan risiko individual maupun risiko sistemik. Oleh karena itu, penelitian ini dilakukan untuk mengukur risiko individual dan risiko sistemik pada sembilan emiten saham subsektor transportasi dan logistik menggunakan data return harian selama periode 3 Juli 2023 hingga 31 Desember 2025. Estimasi VaR dilakukan menggunakan pendekatan Autoregressive Generalized Autoregressive Conditional Heteroskedasticity (AR-GARCH) dengan asumsi distribusi standardized skewed Student-t (sstd). Selanjutnya, estimasi CoVaR dan ΔCoVaR dilakukan menggunakan regresi kuantil dan Quantile Regression Neural Network (QRNN). Seleksi variabel pada model CoVaR dilakukan menggunakan pendekatan Stochastic Search Variable Selection (SSVS). Hasil penelitian menunjukkan estimasi VaR pada kuantil 5% dan 1% menunjukkan bahwa pendekatan AR-GARCH menghasilkan model yang valid pada seluruh saham berdasarkan pengujian proportion of shortfall dan Kupiec test sehingga mampu merepresentasikan risiko individual dengan baik. Pada estimasi risiko sistemik, nilai CoVaR dan ΔCoVaR pada kuantil 1% secara umum lebih besar dibandingkan kuantil 5%, yang menunjukkan peningkatan risiko sistemik pada kondisi pasar yang sangat ekstrem. Hasil backtesting menunjukkan bahwa seluruh model CoVaR berbasis regresi kuantil dan QRNN pada delapan saham kuantil 5% dan 1% menghasilkan keputusan valid. Selain itu, QRNN lebih mampu mengikuti pola pergerakan return ekstrem dan menangkap hubungan nonlinier antarsaham dibandingkan regresi kuantil. Penerapan SSVS membantu model dalam memilih variabel prediktor yang paling relevan sehingga model menjadi lebih sederhana dan efisien. Namun, dibandingkan CoVaR berbasis QRNN, performa model dengan SSVS pada beberapa saham masih cenderung kurang stabil. Hal ini menunjukkan bahwa proses seleksi variabel tidak selalu meningkatkan hasil backtesting secara keseluruhan, meskipun mampu mengurangi kompleksitas model. Secara keseluruhan, model CoVaR berbasis QRNN dapat dikatakan sebagai model terbaik karena memberikan hasil yang paling stabil dan konsisten dalam merepresentasikan risiko ekstrem pada saham subsektor transportasi dan logistik.
======================================================================================================================================
The transportation and logistics sector plays an important role in supporting economic activities but is also exposed to various risks arising from economic and market uncertainties. These risks are reflected in stock return fluctuations, which may lead to both individual and systemic risks. Therefore, this study aims to measure individual and systemic risks in nine transportation and logistics subsector companies using daily stock return data from July 3, 2023, to December 31, 2025. Value-at-Risk (VaR) was estimated using the Autoregressive Generalized Autoregressive Conditional Heteroskedasticity (AR-GARCH) model with a standardized skewed Student-t (sstd) distribution. Furthermore, Conditional Value-at-Risk (CoVaR) and ΔCoVaR were estimated using Quantile Regression (QR) and Quantile Regression Neural Network (QRNN) approaches. Variable selection in the CoVaR model was conducted using the Stochastic Search Variable Selection (SSVS) method. The results show that VaR estimates at the 5% and 1% quantiles produced valid models for all stocks based on the proportion of shortfall and Kupiec tests, indicating that the AR-GARCH approach is capable of adequately representing individual risk. For systemic risk estimation, CoVaR and ΔCoVaR values at the 1% quantile were generally higher than those at the 5% quantile, indicating increased systemic risk under extreme market conditions. Backtesting results revealed that all CoVaR models based on QR and QRNN were valid for the eight selected stocks at both quantile levels. In addition, QRNN was more effective in capturing extreme return movements and nonlinear relationships among stocks than conventional quantile regression. The application of SSVS helped identify the most relevant predictor variables, resulting in simpler and more efficient models. However, compared with the QRNN-based CoVaR model, the performance of models incorporating SSVS was less stable for several stocks. This finding suggests that variable selection does not always improve overall backtesting performance, although it can reduce model complexity. Overall, the QRNN-based CoVaR model demonstrated the most stable and consistent performance in representing extreme risk in transportation and logistics stocks.
| Item Type: | Thesis (Other) |
|---|---|
| Subjects: | Q Science > QA Mathematics > QA274.2 Stochastic analysis Q Science > QA Mathematics > QA280 Box-Jenkins forecasting Q Science > QA Mathematics > QA76.87 Neural networks (Computer Science) |
| Divisions: | Faculty of Science and Data Analytics (SCIENTICS) > Statistics > 49201-(S1) Undergraduate Thesis |
| Depositing User: | Lailli Arifia Dewi |
| Date Deposited: | 30 Jul 2026 03:05 |
| Last Modified: | 30 Jul 2026 03:05 |
| URI: | http://repository.its.ac.id/id/eprint/139966 |
Actions (login required)
 |
View Item |