Optimasi Portofolio Model CVaR-Black-Litterman dengan Integrasi Pandangan Investor Berbasis Inflasi Menggunakan Model Markov Switching Autoregressive

Darmawansyah, Irza Rashad (2026) Optimasi Portofolio Model CVaR-Black-Litterman dengan Integrasi Pandangan Investor Berbasis Inflasi Menggunakan Model Markov Switching Autoregressive. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5006221048-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (7MB) | Request a copy

Abstract

Perkembangan pasar modal Indonesia yang semakin dinamis menuntut strategi optimasi portofolio yang mampu beradaptasi terhadap perubahan kondisi makroekonomi sekaligus mengendalikan risiko kerugian ekstrem. Penelitian ini membahas optimasi portofolio saham IDXBUMN20 dengan mengintegrasikan model Black-Litterman, dan Conditional Value-at-Risk Black-Litterman (CVaR-BL) dengan model Markov Switching Autoregressive (MSAR) berbasis inflasi. Berdasarkan hasil identifikasi dengan batasan dua state dan kandidat orde autoregresif AR(1) hingga AR(5), model MS(2)-AR(2) terpilih sebagai model terbaik karena tidak menunjukkan adanya autokorelasi residual yang kuat berdasarkan uji Durbin - Watson, memenuhi asumsi residual berdistribusi normal, serta memiliki nilai AIC paling kecil. Estimasi parameter menggunakan algoritma Expectation-Maximization (EM) menunjukkan adanya perbedaan karakteristik antara dua rezim inflasi, yaitu rezim stabil dengan rata-rata -0,009 dan varians 0,007, serta rezim fluktuatif dengan rata-rata -0,040 dan varians 0,053. Hasil prediksi satu bulan ke depan dari smoothed probability menunjukkan peluang rezim stabil sebesar 0,714 dan rezim fluktuatif sebesar 0,286. Probabilitas tersebut digunakan untuk membentuk pandangan investor dan ketidakpastian pandangan investor sebagai input model Black-Litterman. Pandangan investor dengan nilai tertinggi pada AGRO.JK sebesar 0,021 dan terendah pada SMGR.JK sebesar 0,004. Ketidakpastian pandangan investor dibentuk berdasarkan varians return berbasis rezim, dengan nilai tertinggi pada AGRO.JK sebesar 0,055 dan terendah pada TLKM.JK sebesar 0,003. Kedua komponen tersebut selanjutnya digunakan sebagai dasar pembentukan posterior expected return pada model Black-Litterman untuk optimasi portofolio. Optimasi portofolio dilakukan menggunakan Black-Litterman berbasis Mean Variance Optimization dan CVaR-BL berbasis Mean-CVaR Optimization pada tingkat kepercayaan 95% dan 99%. Hasil optimasi menunjukkan bahwa Black-Litterman menghasilkan bobot yang lebih terkonsentrasi, sedangkan CVaR-BL menghasilkan portofolio yang lebih terdiversifikasi. Evaluasi kinerja dilakukan secara out-of-sample pada periode Januari hingga Desember 2025 menggunakan cumulative return, Sharpe Ratio, dan maximum drawdown. Hasil evaluasi out-of-sample menunjukkan bahwa CVaR-BL 99% memberikan kinerja terbaik dengan cumulative return sebesar 5,159%, Sharpe Ratio sebesar 3,304%, dan maximum drawdown sebesar -15,737%. Dengan demikian, model CVaR-BL 99% dipilih sebagai portofolio optimal karena mampu menghasilkan return yang lebih tinggi, efisiensi risiko-return yang lebih baik, serta pengendalian risiko kerugian ekstrem yang lebih optimal dibandingkan model lainnya.
=====================================================================================================================================
The increasingly dynamic development of the Indonesian capital market requires portfolio optimization strategies that can adapt to changing macroeconomic conditions while effectively controlling the risk of extreme losses. This study examines the optimization of an IDXBUMN20 stock portfolio by integrating the Black–Litterman model and the Conditional Value-at-Risk Black–Litterman (CVaR-BL) model with an inflation-based Markov Switching Autoregressive (MSAR) model. Based on the identification results using two states and autoregressive order candidates ranging from AR(1) to AR(5), the MS(2)-AR(2) model was selected as the best-performing model because it showed no strong residual autocorrelation based on the Durbin–Watson test, satisfied the assumption of normally distributed residuals, and produced the lowest Akaike Information Criterion (AIC) value. Parameter estimation using the Expectation-Maximization (EM) algorithm revealed distinct characteristics between the two inflation regimes: a stable regime with a mean of −0,009 and a variance of 0,007, and a fluctuating regime with a mean of −0,040 and a variance of 0,053. The one-month-ahead forecast based on smoothed probabilities indicated a probability of 0,714 for the stable regime and 0,286 for the fluctuating regime. These probabilities were used to construct investor views and investor view uncertainty as inputs to the Black–Litterman model. The highest investor view was obtained for AGRO.JK at 0,021, while the lowest was recorded for SMGR.JK at 0,004. Investor view uncertainty was constructed based on regime-dependent return variances, with the highest value observed for AGRO.JK at 0.055 and the lowest for TLKM.JK at 0,003. These two components were subsequently used to determine the posterior expected returns in the Black–Litterman model for portfolio optimization. Portfolio optimization was conducted using the Black–Litterman model based on Mean-Variance Optimization and the CVaR-BL model based on Mean-CVaR Optimization at confidence levels of 95% and 99%. The optimization results showed that the Black–Litterman model produced more concentrated portfolio weights, whereas the CVaR-BL model generated more diversified portfolios. Portfolio performance was evaluated out-of-sample from January to December 2025 using cumulative return, the Sharpe Ratio, and maximum drawdown. The out-of-sample evaluation results demonstrated that the CVaR-BL model at the 99% confidence level achieved the best performance, with a cumulative return of 5,159%, a Sharpe Ratio of 3,304%, and a maximum drawdown of −15,737%. Therefore, the CVaR-BL model at the 99% confidence level was selected as the optimal portfolio model because it generated higher returns, provided better risk–return efficiency, and offered more effective control of extreme loss risk than the other models.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Black-Litterman, CVaR-Black-Litterman, Inflasi, IDXBUMN20, Markov Switching Autoregressive, Black-Litterman, CVaR-Black-Litterman, Inflation, IDXBUMN20, Markov Switching Autoregressive
Subjects:	H Social Sciences > HG Finance > HG4012 Mathematical models H Social Sciences > HG Finance > HG4529.5 Portfolio management H Social Sciences > HG Finance > HG4915 Stocks--Prices Q Science > Q Science (General) > Q180.55.M38 Mathematical models Q Science > QA Mathematics > QA274.7 Markov processes--Mathematical models.
Divisions:	Faculty of Mathematics, Computation, and Data Science > Actuaria > 94203-(S1) Undergraduate Thesis
Depositing User:	Irza Rashad Darmawansyah
Date Deposited:	16 Jul 2026 09:24
Last Modified:	16 Jul 2026 09:24
URI:	http://repository.its.ac.id/id/eprint/135268

Actions (login required)

View Item