Analisis Perbandingan Indikator Ekonomi Pendahulu antara Metode Filterisasi Hodrick-Prescott dengan Christiano-Fitzgerald untuk Siklus Bisnis Indonesia
SURYANDARA, REINARDUS ADHIPUTRA (Adv.: Sri Adiningsih, Prof. Dr., M.Sc.), Sri Adiningsih, Prof. Dr., M.Sc.
Antisipasi akan adanya gejala kerentanan perekonomian (early warning system) sangat dibutuhkan dalam memprediksi pergerakan perekonomian suatu negara pada masa mendatang. Leading Economic Index (LEI) atau indikator ekonomi pendahulu adalah suatu indeks komposit yang berisi beberapa variabel ekonomi tertentu untuk meramalkan kegiatan perekonomian dan pergerakan siklus indeks kompositnya mendahului pergerakan siklus bisnis suatu negara. Proses pembentukan indikator ekonomi pendahulu dan siklus bisnis dilakukan dengan seasonal adjustment (memisahkan antara komponen musiman dari komponen siklikalnya) dan detrending (memisahkan komponen tren dari komponen siklikalnya). Proses seasonal adjustment menggunakan Census X-12 ARIMA. Proses detrending menggunakan dua metode filterisasi yang akan dibandingkan yaitu band pass filter Christiano Fitzgerald dan Hodrick-Prescott filter. Pembentukan siklus bisnis Indonesia didekati dengan data Produk Domestik Bruto (PDB) dengan periode kuartalan. Pembentukan indikator ekonomi pendahulu menggunakan data kuartalan dengan variabel penyusun: base money, cadangan devisa, market capitalization, stock issued value, IHSG, konsumsi semen, penjualan mobil, jumlah penumpang pesawat keberangkatan domestik, jumlah penumpang pesawat keberangkatan internasional, realisasi investasi domestik, realisasi invetsasi asing, dan nilai ekspor non migas
Perbandingan akurasi peramalan LEI kedua filter dari proses detrending dilakukan dengan melihat error yang dihasilkan menggunakan program POM For Windows. Proses perbandingan titik balik siklus bisnis kedua filter dilihat dengan uji grafik. Hasil penelitian menunjukan bahwa band pass filter Christiano Fitzgerald mempunyai error yang lebih kecil dibandingkan dengan Hodrick-Prescott filter dalam pembentukan LEI. Hodrick-Prescott filter menghasilkan titik balik siklus PDB lebih banyak dibandingkan dengan band pass filter Christiano Fitzgerald. Hasil uji korelasi silang menunjukan nilai korelasi silang LEI band pass filter Christiano Fitzgerald adalah 0,87 sedangkan untuk LEI Hodrick-Prescott filter adalah 0,51. Band pass filter Christiano Fitzgerald mampu menangkap hubungan jangka panjang yang lebih baik dibandingkan dengan Hodrick-Prescott filter dalam pembentukan LEI
Metode filterisasi Christiano-Fitzgerald mempunyai keakuratan peramalan yang lebih baik dibandingkan dengan metode filterisasi Hodrick-Prescott dalam proses pembentukan LEI untuk meramalkan pergerakan siklus bisnis di Indonesia.
The anticipation of economic susceptible (early warning system) is urgent in predict the movement of a country's economy in the future. Leading Economic Index (LEI) or a precursor of economic indicators is a composite index that contains some specific economic variables to forecast economic activity and the movement of the composite index cycle precedes a country's business cycle movements. The process of formation LEI and business cycles do with seasonal adjustment (separate the seasonal component from its cyclical component) and detrending (separate the trend component from its cyclical component). Seasonal adjustment process using Census X-12 ARIMA. Detrending process using two filtration methods that will be compared to the band pass filter of Christiano Fitzgerald and Hodrick-Prescott filter. Formation of Indonesian business cycle data is approximated by the Gross Domestic Product (GDP) with quarterly period. Formation of the LEI using quarterly data with constituent variables: base money, foreign exchange reserves, market capitalization, stock issued value, IHSG, cement consumption, car sales, the number of domestic departure passengers, the number of passengers of international departures, domestic investment realization, foreign investment realization , and the value of non-oil exports.
Comparison of forecasting accuracy of LEI both of filter detrending process is done by looking at the error generated using POM program for windows. Comparison process business cycle turning points seen by both filter test chart. The results showed that the Christiano Fitzgerald band pass filter has a smaller error than the Hodrick-Prescott filter in the formation of LEI. Hodrick-Prescott filter produces the GDP cycle turning points more than the Christiano Fitzgerald band pass filter. The test results show the value of cross-correlation cross-correlation LEI Christiano Fitzgerald band pass filter was 0.87 while for LEI Hodrick-Prescott filter is 0.51. Christiano Fitzgerald band pass filter is able to capture the long-term relationship better than the Hodrick-Prescott filter in the formation of LEI.
Christiano-Fitzgerald filtering method has better forecasting accuracy compared with the Hodrick-Prescott filtering methods in the process of formation of LEI to predict business cycle movements in Indonesia.
Kata Kunci : early warning system, leading economic indicators, seasonal adjustment, detrending, Christiano Fitzgerald band pass filter, Hodrick-Prescott filter, business cycle
