Comparison of temporal disaggregation techniques in the case of export of Bangladesh

Abstract

Export plays very an import role in the economy of Bangladesh. We have tested four methods of disaggregation of which one is Boot, Feibes and Lisman (BFL) method mathematical based method. Other three are regression based techniques namely Chow-Lin, Fernandez and Litterman. Yearly export is the only available information for Boot, Feibes and Lisman (BFL) method. It is found that for T-4 growth both BFL first difference (FD) and second difference (SD) of BFL generated much better result than T-1 growth. Secondly, we have disaggregated yearly export of Bangladesh to quarterly export data by Chow-Lin, Fernandez and Litterman using Quantum index of Industrial production (QIP) as the indicator series. Based on RMSE (root mean square error) and figures, it is concluded that the Chow-Lin (MinSS and MaxLog) as well as by Fernandez disaggregated better than Litterman (Maximum likelihood) estimates. It is also found that Chow-Lin (MaxLog) and Fernandez forecasting power is slightly better than the Litterman (ML) approach. This a scholastic work, but the results may be a guideline to temporally disaggregate the annual time series data into quarterly series, which will be beneficial for the countries where high frequency (quarterly) GDP are not available.