摘要：This paper is a report on the development and the performance of the composite leading indicator of the Croatian economy – CROLEI, those purpose is to forecast classical business and growth cycles. The structure of the paper follows the latest CROLEI revision, based on NBER barometric method. After briefly describing the characteristics of CROLEI and its database, the turning points in the reference series (industrial production) are determined by applying Bry-Boschan algorithm. By combining the graphic analysis, Granger causality and Wald exclusion test on potential leading time series and the reference series, the list of 15 best leading series is compiled. Recursive estimation of Granger test is applied to check whether the time series´ leading properties are stabile over time. Then 14 composite leading indicators are constructed from 15 best leading series. In order to decide which composite indicator yields the best forecasts, Granger causality test is used. New CROLEI indicator is composed of 7 series and leads the reference series by 8 months. Additionally, a diffusion index - an auxiliary tool for forecasting business and growth cycles - is constructed.
关键词：leading indicator; forecasting; non-model based approach; growth cycles
