Forecasting the behaviour of manufacturing inventory

Forecasting levels of stocks held by manufacturing industry is problematic. Stocks are the most volatile component of GDP. The data itself is subject to chronic revision. Yet, forecasting inventory changes in the supply chain is crucial for firms trying to manage output. The paper reports a successful approach to forecasting UK manufacturing stock behaviour sponsored by a leading European metals manufacturer. The model exploits the seasonality of stock build-ups and run-downs. (Existing econometric approaches rely on seasonally adjusted data.) The forecasting performance of our model is compared to alternative time series approaches. Use of raw, unadjusted data implies markedly different specifications from those in the established literature. In particular, we find no significant evidence of a structural break in UK stockholding behaviour in our sample period and the absence of cointegration between stocks and output suggests the conventional error correction approach is spurious, as well as giving poor forecasts