TRIPLE EXPONENTIAL SMOOTHING TECHNIQUES: APPLICATION TO KENYA’S INDUSTRIAL INPUTS PRICE INDEX

Abstract

A move towards industrialization is an active ingredient in achieving sustainable economic development owing to the derived benefits of the creation of employment opportunities and enhanced international trade. Through its big four agenda launched on December 12, 2017, Kenya aims to foster the manufacturing sector. One of the industrial- agenda is reducing the costs of industrial inputs. Thus, an accurate predictive model that can be used to gauge the cost of manufacturing inputs ought to be developed. The current study compared the pertinence of two Holt-Winter Exponential Smoothing (HWES) techniques in forecasting Kenya's industrial inputs price data. Unlike simple moving average, where past values are weighted equally, exponential functions assign exponentially decaying weights, over time. The study used secondary data on Kenya's monthly industrial inputs price index from January 1980 to June 2018 extracted from the OECD website. The data had 450 observations and was analyzed using R software. The findings indicated that a hybrid of both the additive and multiplicative HWES model efficiently captures the nonlinearity or seasonality of industrial inputs price index series. Specifically, the “optimal” model was a specification of the multiplicative error, additive trend, and multiplicative seasonality (“MAM”) with a performance accuracy of 2.3% in terms Mean Absolute Percentage Error (MAPE) in making 24 months step-ahead forecasts. The model outperformed the purely additive (2.44%) or multiplicative HWES model (2.55%). The estimated smoothing of alpha, beta and gamma were; 0.9647, 0.1378, and 0.0004, respectively. The prediction future prices movement is beneficial to producers, consumers and policymakers. The 24-period forecast of the industrial inputs the price index indicates a falling trend, and indication that the industrial agenda shows some prospects in the reduction of the cost of inputs