Page 97 - Special Topic Session (STS) - Volume 4
P. 97

STS566 Iluminada T. Sicat
            In  2018,  the  currency  forecast  model  was  again  enhanced  to  add  2  more
            dummy variables to adjust the impact of seasonality in cash demand observed
            in Q2 and Q4, and adjustment terms to account for serial correlation.
               Model 4: (2017 - present) Model 3 + dummy variable for Q2 + dummy
                                  variable for Q4 + AR & MA terms

            Where:

                        DFIN08 – dummy variable for financial crisis

                         t – error correction term

                         dummy variable Q2

                         dummy variable Q4


                        AR & MA terms – to account for serial correlation

            5.  MAPE: Evaluating Reliability of Forecasting Models
                The following presents an assessment of the performance of the models
            used relative to actual. The best performing model is that which produces the
            lowest forecast error on average, based on the mean absolute percent error
            or MAPE. MAPE refers to the variance or difference between the forecasts and
            actual. Lower MAPE indicates the model’s improving goodness of fit. As can
            be gleaned from the table below, MAPE has been declining from a high of
            33% to 6.9%, which can indicate improving forecast performance. Similarly,
            the accuracy of the forecast is even more enhanced with the introduction of
            dummy variables and autoregressive terms in Model 4, yielding a MAPE of
            only around 2% based on in-sample and out-samples estimates.

                   Model       Model 1       Model 2       Model 3       Model 4
                  MAPE          33.3 %        12.9 %         6.9 %        2.0 %

            6.  Other Factors Affecting Currency Demand
                In addition to economic variables, the currency demand framework is also
            underpinned by the need to maintain a level of inventory for pre-cautionary
            needs. In the Philippines, the BSP Monetary Board, in the past, has approved
            to  maintain  2  types  of  safety  cushions  called  the  “buffer”  stock  and
            “contingency reserves”, equivalent each to 3 months average withdrawal for
            the  past  3  years.  The  buffer  stock  is  meant  to  serve  as  cushion  or  cover
            primarily against uncertainty regarding spikes in demand for cash arising from
            unexpected  business  cycle,  or  uncertainty  arising  from  timing  in  supply
            delivery. Meanwhile, the contingency reserves are meant to provide supply in


                                                                86 | I S I   W S C   2 0 1 9
   92   93   94   95   96   97   98   99   100   101   102