Forecasting the Colombian NationaL Government´s total expenditures and total income using temporal aggregation

Abstract

We estimate annual forecasts to the Colombian Government’s total revenues and total expenditures, using temporal aggregation methodology. Particularly, we use monthly data (high frequency data) to estimate a SARIMA model for each account, then we apply the Silvestrani and Veredas(2004) temporal aggregation technique to obtain an annual equivalent Data Generating Process (DGP). Two different forecast methods are compared: i) adding twelve-steps-ahead forecasts from the high frequency DGP and ii) estimating a one-stepahead forecast from the low frequency (annual) DGP. We found that the aggregation technique improves the accuracy of the deficit forecasts. Furthermore, this methodology allows updating the annual forecasts as soon as new monthly data available. Finally, we discuss the applications of this methodology in the private and public sector’s budgeting process.