Evaluación de pronósticos de las reservas internacionales netas en Colombia

Abstract

ResumenDisponer de adecuados pronósticos para las reservas internacionales resulta relevante a fin de sustentar la toma de decisiones por parte de las autoridades monetarias encargadas de su administración. La presente investigación compara la capacidad de pronóstico para dicha variable en Colombia a través de varios modelos con fundamento económico subyacente, así como sistemas estadísticos de series temporales univariadas, multivariadas y de heterocedasticidad condicional, ajenos a la teoría económica. Dada la especificación de la muestra (2000Q1-2014Q3), el mejor modelo resulta ser el de vectores autorregresivos cointegrado en niveles con variables exógenas. Adicionalmente, se estima econométricamente el coeficiente de compensación de corto plazo para Colombia utilizando las perspectivas monetarista, de portafolio y keynesiana, argumentando la hipótesis de que el Banco de la República tiene el poder de afectar la base monetaria por medio de cambios en el activo interno neto. Palabras Claves: reservas internacionales; capacidad de pronóstico; coeficiente de compensación; Banco de la República. AbstractHave adequate forecasts for international reserves is relevant to support decision-making by monetary authorities responsible administration. This research compares the forecasting performance for that variable in Colombia through several models with underlying economic rationale and statistical systems univariate and multivariate time series, and conditional heteroskedasticity, oblivious to 2000Q1-2014Q3), the best model proves to be the vector autoregressive with exogenous variables cointegrate in levels. Additionally, the short term compensation coefficient for Colombia is estimated using monetarist, portfolio and keynesian perspectives, arguing statistically the hypothesis that the Banco de la República has the power to affect the monetary base through changes in net domestic assets.

Forecasting Foreign Exchange Reserves Using Bayesian Model Averaging-Naïve Bayes

“…On the other hand, those works that have studied the reasons why countries accumulate international reserves and the implications derived from this fact (Seghezza, Morelli and Pittaluga, 2017;Pina, 2015;Delatte and Fouquau, 2012;Vujanovic, 2011;Bastourre, Carrera and Ibarlucia, 2009;Aizenman and Lee, 2007;Mohanty and Turner, 2006). Finally, those empirical studies that have developed prediction models of international reserves (Jung and Pyun, 2018;Espinosa, 2016;Gupta, Hammoudeh, Kim and Simo-Kengne, 2014;Li and Yang, 2012).…”

“…They obtained accuracy close to 68%. For its part, Espinosa (2016) it made a forecast of net international reserves in Colombia. In this study, six different quantitative methodologies were presented for the construction of international reserves growth models.…”

“…In the recent literature, some models of prediction of international reserves stand out (Jung and Pyun, 2018;Espinosa, 2016;Gupta, Hammoudeh, Kim and Simo-Kengne, 2014). These models have shown that knowing the factors that explain the behaviour of international reserves is of great interest to predict the future functioning of the economy.…”

“…Likewise, Espinosa (2016) proposed as future research issue to deepen in the estimation of forecasts using Bayesian models, which could be a valid alternative for the prediction of international reserves. Further, the global crisis and it consequences in Europe have demonstrated that capital account risk can strike advanced economies too, putting into question the typical acceptance that they need little reserves.…”

“…This model distinguishes itself for the capacity to predict in all countries, also achieving levels of accuracy of more than 93%. This model has been constructed from a sample of 102 countries, including both emerging and developed, and applying the Bayesian Model Averaging-Naive Bayes classifier (BMA-NB), which shows excellent precision results compared to other methods applied in the previous literature (Jung and Pyun, 2018;Espinosa, 2016;Gupta, Hammoudeh, Kim and Simo-Kengne, 2014).…”