Modelo lineal dinámico para estimar la tasa de desempleo en Bogotá D.C.

Abstract

For this project, the estimation of the unemployment rate in Bogot´a D.C. was developed, andthe relationship with some variables of interest was analyzed, all from the frequentist and Bayesianstatistical point of view. Initially, a general exploration of time series for the unemployment rate wascarried out, where it was evidenced that our variable was not stationary. Therefore, it was decidedto apply logarithms and a difference to transform our series and achieve stationarity, necessaryfor a good estimation. Subsequently, the frequentist estimation was carried out under the Box-Jenkins methodology of various ARIMA models, selecting the best model under the AIC and BICinformation criteria. Afterward, the most important assumptions for the residuals of the chosenmodel were verified. Finally, forecasts were made for six months ahead (July 2022 to December2022) of the unemployment rate in Bogot´a D.C.On the other hand, the estimation of the unemployment rate from the Bayesian point of view wascarried out, starting from the assumption that both the prior and posterior distribution followa normal distribution. Subsequently, the first-order polynomial estimation and forecasts for sixmonths ahead were made.In conclusion, a comparison was made between the frequentist and Bayesian methodology to concludewhich was the best for the estimation of the unemployment rate in Bogot´a D.C. After this,several normal dynamic models were proposed, selecting the best ones and observing their results,thus achieving the best possible model.