Naslov
OPENNESS AND INFLATION NEXUS IN THE US: STATISTICAL LEARNING APPROACHES

Autori
Bošnjak, Mile ; Novak, Ivan ; Vukas, Jurica

Vrsta, podvrsta i kategorija rada
Sažeci sa skupova, sažetak, znanstveni

Izvornik
Conference Proceedings of the International Scientific Conference Technology, Innovation and Stability: New Directions in Finance (TINFIN) / Družić, Gordan ; Šimurina, Nika - Zagreb : Hrvatska akademija znanosti i umjetnosti (HAZU), 2023, 17-32

Skup
Tehnologija, inovacije i stabilnost: novi pravci u financijama (TINFIN)

Mjesto i datum
Zagreb, Hrvatska, 05.05.2022. - 06.05.2022

Vrsta sudjelovanja
Predavanje

Vrsta recenzije
Međunarodna recenzija

Ključne riječi
Inflation rates ; Trade Openess ; Support Vector Regression ; Lasso Regression

Sažetak
Recent empirical literature has recognized statistical learning as a potential approach to compete standard econometrics. However, most of applications rely on data reach environment leading to improved forecasts but confusing its economic interpretability. Generally, critics of statistical learning often consider it as a black box that does not bring a lot to economic reasoning. Forecasting inflation rates was often emphasised as one of the most challenging and important topics due to its effects on wide areas of economics and finance. Various statistical learning approaches have already been applied to forecast inflation, but the link to economic reasoning is still a missing one. Therefore, full potential of statistical learning approach still waits to be illustrated. Wide ranges of variables from economic and finance haves been included as inflation predictors in empirical literature that follows statistical learning approaches, but effects from abroad to forecast inflation rates have been neglected. External effects on inflation rates are generally neglected in case of large economies. However, it is reasonable to believe that effects from abroad matters in conditions of highly globalized world economies. Consequently, besides the missing link to economic reasoning, there seems to be one more weak point. This paper makes a step ahead to overcome existing weak points. While testing trade openness and inflation nexus in case of the US, this paper tries to link results from statistical learning approaches to economic theory. This paper employs statistical learning methods to examine openness and inflation nexus in the US economy. Based on the analytical framework derived from the Philips curve and Romer’s hypothesis, this research examines causality among considered variables and drivers of inflation in the US. The training data sample consists of quarterly data from 1948q2 to 2010q4 while out of sample evaluation was performed on the data sample that ranges from 2011q1 to 2020q3. Based on empirical results from this paper, Support Vector Regression outperformed Lasso Regression in forecasting US inflation rates. While the unemployment rate was found insignificant to improve forecast accuracy of US inflation rates, openness of the US economy to international trade appeared as a significant predictor of inflation rates. Effects from inflation to trade openness were not empirically supported. The results of empirical analysis from this paper suggest that the inflation rate in the US was an externally driven phenomenon.

Izvorni jezik
Engleski

Znanstvena područja
Ekonomija

POVEZANOST RADA