The behavior of the forecast errors: Insights from firm level survey data

Botsis, Alexandros ORCID: 0000-0003-1206-7993 (2021). The behavior of the forecast errors: Insights from firm level survey data. University of Birmingham. Ph.D.

Preview

Botsis2021PhD.pdf
Text
Available under License All rights reserved.
Download (1MB) | Preview

Abstract

Expectations are the quintessence of modern economic theory. Indeed, economic agents base their decisions on their forecasts of the future, which distinguishes the economic science from the natural sciences. Depending on how agents form their expectations, economic models predict vastly different outcomes, which has policy making implications. For instance, money supply cannot affect output when the full information rational expectations hypothesis holds. Therefore, there is a need for a better understanding of the properties of the agents' expectations. To improve our understanding of the agents' expectations, in this thesis, I focus on their forecast errors. I study an innovative combination of data comprising of two datasets. First, the survey data that records the expectations of the firms in Greece's Manufacturing sector. Second, I match the survey responses with the respondents' financial statements. Studying the forecast errors of the firms, I offer some novel insights into the full information rational expectations hypothesis. Namely, I show that only major forecast errors (the least accurate ones) show systematic patterns that lead to the rejection of the rational expectations. Minor forecast errors show no systematic patterns. In order to arrive at this conclusion, I develop a quantification model that allows me to compute quantified forecast errors on sales growth using the survey-based expectations and the realized sales growth of the financial statements. This quantification is pivotal in distinguishing between major and minor forecast errors. The contribution of my quantification model is that it delivers quantitative annual forecasts at the firm-level. Second, I adapt and implement a recent threshold estimator that endogenously estimates where the behavior of the forecast errors changes. Finally, in the last part, I provide a new test for the full information rational expectations which measures the extent of the information inefficiencies in survey data. This test is broadly applicable to all survey-based forecast errors.

Type of Work:

Thesis (Doctorates > Ph.D.)

Award Type:

Doctorates > Ph.D.

Supervisor(s):