The Combined Method of Forecasting the Investments within the Framework of Panel Data Models

Lyudmila O Babeshko,Irina V Orlova,Maria Y Mikhaleva

doi:10.6000/1929-7092.2018.07.67

Lyudmila O Babeshko, Irina V Orlova + Show 1 more

Open Access

https://doi.org/10.6000/1929-7092.2018.07.67

Copy DOI

Abstract

The article is devoted to the panel data modeling of the firm's investments depending on its market value and the size of fixed assets. The Grunfeld’s investment data as provided in R package were used as the initial data. The data frame contains annual observations for 11 firms over 20 years. The main econometric models for panel data (pooled model, fixed effects model, random effects model) were estimated. To make choice the most effective specification of the model the character of effects was tested. The heterogeneity of firms was explained by individual random factors. The comparative analysis of parameters’ estimates was performed using the basic panel data models and their optimal combination in the framework of combined assessment (forecasting). Weight coefficients of hybrid forecasts are assigned as directed by the combined model list in accordance with standard optimality requirements. It was shown that the results of the combined assessment coincided with the estimates of the random effects model.

Highlights

The article is devoted to the panel data modeling of the firm's investments depending on its market value and the size of fixed assets
The comparison of the results of panel modeling and combined forecasts is performed for the model of dependence of the firm's investments (Iit ) on its market value (Fit ) and the value of fixed assets (Cit )
The following are the results of evaluation and testing of the model (1) according to panel data for four companies: "General Electric", "IBM", "Chrysler", "General Motors" for 20 years: Pooled model: I = !66,897+ 0,097 F + 0, 315C, (s)

Summary

Introduction

The article is devoted to the panel data modeling of the firm's investments depending on its market value and the size of fixed assets. The main econometric models for panel data (pooled model, fixed effects model, random effects model) were estimated. The first one is pooled regression model It does not consider individual features of panels:. The second type of panel data models is fixed effects (FE) model with the individual-specific effects μi , i = 1,..., n , yit. The estimators for panel data models differ based on whether they consider the between or within variation in the data. These estimation methods are based on the exclusion of individual and general means. Fixed effects estimator uses the within variation and the individual-specific deviations of variables from their time-averaged values [Green 2012]:

Methods

Results

Conclusion