Latent variable analysis and partial correlation graphs for multivariate time series

Purpose Corporate social responsibility (CSR) has several dimensions that are inherently unobservable or measured with errors. Due to measurement errors of CSR proxies, regression analysis seems inappropriate for investigating the relationship between CSR and firm value. Accounting for CSR measurement errors, the purpose of this paper is to use a latent variable analysis to examine whether CSR affects firm value. Design/methodology/approach This study applies a latent variable model that directly takes into account the measurement errors of CSR proxies. Moreover, the inclusion of firm-fixed effects in the model controls for time-invariant unobservable firm-specific characteristics that may drive both CSR and firm value. CSR is measured by environmental, social, and corporate governance activities. Findings Based on data of US firms between 2002 and 2014, this study finds conflicting evidence of a direct association between each CSR proxy and firm value. When all CSR proxies are incorporated into a latent variable model, CSR significantly positively impacts firm value. Therefore, CSR strategies based on a single measure of CSR or the equal weighting of CSR measures tend to underestimate the influence of CSR on firm value. Practical implications Corporate managers should enhance firm value by simultaneously engaging in environmental, social, and corporate governance activities because there is a synergistic effect with firm value. Furthermore, investors who downplay CSR factors in firm valuation can lead to significant errors in making equity investment choices. Originality/value This study presents a novel examination of the price-earnings ratio in the CSR valuation by using the latent variable model with firm-fixed effects.

Download Full-text

Latent-Variable Analysis of Multivariate Data in Infrared Spectrometry

Near-Infrared Spectroscopy in Food Science and Technology ◽

10.1002/9780470047705.ch6 ◽

2006 ◽

pp. 145-162

Author(s):

Alfred A. Christy ◽

Olav M. Kvalheim

Keyword(s):

Latent Variable ◽

Multivariate Data ◽

Infrared Spectrometry ◽

Latent Variable Analysis ◽

Variable Analysis

Download Full-text

Analytical Profiling and Latent-Variable Analysis of Multivariate Data in Petroleum Geochemistry

The Handbook of Environmental Chemistry - Chemometrics in Environmental Chemistry - Applications ◽

10.1007/978-3-540-49150-7_4 ◽

1995 ◽

pp. 87-133 ◽

Cited By ~ 1

Author(s):

Olav M. Kvalheim ◽

Alfred A. Christy

Keyword(s):

Latent Variable ◽

Multivariate Data ◽

Latent Variable Analysis ◽

Petroleum Geochemistry ◽

Variable Analysis

Download Full-text

Partial Correlation-Based Attention for Multivariate Time Series Forecasting

Proceedings of the AAAI Conference on Artificial Intelligence ◽

10.1609/aaai.v34i10.7132 ◽

2020 ◽

Vol 34 (10) ◽

pp. 13720-13721

Author(s):

Won Kyung Lee

Keyword(s):

Time Series ◽

Partial Correlation ◽

Time Series Data ◽

Multivariate Time Series ◽

Time Series Forecasting ◽

Data Driven ◽

Series Data ◽

Time Lags ◽

Dependency Structure ◽

Agnostic Learning

A multivariate time-series forecasting has great potentials in various domains. However, it is challenging to find dependency structure among the time-series variables and appropriate time-lags for each variable, which change dynamically over time. In this study, I suggest partial correlation-based attention mechanism which overcomes the shortcomings of existing pair-wise comparisons-based attention mechanisms. Moreover, I propose data-driven series-wise multi-resolution convolutional layers to represent the input time-series data for domain agnostic learning.

Download Full-text