Elsevier

Energy Economics

Volume 83, September 2019, Pages 430-444
Energy Economics

Can stale oil price news predict stock returns?

https://doi.org/10.1016/j.eneco.2019.07.022Get rights and content

Highlights

  • We hand-collect time-series data on positive and negative oil price news.

  • Data come from 100 global news sources, covering 59,129 news articles on oil prices.

  • Positive and negative news predict stock returns for at most 12 countries.

  • Together all three oil price measures predict returns for at most 23/45 countries.

  • Discount rate and cash flow channels help explain how oil predicts returns.

Abstract

We hand-collect time-series data on positive and negative oil price news from 100 news sources from around the world, covering 59,129 news articles on oil prices. Using time-series predictive regression models estimated for 45 countries, we show that: (a) positive and negative news predict stock returns for at most 12 countries for which the oil price does not predict returns; and (b) together the three oil price measures predict returns for at most 23/45 countries. Therefore, oil price news turns out to be more powerful in predicting returns in a horserace with oil price. We show that the ability of oil to predict returns is through the discount rate and cash flow channels. Our results survive a battery of robustness tests.

Introduction

This paper examines the role that stale financial news on oil plays in moving stock returns. There is a large and growing literature, which examines whether or not oil prices predict stock market returns. The evidence is conclusive in that it supports that oil prices either negatively or positively predict stock returns. Our aim is different from this literature. We focus on stale oil price news and compare its predictive ability with that of oil prices. In other words, we undertake a horserace between oil prices on the one hand and stale positive and negative news on oil prices on the other. We consider oil price news reported by newspapers and newswire sources as stale news for two reasons. First, news reported in newspapers is typically already known via other sources such as the internet (in particular, social media). Second, even if news reported in newspapers was fresh, just because we aggregate daily news to obtain a monthly oil price news sentiment, what we end up having on hand is stale news.1

Our motivation for considering oil price news has roots in a recent body of influential literature that shows that stale financial news (positive and negative word-based) influences stock market returns; see, inter alia, Garcia (2013), Tetlock (2007), Tetlock et al. (2008), and Gurun and Butler (2012). There is a growing view that stale financial news effects stock markets. Yet we know virtually nothing about how stale oil price news effects the stock markets. It is important to consider this because there is a growing number of studies that show that oil price influences the stock market (see Driesprong et al., 2008; Narayan and Sharma, 2011; Kang et al., 2015; Kilian and Park, 2009).2 If oil price matters to stock markets then news (related to oil prices) should have more information content to influence stock markets as the broader financial news—stock returns literature has demonstrated. We therefore fill an important gap in this literature.

Specifically, our empirical approach follows three steps. First, we hand-collect oil price news data. We search for oil price news articles from a global database of published news articles. We end up with a total of 59,128 news articles on oil prices over the period 1995 to 2014. This news comes from 100 different newspapers and newswire services, both published in print and online, from around the world. We then use the Loughran and McDonald (2011) word dictionary that defines positive words and negative words, and extract from the 59,128 news articles positive and negative oil price news. The distinction between positive and negative financial news relating to oil is important because countries that are net oil exporters will be affected differently from positive and negative news compared to countries that are net oil importers. For example, negative news that reduces oil prices (such as the first news article in Fig. I) will benefit net importer than net exporter countries while positive news that increases oil prices (such as the second news article in Fig. I) will have the opposite effects.

Second, we run predictive regression models treating oil price news (negative and positive) as a predictor of country stock excess returns. We end up comparing the ability of news to predict stock returns with the predictive ability of oil price. We apply predictive regression models to as many as 45 countries for which consistent time-series monthly stock price index data are available. Third, we focus on the robustness of the effect of oil price news on stock returns by: (a) adjusting stock returns for commonly known risk factors; (b) using different specifications for predictive regression models; (c) including contemporaneous variables in predictive regression models; (d) modelling outliers; (e) modelling time-varying heteroscedasticity; (f) accounting for structural breaks; (g) considering newswire source of news as distinct from a newswire and newspaper combined source; and (h) controlling for the 2018 Global Financial Crisis (GFC).

Our approaches contribute the following findings. We find that while both positive and negative news predict returns, they (a) do not predict returns for all countries and (b) positive news predicts returns more than negative news. Our empirical analysis reveals that positive news (negative news) predicts risk-adjusted returns for 10 (5) countries. Positive news moves returns by 0.67% to 3.10% per month, while the magnitude of the effect of negative news is smaller, in the 0.65% to 1.72% range.

Our article contributes to four literatures. One is the stock return predictability literature, which has used a large number of financial ratios and macro variables, considered as the traditional predictors of returns; see, inter alia, Campbell and Thompson (2008) and Rapach et al. (2010). Recent literature has entertained non-traditional predictors though. For example, Hsu, 2009 shows that technological innovations predict stock returns; Rapach et al. (2010) show that technical trading rules predict returns; Rapach et al. (2013) find that the U.S. market returns predict returns of selected industrialized countries; and Chava et al. (2015) find that credit standards predict returns.

The study most closely related to ours is Driesprong et al. (2008), Narayan and Sharma (2011), and Sim and Zhou (2015), who show that oil prices predict stock returns. We join this literature on non-traditional predictors of returns by showing that stale oil price news, as reported by leading newspapers and newswire services from around the world, predicts stock returns. In this regard, our findings complement those obtained by studies on the oil price and stock returns nexus, however, the key difference is that we contribute from the stale news perspective. More specifically, the contribution of our paper here is not that we show, like this literature does, that oil prices predict stock returns; rather, we show that oil price news predicts returns for more countries in our sample compared to oil price.

Our second contribution is to the literature on news (financial news-based on word count) and stock returns more generally. This literature is popular and growing in significance as illustrated in the works of, inter alia, Garcia (2013), Tetlock (2007), Tetlock et al. (2008), and Gurun and Butler (2012). These studies show that word-count based financial news acts as an important source of information in predicting stock returns. We join this literature by showing that not only financial news is helpful in predicting stock returns as these studies so well document, but oil price news (word count-based) is also a source of information for the equity markets. The main implication of this finding is that word count-based oil price news can now be considered as a predictor of stock returns, not for all countries but for sufficient countries in our sample to generate interest.

Our third contribution is to the literature on the effect of stale news. A subset of the literature on the effect of news on stock returns shows that stale news matters for stock returns; see studies such as Huberman and Regev (2001) and Tetlock (2011). To summarize, these studies show that investors trade on stale news. Tetlock (2011) explains how this relationship between stale news and stock returns unfolds. He argues (p. 1481) that: (a) stale news contributes to the speed and volume of information dissemination which subsequently is a source of information efficiency; and (b) readers of stale news have no knowledge of how market participants have reacted to the same news when it was first released, leading them to overreact to stale news. The main message of our empirical analysis is that investors' trade on stale oil price news too, suggesting, like Tetlock (2011) observes, that stales news boosts the speed and volume of information dissemination and it instigates investor overreaction. Therefore, while we already knew that investors traded on stale financial news we now know that they also trade on stale oil price news.

Our final contribution relates to the literature on the determinants of asset pricing. There is a lack of consensus here: studies have generally concluded with mixed evidence on whether stock returns are best characterized by dividend yield forecasts—that is, discount rate channel—or dividend growth—that is, cash flow channel. In this literature, the discount rate channel has accumulated most evidence and is, therefore, more popular. However, a recent finding by Huang, Jiang, Tu, and Zhou (2015) that the source of predictability is through the cash flow channel has reinvigorated the debate. With our analysis, we enter this debate to show that from the perspective of oil price growth rate and oil price positive and negative news the bulk of the evidence on the source of predictability supports the discount rate channel. For example, in 15 countries we find that oil price predictors predict cash flow proxies whereas in only five countries oil price predictors predict the discount rate proxy. It follows that the evidence we provide corroborates those of Huang et al. (2015).

The rest of the paper is organized as follows. Section 2 discusses the data and presents preliminary evidence on the relationship between oil price and stock returns and oil price news and stock returns. This descriptive story is based on three types of returns—namely, raw returns, excess returns, and the Fama and French (1993) risk-adjusted returns. Section 3 presents the main results on the importance of oil price and oil price news on stock return predictability. Section 4 presents results from a range of robustness tests aimed at confirming our main finding on the importance of oil price news (and oil price) in predicting returns.

Section snippets

Data and preliminary evidence

This section has two parts. In the first part, we discuss the data. The main part of this data section is about our oil price news which we hand-collect. We explain this. In the second part, we look at some of the statistical features of our data, in particular the potential relationship between oil price news and stock market returns.

Statistical results

We begin with evidence on return predictability obtained when using excess unadjusted returns. The three predictors are the conventional time-series of oil price (growth rate) and our two news variables—positive oil price news and negative oil price news. We first run time-series predictive regressions of the form:Rt=α+δii=13Oilti+ϑii=13Rti+θii=13VRti+εt

Eq. (1) is motivated by the framework used in Garcia (2013). Here Rt is the excess stock market returns and Oilt denotes either the

Robustness tests

Thus far, we have confirmed that our main results on the importance of oil price news holds in different regression specifications based on adjusting excess returns for risk factors. In this section we undertake a range of additional robustness tests to confirm our main findings. First, we examine whether the in-sample evidence of predictability also holds in out-of-sample tests. Second, we examine whether accounting for contemporaneous effects influences our findings on predictability. Third,

Concluding remarks

This paper is based on 19 years of data on stale oil price news, hand-collected from 100 different news sources from around the world, covering 59,129 news articles on oil prices. This constitutes the first ever and the largest quantitative record of oil price news, allowing for an analysis of the predictive ability of stale oil price news. Using time-series predictive regression models estimated for 45 countries, we show that positive and negative news predict stock returns for 12 countries

References (42)

  • S.P. Kothari et al.

    Stock return variation and expected dividends: a time-series and cross-sectional analysis

    J. Financ. Econ.

    (1992)
  • P.K. Narayan et al.

    Are Indian stock returns predictable?

    J. Bank. Financ.

    (2015)
  • P.K. Narayan et al.

    New evidence on oil price and firm returns

    J. Bank. Financ.

    (2011)
  • P.K. Narayan et al.

    Can institutions and macroeconomic factors predict stock returns in emerging markets?

    Emerg. Mark. Rev.

    (2014)
  • P.K. Narayan et al.

    Can governance quality predict stock market returns? New global evidence

    Pac. Basin Financ. J.

    (2015)
  • N. Sim et al.

    Oil prices, US stock return, and the dependence between their quantiles

    J. Bank. Financ.

    (2015)
  • J. Bai et al.

    Estimating and testing linear models with multiple structural changes

    Econometrica

    (1998)
  • J. Campbell et al.

    The dividend-price ratio and expectations of future dividends and discount factors

    Rev. Financ. Stud.

    (1988)
  • J.Y. Campbell et al.

    Predicting the equity premium out of sample: can anything beat the historical average?

    Rev. Financ. Stud.

    (2008)
  • J. Campbell et al.

    Growth or glamour? Fundamentals and systematic risk in stock returns

    Rev. Financ. Stud.

    (2010)
  • J. Cochrane

    The dog that did not bark: A defense of return predictability

    Rev. Financ. Stud.

    (2008)
  • Cited by (75)

    • Understanding mispricing in the travel and leisure industry

      2023, International Review of Financial Analysis
    • International spillovers of U.S. monetary uncertainty and equity market volatility to China's stock markets

      2023, Journal of Asian Economics
      Citation Excerpt :

      From a global perspective, detecting the direction of causality among these variables has great value; it can offer insights into portfolio diversification and demonstrate how fluctuations in U.S. monetary policy impact China’s stock prices. When exploring how monetary policy causes stock price fluctuations, the financial theory postulates that asset prices reflect expectations regarding the discounted values of firms’ future cash flows, which are frequently affected by macroeconomic fluctuations (Lee et al., 2019; Lee et al., 2022; Narayan, 2019; Narayan et al., 2022). Any monetary shock affects investors’ expectations, which cause fluctuations in stock prices (Adediran & Akpa, 2022; Fama & Schwert, 1977; Gangopadhyay, 2008).

    View all citing articles on Scopus

    An earlier version of this paper was presented as a Keynote speech at the 2017 International Conference on Energy Finance at the X'ian University of Technology (Xian, China).

    View full text