Oil price volatility and real options: 35 years of evidence

• Published date: 01 December 2019
• Author: John Elder
• Date: 01 December 2019
• DOI: http://doi.org/10.1002/fut.22057

J Futures Markets. 2019;39:1549–1564. wileyonlinelibrary.com/journal/fut © 2019 Wiley Periodicals, Inc.

|

1549

Received: 28 December 2018

|

Accepted: 26 August 2019

DOI: 10.1002/fut.22057

RESEARCH ARTICLE

Oil price volatility and real options: 35 years of evidence

John Elder

Department of Finance and Real Estate,

Colorado State University, Fort Collins,

Colorado

Correspondence

John Elder, Department of Finance and

Real Estate, Colorado State University,

Fort Collins, CO 80528‐1272.

Email: john.elder@colostate.edu

Abstract

There has been a surge in interest in the effects of uncertainty on investment

decisions, motivated at least in part by the theory of real options. For example,

Bloom (2009, Econometrica,77, 623–685) shows that higher uncertainty causes

firms to temporarily pause investment and hiring, generating sharp economic

downturns. This paper investigates these effects by examining the response of

disaggregated measures of production to volatility in oil prices. We find that

increased oil price volatility has strong negative effects on the production of

durable goods, such as transportation equipment, and oil exploration, such as

the drilling of oil and gas wells.

KEYWORDS

oil volatility, real options, uncertainty

JEL CLASSIFICATION

E32; G17; G31

1

|

INTRODUCTION

There has been a resurgence of interest in the effects of uncertainty on economic activity. Bloom (2009), in a general

equilibrium model, shows that higher uncertainty causes firms to temporarily pause investment and hiring,

contributing to sharp but brief economic downturns. He finds empirical support for this model with aggregate measures

of uncertainty. Bloom, Bond, and Van Reenen (2007) find that aggregate uncertainty, such as the 1973 oil crisis, reduces

the responsiveness of investment to demand shocks. Bansal and Yaron (2004) show that uncertainty risk may increase

precautionary savings and decrease consumption. Panousi and Papanikolaou (2012) find that high uncertainty tends to

be associated with lower investment in a panel of firms. Baker, Bloom, and Davis (2016) show that uncertainty about

economic policy tends to precede declines in investment and output. Wang, Xu, and Zhong (2018) examine the effect of

economic policy uncertainty on CDS spreads.

An important channel for the effects of uncertainty is the theory of real options. This theory suggests that an increase

in uncertainty about the return to an underlying asset tends to increase the time value of a call option on that asset,

delaying exercise. Uncertainty may then diminish the willingness of individual firms to commit resources to irreversible

investments and, similarly, the willingness of consumers to commit resources on relatively illiquid durables. This was

shown for financial options by Black and Scholes (1973), and popularized in capital budgeting by Dixit and Pindyck

(1994) and others. Giaccotto, Gerson, and Shantaram (2007) analyze real options in the context of a durable good.

The empirical evidence highlighting the importance of uncertainty for aggregate production and investment

decisions is abundant, but empirical evidence at the industry and firm‐level is less so. Moel and Tufano (2002) examine

the behavior of Brazilian mine closings. Grullon, Lyandres, and Zhdanov (2012) examine the implication of real options

for firm‐level returns and volatility. Kellogg (2014) finds oil price volatility causes drilling activity in Texas to decline.

Ritz and Walther (2015) examine the response of banks to funding uncertainty. Carvalho (2018) analyzes the effects of

political uncertainty in Brazil on manufacturing firms.

We extend this literature by using relatively granular and comprehensive industry‐level production data to explore

the theoretical predictions from the real options literature. In particular, we examine the effect of uncertainty on the

subindices of industrial production (IP), as compiled by the Federal Reserve Board of Governors. The broad index

measures domestic industrial output, which accounts for about one‐fifth of U.S. gross domestic product. The subindices

measure production at the two, three, four and five digit North American Industry Classification System (NAICS), as

published by the U.S. Office of Management and Budget. At the three‐digit level, IP is attributed to more than two

dozen industries. As an example of the granularity provided by this data, transportation equipment (NAICS 336)

provides separate categories for the production of motor vehicles (NAICS 3361), which is divided into automobile and

light‐duty motor vehicle (33611) and heavy‐duty trucks (33612), which are distinguished from motor vehicle body and

trailer (3362), motor vehicle parts (3363), and other subindices. This permits a focused empirical analysis with the

opportunity to investigate theoretical implications that are not possible with more aggregated data.

1

We use a statistical measure of uncertainty based on oil prices, as in Bredin, Elder, and Fountas (2011). Alternative

measures of uncertainty include survey data (Bachmann, Elstner, & Sims, 2013) and newspaper citations (Baker et al.,

2016). Measuring uncertainty from commodity prices has some advantages. They are easily quantifiable, market‐based,

and there exists considerable evidence that oil prices have strong real effects, at least since Hamilton (1983). They are

also closely related to other measures of uncertainty, as Bloom (2009) shows that measures of uncertainty based on

newspaper citations are linked to measures based on commodity prices.

Importantly, our empirical model permits oil price volatility to affect production independently of the relation between oil

prices and production. For example, lower oil prices may tend to increase manufacturing production, while greater oil price

volatility may tend to simultaneously dampen manufacturing production. Our model also accommodates production activity

related to oil exploration, such as the drilling of oil and gas wells. In these industries, we would expect lower oil prices to be

associated with less drilling, while greater oil price volatility would tend to reinforce this decline.

Other authors who have examined the role of oil prices in uncertainty and risk include Pindyck (2004), Kellogg

(2014), and Chiang, Hsuan, Hughen, and Sagi (2015). Gao, Hitzemann, Shaliastovich, and Xu (2018) develop a model

that details a mechanism by which firms choose not to invest during periods of high oil price volatility. P2012m develop

a real business cycle model incorporating oil price volatility, finding that oil price volatility produces a temporary

decrease in durables. Elder and Serletis (2010) find that volatility in oil prices has tended to depress some components

of quarterly aggregate investment and durables consumption.

Our empirical results are strong. We find that oil price volatility tends to depress durables production at the two,

three, four, and five digit NAICS industry levels. The decline in output is particularly large for industries that are highly

dependent on energy prices, such as the production of transportation equipment. We also find that oil price volatility

has a negative and significant effect on oil and gas exploration, including the drilling of wells. The magnitude of the

effect we estimate is broadly consistent with predictions from theory, such as Bloom (2009). In particular, we estimate

that a volatility shock at the 90th percentile reduces durables production by about 2.3 percentage points at an annual

rate. The effects in some sectors, such as transportation, is even larger. Our results are robust to changes in the

empirical specification, including different sample periods and different measures of oil volatility, such as stochastic

volatility and implied volatility from option prices.

2

|

EMPIRICAL MODEL

Our empirical model is based on Bredin et al. (2011) and Elder (2004), and is briefly summarized here. The model is

structural vector autoregression, generalized to permit the structural disturbances to be conditionally heteroskedastic

and to permit the conditional variance to interact with the conditional mean

tBy C ΓyΓyΓyH ε= + + + ··· + + Λ() + ,

ttt ptp t1–1 2–2 – 1/2

(1)

where dim(B) = dim(Γ

i

) = (N × N), ε

t

∣ψ

t–1

~iid N(0,H

t

), H

t

is diagonal, and ψ

t–1

denotes the information set at time

t–1, which includes variables dated t–1 and earlier.

1

Guo and Kliesen (2005) is an insightful early paper that examines the effects of oil volatility on aggregated IP. Kliesen (2008), Rentschler (2013) andPinno and Serletis (2013) also examine the effects

of oil volatility, although they examine a small subset of the production measures considered here with different empirical models. For example, these papers do not include more granular data on

motor vehicles and the drilling of oil and gas wells.

1550

|

ELDER