On perceptions of financial volatility in price sequences

ABSTRACT

Stock prices in financial markets rise and fall, sometimes dramatically, thus asset returns exhibit volatility. In finance theory, volatility is synonymous with risk and as such represents the dispersion of asset returns about their central tendency (i.e. mean), measured by the standard deviation of returns. When individuals make investment decisions, influenced by perceptions of risk and volatility, they commonly do so by examining graphs of historic price sequences rather than returns. It is unclear, therefore, whether standard deviation of return is foremost in their mind when making such decisions. We conduct two experiments to examine the factors that may influence perceptions of financial volatility, including standard deviation along with a number of price-based factors. Also of interest is the influence of price sequence regularity on perceived volatility. While standard deviation may have a role to play in perception of volatility, we find evidence that other price-based factors play a far greater role. Furthermore, we report evidence to support the view that the extent to which prices appear irregular is a separate aspect of volatility, distinct from the extent to which prices deviate from central tendency. Also, while partially correlated, individuals do not perceive risk and volatility as synonymous, though they are more closely related in the presence of price sequence irregularity.

Keywords:

• Volatility
• risk
• price sequence
• mean absolute price change
• standard deviation

JEL Classifications:

• G02
• G11

Acknowledgements

The authors thank the editor and two anonymous reviewers for their helpful comments on an earlier version of this paper. All errors are those of the authors. We are grateful to Kevin Keasey for valuable suggestions during the inception of this research. We also thank participants at the 2015 and 2016 Behavioural Finance Working Group conferences, London, the 2016 Behavioural and Experimental Northeast Cluster (BENC) and the Center for the Economic Analysis of Risk (CEAR) workshop, Durham, and the 2015 Subjective Probability, Utility and Decision Making conference, Budapest, along with seminar participants at the Universities of Cardiff, Gothenburg and Warwick.

Disclosure statement

No potential conflict of interest was reported by the authors.

ORCID

Darren Duxbury http://orcid.org/0000-0002-7285-8474

Barbara Summers http://orcid.org/0000-0002-9294-0088

Notes

1 The study of volatility has long held academic interest and has witnessed many advances over the years, as exemplified by the rapidly growing literature on modelling and forecasting ‘realized volatility’ using intra-day data to obtain more accurate and efficient forecasts. See, for example, the many papers published in the ‘realized volatility’ special issue of the Journal of Econometrics edited by Meddahi, Mykland, and Shephard (2011) and more recent studies including Fuertes, Kalotychou, and Todorovic (2015), Andrada-Félix, Fernández-Rodríguez, and Fuertes (2016) and Kourtis, Markellos, and Symeonidis (2016), among others.

2 Earlier, Parkinson (1980) and Kunitomo (1992) propose price-based, extreme value methods for estimating volatility. Such models are shown to provide more efficient volatility estimators than commonly used return-based estimators such as standard deviation.

3 While it is common to use the terms risk and volatility interchangeably in the finance literature, we do not do so here. Our approach is not to adopt specific definitions of risk or volatility, but to let participants reveal, via their ratings of graphical price sequences, what these concepts mean to them. That is, their perceptions of risk and of volatility, whatever they may be. We are then interested in finding which characteristics of the price sequences, individually or in combination, best explain the experimental data. Where no confusion arises we use the terms perception and rating interchangeably when discussing risk and volatility. When reviewing other studies we adopt the nomenclature used in the original study.

4 Gallup poll [http://www.gallup.com/poll/147206/stock-market-investments-lowest-1999.aspx, accessed on 05-09-15].

5 Comparable changes to pension systems have been witnessed in other developed countries such as the UK (Duxbury et al. 2013).

6 The Week Ahead: Greece, China, and the Fed, The Motley Fool, [http://www.fool.com/investing/general/2015/07/06/the-week-ahead-greece-china-and-the-fed.aspx, accessed 04-09-15].

7 Edmond Jackson’s Stockwatch: Is Royal Mail’s growth prospect limited?, Interactive Investor, [http://www.iii.co.uk/articles/168701/edmond-jacksons-stockwatch-royal-mails-growth-prospect-limited, accessed 04-09-15].

8 Many of the price-based factors we examine have their roots in early work by Pinches and Kinney (1971).

9 The exact phrasing used in the experiments was: ‘We would like you to tell us how risky you think these investments are. Please rate each graph on a scale from 0 (no risk at all) to 10 (highest possible risk).’ and ‘We would also like you to tell us how volatile you think the investments are. Please rate each graph on a scale from 0 (not at all volatile) to 10 (extremely volatile).’ Note also, the experimental instrument contained no mention of such terms as ‘dispersion’, ‘standard deviation’, ‘variance’ or any other such statistical term associated with dispersion. Thus, participants were free to adopt their own interpretations of ‘risky’ and ‘volatile’. This was essential given our intention of examining factors that influence perceptions of risk and volatility.

10 See Jiménez-Buedo and Miller (2010) for a convincing argument that the commonly held view of a trade-off between internal and external validity need not hold true. Indeed, they conclude that problems of external or internal validity ‘do not necessarily nor crucially depend … on the artificiality of experimental settings’ (318). It need not be the case, therefore, that external validity, or the generalizability of results, is compromised by the pursuit of experimental control (internal validity).

11 Results are untabulated, but are available from the authors on request.

12 With enhancement the proportion of variation explained by a regression with a particular pair of independent variables is greater than the sum of the proportions of variation explained in regressions with each alone. The terminology in this area is somewhat confusing in that enhancement is also referred to in some literatures as suppression. The intuition behind this name is that the variable that gives rise to the enhancement acts to suppress variance in another variable (say, X₁), enhancing its explanatory power. This comes about because the variable producing enhancement is correlated with elements of X₁ which are not correlated with Y. NumAccelChg actually fulfils the requirement for a classic suppression effect (Horst 1941), having no significant relationship with the outcome itself.

13 Results are untabulated, but are available from the authors on request.

14 Results are untabulated, but are available from the authors on request.

15 As the graphs had been numbered consecutively from 1 in each experiment, numbers for the graphs in experiment two were adjusted to give unique references by adding 20 to each value (so 1 becomes 21, etc.).

16 Results are untabulated, but are available from the authors on request.

17 Results are untabulated, but are available from the authors on request.

18 Similar phenomena have been found in stock market data (Summers, Griffiths, and Hudson 2004).