Skip to main content
SpringerLink
Log in

Daily volume, intraday and overnight returns for volatility prediction: profitability or accuracy?

  • Original Research
  • Published:
Review of Quantitative Finance and Accounting Aims and scope Submit manuscript

Abstract

This article presents a comprehensive analysis of the relative ability of three information sets—daily trading volume, intraday returns and overnight returns—to predict equity volatility. We investigate the extent to which statistical accuracy of one-day-ahead forecasts translates into economic gains for technical traders. Various profitability criteria and utility-based switching fees indicate that the largest gains stem from combining historical daily returns with volume information. Using common statistical loss functions, the largest degree of predictive power is found instead in intraday returns. Our analysis thus reinforces the view that statistical significance does not have a direct mapping onto economic value. As a byproduct, we show that buying the stock when the forecasted volatility is extremely high appears largely profitable, suggesting a strong return-risk relationship in turbulent conditions.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Log in via an institution

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Institutional subscriptions

Fig. 1
Fig. 2

Similar content being viewed by others

Notes

  1. The expression ‘volatility-based technical trading’ is used here to denote trading strategies that are based on buy/sell signals implied from volatility forecasts. This differs from what is called ‘volatility trading’ in the literature, namely, trading strategies that treat volatility, i.e. the VIX index and, more recently, VIX futures, as an asset class (see Hafner and Wallmeier 2007; Konstantinidi et al. 2008). It also differs from ‘volatility timing’ which refers to the use of (co)variance forecasts in dynamic optimal portfolio construction (see Fleming et al. 2003).

  2. Two stocks are listed on Nasdaq (DELL and MSFT) while the remaining are NYSE listed.

  3. We deploy the Ljung–Box statistic to test the null hypothesis of no residual autocorrelation. For 9 stocks the conditional mean equation in (1a) is appropriate. For the remaining 6 stocks (ATT, DELL, GM, IBM, PG and WMT) we employ instead an ARMA(p, q) equation \(r_{t}=\mu +\sum\nolimits_{i=1}^{p} \theta_{i}r_{t-i}+\sum\nolimits_{j=1}^{q}\lambda_{j}\varepsilon_{t-j}\) with appropriate orders p and q so as to whiten the residual sequence. The GARCH equation (2b) for CAT, JPM, KO and MCD has lags r = 2 and s = 1 whereas for all other stocks a GARCH (1,1) suffices to absorb the autocorrelation in squared daily returns.

  4. Thus the out-of-sample period comprises about 1/4 of the total sample as in Koopman et al. (2005), Liu and Maheu (2009), Ghysels et al. (2007) and Fuertes et al. (2009).

  5. For space constraints, we do not report the detailed statistics here but they are tabulated in Appendices A1 and A2 of the longer working paper version of this article (see Fuertes et al. 2013).

  6. A description of these statistical metrics can be found in Fuertes et al. (2009). The R 2 of the Mincer–Zarnowitz levels regression (MZ-R 2 henceforth), \(\tilde{\sigma}_{t}^{2}=a+bh_{t,m}+e_{t},t=1,\ldots,500\), measures the information content of the forecasts; h t,m is unbiased for \(\tilde{\sigma}_{t}^{2}\) if a = 0 and b = 1.

  7. Results for the remaining 11 stocks are qualitatively similar and not reported to preserve space.

  8. For nested models the DM test statistic is non-Gaussian, resulting in undersized tests with low power, so the results are interpreted with caution. The ENC-T test null is that model A encompasses model B and the alternative is that model B contains additional predictive information.

  9. Bali and Peng (2006) are the first to use daily realized variances to examine the risk-return link for the aggregate stock market. In addition, they use risk measures obtained from GARCH models estimated with 5-min returns and daily implied volatilities. All three risk measures suggest that the intertemporal risk-return relation is positive and statistically significant.

  10. For details on the PCSE methodology see Beck and Katz (1995). Wooldridge’s panel test statistic for zero autocorrelation in the residuals of the quintile regressions is insignificant throughout.

  11. For the GARCH model augmented with the squared overnight return, all trading strategies are deployed using the 10:00 a.m. price (instead of the opening 9:30 a.m. price) as the buy or sell price. This is because it is not feasible for an investor to trade on day t + 1 at the open price if that price is precisely required for the GARCH-OVN model to generate the day t + 1 volatility forecast.

  12. The 5-day SMA is created as the simple moving average of day t − 1 to t − 5 closing prices, \(SMA_{t}=\frac{P_{t-5}+P_{t-4}+\cdots+P_{t-1}}{5}\). The 5/20-day DCMA, often suggested to identify short term trends in prices (see Pring 2002) combines a weekly and a monthly trend: if the weekly SMA falls below the monthly SMA then a stop-loss (i.e., sell) signal is generated.

  13. Elkins-McSherry Report, Vol. II (2), May 2005, available at http://www.elkinsmcsherry.com.

  14. Out of 70 settings, only a handful of them show a discrepancy between the SoR and α metrics. In those few exceptions, we proceed by following the metric with the largest incremental gain. For instance, for IBM stock and the Directional-SMA-DCMA strategy, GARCH-VOL is best according to α and GARCH-RPV according to SoR but the incremental gain (vis-à-vis the standard daily-based GARCH model) in α exceeds that in SoR; hence, we count GARCH-VOL as winner.

  15. For completeness, we computed incremental end-of-period values, incremental annualized returns and standard deviation of returns generated by actively investing $100 over the holdout period on the basis of augmented GARCH model forecasts. The different information sets afford positive incremental returns that range from 0.2 to 24.1 % per annum; trading volume remains in the lead. See Appendix A3 in the working paper version of this article (Fuertes et al. 2013).

  16. In each stock-strategy competition reported in Table 5 the best information set is defined as the one leading to the highest switching fees for risk aversion parameter γ = 1 or γ = 10.

  17. More detailed results on the comparison of the strategies with the B&H can be found in Appendix A4 of the working paper version of this article (Fuertes et al. 2013).

  18. We computed the statistical criteria reported in Table 1 separately over high volatility (top quintile) and low volatility (bottom quintile) out-of-sample days. RPV remains in the lead from the point of view of statistical accuracy for extremely low and high volatility days. Conversely, trading volume incurs poor statistical forecast accuracy especially for extreme volatility days.

  19. Each graph contains 60 = 4 × 15 observations corresponding to four models (standard GARCH and three augmented versions) and fifteen assets (14 individual stocks and S&P500 portfolio).

  20. Kendall’s tau is based on the number of concordances and discordances between the variables. If the number of concordances and discordances are roughly the same for all observations, there is no association between the variables. Relatively large numbers of concordances (discordances) suggest a positive (negative) relationship between the variables.

  21. Similar findings emerge from rank correlation measures and scatterplots using the remaining statistical loss functions reported in Table 1 instead of the MSE. Details available upon request.

  22. This disconnect has been documented in other contexts. Abhyankar et al. (2005) show that a utility-based value approach reverses the previous empirical consensus that monetary-fundamental models cannot beat the random walk. Hall and Sirichand (2010) compare the forecast performance of an atheoretic and a theory-informed model of bond returns for portfolio decision making and illustrate the sensitivity of the ranking to the evaluation criteria.

References

  • Abhyankar A, Sarno L, Valente G (2005) Exchange rates and fundamentals: evidence on the economic value of predictability. J Int Econ 66:325–48

    Article  Google Scholar 

  • Andersen T, Bollerslev T (1998) Answering the skeptics: yes, standard volatility models do provide accurate forecasts. Int Econ Rev 39:885–906

    Article  Google Scholar 

  • Bali TG, Peng L (2006) Is there a risk-return trade-off? Evidence from high-frequency data. J Appl Econ 21:1169–98

    Article  Google Scholar 

  • Barndorff-Nielsen OE, Shephard N (2004) Power and bipower variation with stochastic volatility and jumps. J Financ Econ 2:1–37

    Google Scholar 

  • Beck N, Katz JN (1995) What to do (and not to do) with time-series cross-section data. Am Pol Sci Rev 89:634–647

    Article  Google Scholar 

  • Bessembinder H (2003) Issues in assessing trade execution costs. J Financ Mark 6:233–257

    Article  Google Scholar 

  • Blair JB, Poon SH, Taylor SJ (2001) Forecasting S&P100 volatility: the incremental information content of implied volatilities and high frequency index returns. J Financ Econ 105:5–26

    Google Scholar 

  • Brooks C (1998) Predicting stock index volatility using GARCH models: can market volume help. J Forecast 17:57–80

    Article  Google Scholar 

  • Brownlees C, Gallo G (2010) Comparison of volatility measures: a risk management perspective. J Financ Econ 8:29–56

    Google Scholar 

  • Christoffersen PF, Diebold FX (2006) Financial asset returns, direction-of-change forecasting, and volatility dynamics. Manag Sci 52:1273–1287

    Article  Google Scholar 

  • Corrado CJ, Lee SH (1992) Filter rule tests of the economic significance of serial dependence in daily stock returns. J Financ Res 154:369–387

    Article  Google Scholar 

  • Diebold FX, Mariano R (1995) Comparing predictive accuracy. J Bus Econ Stat 13:253–263

    Google Scholar 

  • Donaldson G, Kamstra M (2005) Volatility forecasts trading volume and the ARCH versus option-implied volatility trade-off. J Financ Res 27:519–538

    Article  Google Scholar 

  • Fama E (1970) Efficient capital markets: a review of theory and empirical work. J Financ 25:383–417

    Article  Google Scholar 

  • Fleming J, Kirby C, Ostdiek B (2003) The economic value of volatility timing using ‘realised’ volatility. J Financ Econ 67:473–509

    Article  Google Scholar 

  • Forsberg L, Ghysels E (2007) Why do absolute returns predict volatility so well? J Financ Econometrics 5:31–67

    Google Scholar 

  • Fuertes AM, Izzeldin M, Kalotychou E (2009) On forecasting daily stock volatility: the role of intraday information and market conditions. Int J Forecast 25:259–281

    Article  Google Scholar 

  • Fuertes AM, Kalotychou E, Todorovic N (2013) Volume, intraday or overnight information for volatility-based technical trading? SSRN Working Paper 1438041

  • Fuertes AM, Olmo J (2012) Optimally harnessing inter-day and intra-day information for value-at-risk prediction. Int J Forecast 29:28–42

    Article  Google Scholar 

  • Gallo G (2001) Modelling the impact of overnight surprises on intra-daily volatility. Aust Econ Pap 40:567–580

    Article  Google Scholar 

  • Ghysels E, Sinko A, Valkanov R (2007) MIDAS regressions: further results and new directions. Econ Rev 26:53–90

    Article  Google Scholar 

  • Hafner R, Wallmeier M (2007) Volatility as an asset class: European evidence. Eur J Financ 13:621–644

    Article  Google Scholar 

  • Hall S, Sirichand K (2010) Decision-based forecast evaluation of UK interest rate predictability. University of Leicester Working Paper 10/09

  • Harvey DI, Leybourne SJ, Newbold P (1998) Tests for forecast encompassing. J Bus Econ Stat 16:254–259

    Google Scholar 

  • Harvey CR, Whaley RE (1992) Market volatility prediction and the efficiency of the S&P 100 index option market. J Financ Econ 31:43–73

    Article  Google Scholar 

  • Jacobs M, Onochie J (1998) A bivariate GARCH-in-mean study of the relationship between return variability and trading volume in international futures markets. J Futures Mark 18:379–397

    Article  Google Scholar 

  • Karpoff JM (1987) The relation between price changes and trading volume: a survey. J Financ Quant Anal 22:109–126

    Article  Google Scholar 

  • Koopman SJ, Jungbacker B, Hol E (2005) Forecasting daily variability of the S&P 100 stock index using historical, realised and implied volatility measurements. J Empir Financ 12:445–475

    Article  Google Scholar 

  • Konstantinidi E, Skiadopoulos G, Tzagkaraki E (2008) Can the evolution of implied volatility be forecasted? Evidence from European and US implied volatility indices. J Bank Financ 32:2401–2411

    Article  Google Scholar 

  • Larsen M (2004) A hands-on approach to volatility trading. Futures Magazine, Sept 2004

  • Lasky P (2001) By Garch, volatility trading works. Futures Magazine, July 2001

  • Lee DD, Chan H, Faff RW, Kalev P (2003) Short-term contrarian investing: it is profitable—yes and no. J Multin Financ Manag 13:385–404

    Article  Google Scholar 

  • Leland H (1999) Beyond mean-variance: risk and performance measures for portfolios with nonsymetric distributions. Financ Anal J 55:27–36

    Article  Google Scholar 

  • Liu C, Maheu J (2009) Forecasting realised volatility: a Bayesian model-averaging approach. J Appl Econ 24:709–733

    Article  Google Scholar 

  • Merton RC (1973) An intertemporal capital asset pricing model. Econometrica 41:867–887

    Article  Google Scholar 

  • Najand M, Yung K (1991) A GARCH examination of the relationship between volume and price variability in futures market. J Futures Mark 11:613–621

    Article  Google Scholar 

  • Noh J, Engle RF, Kane A (1994) Forecasting volatility and option prices of the S&P 500 index. J Deriv 2:17–30

    Article  Google Scholar 

  • Northington K (2009) Volatility-based technical analysis: strategies for trading the invisible. Wiley, New York

    Google Scholar 

  • Pesaran MH, Timmermann A (1992) A simple nonparametric test of predictive performance. J Bus Econ Stat 10:561–65

    Google Scholar 

  • Peterson M, Sirri E (2003) Evaluation of the biases in execution cost estimation using trade and quote data. J Financ Mark 6:259–280

    Article  Google Scholar 

  • Pong S, Shackleton M, Taylor M, Xu X (2004) Forecasting sterling/dollar volatility: a comparison of implied volatilities and AR(FI)MA models. J Bank Financ 28:2541–2563

    Article  Google Scholar 

  • Pring MJ (2002) Technical analysis explained: the successful investor’s guide to spotting investment trends and turning points. McGraw-Hill Professional, New York

    Google Scholar 

  • Rahman S, Lee C-F, Ang KP (2002) Intraday return volatility process: evidence from NASDAQ stocks. Rev Quant Financ Acc 19:155–180

    Article  Google Scholar 

  • Rattray S, Balasubramanian V (2003) The new VIX as a market signal: It still works. Goldman Sachs Research Report, Equity Derivatives Strategy

  • Satchell S, Timmermann A (1995) An assessment of the economic value of non-linear foreign exchange rate forecasts. J Forecast 14:477–497

    Article  Google Scholar 

  • Taylor SJ, Xu X (1997) The incremental volatility information in one million foreign exchange quotations. J Empir Financ 4:317–340

    Article  Google Scholar 

  • Tsiakas I (2008) Overnight information and stochastic volatility: a study of European and US stock exchanges. J Bank Financ 32:251–268

    Article  Google Scholar 

  • Wu C, Xu XE (2000) Return volatility, trading imbalance and the information content of volume. Rev Quant Financ Acc 14:131–153

    Article  Google Scholar 

Download references

Acknowledgments

We would like to thank the journal editor, Cheng-Few Lee, and an anonymous referee for very constructive comments. We also acknowledge feedback from Roy Batchelor, Jerry Coakley, Aneel Keswani, Eirini Konstantinidi, Giampiero Gallo, Jöelle Miffre, Lucio Sarno, and participants at the 2011 GdRE International Annual Symposium on Money, Banking and Finance, University of Reading, United Kingdom, 2010 FMA European Conference, Hamburg, Germany, 2010 BAA Conference, Cardiff, 2009 EFMA Conference, Bocconi University, Italy, 2009 ERCIM-CFE Conference, Cyprus and seminar participants at Essex Business School and Cass Business School.

Author information

Authors and Affiliations

  1. Faculty of Finance, Cass Business School, City University London, 106 Bunhill Row, London, EC1Y 8TZ, UK

    Ana-Maria Fuertes, Elena Kalotychou & Natasa Todorovic

Authors
  1. Ana-Maria Fuertes
    View author publications

    You can also search for this author in PubMed Google Scholar

  2. Elena Kalotychou
    View author publications

    You can also search for this author in PubMed Google Scholar

  3. Natasa Todorovic
    View author publications

    You can also search for this author in PubMed Google Scholar

Corresponding author

Correspondence to Elena Kalotychou.

Rights and permissions

Reprints and permissions

About this article

Cite this article

Fuertes, AM., Kalotychou, E. & Todorovic, N. Daily volume, intraday and overnight returns for volatility prediction: profitability or accuracy?. Rev Quant Finan Acc 45, 251–278 (2015). https://doi.org/10.1007/s11156-014-0436-6

Download citation

  • Received:

  • Accepted:

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1007/s11156-014-0436-6

Keywords

JEL Classification

Search

Navigation