Abstract
Contrary to point predictions that only convey information about the central tendency of the target variable, or the best prediction, density predictions take into account the whole shape of the conditional distribution, which means that they provide a characterization of prediction uncertainty. They can also be used to assess out-of-sample predictive power when specific regions of the conditional distribution are emphasized, such as the center or the left tail. We carry out an out-of-sample density prediction study for monthly returns on the Standard & Poor’s 500 index from 1859m9 through 2017m12 with a stochastic volatility benchmark and alternatives to it that include the West Texas Intermediate price of crude oil. Results suggest that models employing certain nonlinear transformations of the price of crude oil help deliver statistically significant density prediction improvements relative to the benchmark. The biggest payoff occurs when predicting the left tail of the conditional distribution. They also generate the earliest signal of a market downturn around the 2008 financial crisis.
Similar content being viewed by others
Notes
There is no shortage of tension in the literature when it comes to the choice between in-sample and out-of-sample evaluation. Lo and MacKinlay (1990), Foster et al. (1997) and Rapach and Wohar (2006) among others argue that out-of-sample analysis is relatively more important as in-sample analysis tends to suffer from data mining. Conversely, Inoue and Kilian (2004) argue that in and out-of-sample tests of predictability are equally reliable against data mining under the null hypothesis of no predictability.
This information is also very useful for making economic decisions. For example, the Bank of England publishes its inflation forecast as a probability distribution in the form of a fan chart, see Britton et al. (1998). Likewise, Alessi et al. (2014) explain how using density forecasts, the New York FED produced measures of macroeconomic risk during the 2008 crisis.
The modern era of oil production in the USA begins in 1859 when Edwin Drake succeeds in producing usable quantities of crude oil for commercial purposes from a 69-foot well in Titusville, Pennsylvania. As a result of Drake’s discovery, WTI crude oil price falls from an average of \(\$9.60\) per barrel in 1860 to 10 cents per barrel by the end of 1861, see also Narayan and Gupta (2015).
There are two major crude oil markets: WTI and Brent Blend. WTI is a blend of several US domestic oil fields. Brent Blend is a combination of crude oil from fifteen different oil fields in the North Sea. Both are priced in US dollars. However, monthly time-series of the price of Brent crude oil is only available starting in the early 1950s. An alternative measure of the price crude oil is the price paid by US refiners purchasing crude oil. However, this series is only available starting in 1974m1.
As argued in Alquist et al. (2013) this is explained by the specific regulatory structure of the crude oil industry imposed by the Texas Railroad Commission and other US state regulatory agencies. More precisely, each month these entities would forecast the demand for crude oil and then set the allowable production levels to meet the demand. As a result, much of the cyclically endogenous component of oil demand was reflected in shifts in quantities rather than prices.
The CPI series can be downloaded from Robert J. Shiller’s web site: http://www.econ.yale.edu/~shiller/.
Specifically, in Hamilton (2011), page 370 it is stated: “...deflating by a particular number, such as the CPI, introduces a new source of measurement error, which could lead to deterioration in the forecasting performance. In any case, it is again quite possible that there are differences in the functional form of predictions based on nominal prices instead of real prices”.
We also explore the robustness of our results to different choice of the truncation lag for \(net^{+}\), \(net^{-}\), gap and net. Overall, we observe that results are similar across the different truncation lags.
An obvious issue with (3.1) is that it ignores the possibility of multiple predictors. However, there is a potential issue of multicollinearity if we were to include more predictors, which explains why researchers generally tend to prefer the single predictor model. In fact, as can be found from the large volume of stock return predictability literature, researchers tend to engage in a horse race among all potential predictors, one by one, see for example, Welch and Goyal (2008) and Westerlund and Narayan (2012). In this study, we choose to keep this tradition and use a single predictor in (3.1).
Stambaugh (1999) and Lewellen (2004) show that \(\rho \ne 0\) is a source of major complication in terms of OLS estimation of (3.1). Particularly, the OLS bias is given as: \(-\varphi \left( 1+3\phi \right) /T\). Hence, while decreasing in T, the bias is increasing in \(\varphi \) and \(\phi \), respectively.
Using Monte Carlo simulations, Westerlund and Narayan (2012) document that there are notable gains to be made by accounting for heteroskedasticity in predictive regressions, such as (3.1) and (3.2). Particularly, compared to the constant conditional volatility counterpart, they find that accounting for heteroskedasticity reduces the out-of-sample root mean square error.
As stated in Johannes et al. (2014), the marginal and predictive distributions of returns that integrate out the unobserved log-volatilities are a scale mixture of Normals, which is leptokurtotic.
For instance, \(N^{-1}\Sigma _{i=1}^{N}\psi ^{\left( i\right) }\) is a simulation consistent estimate of the posterior mean of \(\psi \).
Results are robust to different prior hyperparameter values on the model parameters. It is important to note that we have a relatively large sample, and even when we recursively estimate our models in Sect. 5, each estimation window contains 40 years of data (480 monthly observations). Hence, information from the data will tend to dominate information from the prior.
Our decision to adopt the rolling window procedure is motivated by studies, such as Stock and Watson (1996) and Swanson (1998), where it is argued that by allowing the data generating process to evolve over time, a rolling moving window approach can account for parameter instability in the data generating process. Furthermore, as illustrated in Clark and McCracken (2012), under a expanding window approach, test statistics, such as Diebold and Mariano (1995) have a nonstandard limiting distribution.
We follow Groen et al. (2013) and rely on the one-sided Diebold and Mariano (1995) test due to the nested structure of our models. The one-sided test is also more intuitive because we are interested in evaluating the predictive power afforded by employing the price of crude oil in one direction, namely if it adds any.
References
Alessi L, Ghysels E, Onorante L, Peach R, Potter S (2014) Central bank macroeconomic forecasting during the global financial crisis: the European Central Bank and Federal Reserve Bank of New York experiences. J Bus Econ Stat 32:483–500
Alquist R, Kilian L, Vigfusson RJ (2013) Forecasting the price of oil. In: Timmermann A, Elliott G (eds) Handbook of economic forecasting. Elsevier, Amsterdam
Amihud Y, Hurvich CM (2004) Predictive regressions: a reduced-bias estimation method. J Financ Quant Anal 39:813–841
Andrews DWK, Monahan JC (1992) An improved heteroskedasticity and autocorrelation consistent covariance matrix estimator. Econometrica 60:953–966
Balcilar M, Gupta R, Miller SM (2015) Regime switching model of US crude oil and stock market prices: 1859 to 2013. Energy Econ 49:317–327
Britton E, Fisher P, Whitley J (1998) The Inflation Report projections: understanding the fan chart. Bank Engl Q Bull 38:30–37
Carter C, Kohn R (1994) On gibbs sampling for state-space models. Biometrika 81:541–553
Chan J (2013) Moving average stochastic volatility models with application to inflation forecast. J Econom 176:162–172
Chan J (2017) The stochastic volatility in mean model with time-varying parameters: an application to inflation modeling. J Bus Econ Stat 35:17–28
Chan J, Grant AL (2016) Modeling energy price dynamics: GARCH versus stochastic volatility. Energy Econ 54:182–189
Chen SS (2010) Do higher oil prices push the stock market into bear territory? Energy Econ 32:490–495
Chen NF, Roll R, Ross S (1986) Economic forces and the stock market. J Bus 59:383–403
Clark TE, McCracken MW (2012) Advances in forecast evaluation. In: Timmermann A, Elliott G (eds) Handbook of economic forecasting. Elsevier, Amsterdam
Dangl T, Halling M (2012) Predictive regressions with time-varying coefficients. J Financ Econ 106:157–181
Dickey DA, Fuller WA (1981) Likelihood ratio statistics for autoregressive time series with a unit root. Econometrica 49:1057–1072
Diebold FX, Mariano RS (1995) Comparing predictive accuracy. J Bus Econ Stat 13:253–63
Driesprong G, Jacobsen B, Maat B (2008) Striking oil: another puzzle? J Financ Econ 89:307–327
Durbin J, Koopman SJ (2002) A simple and efficient simulation smoother for state space time series analysis. Biometrika 89:603–615
Elliot G, Stock JH (1994) Inference in time series regression when the order of integration of a regressor is unknown. Econom Theory 10:672–700
Engle RF (1982) Autoregressive conditional heteroskedasticity with estimates of the variance of United Kingdom inflation. Econometrica 50:987–1007
Foster FD, Smith T, Whaley RE (1997) Assessing goodness-of-fit of asset pricing models: the distribution of the maximal $R^{2}$. J Finance 53:591–607
Geweke J (1992) Evaluating the accuracy of sampling-based approaches to the calculation of posterior moments. In: Berger J, Bernardo J, Dawid A, Smith A (eds) Bayesian statistics. Oxford University Press, Oxford
Giacomini R, Rossi B (2010) Forecast comparisons in unstable environments. J Appl Econom 25:595–620
Gil-Alana LA, Gupta R (2014) Persistence and cycles in historical oil price data. Energy Econ 45:511–516
Gil-Alana LA, Gupta R, Olubusoye OE, Yaya OS (2016) Time series analysis of persistence in crude oil price volatility across bull and bear regimes. Energy 109:29–37
Gjerde O, Saettem F (1999) Causal relations among stock returns and macroeconomic variables in a small, open economy. J Int Financ Mark Inst Money 9:61–74
Gneiting T, Ranjan R (2011) Comparing density forecasts using threshold- and quantile-weighted scoring rules. J Bus Econ Stat 29:411–422
Groen JJJ, Richard P, Ravazzolo F (2013) Real-time inflation forecasting in a changing world. J Bus Econ Stat 1:29–44
Hamilton JD (2003) What is an oil shock? J Econom 113:363–398
Hamilton JD (2011) Nonlinearities and the macroeconomic effects of oil prices. Macroecon Dyn 15:472–497
Huang R, Masulis R, Stoll H (1996) Energy shocks and financial markets. J Futures Mark 16:1–27
Hull J, White A (1987) The pricing of options on assets with stochastic volatilities. J Finance 42:281–300
Inoue A, Kilian L (2004) In-sample or out-of-sample tests of predictability: which one should we use? Econom Rev 23:371–402
Jansson M, Moreira M (2006) Optimal inference in regression models with nearly integrated regressors. Econometrica 74:681–714
Johannes M, Korteweg A, Polson N (2014) Sequential learning, predictability, and optimal portfolio returns. J Finance 69:611–644
Jones CM, Kaul G (1996) Oil and stock markets. J. Finance 51:463–491
Jurado K, Ludvigson SC, Ng S (2015) Measuring uncertainty. Am Econ Rev 105:1177–1216
Kilian L, Manganelli S (2008) The central banker as a risk manager: estimating the Federal Reserve’s preferences under Greenspan. J Money Credit Bank 40:1103–1129
Kilian L, Vigfusson RJ (2013) Do oil prices help forecast U.S. real GDP? The role of nonlinearities and asymmetries. J Bus Econ Stat 31:78–93
Kim S, Shephard N, Chib S (1998) Stochastic volatility: likelihood inference and comparison with ARCH models. Rev Econ Stud 65:361–393
Koop G (2003) Bayesian econometrics. Wiley, New York
Koopman SJ, Lucas A, Scharth M (2016) Predicting Time-Varying Parameters with Parameter-Driven and Observation-Driven Models. Review of Economics and Statistics 98:97–110
Lewellen J (2004) Predicting returns with financial ratios. J Financ Econ 74:209–235
Liu L, Ma F, Wang Y (2015) Forecasting excess stock returns with crude oil market data. Energy Econ 48:316–324
Lo AW, MacKinlay AC (1990) Data-snooping biases in tests of financial asset pricing models. Rev Financ Stud 3:431–467
Lux T, Segnon M, Gupta R (2016) Forecasting crude oil price volatility and value-at-risk: evidence from historical and recent data. Energy Econ 56:117–133
Miller J, Ratti R (2009) Crude oil and stock markets: stability, instability and bubbles. Energy Econ 31:559–568
Narayan PK, Gupta R (2015) Has oil price predicted stock returns for over a century? Energy Econ 48:18–23
Narayan PK, Narayan S (2010) Modelling the impact of oil prices on Vietnam’s stock prices. Appl Energy 87:356–361
Narayan PK, Sharma SS (2011) New evidence on oil price and firm returns. J Bank Finance 5:3253–3262
Naser H, Alaali F (2018) Can oil prices help predict US stock market returns? Evidence using a dynamic model averaging (DMA) approach. Empir Econ 55:1757–1777
Nelson CR, Kim M (1993) Predictable stock returns: the role of small sample bias. J Finance 48:641–661
Park J, Ratti RA (2008) Oil price shocks and stock markets in the U.S. and 13 European countries. Energy Econ 30:2587–2608
Phan DHB, Sharma SS, Narayan PK (2015) Stock return forecasting: some new evidence. Int Rev Financ Anal 40:38–51
Rapach DE, Wohar ME (2006) In-sample vs. out-of-sample tests of stock return predictability in the context of data mining. J Empir Finance 13:231–247
Sadorsky P (1999) Oil price shocks and stock market activity. Energy Econ 21:449–469
Stambaugh RF (1999) Predictive regressions. J Financ Econ 54:375–421
Stock JH, Watson MW (1996) Evidence on structural instability in macroeconomic time series relations. J Bus Econ Stat 14:11–30
Swanson NR (1998) Money and output viewed through a rolling window. J Monet Econ 41:455–474
Tourus W, Valkanov R, Yan S (2004) On predicting stock returns with nearly integrated explanatory variables. J Bus 77:937–966
Wei C (2003) Energy, the stock market, and the putty-clay investment model. Am Econ Rev 93:311–23
Welch I, Goyal A (2008) A comprehensive look at the empirical performance of equity premium prediction. Rev Financ Stud 21:1455–1508
Westerlund J, Narayan PK (2012) Does the choice of estimator matter when forecasting returns? J Bank Finance 36:2632–2640
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflict of interest
The author declares that he/she has no conflict of interest.
Human and animal rights
This article does not contain any studies with human participants or animals performed by the author.
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Appendix
Appendix
1.1 Gibbs sampling estimation of the predictive model
To conduct estimation, we rely on Gibbs sampling combined with data augmentation. In other words, we augment the parameter space to include any latent states, and sample from the augmented posterior by sequentially sampling from the conditional posteriors. We refer the reader to Koop (2003) for an introduction to Gibbs sampling. Specifically, we divide the augmented posterior of (3.1)–(3.4), \(p\left( h_{1:T}^{\eta },h_{1:T}^{X},\zeta _{1:T}^{\eta },\zeta _{1:T}^{X},\psi \mid y_{1:T},X_{1:T}\right) \), where \(\psi \) denotes the vector containing the model parameters into the following blocks: The log-volatilities, \(h_{1:T}^{\eta }\) and \(h_{1:T}^{X}\), conditional mean parameters, \(\gamma =\left( \mu ,\phi ,\alpha ,\beta ,\varphi \right) ^{'}\), conditional volatility parameters, \(\gamma ^{\eta }\) and \(\gamma ^{X}\), where \(\gamma ^{\eta }=\left( \mu ^{\eta },\phi ^{\eta },\omega ^{2\eta }\right) ^{'}\), and the mixture component indicators, \(\zeta _{1:T}^{\eta }\) and \(\zeta _{1:T}^{X}\), which we use to generate \(h_{1:T}^{\eta }\) and \(h_{1:T}^{X}\) as following Kim et al. (1998). More precisely, posterior draws from \(p\left( h_{1:T}^{\eta },h_{1:T}^{X},\zeta _{1:T}^{\eta },\zeta _{1:T}^{X},\psi \mid y_{1:T},X_{1:T}\right) \) are obtained by cycling through the following steps:
\(\gamma \sim p\left( \gamma \mid h_{1:T}^{\eta },h_{1:T}^{X},y_{1:T},X_{1:T}\right) \),
\(h_{1:T}^{\eta }\sim p\left( h_{1:T}^{\eta }\mid \zeta _{1:T}^{y},\gamma ^{\eta },\gamma ,y_{1:T},X_{1:T}\right) \),
\(h_{1:T}^{X}\sim p\left( h_{1:T}^{X}\mid \zeta _{1:T}^{X},\gamma ^{X},\gamma ,X_{1:T}\right) \),
\(\zeta _{1:T}^{\eta }\sim p\left( \zeta _{1:T}^{\eta }\mid h_{1:T}^{y},\gamma ^{\eta },\gamma ,y_{1:T},X_{1:T}\right) \),
\(\zeta _{1:T}^{X}\sim p\left( \zeta _{1:T}^{X}\mid h_{1:T}^{X},\gamma ^{X},\gamma ,X_{1:T}\right) \),
\(\gamma ^{\eta }\sim p\left( \gamma ^{\eta }\mid h_{1:T}^{\eta }\right) \),
\(\gamma ^{X}\sim p\left( \gamma ^{X}\mid h_{1:T}^{X}\right) \).
In order to generate \(\gamma \sim p\left( \gamma \mid h_{1:T}^{\eta },h_{1:T}^{X},y_{1:T},x_{1:T}\right) \) in step one, we use simple Gibbs sampling techniques for linear regression models, see Koop (2003) among many other. More precisely, we start by generating \(\mu \) and \(\phi \). Thereafter, we follow Amihud and Hurvich (2004), and construct the corrected estimator of \(\phi \) as follows:
Once we compute (A.1), we can construct \({\hat{\varepsilon }}_{1:T}^{X}\). Conditional on \({\hat{\varepsilon }}_{1:T}^{X}\) and the data, we then generate \(\alpha \), \(\beta \) and \(\varphi \) in one block. Likewise, conditional on \(\gamma \) and the data, we can generate \(h_{1:T}^{\eta }\) and \(h_{1:T}^{X}\) from their respective conditional posteriors using Gibbs sampling techniques for linear state-space models, see Carter and Kohn (1994) and Durbin and Koopman (2002) for textbook treatments. For instance, conditional on \(\mu \), \(\phi \), and \(X_{1:T}\), we have that
Evidently, (A.2)–(A.3) is a nonlinear state-space model as the observation equation, (A.2), is nonlinear in the state variable, \(h_{t}^{X}\). Therefore, to generate \(h_{1:T}^{X}\) from its conditional posterior, we use the auxiliary mixture sampler suggested in Kim et al. (1998). In a nutshell, the idea is to approximate the nonlinear stochastic volatility model using a mixture of linear Gaussian models, which are relatively easier to estimate. More precisely, our goal is to transform (A.2), such that it becomes linear in \(h_{1:T}^{X}\) . We can simply do this by squaring both sides of (A.2) and taking the logarithm. This way, the transformed model,
is linear in \(h_{t}^{X}\). However, the error term in (A.4), \(\log \varepsilon _{t}^{2X}\), does not follow a Normal distribution, but a \(\log \chi _{1}^{2}\) distribution. Consequently, the mechanism for fitting linear Gaussian state-space models cannot be directly applied to (A.4)–(A.5). To tackle this difficulty, we follow Kim et al. (1998) and approximate \(\log \varepsilon _{t}^{2X}\) using a mixture of seven Gaussian densities:
where \(\varpi _{i}\) is the probability of the ith mixture component and \(\Sigma _{i=1}^{7}\varpi _{i}=1\). The values of \(m_{i}\), \(\varsigma _{i}^{2}\) and \(\varpi _{i}\), \(i=1,\ldots ,7\), are all fixed and provided in Kim et al. (1998). We can equivalently write the mixture density in terms of a random indicator variable, \(\zeta _{t}^{X}\in \left\{ 1,\ldots ,7\right\} \), (hence, the name of the approach) as follows:
Equations (A.4)–(A.5) with (A.6)–(A.7) constitute a linear state-space model,
where \(v_{t}=\log \left( X_{t}-\mu -\phi X_{t-1}\right) ^{2}-\left( m_{t}-1.2704\right) \), \(H_{t}=1\), \(\theta _{t}=h_{t}^{X}\), \({\widetilde{\mu }}=\mu ^{X}\left( 1-\phi ^{X}\right) \), \(F=\phi ^{X}\), \(Q=\omega ^{2X}\) and \(R_{t}=\varsigma _{t}^{2}.\) Furthermore, \(m_{t}\) and \(\varsigma _{t}^{2}\) correspond to \(m_{i}\) and \(\varsigma _{i}^{2}\), \(i=1,\ldots ,7\), with the highest probability and are available from the previous Gibbs iteration. Given these quantities, we can simply use a simulation smoother to generate \(h_{1:T}^{X}\) from its conditional posterior, see for example, Kim et al. (1998).
To generate the corresponding mixture component indicator series, \(\zeta _{1:T}^{X}\), we use that each \(\zeta _{t}^{X}\) can be generated independently conditional on \(v_{1:T}\) and \(h_{1:T}^{X}\). Specifically, \(\zeta _{t}^{X}\) is a discrete random variable that follows a seven-point distribution, and it can be easily sampled via the inverse transform method, see Kim et al. (1998) for details. The processes, \(h_{1:T}^{\eta }\) and \(\zeta _{1:T}^{\eta }\), are generated following the exact same procedure.
The conditional volatility parameters, \(\gamma ^{\eta }=\left( \mu ^{\eta },\phi ^{\eta },\omega ^{2\eta }\right) ^{'}\) and \(\gamma ^{X}=\left( \, \mu ^{X},\phi ^{X}, \omega ^{2X}\right) ^{'}\), are generated element-by-element conditional on \(h_{1:T}^{\eta }\) and \(h_{1:T}^{X}\), respectively. The conditional posteriors of \(\mu ^{\eta }\)\(\left( \mu ^{X}\right) \) and \(\omega ^{2\eta }\)\(\left( \omega ^{2X}\right) \) have closed-formed solutions. Lastly, \(\phi ^{\eta }\)\(\left( \phi ^{X}\right) \) is generated using an independence-chain Metropolis-Hastings step, see Kim et al. (1998) and Chan (2013) for more details.
1.2 Convergence diagnostic test
Table 11 reports p values from the convergence diagnostic test suggested in Geweke (1992). The null hypothesis of this test is that the posterior mean of the model parameters based on the first \(10\%\) of the sample of retained draws is the same as the posterior mean of the model parameters based on the last \(50\%\) of the sample of retained draws. Following Koop (2003), a p value greater than 0.05 for all parameters suggests that convergence of the proposed Gibbs sampler has been achieved.
1.3 Six-months ahead density prediction results
Tables 12, 13, 14, 15 and 16 report density prediction results from our models relative to the stochastic volatility benchmark at the 6-months ahead prediction horizon.
Rights and permissions
About this article
Cite this article
Nonejad, N. Does the price of crude oil help predict the conditional distribution of aggregate equity return?. Empir Econ 58, 313–349 (2020). https://doi.org/10.1007/s00181-019-01643-2
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00181-019-01643-2