Abstract
In this chapter, the reader is introduced to time series modeling. The science (and art) of time series modeling reflects the times series models of Professors Box and Jenkins, Granger, and Hendry. Many economic time series follow near-random walks or random walk with drift processes. This chapter uses the time series modeling of real Gross Domestic product, GDP, as a time series of interest. Time series models can be univariate, where a time series is modeled only by its past values, or multivariate, in which an input series leads an output series, such as a composite index of leading economic indicators, LEI, that can be used as an input to a transfer function model of real Gross Domestic Product, GDP. Economic indicators are descriptive and anticipatory time-series data can be used to analyze and forecast changing business conditions. Cyclical indicators are comprehensive series that are systematically related to the business cycle. Business cycles are recurrent sequences of expansions and contractions in aggregate economic activity. Coincident indicators have cyclical movements that approximately correspond with the overall business cycle expansions and contractions. Leading indicators reach their turning points before the corresponding business cycle turns. The lagging indicators reach their turning points after the corresponding turns in the business cycle.
This is a preview of subscription content, log in via an institution.
Buying options
Tax calculation will be finalised at checkout
Purchases are for personal use only
Learn about institutional subscriptionsNotes
- 1.
The transfer function model forecasts are compared to several naïve models in terms of testing which model produces the most accurate forecast of real GDP. No-change forecasts of real GDP and random walk with drift models may useful forecasting benchmarks [145, 247]. Economists have constructed leading economic indicator series to serve as a business barometer of the changing U.S. economy since the time of Wesley C. Mitchell [248]. The purpose of this study is to examine the time series forecasts of composite economic indexes produced by The Conference Board (TCB), and test the hypothesis that the leading indicators are useful as an input to a time series model to forecast real output in the United States.
- 2.
An example of business cycles can be found in the analysis of Mitton Friedman and others who discussed how changes in the money supply lead to rising prices and an initial fall in the rate of interest, and how this results in raising profits, creating a boom. The interest rate later rises, reducing profits, and ending the boom. A financial crisis ensues when businessmen, whose loan collateral is falling as interest rates rise, run to cash and banks fail. The money supply was a series in the leading economic indicators of Persons [200], Burns and Mitchell [46], Friedman and Schwartz [132], and Zarnowitz [337]. The Conference Board index of leading economic indexes, LEI, dropped the money supply when it failed to forecast the Global financial Crisis, GFC, see Levanon, Manini, Ozyildirim, Schaitkin, and Tanchua [216].
- 3.
- 4.
Please see Box and Jenkins, Time Series Analysis, Chapter 3, for the most complete discussion of the ARMA (p,q) models.
- 5.
A stationary AR(p) process can be expressed as an infinite weighted sum of the previous shock variables
$$ {\tilde{Z}}_t={\varphi}^{-1}(B){\alpha}_t. $$In an invertible time series, the current shock variable may be expressed as an infinite weighted sum of the previous values of the series
$$ {\theta}^{-1}(B){\tilde{Z}}_t={\alpha}_t. $$ - 6.
Box and Jenkins, Time Series Analysis. Chapter 6; C.W.J. Granger and Paul Newbold, Forecasting Economic Time Series. Second Edition (New York: Academic Press, 1986), pp. 109–110, 115–117, 206.
- 7.
Granger and Newbo1d, [140]. pp. 185–186.
- 8.
Box and Jenkins, Time Series Analysis. pp. 173–179. Also Jenkins [193]
- 9.
Box and Jenkins, Time Series Analysis, pp. 305–308.
- 10.
Box and Jenkins, op. cit.
- 11.
Guerard [155] reported the estimated autocorrelation and partial autocorrelation functions of U.S. real GDP, 1963- March 2002. The estimated autocorrelation function decays gradually, falling from 0.979 for a one period (quarter lag), 0.958 for a two quarter lag, to 0.584 for a twenty quarter lag, and 0.318 for a 36 quarter lag. The estimated partial autocorrelation function is characterized by the “spike” at a one quarter lag. A virtually identical ACF and PACF patern is presented in exhibit 1; The real GDP series should be first-differenced. The U.S. real GDP series was estimated as a random walk with drift series for the 1963–2002 period and that is the general form for 1947–2015Q2, with the notable exception of outliers in the 2008–2009 period.
- 12.
- 13.
- 14.
Geoffrey Moore passed in 2000. Pami Dua [97] edited a collection of papers to honor Moore, entitled Business Cycles and Economic Growth (Oxford University Press, New York).The reader is referred to papers in the volume by Victor Zarnowitz, John Guerard, Lawrence Klein, and S. Ozmucur, and D. Ivanova and Kajal Lahiri.
- 15.
The reader is referred to Zarnowitz [337] for his seminal development of underlying economic assumption and theory of the LEI and business cycles. Guerard [155, 157] used an autoregressive variation of the random walk with drift model as a forecasting benchmark. We estimate the cross-correlation function of the LEI and real GDP for an initial 32 quarter estimation period, following Thomakos and Guerard [303], and used the 1978-March 2002 period for initial U.S. post-sample evaluation. Thomakos and Guerard [303] compared the forecasting accuracy of four models of U.S. real GDP. The models tested are: (1) the transfer function model in which The Conference Board composite index of leading economic indicators (LEI) is lagged three quarters, denoted TF; (2) a no-change (NoCH) forecast; (3) the simple RWD model; and (4) a simple transfer function model in which The Conference Board composite index of leading economic indicators is lagged one period, denoted TF1. One finds that the three-quarter of lagged LEI transfer function is the most accurate out-of-sample forecasting model for the U.S. real GDP; although there is no statistically significant differences in the rolling one-period-ahead root mean square forecasting errors of the RWD, TF, and TF1 models. The U.S. leading indicators lead Real GDP, as one should expect, and the transfer function model produced lower forecast errors than the univariate model, and a naive benchmark, the no-change model.
The model forecast errors are not statistically different (the t-value of the paired differences of the univariate and TF models is 0.91). The multiple regression models indicate statistical significance in the U.S. composite index of leading economic indicators for the 1978-March 2002 period. One does not find that the transfer function model forecast errors are (statistically) significantly lower than univariate ARIMA model (RWD) errors in a rolling one-period-ahead analysis.
- 16.
- 17.
- 18.
- 19.
The author is indebted to Jurgen Doornik for this suggestion.
- 20.
One must apparently be even more careful with the Box-Pierce test on sums of squared ρ k .
- 21.
OLS estimation suffices to produce unbiased estimates, since all the bivariate models considered are reduced forms. It also allows one to consider variants of one equation without disturbing the forecasting results from the other, and it is computationally simpler. On the other hand, where substantial contemporaneous correlation occurs between the residuals, seemingly-unrelated regressions GLS estimation can be expected to yield noticeably better parameter estimates and post-sample forecasts. All estimation in this study is OLS; a re-estimation of our final bivariate model using GLS might strengthen our conclusions somewhat.
- 22.
Alternatively, one might fit both models to the sample period, produce forecasts of the first post-sample observation, re-estimate both models with that observation added to the sample, forecast the second post-sample observation, and so on until the end of the post-sample period. This would, of course, be more expensive than the approach in the text.
- 23.
The MSFE-reduction testing methodology used here is essentially identical to that of Ashley and Ye [10], which the reader should consult for a more detailed discussion than is given below. In fact, the only differences here are that a noticeably larger number of (substantially more macroeconomically interesting) economic time-series are considered in both the unrestricted and restricted models and that the two different model-identification schemes are employed and compared.
- 24.
For simplicity, we fix the values of initial observations at their actual sample values.
References
Arnott, R. (1985). The use and misuse of consensus earnings. Journal of Portfolio Management, 12, 18–27.
Ashley, R. A., Granger, C. W. J., & Schmalensee, R. L. (1980). Advertising and aggregate consumption: An analysis of causality. Econometrica, 48, 1149–1168.
Ashley, R., & Ye, H. (2012). On the granger causality between median inflation and price dispersion. Applied Economics, 44, 4221–4238.
Bollerslev, T. (1986). Generialized autoregressive conitional heteroskedasticity. Journal of Econometrics, 31, 307–327.
Box, G. E. P., & Cox, D. R. (1964). An analysis of transformations. Journal of the Royal Statistical Society, Series B, 26(2), 211–252.
Box, G. E. P., & Jenkins, G. M. (1970). Time series analysis: Forecasting and control. Oakland: Holden-Day.
Burns, A. F., & Mitchell, W. C. (1946). Measuring business cycles. New York: NBER.
Castle, J., & Shepard, N. (2009). The methodology and practice of econometrics. Oxford: Oxford University Press.
Dhrymes, P. J., & Guerard Jr., J. B. (2017). Returns, risk, portfolio selection, and evaluation. In J. Guerard (Ed.), Portfolio construction, Measurment, and efficiency: Essays in honor of Jack Tteynor. New York: Springer.
Doornik, J. A., & Hendry, D. F. (2009a). Empirical Econometric Modelling. London: Timberlake Consultants, Ltd..
Doornik, J. A., & Hendry, D. F. (2009b). Modelling dynamic systems. London: Timberlake Consultants, Ltd.
Doornik, J. A. (2009). Autometrics. In J. Castle & N. Shepard (Eds.), The methodology and practice of econometrics. Oxford: Oxford University Press.
Doornik, J. A., & Hendry, D. F. (2015). Statistical model selection with “big data”. Cogent Economics & Finance, 3, 1–15.
Dua, P. (2004). Business cycles and economic growth. New Delhi: Oxford University Press in India.
Elton, E. J., Gruber, M. J., Brown, S. J., & Goetzman, W. N. (2007). Modern portfolio theory and investment analysis (7th ed.). New York: Wiley.
Engle, R. F. (1982). Autoregressive conditional heteroscedasticity with estimatea of the variance in United Kingdom inflation. Econometrica, 50, 987–1006.
Engle, R. F. (1995). ARCH: Selected readings. New York: Oxford University Press.
Fox, A. J. (1972). Outliers in time series. Journal of the Royal Statistical Society,Series B, 34, 350–363.
Friedman, M., & Schwartz, A. (1963). Money and business cycles. Review of Economics and Statistics, 45, 32–64.
Granger, C. W. J. (1969). Investigating casual relations by economic models and cross-spectral methods. Econometrica, 37, 424–438.
Granger, C. W. J., & Newbold, P. (1977). Forecasting economic time series. New York: Academic Press.
Granger, C. W. J. (1980b). Testing for causality: A personal viewpoint. Journal of Economic Dynamics and Control, 2, 329–352.
Granger, C. W. J. (1986). Developmentsin the study of Cointegrated economic variables. Oxford Bulletin of Economics and Statistics, 48, 213–228.
Granger, C. W. J. (1989a). Invited review: Combining forecasts-twenty years later. Journal of Forecasting, 8, 167–173.
Guerard Jr., J. B. (2001). A note on the effectiveness of the U.S. leading economic indicators. Indian Economic Review, 36, 251–268.
Guerard Jr., J. B. (2004). The forecasting effectiveness of the U.S. leading economic indicators: Further evidence and initial G7 results. In P. Dua (Ed.), Business cycles and economic growth: An analysis using leading indicators. New York: Oxford University Press.
Hendry, D. F., & Nielsen, B. (2007). Econometric modeling: A likelihood approach. Princeton: Princeton University Press.
Hendry, D. F., & Doornik, J. A. (2014). Empirical model discovery and theory evaluation. Cambridge: MIT Press.
Hoerl, A. E. (1959). Optimum solution of many variables equations. Chemical Engineering Progress, 55, 69–78.
Jagannathan, R., & Ma, T. (2003). Risk reduction in large portfolios: Why imposing the wrong constraints helps. Journal of Finance, 58, 1651–1684.
Jenkins, G. M. (1979). Practical experiences with modelling and forecasting tine series. Jersey/Channel Island: A GJP Publication.
Koopmans, T. C., & Hood, W. C. (1953). The estimation of simultaneous linear economic relationships. In W. C. Hood & T. C. Koopmans (Eds.)., Chapter 6 Studies in econometric method. New York: Wiley.
Lahiri, K., & Wang, J. G. (2013). Evaluating probability forecasts for GDP deciles using alternative methodologies. International Journal of Forecasting, 29, 175–190.
Lee, J. H., & Stefek, D. (2008). Do risk factors eat alphas? Journal of Portfolio Management, 34(4), 12–25.
Levanon, G., Manini, J.-C., Ozyildirim, A., Schaitkin, B., & Tanchua, J. (2016). Using financial indicators to predict turning points in the business cycle: The case of the leading economic index for the United States. International Journal of Forecasting, 31, 427–445.
McCracken, M. W. (2007). Asymptotics for out of sample tests of granger causality. Journal of Econometrics, 140(2), 719–752.
Mills, T. C. (1990). Time series techniques for economists. New York: Cambridge University Press.
Mincer, J., & Zarnowitz, V. (1969). The evaluation of economic forecasts. In J. Mincer (Ed.), Economic forecasts and expectations. New York: Columbia University Press.
Mitchell, W. C. (1913). Business cycles. New York: Burt Franklin reprint.
Mitchell, W. C. (1951). What happens during business cycles: A progress report. New York: NBER.
Moore, G. H. (1961). Business cycle indicators. 2 volumes. Princeton: Princeton University Press.
Nelson, C. R., & Plosser, C. I. (1982). Trends and random walks in macroeconomic time series. Journal of Monetary Economics, 10, 139–162.
Thomakos, D., & Guerard, J. (2004). Naïve, ARIMA, transfer function, and VAR models: A comparison of forecasting performance. The International Journal of Forecasting, 20, 53–67.
Treynor, J. L. (1999). Toward a theory of market value for risky assets. In R. Korajczyk (Ed.), Asset pricing and portfolio performance. London: Risk Books.
Ye, H., Ashley, R. A., & Guerard Jr., J. B. (2015). Comparing the effectiveness of traditional vs. mechanized identification methods in post-sample forecasting for a macroeconomic granger causality analysis. The International Journal of Forecasting, 31, 488–500.
Zarnowitz, V. (1992). Business cycles: Theory, history, indicators, and forecasting. Chicago: University of Chicago Press.
Author information
Authors and Affiliations
Appendix: Advanced Time Series Modeling
Appendix: Advanced Time Series Modeling
Time series modeling can involve predicting the conditional mean of a univariate time series and its conditional variance, its second order moment. Volatility forecasting has become extremely important in derivative pricing and risk management. In this chapter the reader is introduced to several techniques used in time series modeling, the univariate generalized autoregressive conditional heteroscedasticity (GARCH) model, developed by Engle [109, 110] and Bollerslev [35], and vector autoregressive models (VAR). In the previous chapter the reader was introduced to the simple autoregressive model, an AR(1):
If x t is a stationary time series, such as financial returns, then the series can be expressed as its mean plus a white noise term that is identically and identically distributed, denoted iid v(ε) = σ2. If large changes in data are followed by large changes and small changes follow small changes, then there can be volatility clustering, or conditional heteroscedasticity.
A model that allows the conditional variance to be a function of its past history is said to be the bilinear model of Granger and Anderson in Granger and Newbold [140].
The conditional variance is \( {\sigma}^2{y}_{t-1}^2 \) and can be either zero or infinite. Engle [86] states a preferable model is
where \( {h}_t={\alpha}_0+{\alpha}_1{y}_{t-1}^2 \)Equation (A.3) is referred to as the autoregressive conditional heteroscedasticity (ARCH) model.
where p is the autoregressive process and α is the vector of unknown parameters.
Engle writes an ARCH process as being described in equations (A.4) and (A.6). The conditional variance is
The general first-order linear ARCH process, with α0 > 0, is (covariance) stationary if and only if the characteristic equation has all roots outside the unit circle.
A pth-order linear case can be written as:
The Generalized Autoregressive Conditional Heteroscedasticity GARCH (p,q) process can be written as:
where
If p is zero, then the GARCH (p,q) model reduces to ARCH (q) process.
If all roots of 1-β(z) = 0 lie outside the unit series, then
The GARCH (p,q) process can be written:
The GARCH (i,l) process is:
or
and
Rights and permissions
Copyright information
© 2017 Springer International Publishing AG
About this chapter
Cite this chapter
Dhrymes, P. (2017). Time Series Modeling. In: Introductory Econometrics. Springer, Cham. https://doi.org/10.1007/978-3-319-65916-9_7
Download citation
DOI: https://doi.org/10.1007/978-3-319-65916-9_7
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-319-65914-5
Online ISBN: 978-3-319-65916-9
eBook Packages: Economics and FinanceEconomics and Finance (R0)