Skip to main content
main-content

Über dieses Buch

In this book, the authors reject the theorem-proof approach as much as possible, and emphasize the practical application of econometrics. They show with examples how to calculate and interpret the numerical results.

This book begins with students estimating simple univariate models, in a step by step fashion, using the popular Stata software system. Students then test for stationarity, while replicating the actual results from hugely influential papers such as those by Granger and Newbold, and Nelson and Plosser. Readers will learn about structural breaks by replicating papers by Perron, and Zivot and Andrews. They then turn to models of conditional volatility, replicating papers by Bollerslev. Finally, students estimate multi-equation models such as vector autoregressions and vector error-correction mechanisms, replicating the results in influential papers by Sims and Granger.

The book contains many worked-out examples, and many data-driven exercises. While intended primarily for graduate students and advanced undergraduates, practitioners will also find the book useful.

Inhaltsverzeichnis

Frontmatter

Chapter 1. Introduction

Abstract
Econometrics can be used to answer many practical questions in economics and finance:
  • Suppose you own a business. How might you use the previous 10 years’ worth of monthly sales data to predict next month’s sales?
  • You wish to test whether the “Permanent income hypothesis” holds. Can you see whether consumption spending is a relatively constant fraction of national income?
  • You are a financial researcher. You wish to determine whether gold prices lead stock prices, or vice versa. Is there a relationship between these variables? If so, can you use it to make money?
Answering each of these questions requires slightly different statistical apparatuses, yet they all fall under the umbrella of “time series econometrics.”
John D. Levendis

2. ARMA(p,q) Processes

Abstract
A long-standing dream of economists was to build a massive model of the economy. One with hundreds of supply and demand equations. A supply and demand system for each input, intermediate good, and final product. One would only need to estimate the relevant elasticities and a few simple parameters to construct an economic crystal ball. It would be able to make accurate forecasts and useful policy prescriptions. Most economists wished this at one point. Slowly, though, the era of optimism in structural macroeconomic forecasting during the 1950s and 1960s became an era of doubt during the 1970s and 1980s.
John D. Levendis

3. Model Selection in ARMA(p,q) Processes

Abstract
In practice, the form of the underlying process that generated the data is unknown. Should we estimate an AR(p) model, an MA(q) model, or an ARMA(p,q) model? Moreover, what lag lengths of p and q should we choose? We simply do not have good a priori reason to suspect that the data generating process is of one type or another, or a combination of the two. How is a researcher to proceed? Which sort of model should we estimate?
John D. Levendis

4. Stationarity and Invertibility

Abstract
Most time-series methods are only valid if the underlying time-series is stationary. The more stationary something is, the more predictable it is. More specifically, a time-series is stationary if its mean, variance, and autocovariance do not rely on the particular time period.
John D. Levendis

5. Non-stationarity and ARIMA(p,d,q) Processes

Abstract
Up until now we have been looking at time series whose means did not exhibit long-run growth. It is time to drop this assumption. After all, many economic and financial time series do not have a constant mean. Examples include: the US GDP per capita, the US CPI, the Dow-Jones Industrial Index, and the share price of Google.
John D. Levendis

6. Seasonal ARMA(p,q) Processes

Abstract
Many financial and economic time series exhibit a regular cyclicality, periodicity, or “seasonality.” For example, agricultural output follows seasonal variation, flower sales are higher in February, retail sales are higher in December, and beer sales in college towns are lower during the summers.
John D. Levendis

7. Unit Root Tests

Abstract
A process might be non-stationary without being a unit root. The two concepts are related, but they are not identical and it is common to confuse the two. We can have non-stationarity without it being due to a unit root. We could have a seasonal model. Or, we could have a deterministic trend. (We can even have non-stationarity because the variance is changing over time.)
John D. Levendis

8. Structural Breaks

Abstract
In 1976, Robert Lucas offered one of the strongest criticisms of the Cowles Commission large-scale econometric modeling approach. Lucas critiqued Cowles’ presumption that many economic phenomena are structural. They are not.
John D. Levendis

9. ARCH, GARCH and Time-Varying Variance

Abstract
To this point, we have considered non-stationary means, but strictly speaking, non-stationarity could apply to any of the moments of a random variable: the mean, variance, skewness, kurtosis, etc… Finance especially is concerned with the non-stationarity of variance.
John D. Levendis

10. Vector Autoregressions I: Basics

Abstract
If we take the notion of general equilibrium seriously, then everything in the economy is related to everything else. For this reason, it is impossible to say which variable is exogenous. It is possible that all variables are endogenous: they can all be caused by, and simultaneously be the cause of, some other variable.
John D. Levendis

11. Vector Autoregressions II: Extensions

Abstract
In the previous chapter, we covered the basics of reduced form VARs on stationary data. In this chapter, we continue learning about VARs, but we extend the discussion to structural VARs (SVARs) and VARS with integrated variables.
John D. Levendis

12. Cointegration and VECMs

Abstract
The VARs that we looked at in the last chapter were very well suited for describing the short-run relationship between variables, especially if they are stationary. Most economic variables are not stationary, however. This required us to transform the variables, taking first differences, so that they are stationary. In this chapter, we show how to model the long-run relationship between variables in their levels, even if they are integrated. This is possible if two or more variables are “cointegrated.” If two variables are cointegrated, then, rather than taking the first difference of each variable, we can essentially model the difference between the two variables. Loosely speaking.
John D. Levendis

13. Conclusion

Abstract
In this text, we have explored some of the more common time-series econometric techniques. The approach has centered around developing a practical knowledge of the field, learning by replicating basic examples and seminal research. But there is a lot of bad research out there, and you would be best not to replicate the worst practices of the field.
John D. Levendis

14. Correction to: Time Series Econometrics

John D. Levendis

Backmatter

Weitere Informationen

Premium Partner

    Bildnachweise