Abstract
In this chapter, we consider the question of long-run exogeneity and the related issue of identification. In our opinion, detection of the exogenous variables in the long run or the short run is a precursor to any attempt to structurally identify economic or financial phenomena.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Notes
- 1.
We would like to thank Graham Mizon for his discussion of this issue.
- 2.
For the cointegrating exogenous case in Hunter (1992) r 1 = 1 and n 1 = 3.
- 3.
- 4.
Hence, α and β are 6 × 2 dimensioned matrices.
- 5.
The restrictions for strict exogeneity (SE) of the oil price are α ij  = 0 and β ij  = 0 for i, j = 1, 2.
- 6.
In fact the same test can be used by rerunning the Johansen procedure with the variable to be tested for WE being placed first.
- 7.
The restrictions for CE are α 51 = 0, α 61 = 0 and β j2 = 0 for j = 1, …, 4.
- 8.
The WE restrictions for the model in Hunter (1992) are α i1 = 0 for i = 4, 5, 6 and α j2 = ω j1 α 42 +ω j2 α 52 +ω j3 α 62 for j = 1, 2, 3.
- 9.
Strong exogeneity augments the sub-block WE restriction above by β i2 = 0 for i = 1, 2, 3.
- 10.
If an equation has exactly j i  = n 1 − 1 restrictions, then it has enough restrictions to be exactly identified. When j i  > n 1 − 1, it has enough restrictions to be over-identified, but without the appropriate number it will be under or not identified.
- 11.
The Hessian is the second derivative of the likelihood, which provides an estimate of the variance-covariance matrix of the parameters. If some parameters are ill-defined, then the likelihood is flat and the Hessian matrix singular. Then some parameters in the model are not identified. Perfect multi-collinearity is a special case of this and it occurs when two or more variables are related and their parameters cannot be independently estimated and as a result are not identifiable.
- 12.
- 13.
Usually, two equations or blocks of equations with different parameterizations have the same value for their likelihoods. In general, there exists at least one model with r exactly identifying restrictions for each equation with a likelihood value, the same as the unrestricted likelihood.
- 14.
The approach described here was first outlined for the I(1) case in Hunter and Simpson (1995).
- 15.
- 16.
The additional restriction is required to solve for all of the parameters and of the eight restrictions implied by \(\beta _{r}^{{\prime}} = \left [\begin{array}{cccccc} \beta _{11} & \beta _{21} & -\beta _{21} & -\beta _{21} & 0 & \beta _{61}\\ 0 & 0 & 0 & 0 & \beta _{52 } & -\beta _{52} \end{array} \right ]\) only six are binding. The test associated with this structure for β is χ 6 2 = 6. 8291, which is accepted at the 5% level based on a p-value=[0.3369].
- 17.
For every row and column of \(\Pi\) selected, there is an equivalent r dimensioned sub-matrix of α and β. To determine an appropriate orientation of the system the sub-matrices selected need to be of full rank.
- 18.
In the case where more complex restrictions apply, then the general restriction condition and procedure in Doornik and Hendry (2001) apply.
- 19.
Here β is identifiable for the restrictions in (I) when the selected columns of \(\Pi\) yield a matrix A of rank r.
- 20.
For n = 4, a more complex example, the approach discussed above can be shown to identify. Let:
$$\displaystyle{ \beta ^{{\prime}} = \left [\begin{array}{cccc} a&0 & b & c \\ d&e&f &0 \end{array} \right ]\text{ and }B = H_{2} = \left [\begin{array}{cc} a&0\\ d & e \end{array} \right ]. }$$Following Boswijk (1996), identifiability is lost when a normalization is invalid (i.e. a = 0 ⇒ rank(H 2) < r), but with this new restriction [α: β] is over-identified as j = 3 > r 2 − r. Selecting a new orientation, ensuring the generic result associated with Theorem 9 holds, then:
$$\displaystyle{ \beta _{(1)}^{{\prime}} = \left [\begin{array}{cccc} 0 & 0 & b & c \\ d&e&f &0 \end{array} \right ]\text{ and }B_{(1)} = \left [\begin{array}{cc} b & c\\ f &0 \end{array} \right ]. }$$This orientation is rejected when x t  ∼ I(1), f = 0 and α is not identifiable. But the following orientation for x t  ∼ I(1), implies:
$$\displaystyle{ \beta _{(2)}^{{\prime}} = \left [\begin{array}{cccc} 0 & 0 & b & c \\ d&e&0 & 0 \end{array} \right ],\text{ }B_{(2)} = \left [\begin{array}{cc} 0 & b\\ e &0 \end{array} \right ] }$$and rank(B) = r. Now, [α: β (2)] is always empirically identified and identifiable.
- 21.
Notice, that price homogeneity is a long-run property of LEPT. This means that in the short-run agents may mistake relative and absolute price movements. However, long-run pricing that does not satisfy this property would not appear to be consistent with competitive behaviour.
- 22.
The matrices α ij and β ij have the dimensions n i × r j , for i = 1, 2 and j = 1, 2. For example, the matrix β is partitioned into two blocks of columns, β . 1 of dimensions n × r 1, and β . 2 of dimensions n × r 2, then each block is itself cut into two blocks of rows.
- 23.
In the limit there are r such sub-blocks, which leads to the identification case considered by Boswijk (1992) where \(\alpha = \left [\begin{array}{c} I\\ 0 \end{array} \right ].\)
- 24.
The original source of the data is the National Institute of Economic Research, kindly passed on to us by Paul Fisher and Ken Wallis.
- 25.
The model in Hunter (1992) is massively over-identified. It is possible to identify subject to restrictions on both α and β. Here we will concentrate on identification from α alone.
- 26.
The discovery of four valid solutions implies that the model has four over-identifying restrictions.
- 27.
If the determinant is tested for any sub-matrix of α then it is found that no such combination with a non-zero determinant appears to exist.
- 28.
It is also interesting to speculate that the plot of the IRF appears to converge to values that look like the solution to a difference equation. In the cointegration case there is at least one unit root or a single stochastic trend. If there are roots that are complex, and this may arise when there is a second order difference equation, then to preclude a solution that is not real the roots must consist of a complex conjugate pair which may be of unit modulus. These ought to relate back to the roots of the VAR, but this is associated with differences and not simply inverses, because the VAR has real roots, one of which is a unit root.
- 29.
Roots in the paper are the reciprocals of those normally reported, thus a root less than one in modulus is a stationary root. On this basis the roots of the process are, respectively: {0.5, 0.5, 0.5, 0.5}, {0.5, 0.5, 0.95, 0.95}, {0.5, 0.5, 0.99, 0.99}, {0.5, 0.5, 1.0, 1.0}, {1.0, 1.0. 1.0, 1.0}.
- 30.
An alternative study would be one based on perturbations of the cointegrated model, model 4, that retained the common feature, but moved it from being at the unit root to being further outside the unit circle. This would mean that the processes became stationary, and more solidly so, but retained the reduced rank property key to cointegration. In this way, it is possible to isolate two aspects of the problem with potentially different impacts: stationarity and common features (reduced rank).
- 31.
- 32.
Figure 5.2d also shows clearly that, in this case, under-specification of the cointegrating rank is not harmful to forecast performance (including imposing unit roots but non-stationarity), whereas over-specification leads to a deterioration in forecasting performance.
- 33.
Though they do not establish whether it is the imposition of any false restriction that matters, or that of unit roots in particular. This is the point made by Clements and Hendry. They also do not consider if the near unit root is a common feature, or if restricting it to being so would be advantageous.
- 34.
The information criterion can also be written in terms of the eigenvalues of the underlying problem, and hence in terms of the test statistics.
References
Ahn, S. K., Hong, H., & Cho, S. (2015). Estimation of cointegrated models with exogenous variables. Journal of Statistical Computation and Simulation, 85(10), 1945–1962.
Bauwens, L., & Hunter J. (2000). Identifying Long-run Behaviour with Non-Stationary Data. Discussion Paper CORE. The Catholic University: Louvain-La Nueve DP 2000/43.
Bauwens, L., Lubrano, M., & Richard J.-F. (2000). Bayesian inference in dynamic econometric Models. Oxford: Oxford University Press.
Boswijk, H. P. (1992). Cointegration, identification and exogeneity: Inference in structural error correction models. Amsterdam: Thesis Publishers.
Boswijk, H. P. (1996). Cointegration, identification and exogeneity: Inference in structural error correction models. Journal of Business and Economics and Statistics, 14, 153–160.
Burke, S. P. (1996). Some reparameterizations of lag polynomials for dynamic analysis. Oxford Bulletin of Economics and Statistics, 58, 373–389.
Burke, S. P., & Hunter, J. (2011). Long-run equilibrium price targeting. Quantitative and Qualitative Analysis in Social Sciences, 5, 26–36.
Clements, M. P. (2005). Evaluating econometric forecasts of economic and financial variables. Basingstoke: Palgrave-Macmillan.
Clements, M. P., & Hendry, D. F. (1995). Forecasting in cointegrated systems. Journal of Applied Econometrics, 10, 127–146.
Clements, M. P., & Hendry, D. F. (1998). Forecasting economic time series. Cambridge: Cambridge University Press.
Clements, M. P., & Hendry, D. F. (2001). Forecasting non-stationary economic time series. London: The MIT Press.
Dhrymes, P. J. (1984). Mathematics for econometrics. New York: Springer-Verlag.
Doornik, J. A. (1995). Testing general restrictions on the cointegration space. Oxford: Mimeo, Nuffield College.
Doornik, J. A. (1998). Approximations to the asymptotic distribution of cointegration tests. Journal of Economic Surveys, 12, 573–593.
Doornik, J. A., & Hendry, D. F. (2001). PCFIML 10. London: Timberlake Consultants Press.
Doornik, J. A., & Hendry, D. F. (2009). Oxmetrics 6. London: Timberlake Consultants Press.
Doornik, J. A., & Hendry, D. F. (2013). Oxmetrics 7. London: Timberlake Consultants Press.
Dunne, J. P., & Hunter, J. (1998, August). The allocation of government expenditure in the UK: A forward looking dynamic model. Paper presented at the International Institute of Public Finance Conference, Cordoba, Argentina.
Engle, R. F., Hendry, D. F., & Richard, R. F. (1983) Exogeneity. Econometrica, 51, 277–304.
Engle, R. F., & Yoo, B. S. (1987). Forecasting and testing in co-integrated systems. Journal of Econometrics, 35, 143–159.
Engle, R. F., & Yoo, B. S. (1991). Cointegrated time series: An overview with new results (Chapter 12). In R. F. Engle and C. W. J. Granger (Eds.), Long-run economic relationships. Oxford: Oxford University Press.
Ericsson, N. R. (1994). In N. R. Ericsson & J. S. Irons (Eds.), Testing exogeneity: An introduction, in testing exogeneity (pp. 3–38). Oxford: Oxford University Press.
Ericsson, N. R., Hendry, D. F., & Mizon, G. E. (1998). Exogeneity, cointegration and economic policy analysis. Journal of Business and Economics Statistics, 16, 371–387.
Ericsson, N. R., & Irons, J. S. (1994). Testing exogeneity. Oxford: Oxford University Press.
Fisher, P. G., Tanna, S. K., Turner, D. S, Wallis, K. F., & Whitley, J. D. (1990). Econometric evaluation of the exchange rate in models of the UK economy. Economic Journal, 100, 1024–1056.
Florens, J.-P., Mouchart, M., & Rolin, J.-M. (1990). Sequential experiments, Chapter 6 in Elements of bayesian statistics. New York: Marcel Dekker.
Granger, C. W. J., & Lee, T.-H. (1989). Multicointegration. In T. B. Fomby & G. F. Rhodes (Eds.), Co-integration, spurious regressions, and unit roots. Advances in Econometrics (Vol. 8, pp. 71–84). London: JAI Press.
Hall, S. J., & Wickens, M. (1993). Causality in Integrated systems. Centre for Economic Forecasting, Discussion paper 27-93. London Business School.
Hendry, D. F., & Mizon, G. E. (1993). Evaluating dynamic econometric models by encompassing the VAR. Chapter 18 in P. C. B. Phillips (Ed). Models, Methods and Applications of Econometrics: Essays in Honour of A.R. Bergstrom (pp. 272–300). Cambridge, MA: Blackwell Publishers.
Hunter, J. (1990). Cointegrating exogeneity. Economics Letters, 34, 33–35.
Hunter, J. (1992). Tests of cointegrating exogeneity for PPP and uncovered interest rate parity for the UK. Journal of Policy Modelling,Special Issue: Cointegration, Exogeneity and Policy Analysis, 14, 453–463.
Hunter, J., & Burke, S. P. (2007). Common trends, cointegration and competitive price behaviour. In Royal Economic Society Conference, Warwick.
Hunter, J., & Simpson, M. (1995) Exogeneity and identification in a model of the UK effective exchange rate. Paper presented at the EC2 Conference in Aarhus Dec (1995) and the Econometrics Society European Meeting in Istanbul 1996.
Johansen, S. (1991). Estimation and hypothesis testing of cointegrating vectors in gaussian vector autoregressive models. Econometrica, 59, 1551–1580.
Johansen, S. (1992a) Testing weak exogeneity and the order of cointegration in UK money demand data. Journal of Policy Modelling, Special Issue: Cointegration, Exogeneity and Policy Analysis, 14, 313–334.
Johansen. S. (1992b). Cointegration in partial systems and the efficiency of single equation analysis. Journal of Econometrics, 52(3), 389–402.
Johansen, S. (1995a). Likelihood-inference in cointegrated vector auto-regressive models. Oxford: Oxford University Press.
Johansen, S. (1995b). Identifying restrictions of cointegrating vectors. Journal of Econometrics, 69, 111–132.
Johansen, S. (2002). A small sample correction for the test of cointegrating rank in the vector autoregressive model. Econometrica, 70, 1929–1961.
Johansen, S., & Juselius, K. (1990). Maximum likelihood estimation and inference on cointegration - with applications to the demand for money. Oxford Bulletin of Economics and Statistics, 52, 169–210.
Johansen, S., & Juselius, K. (1992). Some structural hypotheses in a multivariate cointegration analysis of the purchasing power parity and the uncovered interest parity for UK. Journal of Econometrics, 53, 211–244.
Johansen, S., & Juselius, K. (1994). Identification of the long-run and the short-run structure: An application to the IS/LM model. Journal of Econometrics, 63, 7–36.
Juselius, K. (1994). Do PPP and UIRP hold in the long-run? - An example of likelihood inference in a multivariate time-series model. Paper presented at Econometric Society European Meeting 1994, Maastricht.
Juselius, K. (1995). Do purchasing power parity and uncovered interest rate parity hold in the long-run?- An example of lilkelihood inference in a multivariate time-series model. Journal of Econometrics, 69, 178–210.
Koopmans, T. C. (1953). In T. C. Koopmans & W. C. Hood (Eds.), Identification problems in economic model construction. In studies in econometric method. Cowles Commission Monograph (Vol. 14). New York: John Wiley and Sons.
Lin, J.-L., & Tsay, R. S. (1996). Co-integration constraint and forecasting: An empirical examination. Journal of Applied Econometrics, 11, 519–538.
Lippi, M., & Reichlin, L. (1994). VAR analysis, non-fundamental representations, Blaschke matrices.Journal of Econometrics, 63, 290–307.
Lütkepohl, H. (2006). New introduction to multiple time series analysis. Berlin: Springer-Verlag.
Mosconi, R., & Giannini, C. (1992). Non-causality in cointegrated systems: representation, estimation and testing. Oxford Bulletin of Economics and Statistics, 54, 399–417.
Parker, S. (1998). Opening a Can of Worms: The Pitfalls of Time Series Regression Analyses of Income Inequality. Brunel University Discussion Paper, 98-11.
Pesaran, M. H., Shin, Y., & Smith, R. J. (2000). Structural analysis of vector error correction models with exogenous I(1) variables. Journal of Econometrics, 97, 293–343.
Podivinsky, J. M., (1992). Small sample properties of tests of linear restrictions on cointegrating vectors and their weights. Economics Letters, 39, 13–18.
Reimers, H.-E. (1992). Comparisons of tests for multivariate cointegration. Statistical Papers, 33, 335–359.
Rothenberg, T. J. (1971). Identification in parametric models. Econometrica, 39, 577–591.
Sargan, J. D. (1975). The Identification and Estimation of Sets of Simultaneous Stochastic Equations. LSE Discussion Paper No A1.
Sargan, J. D. (1983a). Identification and lack of identification. Econometrica, 51, 1605–1633.
Sargan, J. D. (1983b). Identification in models with autoregressive errors. In S. Karlin, T. Amemiya, & L. A. Goodman (Eds.), Studies in econometrics, time series and multivariate statistics (pp. 169–205). New York: Academic Press.
Sargan, J. D. (1988). Lectures on advanced econometric theory. Oxford: Basil Blackwell.
Toda, H. Y., & Phillips, P. C. B. (1994). Vector autoregression and causality: A theoretical overview and simulation study. Econometric Reviews, 13, 259–285.
Yoo, S. (1986). Multi-cointegrated Time Series and Generalised Error-Correction Models. University of San Diego Working Paper.
Author information
Authors and Affiliations
Copyright information
© 2017 The Author(s)
About this chapter
Cite this chapter
Hunter, J., Burke, S.P., Canepa, A. (2017). Structure and Evaluation. In: Multivariate Modelling of Non-Stationary Economic Time Series. Palgrave Texts in Econometrics. Palgrave Macmillan, London. https://doi.org/10.1057/978-1-137-31303-4_5
Download citation
DOI: https://doi.org/10.1057/978-1-137-31303-4_5
Published:
Publisher Name: Palgrave Macmillan, London
Print ISBN: 978-0-230-24330-9
Online ISBN: 978-1-137-31303-4
eBook Packages: Economics and FinanceEconomics and Finance (R0)