Cointegration and long-horizon forecasting

被引:74
作者
Christoffersen, PF [1 ]
Diebold, FX
机构
[1] McGill Univ, Fac Management, Montreal, PQ H3A 1G5, Canada
[2] Univ Penn, Dept Econ, Philadelphia, PA 19104 USA
[3] Natl Bur Econ Res, Cambridge, MA 02138 USA
关键词
integration; loss function; prediction; unit root;
D O I
10.2307/1392613
中图分类号
F [经济];
学科分类号
02 ;
摘要
We consider the forecasting of cointegrated variables, and we show that at long horizons nothing is lost by ignoring cointegration when forecasts are evaluated using standard multivariate forecast accuracy measures. In fact, simple univariate Box-Jenkins forecasts are just as accurate. Our results highlight a potentially important deficiency of standard forecast accuracy measures-they fail to value the maintenance of cointegrating relationships among variables-and we suggest alternatives that explicitly do so.
引用
收藏
页码:450 / 458
页数:9
相关论文
共 31 条
[1]   OLS BIAS IN A NONSTATIONARY AUTOREGRESSION [J].
ABADIR, KM .
ECONOMETRIC THEORY, 1993, 9 (01) :81-93
[2]  
ABADIR KM, 1996, 9614 U YORK
[3]  
Banerjee A., 1993, ADV TEXTS ECONOMETRI, DOI [DOI 10.1093/0198288107.001.0001, 10.1093/0198288107.001.0001]
[4]  
BRANDNER P, 1990, 265 I ADV STUD
[5]   COINTEGRATION AND TESTS OF PRESENT VALUE MODELS [J].
CAMPBELL, JY ;
SHILLER, RJ .
JOURNAL OF POLITICAL ECONOMY, 1987, 95 (05) :1062-1088
[6]   Optimal prediction under asymmetric loss [J].
Christoffersen, PF ;
Diebold, FX .
ECONOMETRIC THEORY, 1997, 13 (06) :808-817
[7]  
Christoffersen PF, 1996, J APPL ECONOM, V11, P561, DOI 10.1002/(SICI)1099-1255(199609)11:5<561::AID-JAE406>3.3.CO
[8]  
2-J
[9]  
CHRISTOFFERSEN PF, 1997, 217 NAT BUR EC RES
[10]   ON THE LIMITATIONS OF COMPARING MEAN-SQUARE FORECAST ERRORS [J].
CLEMENTS, MP ;
HENDRY, DF .
JOURNAL OF FORECASTING, 1993, 12 (08) :617-637