TY - UNPB

T1 - Haavelmo's Probability Approach and the Cointegrated VAR

AU - Juselius, Katarina

N1 - JEL classi…cation: B16, B31, B41, C32, C82

PY - 2012/4

Y1 - 2012/4

N2 - Some key econometric concepts and problems addressed by TrygveHaavelmo and Ragnar Frisch are discussed within the general frame-work of a cointegrated VAR. The focus is on problems typical of time-series data such as multicollinearity, spurious correlation and regres-sion results, time dependent residuals, normalization, reduced rank,model selection, missing variables, simultaneity, autonomy and iden-ti…cation. Speci…cally the paper discusses (1) the conditions underwhich the VAR model represents a full probability formulation of asample of time-series observations, (2) the plausibility of the multivari-ate normality assumption underlying the VAR, (3) cointegration as asolution to the problem of spurious correlation and multicollinearitywhen data contain deterministic and stochastic trends, (4) the exis-tence of a universe, (5) the association between Frisch’s con‡uenceanalysis and cointegrated VAR analysis, (6) simultaneity and iden-ti…cation when data are nonstationary, (7) conditions under whichidenti…ed cointegration relations can be considered structural or au-tonomous, and …nally (8) a formulation of a design of experiment forpassive observations based on theory consistent CVAR scenarios illus-trated with a monetary model for in‡ation.

AB - Some key econometric concepts and problems addressed by TrygveHaavelmo and Ragnar Frisch are discussed within the general frame-work of a cointegrated VAR. The focus is on problems typical of time-series data such as multicollinearity, spurious correlation and regres-sion results, time dependent residuals, normalization, reduced rank,model selection, missing variables, simultaneity, autonomy and iden-ti…cation. Speci…cally the paper discusses (1) the conditions underwhich the VAR model represents a full probability formulation of asample of time-series observations, (2) the plausibility of the multivari-ate normality assumption underlying the VAR, (3) cointegration as asolution to the problem of spurious correlation and multicollinearitywhen data contain deterministic and stochastic trends, (4) the exis-tence of a universe, (5) the association between Frisch’s con‡uenceanalysis and cointegrated VAR analysis, (6) simultaneity and iden-ti…cation when data are nonstationary, (7) conditions under whichidenti…ed cointegration relations can be considered structural or au-tonomous, and …nally (8) a formulation of a design of experiment forpassive observations based on theory consistent CVAR scenarios illus-trated with a monetary model for in‡ation.

KW - Faculty of Social Sciences

KW - Haavelmo

KW - CVAR

KW - autonomy

KW - identi¿cation

KW - passive observations

M3 - Working paper

BT - Haavelmo's Probability Approach and the Cointegrated VAR

PB - Department of Economics, University of Copenhagen

ER -