This survey paper discusses the Cointegrated Vector AutoRegressive (CVAR) methodology and how it has evolved over the past 30 years. It describes major steps in the econometric development, discusses problems to be solved when confronting theory with the data, and, as a solution, proposes a so-called theory-consistent CVAR scenario. A number of early CVAR applications are motivated by the urge to find out why the empirical results did not support Milton Friedman’s concept of monetary inflation. The paper also proposes a method for combining partial CVAR analyses into a large-scale macroeconomic model. It argues that an empirically-based approach to macroeconomics preferably should be based on Keynesian disequilibrium economics, where imperfect knowledge expectations replace so called rational expectations and where the financial sector plays a key role for understanding the long persistent movements in the data. Finally, the paper argues that the CVAR is potentially a candidate for Haavelmo’s “design of experiment for passive observations” and provides several illustrations.

中文翻译：

---

寻找适合数据的理论：个人研究奥德赛

这份调查报告讨论了协整矢量自回归（CVAR）方法及其在过去30年中的发展。它描述了计量经济学发展的主要步骤，讨论了将理论与数据相面对时需要解决的问题，并提出了一种所谓的理论一致的CVAR方案。渴望发现为什么经验结果不支持米尔顿·弗里德曼（Milton Friedman）的货币通胀概念的动机激发了许多早期的CVAR应用。本文还提出了一种将部分CVAR分析结合到大规模宏观经济模型中的方法。它认为，基于经验的宏观经济学方法最好应基于凯恩斯主义的不平衡经济学，不完善的知识期望取代了所谓的理性期望，而金融部门在理解数据长期持续变化中扮演着关键角色。最后，本文认为CVAR可能是Haavelmo的“被动观测实验设计”的候选人，并提供了一些说明。