Learning from learners

Tom Holden

No 1512, School of Economics Discussion Papers from School of Economics, University of Surrey

Abstract: Traditional macroeconomic learning algorithms are misspecified when all agents are learning simultaneously. In this paper, we produce a number of learning algorithms that do not share this failing, and show that this enables them to learn almost any solution, for any parameters, implying learning cannot be used for equilibrium selection. As a by-product, we are able to show that when all agents are learning by traditional methods, all deep structural parameters of standard new-Keynesian models are identified, overturning a key result of Cochrane (2009; 2011). This holds irrespective of whether the central bank is following the Taylor principle, irrespective of whether the implied path is or is not explosive, and irrespective of whether agents’ beliefs converge. If shocks are observed then this result is trivial, so following Cochrane (2009) our analysis is carried out in the more plausible case in which agents do not observe shocks.

Keywords: Identification; Learnability; Limited Information; Indeterminacy; Taylor Rules (search for similar items in EconPapers)
JEL-codes: D83 E3 E52 (search for similar items in EconPapers)
Pages: 39 pages
Date: 2012-10
New Economics Papers: this item is included in nep-mac
References: View references in EconPapers View complete reference list from CitEc
Citations:

Downloads: (external link)
https://repec.som.surrey.ac.uk/2012/DP15-12.pdf (application/pdf)

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:sur:surrec:1512

Access Statistics for this paper

More papers in School of Economics Discussion Papers from School of Economics, University of Surrey Contact information at EDIRC.
Bibliographic data for series maintained by Ioannis Lazopoulos ().