Econometric Analysis of Switching Expectations in UK Inflation*
| Published date | 01 June 2022 |
| Author | Adriana Cornea‐Madeira,João Madeira |
| Date | 01 June 2022 |
| DOI | http://doi.org/10.1111/obes.12479 |
Econometric Analysis of Switching Expectations in
UK Inflation*
ADRIANA CORNEA-MADEIRA and JO
˜
AO MADEIRA
University of York, York YO10 5GD, UK (e-mails: adriana.cornea-madeira@york.ac.uk;
joao.madeira@york.ac.uk)
Abstract
We estimate with UK data a Phillips curve model with backward-looking and forward-
looking methods of determining inflation expectations and with agents switching
between these based on their recent performance. We find that, while on average
backward-looking and forward-looking methods have about equal weight, there are
considerable movements in the weight given to each method. We show this model has
better in-sample fit than other Phillips curve models and this is robust to the
methodology chosen. The model out-of-sample forecasts on certain dates do better than
other Phillips curve models and the Atkeson and Ohanian model.
I. Introduction
In this paper, we examine how inflation expectations have evolved over time in the
UK by switching between backward (the forecast corresponds to lagged inflation) and
forward-looking or fundamentalist (the forecast is based on a discounted sequence of
expected future output gaps, as implied by a purely forward-looking Phillips curve)
rules.
1
We do this by estimating with UK data on inflation and output gap a model of
inflation dynamics, developed by Cornea-Madeira, Hommes and Massaro (2019), that
allows for endogenous switching between different expectation rules. The model
allows agents to switch between backward-looking and forward-looking rules based on
their recent relative forecast performance (as in Brock and Hommes, 1997).
Our motivation is to show not only that endogenous switching between different
expectation rules is a better description of human behaviour (as documented for
example in lab experiments, see Hommes, 2011) than the extreme cases of rational
expectations (which assumes agents have the best possible understanding of how the
JEL Classification numbers: C22, E31, E37.
*We would like to thank the editor Francesco Zanetti and three anonymous reviewers for useful comments
and suggestions. Any errors are ours alone.
1
There have been previous studies which have studied UK inflation using forward-looking expectations (see
e.g. Zanetti, 2011; Faccini, Millard and Zanetti, 2013) or both backward and forward-looking expectations (see
e.g. Batini, Jackson and Nickell, 2005) but to the best of our knowledge no previous work has considered
endogenous switching between forecast rules.
651
©2021 Oxford University and John Wiley & Sons Ltd.
OXFORD BULLETIN OF ECONOMICS AND STATISTICS, 84, 3 (2022) 0305-9049
doi: 10.1111/obes.12479
economy works) or adaptative/naive expectations (which assumes that agents have no
knowledge of how the economy works) but also that endogenous switching will
improve the Phillips curve fittoinflation data (in-sample and out-of-sample).
We find that in the UK on average backward-looking and forward-looking
behaviour have about equal weight.
2
There are however substantial fluctuations in how
expectations are formed over time.
3
In some periods forward-looking behaviour is
dominant (with the weight on the fundamental forecasting rule on occasion exceeding
0.8) while in others backward-looking behaviour is dominant (with the weight on the
fundamental forecasting rule on occasion below 0.1). For example, we found that after
the 1973 and 1979 oil price crises that there was an increase in backward-looking
behaviour. We also found a large and persistent increase in the backward-looking
behaviour in the early years of the XXI century, which seems to be due to a prolonged
period in which UK inflation was below fundamentals (as determined by the present-
value of expected future output gaps). Finally, we show that the cyclical behaviour in
adoption of forecast rules changed over time in the UK. Prior to the inflation targeting
regime (which started at the end of 1992) the estimated series for the weight on the
fundamentalist rule was much more volatile. This indicates that inflation expectations
became more anchored under inflation targeting.
The estimation of the Phillips curve model with endogenous switching is shown to
be quite robust to inclusion of different variables in estimation of the fundamentalist
rule forecast, use of different number of lags in estimation of the fundamentalist rule
forecast and changes to the driving variable in the vector autoregressive (VAR) model
used in estimation of the fundamentalist rule forecast.
The Bayesian information criterion and Akaike information criterion both show that
the Phillips curve model with endogenous switching in forecast rules accounts better
for in-sample inflation dynamics than other Phillips curve models. We also show that
the Phillips curve model with endogenous switching in forecast rules forecasts better
out-of-sample in some periods than any of the competing Phillips curve models. Also,
the Phillips curve model with endogenous switching in forecast rules does better in
out-of-sample forecasting in some periods than the random walk model of Atkeson and
Ohanian (2001), which has been shown to be one of the best econometric models in
forecasting US inflation (Faust and Wright, 2013). Moreover, the Phillips curve model
with endogenous switching in forecast rules has the advantage over econometric
models that it is not as vulnerable to the Lucas (1976) critique (which argues that it
can be misleading to try to predict the effects of a change in economic policy entirely
on the basis of relationships observed in historical data).
Aside from Cornea-Madeira et al. (2019) our work is related to other empirical
applications of the Brock and Hommes (1997) mechanism for endogenous switching in
2
For the US (see Cornea-Madeira et al., 2019) the backward-looking rule has on average about twice the
weight of the forward-looking rule (one reason for this is that inflation is much less persistent in the UK, with
first order autocorrelation only about 0.7, whereas in the US the value is about 0.9 for the same period).
3
This is consistent with experimental evidence at the micro level that found individuals adjust expectations in
response to past forecast errors (see Bloomfield and Hales, 2002; Hommes, 2011) and with the finding of
multiple structural breaks in the coefficient estimates from US macro data of the backward-looking and forward-
looking terms of Phillips curve models (see Zhang, Osborn and Kim, 2008; Hall, Han and Boldea, 2012).
©2021 Oxford University and John Wiley & Sons Ltd.
652 Bulletin
Get this document and AI-powered insights with a free trial of vLex and Vincent AI
Get Started for FreeStart Your Free Trial of vLex and Vincent AI, Your Precision-Engineered Legal Assistant
-
Access comprehensive legal content with no limitations across vLex's unparalleled global legal database
-
Build stronger arguments with verified citations and CERT citator that tracks case history and precedential strength
-
Transform your legal research from hours to minutes with Vincent AI's intelligent search and analysis capabilities
-
Elevate your practice by focusing your expertise where it matters most while Vincent handles the heavy lifting
Start Your Free Trial of vLex and Vincent AI, Your Precision-Engineered Legal Assistant
-
Access comprehensive legal content with no limitations across vLex's unparalleled global legal database
-
Build stronger arguments with verified citations and CERT citator that tracks case history and precedential strength
-
Transform your legal research from hours to minutes with Vincent AI's intelligent search and analysis capabilities
-
Elevate your practice by focusing your expertise where it matters most while Vincent handles the heavy lifting
Start Your Free Trial of vLex and Vincent AI, Your Precision-Engineered Legal Assistant
-
Access comprehensive legal content with no limitations across vLex's unparalleled global legal database
-
Build stronger arguments with verified citations and CERT citator that tracks case history and precedential strength
-
Transform your legal research from hours to minutes with Vincent AI's intelligent search and analysis capabilities
-
Elevate your practice by focusing your expertise where it matters most while Vincent handles the heavy lifting
Start Your Free Trial of vLex and Vincent AI, Your Precision-Engineered Legal Assistant
-
Access comprehensive legal content with no limitations across vLex's unparalleled global legal database
-
Build stronger arguments with verified citations and CERT citator that tracks case history and precedential strength
-
Transform your legal research from hours to minutes with Vincent AI's intelligent search and analysis capabilities
-
Elevate your practice by focusing your expertise where it matters most while Vincent handles the heavy lifting
Start Your Free Trial of vLex and Vincent AI, Your Precision-Engineered Legal Assistant
-
Access comprehensive legal content with no limitations across vLex's unparalleled global legal database
-
Build stronger arguments with verified citations and CERT citator that tracks case history and precedential strength
-
Transform your legal research from hours to minutes with Vincent AI's intelligent search and analysis capabilities
-
Elevate your practice by focusing your expertise where it matters most while Vincent handles the heavy lifting