Imperfect Information and Consumer Inflation Expectations: Evidence from Microdata
| DOI | http://doi.org/10.1111/obes.12189 |
| Author | Michael J. Lamla,Lena Dräger |
| Date | 01 December 2017 |
| Published date | 01 December 2017 |
933
©2017 The Department of Economics, University of Oxford and JohnWiley & Sons Ltd.
doi: 10.1111/obes.12189
Imperfect Information and Consumer Inflation
Expectations: Evidence from Microdata*
Lena Dr ¨
ager† and Michael J. Lamla‡
†Johannes Gutenberg-University Mainz, (e-mail: ldraeger@uni-mainz.de)
‡University of Essex and ETH Zurich, KOF Swiss Economic Institute,
(e-mail: mlamla@essex.ac.uk)
Abstract
This paper explores which factors trigger an adjustment in consumers’ inflation expecta-
tions and looks at the implications regarding forecast errors. We find support for imperfect
information models, as inflation volatility and news trigger an adjustment in expectations.
Furthermore, we document that individual expectations become more accurate if they have
been adjusted.
1. Introduction
Inflation expectations play a central role in modern macroeconomic models and are an
important factor for economic policy. Despite their importance, we still know relatively
little about how people form their expectations. Researchers have proposed a wide arrayof
frameworks to model the expectations formation process. Contributions by Sims (2003),
Mankiw and Reis (2002) or Woodford (2001), revived the interest in information rigidities
and highlight their importance for the process of forming inflation expectations.1Account-
ing for imperfect information allows to solve several empirical puzzles that did not match
the predictions of the full-information rational expectations models, as shown in Ball et al.
(2005).2
Although recent approaches have used micro survey data of professional forecasters
to identify the degree of information rigidities (Andrade and Le Bihan, 2013; An et al.,
2016), so far there is only little evidence for consumers. The idea of this paper is thus
to use the updating behaviour regarding inflation expectations of individual consumers in
JEL Classification numbers: D84, E31.
*The authors would like to thank the editor and two anonymous referees as well as Robert Anderson, Olivier
Coibion, Damjan Pajfar, Ricardo Reis as well as seminar and conference participants at the 2014 ASSA meeting in
Philadelphia, the University of Bonn, the University of Hamburg, the Universityof Essex and the National Bank of
Poland for helpful comments and suggestions. All remaining errors are our own.
1Similarly, the literature on the importance of learning for the expectation formation process discusses how
individuals use past data to learn about the true data generating process over time (Evans and Honkapohja, 2001;
Malmendier and Nagel, 2016).
2For an overviewsee also Mankiw and Reis (2011).
OXFORD BULLETIN OF ECONOMICSAND STATISTICS, 79, 6 (2017) 0305–9049
934 Bulletin
order to provide evidence of imperfect information using a better identification based on
micro survey data. We calculate the updating frequency overtime and test which macroeco-
nomic factors trigger an adjustment of inflation expectations.3In doing so, we empirically
test the relevance of theoretical imperfect information models, such as rational inatten-
tion as proposed by Sims (2003) or sticky information as introduced by Mankiw and Reis
(2002). Furthermore, we explore if updating improves the forecast accuracy of expecta-
tions.
For our analysis we make use of the rotating panel microstructure of the University of
Michigan Survey of Consumers, where a fraction of individuals is re-interviewed after 6
months. This allows us to track individuals and their expectations with two observations
over a period of 6 months. Hence, we can directly calculate the change in individual
expectations and the share of individuals that have adjusted their expectations and, thus,
do not need to rely on identification coming only from the cross section or the aggregated
series.
In Dr¨ager and Lamla (2012) we have shown that there is a pronounced time variation
in the updating frequency. This pronounced time-variation in the qualitative expectation
updating share calls into question the assumption of a constant in standard sticky infor-
mation models and favours noisy information models or time-varying sticky information
models. While this is interesting per se as it is useful for providing guidance in calibrating
the updating probability in sticky information models, it does not tell us why we observe
this time variation.At a micro level, both sticky information and rational inattention models
posit that the optimal degree of attentiveness is a function of signals, such as the volatility
of the forecast variable and news (Sims, 2003; Reis, 2006). In this paper, we thus test
the evidence of imperfect information in consumers’ inflation expectations and derive hy-
potheses from both rational inattention and sticky information models regarding agents’
expectation updating behaviour.
Our results provide evidence in favour of imperfect information, since we find that
measures of the volatility of inflation raise attention and consequently trigger an updating
of inflation expectations by consumers. In addition, we find support for news effects: If
people have heard news on inflation, they are more likely to adjust their expectations.
Moreover, we explore whether adjusting expectations has beneficial implications for
the accuracy of inflation expectations. Besides proving theories of rational attention it is
extremely relevant to check whether adjusting expectations has improved the situation of
individuals implying that they not only adjust expectations but also improve the forecast
and thereby allow for better economic action in response to this adjustment. Indeed, we
can report that forecast accuracy increases if inflation expectations are adjusted.
To our knowledge, this is the first paper that uses the rotating panel dimension of
the microdata in the University of Michigan Survey of Consumers to test for predictions
from theories of expectation formation under imperfect information regarding consumers’
inflation expectations. Nevertheless, there exist several approaches in the literature that
test consumers’ expectation formation for evidence of information frictions in aggregate
data. For the US, Carroll (2003) finds support in aggregate survey data for the conjecture
3Note that we discussed the data set and the calculation of the updating shares in Dr¨ager and Lamla (2012).
Consequently, in this paper we focus on whether we can explain the movements in the updating frequencies and
explore the consequences for forecast accuracy.
©2017 The Department of Economics, University of Oxford and JohnWiley & Sons Ltd
Imperfect information and consumer inflation expectations 935
of Mankiw and Reis (2002) that consumers update their quantitative inflation expectations
roughly once a year.For Europe, D ¨opke et al. (2008a) estimate that consumers update their
inflation expectations once every 18 months.4Mankiw, Reis and Wolfers (2003) as well
as Branch (2007) use aggregate survey data of consumers’ inflation expectations to test
for specific features of the sticky information model, while Lamla and Sarferaz (2012)
document substantial time-variation in the inflation expectation updating behaviour of
German households.
Additionally, there exists evidence of information frictions in micro survey data for
professional forecasters: Using microdata for inflation expectations of professional fore-
casters,Andrade and Le Bihan (2013) report evidence of information frictions as forecasters
show staggered updating of expectations and persistent disagreement. An et al. (2016) use
a similar measure to derive an aggregate inattention measure from professional forecasters’
inattention to several variables and report substantial time-variation in inattention over the
business cycle. Looking at the movements of forecast errors in relation to the variable be-
ing forecasted, Coibion and Gorodnichenko (2012, 2015) document pervasive and robust
evidence consistent with information rigidities and derive a test for information frictions
consistent under both sticky information and models of imperfect information as in Wood-
ford (2001). Similarly, Dovern et al. (2015) use a related approach with both aggregate
and individual survey data of professional forecasts in a large country panel. Our approach
differs from these studies in that we focus on consumers’expectations. Specifically,we em-
ploy the rotating panel feature of the Michigan Survey for Consumers to identify updates
in inflation expectations and derive hypotheses from limited information theories that we
test in the regression analysis. We therefore focus on the determinants of the likelihood of
an individual expectation update.
So far, only a few studies use the rotating panel dimension of the University of
Michigan Survey of Consumers. Souleles (2004) employs the rotating panel to con-
struct individual forecast errors, which are then subjected to rationality tests and eval-
uated with respect to their forecasting power regarding household expenditure. Ander-
son et al. (2010) analyse differences in the formation of consumers’ inflation expecta-
tions and their forecast errors conditional on sociodemographic characteristics. Pfajfar
and Santoro (2013) test the hypotheses of the epidemiology model proposed by Carroll
(2003). Finally, Bachmann et al. (2015) test for a possible link between individual con-
sumers’ inflation expectations and their reported readiness to spend on large consumer
goods.
The paper is structured as follows. We discuss models with imperfect information
in section II, where we derive our hypotheses for the empirical analysis. In section III we
discuss the data set used in the analysis. Empirical results regarding the updating behaviour
of individual inflation expectations and their forecast errors are presented in section IV.
Section V concludes.
II. Theoretical determinants of imperfect information
In order to derive testable hypotheses for individuals’ updating behaviour, we present
two simple models of inflation expectations under imperfect information. Following the
4For evidence regarding the expectationsupdating by professional forecasters, see D ¨opke et al. (2008b).
©2017 The Department of Economics, University of Oxford and JohnWiley & Sons Ltd
To continue reading
Request your trial