down’’ responses to question A12 or A12a are then asked
the follow-up question A12b to obtain their quantitative
expectations. If a respondent says ‘‘don’t know’’ to A12b,
question A12c is used. All of the quantitative responses exceeding
5% are probed with one additional question:
Probe. Let me make sure I have that correct. You said that
you expect prices to go (up/down) during the next 12
months by (X) percent. Is that correct?
Quantitative responses to A12b and A12c are recorded as a
percentage in a whole number. Responses exceeding 50%
(in absolute value) are replaced with 50% before the data
are released. Also, as the order of the questions suggests, it
is not possible to have mismatched quantitative and qualitative
responses; e.g., for a qualitative response of ‘‘go up’’,
the quantitative response must be positive. After the above
questions have been used to obtain consumers’ one-yearahead
expectations, a similar set of questions is used to obtain
consumers’ price outlooks for the next 5–10 years. In
this study, we only quantify the one-year-ahead expectations,
which are more likely to be close to the actual inflation
rates; however, the discussions and the approach
presented in this paper are equally applicable for quantifying
the long run expectations.
The sample used in this study is based on 414 monthly
surveys, covering the period from January 1978 to June
2012. A total of 236,865 individuals are included in our
sample. As we do not use the quantitative data during the
process of quantification of qualitative responses, we retain
all of the non-missing qualitative responses, even if
a corresponding quantitative response is missing.4 Fig. 1
shows the evolution of the mean of the quantitative survey
responses, compared with the actual inflation rate for the
next 12 months, over which the expectations were held.
The inflation rate is defined as the percentage change from
a year ago. The shaded areas represent NBER-dated recessions.
Since these are one-year-ahead expectations from
households instead of professional forecasters, it is not surprising
that they sometimes deviate substantially from the
actual inflation rate. This could be due to unanticipated
shocks, information stickiness, heterogeneity in the interpretation
of the same information, and other imperfections
and irrationalities in individual expectation formation processes
(see, among others, Henzel & Wollmershäuser,
2005; Mankiw & Reis, 2002). Therefore, the Carlson–Parkin
procedure may produce quantified expectations that differ
substantially from the ex post actual inflation rate.
The effectiveness of alternative quantification methods
can be analyzed better using the quantitative inflation expectations
from the survey, rather than the actual inflation
rates from the US Bureau of Labor Statistics (BLS).
There are several reasons for this. First, consumers are
not sophisticated professional forecasters. It is natural for
their expectations to deviate from the official statistics, as