Empir Econ (2017) 53:137–163

DOI 10.1007/s00181-016-1163-8

Applying a microfounded-forecasting approach to

predict Brazilian inflation

Wagner Piazza Gaglianone1 · João Victor Issler2 ·

Silvia Maria Matos3

Received: 22 February 2016 / Accepted: 28 July 2016 / Published online: 27 September 2016
© Springer-Verlag Berlin Heidelberg 2016

Abstract We investigate whether combining forecasts from surveys of expectations

is a helpful empirical strategy for forecasting inflation in Brazil. We employ the FGV–
IBRE Economic Tendency Survey, which consists of monthly qualitative information
from approximately 2000 consumers since 2006, and also the Focus Survey of the
Central Bank of Brazil, with daily forecasts since 1999 from roughly 250 profes-
sional forecasters. Natural candidates to win a forecast competition in the literature
of surveys of expectations are the (consensus) cross-sectional average forecasts (AF).

The authors thank the editor Badi H. Baltagi, the anonymous referee and the seminar participants of the
XVIII Annual Inflation Targeting Seminar of the Banco Central do Brasil for the helpful comments and
suggestions. The views expressed in the paper are those of the authors and do not necessarily reflect those
of the Banco Central do Brasil (BCB) or of Getulio Vargas Foundation (FGV). We are also grateful to the
Investor Relations and Special Studies Department (Gerin) of the BCB and to the Brazilian Institute of
Economics (IBRE) of the FGV for kindly providing the data used in this paper. The collection and
manipulation of data from the Market Expectations System of the BCB is conducted exclusively by the
staff of the BCB. Gaglianone, Issler and Matos gratefully acknowledge the support from CNPq, FAPERJ,
INCT and FGV on different grants. The excellent research assistance given by Francisco Luis Lima,
Marcia Waleria Machado and Andrea Virgínia Machado are gratefully acknowledged.

B João Victor Issler

jissler@fgv.br
Wagner Piazza Gaglianone
wagner.gaglianone@bcb.gov.br
Silvia Maria Matos
silvia.matos@fgv.br

1 Research Department, Central Bank of Brazil, Rio de Janeiro, Brazil

2 Graduate School of Economics, Getulio Vargas Foundation, Praia de Botafogo 190, s.1104,
Rio de Janeiro, RJ 22.253-900, Brazil
3 Instituto Brasileiro de Economia – IBRE, Getulio Vargas Foundation, Rio de Janeiro, Brazil

123
138 W. P. Gaglianone et al.

We first show that these forecasts are a bias-ridden version of the conditional expec-
tation of inflation using the no-bias tests proposed in Issler and Lima (J Econom
152(2):153–164, 2009) and Gaglianone and Issler (Microfounded forecasting, 2015).
The results reveal interesting data features: Consumers systematically overestimate
inflation (by 2.01 p.p., on average), whereas market agents underestimate it (by 0.68
p.p. over the same sample). Next, we employ a pseudo out-of-sample analysis to eval-
uate different forecasting methods: the AR(1) model, the Granger and Ramanathan
(J Forecast 3:197–204, 1984) forecast combination (GR) technique, a Phillips-curve
based method, the Capistrán and Timmermann (J Bus Econ Stat 27:428–440, 2009)
combination method, the consensus forecast (AF), the bias-corrected average forecast
(BCAF), and the extended BCAF. Results reveal that: (i) the MSE of the AR(1) model
is higher compared to the GR (and usually lower compared to the AF); and (ii) the
extended BCAF is more accurate than the BCAF, which, in turn, dominates the AF.
This validates the view that the bias corrections are a useful device for forecasting
using surveys. The Phillips-curve based method has a median performance in terms of
MSE, and the Capistrán and Timmermann (2009) combination method fares slightly
worse.

Keywords Consensus forecasts · Forecast combination · Common features ·

Panel data

JEL Classification C14 · C33 · E37

1 Introduction

From a theoretical and empirical point-of-view, Bates and Granger (1969) made the
econometric profession aware of the benefits of forecast combination. Their empirical
results were later confirmed by a variety of time-series studies—e.g., Granger and
Ramanathan (1984), Palm and Zellner (1992), Stock and Watson (2002a, b, 2006),
Timmermann (2006), and Genre et al. (2013)—where the simple average of forecasts
is shown to perform relatively well.
In a pioneering study on forecasting using panel-data techniques, Davies and Lahiri
(1995) proposed testing forecast rationality using information across forecasts and
forecast horizons by employing a three-way decomposition.1 This lead to a subsequent
literature on what could be labeled a panel-data approach to forecasting; see Baltagi
(2013) for a broader discussion on this topic. Some of these studies focused on forecast
combination and others on density-forecast estimation, e.g., Issler and Lima (2009),
Gaglianone et al. (2011), Gaglianone and Lima (2012, 2014), Gaglianone and Issler
(2015), and Lahiri et al. (2015).
In Issler and Lima and Gaglianone and Issler, the main idea is that the consensus
forecast (the cross-sectional average of individual forecasts) is a bias-ridden version

1 Prior to that, Palm and Zellner (1992) also employ a two-way decomposition to discuss forecast com-
bination in a Bayesian and a non-Bayesian setup. But, the main focus of their paper is not on panel-data
techniques, but on the usefulness of a Bayesian approach to forecasting.

123
Applying a microfounded-forecasting approach to predict… 139

of the conditional expectation, which is the optimal forecast under a mean-squared-

error (MSE) risk function. Once these potential biases are properly removed, one can
construct forecasting devices that are optimal in the limit, i.e., that can approach the
conditional expectation asymptotically.
In this paper, we show empirically that combining forecasts in surveys is a promising
empirical strategy to increase forecast accuracy. Our focus is on Brazilian inflation, a
key variable on the Inflation Targeting Regime , implemented by the Central Bank of
Brazil (BCB). In Brazil, there are two main surveys of expectations regarding inflation.
The first is the Focus Survey—a survey of professional forecasters—kept by the BCB.
The second is the FGV–IBRE Economic Tendency Survey—a survey of individual
consumers. Both include forecasts for the official index used in the Inflation Targeting
Regime—the IPCA, a Broad Consumer Price Index (CPI). Using information in these
surveys, we employ the techniques in Issler and Lima and in Gaglianone and Issler to
construct potentially optimal forecasts under MSE risk. The method discussed in the
former shows that the mean of the consensus forecast is biased. The method discussed
in the latter is microfounded and shows that the optimal forecast decision faced by an
individual agent is related to the conditional expectation by an affine structure. This
leads to the fact that the consensus forecast has two sources of bias—an intercept
bias and a slope bias. Fortunately, they can be estimated to filter out the conditional
expectation from a database of survey responses.
Using bias correction for the consensus forecast is a relevant empirical issue, since
there is a growing forecasting literature in macroeconomics where the consensus fore-
cast is used with no correction and rationality tests employing it usually find a result
opposite to rationality; see, e.g., Coibion and Gorodnichenko (2012). Below, we show
some evidence of these phenomena and discuss some important implications regarding
rationality tests.
The usefulness of survey data for forecasting has been recently shown in a vari-
ety of studies. For example, Ang et al. (2007) find that true out-of-sample survey
forecasts (e.g., Michigan; Livingston) outperform a large number of out-of-sample
single-equation and multivariate time-series competitors. Faust and Wright (2013)
argue that subjective forecasts of inflation based on surveys seem to outperform model-
based forecasts in certain dimensions, often by a wide margin. In a world where there
is an increasing availability of reliable survey data provided electronically, it is inter-
esting to examine how one could efficiently use it. This is exactly the objective of this
paper, where we exploit the fact that the Focus Survey has information on more than
250 professional forecasters and that the Economic Tendency Survey keeps responses
of about 1500 respondents on inflation expectations in every month.
The rest of the paper is divided as follows. Section 2 discusses the econometric
techniques employed in this paper, with some focus on the work of Issler and Lima
(2009) and of Gaglianone and Issler (2015), but covering the adjacent literature as well.
Section 3 presents a real-time forecasting exercise with data from the FGV survey of
consumer expectations on inflation and from the Focus survey of the BCB on inflation
expectations of professional forecasters. It also discusses rationality tests using the
consensus forecasts and its possible shortcomings. Section 4 concludes.

123
140 W. P. Gaglianone et al.

2 Econometric methodology

Forecast combination has proved to be a valuable tool at least since Bates and Granger
(1969). Palm and Zellner (1992) and Davies and Lahiri (1995) pioneered the use of
panel-data techniques in forecasting. This section discusses in some detail the forecast-
combination approach put forth by Gaglianone and Issler (2015) on how to combine
survey expectations to obtain optimal forecasts in a panel-data context. Some parts of
the material therein can also be found in Issler and Lima (2009) and in Lahiri et al.
(2015).
The techniques discussed in this section are appropriate for forecasting a weakly
stationary and ergodic univariate process {yt } using a large number of forecasts, coming
from forecast surveys (expectations) on the variable in question – yt . Some (or all) of
these responses can be generated by using econometric models, but then we have no
knowledge of them. We label individual forecasts of yt , computed using information
h , i = 1, 2, . . . , N , h = 1, 2, . . . , H , and t = 1, 2, . . . , T .
sets lagged h periods, by f i,t
Therefore, f i,t are h-step-ahead forecasts of yt , formed at period t − h, and N is the
h

number of respondents of this opinion poll regarding yt . Gaglianone and Issler show
that, in a variety of interesting cases, optimal forecasts are related to Et−h (yt )—the
conditional expectation of yt , computed using information lagged h periods—by an
affine function:

h
f i,t = kih + βih Et−h (yt ) + εi,t
h
. (1)

This result is somewhat expected given the pioneering work of Granger (1969). He
shows that, under a mean-squared-error (MSE) risk function and proper regularity
conditions, the optimal forecast is equal to Et−h (yt ). Moreover, if the risk function is
asymmetric and proper regularity conditions are met, then Granger showed that the
optimal forecast is equal to Et−h (yt ) + kih ; see also the later developments in Christof-
fersen and Diebold (1997), Elliott and Timmermann (2004), Patton and Timmermann
(2007), and Elliott et al. (2008), for example.
Gaglianone and Issler consider a setup with two layers of decision making. In the 
first layer, individuals (survey respondents) form their optimal point forecasts f i,t h

of a random variable yt by using a specific loss function. They allow for the existence
of asymmetry of the loss function and different assumptions about knowledge of the
DGP of yt . The optimal forecasts f i,t h will be available as survey results, where the

number of respondents is potentially large, i.e., N → ∞, and these surveys can be

periodically taken on a large number of different occasions, i.e., T → ∞. In the
second layer of decision making, an econometrician will be the final user of this large
number of forecasts, operating under an MSE risk function. Her/his challenge is to
uncover Et−h (yt )—the optimal forecast of the second layer of decision making.
Gaglianone and Issler employ the basic assumptions in the econometric forecast
literature to derive a consistent estimate of Et−h (yt ), which can be obtained by using
Hansen’s (1982) generalized method of moments (GMM). Start by taking a cross-
sectional average of (1), after using:

123
Applying a microfounded-forecasting approach to predict… 141

yt = Et−h (yt ) − ηth , (2)

noting that ηth is a martingale-difference sequence by construction, i.e., Et−h (ηth ) = 0.

Under suitable assumptions, they show that the following system of equations can be
used to estimate the key parameters k h and β h in:
  
E f ·,th − k h − β h yt ⊗ z t−s = 0, h = 1, 2, . . . , H, (3)

N
where z t−s is a vector of instruments, dated t − s or older, s ≥ h, f ·,th = N1 i=1 h,
f i,t
 N  N
k h = N1 i=1 kih and β h = N1 i=1 βih , for each h. Over-identification requires that
dim(z t−s ) > 2.
Gaglianone and Issler argue that there is no need to estimate individual coefficients
kih and βih —only their means to be able to identify and estimate Et−h (yt ). In doing
so, they avoid the curse of dimensionality given that N → ∞. As long as these cross-
sectional averages converge, GMM using time-series restrictions delivers consistent
estimates of the respective parameter means.
Current surveys, however, usually approximate better the case where T → ∞,
while N is small or diverges at a smaller rate than T . Under additional conditions on the
h = 1  N f h , k h = 1  N k h and β h = 1  N β h ,
cross-sectional averages f ·,t N i=1 i,t N i=1 i N i=1 i
Gaglianone and Issler show that one can still estimate Et−h (yt ) consistently, when
T → ∞ first, and then N → ∞.
One way to exploit all possible moment conditions implicit in (3) is to stack all the
restrictions across h (finite) as:
⎡⎛ ⎞ ⎤
1 − k1 − β1 y
f ·,t t
⎢⎜ 2 ⎟ ⎥
⎢⎜ f ·,t − k 2 − β 2 yt ⎟ ⎥
E⎢
⎢⎜
⎜ .. ⎟
⎟ ⊗ z ⎥
t−s ⎥ = 0, (4)
⎣⎝ . ⎠ ⎦
f ·,t − k − β yt
H H H

where the problem collapses to one where we have H ×dim(z t−s ) restrictions and 2H
parameters to estimate. As before, over-identification requires that dim(z t−s ) > 2.
Given a choice of H , GMM estimation of (4) is efficient.2 A less efficient alternative
to estimate the whole stacked system (4) is to estimate separately (3) for every horizon

2 One could impose additional restrictions across coefficients k h and β h using the term structure of
forecasts. Given stationarity of yt : yt = εt + ψ1 εt−1 + · · · + ψh−1 εt−h+1 + ψh εt−h + · · · , which
gives: Et−(h−1) (yt ) = Et−h (yt ) + ψh−1 εt−(h−1) , h = 1, 2, . . . , H, showing that the expectation revi-
sion Et−(h−1) (yt ) − Et−h (yt ) is unforecastable using information up to t − h. In a GMM context, this
 N fi,th −k h
implies additional moment restrictions across horizons if we substitute Et−h (yt ) by N1 i=1 and
βh
 N fi,t
h−1
−k h−1
Et−(h−1) (yt ) by N1 i=1 . Notice that, for large enough N , we can approximate reasonably
β
h−1
well these conditional expectations using these averages, as proposed in Gaglianone–Issler. They did not
consider these additional restrictions into GMM estimation, but that could certainly be exploited in the
future.

123
142 W. P. Gaglianone et al.

h, which could also be attempted for computational reasons. As is usual within GMM
estimation, one can perform a standard misspecification test of correlation between
errors and instruments by employing the J-test proposed by Hansen (1982).
 
Using GMM estimates, k h and β h , obtained when T → ∞, Gaglianone and Issler
show the critical result that:


1  f i,t − k h p
N h
−→ Et−h (yt ) , (5)
N 
i=1 βh

in a variety of different setups: when T → ∞ first, and then N → ∞; when N → ∞

first, and then T → ∞; and when N , T → ∞ in a joint limit framework, all using
the method in Phillips and Moon (1999). The conditions under which (5) holds vary
according to the asymptotic setup. N
Equation (5) offers a way to filter the consensus forecast N1 i=1 h extracting
f i,t
Et−h (yt ) in the limit. Thus, it delivers an estimate of the optimal forecast in the second
layer of decision making. It can be viewed immediately as a forecast-combination
method.
A paper that delivers a similar result, although suggests a different estimation
method, is Capistrán and Timmermann (2009). Start with Eq. (1) and take the cross-
sectional average to obtain3 :

1  h 1  h 
N N
f i,t = ki + βih Et−h (yt ) + εi,t
h
, or, in standard notation,
N N
i=1 i=1
h
f ·,t = k h + β h Et−h (yt ) + ε·,t
h (6)

Combine (6) with (2) to obtain, for large enough N :

kh 1
yt = − + f ·,th − ηth . (7)
βh βh

The approach in Capistrán and Timmermann yields a single equation where k h and
β h can be estimated to filter Et−h (yt ) in a similar fashion to that in (5). As can be
seen by comparing (7) and (3), these two equations identify k h and β h . A potential
advantage for the method in Gaglianone and Issler using (4) is that it allows stacking
moment conditions across horizons and joint estimation of the bias terms. Thus, it
benefits from the efficiency gains associated with employing a multi-equation setup
as opposed to the single-equation setup in Capistrán and Timmermann. In the empirical
section, we shall compare directly these two methods regarding their out-of-sample
mean-squared error.

3 We thank an anonymous referee for pointing this out to us.

123
Applying a microfounded-forecasting approach to predict… 143

In an out-of-sample forecasting exercise involving surveys, there are usually three

consecutive distinct time sub-periods, where time is indexed by t = 1, 2, . . . , T1 , . . . ,
T2 , . . . , T . The first sub-period E is labeled as the “estimation sample,” where models
are usually fitted to forecast yt in the subsequent period, if that is the case. The
number of observations in it is E = T1 = κ1 · T , comprising (t = 1, 2, . . . , T1 ).
For the other two, we follow the standard notation in West (1996). The sub-period R
(for regression) is labeled as the post-model-estimation or “training sample,” where
realizations of yt are usually confronted with forecasts produced in the estimation
sample, and weights and bias-correction terms are estimated, if that is the case. It has
R = T2 − T1 = κ2 · T observations in it, comprising (t = T1 + 1, . . . , T2 ). The final
sub-period is P (for prediction), where genuine out-of-sample forecast is entertained.
It has P = T − T2 = κ3 · T observations in it, comprising (t = T2 + 1, . . . , T ).
Notice that 0 < κ1 , κ2 , κ3 < 1 , κ1 + κ2 + κ3 = 1, and that the number of observations
in these three sub-periods keep a fixed proportion with T —respectively, κ1 , κ2 and
κ3 —being all O (T ).
It is worth mentioning that the method proposed by Gaglianone and Issler extends
the previous literature of forecasting within a panel-data framework, e.g., Palm and
Zellner (1992), Davies and Lahiri (1995), Issler and Lima (2009), and Lahiri et al.
(2015).
For example, their setup encompasses that of Davies and Lahiri (1995), reproduced
below with our notation:
 
yt − f i,t
h
= − kih + ηth + εi,t
h
, (8)

who imposed βih = 1 for all i = 1, . . . , N and kih = ki for all h = 1, . . . , H . Also,
it generalizes the results in Issler and Lima, where slopes are restricted as βih = 1 for
all i = 1, . . . , N . Here, we have two sources of bias correction: intercept and slope.
Notice that both arise from a structural affine function that links individual forecasts
to the conditional expectation. In itself, this provides a general framework that can be
used whenever a panel of forecasts is available.
The method proposed by Gaglianone and Issler is appropriate to cover any type
of survey where potentially the number of observations T is large, encompassing the
cases where the number of survey respondents and of time observations is large—big
data, and also the case where the number of time observations is large but the number
of respondents is fixed—standard continuous macroeconomic surveys.
The way Gaglianone and Issler identify the conditional expectation can be viewed
as a combination of cross-sectional averages with standard GMM moment restric-
tions, where the affine structure offers natural orthogonality restrictions allowing
the estimation of bias-correction terms. They circumvent the curse of dimension-
ality that arises from the factor structure (large N ) by employing these cross-sectional
averages.

123
144 W. P. Gaglianone et al.

3 Empirical application

3.1 Data

3.1.1 FGV’s consumer survey

The FGV–IBRE Economic Tendency Survey compiles business and consumers expec-
tations of key economic series in Brazil. The Brazilian Institute of Economics (IBRE)
is a pioneer in surveys, and this one runs since 1966. Since September, 2005, FGV–
IBRE conducts a monthly consumer survey, which consists of qualitative information
on household consumption, savings, financial variables, employment, etc. The survey
has a country-wide coverage (seven major state capitals) with approximately 2000
consumers. Survey respondents are classified into four classes of household monthly
income as follows4 : Income Level 1—Up to R$2100; Income Level 2—Between R$
2100.01 and R$ 4800.00; Income Level 3—Between R$4800.01 and R$ 9600.00;
Income Level 4—More than R$ 9600.01. Survey information can also be broken down
by different classes of education: Group 1—No education or incomplete first year;
Group 2—Complete first year or incomplete primary education; Group 3—Complete
primary education or incomplete secondary education; Group 4—Complete secondary
education or incomplete undergraduate; Group 5—Complete undergraduate; and
Group 6—Graduate studies.
The panel is unbalanced, and on average, each household is interviewed 7.81 times
(months) per year. We possess microdata at the individual level for this survey, begin-
ning in January 2006 through to May 2015 (T = 113 months). Overall, our sample
has N × T = 164,479 responses. Decomposing our sample into N and T gives the
following breakdown: T = 113 months, and an average of approximately N = 1456
individuals per month. The key question of interest for us is the following: “In your
opinion, how much will be Brazilian inflation over the next 12 months?” We excluded
outliers in the data set whenever the respondent answered that inflation would be
greater than 100 % in the next 12 months. In our sample, it was rare to observe 12-
month inflation greater than 10 %, so we considered this type of response completely
out of scope.5

3.1.2 Focus survey of the Central Bank of Brazil

The Focus Survey of forecasts of the Central Bank of Brazil (BCB) contains daily
(working days) forecasts from roughly 250 registered institutions since 1999, the
year when Brazil implemented its Inflation-Targeting Regime.6 About 100 of these
institutions are actively feeding the database with forecasts on any given day. Institu-
tions include professional forecasters—commercial banks, asset management firms,

4 See consumer survey: methodological features (April 2013), available at http://portalibre.fgv.br.

5 With this criterium, we excluded 349 observations of a total of 164,479.
6 The Focus survey is widely used in Brazil, and its excellence has been internationally recognized. In
2010, the survey received the Certificate of Innovation Statistics, due to a second place in the II Regional
Award for Innovation in Statistics in Latin America and Caribbean, offered by the World Bank.

123
Applying a microfounded-forecasting approach to predict… 145

consulting firms, non-financial firms or institutions, academics, etc. Participants can

provide forecasts for a large number of economic variables, e.g., inflation using dif-
ferent price indices, interest and exchange rates, GDP, industrial production, balance
of payments accounts, and fiscal variables and for different forecast horizons.
Our focus here is on Brazilian inflation, measured by the Brazilian Consumer Price
Index—IPCA, which is the official inflation target of the Brazilian Inflation-Targeting
Regime. Our sample covers daily inflation forecasts collected from January 2nd, 2004
until May, 28th, 2015 (2861 workdays).7 In each day t, t = 1, . . . , T , survey respon-
dent i, i = 1, . . . , N , may inform her/his expectations regarding inflation rates all
the way up to the next 18 months, as well as for the next twelve months and for the
next 5 years on a year-end basis. The data set regarding the twelve-month-ahead infla-
tion forecasts forms an unbalanced panel (N × T ) containing an amount of 244,043
observations. Decomposing our total number of observations into N and T gives the
following breakdown: T = 2861 daily observations and an average of N = 85.3 fore-
casters in our sample. For more information on this data set, see Carvalho and Minella
(2012) and Marques (2013).
Despite the fact that the Focus Survey has a longer time span and a higher frequency
than FGV’s Consumer Survey (working day vs. monthly), our sample size is, most of
the time, constrained by the span and frequency of the latter.

3.1.3 Consensus forecasts

Our target variable yt —is Brazilian inflation—as measured by the Broad Consumer
Price Index (IPCA) collected at the monthly frequency. The consumer forecasts f i,t h

from the FGV survey regarding expected inflation rate over the next 12 months (h = 12
months) are cross-sectionally averaged, forming the so-called consensus forecasts
h = 1  N f h . Besides the average over the entire data set (approximately 1500
f ·,t N i=1 i,t
households), labeled hereafter as “cons_all”, the individual forecasts are also aver-
aged by selected demographic groups, according to the education level of the survey
participant or to its level of income. The breakdown by education yields the following
consensuses: cons_educ_1_3; cons_educ_4 and cons_educ_5_6, respectively, educa-
tional levels groups 1 through 3, level 4, and group level 5 through 6. For income, we
breakdown the groups numbered from 1 to 4, respectively: cons_inc_1; cons_inc_2;
cons_inc_3; cons_inc_4. Figure 1 plots these eight time series of aggregate (cross-
sectional mean) forecasts.
There is considerable heterogeneity across groups, either from an education or
income point of view. Low-income households usually forecast higher inflation rates
vis-à-vis the forecasts of higher-income families. Obviously, inflation is not perceived
equally by all families, each with its own consumption bundle. Low-income house-
holds spend more on food, while higher-income households spend a larger proportion
of their budget on housing, education, and leisure. By construction, the IPCA index

7 Since the survey had a small cross-sectional coverage (small N ) in its first years, we only considered
post-2003 data.

123
146 W. P. Gaglianone et al.

14.50

12.50

10.50

8.50

6.50

4.50 Jul-08

Jul-13
Jun-06

Jun-11
Nov-06
Apr-07

Nov-11
Apr-12
Dec-08

Dec-13
Sep-07
Feb-08

May-09

Mar-10
Aug-10

Sep-12
Feb-13

May-14

Mar-15
Jan-06

Oct-09

Jan-11

Oct-14
cons_educ_1_3 cons_educ_4 cons_educ_5_6
cons_inc_1 cons_inc_2 cons_inc_3
cons_inc_4 cons_all
h from the FGV consumer survey (% 12-months-ahead)
Fig. 1 Consensus inflation forecasts f ·,t

11.00
10.00
9.00
8.00
7.00
6.00
5.00
4.00
3.00
2.00
Oct-09

Oct-14
Jun-06

Apr-07

Dec-08

Mar-10

Jun-11

Apr-12

Dec-13

Mar-15
Nov-06

Jul-08

May-09

Nov-11

Jul-13

May-14
Jan-06

Jan-11
Sep-07
Feb-08

Sep-12
Feb-13
Aug-10

cons_all focus_day10 IPCA (12-months-ahead)

h with the actual inflation rate y
Fig. 2 Comparison of the consumer and market inflation forecasts f ·,t t
(% 12 months)

follows closely the cost-of-living of households with income well above the income
level of our categories 1 and 2.
Figure 2 compares the consensuses of consumer forecasts (cons_all), professional
(or market forecasts from the Focus Survey), and the twelve-month-ahead IPCA
inflation rate. The latter is exactly what these forecasts are supposed to track. Mar-
ket forecasts are collected at the 10th and 15th day of each month (focus_day10,
focus_day15). This makes the focus consensus consistent with the days in which the

123
Applying a microfounded-forecasting approach to predict… 147

data are collected for the FGV consumer survey. On average, the consumer consensus
has a positive bias, while the consensus of professional forecasters has a negative bias.
It is also worth noting that the behavior of both consensuses is very different at the end
of the sample period. While the consumer consensus follows the observed increase in
inflation, the consensus of professional forecasters gets it completely wrong, predict-
ing a decrease in inflation.

3.2 Empirical results

3.2.1 Descriptive statistics

Table 1 shows the pairwise sample correlations of the consensuses forecasts. Notice
that all series are highly correlated. The consensus forecasts of consumers and of
market agents show a sample correlation of 0.80.
Table 2 presents the sample mean and the respective estimation of the asymptotic
standard error based on the number of sample observations T = 113. Next, we inves-
tigate whether or not the consensuses forecasts Granger Cause future inflation using
Sims’ test. More than half of the consensuses forecasts for consumers Granger Cause
future inflation, i.e., are leading indicators of future inflation. This is also true for
the consensus including all consumers. The same is true for both the consensuses of
professional forecasters. Therefore, they will be useful as tools in forecasting future
inflation.

3.2.2 Bias-correction devices

In Fig. 2, we provided soft evidence that the consensuses forecasts are biased as
predictors of future inflation. Next, we formally test for zero bias employing the bias-
corrected average forecast (BCAF) approach of Issler and Lima (2009). There, the
estimate of the out-of-sample additive forecast bias k h is:

N  
1  
T2

kh = yt − f i,t
h
, (9)
NR
t=T1 +1 i=1

where its robust standard error is computed taking into account possible spatial and
time dependence. Here, we employ the whole sample in estimating k h : t = 1, 2, . . . , T ,
where the consensus data runs from January 2006 to July 2014, whereas the IPCA
sample runs from December 2006 to June 2015.
Results of the zero-bias test are presented in Table 3. As expected from the plot in
Fig. 2, all consumer forecasts showed a negative bias, i.e., consumers overestimated
future inflation, while all professional forecasts underestimated future inflation. The
formal evidence provided below calls for the use of bias-correction devices in con-
structing optimal forecasts.
A possible shortcoming of the test in Table 3 is that it is implicitly assumed that
β = 1. However, Gaglianone and Issler (2015) showed that, in general, there is
h

123
 148

123
Table 1 Sample correlations

Variable cons_all cons_ed13 cons_ed4 cons_ed56 cons_inc1 cons_inc2 cons_inc3 cons_inc4 focus_d10 focus_d15

cons_all 1.000
cons_ed13 0.768 1.000
cons_ed4 0.950 0.694 1.000
cons_ed56 0.986 0.702 0.904 1.000
cons_inc1 0.874 0.808 0.900 0.822 1.000
cons_inc2 0.945 0.708 0.911 0.930 0.785 1.000
cons_inc3 0.957 0.667 0.909 0.965 0.802 0.892 1.000
cons_inc4 0.927 0.662 0.832 0.950 0.785 0.812 0.890 1.000
focus_d10 0.803 0.543 0.808 0.816 0.779 0.771 0.867 0.753 1.000
focus_d15 0.803 0.542 0.806 0.819 0.776 0.768 0.869 0.758 0.999 1.000

Full sample: January 2006 to May 2015 (T = 113 observations)

W. P. Gaglianone et al.
Applying a microfounded-forecasting approach to predict… 149

Table 2 Mean forecast, SE, and Granger Causality test

Variable Mean SE GC test

cons_all 7.78 0.09 0.0296**

cons_educ_1_3 9.21 0.11 0.5110
cons_educ_4 8.23 0.10 0.0219**
cons_educ_5_6 7.38 0.09 0.0423**
cons_inc_1 9.32 0.15 0.0175**
cons_inc_2 8.01 0.10 0.1171
cons_inc_3 7.41 0.10 0.1257
cons_inc_4 7.10 0.08 0.2206
focus_day10 5.05 0.08 0.0306**
focus_day15 5.06 0.08 0.0211**

Full sample: January 2006 to May 2015 (T = 113 observations) √

Confidence interval based on asymptotic values: SE = sample SD/ T
The last column shows the p values of the Granger Causality test (12 lags) based on the null hypothesis:
Forecast does not GC IPCA
** Rejection at a 5 % significance level

Table 3 BCAF additive bias k h test (Issler and Lima 2009)

Variable Mean Bias Robust SE Zero-bias test (p value)

cons_all 7.62 −2.01 0.26 0.000

cons_educ_1_3 9.01 −3.41 0.27 0.000
cons_educ_4 8.06 −2.45 0.26 0.000
cons_educ_5_6 7.21 −1.61 0.26 0.000
cons_inc_1 9.02 −3.41 0.26 0.000
cons_inc_2 7.83 −2.22 0.28 0.000
cons_inc_3 7.24 −1.64 0.26 0.000
cons_inc_4 6.93 −1.33 0.25 0.000
focus_day10 4.92 0.68 0.21 0.001
focus_day15 4.93 0.68 0.21 0.001

Consensus sample: January 2006 to July 2014. IPCA sample: December 2006 to June 2015
(103 observations). Average IPCA (in 12 months) = 5.61 %. Unconditional (additive) bias is defined as:
IPCA—mean (forecast). The zero-bias test is based on the null hypothesis H0 : k h = 0
Robust SE employs Newey–West. See Issler and Lima (2009)

an intercept and a slope bias term for the consensus forecast, suggesting the use of
the Extended BCAF setup, where the parameters k h and β h are estimated by the
generalized method of moments (GMM) based on yt and f i,t h .8

Results of GMM estimation and of the no-bias test are presented in Table 4. We
use a set of instruments z t−s containing two lags of the consensus forecasts f ·,th , up

8 The “iterative” procedure of Hansen et al. (1996) is employed in the GMM estimation, and the initial
weight matrix is the identity.

123
 150

123
Table 4 Extended BCAF estimation and No Bias test
 
Variable kh Robust t stat p βh Robust t stat p OIR test p No Bias
SE value SE value value test p value

cons_all −0.11 2.59 0.97 1.34 0.46 0.00 0.115 0.000

cons_educ_1_3 2.77 3.40 0.41 1.08 0.59 0.07 0.207 0.000
cons_educ_4 0.65 2.12 0.76 1.30 0.37 0.00 0.103 0.000
cons_educ_5_6 −1.64 2.77 0.55 1.54 0.49 0.00 0.197 0.000
cons_inc_1 0.33 1.54 0.83 1.54 0.27 0.00 0.128 0.000
cons_inc_2 0.03 2.60 0.99 1.37 0.46 0.00 0.060 0.000
cons_inc_3 −2.12 2.65 0.42 1.64 0.47 0.00 0.170 0.000
cons_inc_4 −1.47 2.75 0.59 1.44 0.49 0.00 0.268 0.000
focus_day10 −2.28 1.29 0.08 1.26 0.22 0.00 0.704 0.000
focus_day15 −2.24 1.26 0.08 1.25 0.22 0.00 0.702 0.000

Consensus sample: January 2006 to July 2014. IPCA sample: December 2006 to June 2015 (103 observations). Robust standard errors (SE) from a GMM estimation. The
column “OIR test” denotes the p values from the Over-Identifying restriction (OIR) J test due to Hansen (1982)
The column “No-Bias test” shows the p values based on the null hypothesis H0 : [k h = 0; β h = 1]
W. P. Gaglianone et al.
Applying a microfounded-forecasting approach to predict… 151

Table 5 Forecast combination (Granger and Ramanathan 1984) IPCAt+12 = α + β 1 cons_allt + β 2

focus_day10t + ε t+12

Parameter Unrestricted regression Restricted regression

Full sample Last 5 years Full sample Last 5 years


α 4.723** 1.032 2.578*** 0.646
(2.192) (2.627) (0.851) (0.769)
β1 −0.865** 0.064 −0.703** 0.084
(0.378) (0.396) (0.326) (0.316)
β2 1.518*** 0.873*** − −
(0.336) (0.277)
Adjusted R 2 0.385 0.279 0.348 0.289

Consensus sample: January 2006 to July 2014. IPCA sample: December 2006 to June 2015
Robust (HAC) standard errors in parentheses. Restricted regression imposes β1 +β 2 = 1
*, **, and *** indicate, respectively, significance at 10, 5, and 1 % level

to four lags of the output gap9 and two lags of the Commodity Price Index (IC-Br),
which weighting structure is designed to measure the impact of commodity prices on
Brazilian consumer inflation10 ; see Appendix for results using an alternative set of
instruments. Hansen’s Over-Identifying Restriction (OIR) test is employed in order
to check the validity of GMM estimates. At the usual significance levels, we do not
observe any rejection for the OIR test. The no-bias test consists of a joint test of the
null hypothesis H0 : [k h = 0; β h = 1]. We reject the null of no-bias on all occasions,
showing that the consensus forecasts are a bias-ridden version of the conditional
expectation Et−h (yt ).
From Eq. (5), note that, under the null H0 : [k h = 0; β h = 1], the aggregate forecast
h should converge in probability to the conditional expectation E
f ·,t t−h (yt ), with no
need for bias correction. However, if the null is rejected, we should expect a corrected
version of the consensus forecast to produce superior forecasts vis- à-vis the consensus
forecasts. These are exactly our findings below.

3.2.3 Forecast combination of consumer and market expectations

Next, we apply the Granger and Ramanathan (1984) forecast combination technique
for the entire sample. It consists of a standard OLS regression (including an intercept)
of IPCA inflation on the two consensus forecasts—consumer and market, respectively:
cons_all and focus_day10. We run those regressions in two different ways: When the
two slope coefficients are unrestricted and when they are restricted to add up to unity.
Also, we cover two different sample periods: the whole sample and only the last five
years. Results are presented in Table 5. Notice that, for the last 5 years, the consumer
consensus is not significant, which does not happen for the whole sample.

9 The Hodrick-Prescott (HP) filtered output gap is computed from the log real monthly GDP (IBRE/FGV).
10 The adopted set of instruments is slightly modified for the lower education groups (cons_educ_1_3) and
the market forecasts (focus_day10; focus_day15).

123
152 W. P. Gaglianone et al.

The estimated slopes can be interpreted as optimal weights under a MSE loss
function. The combined (consumer-market) forecast is given by:

h
 α + β1 cons_allt + β2 focus_day10t .
f cm t = 

3.2.4 Out-of-sample forecasting exercise

In this section, we conduct a truly out-of-sample forecasting exercise, incorporating

what we have learned from the previous sections. We compare different forecast meth-
ods by computing their out-of-sample mean-squared error (MSE) of forecasting. Our
whole sample for the experiment runs from January 2006 through July 2014 for the
forecasts and from February, 2006 through June 2015 for the realizations of inflation,
comprising 103 and 113 observations, respectively. We leave the last 60 observations
(5 years, comprising July 2010 through June 2015) for a pseudo out-of-sample evalua-
tion of several forecasting strategies (or methods), which are estimated with a growing
window. Thus, model parameters are estimated recursively as we incorporate every
new time-series observation, one at a time. These forecasting methods include: a “best”
model, chosen by information criteria, a Phillips-curve based forecast, the Granger
and Ramanathan (1984) forecast-combination technique, the Capistrán and Timmer-
mann (2009) consensus regression, the average forecast (or consensus forecast), the
bias-corrected average forecast (BCAF), proposed by Issler and Lima (2009), and the
extended BCAF, proposed by Gaglianone and Issler (2015). For the last four strategies,
we also present the MSEs by category of income and education.
The best model using information criteria is the AR(1) model for the monthly
inflation rate, see Cordeiro et al. (2015) and Gaglianone et al. (2016). The AR(1)
model for the monthly inflation rate yt is as follows:

yt = c + βyt−1 + εt . (10)
 
T 
cT ; β
From a sample with t = 1, . . . , T observations, the respective estimates 
are computed. The respective h-step-ahead forecast (h > 1) is given by

 h 
h
 i−1
y
T +h = 
β T y T + cT β
 T , (11)
i=1

and the cumulative twelve-month rate ahead yT12m_rate

+12 , in percentage is:
   
 12 y
T +h
yT12m_rate
+12 = 100  1+ −1 . (12)
h=1 100

We also estimate a Phillips-curve model (PC), which has a long tradition in forecast-
ing inflation (for a comprehensive survey, see Stock and Watson 2009). The Brazilian
Central Bank (BCB) has had several years of experience fitting different versions of the
Phillips curve, and we benefit from that in our forecasting exercise. Here, we consider

123
Applying a microfounded-forecasting approach to predict… 153

a hybrid (New Keynesian) version of the Phillips curve, which includes backward and
forward looking terms, lagged imported inflation (to consider a pass-through channel),
and a lagged output gap. In the BCB experience, what works best for forecasting IPCA
inflation is to disaggregate inflation into two main components: market prices (roughly
75 % of IPCA) and the so-called regulated and monitored prices (approximately 25 %
of IPCA), which contains consumer items that are relatively insensitive to domestic
demand and supply conditions or that are in some way regulated by a public agency.
Since these two groups have completely different time dynamics, they are modeled
separately. We first estimate a Phillips curve for inflation of the market prices basket,
as follows:
exp imp
πtmarket = α0 + α1 πt−1
market
+ α2 πt|t−1 + α3 πt−4 + α4 gt−1 + εt , (13)

exp
where πtmarket is the monthly inflation of market prices, πt|t−1 denotes the one-month
imp
ahead expected inflation (from the Focus survey), πt is what has been labeled as
imported inflation (R$/US$ exchange rate variation plus the U.S. PPI), and gt is the
output gap.11 We also impose the coefficient restriction: α1 +α2 +α3 = 1, to guarantee
a vertical long-run Phillips curve. For the so-called monitored prices, we estimate an
auxiliary AR model. The usual criteria for lag order selection indicate one lag only (and
the estimated autoregressive coefficient is 0.53). Then, we use each model to produce
h-step ahead point forecasts (h = 1, . . . , 12 months) which are, then, aggregated
through time to form a 12-month cumulated forecast for each component.12 Both
forecasts are then aggregated (using corresponding CPI weights) in order to generate
an aggregated 12-month ahead point forecast for the whole IPCA inflation.
Table 6 presents the comparison of the out-of-sample MSE of the different forecast
strategies listed above. First, the MSE of the AR(1) model is higher when com-
pared to that of the Granger and Ramanathan combination of the consensuses of
consumers and of professional forecasters, and it is lower than that of the average
forecasts—exceptions are the consensus for the highly educated forecasters, higher
income forecasters, and professional forecasters (Focus Survey). Second, when we
compare the MSE of the average forecast, BCAF and Extended BCAF, there is a
clear pecking order: The extended BCAF is more accurate than the BCAF, which, in
turn, dominates the average forecast. This validates the view that the bias corrections
performed either by the BCAF, or by the extended BCAF, are a useful device for fore-
casting using surveys. Third, when we compare the best forecasts for all consumers
(Extended BCAF, cons_all) with the best forecasts for all professional forecasters
(Extended BCAF, focus_day15), we observe a reduction in MSE in favor of profes-

11 The output gap is based on the seasonally adjusted IBC-BR index of economic activity. The Hodrick–
Prescott (HP) filter is employed to generate the output gap in a recursive estimation scheme, that is, we
re-construct the entire output gap series for each new observation added to the estimation sample along the
out-of-sample exercise (and then re-estimate the Phillips curve to construct new h-step ahead forecasts).
This way, we perform a “pseudo” real-time PC-forecast (based on the last available vintage of the IBC-BR).
12 The h-step ahead forecasts are constructed using an iterated procedure, which is trivial in the case of
the AR model. In the case of the PC and h >1, we use the previous PC-forecast (from h − 1), random
walk forecasts for the expected and imported inflation rates and the AR(1) forecast for the output gap
(autoregressive coefficient of 0.95, based on in-sample estimations).

123
154 W. P. Gaglianone et al.

sional forecasters of 22.5 %, which is sizable. This happens despite the fact that, for
consumers, we are averaging the forecast of approximately 1500 individuals, whereas
for professional forecasters, we employ about 85 individuals. This may be a sign that it
matters to employ informed and well-trained professionals in forecasting inflation—a
relief for the profession as a whole.
Forecasts using the Phillips curve ranked relatively well here—a median result—
better than the AR(1) model, the average forecast, and the Capistrán and Timmermann
method, but worse than the Granger–Ramanathan combination, the BCAF, and the
extended BCAF. The Capistrán and Timmermann (2009) consensus regression did
slightly worse than the Phillips curve. A direct comparison of Capistrán and Timmer-
mann’s method with the extended BCAF is interesting here, since, in principle, both
could potentially identify the intercept and slope biases present in the consensus fore-
cast, respectively, denoted by k h and β h in Sect. 2. Indeed, the MSEs of the Extended
BCAF and that of the Capistrán and Timmermann method are very different across
all levels of education and income for consumers, as well as for all days in the Focus
Survey. When we consider all consumers, the MSE of the extended BCAF is 0.42 of
that of Capistrán and Timmermann’s method, a sizable reduction.
In order to check whether the differences in MSE listed in Table 6 are significant, we
employ the equal predictive accuracy test of Clark and West (2007) for nested models,
and the equal predictive ability test of Diebold and Mariano (1995) for non-nested
models. All MSE comparisons are direct between the MSE of the extended BCAF and
that of any specific cell in Table 6. In most cases, the extended BCAF produces an MSE
that is statistically different (and smaller) than that of the direct competitor. Details are
as follows: The average forecast and the BCAF have an inferior out-of-sample MSE
vis-à-vis the extended BCAF forecast. Comparisons of the extended BCAF with the
AR(1) model using a Diebold–Mariano test for equal variances shows that the former
is statistically superior to the latter (at the 10 % significance level) in almost all cases.
The Extended BCAF for the market forecasts also statistically improves (at a 10 %
level) the out-of-sample accuracy compared to the Granger and Ramanathan (1984)
forecast combination approach. Also, it statistically improves on the Capistrán and
Timmermann (2009) consensus-forecasting approach almost everywhere.
All in all, if we had to suggest a single forecasting strategy for Brazilian inflation
one-year ahead, we would suggest the use of extended BCAF based on the Focus
Survey consensus (15th working day).

3.2.5 Rationality tests

In testing for rationality, we employed two alternative approaches. The first is based
on the well-known Mincer–Zarnowitz regression. It is only valid if one assumes
that the forecaster minimizes the MSE risk function. For other functions, the test
no longer applies,13 and one must find alternative ways of testing rationality. Here,

13 For example, an asymmetric loss function (e.g., lin-lin) could reflect the forecaster’s different weighting
scheme of positive forecast errors vis-à-vis the negative ones. In this setup, a non-trivial forecast bias
may arise, for instance, reflecting the asymmetric loss function (instead of lack of rationality). For further

123
Applying a microfounded-forecasting approach to predict… 155

Table 6 Forecast evaluation: mean-squared error (MSE) out of sample

Variable AR(1) PC GR CT Average BCAF Extended

forecast BCAF

cons_all 2.344* 1.240 0.921 1.795* 3.406* 1.026 0.760

[0.070] [0.274] [0.311] (0.000) (0.000) (0.129)
cons_educ_1_3 2.344* 1.240 0.921 1.757* 10.148* 1.927* 1.417
[0.074] [0.638] [0.892] (0.000) (0.000) (0.017)
cons_educ_4 2.344* 1.240 0.921 1.789* 5.658* 0.863* 0.745
[0.056] [0.232] [0.257] (0.000) (0.000) (0.092)
cons_educ_5_6 2.344* 1.240 0.921 1.689* 2.287* 1.047 0.758
[0.066] [0.264] [0.315] (0.000) (0.008) (0.196)
cons_inc_1 2.344 1.240 0.921 1.568* 11.845* 1.149* 0.740
[0.128] [0.200] [0.345] (0.000) (0.000) (0.001)
cons_inc_2 2.344* 1.240 0.921 1.868* 4.650* 0.939 0.783
[0.047] [0.255] [0.354] (0.000) (0.000) (0.103)
cons_inc_3 2.344* 1.240 0.921 1.539* 2.784* 0.878 0.768
[0.059] [0.288] [0.308] (0.000) (0.001) (0.251)
cons_inc_4 2.344* 1.240 0.921 1.618* 1.523* 1.308 0.787
[0.086] [0.284] [0.344] (0.000) (0.050) (0.174)
focus_day10 2.344* 1.240 0.921* 1.104 1.295 0.624 0.590
[0.050] [0.209] [0.098] (0.248) (0.289) (0.454)
focus_day15 2.344* 1.240 0.921* 1.075 1.282 0.624 0.589
[0.052] [0.205] [0.099] (0.247) (0.296) (0.443)

Forecast evaluation sample = last 60 observations—July 2010 through June 2015. PC denotes the Phillips
curve forecast. GR means the combined forecast of Granger and Ramanathan (1984), based on an OLS
regression with intercept and “cons_all” and “focus_day10” forecasts as regressors. CT denotes the forecast
from the OLS regression of Capistrán and Timmermann (2009) of inflation on the consensus and an intercept.
The second, third, and fourth columns show [in brackets] the p values of the equal predictive ability test of
Diebold–Mariano (1995) between the AR(1), PC, and GR, respectively, compared to the Extended BCAF
(on each row). The fifth, sixth, and seventh columns show (in parenthesis) the p values of the equal predictive
accuracy test of Clark and West (2007), which compares the CT forecast, the average forecast and the BCAF,
respectively, with the Extended BCAF (on each row). In all cases, * indicates a rejection of the null at a 10
% level

we employed the quantile Mincer–Zarnowitz regression test proposed by Gaglianone

et al. (2011).
h be the respective
Let yt be the 12-month inflation rate measured by IPCA, f i,t
inflation forecast of survey participant i formed at period t −h, and f ·,th be the consensus
forecast. The cross-sectional dimension is considered in two ways: (i) disaggregated
data, with an individual OLS regression for each i = 1, . . . , N 14 ; and (ii) aggregated

Footnote 13 continued
details, see Batchelor (2007) which presents several arguments consistent with forecasters having a
nonquadratic loss function.
14 In each test, with disaggregated data (individual OLS regressions), we only considered survey partici-
pants with more than 20 available forecasts.

123
156 W. P. Gaglianone et al.

Table 7 Rationality (MZ) tests for consumers

Variable Individual Consensus

forecasts (% of forecasts (MZ
rational) test, p value)

cons_all 35 0.000
cons_educ_1_3 26 0.000
cons_educ_4 31 0.000
cons_educ_5_6 40 0.000
cons_inc_1 22 0.000
cons_inc_2 31 0.000
cons_inc_3 35 0.000
cons_inc_4 37 0.000

In the second column, the % of rational forecasters is based on p value >0.05 in the MZ
individual rationality test and on consumers with at least 20 observations in time dimension
Consensus sample: January 2006 to July 2014. IPCA sample: December 2006 to June 2015
(103 observations). Robust standard errors (SE) from a Newey–West covariance matrix

data (consensus forecast), with OLS regression. We focus on the widely employed
rationality test of Mincer and Zarnowitz (1969), hereafter MZ. First, we consider
disaggregate data, with an individual OLS regression for each i = 1, 2, . . . , N . The
individual-forecasts OLS regressions are given as follows:

yt = αi + βi f i,t
h
+ εi,t , (14)

h denotes individual i forecast of

where yt denotes observed inflation on period t, and fi,t
yt using only information up to period t − h. The MZ test is based on a Wald test of the
joint null hypothesis of rationality, H0 : [αi = 0; βi = 1] for each i. We then compute
the proportion of agents for which we do not reject the null of rationality. Next, we
consider aggregate data (consensus forecast), with OLS regression in time-series data.
The consensus-forecast OLS regression is given by:

yt = α + β f ·,th + εt , (15)

where the rationality test is based on a Wald test of the joint null hypothesis H0 : [α =
0; β = 1]. In testing all the null hypotheses listed above, we have employed robust
(HAC) standard errors. Results for consumer forecasts are summarized in Table 7.15
When we consider tests using Mincer–Zarnowitz regressions with disaggregate
data, individual OLS regressions for each i show that 35 % of the consumers indeed
pass rationality tests at the 5 % significance level. Moreover, this percentage increases
to 40 % and 37 % for the consensuses of higher educated consumers and for those
with higher income, respectively. However, when we analyze test results using the

15 We leave the rationality tests for market forecasters for future research, since the focus here is on
leveraging the consumer data for forecasting purposes.

123
Applying a microfounded-forecasting approach to predict… 157

consensus forecasts with OLS regressions, the null is always rejected at the 5 % level,
which is consistent with results previously found in the literature, where rationality
was rejected overwhelmingly in Mincer–Zarnowitz tests (Coibion and Gorodnichenko
2012), and the results are shown in Tables 3 and 4, in which the null of zero bias is
strongly rejected.16
Next, we test for rationality by employing the quantile Mincer–Zarnowitz regression
test proposed by Gaglianone et al. (2011). Similarly to the OLS-MZ regression, the
idea is to estimate a quantile regression (QR), for a given quantile level τ , of the form:
Q τ (yt |Fi,t−h ) = θ0,i (τ ) + θ1,i (τ ) f i,t
h , in the case of individual forecasts, and of the

form: Q τ (yt |Ft−h ) = θ0 (τ ) + θ1 (τ ) f ·,t

h for the consensus forecast.

The QR-MZ test is performed by using a Wald test of the joint null hypothesis H0 :
[θ0,i (τ ) = 0; θ1,i (τ ) = 1] at a given level τ ∈ (0, 1), for individual forecasts, or H0 :
[θ0 (τ ) = 0; θ1 (τ ) = 1] for the consensus forecast. The results are presented in Table 10
of the Appendix. Overall, there is an increase in the percentage of rational forecasters
as long as the quantile level τ increases, suggesting that consumer’s rationality in
Brazil is more likely to occur at the right tail of the inflation conditional distribution.17
This result is perfectly in line with the fact that consumers systematically overestimate
inflation (which might be a sign that consumers assign more weight to positive forecast
errors than to negative ones). With respect to the consensus forecast, similar to the
result from the OLS-MZ test, we again strongly reject the null hypothesis of rationality
(at a 1 % significance level) in the QR-MZ sense, for all considered cases.
We can summarize our results in this section as follows: We provide evidence that
rationality test results based on a time series of individual forecasts are very different
from those using the time series of the aggregate measure represented by the consensus
forecast. When testing individuals separately, we find evidence of rationality for a
subset of consumers, whereas we find the opposite for the consensus forecasts. These
results are stronger when we focus on the mean forecast (standard MZ test), but are still
present when we employ the quantile Mincer–Zarnowitz regression test proposed by
Gaglianone et al. (2011). One possible explanation is that the group of “non-rational”
consumers contaminates the result of the consensus.

4 Conclusions

In this paper, we investigate empirically whether or not combining forecasts from sur-
veys of expectations is a helpful strategy for forecasting one-year-ahead CPI inflation
for Brazil. Combining forecasts has been a promising strategy at least since the sem-
inal paper of Bates and Granger (1969). The techniques used in this paper have their
roots in the pioneering work of Davies and Lahiri (1995) on what could be labeled a

16 Results in Table 7 reveal that there might be some useful information (for forecasting purposes) embodied
in the consumer consensus forecast. Also, in forming the consensuses, one might consider, for instance,
taking into account only those rational consumers when forming a “filtered-consensus” forecast, avoiding
mixing information from the non-rational forecasters. We leave this for future research.
17 Regarding income level 1, the % of rational consumers jumps from 30 % (at τ = 0.3) to 62 % (at
τ = 0.7).

123
158 W. P. Gaglianone et al.

panel-data approach to forecasting. In this literature, some of the studies focused on

forecast combination and others on density-forecast estimation, e.g., Issler and Lima
(2009), Gaglianone et al. (2011), Gaglianone and Lima (2012, 2014), Gaglianone and
Issler (2015), and Lahiri et al. (2015); see Baltagi (2013) for a broader discussion on
this topic.
We investigate inflation forecasts from the Brazilian Consumer Survey conducted
by FGV–IBRE, which consists of monthly quantitative and qualitative information
from consumers in seven major capitals of the country, with a representative sample
of the population and the rate of consumption of each capital with approximately
2000 consumers. Our sample ranges from January 2006 up to May 2015, forming an
unbalanced panel with N T = 164,479 responses (T = 113 months; and an average of
N = 1456 individuals per month). We also use The Focus Survey of the Central Bank
of Brazil. It contains daily (working days) forecasts from roughly 250 registered insti-
tutions since 1999, the year when Brazil implemented its Inflation-Targeting Regime.
About 100 of these institutions are actively feeding the database with forecasts on
any given day. Institutions include professional forecasters working in commercial
banks, asset management firms, consulting firms, non-financial firms or institutions,
academics, etc.
Natural candidates to win a forecasting competition in the literature of surveys of
expectations are the so-called consensus forecasts, i.e., the cross-sectional average
among survey respondents. In an exploratory investigation, we first show that these
consensus forecasts are a bias-ridden version of the conditional expectation of one-
year ahead inflation. No-bias tests are conducted for the intercept and slope using the
methods in Issler and Lima (2009) and Gaglianone and Issler (2015). The bias-test
results reveal interesting features of the data: Consumers systematically overestimate
inflation by 2.01 p.p., on average, whereas professional forecasters underestimate it by
0.68 p.p., on average, over the same sample. Furthermore, we show that these biases
lead to rejection in the Mincer–Zarnowitz tests of rationality using the consensus
forecasts for consumers, although from 22 to 40 % of consumers pass rationality tests
at the individual level. We also apply the quantile Mincer–Zarnowitz regression test
proposed by Gaglianone et al. (2011), where we found an increase in the percentage of
rational forecasters as long as the quantile level τ increases, suggesting that consumer’s
rationality in Brazil is more likely to occur at the right tail of the inflation conditional
distribution.
In a pseudo out-of-sample analysis, we evaluate seven different forecasting meth-
ods consisting of: A “best” model chosen by information criteria, the Granger
and Ramanathan (1984) forecast combination technique, the Phillips-curve based
forecasts, the Capistrán and Timmermann (2009) consensus-forecast method, the con-
sensus forecast, the bias-corrected average forecast (BCAF), and the extended BCAF;
we found very interesting results. The out-of-sample MSE of the AR(1) model is higher
to that of the Granger and Ramanathan combination of the consensuses of consumers
and of professional forecasters. It is also lower than the out-of-sample MSE of the
average forecasts—with a few exceptions. When we compare the MSE of the average
forecast, BCAF, and Extended BCAF, there is a clear pecking order: The extended
BCAF is more accurate than the BCAF, which, in turn, dominates the average forecast.
This validates the view that the bias corrections performed either by the BCAF, or by

123
Applying a microfounded-forecasting approach to predict… 159

the extended BCAF, are a useful device for forecasting using surveys. The results for
the Capistrán and Timmermann (2009) consensus-forecast method are mixed. It beats
the average forecast and the AR(1) model, but it is beaten by all the other combina-
tion methods and also by Phillips-curve based forecasts. The latter achieved a median
result, since it was beaten by the Granger and Ramanathan combination, the BCAF,
and the Extended BCAF.
Finally, when we compare the best forecasts for all consumers (Extended BCAF,
cons_all) with the best forecasts for all professional forecasters (Extended BCAF,
focus_day15), we observe a reduction in MSE in favor of professional forecasters of
22.5 %, which is sizable. This happens despite the fact that, for consumers, we are
averaging the forecast of approximately 1500 individuals, whereas for professional
forecasters, we employ a little less than 100 individuals. This may be a sign that it
matters to employ informed and well-trained professionals in forecasting inflation—a
relief for the profession.
All in all, if we had to suggest a single forecasting strategy for Brazilian inflation
one-year ahead, we would suggest the use of extended BCAF based on the Focus
Survey consensus (15th working day).

Appendix

In the Appendix, Tables 8–11, we provide complementary empirical evidence sup-

porting some of the results presented in the text.

123
 160

123
Table 8 Extended BCAF estimation
 
Variable kh Robust SE t stat p βh Robust SE t stat p OIR test p No Bias
value value value test p value

cons_all −1.04 2.91 0.72 1.49 0.51 0.00 0.121 0.000

cons_educ_1_3 1.53 4.90 0.76 1.29 0.85 0.13 0.277 0.000
cons_educ_4 −2.83 3.68 0.44 1.89 0.65 0.00 0.297 0.000
cons_educ_5_6 −3.47 3.76 0.36 1.85 0.66 0.01 0.378 0.000
cons_inc_1 −2.20 2.65 0.41 1.95 0.46 0.00 0.132 0.000
cons_inc_2 −14.58 13.76 0.29 3.90 2.42 0.11 0.961 0.000
cons_inc_3 −7.75 5.31 0.14 2.61 0.93 0.01 0.899 0.000
cons_inc_4 −2.07 3.11 0.51 1.54 0.55 0.00 0.270 0.000
focus_day10 −3.04 1.50 0.04 1.39 0.26 0.00 0.994 0.000
focus_day15 −2.95 1.45 0.04 1.38 0.25 0.00 0.989 0.000

GMM estimation based on the set of instruments: (lagged 1–5 months) aggregate forecasts
Consensus sample: January 2006 to July 2014. IPCA sample: December 2006 to June 2015 (103 observations). Robust standard errors (SE) from a GMM estimation of the
BCAF
Set of instruments: lagged forecasts, from 1 to 5 months. The column “OIR test” denotes the p values from the over-identifying restriction (OIR) J test due to Hansen (1982)
The column “No-Bias test” shows the p values based on the null hypothesis H o: [k h = 0;β h = 1]
W. P. Gaglianone et al.
 t
h +
Table 9 Consensus forecast OLS regression of Capistrán and Timmermann (2009): yt = a + b f ·,t

Variable a
 NW t stat p 
b NW SE t stat p Adjusted No Bias
h
f ·,t SE value value R2 test p value

cons_all 3.36 2.26 0.14 0.30 0.31 0.34 0.0258 0.0000

cons_educ_1_3 4.51 1.54 0.00 0.12 0.18 0.50 0.0009 0.0000
cons_educ_4 2.95 1.84 0.11 0.33 0.24 0.16 0.0505 0.0000
cons_educ_5_6 3.11 2.03 0.13 0.35 0.29 0.23 0.0438 0.0000
cons_inc_1 2.45 1.48 0.10 0.35 0.16 0.04 0.1253 0.0000
cons_inc_2 3.90 1.89 0.04 0.22 0.25 0.38 0.0204 0.0000
Applying a microfounded-forecasting approach to predict…

cons_inc_3 2.76 1.67 0.10 0.39 0.23 0.09 0.0869 0.0000

cons_inc_4 3.11 2.04 0.13 0.36 0.31 0.24 0.0394 0.0000
focus_day10 1.49 1.46 0.31 0.84 0.28 0.00 0.2689 0.0048
focus_day15 1.44 1.45 0.32 0.85 0.28 0.00 0.2749 0.0052

Consensus sample: January 2006 to July 2014. IPCA sample: December 2006 to June 2015 (103 observations). Newey–West (1987)’s standard errors (SE) The column
“No-Bias test” shows the p values based on the joint null hypothesis Ho : [a = 0; b = 1]

123
161
162 W. P. Gaglianone et al.

Table 10 Rationality (QR-MZ) tests for consumers

Variable τ = 0.3 (%) τ = 0.4 (%) τ = 0.5 (%) τ = 0.6 (%) τ = 0.7 (%)

cons_all 36 41 46 49 47
cons_educ_1_3 34 37 42 47 49
cons_educ_4 30 35 44 53 58
cons_educ_5_6 40 46 46 47 42
cons_inc_1 30 34 42 57 62
cons_inc_2 27 32 51 56 58
cons_inc_3 41 43 49 54 55
cons_inc_4 38 44 45 43 38

Table entries are the % of rational forecasters based on p value > 0.05 in the QR-MZ individual rationality
test and on consumers with at least 20 observations in time dimension
Forecasts’ sample: January 2006 to July 2014. IPCA sample: December 2006 to June 2015

exp imp
Table 11 Phillips curve estimation πtmar ket = α0 + α1 πt−1
mar ket + α π
2 t|t−1 + α3 πt−4 + α4 gt−1 + εt

Parameter Full sample


α0 0.0005 **
(0.0002)

α1 0.6054 ***
(0.0914)

α2 0.3787 ***
(0.0914)

α3 0.0160 **
(0.0074)

α4 0.0156 *
(0.0081)
Adjusted R 2 0.394

Monthly frequency. Sample: January 2006 to June 2015. Robust (HAC) standard errors in parentheses
Restricted regression imposes: α1 + α 2 + α 3 = 1
*, **, and ***indicate, respectively, significance at 10, 5, and 1 % level

References
Ang A, Bekaert G, Wei M (2007) Do macro variables, asset markets or surveys forecast inflation better? J
Monet Econ 54:1163–1212
Baltagi BH (2013) Panel Data Forecasting. In: Elliott G, Timmermann A (eds) Handbook of economic
forecasting, vol 2. Elsevier, New York, pp 995–1024
Batchelor R (2007) Bias in macroeconomic forecasts. Int J Forecast 23:189–203
Bates JM, Granger CWJ (1969) The combination of forecasts. Oper Res Q 20:309–325
Capistrán C, Timmermann A (2009) Forecast combination with entry and exit of experts. J Bus Econ Stat
27:428–440
Carvalho FA, Minella A (2012) Survey forecasts in Brazil: a prismatic assessment of epidemiology, perfor-
mance, and determinants. J Int Money Finance 31(6):1371–1391
Christoffersen PF, Diebold FX (1997) Optimal prediction under asymmetric loss. Econ Theory 13:808–817
Clark TE, West KD (2007) Approximately normal tests for equal predictive accuracy in nested models. J
Econom 138:291–311
Coibion O, Gorodnichenko Y (2012) What can survey forecasts tell us about information rigidities? J Polit
Econ 120:116–159

123
Applying a microfounded-forecasting approach to predict… 163

Cordeiro YAC, Gaglianone WP, Issler JV (2015) Inattention in individual expectations. Working paper no.
395, Central Bank of Brazil
Davies A, Lahiri K (1995) A new framework for analyzing survey forecasts using three-dimensional panel
data. J Econom 68:205–227
Diebold FX, Mariano RS (1995) Comparing predictive accuracy. J Bus Econ Stat 13:253–263
Elliott G, Komunjer I, Timmermann A (2008) Biases in macroeconomic forecasts: irrationality or asym-
metric loss? J Eur Econ Assoc 6(1):122–157
Elliott G, Timmermann A (2004) Optimal forecast combinations under general loss functions and forecast
error distributions. J Econom 122:47–79
Faust J, Wright JH (2013) Forecasting inflation. Handbook of economic forecasting, vol 2A, Chap 1, pp
3–56. Ed. Elsevier B.V
Gaglianone WP, Giacomini R, Issler JV, Skreta V (2016) Incentive-driven inattention. Getulio Vargas
Foundation, Mimeo
Gaglianone WP, Issler JV (2015) Microfounded forecasting. Ensaios Econômicos EPGE no. 766, Getulio
Vargas Foundation
Gaglianone WP, Lima LR (2012) Constructing density forecasts from quantile regressions. J Money Credit
Bank 44(8):1589–1607
Gaglianone WP, Lima LR (2014) Constructing optimal density forecasts from point forecast combinations.
J Appl Econom 29(5):736–757
Gaglianone WP, Lima LR, Linton O, Smith DR (2011) Evaluating value-at-risk models via quantile regres-
sion. J Bus Econ Stat 29(1):150–160
Genre V, Kenny G, Meyler A, Timmermann A (2013) Combining expert forecasts: can anything beat the
simple average? Int J Forecast 29:108–121
Granger CWJ (1969) Prediction with a generalized cost of error function. Oper Res Q 20(2):199–207
Granger CWJ, Ramanathan R (1984) Improved methods of combining forecasting. J Forecast 3:197–204
Hansen LP (1982) Large sample properties of generalized method of moments estimators. Econometrica
50:1029–1054
Hansen LP, Heaton J, Yaron A (1996) Finite-sample properties of some alternative GMM estimators. J Bus
Econ Stat 14:262–280
Issler JV, Lima LR (2009) A panel data approach to economic forecasting: the bias-corrected average
forecast. J Econom 152(2):153–164
Lahiri K, Peng H, Sheng X (2015) Measuring uncertainty of a combined forecast and some tests for
forecaster heterogeneity. CESifo working paper no. 5468
Marques ABC (2013) Central Bank of Brazil’s market expectations system: a tool for monetary policy.
Bank for International Settlements, vol 36 of IFC Bulletins, pp 304–324
Mincer JA, Zarnowitz V (1969) The evaluation of economic forecasts. In: Mincer Jacob (ed) Econmic
forecasts and expectations. National Bureau of Economic Research, New York
Palm FC, Zellner A (1992) To combine or not to combine? Issues of combining forecasts. J Forecast
11(8):687–701
Patton AJ, Timmermann A (2007) Testing forecast optimality under unknown loss. J Am Stat Assoc
102(480):1172–1184
Phillips PCB, Moon HR (1999) Linear regression limit theory for nonstationary panel data. Econometrica
67:1057–1111
Stock J, Watson M (2002a) Macroeconomic forecasting using diffusion indexes. J Bus Econ Stat 20:147–162
Stock J, Watson M (2002b) Forecasting using principal components from a large number of predictors. J
Am Stat Assoc 97(460):1167–1179
Stock J, Watson M (2006) Forecasting with many predictors. In: Elliott G, Granger CWJ, Timmermann A
(eds) Handbook of economic forecasting. North-Holland, Amsterdam, pp 515–554
Stock J, Watson M (2009) Phillips curve inflation forecasts. In: Fuhrer Jeffrey, Kodrzycki Yolanda, Little
Jane, Olivei Giovanni (eds) Understanding inflation and the implications for monetary policy. MIT
Press, Cambridge
Timmermann A (2006) Forecast combinations. In: Elliott G, Granger CWJ, Timmermann A (eds) Handbook
of economic forecasting. North-Holland, Amsterdam, pp 135–196
West K (1996) Asymptotic inference about predictive ability. Econometrica 64:1067–1084

123
Empirical Economics is a copyright of Springer, 2017. All Rights Reserved.