Economic Modelling 30 (2013) 670–684

Contents lists available at SciVerse ScienceDirect

Economic Modelling
journal homepage: www.elsevier.com/locate/ecmod

Credibility and monetary transmission channels under inflation

targeting: An econometric analysis from a developing country
Gabriel Caldas Montes
Fluminense Federal University, Department of Economics; National Council for Scientific and Technological Development (CNPq), Rua Tiradentes, 17, Ingá, Niterói, Rio de Janeiro,
CEP: 24.210-510, Brazil

a r t i c l e i n f o a b s t r a c t

Article history: After more than two decades of inflation targeting in the world, it is important to evaluate if the adoption of this
Accepted 21 September 2012 regime in a relevant developing country contributed to the creation of a better environment for the process of
entrepreneurs' expectations formation. Brazil is part of an important group of developing countries (BRIC — Brazil,
JEL classification:
Russia, India and China) and represents a potential laboratory experiment in which the effects of an adoption of
E22
inflation targeting after more than a decade can be evaluated. Not enough is known about the consequences of in-
E24
E43
flation targeting credibility on both monetary policy and monetary policy transmission channels in developing
E52 countries that adopted inflation targeting. Hence, the paper analyzes the effects of credibility on the conduct of
monetary policy, and the transmission channels through which monetary policy affect the economy. The study
Keywords: found that inflation targeting has represented an important strategy for the achievement of a more stable macroeco-
Entrepreneur nomic environment, which has been of great importance to investments and employment creation. Besides, the
Expectation study found the following novelties: inflation targeting credibility positively affects entrepreneurs' expectations,
Investment expectations of entrepreneurs affect investment decisions and employment creation and the expectations of
Employment
entrepreneurs are affected by the traditional transmission channels.
Central bank
Interest rate
© 2012 Elsevier B.V. All rights reserved.

1. Introduction mechanism for different countries. Based on many studies assessing

monetary transmission of developed countries (e.g., Christiano et al.,
At the start of 2011, 27 central banks were considered fully fledged 1999; Peersman and Smets, 2003) some consensus of how the mone-
inflation targeters, and several others were in the process of establishing tary transmission mechanism works has been reached regarding these
a full inflation-targeting (Hammond, 2011). The explicit purpose of infla- developed countries.
tion targeting is to ensure an environment of low and stable inflation and However, not enough is known about the consequences of inflation
to serve as a nominal anchor for the process of expectations formation. In targeting credibility on both monetary policy and the transmission of
this regime, the way to conduct monetary policy (more aggressively or monetary policy to the economy in developing countries that adopted
more smoothly) is closely related to the credibility of the regime, i.e., to inflation targeting. In this sense, some important questions are raised.
the power of the nominal anchor. The belief in the capacity of the regime Does this consensus view on the monetary transmission mechanism
to anchor inflation expectations is important because a higher credibility hold for a developing country (even under inflation targeting)? Is cred-
requires smaller changes in interest rates to control inflation, thereby ibility important for creating a more stable macroeconomic environ-
giving rise to minor variations also on employment and output. Thus, it ment, able to enhance monetary policy effectiveness and stimulate
is expected that a high credibility is associated with a lower volatility of investments in the economy? Do the expectations of entrepreneurs
the interest rate for the achievement of the inflation target, and as a play an important role as a transmission channel and are these expecta-
consequence, with a lower volatility of both output and employment. tions affected by the other transmission mechanisms?
Monetary policy does not directly affect the performance of macro- Therefore, after more than two decades of inflation targeting in the
economic variables such as employment, output and inflation. Until the world, it is important to evaluate if the adoption of this regime by a
macroeconomic performance is affected, monetary policy acts first on representative developing country contributed to the improvement of
the decisions of economic agents and through different transmission credibility and the creation of a better environment for the process of
channels. There is a vast literature studying the monetary transmission entrepreneurs' expectations formation, investment expenditures and em-
ployment creation. Different from other approaches, in this paper it is also
analyzed: (i) the influence of credibility on monetary policy, (ii) the influ-
ence of the four traditional transmission channels (interest rate, credit,
E-mail address: gabrielmontesuff@yahoo.com.br. exchange rate and asset prices) on entrepreneurs' expectations, and on

0264-9993/$ – see front matter © 2012 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.econmod.2012.09.035
 G.C. Montes / Economic Modelling 30 (2013) 670–684 671

investments, and (iii) the importance of investments for employment required by the central bank in terms of changes in interest rate.
creation. The work of de Mendonça (2009) found evidence that credibility
Regarding developing countries, Brazil is part of an important group – represented an important element responsible for the disinflationary
BRIC (Brazil, Russia, India and China) – and represents a potential process in Brazil, after the adoption of inflation targeting.
laboratory experiment in which the effects of an adoption of inflation With minor fluctuations in interest rate, due to greater credibility,
targeting after more than a decade can be evaluated. Hence, considering investment decisions are taken with greater confidence, leading to a
the Brazilian experience, this paper estimates the effects of inflation process of employment creation more robust.
targeting credibility on the conduct of monetary policy through changes The work of de Mendonça and Lima (2011) presents an empirical
in the basic short-term interest rate (Selic), and the transmission analysis through Generalized Method of Moments in order to find the
channels through which monetary policy can affect investment expendi- macroeconomic determinants of investment under inflation targeting
tures and employment. in Brazil. The period covered in the analysis is from January 2000 to
Hence, the main contribution of this study is the presentation of September 2009. The work tests the influence of GDP, real interest
empirical evidence for the influence of credibility on monetary policy rate, credit, inflation, exchange rate, public debt and credibility on
and, as a consequence, for the transmission of monetary policy to the private investment. According to them, the main idea of the link
economy through the traditional channels (the interest rate channel, between credibility and investment is that when credibility is high,
the credit channel, the exchange rate channel, the asset prices channel the capacity of anchoring inflation expectations improves, as a conse-
and the expectations channel), with emphasis on the expectations quence, the need to increase interest rates to promote inflation stabi-
channel, after the adoption of inflation targeting in Brazil. lization then decreases. In brief, it is expected that there exists a
For this purpose, the paper is organized as follows. Section 2 presents positive relation between credibility and investment. The results
the relation between credibility and monetary policy conduction and suggest that the variables capable of promoting private investment
explains how monetary policy affects investment decisions and, as a include output, credit, and credibility. In particular, the effect on pri-
consequence, the amount of employment on the economy through the vate investment of an increase in inflation target credibility is positive
monetary policy transmission mechanisms. Section 3 provides empirical and has statistical significance. An increase in credibility enhances
evidence about the monetary policy transmission mechanisms for the planning capacity due to lower uncertainty in the economy and con-
Brazilian economy, based on impulse response functions through vector comitant increases in investment. In brief, the success of inflation
autoregression (VAR) and estimates the relations through ordinary least targeting in the Brazilian economy is an important mechanism to pro-
square (OLS) and generalized method of moments (GMM). Section 4 mote private investment, i.e., the work found that successful inflation
concludes the paper. targeting creates a stable macroeconomic environment that promotes
private investment.
2. Credibility and the transmission channels of monetary policy Under inflation targeting, the effectiveness of monetary policy
that affect investment and employment depends on the transmission channels of monetary policy. According
to Mukherjee and Bhattacharya (2011), the literature to date empha-
Since the seminal work of Kydland and Prescott (1977), credibility sizes four key transmission channels: the traditional interest rate
plays a key role in the conduct of monetary policy. According to them, channel, the credit or loan supply channel, the exchange rate channel,
for a monetary policy being considered credible it must follow a rule and the asset price channel. However, most of the private economic
or a strategy in which the agents will believe that the monetary decisions are taken under uncertainty and, as a consequence, are
authority is not going to renege. Hence, a policy is credible when guided by expectations. Investment decisions, for example, which
the public believes on the rule/strategy and, through expectations, entrepreneurs have to take, are often based on several expectations
on the results the monetary authority is attempting to reach. concerning different aspects, such as the influence of economic poli-
As expectations are considered an important monetary transmission cies and the state of the economy on their business.
channel, the lack of credibility may hinder the reach of a certain goal due Regarding the effect of monetary policy on the economy (Fig. 1),
to expectations being formed in a context of distrust. Under inflation five transmission channels can be recognized: 1) interest rate channel,
targeting, for example, expectations and credibility are fundamental to 2) exchange rate channel, 3) credit channel, 4) asset prices channel,
the success of monetary policy. and 5) expectation channel.
When the inflation target is credible and the monetary authority Under inflation targeting, changes in the short-term interest rate
presents an increased capacity of affecting the public expectation, depends on the credibility of the regime and affect investment
less effort is necessary to achieve the inflation target. The work of decisions (and, as a consequence, employment) through a large set
de Mendonça and de Guimarães e Souza (2009) used OLS estimations of variables, including the real cost of capital, availability of bank
to test the impact of different credibility indices on the volatility of credit, the exchange rate, wealth and expectations.
interest rates. They argued that a credible monetary policy implies The traditional interest rate channel derives from standard
less effort by the Central Bank of Brazil for the achievement of the Keynesian theory. The basic assumption is that prices are sticky and
inflation target due to the increased capacity of affecting the public adjust to a monetary shock with a delay. Thus adjustments to the
expectation. In this sense, they suggest that a high credibility, ceteris short-term nominal policy interest rate lead to changes in the
paribus, is associated with a lower volatility of the interest rate for the short-term real interest rate. Movements in real interest rates in
achievement of the inflation target. Their estimates found, for the turn influence the decisions of economic agents. More specifically,
Brazilian economy, empirical evidence that confirms the hypothesis changes in the cost of capital affect the investment decision of firms.
that higher credibility implies lower variations in the interest rate Thus, it is expected that when the real interest rate increases the
for controlling inflation. cost of capital is raised, which decreases investments. Besides, when
According to Calderón et al. (2004), the cyclical properties of the real interest rate increases and turns more volatile, expectations
macroeconomic policies depend critically on policy credibility. The of entrepreneurs deteriorate, postponing investment decisions.
evidence they found for emerging countries supports that countries The study of Minella and Souza-Sobrinho (2009) developed,
with higher credibility are able to conduct countercyclical fiscal and estimated and used a semi-structural model for the inflation targeting
monetary policies. period to decompose the effects into four important individual channels
The gain in credibility is a powerful instrument for reducing infla- (the household interest rate channel, the firm interest rate channel, the
tion without increasing social costs. This is because expectations are expectation channel and the exchange rate channel). They estimated
more strongly anchored to the inflation target and thus less effort is the model equation by equation with two stage least squares or
672 G.C. Montes / Economic Modelling 30 (2013) 670–684

Author’s elaboration

Fig. 1. Credibility and monetary policy transmission channels.

ordinary least squares, using quarterly data since the implementation of as true that higher domestic interest rates lead to a currency apprecia-
inflation targeting regime (1999Q3 to 2008Q2). They found that the tion (see, for example, Mishkin, 1995). However, under certain condi-
household interest rate channel plays the most important role in tions, higher interest rates may not render the domestic currency
explaining output dynamics. Regarding inflation, the main transmission more attractive. According to Gonçalves and Guimarães (2011, p. 34)
mechanisms are the household and the exchange rate channels. When the explanation is simple: “debt default is associated with output losses.
they added a proxy for the expectation channel, this channel became Those losses reduce the amount of collectable taxes and hence the
key to understand the behavior of inflation. Concerning future re- resources available for repayment. Higher interest rates imply higher
searches, they argued that other channels may play an important role odds of a default. Thus, increasing interest rates has two opposite
as the Brazilian economy develops and financial and credit markets effects: the first is to increase repayment if debt is honored, and the
deepen. Among the potential competing channels, they highlighted second is to reduce the expected amount of available resources due to
the wealth and credit channels. the higher expected costs stemming from a greater probability of
Regarding the credit channel, this channel can be split into two default”. Theoretical models show that if interest rates adversely
different channels: the bank lending channel and the balance sheet affect default risk or monetization risk, a monetary tightening may
channel. Concerning the bank lending channel, in particular, contrac- render the domestic currency less attractive. Sargent and Wallace
tionary monetary policy results in lower usable reserve assets at com- (1981) show how raising interest rates might lead to increased
mercial banks. Moreover, a tighter monetary policy usually leads to expected inflation if households anticipate debt will eventually need
lower deposits at commercial banks through its impact on economic to be monetized. Blanchard (2005) studied the Brazilian case and
activity. These factors lower the supply of bank loans available to argued that for high debt levels and risk premia, higher interest rates
firms, thereby adversely affecting investment decisions. may spark currency depreciations. In the model of Akemann and
The transmission mechanism of monetary policy through the credit Kanczuk (2005), the effect of monetary policy tightenings causes cur-
channel for the period of inflation targeting in Brazil was approached by rency depreciations when interest rates and the level of indebtedness
different empirical studies (e.g., Auel and de Mendonça, 2011; Catão are high enough. For the Brazilian economy, Gonçalves and Guimarães
and Pagan, 2009; Catão et al., 2008; Coelho et al., 2010; de Mello and (2011) argued that the negative effect of higher interest rates on the
Pisu, 2010; Minella and Souza-Sobrinho, 2009). Among these studies, exchange rate is not necessarily confined to episodes of attacks to cur-
the work of Auel and de Mendonça (2011) found important evidence rency pegs. They found evidence that, during the period 2000–2006,
and must be highlighted. monetary policy tightenings led, on average, to currency depreciations.
The work of Auel and de Mendonça (2011) analyzes the macro- In relation to the influence of exchange rate on investments, it is
economic relevance of the credit channel in Brazil. Based on data expected that exchange rate appreciations have a positive effect on
from 2002 to 2009, three sets of GMM models are considered, a investments. The explanation is simple: if the domestic producer relies
GMM system model is built and a VAR analysis is made. The findings on imported inputs in production and investment, an appreciation of
denote that the effects of economic shocks on credit supply and on the exchange rate implies a decrease in the marginal cost of production,
credit spread are in accordance with the credit channel theory. which could lead to an expansion in investment activity.
Besides, it is observed that shocks on the interest rate are not trans- In relation to the asset price channel two channels are often em-
mitted directly to the economy but through the credit channels. In phasized in studies about the monetary transmission mechanism:
relation to the operation of credit channels for monetary policy, the these involve Tobin's q theory of investment and wealth effects on
work of Auel and de Mendonça (2011) points that empirical evidence consumption. Tobin's q theory provides an explanation about how
for emerging economies is scarce. monetary policy affects the economy through its effects on the
Regarding the exchange rate channel, in the case of an emerging valuation of equities. Tobin (1969) defines q as the market value
market economy with a flexible exchange rate regime, the precise of firms divided by the replacement cost of capital. If q is high, the
impact of monetary policy on exchange rate is uncertain and will market price of firms is high relative to the replacement cost of
depend on expectations about domestic and foreign interest rates, capital, and new plant and equipment capital is cheap relative to
inflation and government indebtedness. Economists usually assume the market value of business firms. Companies can then issue equity
 G.C. Montes / Economic Modelling 30 (2013) 670–684 673

and get a high price for it relative to the cost of the plant and equip- in order to smooth out fluctuations in the level of economic activity.
ment they are buying. Thus investment spending will rise because This delicate balance of objectives depends crucially on the credibility
firms can buy a lot of new investment goods with only a small of the regime, which between a crisis and another, has been consoli-
issue of equity. In this sense, monetary policy may affect equity dating since its implementation.
prices by rising (falling) the basic interest rate, making bonds In general, in times of crisis, the regime has been very effective
more (less) attractive relative to equities, thereby causing the and as a consequence the macroeconomic environment has become
price of equities to fall (raise). As a consequence, lower (higher) more stable. The macroeconomic stability was crucial for creating a
equity prices will lead to a lower (higher) q, and thus to lower favorable environment for investment decisions and employment
(higher) investment spending. creation.
An alternative channel for the transmission of monetary policy There is empirical evidence about the impact of inflation targeting
through the asset prices is the wealth effect on consumption. This credibility on short-term interest rates in Brazil, which suggest that
channel was built based on the life-cycle model of consumption when the credibility of the regime of inflation targeting increases, the
developed by Ando and Modigliani (1963), in which households' level and volatility of the basic interest rate decrease (de Mendonça,
wealth is a key determinant of consumption spending. The connec- 2007a; de Mendonça and de Guimarães e Souza, 2009). Moreover,
tion to monetary policy comes via the link between interest rates there is evidence that successful inflation targeting creates a stable
and asset prices, for example, as equity prices raise because the cen- macroeconomic environment that promotes private investment (de
tral bank reduced the interest rate, share-owning households become Mendonça and Lima, 2011). However, empirical evidence concerning
wealthier and may choose to increase their consumption. Conversely, the relation between credibility and the transmission mechanisms of
when equity prices fall, households may reduce consumption. It is monetary policy are scarce, and, in particular, the transmission through
important to note that the asset price channel through the wealth ef- expectations and confidence of entrepreneurs was never explored.
fect is relevant to countries like United States and England. However, Hence, an important point that must be looked into is the effect of
in the case of the Brazilian economy, the participation of consumers credibility and monetary policy on entrepreneur sentiments and, as a
in the stock market is negligible, which makes this transmission consequence, on investment expenditures and employment creation.
mechanism of little relevance in Brazil (Tomazzia and Meurer, 2009). In the present study, the period of analysis runs from 2001Q4 to
According to Mishkin (2007, p. 59), “Fluctuations of the stock 2011Q2 (the justification for using this period is because the market
market, which are influenced by monetary policy, have important expectations series available from the Central Bank of Brazil started
impacts on the aggregate economy. Transmission mechanisms involving in the last quarter of 2001). Due to the fact that “Industrial entrepre-
the stock market are of three types: 1) stock market effects on investment, neur confidence index” and “investment” are quarterly series; the
2) firm balance-sheet effects, 3) household wealth effects and 4) household analysis is done with quarterly data. In this sense, since the other
liquidity effects”. In the present paper, the analysis for the asset price series are on a monthly basis, thus, these series were turned into
transmission mechanism is based on Tobin's q theory of investment, quarterly through the calculation of the average. All series were
i.e., it is considered the stock market effects on investment. For obtained from the Central Bank of Brazil (CBB) site. The analysis
more details about the role of asset prices in monetary policy, see was performed using the software EViews 7. The (quarterly) series
Mishkin (2007, chapter 3). used are:
Most studies approach the expectation channel based on financial
market expectations about inflation, the path of future policy actions, • Selic interest rate (Selic) — (series number 4189 — CBB). This series
long-term interest rates and exchange rate. In this paper, unlike other is the “Interest rate - Selic accumulated in the month in annual
approaches made so far, the analysis of the expectation channel is terms - % p.y”.
done based on the expectations of entrepreneurs. These expectations • Real interest rate (real_ir) — for this series it was used the “Interest
consider the perception of entrepreneurs regarding the actual and rate - Selic accumulated in the month in annual terms - % p.y”
future state of the economy and their own companies. In this sense, (series number 4189 — CBB) and the “National consumer price
it is expected that the expectations of entrepreneurs are influenced index (IPCA) - in 12 months - %” (series number 13522 — CBB).
by monetary policy as well as by the other transmission channels, • Exchange rate (exch_rate) — this series is the average between two
i.e., by real interest rate, credit, exchange rate and the value of compa- series: “Exchange rate - Free - United States dollar (purchase) - end of
nies listed at the stock market. Investment decisions are positively period” (series number 3695 — CBB) and “Exchange rate - Free - United
affected by the expectations of entrepreneurs, i.e., when entrepre- States dollar (sale) - end of period” (series number 3696 — CBB).
neurs reveal optimism and confidence regarding the macroeconomic • Credit as a proportion of GDP (credit_gdp) — (series number 17473 —
environment as well as their own businesses the volume of invest- CBB). This series is the “Total credit operations in the financial system
ments increase. to private sector/GDP - %”.
• Value of companies listed at Bovespa (vc_bovespa) — (series
3. Empirical evidence number 7849 — CBB). This series is the “Value of companies listed
at Bovespa — c.m.u. (million)”.
Since January 1999, Brazil operates under a floating exchange rate • Investment (gfcf): the series of investment comes from series (7338 —
regime. The inflation targeting regime was adopted in June of the CBB) “Quarterly GDP - seasonally adjusted data (1995=100) -
same year, opening a period of transparent and accountable economic Investment – Index”.
policies. The regime of inflation targeting represented an important • Ratio between employed people and economically active population
advance in terms of commitment technology for the Central Bank of (empl_eap) — this series was constructed using the following series:
Brazil since it helped in the task of stabilizing the economy. The adop- (series number 10801) “Employed people - Total - Units (thousand)”
tion of the regime represented a clear strategy concerned about guiding and (series number 10810) “Economically active population - Units
public expectations. (thousand)”.
The Brazilian experience presented important stress tests for a re- • Industrial entrepreneur confidence index (IECI): (series number
gime of inflation targeting; according to Minella et al. (2003, p. 1039), 7341 — CBB) this series is the “Business confidence index - General –
“The Brazilian experience has been a successful stress test for the Index”. It is worth noting that the industrial entrepreneur confidence
inflation-targeting framework”. In many difficult times, the Central index (ieci) is elaborated with the assistance of the Federations of In-
Bank of Brazil has been able to act in accordance with the universally dustries of 24 Brazilian States. The index is obtained on the basis of a
accepted practice to accommodate temporary deviations from target, survey that assesses the attitudes and stances of businessmen and
674 G.C. Montes / Economic Modelling 30 (2013) 670–684

their expectations in regard to the coming six months in terms of Table 1

Brazilian economy and of their own companies. The index is Correlations.

constructed based on two other indexes, one considering actual Variables CI

conditions and the other considering expectations for the next
CI 1.00
six months. The icei varies in the 0–100 interval. SELIC −0.38
• Inflation targeting credibility (CI): The measure of credibility is
based on the index developed by de Mendonça (2007a). The Variables SELIC
credibility index has a value equal to 1 when the annual expected SELIC 1.00
inflation (E(π)) is equal to the inflation target and decreases in a REAL_IR 0.72
linear way while inflationary expectation deviates from the an- IECI −0.41
CREDIT_GDP −0.82
nounced target. Therefore, the credibility index shows a value EXCH_RATE 0.80
between 0 and 1 strictly if the expected inflation is situated be- VC_BOVESPA −0.87
tween the maximum and minimum limits (πt*) established for
each year and assumes a value equal to 0 when the expected in- Variables GFCF
flation exceeds one of these limits. The index uses the series of GFCF 1.00
inflation expectations obtained from the CBB site and, the infla- REAL_IR −0.77
tion target defined by the monetary authority and the tolerance IECI 0.48
CREDIT_GDP 0.94
bands. Hence, EXCH_RATE −0.84
8 9 VC_BOVESPA 0.94
>
> 1 if ΕðπÞ ¼ πt >
>
< 1 =
CI ¼ 1−  ½ΕðπÞ−πt  if πt Min < ΕðπÞ < πt Max : Variables EMPL_EAP
>
> π −π >
>
: t t
  ; EMPL_EAP 1.00
0 if ΕðπÞ≥ πt Max or ΕðπÞ≤ πt Min GFCF 0.95

Author's elaboration.
Although different indexes of credibility have been proposed – as
summarized in the works of de Mendonça and de Guimarães e Souza
(2009), Nahon and Meurer (2009) and Garcia and Guillén (2009) –
and therefore there is a variety of indexes of credibility capable of effect of the traditional transmission mechanisms on entrepreneurs'
being used in empirical analyses, the present work does not seek to expectations and the impact of the transmission mechanisms on
analyze the influence and power of each index on monetary policy investments. In order to validate the VAR, diagnostic tests for the
in Brazil – although such research is important. Thus, the option for absence of autocorrelation, heteroscedasticity and non-normality
using the index proposed by de Mendonça (2007a) is due to the in the error process are performed. Due to the fact that impulse re-
following arguments: (i) the index is recognized by international lit- sponses that are calculated from unrestricted VARs with roots near
erature, being this index used in several applied studies, (ii) simplic- unity and cointegrating relations among the variables have long
ity of understanding and preparation, (iii) the index captures the period estimated impulse responses that may be inconsistent and,
changes and fluctuations in credibility in a way compatible with the thus, “policy analysis that is undertaken from unrestricted VARs using
regime of inflation targeting adopted in Brazil, i.e., the index uses estimated impulse responses can be expected to be inherently uncertain
predetermined tolerance bands, and not ad-hoc tolerance bands as even in large samples as the horizon increases” (Phillips, 1998, p. 23),
proposed by other indices, and (iv) the index is rigorous enough it is commonly suggested the estimation of a reduced rank VAR
and punishes appropriately deviations of inflation expectations in when the data are cointegrated. In other words, before conducting
relation to the inflation target. impulse response analysis one should check the cointegrating prop-
Considering the period of the analysis, Table 1 below presents the erties of the data and restrict the VAR. However, according to
correlations between the variables of the analysis. Mitchell (2000), “there are practical dangers associated with following
Fig. 2 presents the evolution of the Brazilian macroeconomic this advice. When the data are ‘nearly’ cointegrated then not just does
variables. the VAR in levels provide misleading inference about long-run impulse
responses but so does the reduced rank VAR”. In this sense, based
3.1. Methodology on the results presented by Phillips (1998) and Mitchell (2000), a
Vector Error Correction Model (VECM) was also incorporated in
Since this paper is concerned with the effects of central bank cred- order to give robustness for the VAR analysis.
ibility under inflation targeting on the conduct of monetary policy - Estimations of single equations through Ordinary Least Squares
and its transmission channels in Brazil, an econometric analysis is re- (OLS) and Generalized Method of Moments (GMM) — The esti-
alized. It is important to highlight that one of the main contributions mates make use of OLS and GMM. One reason for using GMM is
of the paper is the use of entrepreneurs' expectations in the study. that while OLS estimates often have problems of serial autocorrela-
The econometric analysis is performed to find out the influence of tion, heteroskedasticity or non-linearity, which is typical in macro-
the following: credibility on monetary policy, the traditional trans- economic time series, this method provides consistent estimators
mission mechanisms on entrepreneurs' expectations and on invest- for the regression (Hansen, 1982). As pointed out by Wooldridge
ments, and of investments on job creation. For this, the following (2001, p. 95), “to obtain a more efficient estimator than two-stage
methods are applied: least squares (or ordinary least squares), one must have overriding
restrictions”. The weighting matrix in the equation was chosen to
- Impulse response through Vector Autoregression (VAR) and Vector enable the GMM estimates to be robust, considering the possible
Error Correction Model (VECM) for a dynamic analysis of the trans- presence of heteroskedasticity and autocorrelation of unknown
mission mechanisms — This method is applied with the objective of form. Regarding the influence of credibility on monetary policy,
verifying the monetary policy transmission mechanism. In this the analysis seeks to provide robustness for the finding of the VAR
sense, through impulse response analysis, it is possible to observe and also for the results found by de Mendonça and de Guimarães
the influence of credibility on monetary policy (Selic), the influence e Souza (2009). Thus, the estimated equation for the relation be-
of monetary policy (Selic) on the transmission mechanisms, the tween credibility and monetary policy is inspired on the analysis
 G.C. Montes / Economic Modelling 30 (2013) 670–684 675

CI SELIC REAL_IR
1.0 28 14

12
0.8 24

10
0.6 20
8
0.4 16
6

0.2 12
4

0.0 8 2
01 02 03 04 05 06 07 08 09 10 11 01 02 03 04 05 06 07 08 09 10 11 01 02 03 04 05 06 07 08 09 10 11

IECI CREDIT_GDP EXCH_RATE

70 50 4.0

45
65 3.5

40
60 3.0
35
55 2.5
30

50 2.0
25

45 20 1.5
01 02 03 04 05 06 07 08 09 10 11 01 02 03 04 05 06 07 08 09 10 11 01 02 03 04 05 06 07 08 09 10 11

VC_BOVESPA GFCF EMPL_EAP

3,000,000 180 74

2,500,000 72
160

2,000,000 70
140
1,500,000 68
120
1,000,000 66

100
500,000 64

0 80 62
01 02 03 04 05 06 07 08 09 10 11 01 02 03 04 05 06 07 08 09 10 11 01 02 03 04 05 06 07 08 09 10 11

Fig. 2. Macroeconomic variables.

of de Mendonça and de Guimarães e Souza (2009); however, the test was also considered, because it is generally recommended for
present study also makes use of GMM in order to give robustness. analysis with small samples. According to the KPSS test (Table 2b)
In turn, aiming at giving robustness for the findings of the VAR, all series presented stationarity. It is important to note that the
two equations are estimated (through OLS and GMM) concerning KPSS test tends to be biased for this type of result. In this sense, for
how investment is affected by the transmission mechanisms and those equations presenting series that are not cointegrated, the esti-
how entrepreneurs' expectations are affected by credibility, mone- mates were done not only with the series in first difference, but also
tary policy and the traditional mechanisms. Moreover, one last with the series in level — as done in the work of de Mendonça and
equation is estimated (through OLS and GMM) for the influence of de Guimarães e Souza (2009).
investment on job creation. It is important to highlight that all OLS
estimates are followed by diagnostic tests for the absence of auto- 3.2. VAR analysis
correlation, heteroscedasticity, non-normality and model specifica-
tion (Ramsey RESET test), and; for all GMM estimates, the Since the seminal work of Sims (1980), it has become a common
instrumental variables are presented and a standard J-test was practice to estimate the effects of the monetary policy on the econo-
performed with the objective of testing the selection of such instru- my using vector autoregressions (VAR). In a general way the dynamic
mental variables. analysis of vector autoregressive models is made through methods
such as impulse response functions because it permits evaluation of
Fig. 3 below presents the transmission mechanisms and summarizes the impulse on key variables caused by shocks (or innovations) pro-
the methods used for each analysis. voked by residual variables over time (Sims, 1980). As pointed out
A first condition to be analyzed before applying the econometric by Lutkenpohl (1991), the conventional method applies “orthogonal-
analysis is to check if series have unit root. Therefore, the unit-root ity assumption” and thus the result may depend on the ordering of
test Phillips–Perron (PP) was applied (Table 2a). Besides, the KPSS variables in the VAR. The works of Koop et al. (1996) and Pesaran
676 G.C. Montes / Economic Modelling 30 (2013) 670–684

VAR estimates (impulse response)

Transmission channels:

REAL_IR

IECI

Credibility affects SELIC when SELIC changes , then, CREDIT_GDP change, and, as a consequence, affect GFCF and EMPL_EAP

EXCH_RATE
OLS and GMM estimates
VC_BOVESPA OLS and GMM estimates

OLS and GMM estimates

Author’s elaboration.
Fig. 3. Methods of the analysis.

and Shin (1998) developed the idea of the generalized impulse Fig. 4 presents the results of the generalized impulse response
response function as a manner of eliminating the problem of the or- functions for a time horizon of 10 months. Fig. 5 presents the VAR
dering of variables in the VAR. The main argument is that the stability. The diagnostic tests for the absence of autocorrelation,
generalized impulse responses are invariant to any re-ordering of heteroscedasticity or non-normality in the error process are
the variables in the VAR. presented at the appendix A (see Tables A.1, A.2 and A.3). Besides,
The VAR analysis was performed for the following set of series — CI, due to the fact that the Johansen cointegration tests suggest at
SELIC, IECI, REAL_IR, CREDIT_GDP, EXCH_RATE, VC_BOVESPA and GFCF. least four cointegration vectors, a Vector Error Correction Model
The unit-root test Phillips–Perron (PP) indicates that all series are I(1). (VECM) is used in order to give robustness for the VAR analysis
However, it is possible to estimate using all series in level, if they are (see Fig. A.1 at the Appendix A). It is important to note that the
cointegrated. Hence, the cointegration test proposed by Johansen results of the VECM corroborate the results of the VAR.
(1991) was performed for the set of series — CI, SELIC, IECI, REAL_IR, According to Fig. 4, the Selic interest rate is negatively affected, with
CREDIT_GDP, EXCH_RATE, VC_BOVESPA and GFCF. statistical significance, by an unexpected positive shock on credibility.
For the definition of the VAR lag order, the criterion of Schwarz (SC) This evidence is consistent with the empirical literature for the Brazilian
was used. Table 3 indicates that the VAR order is 1. The cointegration economy, which establishes that a higher credibility implies lower
test, based on the significance of the estimated eigenvalues, indicates variations in the interest rate for controlling inflation (de Mendonça
that the trace statistic rejects the non-cointegration hypothesis and de Guimarães e Souza, 2009), and, as a consequence, a lower level
(Table 4). In this sense, it is possible to estimate the VAR using all series for the interest rate (de Mendonça, 2007b).
in level. The results suggest, with statistical significance, that an unex-
pected positive shock on Selic interest rate causes: (i) an increase
in real interest rate, (ii) a decrease in the expectations of entrepre-
Table 2a neurs, (iii) a decrease in credit, (iv) a currency depreciation, and
Phillips–Perron (PP) test. (v) a decrease in the value of companies listed at Bovespa.
Series PP It is important to note that the evidence found for the exchange
rate channel reinforce the findings of Akemann and Kanczuk (2005)
Bandwidth Test 1% critical values 5% critical values
and Gonçalves and Guimarães (2011) that the effect of monetary
gfcf 7 −1.939 −4.219 −3.533
d(gfcf) 34 −4.593 −4.227 −3.537
ieci 7 −3.270 −4.219 −3.533
d(ieci) 33 −10.722 −4.227 −3.537 Table 2b
real_ir 2 −1.841 −4.219 −3.533 KPSS test.
d(real_ir) 1 −4.876 −4.227 −3.537
credit_gdp 2 −2.579 −4.219 −3.533 Series KPSS
d(credit_gdp) 2 −3.879 −4.227 −3.537
Bandwidth Test 1% critical values 5% critical values
exch_rate 9 −2.814 −4.219 −3.533
d(exch_rate) 15 −4.242 −4.227 −3.537 gfcf 3 0.121 0.216 0.146
vc_bovespa 1 −2.507 −4.219 −3.533 ieci 0 0.085 0.216 0.146
d(vc_bovespa) 5 −3.698 −4.227 −3.537 real_ir 5 0.124 0.216 0.146
selic 4 −2.396 −4.219 −3.533 credit_gdp 5 0.178 0.216 0.146
d(selic) 15 −2.242 −2.629 −1.950 exch_rate 2 0.102 0.216 0.146
empl_eap 1 −2.390 −4.219 −3.533 vc_bovespa 3 0.074 0.216 0.146
d(empl_eap) 6 −4.828 −4.227 −3.537 selic 2 0.065 0.216 0.146
ci 13 −2.625 −4.219 −3.533 empl_eap 4 0.091 0.216 0.146
d(ci) 36 −8.185 −4.227 −3.537 ci 3 0.093 0.216 0.146

Author's estimates. Phillips–Perron test (PP): the number of lags for each series was Author's estimates. KPSS: the number of lags for each series was defined according to
defined according to the Newey–West Bandwidth, i.e., lag is the lag truncation chosen the Newey–West Bandwidth, i.e., lag is the lag truncation chosen for the Bartlett Ker-
for the Bartlett Kernel. Intercept and linear trend were used for all series. nel. Based on Schwarz criterion, intercept and linear trend were used for all series.
 G.C. Montes / Economic Modelling 30 (2013) 670–684 677

Table 3 following estimates: (i) the Influence of credibility on monetary pol-

VAR lag order. icy interest rate, (ii) the influence of the transmission mechanisms on
SC investments and (iii) the influence of investments on employment.
The regime of inflation targeting aims to anchor market expecta-
With constant Without constant
tions about the behavior of inflation and to stabilize the economy.
0 54.60
Empirical evidence suggests that a higher credibility requires smaller
1 45.03 45.73
2 45.90 46.11 changes in interest rates to control inflation, thus resulting in minor
variations of employment and output (Amano et al., 1999 and de
Author's estimates.
Mendonça and de Guimarães e Souza, 2009).
In order to analyze the influence of inflation targeting credibility
on Selic interest rate, the following functions were estimated:

policy tightenings causes currency depreciations when interest rates dðselicÞ ¼ f ½dðCIÞ ð1aÞ
and the level of indebtedness are high enough.
In relation to the influence of the monetary transmission mecha- selic ¼ f ðCIÞ ð1bÞ
nisms on expectations, it is worth highlighting the effects of the real
interest rate, the exchange rate and the asset prices. The results
Eq. (1a) assumes this form due to the fact that, according to the
suggest that when the real interest rate increases and the exchange
unit root test (PP), both selic and CI are I(1) and the series are not
rate devaluates, expectations deteriorate. In turn, when the value of
cointegrated (see Tables A.4 and A.5 at the Appendix A), thus, the
companies listed at Bovespa increases, the expectations of the entre-
estimates were done with series in first differences. But, due to the
preneurs increase. Due to lack of statistical significance for the influ-
fact that the stationarity test (KPSS) points that both series are I(0),
ence of credibility on the expectations of entrepreneurs, this graph
thus Eq. (1b) was also estimated — as de Mendonça and de
was removed.
Guimarães e Souza (2009) have done. The hypothesis adopted is
Regarding the influence on investments, the evidence points that:
that departures of inflation expectations from inflation target –
(i) both monetary policy and real interest rate exert negative effects,
which means reduction of credibility – imply an increase in the
but statistical significance is observed just for the influence of mone-
variation of basic interest rate (Selic). Thus, regarding the influence of
tary policy, (ii) unexpected positive shocks on expectations and on
inflation targeting credibility on Selic interest rate, Tables 5a and 5b
the value of companies listed at Bovespa cause, with statistical signif-
present the estimates by OLS and GMM.
icance, an increase in investments and (iii) currency depreciations
For the OLS estimates, Tables 5a and 5b show that the F-statistics
make imported capital goods more expensive and therefore discour-
indicate that the regressions are significant; besides, the outcomes
age investments.
of the Ramsey RESET tests indicate that the estimations do not
These findings reinforce and complement some results presented
present problem of model specification. In terms of GMM, the results
by the work of de Mendonça and Lima (2011). According to them,
of the J-test indicate that the models are correctly specified.
the success of inflation targeting in the Brazilian economy is an
In terms of the GMM technique, the use of instruments needed to
important mechanism to promote private investment, i.e., successful
be dated to the period t-1 or earlier as a condition for predicting the
inflation targeting creates a stable macroeconomic environment that
contemporaneous variables, this procedure follows Johnston (1984).
promotes private investment. The analysis also complements and
Moreover, Cragg (1983) pointed out that overidentification analysis
reinforces the work of Auel and de Mendonça (2011) which found
has an important role in the selection of instrumental variables to
that shocks on the interest rate are transmitted to the economy
improve the efficiency of the estimators. Hence, a standard J-test
through the credit channels. However, Auel and de Mendonça
was performed with the objective of testing this property for the va-
(2011) did not found evidence for the transmission through the
lidity of the overidentifying restrictions (Hansen, 1982). The GMM
credit supply; in fact, they found evidence that the transmission
estimates of the Eq. (1a) apply the following instrumental variables:
works through the credit spread.
constant, d(SELIC)(−1 to − 6) and d(CI)(− 2 to − 5). In turn, The
GMM estimates of the Eq. (1b) apply the following instrumental
variables: constant, SELIC(− 1 to − 6) and CI(− 2 to − 6).
3.3. Estimates
The estimates for Eqs. (1a) and (1b) point to the same type of rela-
tionship. The estimates reveal, with statistical significance, that when
In order to understand the influence of credibility on monetary
credibility increases, the Selic interest rate decreases, and vice-versa.
policy, and as a consequence, the effect of the monetary transmission
In order to analyze how entrepreneurs' expectations are affected,
mechanisms on the economy, the study was performed based on the
the following function was estimated:
 
IECI ¼ f SELIC; REAL–IR; CREDIT –GDP; EXCH –RATE; VC –BOVESPA; CI :
ð2Þ
Table 4
Johansen cointegration test.
Although the KPSS tests indicate that the series are I (0), the PP
Trend assumption: No deterministic trend
tests indicate that the series are I(1). If the results of the KPSS tests
Hypothesized no. of Eigenvalue Trace statistic 0.05 critical value Prob.⁎⁎ are considered, then the estimates are made with the series in level
CE(s)
without concerns. However, if the results considered are those of
None⁎ 0.871 232.245 143.669 0.000 the PP tests, then it is possible to estimate using all series in level, if
At most 1⁎ 0.765 156.479 111.781 0.000
all are I(1) and if they are cointegrated. Hence, the cointegration
At most 2⁎ 0.639 102.974 83.937 0.001
At most 3⁎ 0.546 65.230 60.061 0.017 test was performed for the set of series — IECI, SELIC, REAL_IR,
At most 4 0.384 36.049 40.175 0.123 CREDIT_GDP, EXCH_RATE, VC_BOVESPA and CI. The choice of the VAR
Author's estimates.
lag order was determined using the Schwarz (SC) information criteri-
⁎ Denotes rejection of the hypothesis at the 0.05 level. on (Table 6). The cointegration test indicates that the trace statistic
⁎⁎ MacKinnon–Haug–Michelis (1999) p-values. rejects the non-cointegration hypothesis (Table 7).
678 G.C. Montes / Economic Modelling 30 (2013) 670–684

Response to Generalized One S.D. Innovations ±2 S.E.

Response of EXCH_RATE Response of GFCF to
Response of SELIC to CI to SELIC Response of IECI to REAL_IR Response of GFCF to SELIC EXCH_RATE
2 .4 8 8 8

6
1 .2 4 4
4

0 .0 2 0 0

0
-1 -.2 -4 -4
-2

-2 -.4 -4 -8 -8
1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10

Response of VC_BOVESPA Response of IECI to Response of GFCF to

Response of REAL_IR to SELIC to SELIC CREDIT_GDP Response of GFCF to REAL_IR VC_BOVESPA
400,000 8 8 8
1.0
6
200,000 4 4
0.5 4

0 2 0 0
0.0
0
-200,000 -4 -4
-0.5 -2

-400,000 -4 -8 -8
1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10

Response of IECI to
Response of IECI to SELIC EXCH_RATE Response of GFCF to IECI
8 8 8

6 6
4
4 4

2 2 0

0 0
-4
-2 -2

-4 -4 -8
1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10

Response of CREDIT_GDP to Response of IECI to Response of GFCF to

SELIC VC_BOVESPA CREDIT_GDP
4 8 8

3 6
4
2 4

1 2 0

0 0
-4
-1 -2

-2 -4 -8
1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10

Author’s elaboration.
Fig. 4. Impulse response.

Regarding the influence of the monetary transmission mechanisms as The estimates in Table 8 reveal, with statistical significance, that
well as the influence of both credibility and monetary policy on entrepre- both the real interest rate and the Selic interest rate negatively affect
neurs' expectations, Table 8 presents the estimates by OLS and GMM. The the expectations of entrepreneurs. These results corroborate the
GMM estimate (Table 8) applies the following instrumental variables: dynamic analysis made through VAR.
constant, IECI(−1 to −4), SELIC(−4 to −8), REAL_IR(−2 to −4), Regarding the influence of credit on the expectations of entrepre-
CREDIT_GDP(−2 to −6) EXCH_RATE(−3 to −4), CI(−2) and neurs, the estimated relationships are ambiguous and do not present
VC_BOVESPA(−2 to −7). The estimates include a dummy variable statistical significance.
in order to capture the subprime crisis, called dummy_sub. This dummy The estimates for the influence of the exchange rate on the expec-
variable captures the effect of the subprime crisis, it assumes value 1 tations of entrepreneurs found evidence of a positive relationship,
from 2008Q4 to 2009Q2, and zero otherwise. with statistical significance. This result contradicts the result found
According to the estimates, the dummy variable is negative and is by the VAR analysis. However, it is possible to give a plausible expla-
statistically significant in both specifications. nation for this contradiction in order to make sense. Changes in
Regarding the OLS estimate, Table 8 shows that the F-statistic exchange rates affect business decisions and therefore entrepreneurs'
indicates that the regression is significant; besides, the outcomes of expectations in the following ways: (i) being the Marshall–Lerner
the Ramsey RESET test indicate that the estimation does not present condition respected, currency depreciations stimulate exports and
problem of model specification. In terms of GMM, the result of the discourage imports, thus, the expectations of exporters (entrepre-
J-test indicates that the model is correctly specified. neurs) become optimistic; (ii) in turn, currency depreciations
 G.C. Montes / Economic Modelling 30 (2013) 670–684 679

Inverse Roots of AR Table 5b

OLS and GMM estimates.
Characteristic Polynomial
1.5 Dependent variable: selic

Explanatory variable OLS estimate GMM estimate

1.0 Selic Selic

Constant 22.938*** 23.440***

(2.926) (1.982)
0.5 [7.837] [11.822]
(CI)-1 −11.528*** −12.920**
(3.882) (2.943)
[−2.969] [−4.389]
0.0 R2 0.31 0.17
Adjusted R2 0.29 0.15
F-statistic 16.33
-0.5 Prob (F-statistic) 0.00
Ramsey (1); p-value 0.05
Ramsey (2); p-value 0.08
Serial correlation LM Test (1); p-value 0.00
-1.0 Serial correlation LM Test (2); p-value 0.00
Heteroskedasticity (ARCH) (1) 0.00
Heteroskedasticity (ARCH) (2) 0.00
-1.5 Normality (Jarque-Bera); probability 0.28
-1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 J-statistic 8.16
Prob (J-statistic) 0.61
Author’s elaboration. Rank 12

Author's estimates. NOTE: the reported t-statistics in the OLS estimates are based on
Fig. 5. VAR stability.
the estimator of Newey and West (1987), which is consistent in the presence of both
heteroskedasticity and autocorrelation of unknown form. Marginal significance levels:
*** denotes 0.01, ** denotes 0.05 and * denotes 0.1.
discourage imports of capital goods and make imported inputs more
expensive (which could boost inflation) leading, in this sense, the
expectations of entrepreneurs to deteriorate. The estimates for the influence of credibility on entrepreneur's
The estimates reveal, with statistical significance, a positive relation expectations reveal that credibility plays a key role in the analysis.
between entrepreneurs' expectations and the value of companies listed Despite that the VAR analysis does not present statistical significance
at Bovespa, i.e., the asset price channel works as expected. This result for the influence of credibility on entrepreneurs' expectations, the
reinforces the evidence found in the dynamic analysis using VAR. estimates through OLS and GMM present evidence of a positive
When the value of companies in the stock market increases, entrepre- relation and with statistical significance.
neurs' confidence in the economy increases and their expectations Regarding the influence of the transmission mechanisms on
become optimistic. Fig. 2 above shows, for the period of the subprime investments, the following function was estimated:
crisis, that the value of companies in the stock market (VC_BOVESPA),
the expectations and confidence of entrepreneurs (IECI) and invest-
ments (GFCF) presented the same behavior.  
GFCF ¼ f IECI; REAL–IR; CREDIT –GDP; EXCH –RATE; VC –BOVESPA : ð3Þ
Table 5a
OLS and GMM estimates.

Dependent variable: d(selic) The cointegration test was performed for the set of series — GFCF, IECI,
REAL_IR, CREDIT_GDP, EXCH_RATE, and VC_BOVESPA. It is observed that
Explanatory variable OLS estimate GMM estimate
the VAR lag order is 1 (Table 9). The cointegration test indicates that
d(selic)ols d(selic)gmm the trace statistic rejects the non-cointegration hypothesis (Table 10).
Constant −0.157 −0.337* In relation to the influence of the transmission mechanisms on
(0.297) (0.195) investments, Table 11 presents the estimates by OLS and GMM. The
[−0.528] [−1.731]
GMM estimates (Table 11) apply the following instrumental variables:
d(CI)-1 −5.045*** −7.243***
(1.339) (0.794) constant, GFCF(−1 to −3), IECI(−2 to −5), REAL_IR(−3 to −4),
[−3.767] [−9.113] CREDIT_GDP(−2 to −3), EXCH_RATE(−3 to −5) and VC_
R2 0.35 0.20 BOVESPA(−1). Once again, the estimates include a dummy variable in
Adjusted R2 0.33 0.18 order to capture the subprime crisis. According to the estimates in
F-statistic 18.79
Prob (F-statistic) 0.00
Table 11, the dummy variable is negative and is statistically significant
Ramsey (1); p-value 0.63 in both specifications.
Ramsey (2); p-value 0.86
Serial correlation LM Test (1); p-value 0.001
Serial correlation LM Test (2); p-value 0.007 Table 6
Heteroskedasticity (ARCH) (1) 0.44 VAR lag order.
Heteroskedasticity (ARCH) (2) 0.25
Normality (Jarque-Bera); probability 0.60 VAR order SC (Schwarz criterion)
J-statistic 6.88
With constant Without constant
Prob (J-statistic) 0.64
Rank 11 0 48.111
1 39.165 40.029
Author's estimates. NOTE: the reported t-statistics in the OLS estimates are based on
2 40.466 41.024
the estimator of Newey and West (1987), which is consistent in the presence of both
3 39.899 40.573
heteroskedasticity and autocorrelation of unknown form. Marginal significance levels:
*** denotes 0.01, ** denotes 0.05 and * denotes 0.1. Author's estimates.
680 G.C. Montes / Economic Modelling 30 (2013) 670–684

Table 7 increases). Besides, investments are also stimulated when credit

Johansen cointegration test. expands, the value of companies listed at Bovespa increases, the
Trend assumption: No deterministic trend currency appreciates and the real interest rate decreases.
In order to analyze the influence of investment on employment,
Hypothesized no. of Eigenvalue Trace statistic 0.05 critical value Prob.⁎⁎
CE(s) the following functions were estimated:
None⁎ 0.776 150.868 111.781 0.000  
d empl–eap ¼ f ½dðgf cf Þ ð4aÞ
At most 1⁎ 0.614 95.514 83.937 0.006
At most 2⁎ 0.500 60.282 60.061 0.048
At most 3 0.383 34.622 40.175 0.162 empl–eap ¼ f ðgf cf Þ: ð4bÞ
Author's estimates.
⁎ Denotes rejection of the hypothesis at the 0.05 level. Eq. (4a) assumes this form due to the fact that, according to the
⁎⁎ MacKinnon–Haug–Michelis (1999) p-values. unit root test (PP), both empl_eap and gfcf are I(1) and the series
are not cointegrated (see Tables A.6 and A.7 at the Appendix A),
thus, the estimates were done with series in first differences. But,
due to the fact that the stationarity test (KPSS) points that both series
Concerning the OLS estimate, Table 11 shows that the F-statistic are I(0), thus Eq. (1b) was also estimated in order to give robustness
indicates that the regression is significant; besides, the outcomes of for the analysis.
the Ramsey RESET test indicate that the estimation does not present Regarding the influence of investment on employment, Tables 12a
problem of model specification. In terms of GMM, the result of the and 12b present the estimates by OLS and GMM. The GMM estimates of
J-test indicates that the model is correctly specified. Eq. (4a) apply the following instruments: constant, d(empl_eap)(−1 to
As can be seen, the signs of the coefficients of the explanatory vari- −3), d(gfcf)(−1), d(selic) and dummy_sub. In turn, the GMM estimates
ables are those to be expected, in all specifications. It is observed, with of Eq. (4b) apply the following instruments: constant, empl_eap(−1 to
statistical significance for all estimates, that when the expectations of −4), gfcf(−1 to −4), selic and dummy_sub.
the entrepreneurs on the economy and their own businesses become The results reveal, as theoretically expected and with statistical sig-
optimistic (i.e., IECI increases), investments are stimulated (i.e., GFCF nificance, the positive relation between investment and employment.

4. Conclusion
Table 8
OLS and GMM estimates: IECI.
Under inflation targeting, the main instrument that the central
Dependent variable: IECI bank controls to affect the economy is the basic interest rate. However,
Explanatory variable OLS estimate GMM estimate
the effects of monetary policy on the economy (for instance, on invest-
ment and employment) do not occur directly, but through the impact
IECIols IECIgmm
on other economic variables. The widespread adoption of inflation
Constant 61.456*** 29.682*** targeting regimes by developing economies has generated considerable
(18.830) (4.281)
interest in the channels through which monetary policy shocks affect
[3.263] [6.933]
SELIC-3 −0.687* −0.722*** economic aggregates. However, there has been a paucity of empirical
(0.349) (0.027) research for developing economies relative to the large literature on
[−1.966] [−25.792] advanced countries.
REAL_IR-1 −0.863* −0.210** The present study approached the transmission mechanisms of
(0.442) (0.081)
[−1.952] [−2.571]
monetary policy in a different way to existing works. Most of the
CREDIT_GDP-1 −0.367 0.0014 empirical studies for the transmission mechanisms have focused on
(0.260) (0.038) the credit channel, the interest rate channel and the exchange rate
[−1.414] [0.039] channel. There are few studies that explore the expectations channel,
EXCH_RATE-2 6.561** 12.838***
and when they do, generally use expectations for both inflation and
(2.920) (0.735)
[2.246] [17.452] interest rates. (i.e., through the term structure of interest rates and
VC_BOVESPA-1 0.0000046* 0.0000054*** the Phillips curve). Moreover, most of existing works is focused on
(0.0000025) (0.00000033) the analysis of developed countries. In this sense, this work differs
[1.848] [16.296] from the others in the following aspects: 1) it analyzes not just one
CI-1 7.274⁎ 6.626***
monetary transmission mechanism, but those that are traditionally
(3.889) (0.463)
[1.870] [14.308] recognized by the literature; 2) it incorporates the importance of
dummy_sub −7.971*** −12.737*** the credibility of the regime of inflation targeting for the conduct
(2.486) (0.271) of monetary policy and for the expectations of entrepreneurs; 3) it
[−3.205] [−46.844]
explores the interdependence between monetary transmission
R2 0.68 0.53
Adjusted R2 0.61 0.40 mechanisms, when, for instance, analyzing the influence that exerts
F-statistic 8.83
Prob (F-statistic) 0.00
Ramsey (1); p-value 0.45
Ramsey (2); p-value 0.39 Table 9
Serial correlation LM Test (1); p-value 0.28 VAR lag order.
Serial correlation LM Test (2); p-value 0.03
Heteroskedasticity (ARCH) (1) 0.35 VAR order SC (Schwarz criterion)
Heteroskedasticity (ARCH) (2) 0.60
With constant Without constant
Normality (Jarque-Bera); probability 0.74
J-statistic 7.12 0 50.921
Prob (J-statistic) 0.99 1 43.482 43.558
Rank 27 2 44.639 44.862
3 44.433 45.107
Author's estimates. Note: Marginal significance levels: *** denotes 0.01, ** denotes 0.05
and * denotes 0.1. Author's estimates.
 G.C. Montes / Economic Modelling 30 (2013) 670–684 681

Table 10 Table 12a

Johansen cointegration test. OLS and GMM estimates.

Trend assumption: No deterministic trend Dependent variable: d(empl_eap)

Hypothesized no. of Eigenvalue Trace statistic 0.05 critical value Prob.⁎⁎ Explanatory variable OLS estimate GMM estimate
CE(s)
d(empl_eap)ols d(empl_eap)gmm
None⁎ 0.727 118.031 83.937 0.000
Constant 0.122 0.101
At most 1⁎ 0.626 70.048 60.061 0.006
(0.080) (0.068)
At most 2 0.406 33.705 40.175 0.192
[1.514] [1.478]
Author's estimates. d(gfcf) 0.035*** 0.037***
⁎ Denotes rejection of the hypothesis at the 0.05 level. (0.012) (0.008)
⁎⁎ MacKinnon–Haug–Michelis (1999) p-values. [2.874] [4.362]
R2 0.18 0.19
Adjusted R2 0.16 0.16
F-statistic 8.26
the interest rate, the exchange rate, the credit and the asset prices on Prob (F-statistic) 0.00
the expectations of entrepreneurs; 4) it also examines the influence Ramsey (1); p-value 0.79
of monetary policy, through the monetary transmission channels, Ramsey (2); p-value 0.82
on investments in fixed capital and the effects on employment, and; Serial correlation LM Test (1); p-value 0.32
Serial correlation LM Test (2); p-value 0.54
5) the work analyzes the Brazilian case, an important developing Heteroskedasticity (ARCH) (1) 0.29
country that currently uses the regime of inflation targeting. Heteroskedasticity (ARCH) (2) 0.38
Comparing with the existent literature for both developed and Normality (Jarque-Bera); probability 0.99
developing economies, this study goes beyond and presents new J-statistic 4.19
Prob (J-statistic) 0.52
evidence since it also approaches the role that the expectations of
Rank 7
entrepreneurs play as a monetary transmission channel. Moreover,
the study highlighted the importance of credibility to the effort of Author's estimates. Note: Marginal significance levels: *** denotes 0.01, ** denotes 0.05
and * denotes 0.1.
monetary policy implemented and how this affects the expectations
of entrepreneurs and investment decisions through the monetary

transmission mechanisms. Regarding the importance of credibility

for the expectations of entrepreneurs, and the influence of monetary
Table 11
OLS and GMM estimates: GFCF and transmission channels. policy on entrepreneurs' expectations (IECI) through the monetary
transmission, it is worth noting that the expectations of entrepre-
Dependent variable: GFCF
neurs react as expected to changes in real interest rate and asset
Explanatory variable OLS estimate GMM estimate prices, i.e., entrepreneurs become optimistic when the real interest
GFCFols GFCFgmm rate decreases and the value of companies listed at Bovespa increases.
Constant 81.156*** 78.871***
In addition, inflation targeting credibility positively affects the expecta-
(16.114) (11.456) tions of entrepreneurs since it reflects monetary stability, which is
[5.036] [6.884] crucial for the creation of a favorable business environment.
IECI-1 0.358** 0.433***
(0.174) (0.124)
[2.049] [3.476]
REAL_IR-2 −1.112*** −0.824**
(0.370) (0.302) Table 12b
[−3.000] [−2.725] OLS and GMM estimates.
CREDIT_GDP-1 1.478*** 1.951***
Dependent variable: empl_eap
(0.260) (0.198)
[5.678] [9.842] Explanatory variable OLS estimate GMM estimate
EXCH_RATE-2 −10.768*** −15.028***
empl_eap empl_eap
(3.031) (1.669)
[−3.552] [−9.003] Constant 56.669*** 56.697***
VC_BOVESPA 0.0000112*** 0.00000542*** (0.864) (0.839)
(0.00000226) (0.00000153) [65.557] [67.576]
[4.956] [3.552] gfcf 0.088*** 0.087***
dummy_sub −12.891** −17.415*** (0.005) (0.005)
(5.866) (2.904) [15.093] [16.177]
[−2.197] [−5.995] R2 0.90 0.88
R2 0.97 0.97 Adjusted R2 0.89 0.88
Adjusted R2 0.96 0.96 F-statistic 331.76
F-statistic 182.44 Prob (F-statistic) 0.00
Prob (F-statistic) 0.00 Ramsey (1); p-value 0.05
Ramsey (lags: 2); p-value 0.18 Ramsey (2); p-value 0.002
Serial correlation LM Test (1); p-value 0.22 Serial correlation LM Test (1); p-value 0.00
Serial correlation LM Test (2); p-value 0.26 Serial correlation LM Test (2); p-value 0.00
Heteroskedasticity (ARCH) (1) 0.003 Heteroskedasticity (ARCH) (1) 0.00
Heteroskedasticity (ARCH) (2) 0.014 Heteroskedasticity (ARCH) (2) 0.00
Normality (Jarque-Bera); probability 0.40 Normality (Jarque-Bera); probability 0.64
J-statistic 5.26 J-statistic 6.35
Prob (J-statistic) 0.81 Prob (J-statistic) 0.70
Rank 16 Rank 11

Author's estimates. NOTE: the reported t-statistics in the OLS estimates are based Author's estimates. NOTE: the reported t-statistics in the OLS estimates are based
on the estimator of White (1980), which is consistent in the presence of on the estimator of White (1980), which is consistent in the presence of
heteroskedasticity. Marginal significance levels: *** denotes 0.01, ** denotes 0.05 and heteroskedasticity. Marginal significance levels: *** denotes 0.01, ** denotes 0.05 and
* denotes 0.1. * denotes 0.1.
682 G.C. Montes / Economic Modelling 30 (2013) 670–684

The study found evidence that the maturing of the regime of infla- An important implication of this study is that, in inflation targeting
tion targeting measured by the gains in terms of credibility has repre- emerging economies, such as that of Brazil, following a committed mon-
sented an important aspect for the achievement of a more stable etary policy to price stability which increases the credibility of the re-
macroeconomic environment, which has presented lower levels of gime of inflation targeting and promoting macroeconomic stability
interest rate, currency appreciation, credit growth, better perfor- represents a good strategy for improving the transmission mechanisms
mance of capital markets and better business expectations. All these of monetary policy, thus stimulating investments and employment.
aspects have been of great importance to the rise of investments in What the findings do indicate is that developing credibility is crucial
the Brazilian economy and, as a consequence, for the development for emerging economies that are trying to grow, but with inflation
of the country and employment creation. being kept under control.

Appendix A

Table A.1
VAR residual serial correlation LM tests.

Lags LM-Stat Prob

1 105.610 0.001
2 81.220 0.072
3 55.105 0.778

Table A.2
VAR residual heteroskedasticity tests.

Joint test:

Chi-sq df Prob.

595.6702 576 0.28

Table A.3
Normality.

Component Jarque-Bera df Prob.

1 0.507893 2 0.7757
2 1.305901 2 0.5205
3 0.966833 2 0.6167
4 1.155825 2 0.5611
5 0.54347 2 0.7621
6 0.327534 2 0.8489
7 0.278127 2 0.8702
8 0.889627 2 0.6409
Joint 5.975209 16 0.9884
 G.C. Montes / Economic Modelling 30 (2013) 670–684 683

Fig. A.1. VECM.

Table A.4
VAR order.

VAR order SC (Schwarz criterion)

With c Without c

0 5.721
1 2.650 3.18
2 2.405 2.494
3 2.427 2.55

Table A.5
Johansen cointegration test.

Trend assumption: No deterministic trend (restricted constant)

Hypothesized no. of CE(s) Eigenvalue Trace statistic 0.05 critical value Prob.⁎

None 0.307681 16.06956 20.26184 0.1712

At most 1 0.064434 2.464345 9.164546 0.6846

Trace test indicates no cointegration at the 0.05 level.

⁎ MacKinnon–Haug–Michelis (1999) p-values.
684 G.C. Montes / Economic Modelling 30 (2013) 670–684

Table A.6
VAR order.

VAR order SC (Schwarz criterion)

With c Without c

0 12.003
1 8.289 8.266
2 8.240 8.34
3 8.417 8.47

Table A.7
Johansen cointegration test.

Trend assumption: No deterministic trend (restricted constant)

Hypothesized no. of CE(s) Eigenvalue Trace statistic 0.05 critical value Prob.⁎

None 0.277268 15.56944 20.26184 0.1955

At most 1 0.091608 3.554923 9.164546 0.4824

Trace test indicates no cointegration at the 0.05 level.

⁎ MacKinnon–Haug–Michelis (1999) p-values.

References Johansen, S., 1991. Estimation and hypothesis testing of cointegration vectors in Gaussian
vector autoregressive models. Econometrica 59 (6), 1551–1580.
Akemann, M., Kanczuk, F., 2005. Sovereign default and the sustainability interest rate Johnston, J., 1984. Econometric Methods, 3rd ed. McGraw-Hill Book Co, Singapore.
risk effect. Journal of Development Economics 76, 53–69. Koop, G., Pesaran, M.H., Potter, S.M., 1996. Impulse response analysis in non-linear
Amano, R., Coletti, D., Macklem, T., 1999. Monetary rules when economic behaviour multivariate models. Journal of Econometrics 74 (1), 119–147.
changes. Bank of Canada Working Paper n. 99–8. Kydland, F.E., Prescott, E.C., 1977. Rules rather than discretion: the inconsistency of
Ando, A., Modigliani, F., 1963. The life cycle hypothesis of saving: aggregate implications optimal plans. Journal of Political Economy 85 (3), 473–492.
and tests. The American Economic Review 53 (1), 55–84. Lutkenpohl, H., 1991. Introduction to Multiple Time Series Analysis. Springer, Berlin.
Auel, M.C., de Mendonça, H.F., 2011. Macroeconomic relevance of credit channels: evidence MacKinnon, J.G., Haug, A.A., Michelis, L., 1999. Numerical distribution functions of like-
from an emerging economy under inflation targeting. Economic Modelling 28, 965–979. lihood ratio tests for cointegration. Journal of Applied Econometrics 14 (5).
Blanchard, O., 2005. Fiscal dominance and inflation targeting: lessons from Brazil. In: Minella, A., Souza-Sobrinho, N.F., 2009. Monetary channels in Brazil through the lens of
Giavazzi, F., Goldfajn, I., Herrera, S. (Eds.), Inflation Targeting, Debt and the Brazil- a semi-structural model. Central Bank of Brazil Working Paper Series, n. 181, April.
ian Experience, 1999 to 2003. MIT Press. Minella, A., de Freitas, P.S., Goldfajn, I., Muinhos, M.K., 2003. Inflation targeting in Brazil:
Calderón, C., Duncan, R., Schmidt-Hebbel, K., 2004. The role of credibility in the cyclical constructing credibility under exchange rate volatility. Journal of International
properties of macroeconomic policies in emerging economies. Review of World Money and Finance 22 (7), 1015–1040.
Economics 140 (4), 613–633. Mishkin, F.S., 1995. Symposium on the monetary transmission mechanism. Journal of
Catão, L.A.V., Pagan, A., 2009. The credit channel and monetary transmission in Brazil Economic Perspectives 9 (4), 3–10.
and Chile: a structured VAR approach. National Centre for Econometric Research, Mishkin, F.S., 2007. Monetary Policy Strategy. The MIT Press, Cambridge, Massachusetts.
NCER Working Paper Series 53, March. Mitchell, J., 2000. The importance of long run structure for impulse response analysis in
Catão, L.A.V., Laxton, D., Pagan, A., 2008. Monetary transmission in an emerging targeter: VAR models. NIESR Discussion Papers, n.172.
the case of Brazil. International Monetary Fund, IMF Working Papers: 08/191. Mukherjee, S., Bhattacharya, R., 2011. Inflation targeting and monetary policy transmis-
Christiano, L.J., Eichenbaum, M., Evans, C.L., 1999. Monetary policy shocks: what have we sion mechanisms in emerging market economies. IMF Working Paper (WP/11/229),
learned and to what end?, In: Taylor, J.B., Woodford, M. (Eds.), edition 1. Handbook of October.
Macroeconomics, vol. 1. Elsevier, pp. 65–148. chapter 2. Nahon, B.F., Meurer, R., 2009. Measuring Brazilian Central Bank credibility under infla-
Coelho, C.A., de Mello, J.M.P., Garcia, M.G.P., 2010. Identifying the bank lending channel tion targeting. International Research Journal of Finance and Economics 27, 72–81.
in Brazil through data frequency. Economía: Journal of the Latin American and Newey, W.K., West, K.D., 1987. A simple, positive semidefinite, heteroskedasticity and
Caribbean Economic Association 10 (2), 47–79. autocorrelation consistent covariance matrix. Econometrica 55 (3), 703–708.
Cragg, J.G., 1983. More efficient estimation in the presence of heteroscedasticity of Peersman, G., Smets, F., 2003. The monetary transmission mechanism in the euro area:
unknown form. Econometrica 51 (3), 751–763. evidence from VAR analysis. In: Angeloni, Kashyap, Mojon (Eds.), Monetary Policy
de Mello, L., Pisu, M., 2010. The bank lending channel of monetary transmission in Brazil: Transmission in the Euro Area. Cambridge University Press, pp. 56–74 (chapter 2).
a VECM approach. The Quarterly Review of Economics and Finance 50 (1), 50–60. Pesaran, M.H., Shin, Y., 1998. Generalized impulse response analysis in linear multivariate
de Mendonça, H.F., 2007a. Towards credibility from inflation targeting: the Brazilian models. Economics Letters 58 (1), 17–29.
experience. Applied Economics 39 (20), 2599–2615. Phillips, P.C.B., 1998. Impulse response and forecast error variance asymptotics in
de Mendonça, H.F., 2007b. Empirical evidence from fourteen countries with explicit nonstationary VARs. Journal of Econometrics 83, 21–56.
inflation targeting. Applied Economics Letters 14 (8), 573–576. Sargent, T., Wallace, N., 1981. Some unpleasant monetarist arithmetic. Federal Reserve
de Mendonça, H.F., 2009. Output-inflation and unemployment-inflation trade-offs under Bank of Minneapolis Quarterly Review 5 (3), 1–19.
inflation targeting: evidence from Brazil. Journal of Economic Studies 36 (1), 66–82. Sims, C., 1980. Macroeconomics and reality. Econometrica 48 (1), 1–48.
de Mendonça, H.F., de Guimarães e Souza, G.J.G., 2009. Inflation targeting credibility and Tobin, J., 1969. A general equilibrium approach to monetary theory. Journal of Money,
reputation: the consequences for the interest rate. Economic Modelling 26, 1228–1238. Credit, and Banking 1 (1), 15–29.
de Mendonça, H.F., Lima, T.R.V., 2011. Macroeconomic determinants of investment Tomazzia, E.C., Meurer, R., 2009. O mecanismo de transmissão da política monetária no
under inflation targeting: empirical evidence from the Brazilian economy. Latin Brasil: uma análise em VAR por setor industrial. Economia Aplicada 13 (4),
American Business Review 12 (1), 25–38. 371–398.
Gonçalves, C.E., Guimarães, B., 2011. Monetary policy, default risk and the exchange White, H., 1980. A heteroskedasticity-consistent covariance matrix estimator and a
rate. Revista Brasileira de Economia 65 (1), 33–45. direct test for heteroskedasticity. Econometrica 48 (4), 817–838.
Hammond, G., 2011. State of the art of inflation targeting. Bank of England, CCBS Wooldridge, J.M., 2001. Applications of generalized method of moments estimation. Journal
Handbook No. 29 — February. of Economic Perspectives 15 (4), 87–100.
Hansen, L.P., 1982. Large sample properties of generalized method of moments estimators.
Econometrica 50 (4), 1029–1054.