Mexico's Monetary Policy Communication and Money Markets

Central bank communication is becoming a key aspect of monetary policy. How much financial markets listen and, possibly, understand Banco de Mexico's communication on its monetary policy stance should be a key consideration for the central bank to further modernize its monetary policy toolkit. In this paper, we tackle this issue empirically by using our own index of the tone of communication based on Banco de Mexico's speeches and statements and find that money markets do not only listen but they also understand the stance of monetary policy conveyed in the central bank's words. Regarding the ability to listen, first, we find that both the volatility and volume in the money market rates change right after communication from Banco de Mexico's governing body. As for the markets' understanding, we find a statistically significant rise in money market interbank rates the more hawkish communication is. All in all, our results show strong evidence of effective oral and written communication from the Central Bank towards Mexico's money markets.


Introduction
Central bank communication became topical, due the liberalization of financial markets first in developed markets and, only later, in emerging ones. Revived interest in such communication was triggered by the recent widespread attempts by central banks to provide "forward guidance".
In the 1990s, numerous central banks in OECD countries started improving their communication, using different means as a function of their target audience, be it the public or financial markets. With respect to the latter, central banks now publish their own assessment of the economic outlook and even hints to their future monetary policy action (BIS, 2009;Filardo et al., 2008). Academic research (surveyed by Blinder et al. 2008) has provided increasing evidence that communication represents a powerful tool for central banks to conduct a more predictable monetary policy, the more so the more developed the financial system where such monetary policy is conducted. The rationale for such a role of communication lies in the final goal of central bank communication: managing financial markets' expectations, which is easier with forward-looking financial markets (Garcia-Herrero and Remolona, 2008). Accordingly, central bank communication gets closer to an instrument to conduct monetary policy than a proof of transparency. Thus Svensson (2004) emphasized that, "monetary policy is to a large extent the management of expectations." As an OECD member, Mexico's monetary policy framework is gradually getting closer to that of its counterparts. This is not only the case for the Banco de Mexico's instruments but also for its communication. The Central Bank of Mexico (Banco de Mexico and Banxico from now on) publishes statements summarizing the meetings of its Monetary Policy Committee (MPC). Its governing body has also become much more active in delivering speeches, which are focused on monetary policy issues and the outlook of the Mexican economy.
Both Banxico´s statements and speeches point to a much more transparent central bank but they are not enough to assess how effective the Bank is in communicating its messages to the financial markets. In fact, a central bank can be extremely transparent but markets may not listen to it. It is also conceivable that markets do listen but they do not understand the message. A major implication is that the ability of a central bank to communicate should be measured by the reaction of money and financial markets to central-bank messages in the intended direction.
The objective of this paper is to assess whether Mexico's money markets "listen" to Banxico's messages and, possibly, whether they "understand" them. We interpret "listening" as it is generally the case in the literature of central bank communication, namely as the central bank's ability to "move" the markets when making a certain verbal or written statement. A standard way to measure this phenomenon is to assess empirically whether the volatility of interbank rates changes right after a speech from the Central Bank's governing body or the publication of a written statement by its MPC. We also consider the effects of the central bank's words on trade volume in the repo market since this may represent an indirect conduit to volatility, which has been all but neglected in the communication literature.
The present study contains three innovative features: First, to the best of our knowledge, it constructs the first comprehensive measure we are aware of Banxico´s communication on the monetary policy stance. Such a measure includes both written statements and speeches by Mexican central bank's officials. More importantly, it codes the statements and speeches according to their indication of the monetary policy stance (dovish, neutral or hawkish). The second innovation of the paper is the use of high-frequency data to analyze the impact of the Banco de Mexico communication on the interbank money market, well beyond what can be found in the literature. In fact, the paper not only assesses the direct effect of Banco de Mexico communication but also uses microstructure data to analyze the indirect effects of Banco de Mexico communication through interbank trading volume.
The paper is structured as follows. Section 2 reviews the related literature and describes the paper's objective in some more detail. Section 3 presents the methodology as well as the data compilation, with particular attention to our measure of central-bank communication. It, finally, presents some stylized facts on Mexico's money markets. Section 4 shows our empirical results and interpretations. Finally Section 5 draws some policy conclusions.
When studying the impact of central bank communication, it is important to determine the market on which to measure the impact. The logical first step is to start by the upstream part of the transmission mechanism and focus on the effects of central-bank signals on money markets. From money markets, central bank words should eventually also affect financial markets, including the longer end of the yield curve and even stock markets. In our paper, we shall follow this logical order and focus on the impact of the Banco de Mexico communication on Mexico's money markets.
Beyond the choice of the market, it is also important to consider the different ways in which a certain market may be affected. As previously explained, a weak way to measure the impact of central bank communication would be its effect on volatility (what we call "listening") while a change in the right direction (i.e., in line with the tone of Working Paper May 2015 the central bank messages) would imply that the central bank is actually "understood". One more important issue is how changes in expectations triggered by communication translate into trades (i.e., volumes) and not only prices. Indeed, available literature, focusing on the U.S. Treasury bond market's response to FED announcements, stresses that the latter is composed of two distinct stages of adjustment (Fleming and Remolona, 1999). A short-lived first stage is a response to pure public information. The release of an announcement triggers a large and immediate (ultra-high frequency, such as intraday) price reaction which does not require trading, in as much as it is driven by inventory control. In a prolonged (lower frequency, such as daily) second stage, trading volume surges in a persistent way, due to a disagreement among investors about the implications for prices of the newly-released information. An interesting maintained hypothesis to be tested is, thus, whether communication impacts both the volume and the volatility in the interbank market.
Empirical analysis on the effectiveness of Banco de Mexico's communication on financial markets is scarce, and to our knowledge there are no analyses that incorporate the oral communication, namely the speeches. Related papers include López-Marmolejo (2013) and Herrerías and Gurrola (2012). López-Marmolejo built an index of monetary policy communication based on central bank written statements in order to analyze whether analysts and financial markets´ monetary policy expectations understand Banco de Mexico communications' tone. This study considers the period 2005-2011 and finds that Banco de Mexico's communication provides signs to the market that helps it forecast the policy interest rate and that both the financial markets and private economic analysts understand the communication's tone and incorporate such tone into their interest rates expectations. As a result, market expectations on interest rates are more consistent with Banxico's view, and are closer to their finally observed levels. Herrerías and Gurrola (2012) study the reaction of short-term interest rate futures to monetary policy announcements and show that, in a market dominated by institutional investors trading solely for hedging purposes, the interest rate futures prices are adjusted on announcement dates, not being fully incorporated in advance. They also find that the change to an interest rate operational target modifies the magnitude of the changes in the volatility, volume and prices of the futures contracts resulting from the announcement.
Although not directly related to communication, there are a number of papers on Mexico's monetary policy and interest rates worth reviewing as a way to better choose our control variables. For example, Cermeño et al. (2012) and Torres (2003) study the determinants of the monetary policy rate in Mexico, finding that inflation and output gaps are relevant determinants, and that interest rates are determined "more" by a forwardrather than in a backward-looking behavior. On the determinants of the term structure of interest rates, García-Verdú (2011) finds evidence that some properties of the term structure are associated with macroeconomic variables such as the monetary policy rate and economic activity.

Working Paper
May 2015

Methodology and Data
In order to assess empirically the impact of communication on Mexico's money markets we consider a number of issues. First, as previously mentioned, we differentiate between "listening" and "understanding", by studying separately the effects on volatility and mean interest rate changes. Second, we not only look at direct effects (measured by prices) but also indirect effects through volumes, which have usually been neglected in the communication literature. In addition we carefully choose our data to represent the most active and liquid segment of the Mexican money market, namely the overnight repo market.

Methodology
To evaluate the impact of Banxico communication on Mexico's money markets, we adopt an encompassing approach, using estimates of the mean and volatility equations of the repo overnight rate for the different aspects of communication. As a first step, we focus on whether money markets listen to the Banco de Mexico by assessing empirically whether central bank communication influences the volatility of financial variables, as in Connolly and Kohler (2004) and Ehrmann and Fratzscher (2007). The underlying hypothesis is that the volatility of asset returns should change in as far as there is newand relevantinformation being provided, everything else equal. A more debatable issue is whether volatility should increase or decrease. This very much depends on how much the new information surprises the markets or actually calms them. The latter may be more common if the situation prior to the central bank launching a message was very uncertain. Within the same framework, we also estimate the impact of communication on the volume of the overnight repo market, as an indirect channel affecting volatility.
Measuring volatility has always been a challenging task for financial economists. Previous literature modeling the impact of communication on volatility in the money market has mostly focused on conditional volatility models following a GARCH framework (Blinder et al. 2008). However this only considers the difference in the closing money market rates. In this paper, we go deeper in measuring volatility by using the logarithm of the price range, defined as the difference between the highest and lowest intraday interest rates. The range is a superior volatility proxy to absolute or square returns (Alizadeh, Brandt and Diebold, 2002; since it takes into account the intraday sample path. The range is also less subject to microstructure noise than realized volatility (as championed by Andersen et al. 2001). Furthermore, the logarithm of the range is a preferable measure to the simple range, in particular since it filters out outliers .
In a second step, we test whether money markets understand the Banxico's words, following Ehrmanm and Fratzscher (2007). This means that we assess whether our index of the change in the monetary policy stance, based on the Banxico words, affects mean repo rate changes in the intended direction. In other words, we investigate whether hawkish (dovish) communication led to an increase (decrease) in the repo rate. Such an exercise obviously introduces the researcher's subjective judgment and, therefore, a potential measurement error.
Beyond monetary policy communication, we need to control for the key variables which may affect Mexico's money market. The instruments used by the Mexican central bank, which constitute the deeds of monetary policy, are the most important controls since we aim to assess the complementary of words and deeds when conducting monetary policy. In addition, we also include relevant time-series controls, such as seasonality and ad-hoc events.

May 2015
Regarding monetary policy instruments, we include changes in the policy rate controlled by Banxico (PR), and net central bank auctions, which represent net liquidity withdrawal (). We control for central bank intervention in the foreign exchange market with a dummy on days of intervention (INTERV). We allow for calendar effects such as end-(WE) and beginning (WB) of-week effects. Finally, we control for key data releases, namely Mexican GDP and CPI announcements with a dummy (MACRO) which takes a value of one on days of GDP growth or inflation releases, and zero otherwise, and similarly for the dummy representing US macro announcements on GDP or CPI (USNEWS). The change in the Fed Funds rate was also considered as a control variable but it was not significant and statistical tests such as the Likelihood ratio and the Bayesian Schwarz criterion rejected its inclusion.
Based on all the issues above, we model the repo rate in two ways: the volatility equation (1) and the mean equation (5). Besides, we also estimate a volume (V) equation (4), to investigate the possibility of an indirect channel for central bank communication to affect the money market.
The volatility equation addresses whether money markets listen to the Banxico, using the log range of the overnight repo rate as a proxy for volatility. We model the heteroscedasticity of the residuals by using a GARCH model. Controlling for the above-mentioned determinants, the logarithm of the range of the repo rate may be explained as stated in equations (1) and (2) or (3) below: ln(range) t = a 0 + a 1 ln(range t-1 ) + a 2 APR t + a 3 ANETA t + a 4 WE t + a 5 WB t + a 6 COM t-j + a 7 MACRO t + a 9 USNEWS t + a 10 INTERV t + a 11 ln(V) t + (h t ) Where h denotes conditional variance, the  coefficient is for the ARCH term, and  the GARCH term, while  t is a unit-variance, serially uncorrelated, zero mean, i.i.d., error term. In addition, in the threshold GARCH model of Zakoian (1994), (equation 2), bad news (negative residual v) have a different effect (+) than good news (), while in the Exponential GARCH model (equation 3), a negative shock is allowed to exert a larger impact than a positive shock, as in Nelson's (1991). Finally, the letter A before some of the control variables stands for absolute value. As an example, APR stands for the absolute value of the change in the policy rate. We note COM the communication variable to be detailed later. Due to the ever-present non-normality in the residuals, we use the Generalized-Error Distribution suggested by Nelson (1991), which encompasses several other distributions depending on the value of the tail-thickness parameter.
An extensive series of work in financial economics provide evidence of a positive correlation between contemporaneous trading volume and volatility. This positive correlation was rationalized initially by Clark (1973), who put forward the Mixture of Distribution Hypothesis (MDH), with further refinements provided by Epps and Epps (1976), Tauchen and Pitts (1983) and Harris and Raviv (1993). The MDH assumes that asset price movement and trading volume are both determined by an unobservable flow of information. In line with the mixture of distribution hypothesis, we thus include daily volume as a potential determinant of volatility.
We now specify the model to estimate the impact of Banxico communication on the transaction volume of the repo market. In this way, we can investigate whether there is an additional indirect effect of communication on

Working Paper
May 2015 the repo market. We use the log of the daily volume of the overnight repo market (V) controlling for the abovementioned determinants as shown in equations (4) below together with (2) or (3): where the same notations and specification are used as in equation (1). Lastly, to assess whether money markets understand the Banxico, we model the intraday change in the repo rate (close minus open, noted id, as motivated below) -to be explained by Banxico words as well as deeds and other control variables-as shown in equation (5) below: with the same notations as in equations (1). A GARCH (equation 6) or an EGARCH (equation 3) model will be used since no evidence of a threshold effect was manifest in the conditional variance of the change in the repo rate.

Data
There are two potential problems when evaluating the effectiveness of central bank communication. One is a sample-selection problem and the other one is a potential measurement error. The former is more likely to occur when oral communication is measured by an intermediary, namely newswire reports and not at the source. While newswire are the source used in the existing literature (see Jansen and De Haan, 2005  Each speech is selected only once, on the basis of first recording. We construct three variables. The first is a dummy which takes the value of 1 on the day the speech is issued, and zero otherwise (ACOM). The second and third intend to characterize the speech based on its information content. Both variables take a higher value the more hawkish the speech and a lower value the more dovish the speech. Our directional measures of the monetary policy stance is a discrete variable with five possible values: 2 (very hawkish), 1 (hawkish), 0 (neutral), -1 (dovish) or -2 (very dovish). Overall we have 35 speeches with a non-zero coding. In order to construct an encompassing measure of communication, we code written quarterly reports, quasi-monthly decision announcements and their minutes by the Banxico in a similar way to speeches and include in our overall measure of communication (COM dummies) the 136 written statements. It did not prove possible to examine separately the separate effects of speeches and written statements since the latter frequently occur either on the same day or on an adjacent day to speeches. From Figure 4, it seems that the monetary policy communication measure follows to some extent the interest rate target. But analyzing this relationship requires further analysis which will be implemented in section 4.
As for the endogenous variable, namely the money market rate, we opt for the repo rate which does remain the benchmark for money market participants. We focus on the collateralized ('Buy Out') repo market, since the uncollateralized part is very narrow and only developed recently. The repo market involves maturities from overnight to more than three months. The relevance of the different maturities can be gauged by examining turnover on the collateralized repo market. As shown in Figure 1, the bulk of volume for such a repo market is concentrated in the low end of the maturity spectrum. In the repo operations, those collateralized by government debt represent around 75% and those collateralized by debt from banking, corporate and other institutions are around 25% in 2014. Accordingly we will focus our subsequent analysis only on the shortest maturity, namely the overnight, repo rate (excluding the observations during weekends and holidays). This will also ensure that our estimations do not suffer from an overlapping observation bias since the frequency of our data is also daily. Descriptive statistics for the repo rate, the daily range and volume over the full sample as well as subsamples are presented in Appendix 1,

Phases in recent monetary policy in Mexico
In the period September 1995 to January 2008 the Bank of Mexico's objective was banks' balances, also known as "short" or "el corto" in spanish. This mechanism aimed at achieving a certain balance of the aggregate current accounts that the banks hold with the central bank. It is called "short" to refer to a negative aggregate balance objective, and when implementing it, the central bank injects all the liquidity the system needs during the established period, but a share of it, the amount "short" is provided at punitive interest rates. In its first stage, from September 13, 1995 to April 9, 2003, the objective was set in terms of the accumulated balances at the end of month, while in the second stage, from April 10, 2003 to January 20, 2008, the objective became the accumulated balances of the banks at the end of the day. The implementation of monetary policy through a target rate for the overnight interbank operations began on January 21st, 2008. 1 Nevertheless, in the periods in which the "short" was used, there was an implicit rate observed, resulting from the decision on the movement in the balance of the current accounts of the banks with the Bank of Mexico, the funding rate. In addition, from August 26, 2005 the monetary policy announcements began including the rates the Bank of Mexico considered consistent with the "short" levels prevailing. Therefore, in the periods when the monetary rate was not explicitly fixed, but the balances were, the implicit rate resulting was the interbank offered rate or funding rate. Figure 4 shows the interbank offered rate under different operational mechanisms of monetary policy. As mentioned above, the communication variable is calculated beginning in 2005 because the communication policy improved considerably and has been more homogeneous from this year on, but also because from this same date the monetary rates have been more stable, which improves the ability of financial markets to predict the central bank moves.
For example, one sub-sample analyzed comprises the period when the operational mechanism was "el corto", from January 1st 2005 to January 20th, 2008, and another is the period when the operational mechanism became a target rate for the overnight interbank interest rate, from January 21st 2008 onwards. Also we consider sub-samples accordingly to the economic cycle. In particular, it is important to consider the 2008 crisis that generated a strong recession in 2009 (-4.7% for annual GDP growth) and financial volatility that generated high economic and financial uncertainty. In terms of GDP growth, the crisis began in the third quarter of 2008 with a quarterly contraction (-0.1%). Thus, we separate the sample before and after the 2008 crisis i.e. 2005 to Sept-08 and Jan-2009 onwards.
The stance implied by communication ( Figure 5) over these 5 sub-samples matches developments in Mexican inflation and economic growth as plotted in Figure 6. In our modeling of the impact of communication on repo rates, we should leave out overnight (open minus previous day' close) changes, and focus on intraday (close minus same day open) changes since changes in the policy rates are announced at 9:00 am, i.e. half an hour after the market opens, and thus cannot have influenced the overnight repo changes, which took place earlier.

Effects of Banxico communication on money markets
In this section we show the results of our model estimations for the two key questions addressed in this paper. For the former, whether monetary markets listen to the Banxico words, we investigate both the direct and indirect channel of influence as already explained.

Do money markets listen to the Banxico? Repo volatility and volume
We test the hypothesis that the volatility of repo rate (from equation 1 above) may move in a statistically significant way right after Banxico communication. As far as the expected sign is concerned, existing literature has long preferred to rationalize increases in volatility in response to communication as a confirmation that markets listen to central-bank communication. The underlying justification was that messages from central bankers convey new information and that the latter tends to move markets. Another view, expressed more recently, would rather interpret a fall in volatility as an indication that central bank communication is able to calm markets. The explanation of the latter phenomenon, suggested by Geraats (2002), relies on the presence of a degree of uncertainty before the central bank speaks and also the clarity of its speech. In her view central banks may either confuse markets (increasing volatility) or clarify the situation (reducing it). Table 1 (Panel A) confirm the hypothesis that money markets listen to the Banxico communication in as far as the daily log range 2 changes in a very significant way right after communication is delivered over all sub-samples from January 2005 to August 2013, which indicates that the Banxico words do convey information that markets perceive as relevant. This is in line with Remolona and Fleming (1999) and Ehrmann and Fratzscher (2007). The sign of the impact on volatility is negative in all cases.

Results in
As for the control variables (Appendix 3, Table A.3.1), the volatility of the REPO rate moves very significantly only in the most recent subsample when the money rate is changed. Net auctions are effective in influencing volatility, both pre and post the 2008 crisis. However, a change in sign seems to have occurred during 2008, since the effect was negative up to January 2008 and became positive thereafter. Volume raises volatility, but significantly only after May 2006.   Table 1), and this effect was often significant. This is in line with the negative response of volatility to communication detected above. By contrast, after Lehman volume seems to have become insensitive to communication. When including the control variables, the range seldom has a significant impact on the volume of the REPO market, 3 as shown in panel B of Table 1. This may be due to the fact that some of these variables are themselves determinants of the range (Appendix 3, Table A.3.2). Among the monetary policy instruments used as control variables, central bank auctions have had an significant effect only since January 2009. This effect is actually negative on the volume of the REPO market (reinforcing the direct effect found above for volatility). By contrast, the most significant effect of changes in the money rate is apparent over the 2005-Sept. 2008 period, while the effect is only weak thereafter. The significant effect of Macro news appears in the most recent sub-sample. Also a dummy was included in the estimation of the Range in order to control for an outlier on 24rd March 2008. This outlier seems to be the result of an inflation reading above market expectations and an important liquidity reduction by the central bank (almost five times the average for the all sample).

Do money markets understand the Banxico?
As a second step, we regress the mean of repo rate changes (close minus open) on Banxico communication (following equation (5) above). We focus again on the repo with most liquid maturity, the overnight repo. We use the five-category variable constructed to summarize the directional intent and intensity of communication (COM).    (6). Close minus open repo in all cases in the mean equation. The COM dummy for communication and the SPEECH dummy for speeches only has a FIVE-pronged classification as very dovish (-2), dovish (-1), neutral (0), hawkish (+1), and very hawkish (+2). These dummies are included alternatively in the equation. GARCH estimation with Generalized Error Distribution. . Only significant effects are reported: *** significant at the 1% on the basis of the z-statistics.

Effect of Banxico communication on intraday change in REPO-rate
We find highly-significant changes in the repo rates after the delivery of oral and written communication ( Table 2, row 1). In other words, the more hawkish the communication by Mexico's central bankers, the larger tends to be the immediate rise in the REPO rate during the day. In fact, the result holds for all subsamples, except for an initial negative influence. These results show that Mexico's money market does pay good attention to the Banxico's communication and understand its messages.
When looking at monetary policy instruments, the common feature is the contrast between the pre-and post-January 2009 subsamples. Even though changes in the money rate have a consistent positive impact across Working Paper May 2015 subsamples, that effect almost doubles in the second subsample, starting in January 2009. However, such a rise occurs even earlier, as testified by the sample starting in May 2006. With respect to auctions, the contrast between subsamples is of a different nature, since the effect on the repo rate is negative in the first subsample, but becomes positive also after May 2006. Recall NETA corresponds to liquidity withdrawal, computed as Deposit auction minus Liquidity auction, where Deposit auction reduces liquidity and Liquidity auction increases it. We thus expect a Net Liquidity Withdrawal to increase rates and therefore a positive coefficient. A positive coefficient multiplied by a negative NETA variable (i.e. more liquidity = Liquidity auction larger than the Deposit auction) should generate lower rates. Therefore, the positive signs after May 2006 seems reasonable and comprises most of the sample (only 2005 is left out) including the most recent period, the period when inflation began staying more regularly below 4% (the upper part of the central bank´s target) and the rates took a downward trend (moving away permanently from the levels around 10%).
In the Repo change estimation dummies were also included to account for a few outliers. These outliers are

Robustness
One could argue that we should control not for the change in the policy rate but rather its surprise component, i.e. the excess of the actual versus expected policy rate. Accordingly, we reestimate our mean model substituting the latter for the former control variable. As shown in Panel B of Table 2, the impact of communication on the intraday change in the repo rate is unchanged, except for the welcome fact that the surprising negative effect in the initial subperiod (2005)(2006), with the change in the policy rate as control variable, is replaced by, a much more intuitive, positive effect with the surprise as control.
Over the second major subsample (

Conclusions
This paper provides the first evidence of the ability of Banxico communication to steer money markets. We focus on the most liquid segment of the REPO market, the one-day maturity and cover an 8 year period (from early 2005 to the summer of 2013). For Banxico communication we build an indicator of the tone of monetary policy based on the speeches delivered by Banxico senior officials but we also control for MPC written statements.
As a first step, we provide evidence that money market operators do listen to Banxico, since its communication generates changes in market volatility (measured by the log range) in a significant way. This result is similar to those previously found for the U.S., the U.K., and some other O.E.C.D. countries (Kohn and Sack, 2004;or Connolly and Kohler, 2004;and Reeves and Savicki, 2007). Going beyond that literature, we also quantify the influence of Banxico communication on the money market transaction volume. We conclude that volume also responds significantly to communication, strengthening our a priori that Mexican money markets do listen to the central bank.
As a second step we address a more elaborate question, namely whether money markets understand Banxico's words. In particular we test empirically whether money markets perceive the stance of monetary policy which the Banxico intends to convey in its speeches or written statements.. Our findings imply that intraday REPO rate changes generally respond to Banxico's communication in the expected direction, namely more hawkish communication leads to increases in money market rates. Such finding is robust when we control either for relevant factors such as the actual changes in monetary policy (i.e., in the policy rate) and for macroeconomic news both in Mexico and in the US.
In sum, we conclude that Banxico words (and not only deeds) have a significant impact on Mexico's money market both in direct and indirect ways. Furthermore, Mexican money market participants not only react to the Banxico words but also appear to understand the directional intent of Banxico statements and speeches. These findings can be interpreted as the Banxico being a very predictable central bank since they are similar to those for peer central banks in larger O.E.C.D. economies, such as the U.S., the U.K. and the euro area.

Working Paper
May 2015 Appendix 1: Descriptive statistics and unit-root tests

Log Range
MZt -5.08*** -4.91*** -3.68*** -4.49*** -12.1*** The growth rate of economic activity in Mexico has slowed. In this environment, the output gap has remained near zero. In addition, the labor market has continued showing a recovery. However, downside risks for the Mexican economy persist as a reflection of macroeconomic imbalances.

MZt
In recent months, headline inflation in Mexico continued declining significantly to close 2012 below 4 percent, as the Bank of Mexico anticipated. Annual core inflation fell to settle at 2.90 percent in December, a historically low level. It is expected that in 2013, annual headline inflation will fall below the levels recorded in 2012 and close to the permanent objective of 3 percent. In the case of annual core inflation it is anticipated that it will locate below 3 percent. In sum, it is considered that the balance of risks for inflation has improved.
If the environment described consolidates it could be advisable to reduce the objective of the overnight interbank interest rate to facilitate the adjustment of the economy to a state of lower economic growth and lower inflation. In sum, downside risks to economic activity in Mexico prevail, derived mainly from the possibility that the recent slowdown in the U.S. economy could intensify.
Annual headline inflation in Mexico in March and the first half of April registered a major hike following a higher incidence of the non-core component. This increase was influenced by effects of the CPI base of comparison, as the Bank of Mexico had anticipated, and by the occurrence of clearly identified supply shocks.
The process of price formation has not been affected and remains consistent with the convergence of inflation to the target of 3 percent. This is evident in the performance of core inflation, which has remained very close to 3 percent as a result of two factors. During the first quarter, economic activity in Mexico did not suffer significantly from the weakening of the U.S. economy.
The recent dynamics of inflation are worrying because it responds to the stress of the same external pressures that have been commented in the past. So, like most central banks, the Bank of Mexico has had to revise upwards its inflation projections. It is also likely that some disparities between certain domestic prices and their world benchmarks will gradually correct. Mexico shows no inflationary pressures on the demand side. Even an eventual weakening of aggregate expenditure could be a factor that would mitigate upward pressures. Inflation expectations over the medium term remain well anchored, and contamination of the prices not related with the hike of commodity prices has been limited. However, the balance of risks to inflation has worsened. The path of inflation for the rest of the year and early 2009 could be somewhat higher than expected in the latest Inflation Report. Coded: 2   *** Z-statistic significant at 1% level (** 5%; * 10%). This corresponds to the estimation of equation (1) and (2) or (3) in the text with corresponding definitions of variables and parameters.