Itaú BBA - Uncovering Latam central bank’s reaction functions

Macro Vision

< Back

Uncovering Latam central bank’s reaction functions

April 6, 2017

Taylor Rules predict more accommodative monetary conditions in the region

Please open the attached pdf to read the full report .

We estimate reaction functions (“Taylor Rules”) for the central banks of Mexico, Chile, Colombia and Peru. These Taylor Rules not only provide evidence on whether current monetary policy stances are being tight or loose relatively to historical reaction, but also predict policy rate paths under our baseline scenarios for inflation and economic activity.

According to our results, the Central Bank of Mexico is currently being more conservative, or inflation-risk averse, than in the past, with the policy rate set above the level predicted by the Taylor Rule.  The Taylor Rules predict rate cuts for Colombia, Chile, and Peru in 2017, in line with our own forecasts. Monetary conditions would turn significantly more accommodative in Colombia, where real rates are still at high levels. 

In this report, we discuss the policy rate forecasts obtained from Taylor-rule models that we have constructed for Mexico, Colombia, Chile and Peru, assuming growth and inflation evolves in-line with our expectations.Our approach consists of regressing the policy rate (actually, its deviation from a neutral level) on a forward-looking deviation from the inflation target and the output gap. The first explanatory variable is defined as 12-month inflation expectations minus the inflation target set by the country’s central bank (when forecasting policy rates, we will assume that market expectations follow our own projections for inflation). The second variable is calculated as the percentage difference between the 12-month average monthly GDP proxy and its trend (which represents potential output). Peru has a somewhat different reaction function because the dollarization of its economy (almost 30% of credit in Peru is denominated in U.S. dollars) makes exchange-rate behavior a key variable in policy rate decisions. So besides inflation and activity variables, we also use the 30-day rolling average of the implicit yield in three-month USDPEN forward contracts (the so-called “forward points”) as an explanatory variable. For Mexico, we use two equations: a conventional Taylor Rule, with inflation expectations and the output gap as explanatory variables, and another equation (using a more recent sample) which adds the fed funds rate as one of the explanatory variables, consistent with the communication adopted by Banxico since the Fed’s liftoff debate begun.

Mechanically, the equation is set up such that when both the output gap and the deviation from the inflation target are zero, then the policy rate is equal to the neutral nominal rate (defined as the neutral real policy rate plus the inflation target). We calculate the neutral real rate as the trend (Hodrick-Prescott filtered) of the real monetary policy rate (which is defined as the monetary policy rate minus 12-month inflation expectations). When we make forecasts, we will assume that neutral rates converge to their average historical level by the end of 2018. All the models have high explanatory power (R2 of around 90%), and all individual coefficients show the expected signs and are statistically significant at the 99% confidence level.

The intuition behind Taylor rules is straightforward: policy rates are set according to the levels of (expected) inflation and slack in the economy. So, when inflation expectations are rising/falling, the central bank will raise/cut interest rates (given the amount of slack in the economy). Alternatively, if growth is above/below potential, the central bank will raise/cut interest rates (given the level of inflation expectations). All the countries covered in our analysis have de jure inflation mandates, but their central banks are not insensitive about economic growth, as growth is a leading indicator of inflation and monetary policy acts on prices with lags.

For Mexico, our two Taylor-rule models (with and without the Fed Funds rate as an explanatory variable) indicate the central bank is currently being more conservative than in the past – that is, the policy rate is above the level predicted by the Taylor rule. Also, lower inflation expectations ahead and weaker GDP growth make the conventional Taylor-rule model (that is, without using the Fed Funds rate) predict an even lower policy rate in the short-term. However, we still see some more tightening (we expect two extra 25-bp rate increases this year), before the central bank starts to “normalize” the policy rate in 2018. In fact, if we include the fed funds rate as an additional explanatory variable (for a shorter sample, January 2009 to March 2017), the gap between Mexico’s actual policy rate and its predicted value is smaller. Moreover, if we assume future Fed rate hikes (75-bp more in 2017, and 100-bps in 2018), the predicted policy rate path actually trends upward, much closer to our actual forecast for year-end 2018 (6.5%). This “follow-the-fed” behavior, however, might be atypical. Once the uncertainty affecting the Mexican economy dissipates, it is likely that Banxico will decouple from the U.S. monetary policy (especially considering that Mexico already runs a tight monetary policy), which is consistent with Governor Carstens recent rhetoric.  

Our Taylor-rule model predicts that monetary policy will be loosened further in Colombia – in line with our call – as the economy grows at below-potential rates and inflation gradually converges to the target. In a context of high real interest rates, the negative output gap seems to be widening, and Banrep is starting to look through the transitory upward pressure on inflation stemming from tax hikes. In fact, the most recent monetary policy decision statement already indicates that the board is shifting its focus to meeting the 3% target in 2018, rather than the previous goal of bringing inflation to below 4% (the upper bound of the range around the target) this year.

For Chile, our Taylor-rule model suggests that there is room for bringing the policy rate down to our already below-consensus 2.5% call. However, it also predicts that a normalization of policy might start sooner than we currently expect as inflation expectations rise back to the 3% target and considering the current policy rate is already significantly below the neutral level.

With a full-fledged “coastal El Niño” baring its teeth, the central bank of Peru will likely cut rates in 2017. Flooding and landslides have destroyed agricultural crops and transportation infrastructure, causing a spike in food prices which made annual headline inflation shot up in March (to 4%, from 3.3% in February). We believe that the BCRP will stop short of cutting rates at its upcoming meetings (at least those in April and May) out of concern that the food price shock might have second-round effects on inflation expectations. Nevertheless, during an interview following the presentation of Q1’s Inflation Report, the central bank governor indicated that rate cuts are likely. In fact, reserve requirements for local currency, usually used as a first line of defense, were lowered to 5% (from 6%) in April. In this context, we foresee lower rates down the road. This is consistent with the results of our Taylor rule. On the exchange-rate side, volatility is currently moderate and will probably not make much of a difference in the BCRP’s short-term policy decisions.

Overall, the results of our estimations point to lower interest rates in the four countries. Yet Mexico’s central bank seems to be taking no chances at the current juncture: it is maintaining a much higher reference rate than what our Taylor-rule models suggest – perhaps as reinsurance against the risk of further MXN depreciation. The policy rates in Chile and Peru are currently only a bit higher than what is predicted by their respective Taylor-rule models, while Colombia’s is in line with our model output.


João Pedro Resende

Alexander Müller



Please open the attached pdf to read the full report .

< Back