In the previous post I discussed what the SARB intends to do with interest rates. The MPR 2015 documented hinted at an increase in many places. This is exactly what the SARB did - it increased the repo rate by 25 basis points to 6 percent. It is strange considering that the latest inflation is 4.7%...
The hike could only mean that the SARB anticipates higher inflation from various underlying pressures and that it thinks it right to act now (or the SARB is simply importing foreign monetary policy decisions...Brazil!). It should be interesting to see how the economy responds - watch out mortgage defaults, debt service costs and household consumption expenditure.
Steps to a new world
Sunday, 26 July 2015
Wednesday, 22 July 2015
Tickle tickle - how monetary policy announcements could go wrong
Interest rate announcements and forward guidance
In this post I look at a potential problem central bankers face. Hint: it has to do with forward guidance - the way a central bank communicates its interest rate decisions to us. I picked the South African Reserve Bank's (SARB) latest (April 2015) Monetary Policy Review (MPR) to highlight these issues. The MPR, I think, tells us that everyone should be aware of a rate tightening cycle. Is it nice of them to "warn" us of an impending rate hike? Does it intend to adjust our expectations? Or can it backfire? Let's start by looking at some of its communication tools:
The fan chart:
Fan charts look pretty cool. Fan charts are sometimes used as a forecasting tool and depicts various paths of a variable with a confidence interval. It looks something like the figure below (source: 2015 MPR, SARB). The dark line is the median of the forecast while the lighter areas represent increasing confidence intervals. An easy way to think of this is by inverting the figure and plot a bell shaped curve where the light areas measure the standard deviation.
Why would a central bank produce such a figure ? 1.) It acknowledges that models are only partial representations of reality and that significant forecast errors exist (inflation is forecast to be anywhere from 3% to about 11% in 2017); 2.) to communicate what it might possibly do with interest rates (only if the central bank targets inflation). The fan chart produced below is a split-normal distribution (i.e. it is not perfectly symmetrical on both sides of the median). In fact all the "risk" to the forecast seems to be on higher inflation. Depending on how seriously the SARB takes these forecasts, it could mean that is pricing in a high probability of an interest rate increase.
Announcements or hints
In the latest MPR from the SARB: "Underlying inflationary pressures are resilient and expectations have converged at the top of the target range. This makes a sustained breach of the inflation target more likely...Given all these factors, monetary policy remains in a tightening cycle". There might be a slight contradiction in this statement - if expectations have converged at the top of the target then why would we have a sustained breach of the target? Converged expectations mean anchored expectations (admittedly at the upper limit of the target). Some models allow for backward indexation in the hybrid Phillips curve (more on this later), which means that past inflation and future inflation affects current inflation.
Sounds like the SARB wants to increase interest rates...
This sounds to me like the SARB is seriously thinking of increasing interest rates. Is this what we call forward guidance? The SARB does not make an explicit statement regarding interest rates. The fact that the fan chart for inflation ranges between 3% and 11% suggests that there is huge monetary policy uncertainty (monetary policy uncertainty is a shock over and above actual changes to interest rates see this).
On the one hand the SARB in their models often target expected inflation as opposed to realised inflation. If they believe inflation in the future (say 18 months from now) is above the target level they might very well increase interest rates. However, if they announce that they are in a tightening cycle and consumers and firms believe them, then economic participants might decrease consumption. Especially when higher expected interest rates affect credit and investment decisions. In this case expected inflation should decrease. Thus, consumer and firm decisions are conditional not only future inflation but also future interest rates.
This makes monetary policy incredibly difficult and somewhat counteracts forward guidance. Since the information set that monetary policy makers face is too big to make useful numerical estimates of inflation, they often revert to simple models. The modelling team at a central bank will usually produce inflation and output forecasts and then the MPC will decide on the interest rate path. It would be interesting to do a study the deviations of the policy advice coming from these models from the final MPC decision.
How do we create such a figure?
With a model that captures economy-wide effects. The smallest of such a model includes an equation for interest rates (Taylor rule), for inflation (Phillips curve equation) and output (IS equation).
So what?
Here is the tricky part: Inflation has been pretty stable around the 6% mark for a while. This has happened despite many negative shocks hitting the economy. Interest rates have also remained pretty much at a similar level. Is this proof for the Neo-Fisherian school (inflation in the long-run follows the monetary policy rate)? Thus when interest rates go up, there might be a little fall in inflation, but then inflation increases. This could be further substantiated by looking at stationary path of inflation. It seems as though inflation is a mean reverting process, but it takes a long time to revert to its mean following a shock.
Using the equations specified above, I illustrate two features of monetary policy: 1.) what happens to the economy when the SARB increases interest rates temporarily where the shock is unanticipated and 2.) where there is a permanent shock while it is anticipated (people have perfect foresight; or the SARB guides the public's expectations).
Scenario 1 is depicted in Figure 1. The results seem pretty standard: An increase in the repo rate decreases inflation. However, Figure 2 shows that an increase in interest rates lead to an increase in inflation. Hmmm...why? This is the idea of the Fischer identity in most central bank models: r=i-p (where r is the real rate, i is the nominal rate and p is inflation). Once we rearrange this identity and make p the object of interest, an increase in i will increase p and vice versa.
Figure 1
What does all of this mean for forward guidance?
There are some who claim that the SARB has not been able to anchor expectations (Kabundi and Schaling, 2013) and some who claim that the interest rates do not affect inflation as expected (Bonga-Bonga and Kabundi, 2015). If these claims are right then it should matter little what the SARB communicates to the public. Inflation adjustments will simply be backward looking.
If the SARB is very credible and can influence expectations then it still faces some tough decisions. With a perfect foresight model inflation could simply follow interest rates (assuming that the Fisher identity holds). Or people might see the possible hike as a signal for higher inflation in which case firms adjust prices to higher inflation expectations. But, we see that short run unanticipated shocks reduce inflation. The Fisher effect works only really in the long run. The SARB nudges interest rates to control for various shocks, and as such perfect foresight models are not always ideally suited to reality. Models with bounded rationality or learning might be better suited to analyse the economy where some foresight (i.e. SARB forward guidance exist).
On the other hand forward guidance may lead to a reduction in inflation as the potential hike makes consumers and investors cut back on credit and spending now, thus lowering inflation expectations. If the SARB then does not react to its initial "guidance" then it could hurt its credibility and may not be able to anchor inflation expectations.
Perhaps the SARB is much smarter than we think- it knows how to influence our expectations and thus are perfectly aware of what the effects of policy announcement are. I for one am waiting for a paper on this.
References
Kabundi, A. and Schaling, E.(2013). Inflation and inflation expectations in South Africa: an Attempt at explanation. South African Journal of Economics, 81(3): 346-355.
Bonga-Bonga, L. and Kabundi, A. (2015). Monetary policy instrument and inflation in South Africa: Structural Vector Error Correction Model approach. Munich Personal Repec Archive, MPRA Paper No. 63731.
In this post I look at a potential problem central bankers face. Hint: it has to do with forward guidance - the way a central bank communicates its interest rate decisions to us. I picked the South African Reserve Bank's (SARB) latest (April 2015) Monetary Policy Review (MPR) to highlight these issues. The MPR, I think, tells us that everyone should be aware of a rate tightening cycle. Is it nice of them to "warn" us of an impending rate hike? Does it intend to adjust our expectations? Or can it backfire? Let's start by looking at some of its communication tools:
The fan chart:
Fan charts look pretty cool. Fan charts are sometimes used as a forecasting tool and depicts various paths of a variable with a confidence interval. It looks something like the figure below (source: 2015 MPR, SARB). The dark line is the median of the forecast while the lighter areas represent increasing confidence intervals. An easy way to think of this is by inverting the figure and plot a bell shaped curve where the light areas measure the standard deviation.
Why would a central bank produce such a figure ? 1.) It acknowledges that models are only partial representations of reality and that significant forecast errors exist (inflation is forecast to be anywhere from 3% to about 11% in 2017); 2.) to communicate what it might possibly do with interest rates (only if the central bank targets inflation). The fan chart produced below is a split-normal distribution (i.e. it is not perfectly symmetrical on both sides of the median). In fact all the "risk" to the forecast seems to be on higher inflation. Depending on how seriously the SARB takes these forecasts, it could mean that is pricing in a high probability of an interest rate increase.
In the latest MPR from the SARB: "Underlying inflationary pressures are resilient and expectations have converged at the top of the target range. This makes a sustained breach of the inflation target more likely...Given all these factors, monetary policy remains in a tightening cycle". There might be a slight contradiction in this statement - if expectations have converged at the top of the target then why would we have a sustained breach of the target? Converged expectations mean anchored expectations (admittedly at the upper limit of the target). Some models allow for backward indexation in the hybrid Phillips curve (more on this later), which means that past inflation and future inflation affects current inflation.
Sounds like the SARB wants to increase interest rates...
This sounds to me like the SARB is seriously thinking of increasing interest rates. Is this what we call forward guidance? The SARB does not make an explicit statement regarding interest rates. The fact that the fan chart for inflation ranges between 3% and 11% suggests that there is huge monetary policy uncertainty (monetary policy uncertainty is a shock over and above actual changes to interest rates see this).
On the one hand the SARB in their models often target expected inflation as opposed to realised inflation. If they believe inflation in the future (say 18 months from now) is above the target level they might very well increase interest rates. However, if they announce that they are in a tightening cycle and consumers and firms believe them, then economic participants might decrease consumption. Especially when higher expected interest rates affect credit and investment decisions. In this case expected inflation should decrease. Thus, consumer and firm decisions are conditional not only future inflation but also future interest rates.
This makes monetary policy incredibly difficult and somewhat counteracts forward guidance. Since the information set that monetary policy makers face is too big to make useful numerical estimates of inflation, they often revert to simple models. The modelling team at a central bank will usually produce inflation and output forecasts and then the MPC will decide on the interest rate path. It would be interesting to do a study the deviations of the policy advice coming from these models from the final MPC decision.
How do we create such a figure?
With a model that captures economy-wide effects. The smallest of such a model includes an equation for interest rates (Taylor rule), for inflation (Phillips curve equation) and output (IS equation).
- Taylor rule: Interest rates are set according to a rule that weights both inflation and output (usually output deviation from potential output) and a disturbance term (which could capture monetary policy uncertainty depending on how the model is solved). The parameters that fit the equation are of particular interest. In many estimates of the equation, interest rates increase by more than 1% for a 1% increase in inflation. An estimate lower than 1 often yield indeterminate equilibria (not a single path solution for the variables).
- Phillips curve: Inflation is a function of future (and sometimes past) inflation and output. If output is above potential output than inflation increases.
- IS curve: Output is a function of the real interest rate (repo rate minus inflation) and future (sometimes past too) output.
To generate the figure the SARB has to make an assumption of the underlying distribution of the disturbance terms. They may decide that all shocks can be drawn from a multivariate normal distribution and solve the model many times. With an adequate amount of solutions they can plot the distribution of the forecasts for each variable.
So what?
Here is the tricky part: Inflation has been pretty stable around the 6% mark for a while. This has happened despite many negative shocks hitting the economy. Interest rates have also remained pretty much at a similar level. Is this proof for the Neo-Fisherian school (inflation in the long-run follows the monetary policy rate)? Thus when interest rates go up, there might be a little fall in inflation, but then inflation increases. This could be further substantiated by looking at stationary path of inflation. It seems as though inflation is a mean reverting process, but it takes a long time to revert to its mean following a shock.
Using the equations specified above, I illustrate two features of monetary policy: 1.) what happens to the economy when the SARB increases interest rates temporarily where the shock is unanticipated and 2.) where there is a permanent shock while it is anticipated (people have perfect foresight; or the SARB guides the public's expectations).
Scenario 1 is depicted in Figure 1. The results seem pretty standard: An increase in the repo rate decreases inflation. However, Figure 2 shows that an increase in interest rates lead to an increase in inflation. Hmmm...why? This is the idea of the Fischer identity in most central bank models: r=i-p (where r is the real rate, i is the nominal rate and p is inflation). Once we rearrange this identity and make p the object of interest, an increase in i will increase p and vice versa.
Figure 1
Figure 2
There are some who claim that the SARB has not been able to anchor expectations (Kabundi and Schaling, 2013) and some who claim that the interest rates do not affect inflation as expected (Bonga-Bonga and Kabundi, 2015). If these claims are right then it should matter little what the SARB communicates to the public. Inflation adjustments will simply be backward looking.
If the SARB is very credible and can influence expectations then it still faces some tough decisions. With a perfect foresight model inflation could simply follow interest rates (assuming that the Fisher identity holds). Or people might see the possible hike as a signal for higher inflation in which case firms adjust prices to higher inflation expectations. But, we see that short run unanticipated shocks reduce inflation. The Fisher effect works only really in the long run. The SARB nudges interest rates to control for various shocks, and as such perfect foresight models are not always ideally suited to reality. Models with bounded rationality or learning might be better suited to analyse the economy where some foresight (i.e. SARB forward guidance exist).
On the other hand forward guidance may lead to a reduction in inflation as the potential hike makes consumers and investors cut back on credit and spending now, thus lowering inflation expectations. If the SARB then does not react to its initial "guidance" then it could hurt its credibility and may not be able to anchor inflation expectations.
Perhaps the SARB is much smarter than we think- it knows how to influence our expectations and thus are perfectly aware of what the effects of policy announcement are. I for one am waiting for a paper on this.
References
Kabundi, A. and Schaling, E.(2013). Inflation and inflation expectations in South Africa: an Attempt at explanation. South African Journal of Economics, 81(3): 346-355.
Bonga-Bonga, L. and Kabundi, A. (2015). Monetary policy instrument and inflation in South Africa: Structural Vector Error Correction Model approach. Munich Personal Repec Archive, MPRA Paper No. 63731.
Subscribe to:
Posts (Atom)