Wednesday, January 17, 2018

Bank of Canada meeting

As expected (see my post from yesterday), the Bank of Canada increased its target interest rate to 1.25% today. The risk they see on the horizon is a potential collapse in NAFTA.

In the first paragraph of the press release, the Committee summarizes its reasons:
Recent data have been strong, inflation is close to target, and the economy is operating roughly at capacity.
This might seem like a justification for doing nothing. The Bank has achieved its goals, so no action is warranted. This only makes sense if Bank people are forecasting that an economy operating "roughly at capacity" will wake up the Phillips curve and cause more inflation, which they think they should tamp down with higher interest rates - now, not when the inflation happens.

But, further on in the press release is this:
Recent data show that labour market slack is being absorbed more quickly than anticipated.
That seems inconsistent with the quote above. How can the economy be operating "roughly at capacity," with labor market slack? I'm running roughly as fast as I can, but I continue to run faster!

Finally, the forward guidance hasn't changed:
While the economic outlook is expected to warrant higher interest rates over time, some continued monetary policy accommodation will likely be needed to keep the economy operating close to potential and inflation on target.
Still no clarification as to why these higher interest rates should be warranted, and under what conditions rates will or will not rise in the future. And why is the current policy seen as "accommodative?" Short-term nominal interest rates may be unusually low, but it's generally accepted that the real effects of monetary policy actions dissipate over time. So, the real effects of monetary policy we should be seeing now are the effects of interest rate hikes last year. Certainly that's not accommodative. In terms of the effects on inflation of interest rate increases, again I think the Bank of Canada has the sign wrong - though they have good company in that belief. Inflation control is about moving the central bank's nominal interest rate target in the direction you want inflation to go. That's what the weight of theory and empirical evidence tells us.


  1. The Bank of Canada has employed a structural model of the Canadian economy to predict the response to monetary policy since the first introduction of the Quarterly Projection Model ("QPM") in 1993 (cf., P. Duguay and S. Poloz, “The Role of Economic Projections in Canadian Monetary Policy Formulation,” Canadian Public Policy 20, 189-199; and also, S. Poloz is the same S. Poloz now Governor of the Bank of Canada.

    The original QPM was updated to "ToTEM" and then to "ToTEM II". "ToTEM II" is the economic modelling formulation is use today (insofar as I am aware). It is described in a Bank of Canada technical report by José Dorich, Michael Johnston, Rhys Mendes,
    Stephen Murchison and Yang Zhang titled "ToTEM II: An Updated Version of the Bank of Canada’s Quarterly Projection Model", Technical Report No. 100, October 2013 ( ).

    The manipulated variable is the short-term risk-free interest rate (nominal) which is set according to an "augmented Taylor rule" (eqn. 1.54, p. 27).

    The term "Phillips curve" makes ten appearances in the technical paper in reference to aggregated equations for the supply and demand curves for the economic sectors modelled in ToTEM II. From this, it is safe to conclude that the Bank of Canada following the ToTEM II model output would appear to be setting monetary policy within a Phillips curve framework although whether this is intentional or simply an unintended result of the modelling approach is not clear from the technical paper.

    The augmented Taylor rule for the manipulated variable, the short-term risk-free (nominal) interest rate, Rᵣ(t), (eqn. 1.54, p. 27) is
    Rᵣ(t) = Θᵣ∙Rᵣ(t−1) + (1 − Θᵣ)∙(ȓ + πͦ (t) + Θᵨ∙(π(t+2) − πͦ (t)) + Θᵧ∙ŷ(t)) (eqn. 1.54)
    Θᵣ is the exponential smoothing parameter
    ȓ is the "steady-state" long-term real interest rate
    πͦ (t) is the targeted annual rate of inflation
    π(t+2) is the projected annual rate of inflation two periods ahead
    Θᵨ is the constant of proportionality (gain factor) on deviations of inflation
    ŷ(t) is the output gap (however specified)
    Θᵧ is the constant of proportionality (gain factor) on deviations of output

    The augmented specification differs from the original Taylor rule formula in its use of exponential smoothing in the calculation of the short-term risk-free interest rate. Selection of the value the smoothing parameter Θᵣ will determine the rate of change in the Bank rate. A controls engineer would anticipate the action to be a first-order lag in response to a step-change in the controlled variables (in this case, π(t+2) and ŷ(t)). The choice of the parameter values leads to a critically-damped response to a typical second-order linear system, as shown in the graphs presented in the technical paper.

    Based on a review of the technical paper, the Bank's decision to raise the level of the short-term risk-free interest rate is motivated by QPM output and the reliance on an exponentially-smoothed Taylor rule generated value for the manipulated variable Rᵣ(t). As noted above, exponential-smoothing gives rise to a first order lag. In order for the controller response to be effective, the manipulated variable must continue to change until the controlled system state returns to zero (i.e., the control point for the controlled variables). There are pros and cons to this form of controller. Evidently, the Bank of Canada researchers are satisfied that this controller specification is appropriate for the Canadian economy which is small (relatively speaking) and open (i.e., Canadian monetary policy cannot effect World interest rates in the way that the US Federal Reserve can).

    Is there a better way? Perhaps.

    1. 1. The Bank of Canada was in the policy simulation business long before 1993. Poloz gives you some of the history in this speech:

      In the 1960s people at the Bank worked on the RDX model, which became RDX2, and by 1980, when I was there, that had morphed into RDXF - mainly a forecasting model, but we sometimes used it for policy simulations too.

      2. All of these models - old fashioned large-scale macroeconometric models, modern New Keynesian quantitative models - have Phillips curves in them. No surprise there.

      3. Though I'm sure the Bank of Canada's current model of favor plays some role in the policymaking exercise, I would be very surprised if they're literally relying on the model to tell them what to do, in the sense of: The model tells me we have to move up 60 basis points to hit the 2% inflation target in 6 months, or some such, so that's what we do. They're much more qualitative than that, and they use a lot more evidence. There are also many factors - NAFTA for example - that matter, but aren't in the model.