A perspective from the newly-released model

Somewhat belatedly, the Reserve Bank last month released a Discussion Paper outlining the features of the Bank’s relatively new forecasting and policy model, NZSIM.  I’m signed up to receive the email advisories when such papers are released, but it appears that on this occasion no advisory was sent out (an oversight apparently).  A commenter yesterday pointed me to the Discussion Paper.

The Reserve Bank has long prided itself on its formal macroeconomic models.   This dates back at least to the days of Roderick Deane, Chief Economist and later Deputy Governor of the Bank in the 1970s and early 1980s and one of the greatest figures in the history of the Reserve Bank.    The Reserve Bank was one of the early central bank adopters of formalised models, although most other advanced country central banks now use them in some role or another.  Historically, the Reserve Bank of Australia has tended to be towards the sceptical end on the role for economy-wide models in policymaking.

Maintaining such models has been a heavy investment for a small institution, especially as on at least a couple of occasions (over decades) the models have been junked almost as soon as they were finished.

And views on quite how large a role the models have ever played in the policy side of the Bank probably differ from observer to observer.  I was closely involved for a long time, and I tend towards the sceptical end.  The Bank had an unwarranted reputation for being somewhat in the thrall of whichever model it was using at the time.  I will always remember the time, fifteen years or so ago, when Glenn Stevens came over and spent several days observing our quarterly forecasting and policy round, and emerged commenting that he hadn’t realised that the Reserve Bank of New Zealand was really quite so pragmatic.

Structural models of the entire economy tend not to be overly useful for the sort of near-term forecasting (and backcasting and nowcasting) that largely shapes real-world monetary policy setting.  The current Deputy Governor, Grant Spencer, made this point well when, as an outsider speaking at a workshop to launch an earlier model in the 1990s, he noted that the technology was likely to be more useful for policy simulations (“what happens if we apply some shock to the system”) than for forecasting. There are simply too many institutional and data-related details that will be known to the forecaster at any particular time, but can’t be captured in a structural model, a deliberately stylised representation of the economy.  And that is even before one asks questions about anyone’s ability to forecast the economy more than a quarter or two ahead.

A good structural model captures the key features of how the designers think the economy works. But in an official agency, it is only likely to be useful if it reflects the key elements of how the decision-makers think the economy works   If it doesn’t, then over time either the model itself has to be adapted, or it will fall into disuse  (perhaps serving as an adding-up framework, and as a technology for generating nice charts and tables quickly –  which NZSIM was doing –  but with the structure of the model overridden pretty much all the time).  Obviously I’m no longer close enough to know what role NZSIM is playing in Graeme Wheeler’s deliberations (whether on forecasting or scenario analysis) but I’d be surprised if it was terribly large.  Apart from anything, for example, in this model, immigration is not explicitly treated, and fiscal policy changes never alter the deficit (any change in spending is automatically financed by a change in lump sum taxes)

Nonetheless, it is good to have the model Discussion Paper in the public domain.  As former Bank of England official Tony Yates has highlighted, (and here) the benchmark in this area remains the Federal Reserve

The Fed recently made its workhorse model FRB-US downloadable, with a dataset, code, everything you need to take a close look at what Governors say and what the staff have been doing for them.  The Bank of England should do the same.

Perhaps one could say the same about the Reserve Bank.  Having that additional material wouldn’t greatly interest me personally, but there are other people outside the Reserve Bank with considerable modelling background and experience for whom it could be useful, as part of further strengthening the external scrutiny of the Reserve Bank.   It can be useful to have a better sense of whether differences from the Bank arise because of different inputs (exogenous variables) or different assumptions about how the economy works.

But for now, I just wanted to highlight one chart in the Discussion Paper.  It shows the responses of a variety of variables to a 1 percentage point “monetary policy shock” –  roughly, a change in the policy rate of 100 basis points different than would the “policy rule” in the model would suggest.  There is nothing special about that particular policy rule –  indeed, I doubt it has ever been discussed in any detail at the Bank’s Monetary Policy Committee  – but also nothing especially objectionable about it.

impulse responses

But it is interesting because one could think of last year’s OCR tightenings as a 100 basis point monetary policy shock.  No doubt it won’t have been quite that in the formal model sense, but many people would now subscribe to the view that the tightening was largely (or completely) unnecessary.  Certainly, it has now been fully reversed, at least in nominal terms (real interest rates are still higher than they were when the tightenings began in March last year).

Within this model, a representation of the economy that the Bank is content to use in its internal processes and to describe as “the” new forecasting and policy model”, a 100 basis point monetary policy shock, that is unwound after a year or so, lowers inflation by about 0.2 per cent (roughly evenly split between tradables and non-tradables).  But it also has real economy effects:  after about five quarters, consumption is almost 1 per cent lower than it otherwise would be, and GDP is almost 0.6 per cent lower than it otherwise would be.

In a more formal way, it makes much the same point that the Minister of Finance was making the other day.

Finance Minister Bill English says the Reserve Bank raised interest rates “a bit too far” in 2014, contributing to slow economic growth at the start of the year.

“It’s one of the factors, along with dairy prices, that probably led to a much flatter 2015 than we had expected,” English told Bloomberg Television on Thursday evening.

“In retrospect, they lifted them a bit far” and “had to go back”, English said.

Graeme Wheeler’s experts might object to my characterisation of last year as a “monetary policy shock” in this sense, but the increases were clearly unnecessary and have been reversed.  Whether or not they could be justified at the time, the fact that they were unnecessary with hindsight means there will have been some short-term real economic cost.  The Bank’s model provides one way  –  using their view of how the economy works – of trying to get a plausible fix on the size of that cost. The model doesn’t have the unemployment rate within it, but  –  all else equal –  an additional 0.6 per cent of GDP might have been enough to have prevented the unemployment rate rising from 5.6 per cent in September 2014 to 6 per cent in September this year.

8 thoughts on “A perspective from the newly-released model

  1. 100 basis points is equivalent to a 40% increase in interest costs to a business. For property developers like the Panama Road Developments Springpark project with 420 houses being built, it spells collapse.


    • Correction, 40% on the 90 bank bill rate but for a developer on 10% interest rate, 100 basis points is equivalent to a 10% rise in interest costs.


    • I’d endorse that approach, but to take effect from the day NZ becomes part of Australia.

      Dealing with the losses that arise in bank failures is almost inevitably a highly political distributional issue – and as we know the Australian play hard ball in many areas. Of course, one could mount an argument to simply apply the same bank supervision rules in NZ as in Aus, and simply keep a reserve army to deal with (exceptionally rare) severe failures, but I doubt that would prove sustainable.


  2. Michael, you’re looking at unanticipated monetary policy shocks here, when what you’re describing is much more like anticipated shocks since the rate rises (at least the direction/path, if not exactly the timing or size) were announced in advance.

    In rational expectations models like this, anticipated shocks have much more muted medium-term effects since agents’ responses are front-loaded. You’d expect to see GDP drop much sooner, and stay lower for a shorter amount of time relative to the impulse responses you’ve plotted here.


    • Fair point thanks James

      Interesting question is whether the 2014/15 episode should be thought of as anticipated or unanticipated. The OCR increases were certainly well-signaled and expected – and had widespread support at the start. But as it became apparent that they had been unnecessary – whether because circumstances changed in ways that weren’t foreseen, or just because core inflation pressures were always weaker than the consensus recognized – couldn’t one argue that there was an “unanticipated monetary shock” (the OCR ended up hgher than it needed to be, and higher than people – looking ahead – had thought would be warranted)?


      • I think it’s an interesting question, too. It’s also worth asking whether instead of shocks, this is better modeled as a change in the rule (perhaps a fall in the inflation target, or rise in the strength of response to inflation/activity).

        WRT the anticipated/unanticipated shock question, I think it’s still better to think of it as anticipated. As long as the Bank’s signaled rate rises were credible, the shocks should be thought of as anticipated, even if the shocks were starting to be thought of as mistakes. As long as agents believe the Bank will go ahead with its publicly stated plans – no matter how ill-advised they may be – those plans are still anticipated. If the Bank lost credibility, though, and the plans weren’t believed, then the path of interest rates would be less anticipated.

        Do you think the Bank’s credibility was at stake here?


  3. I don’t think “credibility” in the sense of ‘have they abandoned the inflation target” was seriously in question (altho perhaps in some circles the personal credibility/reputation of the Governor might have come into question).

    I’m still not really persuaded by the anticipated line: the OCR (actual and projected) was set on some expectations that didn’t play out. With perfect foresight, they’d have set policy differently (and markets would have expected something different, so in that sense the “shock” was in the realization that the things driving the policy rule weren’t actually where people thought they’d have been.

    (All that said, I’m certainly not wedded to those impulse responses (level or paths), or even to the scale of the “shock” – I was mainly interested in just illustrating why it can be nice to have the model out there so that, inter alia, discussions like this can go on.)


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s