The BoE should follow the Fed and make its model downloadable and forecast judgements open to scrutiny

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.

Recently, they published a working paper which includes an equation listing for its new COMPASS model, and an account of how this model sits within the so-called ‘suite’ of models.  On the face of it, you might say ‘Great!  How generous and transparent of them!  What a forward-thinking helpful central bank!’

But, hang on a moment.

They publish the equation listing to set someone the challenge of coding the model up and solving it using their own software?  Why?  To the cynical it looks just like a device to make it look like they are being transparent, while hoping that no-one will find it worth the trouble.  And coding up the model and solving it is the easy part.  Then there is the difficulty of collecting all the data and following the complex data transformations used by the staff.  And then replicating the delicate steps used to estimate it.  And then figuring out the judgements that the MPC have layered on top of the model, derived from their own wisdom, or insights from the ‘suite’.  We’ve nothing to hide, they might say.  It’s all there.  Well, if not, why try to discourage people from checking?

I know that the BoE worries about making this code and data downloadable.  They worry about this leading to them assuming a software and data support role that they don’t want.  So what?  Of course the Bank should assist those who are to hold it to account.  They also worry that it poses operational risks [what if they cock it up?].   And they worry about the scrutiny of the forecast it would bring.  Wouldn’t disclosing the model and judgement just help those who want to make mischief make it?

These concerns are flimsy set against the many arguments for it.  Such a large public body  with so many important powers has to be held to account, and anything that can help that process is vital for the democratic process and for the credibility of policy itself.  If you doubt that, reflect for a moment that there are many occasions when it would serve the MPC well to layer on a heavy judgement to make it look like inflation is going to come back to target when their model’s say it is not.  Purely hypothetically:  imagine an economy seemingly stuck at the zero bound,with inflation drifting slowly below target.  Its models might say [as the Bank’s New Keynesian model could well say] help!  Inflation is heading down to the liquidity trap steady-state and there is nothing but loose fiscal policy that will get us out of it!  The temptation to hide this with a healthy dose of judgement tweaking inflation firmly back towards target would be very hard to resist.

Further, it’s taxpayers ultimately that have funded the building of the model.  [Have a look:  it’s there as a chunky few million quid in the BoE annual Reports].  It seems perverse to make them pay again to get any use out of it [by forcing them to fund their own efforts to replicate the BoE model].

Keeping the code, toolkit and data locked away is against the spirit of modern scientific ethics, contrary to the hints in the new BoE structure that aspires to prioritise research.  Journals now insist that researchers deposit data and code so that research can be replicated and challenged.  The BoE isn’t a journal.  But it is hoping it gets kudos from using this big beast of a DSGE model, and trumpets the beauty of its suite software, so it should allow others to scrutinise it to check just what this thing that cost so much is being used for.  If we had past vintages, we could take a close look at past miracles, like how the switch from one model [BEQM] to another [COMPASS] took place without a bobble or a quiver in the Inflation Report forecast profile.  [It’s easy to hazard a guess:  the judgement is pivotal, and was modified to make sure that the signal from the model had little or no effect on the profile.]

Almost of all of the knowledge in the model has been lifted from research done by others outside the bank, in academia and in other central banks.  [DSGE models trace a lineage from Lucas, Kydland, Prescott, Christiano, Eichenbaum, Evans, Smets and Wouters.  The technology to estimate them comes from Kalman, Sargent, Sims, Schorfheide, Gibbs, Metropolis and Hastings].  The Bank has borrowed enthusiastically from free open source code libraries to hone its own toolkit for solving COMPASS and the other models.  It is odd that having taken so gladly, that it chooses to keep that toolkit from the community it relied on to produce it.

Having this transparency would be particularly important now.  The MPC is making a lot of its ability to manipulate expectations of future rates, via forward guidance.  This throws a particularly keen spotlight on the forecast.  The MPC is using what looks like a model of rational expectations:  a model in which agents inside the model can manipulate and compute its equilibrium.  This is somewhat ironic, since the BoE withholds the means for people to do just that!  Is it relying on the properties of RE for its estimates of the impact of forward guidance?  How has it dealt with the known pathologies of these models under fixed interest rates [the models frequently go haywire, unless arbitrarily fixed to stop it, generating alternatively massive inflations or deflations for small changes in the horizon at which rates are fixed].  Why rely on rational expectations in this case, embarking on a new policy, in novel economic circumstances, precisely those when this should be the worst possible approximation to how expectations would actually behave?  ‘Ah, we’ve dealt with all that with our wise judgements’, the MPC might say.  ‘We don’t take the model so literally.  It’s just one of many [indeed a suite] of sources of information.’  Well, to make sense of this, and to understand what this judgement is doing, we would need to know more about it.  Otherwise we might assume it’s all smoke, mirrors and bluff.

So, in short, the Bank should follow the Fed, make the code, dataset and software available.   Senior management in the Bank often complain at how lethargically the academic community engages with the issues that policymakers are grappling with.  These would be steps that could encourage such engagement.  Who knows, doing so might generate insights that the Bank’s economists could use that they would not otherwise get.

Advertisements
This entry was posted in Uncategorized. Bookmark the permalink.

3 Responses to The BoE should follow the Fed and make its model downloadable and forecast judgements open to scrutiny

  1. Costas Milas says:

    Tony,
    Very sharp your piece asking the BoE to make their model/codes etc available. If I am not mistaken, however, BoE policymakers often note that their model forecasts are “adjusted” to reflect the converging view of the MPC members. So, there might be some good reasons (from their point of view) for not releasing data and codes as this would involve them explaining in detail the “adjustment factor” which might change (and rightly so) as circumstances do change.

    Incidentally, in this recently published (by the Journal of International Money and Finance) paper
    (http://www.sciencedirect.com/science/article/pii/S026156061400031X)
    colleagues and I show that a model which uses (a) money divisia and (b) stock illiquidity is the only one to beat the BoE GDP growth forecasts. The paper was sponsored by the BoE’s Research Donation Committee (we gratefully acknowledge this in the paper). We were hoping that they might be taking on board some of our recommendations as, at the end of the day, they sponsored the project.

  2. Nick says:

    Dear Tony, I think this could be an additional step to give incentives to the BoE to be more transparent. However, to be fair this problem is not exclusive to the BoE. In general central banks have incentives to communicate strategically (and misrepresent private information) rather than in a transparent manner, due to the effect of public announcements on market expectations and, in turn, the effect of expectations on economic stability. A more detailed explanation and empirical evidence supporting it can be found in a recently published journal article in the Contemporary Economic Policy http://onlinelibrary.wiley.com/doi/10.1111/coep.12049/abstract.
    About strategic communication on the FED you can see http://www.sciencedirect.com/science/article/pii/S0176268011000176 and http://www.centreformacroeconomics.ac.uk/pdf/CFMDP2014-11-Paper.pdf

  3. Pingback: A perspective from the newly-released model | croaking cassandra

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