Assessing the official Brexit simulations

The release of three sets of Brexit impact studies this week showed quite clearly that there is no such thing as a Brexit dividend. To use Boris Johnson's famous metaphor, the choice is between having a cake and eating it - we can't do both. Like the slow motion car crash that I have long used to describe the whole Brexit process, the car is about to hit the tree.

There can be no doubt that the quality of the analysis underpinning each of  the studies is first rate and although the teams involved were able to devote considerably more resources to their analysis than I was, their numbers are not too dissimilar to my own.

Starting first with the analysis produced by the National Institute, their simulations suggest that the Withdrawal Agreement which is to be put before parliament in two weeks' time will cost 4% of GDP compared to the case where the UK remains in the EU, largely due to a reduction in trade with the EU, with services particularly affected. As they put it, "this is roughly equivalent to losing the annual output of Wales or the output of the financial services industry in London." To put it in terms of the cost to individuals, this reduces income per head by around 3% which is the equivalent of £1000 per person. 

The government's own simulations also represented a considerable intellectual tour de force. They were based in four scenarios: (i) The Chequers Plan set out in the summer (and rejected by the EU); (ii) a Free Trade Agreement; (iii) EEA membership and (iv) no-deal. Unlike NIESR it did not consider the agreement reached last week – the FTA is probably the closest we are likely to get. It was also useful that the modelling strategy separated trade costs from the impact of changing migration assumptions. In terms of results, the discredited Chequers Plan implies the lowest output losses relative to the baseline of remaining in the EU, with output losses between 0.6% and 2.1% of the baseline (depending on assumptions used for the extent to which non-tariff barriers are raised – see below). EEA membership implies a loss around 1.4% of baseline GDP whilst an FTA or no-deal could lead to losses of 6.7% and 9.3% respectively on the assumption of no new net migration (in line with the government's plan).

These numbers focus only on the longer-term costs (15 years). But the worst case outcomes imply huge short-term disruptions and it may be the case that there are even bigger losses in the near-term which are partially offset in the longer term. In other words, if there are big short-term output losses this will matter more to people's perception of the Brexit costs than the long-term outcomes. 

The BoE’s analysis focuses on the near-term losses up to five years ahead and models a number of possible outcomes ranging from a disorderly Brexit, a disruptive Brexit; a less close relationship with the EU and a close relationship. In the latter two cases, the impact relative to baseline is minimal. But in the disorderly Brexit scenario, the term premium on UK government bond yields rises by 100bps and sterling collapses by another 25%, in addition to the 9% since June 2016. In this case the inflation rate spikes up to 7.5% whilst output falls by between 7¾% and 10½% which drives the unemployment rate up to 6.5% (see table below).

Indeed, the BoE's simulations dominated the headlines this morning. What was less remarked upon was why the BoE predicted such a large short-term decline in output. It transpires that the worst case scenarios point to a big rise in interest rates to combat a surge in exchange rate-induced inflation (Bank Rate rises to 5.5%). This apparently counter-intuitive result is all to do with the structure of the models used by the BoE. The main simulation tool is a so-called Dynamic Stochastic General Equilibrium (DSGE) model. One of these days I will look at them in more detail but suffice to say they assume that Brexit represents a supply shock that results in a slowdown in potential GDP growth and thus to a narrowing of the output gap. This in turn leads to a rise in inflation expectations at the same time as measured price inflation is picking up, and the model assumes that the BoE will respond by tightening policy, which of course exacerbates the effect of the initial Brexit shock on output.

It is a moot point as to whether the BoE will really respond in such a way. My guess is that it will not and will act in a similar fashion to summer 2016 by assuming that inflation is a one-off exogenous shock that will eventually fade. After all, households will not react kindly to a big income squeeze if interest rates are being raised and I maintain that the BoE will be forced to ease monetary policy rather than tighten it.

Perhaps what all this analysis suggests is that we have to be aware of the different types of models and in-built reaction functions when interpreting the simulation results. Although all the modellers have been explicit about the assumptions and models used, it is difficult enough for specialists to figure out exactly what is going on. Thus the average person on the street has no chance. It is therefore easy for charlatans like Jacob Rees-Mogg to dismiss the carefully constructed analysis as part of Project Fear.

It may indeed be true that the worst case outcomes do not materialise. But for his ilk to dismiss the costs associated with a no-deal Brexit is irresponsible. It is certainly unworthy of JRM to dismiss Mark Carney as "a second tier Canadian politician", particularly when he himself has never held a frontline political position and struggles to count to 48. The sky may not fall in next March, in which case it will be the result of a degree of preparedness on the part of responsible adults such as Carney. But if it does, we know where to come looking to apportion the blame.