(Reprinted with permission)
In recent weeks there has been a relentless stream of output from modelling groups on the topic of Brexit- all of it negative. This has included long term and short term reports from not merely the Treasury but also CEP at LSE, PWC, Oxford Economics, the NIESR, the OECD and the IMF. All of these except the IMF have made available fairly good information about their methods and models. The Institute for Fiscal Studies (IFS) has usefully done a survey of this work- IFS, 2016. They note that apart from Economists for Brexit (EFB) and Open Europe, the consensus is that Brexit will reduce GDP relative to a base forecast both in the long term and the short term. EFB by contrast forecast a rise in GDP in both short and long run while Open Europe find a range, depending on post-Brexit policies from small negative to small positive in the long run (they do not look at the short term). The IFS table is shown below.
The IFS look at these forecasts and note that the common element in the consensus outside EFB and Open Europe is that after Brexit under the WTO option the UK continues to maintain protectionist tariffs and other trade barriers against the rest of the world, including the EU.
By contrast EFB assumes unilateral free trade after leaving the EU. Open Europe considers a range of policies, which at the one extreme appear to involve more free market, free trade assumptions; we leave them on one side in what follows. The IFS comments that unilateral free trade is ‘unlikely’ to be politically feasible and in its subsequent calculations of the effects on the public finances takes the mean of the negative consensus.
At a conference on Friday May 27 in the NIESR all of these groups were represented apart from Open Europe and HM Treasury. The assumptions made by HM Treasury were in fact clarified by George Osborne to the Treasury Select Committee: after Brexit there would be general UK protection against the rest of the world, assumed to be in line with past experience of the countries before they entered the EU.
The groups who spoke at the NIESR conference confirmed that they had assumed UK protectionism after Brexit. The extent assumed differed according to the methods used. In fact, some groups used gravity equations (NIESR, CEP), others used Computable General Equilibrium (CGE) models (PWC, Oxford, OECD, also CEP in a new version). In one experiment CEP reported assuming that the UK reduced tariffs by 3% after Brexit, but this did not appear to correspond to the trade barriers the EU imposed on the UK after leaving, which had major effects on UK export trade; nor does it correspond to agricultural protection, which the OECD estimates at just under 20%.
By contrast EFB assumed that the EU trade barriers would be by 2030 around 10% on agriculture and manufacturing- EFB estimates are that recent protection levels are around twice that but it made the cautious assumption that due to international pressures that have gradually brought down trade barriers over the past two decades, protection rates would fall. EFB then assumed that on Brexit these barriers would be unilaterally abolished by the UK on its trade, so that import prices would fall to world levels in agriculture and manufacturing. Exporters of these products would also find their prices falling to world levels from preferentially higher levels. EFB used a standard CGE world trade model and this caused rising consumer living standards and a reallocation of output towards the unprotected service sector. As noted, this would include hi-tech manufacturing so that this transition would include the moving of manufacturing into this form.
Plainly there are other ways in which the various groups differed in their assumptions, including on migration and regulation.
These differences, as well as the differences in modelling methods, can account for detailed differences in the overall Brexit effect on long term output. However, these merely varied the effects on output around an overall negative mean. This negative mean appears to come from the assumption about the absence of UK unilateral free trade, much as indicated by the IFS.
Notice that by pursuing protectionism post-Brexit the UK would not only close down free trade with the EU regionally for its exporters, but fail to increase (or even reduce) free trade for its importers; thus this assumption actually moves away from free trade. All the modelling groups asserted that in their models moving towards free trade enhanced output; so the Brexit trade assumption would be expected, by reducing free trade, to reduce output. Plainly it does in all these models, regardless of their specification.
The question therefore that arises about the long-term consensus results of such a negative mean effect is whether it is reasonable to assume that unilateral free trade is indeed ‘politically infeasible’ and therefore to make this assumption the basis for forecasting.
First, suppose it was infeasible politically in the absence of full information about the consequences of different policies, on which people could form opinions and vote. Would it not be right for economists to cost such an option and provide information on it to the electorate? What the consensus modellers have done is simply to suppress this option and provide information only on the damaging Brexit WTO option.
Second, is unilateral free trade in fact politically infeasible? One major country, New Zealand, actually carried this policy out in the late 1980s- the Douglas reforms. Furthermore 92% of UK workers are employed outside the protected sectors of agriculture and manufacturing and would have a clear interest in lower consumer prices as well as a more efficient economy. It is often forgotten not only that protectionist policies are designed behind closed doors, without consulting or even informing voter-consumers, but that UK workers have been through substantial supply-side reforms in the past four decades; manufacturing employment has dropped from 35% of the workforce in 1970 to 8% today.
Many of those 92% have adjusted already to the same competition that would now be stronger post-Brexit. Given the choice of lower prices or continued protection they may well choose lower prices and opt to pursue policies designed to help manufacturing and agriculture adjust through higher productivity together with help for particular firms or subsectors on grounds of non-economic needs.
The policies involved could include lower energy costs and hi-tech infrastructure such as broadband and transport links. As the US has shown with steel there is scope for anti-dumping action as permitted by the WTO to deal with special situations.
Third, unilateral free trade would be a stimulus to our EU partners to negotiate sensible transitional arrangements for particular sectors such as the car industry. They will not be keen to subject EU car exporters, for example, to a large fall in UK car prices. It has always been envisaged by the Brexit side that there would be constructive negotiations on such matters. But if the UK simply goes to protection there is little incentive for the EU partners to negotiate since they will see the self-damage created and wait for the UK to change its mind.
In short the decision of the consensus to assume that the UK, instead of using Brexit as an opportunity to move to general free trade, pursues general protectionism which in effect reduces the extent of free trade, is important in obtaining its negative results for output. It is also hard to justify both ethically, in that information has been denied to voters and practically, in that unilateral free trade policies are not merely an optimal but also a natural choice for UK voters.
The short term forecasts
It remains to discuss the short-term forecasts made by these modellers. What emerged from the NIESR conference is that the method used by all modellers, with some minor variation, was to inject the long term effects on output productivity into a macro model for the short term and assume Rational Expectations (i.e. people understand the effects of new policies) to allow knowledge of these to affect consumers and firms’ short term spending decisions.
On top of this, modelling groups made a variety of ad hoc assumptions about the rise of ‘uncertainty’ (about post-Brexit policies)- on the rise in credit and other financial costs. Thus whether there would be a recession or not in the short term depended for these groups on a) the extent of the long term hit to productivity, and b) the extent of higher uncertainty.
Against this the EFB argue that uncertainty is a function of how quickly new post-Brexit policies are decided and explained. We can see no reason for delays in this under our WTO option since the decisions will in practice be entirely in UK hands; negotiations would be embarked on whose outcomes could easily be predicted and explained to the relevant industries. Under the WTO option control lies in UK hands. Thus policy uncertainty would be closed down. There is of course always some uncertainty in policy of all sorts; but it would not be out of the ordinary.
Thus it is that the EFB short term forecast is not at all dramatic. Output improves gradually towards the long run trade effect plus the regulative effect calculated separately. Other factors cut in, such as a falling exchange rate and rising competitiveness as workers trade rising living standards for some fall in real wages to boost employment.
In a previous report on the Treasury’s methods for estimating long and short term effects we set out a lengthy critique. We also made a similar critique of the CEP modelling approach. We favoured the use of CGE (structural) models. However we note that other groups have used CGE methods and get to some extent similarly negative results. What has emerged from considering all these approaches used by different modelling groups in the consensus is that they all assume post-Brexit the pursuit of protectionist policies on imports by the UK. This reduction in the scope of free trade predictably would damage UK output and productivity whatever methodology is used.
The key difference in EFB is the use of the unilateral free trade assumption under which Brexit is a move towards free trade. This not only gives a long-term boost to output but it also boosts the short term outlook by the standard route of expectations; with suitable policies it can be generally popular and beneficial across all sectors. It also enables the UK to be strong in negotiations and take control of its own policy environment independently of any actions by our EU neighbours. This in turn closes down short term ‘policy uncertainty’, avoiding the ad hoc rises in credit and other financial costs in the short term.
In short, the consensus has misrepresented the post-Brexit potential outlook quite seriously to UK voters in this referendum. It is the intention of EFB to bring this clearly to these voters’ attention.
Appendix on models and methods of the modelling groups:
The modellers whose results I review here have used a wide variety of modelling methods for both the short and long term.
By way of introduction EFB uses a CGE model of world trade under full competition with a 4x4x4 structure, four ‘countries’ (UK/rest EU/NAFTA/RestofWorld), four sectors (agriculture/manufacturing/services/nontraded), and four factors (land/capital/skilled labour/unskilled labour). This structure does rather well at accounting for long term trends in trade, employment and production (see Minford et al, 1997) in response to globalisation and technology shocks. For the short run EFB use the Liverpool Model supplemented by recent research on DSGE model behaviour- more details in EFB, 2016.
The long term modelling methods of each group fall into two types: gravity equations and CGE modelling.
Gravity equations are used by: HMT, NIESR (and in earlier versions CEP-LSE, but not in latest)
CGE models are used by: PWC, Oxford (GTAP), Open Europe (GTAP), OECD (GTAP for OECD Metro), more recently CEP(LSE) have developed a CGE model underlying their gravity equations- which appears to be the one they favour and show for their results in IFS.
EFB comments on gravity equations are critical: see Minford et al, 2015. Gravity models give large reactions as can be seen in IFS table for HMT and NIESR WTO+.
EFB comments on CGE models: in principle we support this approach but details of these models differ substantially and in particular the GTAP model is proprietary so that knowing what exactly is in which version is hard. In general these models embody reactions made quite ‘sticky’ by imperfect competition: this means that the size of their reactions is generally smaller than that of gravity models or the EFB CGE model which assumes full competition (this seems to us appropriate in a long run model).
Short run modelling by the groups also differ. Three groups- HMT, NIESR and OECD- take long run results and embed them in NIGEM (the NIESR’s macro multi-country model) as changes in long run productivity; they then run NIGEM under rational expectations while adding uncertainty premium effects on various credit and financial cost variables. OECD drops rational expectations in favour of ‘backward-looking’ expectations whereby people slowly learn about the future from events; it argues this goes better with uncertainty effects. These short run uncertainty effects are fairly large in all these models.
LSE and PWC use the same CGE model but ‘dynamically’ for the short term. They too add uncertainty effects ad hoc into investment etc. Oxford Economics use their own short term macro model which is proprietary; they add uncertainty effects similarly to those using NIGEM.
It can be seen that there is a wide variety of both short term and long term modelling methods being used across these groups. As far as generalisation is possible, it would seem that all of them get quite similar uncertainty effects but this may well be because these effects are essentially ad hoc (i.e. we have no clear basis on which to impute them for any shock) and so they have sought to justify similar effects through a wide variety of ad hoc assumptions.
For long term results there seems to be a tendency for the CGE models these groups have used to give smaller effects than the gravity models used. This may be because CGE models under imperfect competition react rather modestly to shocks.
References to modelling group work- much of this work is recent and so written up in memos and informally on slides for presentations. Those interested in following up should address the institutions involved directly for details of all their modelling work.
Institute of Fiscal Studies, 2016, Brexit and the UK’s Finances. IFS Report 116, Institue of Fiscal Studies, London.
Minford, Patrick, Eric Nowell, and Jonathan Riley, 1997,’Trade, Technology and Labour Markets in the World Economy, 1970-90: A Computable General Equilibrium Analysis, The Journal of Development Studies, Vol. 34, No. 2 December 1997, pp.1-34.
Minford, Patrick, Gupta, S., Le, V.P.M., Mahambare, V., Xu, Y., 2015, ‘Should Britain leave the EU? An economic analysis of a troubled relationship’, second edition, Edward Elgar.
EFB publications- our forecasts for Brexit, our critique of the HMT reports, and our response to an LSE press release – can be found on http://www.economistsforbrexit.co.uk