The Civil Service reportedly has redone the Treasury’s Brexit long-term forecasts with a new approach, so say numerous leaks via Buzzfeed and elsewhere. ‘Officials believe the methodology for the new assessment is better than that used for similar analyses before the referendum,’ reports Buzzfeed. This new approach has, it seems, dumped the old Treasury calculations and methodology published in the original Treasury Project Fear report during the referendum. Plainly, the criticisms of this old approach – persistently so from us at Economists for Free Trade – have hit home; if so, that is real progress. Under its old approach, the Treasury used something they called the ‘gravity approach’. This approach consisted of four steps: A set of trade correlations amongst a variety of countries between trade and GDP that noted their membership of different trade blocs. To get the trade effects of Brexit, they put into these correlations estimates of the estimated EU-membership and WTO-option effects, which are assumed to correspond to the different policy possibilities. These effects on trade were then fed into a set of correlations between trade and Foreign Direct Investment (FDI), from data on many countries and time periods. This produced the FDI effects. These FDI effects were then fed into a set of correlations between FDI and productivity, from data on many UK industries over many time periods. This produced the productivity effects. These productivity effects were then fed into a standard macro model of the UK economy; a popular one is the National Institute Global Econometric Model (NIGEM), operated by the National Institute of Economic and Social Research (NIESR). These productivity effects feed into GDP ‘potential’ and trigger the final macroeconomic forecasts that are publicised. As the Treasury or Civil Service seems now to have conceded, this procedure makes no sense because all these relationships are ‘correlations’ – correlations do not reveal causation. We have a correlation between unemployment and crime; but it would be dangerous to use it to predict unemployment from data on crime. This is because both these data series are impacted by a complex causal system involving a lot of other factors. So now the Civil Service seems to have adopted a full world causal trade model. We would guess (since nothing else would make any sense at all) that the new approach is to use an existing reputable world trade model (a so-called ‘computable general equilibrium’ or CGE model) and to put into it the Civil Service version of ‘Brexit policy assumptions’. The Government/Treasury/Civil Service should have done this all along, as some others like Open Europe did, with much less controversial results than the Treasury’s original efforts. We do not know what model the Civil Service has used. However, we do know the results from one such model that is widely used around the world. This is the Global Trade Analysis Project (GTAP) model, produced by the GTAP centre at Purdue University in Indiana, USA. This model is the world leader in such analysis, having been continuously developed since 1993 by universities, governments and international bodies – so it is the blindingly obvious choice. Fortunately for us who want to know more about its Brexit implications, it was used by Open Europe in 2015, in a version with 57 sectors and 28 country groupings, containing all countries. It is likely the Civil Service is familiar with this model. However, such a model does only half of the necessary job. The undone bit, as we will see, is the testing against UK facts to see if it does well on the UK, with which we are concerned. The Brexit effects according to GTAP It is of great interest to ask what a full model like GTAP would say about Brexit under ‘benchmark’ policy assumptions. Fortunately, a group of economists working for Open Europe, as mentioned above, have done this for us, Ciuriak and Xiao. Let us summarise the benchmark trade policy assumptions of the current government as ‘general free trade with the non-EU world’ plus a ‘close relationship with the EU’ such as Canada-plus. We can approximate ‘general free trade’ by ‘unilateral free trade’ (UFT), which has been simulated by Ciuriak and Xiao. UFT approximates the combined effect of many FTAs with the rest of the non-EU world in eliminating protection of food and manufactures, because in an FTA you seek opening of other markets in return for opening yours. Other countries therefore demand you eliminate your protection in exchange for their eliminating theirs: so your own trade barriers go as they would in UFT. Because we are effectively costing only the effect of reducing our own trade barriers and not adding in any gains of reduced trade barriers of others on our products, UFT can be thought of as giving a lower bound on the gains from free trade policy. Ciuriak and Xiao do a ‘UK UFT simulation’ which indeed shows gains for the UK, but their calculated gain is only 0.8% of GDP; so why so low? They assume that tariffs and trade barriers eliminated are just 4%. However, we have estimated, rather uncontroversially (see chapter 4 [the article I co-authored in 2015 in the second edition] of Should Britain leave the EU? An economic analysis of a troubled relationship), that trade barriers erected by the EU for food and manufactures are both 20% when you include non-tariff barriers; 4% is simply the tariff barriers alone. In other words, Ciuriak and Xiao assume post-Brexit we do not eliminate the non-tariff barriers set up by the EU against the world. But, of course, to achieve free trade we would do this! Assuming our estimate of 20% total trade barriers, you could multiply Ciuriak’s and Xiao’s results by 5, which gives 4% of GDP. In fact, in our work we conservatively assume these barriers will fall to 10% over the long term, which would give a welfare gain of 2% of GDP in the GTAP model. The key point is that all of the results show a positive gain for GDP, in contrast to the stonking 8% loss claimed by the new Civil Service analysis. Ciuriak and Xiao also do a ‘Brefta’ scenario which is essentially Canada-plus. Here they assume big costs at the UK-ROW ‘border’ – rules of origin and customs checks. The cost of these to the UK comes out at 1% of GDP. We query this assumption of border costs: how do they arise when the WTO’s Trade Facilitation Agreement mandates that borders must be virtual? (98% of the trade of 15 developed countries in 2016 required no physical inspection and the median clearance time of the other 2% was one day according to the World Bank Logistics Performance Index 2016.) Such costs seem to assume that either the UK or the EU would act illegally at the border, which is entirely implausible; both the UK and the EU are sticklers for legality, the EU because its whole structure is based on treaty law and the UK because its political philosophy is bound up with the rule of law. Furthermore, the Chief Executive of HMRC has testified on seven occasions to select committees that the UK will be ready in March 2019 to operate such a system. On our assessment these border costs are nil. So what this study finds – under its benchmark policy assumptions – is a loss of 0.2% of GDP. With reasonable policy benchmark assumptions this would become +2% of GDP. However, the finding of Economists for Free Trade (see From Project Fear to Project Prosperity) is +4% of GDP on the policy benchmark assumptions. So why is this bigger and why is it right? For this we need to ask whether the GTAP model, designed as it is to compute the effects of general world trade barrier changes, is particularly accurate for the UK. The UK Dimension Tested It is important to test one’s ‘Brexit model’ statistically against the UK trade facts to see whether it fits them or not. To do this, you need to set up a model of UK trade holding the rest of the world as constant; that way we can focus on what it says about the UK. The results of a year-long project in which we did this can be found in the 2017 publication, Classical or gravity? Which trade model best matches the UK facts? which I wrote with Yongdeng Xu. We found that a ‘gravity model’ of UK trade does not match the facts statistically, while the ‘classical’ CGE model we use does. On our CGE model, the Brexit gain to the UK is +4% of GDP; on the full Brexit policy assumptions even a properly constructed gravity version, in fact, also says the same. What really matters, it seems, is whether the model is ‘tuned’ to fit UK trade facts faithfully. You can think of that as a bit like rejigging the GTAP model so it fits the UK better, which has not been done as it is a long, complex process. Conclusions So in sum what we have found is that the Civil Service seems now to be using a world trade CGE model which is a defensible and improved methodology; but it has (a) not been tuned to fit UK trade facts and (b) it has used absurdly pessimistic Brexit policy assumptions to ‘cook’ its anti-Brexit results. According to our trade model, which fits the UK facts and assumes the Government’s announced policy assumptions, there would be a gain from Brexit of +4%. According to GTAP and Ciuriak and Xiao’s policy assumptions, Brexit costs -0.2% of UK GDP. But put in the right policy assumptions to GTAP and you get +2%. Now put in the policy assumptions of the Civil Service and it is reported we get -5%! Since we only export 12% of our GDP to the EU, one is expected to believe that we will lose the value of almost half of our EU exports. Well, there’s cooking for you!