Almost every week, the media carries a story based on a new study about the large damage being wrought to the UK economy as a result of Brexit. This tsunami of ‘evidence’ is bound to have affected public views on Brexit and polls now show a majority of voters supporting rejoining the EU. We should not place too much weight on this, since polling by Professor Matt Goodwin of Kent University found that if respondents were told that rejoining would involve adopting the Euro and resuming free migration etc., support falls to close to levels seen in 2016. Even so, the barrage of flawed evidence is damaging the pro-Brexit case and should be countered.
The latest example appeared last week when much of the press and media, including the BBC, blazoned the headline that lower investment in the UK since the Brexit referendum had resulted in a loss of £1000 a year to each family. The claim came from Jonathan Haskell, a member of the Bank of England’s Monetary Policy Committee which sets the Bank’s base interest rate. Talking to City-AM Haskell said that ‘the sharp reduction after the Brexit referendum has dealt a £1000 blow to each UK household’. He asserts that the productivity penalty caused by a huge drop in capital spending since the 2016 referendum has led the UK economy to be £29 billion smaller than it would otherwise have been.
Haskell followed up his interview with a twitter thread explaining how he reached this conclusion. This thread reveals that his analysis was based on assumptions rather than hard evidence. He said that business investment has ‘flatlined since the referendum in 2016’ and he estimated how much higher it would be today if it had continued to grow at historic rates. He notes the rapid growth in business investment in the recovery following the banking crisis of 2008/9, but feels (rightly) that rapid growth was unlikely to have persisted beyond 2016. He plumps instead for a long-term average growth rate of 2% per annum starting from the peak of the recovery in 2016. His assumed growth path between 2016 and 2019, he says, is similar to that in other major economies.
This is the core of Haskell’s calculation. He asks what would be the consequences for GDP and family incomes if business investment had grown at 2.0% per annum until 2019 from the peak of its post-2009 recovery in 2016 instead of flatlining after 2016 as it actually did. Beyond 2019 he assumes that business investment would have fallen due to covid and takes the observed quarterly growth rates over the period 2020-22 but adjusts these upwards by the equivalent of 2% each year. He then calculates the economic losses reported above by comparing his assumed growth path with what actually happened.
He calculates a cumulative loss of business investment amounting to 3.5% of GDP and then makes a simple calculation to estimate what the extra investment would mean for GDP. To calculate the impact on family incomes he simply divides the extra GDP (£29 billion) by the number of UK families to get his figure of £1000 per family.
The first thing to note is that is calculation is conditional. He is saying that if business investment had grown at a faster rate than it actually did, then most people in the UK would be better off. This was not, however, how his conclusions were reported. The BBC headline for instance was “Brexit hit UK investment by £29 billion says Bank of England Policymaker”. The BBC quoted Haskell as saying that ‘a wave of investment stopped in its tracks following the 2016 vote’ and that ‘the UK had suffered much more of a productivity slowdown because of Brexit’. Fourteen paragraphs into its online account the BBC says that Haskell “referred to a calculation to show what the UK economy could have looked like if it has continued growing at the rate it had been before the referendum compared to what it is currently growing at”. This is the only mention of conditionality. Most of the reports, both online and on its TV news, presented Haskell’s conclusion as fact rather than supposition. Some other media followed the same approach without any mention of conditionality. The Guardian did briefly describe what Haskell had done but did not stress the conditionality and concluded that the work ‘is likely to fuel concerns that the Brexit vote has done irreparable harm to the UK economy’. Only the Spectator described Haskell’s analysis as deeply flawed.
Interestingly, the Treasury was quoted as saying that it did not recognise Haskell’s figures. The Treasury has been under orders to avoid any predictions about Brexit ever since a group of pro-Brexit economists (including myself) met Chancellor Sajid Javid and Treasury Secretary Rishi Sunak in Downing St. in 2019 to criticise the Treasury’s Brexit modelling. This inhibition on joining the attacks on Brexit appears to be still holding.
Pointing out the conditionality and the assumptions underlying any estimates of the impact of Brexit is important, but so too is the reasonableness of the assumptions themselves. Is what Haskell did reasonable? Well no, it isn’t. There are several things wrong with Haskell’s analysis:
- Firstly, Haskell used out of date statistics. The ONS has recently revised its data for business investment adding 10% to the most recent year. The trend now looks more favourable and has regained its pre-covid level. Using the revised data would reduce Haskell’s estimates of reduced investment by 12%.
- Secondly, business investment is inherently cyclical but Haskell assumed that, after a long cyclical upswing 2009-16, investment would continue to grow for years interrupted only by the Covid pandemic.
- Haskell’s trend for the 1997-2016 period is very much higher than the observed level of investment through the second half of this period which coincides with the banking crisis and its aftermath (see chart above). His trend is close to that of the boom years before 2008. GDP and productivity have not continued their pre-2008 trends and are now 20% below the pre-crisis trend. In the period 2009-16, i.e. between the banking crisis and the Brexit referendum, business investment recovered much faster than GDP. The ratio of business investment to GDP thus experienced its largest sustained increase since the 1960’s, against the long-term tendency for this ratio to decline. It seems unlikely that this counter-tendency rise in the ratio would have continued much longer.
- Haskell’s assumed trend of 2% per annum is the recorded growth rate in the level of investment between two points in time. These two points are 1997q3 and 2016q2. This means that his trend must hit the level actually recorded in 2016q2 and as the chart above shows, this is the case. There is thus no possibility, using this approach, for the mid-2016 level of investment to be above or below the long-term trend. It has to be on-trend by assumption. This is important since if, for instance, the mid-2016 level was well above the long-term trend then we might have expected a cyclical downturn to begin soon afterwards. The observed flatlining of business investment would not then have been remarkable. Nor would it be indicative of a Brexit effect.
If instead of using a simple growth trend from the start to the end of the 1997q1-2016q2 period, we fit a best-fit trend to the data through the period, the result (shown in the chart below) is very different. The trend growth rate is much lower (at 1.3% per annum instead of 2% per annum) and the observed mid-2016 level of investment is well above this trend. In this context, the flatlining of investment after the referendum looks quite positive. Investment later declined during the Covid lockdowns but by the end of 2022 has recovered to a level once again above this trend. Haskell predicts that investment would have grown faster during the pandemic and afterwards, but this is pure assumption.
A similarly subdued trend is calculated if we use a longer trend. The ONS data for business investment exists only back to to 1996 and includes only part of the investment of public corporations. ONS data for earlier years is described as ‘indicative’ and based on ‘limited methods’ .Because of the privatisations of previously nationalised industries the ONS definition of what comprised ‘business’ changed a great deal in earlier years. For this analysis we have combined private business with all investment by government owned corporations. Using this definition, the growth of corporate investment from 1971-2016 is 1.3% per annum, or more reliably 1.6%per annum from 1987-2016, in both cases well below Haskell’s 2% per annum. A simulation from 2016 using our UKMOD macro-economic model predicts a downturn after 2016. Again, the conclusion is that flatlining investment from 2016-19 was a rather better performance than might have been expected.
Haskell also argues, correctly, that his counterfactual growth trend for investment is similar to other major economies between 2016 and 2019 and hence the UK had the slowest growing investment of the G7 economies in these years. As we argued above, this followed a six-year period in which the UK had the most rapid growth among these economies except for the USA. Several of the major economies were thus catching up with the UK in the years immediately after the referendum. Since 2019, when the UK actually left the EU, investment in the UK has grown as fast as as the other G7 economies except for the USA and France.
We can illustrate this by comparing the UK and Germany. German investment continued to grow after mid-2016 unlike the UK following a long period in which German growth lagged well behind the UK and can be considered a period of German catch-up. German investment subsequently fell less than in the UK during the pandemic but has subsequently recovered less. The chart below (which goes up to 2022Q4) shows little sign of any sustained negative impact of Brexit when compared with Germany. Investment in the two countries has advanced at a similar long-term rate, albeit with somewhat different cycles.
In conclusion, Haskell’s calculation of £1000 hit per family due to the impact of Brexit on business investment is an estimate conditional on a projection which is itself based on a number of unconvincing assumptions. Business investment increased over a 26-quarter period up to mid-2016 with no recession in investment. Haskell’s assumption that investment would have continued this expansion for another 15 quarters, right up to the pandemic is pure assumption. So is his further assumption that without the Brexit referendum, the performance during and after the pandemic would have been better than actually recorded. We cannot know what would have happened in the absence of Brexit and Haskell makes no attempt to defend his assumption. Despite all of this, the £1000 hit was widely reported as uncontroversial. Such superficial analysis and sloppy reporting really must stop.
Notes: Haskell twitter Jonathan Haskel on Twitter: “The costs of Brexit – a simulation using official data, following interview with @M_C_Klein. These are new calculations I’ve done with @JoshMartin_econ, and reflect personal views and not those of the @bankofengland or the Monetary Policy Committee. 1/10” / Twitter