Are restrictions on free trade a better idea than generally thought? I ask because, despite his lauding of it, it is reported that Johnson will impose customs checks upon goods imported from the EU. This lends credence to the estimate (pdf) by The UK in a Changing Europe that Johnson's plan might eventually cut UK GDP by over six per cent.

But might this estimate miss an important mechanism and thus over-state the costs of us leaving the single market?

What follows cuts against all my instincts in favour of free trade and EU membership. My gut tells me to agree with Martin Wolf's claim that Johnson's proposals mean shooting ourselves in both feet. But we should always question our beliefs. And there is, I suspect, more to be said in favour of trade frictions than we have heard so far.

My story starts in a strange place, with the Feldstein-Horioka puzzle. This is the fact (pdf), pointed out in 1979, that net capital flows between countries are much smaller than you would imagine or, to put it another way, that current account imbalances are surprisingly small.

One reason for this is that financial markets, on their own, cannot achieve any net transfer of capital between countries This is simply because every seller needs a buyer: I can only dump my holdings of UK equities if somebody buys them. My effort to shift capital out of the UK is therefore offset by somebody else shifting it in. Instead, a net transfer of capital out of the UK requires that there be a trade surplus. This, by definition, entails an offsetting outflow of capital: if we are selling more goods and services to foreigners we are earning more than we are spending - ie saving, and these savings must be invested abroad.

So, what is it that limits trade imbalances and therefore current account imbalances? The answer, said Maurice Obstfeld and Kenneth Rogoff in a famous 2001 paper (pdf), are trade frictions. By these they mean not just tariffs but also non-tariff barriers such as the red tape of border checks, regulatory differences or simply customers' preferences (pdf) for home-made goods. It's difficult for a country to export or import a lot simply because trade costs (pdf) are high. Trade imbalances, therefore, tend to be small.

One Big Fact supports Obstfeld and Rogoff's theory. It's that the EU's single market reduced trade costs - and this led to a rise in countries' external imbalances. For example between 1980 and 1997 (when Schengen came into effect) Germany ran an annual average current account surplus of just 0.9% of GDP. Since then, though, it's run an average surplus of 4.7% of GDP. And since 1997 the external deficits of Spain and Greece have been twice what they were in the previous 17 years.

Trade frictions, therefore, reduce current account imbalances.

Which can be a good thing. To see why, remember that a current account deficit means - by definition - that a country's domestic investment exceeds its domestic saving. But think about what this means for its banking system. If investment exceeds savings, the growth in bank lending might well exceed the growth in deposits or the purchase of bank equity by domestic savers. Which means banks might well be becoming riskier - either because they are becoming more highly leveraged, or are more dependent upon wholesale funding to plug the growing gap between loans and deposits. It is for this reason that the NIESR (pdf) and Dallas Fed (pdf) have both found that big current account imbalances help predict financial crises.

The single market, therefore, might well have led to the euro crisis by fuelling greater external imbalances.

Was it really an accident that developed economies saw almost no banking crises during the Bretton Woods era, when current account imbalances were small, but saw plenty before and since?

In limiting current account deficits, then, trade frictions help preserve financial stability.

This is a great prize. Coen Teulings and Nick Zubanov have shown that financial crises lead to a huge and permanent loss of GDP. And this can be multiplied, because crises can lead to bad policy. Nick Crafts points out that the last one gave us austerity and hence Brexit. "The economic costs of the banking crisis are much larger than is usually supposed" he says. An important transmission mechanism here has been proven by Ben Friedman: economic stagnation, he has shown, leads to intolerance, xenophobia and anti-democratic sentiment.

Against all this is the fact that trade frictions create lots of deadweight losses. But even summed across countless goods, these can be small. So much, in fact, that Arnaud Costinot and Andres Rodriguez-Clare have estimated (pdf) that if the US were to shift to autarky it would only cost it 2-8% of GDP: for the UK, the cost would be around four times as much but for the EU much the same. As James Tobin used to say - rightly - "it takes a heap of Harberger triangles to fill an Okun gap." And an Okun gap is what we get from a banking crisis.

Of course, Harberger triangles aren't the only cost of trade frictions. There is also the fact that, in reducing competition, they retard productivity and innovation. But banking crises also do so. It's a wash.

So, what can be said against this argument for trade frictions? I don't think it's sufficient to say that banks are sufficiently strong now that a crisis is unlikely. Even if this is true now, it might not remain the case in coming years. Over long enough periods, small probabilities become big ones.

Instead, you could argue that we are trading off a certain loss - all those Harberger triangles - for what is only a reduced probability of disaster. Your attitude to this trade-off will depend upon your taste for risk.

A stronger argument though is simply that there is a cheaper way than trade frictions of reducing the probability of a crisis - to impose higher capital requirements upon banks. But what if the power of the banking lobby prevents this?

All this suggests there might be more to be said for trade frictions than supposed. Which poses the question: why have we not heard an argument along these lines, when we have heard so many much weaker ones?