Have they learnt nothing?” goes the cry, as profits and pay at Goldman Sachs surge to near record levels.

The investment banks that have survived the crisis have returned to their bad old ways, paying outrageously high bonuses, that encourage excessive risk-taking at institutions that are too big to fail.

Much of this is nonsense. Profits certainly have recovered remarkably quickly and if the markets hold up for the rest of the year the bonuses will surely follow. The size of these payments will be very hard for many ordinary mortals to stomach. After all, Goldman has benefited massively from taxpayer-funded support for the financial system that Wall Street banks helped to undermine.

But, even if you believe big bonuses were a significant contributor to the crisis, there is no evidence that Goldman is encouraging excessive risk-taking.

The vast bulk of the money it is making comes from straightforward stuff such as making markets in commodities — where margins have shot up because competitors have pulled back — or helping companies to raise money from share issues. The bonuses it pays will conform to the rules drawn up by the Financial Services Authority with all the deferrals, clawbacks and the like.

But, say the critics, the fact that Goldman Sachs is too big to fail must inevitably encourage it to take excessive risks. If management screws up, the bank will be bailed out. It is extraordinary how prevalent this idea is among financially sophisticated people. If a bank is bailed out, the depositors will be protected and possibly the debt holders. But not the shareholders or the management. They will be no better off than if the bank was allowed to fail.

Two eminent Harvard professors, Lucian Bebchuk and Holger Spamann, have just published a widely praised paper in which they argue that by their nature, shareholder-owned, limited liability banks are excessively risky. Because they are so highly geared, shareholders and management would be perfectly rational to “go for broke”. The upside is so massive it outweighs the risk of the occasional wipeout.

The flaw in this analysis is that it assumes the only thing people running banks have to lose is money. This is manifestly untrue. When Lehman went down, Dick Fuld, its chief executive, lost stock that a year earlier had been worth $700 million. He also lost for ever something he probably valued a great deal more — a reputation as a Titan of American business a lifetime in the making. Only an economist could think it was rational for someone to risk that.

Goldman may be making most of its money out of plain vanilla stuff these days but it hasn’t given up on whizzy financial engineering, however unfashionable. It has come up with an ingenious-looking but complex wheeze to help RSA reduce risk in its pension funds.

Ready with the wet towels? Here we go. If liabilities in the funds blow out because former RSA staff live longer than expected or because of a nasty dose of inflation, Goldman will pick up the tab in return for a small regular premium.

Now the clever bit: the money to pay the premium is raised by swapping the income stream from the funds’ short-dated gilts with the higher-yielding income stream of longer-dated gilts held by Goldman.

Counter-party risk — the danger of Goldman going bust — is largely eliminated by Goldman having to put up more collateral if the bet moves against it.

Goldman, RSA and the pension fund trustees are all delighted with the pioneering deal, which its supporters believe could be a model for other pension funds. Market movements have made the alternative way of de-risking pension funds, fund buyouts, less attractive.

Deals where everyone turns out to be a winner are rare, even when Goldman’s rocket scientists are set to work. Any flaw in this piece of financial prestidigitation is not immediately obvious — though that doesn’t mean there isn’t one.

Times Recommends

• Lies and cover-up at MG Rover exposed by report
• High oil prices push manufacturing costs up
• Primark founder steps down as chief executive