Physics and economics are bedeviled by the phenomenon known as ''hysteresis''. It's where, once made, a change can't be easily undone.
In the physical world it applies to breaking an egg. It's impossible to run the sequence of events backwards - to unscramble the egg.
In the world of social policy it applies to extending to the well-off benefits that were previously reserved for the poor.
The baby bonus is an example.
By the end of the Hawke/Keating government in 1996 Australia had one of the tightest social security systems in the world. On one estimate, 92 per cent of government payments went to the half of the population that earned the least. By the end of the Howard government a decade later many more of the payments were spilling over to the top half. A lesser 87 per cent went to the bottom half.
Professor Peter Whiteford of the Australian National University, who did the calculations, is keen to point out that even after the end of the Howard era, Australian payments were still the most tightly targeted of the rich nations that make up the Organisation for Economic Co-operation and Development. He ought to know. For many years he was in charge of social policy statistics at the OECD.
The baby bonus was one of Howard's most brazen spillovers. Instead of being directed to the parents who had earned the least it was funnelled to those who had earned the most. Mothers who had enjoyed high incomes in the year before their child was born were able to claim $2500 per year. Mothers who had been on low incomes got $500.
Two years later it became a fairer flat payment of $3000, which later climbed to $5000. But that was as fair as it got. It was paid to millionaires as well as the very poor. No means test, no cut off. Once extended to the well-off it could not easily be prised from their hands.
Kevin Rudd developed courage in 2008 and denied it to families earning more than $150,000 per year.
For his trouble he was accused of being anti-children.
''Every mother loves her baby,'' declared opposition leader Brendan Nelson. ''Every baby is valued and Mr Rudd should value all babies equally. We should not live in an Australia where Mr Rudd thinks that some babies are more valuable than others.''
Late last year, as the government's budget woes worsened, it decided to trim the bonus for second and subsequent babies. They were each to attract a payment of $3000 instead of $5000. It was rewarded with a reference to China's one-child policy.
Middle earners are also alarmed at the thought of high earners losing benefits ... They think one day that might be them.
''The government seems to want to penalise anyone that has a second or third child,'' Coalition Treasury spokesman Joe Hockey said. ''I think that worked quite well in China, didn't it?''
The private health insurance rebate was the baby bonus writ large. Overwhelmingly taken up by the middle- to high-income earners who could afford private health insurance it was budgeted to cost the Howard government $1 billion in its first full year of operation. Thirteen years on it is costing $5 billion per year. The Bureau of Statistics says Australia's highest-earning families (the top 20 per cent) get $16 a week from the rebate. The lowest earners get $5.
Labor has hacked away at the rebate since taking office, most recently means testing it so families earning more than $260,000 get nothing. That high threshold says a lot about how scared it has become about taking away a benefit well-off families should arguably never have received. It's begun a review of alternative ''natural'' therapies that attract the rebate with a view to knocking them out. (The Howard government can take credit for earlier knocking out insurance for running shoes, gym memberships and relaxation tapes.) This week the Australian Council of Social Service went further and proposed axing so-called ''extras'' cover altogether.
''Why should the government help fund high-end dental work for people with a chance of affording it, when it doesn't yet properly fund dental work for people who can't?'' asks ACOSS chief executive Cassandra Goldie.
But in assembling $6 billion of proposed savings for the May budget ACOSS has eschewed even bigger measures such as getting rid of the private health insurance rebate in entirety. ''We would need to convince powerful people,'' Goldie says. ''We would like to do more, but we need to build a consensus.''
Meanwhile, Labor has been quietly eating away at the rebate's foundations. Until now it has paid a certain proportion of whatever the health insurance funds have charged. As medical costs have soared (by 7 per cent a year) so have the payouts.
But from April next year the payout will climb only by the rate of inflation, typically just 2 to 3 per cent each year.
Over time this will mean the government will pay out much less than it would have to families that aren't particularly needy. A Fairfax Media analysis suggests that in a decade the government could be paying 30 per cent less. It will have done it without ruffling too many feathers, as it needs to if it is to take away benefits from middle Australia.
Labor has done the same sort of thing with pensions. The Howard government made it easier for middle earners to keep at least a part pension by cutting the rate at which the income test took it away from 50¢ in the dollar to 40¢. Labor has changed it back to 50¢.
The ANU's Whiteford wants Labor to go further and cut the rate at which the pension increases each six months. At the moment it climbs with male earnings. The next increase, due this month, will be $35 per fortnight. But unemployment benefits increase more slowly, with the consumer price index. It's been happening for so long that the pension is now 57 per cent bigger than the Newstart allowance and climbing. The Henry tax review recommended finding a midpoint and increasing both at a rate somewhere in between the consumer price index and wages.
''It won't be popular, but in the long run it's incredibly important,'' Whiteford says.
Not being popular is the roadblock facing every attempt to remove a piece of middle-class welfare once it is introduced. Richard Denniss of The Australia Institute uses the new insights of experimental economics to explain why losing a benefit matters far more than gaining one.
''Imagine how good you would feel if you found a $20 note on the footpath,'' he says. ''Now imagine how bad you would feel if you lost one on the footpath. The sums involved are the same, but the loss hurts much more than the gain helps. It's true of children when you take away their toys.''
And the number of middle-income Australians with benefits to lose has grown. They are better organised than low-income Australians, better able to use the political process and sometimes have entire industries such as superannuation and private health insurance behind them.
It took just a week for the Gillard government to back down on a plan floated in January to tax presently untaxed super payouts on balances of more than $1 million. A replacement plan that would more highly tax the earnings of funds held by the top 1 per cent of earners has come under attack on the unlikely ground that it might discourage saving. Treasury calculations suggest the top 1 per cent get an average super tax break of $19,200.
Middle earners get $800. Low earners get nothing.
The campaign against taxing super payouts was all the odder because an earlier Howard government measure gave all Australians over 65 an effective tax-free threshold of $60,000. Even if Labor had taxed super payouts a retiree would need to get $60,000 per year from super before he or she paid a thing.
''It's not just that middle earners are alarmed at the thought of losing benefits, they're also alarmed at the thought of high earners losing benefits because they think one day that might be them,'' Denniss says. ''It gives high earners allies.''
And it makes radical measures unimaginable.
Whiteford would include the family home in the pension means test.
''It would be political death, but aged pensions are where we spend the money,'' he says.
''You could do it at a home valuation where most people wouldn't be affected. It would stop really wealthy people getting the pension where others didn't, but it would invite hysteria from oppositions and from the media. It would be - as they say - courageous.''
Subjecting expensive family homes to the capital gains tax would also be courageous, as would ending the 50 per cent tax holiday for capital gains, neither of which ACOSS is pursuing.
Because middle-class welfare is hard to unwind, Labor has been moving slowly and for the most part quietly. Future governments might thank it, but not many people will thank it today.