The majority of workers don’t have employers who offer them retirement accounts, and the ones who do are lucky: They’re the ones the U.S. Treasury helps the most. In a given year, more than $120 billion in federal money goes towards subsidizing retirement plans, mostly 401(k)s and IRAs. States have some similar rules, which result in an additional $20 billion a year of these subsidies, which come in the form of tax breaks that are the second-biggest tax expenditure, after health insurance. (These tax breaks are larger even than those devoted to mortgage deductions.)

On the surface, this seems like a good thing: The government recognizes that saving up for retirement is important, so it’s incentivizing workers to do that by letting them pay less tax on what they put into their retirement accounts. The problem, though, is how the perks of these tax breaks are distributed: Most of the subsidies go to the top earners, the middle class gets the meager remainder, and taxpayers across the board foot the bill.

The reason that these benefits are distributed so unevenly has to do with the fact that they come in the form of tax breaks. For someone who earns a lot of money and is thus in a higher tax bracket, not having to pay taxes on, say, $2,000 is going to be a lot more valuable than it is for someone who earns less money and is thus in a lower tax bracket. As a result, the highest earners, who tend to max out their contributions to retirement accounts, get about $7,000 per year in tax breaks, and the middle class averages about $400 per year. Low-wage workers, who usually aren’t putting money into these accounts, usually get nothing.

These subsidies are so ineffective that 80 percent of the money devoted to them goes to only 20 percent of workers. Two examples illustrate how that happens.

Take someone who falls into the 15-percent tax bracket. She’s not a stranger—77 percent of Americans are in that bracket or below. If she contributes $100 to her retirement account, the tax break works such that she doesn’t pay tax on 15 percent of what she’s contributing—she’s only taxed on $85, which means she’ll pay about $6 less in tax than she otherwise would have. That’s nice for her, and it’s not a big loss for the government.

But the math is different at the higher end of the spectrum. Take someone who is taxed in the 35-percent bracket. Just like everyone else, the benefit she gets from contributing to retirement accounts is that she won’t have to pay taxes on some of what she stashes away. But because she won’t have to pay 35-percent tax on part of it, that’s worth more than for someone who is in a lower bracket. So if she contributes $200 (it’s not a leap to say she’ll deposit twice as much as the first example worker) she’ll get a tax benefit of about $34—not two times $6, but almost three times as much.

On top of that, when she withdraws this money during retirement and pays taxes on it, she’ll likely fall under a lower tax-bracket, perhaps closer to 15 percent. So her contributions were deducted at a rate of 35 percent, but her withdrawals are only taxed at 15 percent. Taking this into account, the value of the tax subsidy balloons to nearly $75. This all adds up quickly as one moves up the earning spectrum, so it’s not surprising that 80 percent of the tax subsidy is going to households whose earnings are the in the top fifth.

This seems unfair, but a counter argument might be that even higher earners need to have retirement savings, and isn’t $120 billion worth it to society to encourage people to put money toward their future security? It’s a nice idea, but research suggests that these high-earning households would have put money into their retirement accounts even if they hadn’t gotten the tax break. The subsidy is a nice present to people who don’t need it at the expense of those who do.

At the moment, then, the tax breaks’ main accomplishment appears to be creating jobs for the accountants who shuffle around their clients’ funds to avoid paying taxes. It should be put to better use. $120 billion a year, divvied up equally, works out to about $800 per taxpayer. What if that amount was what the government put into everyone’s 401(k)s, instead of letting the amount be determined by whether a worker falls under the 35-percent bracket, the 15-percent bracket, or another?

The benefits of this would pile up over time. If $800 a year went in early on in a worker’s career, she’d end up 40 years later with something in the neighborhood of $100,000—even if she didn’t put in any money herself. While that still falls short—it’s recommended to have five to eight times that much saved up—it would give the 60 million people without retirement accounts a starting point.

This reform wouldn’t, on its own, fix a broken retirement system, but it’d be a start—one that wouldn’t even require the government to spend any more money than it currently does.