The relative paucity of the modern welfare state in the United States is a well-known fact among researchers. Compared to rich countries in Europe, the United States spends far less on social insurance programs and other social programs such as education. But these large disparities decrease once the private-sector side of the U.S. welfare state is included in the analysis. Yale University professor Jacob Hacker calls this the “divided welfare state,” where in many instances the U.S. tax code is now the main vehicle for social policy in retirement, college savings, and housing.

How well has this “submerged state” worked? At least in these three areas, the effectiveness of the tax code, via deductions and credits, is questionable. Consider the state of the private-sector retirement system in the United States. Over the past 30 years, workers have come to rely more and more on defined contribution plans such as a 401(k) account. Workers who contribute to these voluntary plans are able to deduct the value of their contributions on their tax returns and reduce their taxable income. This means the value of the deduction is much higher for high earners and the tax savings from the deduction go disproportionally to those at the top of the income ladder.

Or consider the submerged state approach to high college tuitions. The cost of college is skyrocketing, leading to a proliferation of tax-sheltered savings plans. These plans, including the popular 529 savings plan, attempt to help families save enough money to pay for college. But the benefits from these plans go disproportionately to those at the top. A paper by University of Michigan economist Susan M. Dynarski finds that the benefits from tax-advantaged savings plans are highly and positively correlated with income. Those at the top are getting the most.

The mortgage-interest tax deduction is another example of policy being run through the tax code. The tax code allows homeowners to deduct the cost of the interest of their mortgage against their taxable income. The intended purpose was to encourage homeownership, which was viewed positively for a variety of reasons. But research finds that the value goes mostly to wealthier Americans while doing little to increase homeownership. If anything, it encourages those that already are buying a home to buy an even-bigger home.

To be sure, the creation of this network of tax credits and tax expenditures wasn’t without reason. Political realities necessitated the use of the tax code to achieve these ends. And these programs have done real good. But as the evidence shows, they are far from optimal.

The record of using the tax code to do tasks traditionally associated with the welfare state is clearly mixed. At best, it works like a Rube Goldberg machine that attacks a problem by hoping that a chain reaction will do the job. At worse, the machine doesn’t work for the broad majority of the population. The relevant question is now how to re-engineer it for future, more efficient use.