Rachel Cohen has a piece at Vox today about a conservative welfare reform aimed at removing marriage penalties. This comes up a lot in conservative welfare circles, especially during and after Presidential elections. It is doubtful that major policy changes will be made in this area, but it is a topic worth discussing, both in itself and also as a way of illustrating the superiority of a different way of doing tax and welfare policy.

The basic problem is that, whenever you create a welfare benefit that is restricted to families or households with income below a certain level, you make it so that eligibility turns, in part, on how many earners live in a given family or household. If family income needs to be below $20,000 per year to be eligible for a benefit, then a single-earner family earning $15,000 is eligible. But if two single-earners making $15,000 get married, their combined income would go to $30,000 and they would become ineligible for the benefit.

This is distributively unfair and has the effect of preventing some people from marrying. I actually know people in my family who, for many years, cohabitated but didn't marry because doing so would result in the loss of Medicaid eligibility.

The proposal that is typically offered to solve this problem is to have one set of eligibility thresholds for singles and then a higher set of eligibility thresholds for married people. Generally, in these reforms, the married-couple eligibility thresholds are set at twice the level of the eligibility thresholds for singles.

But this doubling-of-thresholds reform can actually generate similar fairness problems in the other direction. We see this already with the federal income tax, which is set up this way. When a single person earning $200,000 and a single person earning $0 marry, their combined tax bill is cut in half. This is distributively unfair to individuals who stay single and has the effect of incentivizing marriages between highly unequal earners but not between equal earners.

There is a way to set up your fiscal and welfare institutions that avoids both marriage penalties and marriage bonuses, though this approach is rarely mentioned in the US policy discourse. The solution has three pieces to it:

1. Tax all income personally. In our current system, Medicare, Social Security, and Unemployment taxes are assessed on each individual's personal income while the federal income tax is assessed on tax units that combine spousal income and then apply different rates based on whether an individual is filing singly, jointly, or as a single parent. We should change the federal income tax so that it is assessed on each individual's personal income just like Medicare, Social Security, and Unemployment taxes are. This would make it so that tax situations do not change based on marriage.
   
2. Provide welfare benefits personally. Welfare benefit eligibility should also be based on the circumstances of specific individuals not families. Unemployed people should get unemployment benefits regardless of how much money others in their family or household earn. The same for disabled people, elderly people, children, and so on. Some of our benefits already work like this, but others do not. By providing benefits to individuals based on their personal circumstances, you make it so that their benefit situations do not change based on marriage.
   
3. Eliminate means tests. Welfare benefit eligibility and levels should be based on an individual's current status (e.g. whether they are unemployed, disabled, on leave, a student) and, for income-smoothing benefits, their prior earnings. As with (2), this ensures that family composition does not end up affecting benefit eligibility.

Of course, this just describes the ideal form of the universal social democratic welfare state. This should come as no surprise. That approach to welfare and tax policy is simply optimal across the board and departures from it invariably generate these sticky problems that are impossible to resolve but nevertheless give rise to an endless cycle of policymaker headscratching and kludgy half-fixes.