Senator Mike Lee's tax plan

9/19/2013 05:35:00 PM
Everyone's abuzz about a tax proposal from republican Senator Mike Lee. Here are the basics:
  • \$2,500 per-child tax credit (in addition to existing \$1,000 credit)
  • \$2,000 per-person tax credit (non-refundable, replaces personal exemption)
  •  eliminate all other deductions, including state and local tax deduction 
  • only two tax brackets: 15 percent and 35 percent
First things first. No one has any idea what the budget impact of this would be--it has not been scored by the CBO (can't score it until someone names income thresholds) but my guess is it would reduce revenues. Though it eliminates some deductions, it also creates massive new ones while presumably lowering most people's marginal tax rates (unless the income threshold for the 35 percent rate is very low). As far as I know, no one on either side is claiming this would be anything but revenue-reducing.

Matt Yglesias Elaine Maag reports Lee would support making the per-child credit a spending program, but this conflicts with what I've read on the plan, which says the credit would be non-refundable and therefore not the same as spending. 

Now, I could support a per-child tax credit for low income families. But I'm still trying to wrap my head around the conservative argument for it. They claim that the current tax code is biased against parents, since non-parents free-ride on the social insurance contributions (ie Social Security and Medicare) of children they didn't parent. The idea is that parents pay twice: they contribute to Social Security and Medicare through payroll taxes, but also pay the expenses involved in parenting children, who will eventually grow up and pay into Social Security and Medicare too. This strikes me as silly: surely we should structure Social Security and Medicare to accommodate the fertility rate, and not the other way around. It also seems like double counting the contributions of parents: the average american pays in about as much as he receives from Social Security over his lifetime--hence, non-parents aren't taking money from parents. But, this is a philosophical issue, not a policy one. A scientist can potentially discuss the welfare impacts of a policy, but has no authority to speak to its fairness. On welfare impacts, I think a child tax credit would be welfare-improving because (as Yglesias notes) for two families with the same incomes, the one with more children is going to have a larger marginal utility of wealth. But, true to Mirrleesian form, we should weigh this redistributionary gain against the incentive effects--there are potentially large negative externalities from over-population, so we may not want to incentivize parents to have too many children.

I have two main criticisms of this proposal. The first is that we should keep the state and local tax deductions. I explain why here. In summary, not deducting state and local taxes at the federal level basically ensures that some people will face a larger than 100% marginal tax rate. That is clearly and unambiguously to the right of the peak of the Laffer curve, and therefore inefficient.

The second criticism is that we should increase, not decrease, the number of tax brackets. Now, I'm aware of a paper in the optimal tax literature that argues that an optimal progressive tax can be "approximated" by a two-bracket system, but my question is, What's the point? Everyone assumes that fewer tax brackets represents a "simpler" tax code, and we all want a simpler tax code. But I disagree that this really is simpler--no one calculates tax brackets by hand! Most of us do taxes online for free--enter a number, it spits out your tax liabilities--that's not going to be any simpler no matter how many or how few brackets there are. And it was always this easy, even before computers: every year the IRS issues a tax table--your income is listed in one column, the taxes you owe in the other--again, this can't get any simpler.

My point here is that the complexity of the tax code comes from the myriad of different tax forms, deductions, alternative minimums, payroll versus income taxes, but NOT from the number of brackets in the Federal Income Tax. Thus, we are strictly worse off with a "simpler" two-bracket system than under the optimized Mirrleesian tax function. If, however, you irrationally value algebraic simplicity over real honest-to-goodness simplicity, then may I suggest a progressive tax function with absolutely no tax brackets: $$\tau=cY-aY^b$$ where $\tau$ is the taxes you owe, $Y$ is your income, and  $a,b,$ and $c$ are parameters satisfying $c,a>0$ and $1>b>0$. In one of my papers, I proved a theorem showing that any tax code based on discrete marginal tax brackets can be well approximated by a function of the form above.