The Corner

Do Marginal Tax Rates Matter for Low-Income People?

Megan McArdle has made the very sensible point that when poor people face extremely high effective tax rates if they take a job (e.g., when I take a job that pays $285 per week, I lose my welfare check plus other benefits worth more than $285 per week), this probably reduces the likelihood of getting a job.

The obvious approach is to create some kind of sliding benefit scale, so that for each extra dollar I earn, I get my benefits reduced by something less than one dollar. Everybody has stories they love to tell about this, and various economists have various models that purport to estimate this effect, but the best way to answer the question of what effect this would really have is to randomly assign some welfare recipients to a tax schedule that does have as high marginal tax rates, and see if this results in more of them taking jobs than of those who face the existing tax rates.

The best such randomized experiment in a comparable situation that I am aware of is the Self-Sufficiency Project (SSP) in Canada in the 1990s. Basically, a group of welfare recipients were randomly split into a treatment group that received a supplementary payment that reduced the effective marginal tax rate on income from 100 percent to 50 percent for a three -year period, and were compared to a control group that was randomly selected to receive no change to the current program. (The details are available in an excellent program final report from 2002.)

The findings complicate almost everybody’s argument:

People respond to incentives.  During the period of the reduced marginal tax rate, reported work earnings and reported income rose for the test group vs. control during the experimental period. Score one for the supply-siders (and common sense).

Marginal is not average.  At the peak effect of the program (16 months after random assignment), about 30 percent of the treatment group were employed full-time versus 15 percent of the control group. Anecdotes about X heroic poor person, or your self-analysis of your likely response to a change in marginal rates, or your speculations about what you would do if you were an entrepreneur, doctor, or dockworker don’t mean much. The whole effect here is driven by 15 percent of the treatment population — the vast majority did very little different than they would have done otherwise, yet the aggregate effects are material. The same thing applies to discussion higher up the income scale.

This cost taxpayers more money, not less.  In round numbers, as compared to control the treatment increased total reported take-home earnings by about $200 CD per month, about $100 CD of which was greater reported wage income, and about $100 CD of which was the supplemental cash transfer from the government (i.e., all the people in Canada who pay taxes) used to reduce the effective tax rate for the welfare recipients in the program.

The effect disappeared after the program ended.  After the program period (for complicated reasons, about five years after program entry), the treatment group had about the same level of reported employment and income as the control group. On one hand, this is further evidence that marginal tax rates matter for people in this situation, but on the other, it also indicates that the program failed to achieve its stated goal of “lift off” into self-sufficiency — that is, transition off the dole and into the workforce. This implies that applying this program as an operational policy would result in a perpetual increase in the welfare cost per family, in return for more work.

Grandiose solutions rarely work.  The closest analog to this kind of an experiment in the U.S. (with important differences) were the Negative Income Tax experiments of the 1960s and 70s.  For reasons I’ve gone into elsewhere, these mostly demonstrated that there is not a politically feasible way to guarantee a minimum income with a sliding implicit marginal tax rates. As I’ll get into in a lot of detail in my forthcoming book, the only method that has consistently demonstrated in the U.S. that it can humanely get people off welfare and into jobs was the workfare movement of the 1990s (which it did without increasing government costs, by the way). 

Jim Manzi is CEO of Applied Predictive Technologies (APT), an applied artificial intelligence software company.


The Latest