Should the US raise taxes on the top 1 percent of earners? One of the country’s top supercomputers has been tasked with answering this question. Early findings are sparking debate about effective taxation strategies and the role of quantitative modeling in guiding economic policy.
The top one percent of households, those that make $365,000 per year or more, pay a marginal tax rate of 42.5 percent (2010 figures). Prominent economists Peter Diamond and Emmanuel Saez recently proposed raising the marginal tax rate on these top earners from 42.5 percent to 73 percent, arguing that taxing top earners at this higher rate produces the maximum government revenue.
A different team of economists, Alejandro Badel and Mark Huggett, elected to take a deeper look at this issue. They used numerical methods to simulate a complex economy to assess the consequences of increasing the tax rates on top earners. Using the Stampede supercomputer through an XSEDE allocation, the researchers came to the conclusion that the optimal tax rate is much lower than the 73 percent target. They also found that increasing tax rates would negatively impact potential top earners.
The effort led to their collaborative research with the Texas Advanced Computing Center and subsequent working paper, Taxing Top Earners: A Human Capital Perspective.
The model economy developed by Badel and Huggett is calibrated to match the US earnings distribution, where earnings and wealth vary widely across the population.
“When one builds an economic model, a typical assumption is that agents make best choices to maximize their welfare,” said Huggett an economics researcher at Georgetown University. “Solving a maximization problem by computer can take a particularly long time if the number of possible states of individuals are very large.”
“We need resources like Stampede to first compute economic agent decisions about working, learning, and saving, and then simulate those results per the millions of households in the U.S. to determine aggregate behavior,” added Badel, an economist at the Federal Reserve Bank of St. Louis.
The vastly different results arrived at by the two research groups were partly due a consideration of human capital accumulation, which if reduced in the high income group would negatively impact government revenue. According to the argument, potential high earners invest heavily in skills development even if it comes at the cost of revenue, but higher taxes would make them less willing to make this trade-off.
Badel and Huggett are still running simulations to determine the optimal tax rate, but initial simulations of the model economy arrived at 53 percent as the optimal tax rate for top earners, higher than the current rate of 42.5 percent but lower than the 73 percent rate proposed by Diamond and Saez.
Besides stimulating conversation about income disparity and government policy, the research has economists and policymakers discussing the relative merits of complex quantitative models versus more traditional economic models.
Makeda Easter, Science and Technology Writer for the Texas Advanced Computing Center (TACC), has the full story.