I show that eliminating tax havens has unintended consequences for the domestic economy by studying a policy that limited profit shifting by U.S. multinationals. The repeal of section 936 of the Internal Revenue Code phased out the ability of U.S. multinationals to shift profits to affiliates in Puerto Rico. Under section 936, U.S. corporations had an incentive to shift profits to Puerto Rico, since the section essentially excluded the income of Puerto Rican affiliates from corporate taxes. Its repeal raised the effective corporate tax rate of exposed firms by about 6 percentage points. Since firms exposed to section 936 employed close to 11 million workers in the continental United States and were responsible for 16.1 percent of all investment by publicly listed firms, this fiscal shock had the potential to cause lasting disruptions in the U.S. economy.
Two sets of empirical findings demonstrate that repealing section 936 had real effects on the U.S. economy. First, I use firm‐level data to show that section 936 firms responded to the repeal by lowering domestic investment and employment. I find that, relative to nonexposed firms, section 936 firms reduced global investment by 10–14 percent, increased the share of investment abroad by 12–16 percent, and reduced U.S. employment by 6.7 percent. These firm‐level responses had significant effects on the U.S. economy because exposed firms reduced their payroll by 720,000 jobs.
My second empirical finding is that firm‐level effects are geographically concentrated and have persistent impact on more exposed areas. I find that, 15 years after the repeal of section 936, more exposed local labor markets saw slower growth in employment, income, wages, and home values, and that these areas became more reliant on government transfers. In particular, moving a location from the 25th to the 75th percentile of the distribution of exposure to section 936 reduced employment growth by 7.2 percentage points—about 35 percent of the growth in nonexposed locations.
That limits to profit shifting have significant, unintended consequences on the real economy revises the view that profit shifting mainly distorts revenue. In addition, these results reveal the importance of including regional effects in cost‐benefit analyses of tax reforms, as fiscal shocks can have acute and long‐lasting effects on local economies.
I establish these findings in three steps: I first construct a model of multinational investment and profit shifting. Since profits generated in high‐tax countries may be shifted to low tax locations, firms with access to tax havens face a lower effective cost of investing in high‐tax countries. Repealing section 936 eliminates access to a tax haven and raises the effective cost of investing in the United States. The model predicts that multinationals would respond by lowering investment and employment, and by shifting economic activity to foreign affiliates.
Second, I test these predictions using firm‐level datasets. I compare the investment and employment decisions of firms exposed to section 936 to those of similar control firms. Using data from Compustat, I estimate that exposed firms decreased parent‐level investment by 10–14 percent. While this is a substantial response, it implies a semi‐elasticity of investment to taxes of −2.3, which is consistent with estimates from the literature. In addition to lowering overall investment, firms reduced investment in the United States by increasing their share of investment in foreign affiliates. These investment responses were accompanied by a decline in firm‐level employment. Using the National Establishment Time Series (NETS) database, I find that exposed firms reduce U.S. employment by 6.7 percent relative to control firms. The data fit the predictions of the model and show that firms viewed the repeal of section 936 as an increase to their effective costs of investing in the United States.
Third, I show that firm‐level responses had persistent effects on local labor markets. The establishment networks of firms exposed to section 936 allow me to measure the degree to which each local economy was affected by this limit on profit shifting. I relate the fraction of establishments in a given county that could benefit from section 936 to data on employment growth at the industry‐county‐year‐level from the Quarterly Census of Employment and Wages.
I find that increasing a county’s exposure to section 936 from the 25th to the 75th percentile of the exposure distribution reduces employment growth by 7.2 percentage points and income growth by 12.5 percentage points. I also find a decrease in wage rates that is concentrated in the wages of workers without a college degree as well as decreases in housing prices and rental costs. Finally, more exposed areas receive an additional $16 in unemployment benefits per person and $30 in income replacement programs per capita. Overall, the repeal of section 936 had acute and long‐lasting impacts on U.S. local labor markets.
The key identifying assumption behind the local labor market effects is that exposure to section 936 is not correlated with other shocks that could affect local labor markets. Several pieces of evidence justify this assumption: First, I show that locations with different degrees of exposure have parallel trends prior to the repeal of section 936. Second, the dynamic effects match the phase‐out of section 936 and the effects stabilize after it is fully phased out. Third, the preferred specification controls for industry‐year fixed effects rule out concerns that the observed change in employment is due to industry‐specific trends. Fourth, exposure to section 936 is not related to a battery of potential confounders, and the results are robust enough to include them as controls. Fifth, I find similar effects when I analyze different geographical units (counties, commuting zones, and Consistent Public Use Microdata Areas); when I study changes in the employment‐to‐population ratio, which rules out the possibility that migration accommodates labor demand shocks; and when using data from several sources. Sixth, “placebo tests” show that the results are not driven by exposure to firms with similar demographics as section 936 corporations, which rules out that the results are driven by changes in the firm‐size distribution; or by exposure to the pharmaceutical industry, which was a heavy user of section 936. Facing these facts, the most likely scenario is that the economic decline in more exposed local labor markets is driven by the firm‐level responses I document in my research.
I conclude my analysis of section 936 by reconciling the firm‐level effects on employment with the local labor market effects. If firm‐level responses to the repeal of section 936 interacted with agglomeration effects or affected suppliers of inputs or local nontradable industries, then layoffs at section 936 corporations could have had spillover effects that amplified job losses within local labor markets. The results suggest that every section 936 layoff led to the loss of an additional 3.8 jobs in the local labor market. In addition, an event study across industries shows that tradable employment was the first to fall in response to the repeal of section 936 and was followed by the nontradable and construction sectors.
The key takeaway of my research is that limits to profit shifting affect real economic activity in the United States. While this result does not mean that tax havens are desirable, it points to a tradeoff between tax enforcement and economic outcomes.
This research brief is based on Juan Carlos Suárez Serrato, “Unintended Consequences of Eliminating Tax Havens,” NBER Working Paper No. 24850, July 2018, http://www.nber.org/papers/w24850.