The use of data at the plant level is an “important innovation,” the author argues, with many advantages over more aggregated data — including firm-level information — because many companies that petition for antidumping protection are large manufacturers that produce a wide array of items. By examining individual plants, which tend to produce a narrower range of products, the author could more accurately match up products named in antidumping investigations and the specific facilities that make them.
Similarly, the author could drill down to determine which plants did and did not receive protection; the Census data reports the full list of products manufactured at each plant, their value, and often their quantity, as well as which goods were shipped where. “The availability of this plant-product-level data represents an additional level of disaggregation beyond the ‘major industry’ codes generally used to identify plants and firms in micro-level datasets,” the author writes.
In his regression analysis, the author found that antidumping protection leads to an average increase in revenue productivity of 3 to 7 percent, depending on the industry. However, the effect of antidumping tariffs on physical productivity is vastly different.
The number of products made actually declines by a greater amount as the effective rate of antidumping duties protecting the plant rises. These rates vary across importing countries, in line with the importance and share of the market commanded by the product in question. For every 1-percentage-point increase in the effective antidumping duty rate, physical productivity declines by 3 percent, according to the author. Furthermore, the author found that increases in the duty rates are connected to both price hikes and markups. On average, he writes, a 1-percentage-point hike in the antidumping rate leads to an average increase in the markup cost of 0.3 percentage points.
Finally, the author explored whether antidumping protection revealed a hidden danger for firms: namely, that the artificial price hikes allowed for the prolonged operation of low-output plants that would otherwise have ceased production. The author examined, first, whether tariff protection made plants less likely to stop producing a guarded product, and second, whether plants that halted manufacturing had lower productivity levels than those that continued to operate.
The data set allowed the author to define “production-stopping” by a plant’s choice to either close its doors or merely drop a product. This enabled the study to account for plants that produce many goods and drop a protected product but remain open.
The author found that antidumping protection makes plants between 10 and 15 percent less likely to halt production of a particular product; plants that do stop production are 15 percent less productive in terms of revenue and 10 percent less productive in terms of physical units shipped.
“Combining the results from these two steps, I find that antidumping protection allows more plants to continue production of protected products than would be the case without protection, which leaves low productivity plants active in the industry,” the author writes. “This continued operation by low productivity plants, therefore, contributes to the relative decline in physical productivity found.”
The findings provide new insights about how firms in a developed economy respond to major tariff shocks, the author concludes. More generally, the results indicate that antidumping protections can be a double-edged sword. Because tariffs are more likely to lead to higher prices than higher productivity, companies must make sure to allocate resources to plants that remain efficient and that are not compensating for lower productivity with temporary price hikes.
The apparent increases in revenue for U.S. firms protected by antidumping duties are driven primarily by price hikes and markups. Physical productivity at plants actually decreases, and antidumping protection brings with it the danger of allowing underperforming plants to continue operations.