Towards Better Economic Modeling of Trade Agreements

October 25, 2020

Summary of CPA Testimony to the ITC on Oct. 6, 2020

By Jeff Ferry, CPA Chief Economist

On Oct. 6th, CPA submitted oral and written testimony to the US International Trade Commission on the economic modeling of trade agreements (TA’s).

Our testimony focused on how economic modeling of trade agreements can be improved to deliver more accurate results. We included two specific modifications. First was a more accurate estimate of how much US imports rise in response to a cut in US tariffs. Our second modification involved estimating the amount of unemployment created by increased import penetration. Our case study demonstrated that these modifications produce a more accurate model when comparing against the actual economic results of, in this case, the trade agreement with South Korea.

We were responding to the ITC’s request for comment relating to their investigation into the performance of trade agreements. The investigation was required by Congress in the  Bipartisan Congressional Trade Priorities and Accountability Act of 2015. This investigation focuses on all trade agreements from NAFTA and the Uruguay Round (completed in 1993) through to today. The ITC will submit the results of its investigation in a report to the House Ways and Means Committee and the Senate Finance Committee by next June 29, 2021.

Many critics of trade agreements have argued that the agreements have worsened rather than improved economic outcomes in the US. In our testimony, we took a different tack from the past and focused on the economic modeling of the agreements, because those economic models and their forecasts form an important part of the debate around each trade agreement. And we went further than we and others have gone in the past, to actually show how economic models of trade agreements can be improved and turned into forecasting tools that are worth paying attention to.

In our ITC testimony (the full text is available here), we set the stage by recounting the history of how economic models of TA’s have consistently been wildly over-optimistic. These models have forecasted improved US trade balances and boosts to US gross domestic product from virtually every trade agreement they have modeled. Instead, major trade agreements like NAFTA have delivered much larger trade deficits for the US, along with millions of lost jobs, deindustrialization, and other economic harm. We quoted respected trade economist Professor Timothy Kehoe who, despite being a supporter and user of trade models, commented that they have proved to have “zero predictive accuracy.”

We explained the main reason why trade models have failed to demonstrate predictive accuracy. Trade modeling is done using the GTAP model developed at Purdue University and now widely used by economists worldwide. The standard GTAP model allows trade to influence national economies through only one channel, that of prices. GTAP was designed to model the effect of a tariff reduction reducing prices of imports, which then stimulates increased consumption of those goods, expanding the national economy. Conversely, higher tariffs would raise prices, which would reduce demand and negatively impact the size of the national economy. All the other possible effects from a trade agreement are excluded from the model. Standard GTAP excludes the possibility of a trade agreement reducing employment when imports take a larger share of the domestic market. It excludes the effect of a TA on foreign direct investment, such as when US auto companies responded to NAFTA by moving the production of millions of automobiles from the US to Mexico. It also excludes the effect of a TA on stimulating the import of products or product categories that were not imported before. In short, the standard GTAP model is effectively designed from the ground up to show only benefits from lower tariffs.

We prepared this testimony in partnership with Badri Gopalakrishnan. A PhD economist, Badri is an expert in GTAP. He played a key role in developing GTAP as a fellow at Purdue and his contribution is recognized on the GTAP website. Badri and CPA share the conviction that economics can play a positive role in understanding trade, provided that economic models (and ultimately economic theory) are modified to better reflect reality.

Improving the Model

We modified the standard GTAP model in two ways to reflect reality. Our first modification involved elasticities of demand. This technical term refers to the change in demand or sales for each 1% change in the price of a good. The standard GTAP model assumes that all nations respond to a change in the price of an import by the same amount, i.e with the same elasticity. That is not the case in the real world.

For our case study, we used the KORUS trade agreement with South Korea, which took effect in 2012. Looking at the actual data on imports in the US and Korea, we found that the US economy has an elasticity of demand for imports 7.5 times higher than the Korean economy. So a 1% reduction in import prices would have a 7.5 times larger effect on US imports than on Korean imports. We modified the elasticities built into the model accordingly. We pointed out that in the lengthy ITC report on the KORUS deal, the ITC noted that the South Korean economy has an “anti-import bias.” However, it did not modify its GTAP model to reflect that bias.

Secondly, the assumption that trade agreements do not cause any unemployment is highly unrealistic. We modified this assumption too. We were able to use the data from standard GTAP modeling of KORUS showing that 582,000 lesser-educated workers would change jobs as a result of the KORUS agreement. We estimate that one third of those workers, or 194,000 workers, would remain unemployed for a period up to two years.

Both of our assumptions lead to loss for the total US economy which overwhelms the gains from lower prices. Our model showed a worsening of the US-Korea trade deficit by $9.9 billion a year, just shy of the actual result of a bilateral deficit that was $10 billion worse (looking at the average of the years 2015 to 2019). This widening of the trade deficit translates to a loss of 60,000 jobs and a decline in GDP of $4.4 billion.

Our work shows that there is nothing inherently wrong with trade models that makes them misleading. They can be adapted to better reflect the way different national economies behave and produce better forecasts. With models customized to reflect the real world, we can begin asking the relevant questions, such as:

  • Which trade agreements will actually benefit the US economy and which ones will not, and why?
  • Which industries will benefit from trade agreements, and which industries will lose out?
  • Which industries will suffer unemployment, and which industries will hire more workers?


At the testimony, the ITC showed great interest in our presentation. Several commissioners asked us questions about our modeling. ITC staff followed up afterwards with further written questions, which Badri and I are now addressing. We are looking forward to a fruitful dialogue with our colleagues at the ITC.


trade Trade and Tariffs

Showing 2 reactions

  • Franklin Kirkland
  • Jeff Ferry