How much priority should the United States place on preserving NAFTA? Have certain states benefited more than others from NAFTA? Economists have spent the past 20 years working to answer these questions. They are hard to answer because of all the confounding factors.
A rough summary of the evidence is that NAFTA is responsible for about 25% of current U.S.-Mexico trade, and that it helped U.S. exports more than Mexico’s. NAFTA may have shifted some economic activity to border states, but not as much as expected. Economic growth rates — including strong growth in Sunbelt states — and exchange rates account for much more trade growth than do changes in trade policy.
Overall Impact on Trade
To measure the impact of NAFTA, we need a way to disentangle these other trends. Many studies have tried to do this — and counter-intuitively many found no NAFTA effect on trade (Table 1). The three most recent studies represent advances in methodology that stand out above the rest, and they find a modest positive impact of NAFTA on trade (1).
Table 1 presents the results of every study that has examined NAFTA directly (2). The original studies present their results in many formats: real and nominal trade growth rate due to NAFTA for different subsets of traded goods and services, growth of a country’s trade share in its partner’s market, change in the rate of trade growth due to NAFTA, and percent increase in trade. Confusingly, these sound similar but are distinct. For this reason I convert them all into the same units, the percent of end-of-period trade due to NAFTA. For the question at hand today, how important has NAFTA been, those are the units that most directly provide an answer.
Another important point in interpreting the body of research on this topic is that the share of trade due to NAFTA was measured up to different stopping points in the various studies. How does that shape the appropriate interpretation of the point estimates? Presumably NAFTA’s influence on trade has not increased since 2005, the most recent stopping point. Most evidence indicates the impact of NAFTA tapered off after 2000, when most tariff reductions had been phased in and China joined the World Trade Organization (WTO). Figure 1 shows that Mexico’s share of U.S. trade in both directions plateaued around that time, only rising again after the global financial crisis.
Neither has NAFTA’s influence on trade likely gone to zero. A simple approach would be to take these point estimates as the amount of today’s trade due to NAFTA. That assumes the impact of NAFTA has continued proportionally since the studies stopped. This should probably be considered an upper bound.
Table 1 — Estimates of the Impact on NAFTA on U.S.-Mexico Trade
| Author(s) | NAFTA Share of U.S. Exports to Mexico (% of end-of-period total) |
NAFTA Share of Mexico Exports to U.S. (% of end-of-period total) |
Data Time frame |
NAFTA Identification | Data Type |
|---|---|---|---|---|---|
| General Equilibrium Models | |||||
| Caliendo & Parro 2015 | 48 | 27 | 1993-2005 | Tariff Changes | Sector-level goods trade with intermediate goods |
| Romans 2007 (NAFTA share of total trade) |
9 | 1989-1999 | Tariff Changes | Product-level goods trade |
|
| Gravity Models | |||||
| Zylkin 2016 | 19 | 4 | 1990-2002 | Dummy | Industry-level goods trade |
| Coughlin & Wall 2003 | 6 | 1988-1997 | Dummy | State-level goods trade |
|
| Lederman, Maloney & Serven 2005 | 0 | 0 | 1980-2000 | Dummy | Country-level non-fuel goods trade |
| Montenegro & Soloaga 2006 | 0-40 | 0 | 1988-2003 | Dummy | Country-level non-fuel goods trade |
| Krueger 1999 | 0 | 0 | 1987-1997 | Dummy | Industry-level goods trade |
| Trade Demand Equations | |||||
| McDaniel & Agama 2003 | 3 | 1994-2001 | Dummy | Country-level goods trade |
|
| McDaniel & Agama 2003 | 2 | 1983-2001 | Dummy | Country-level goods trade |
|
| USITC 2003 | 14 | 1989-1999 | Tariff preference: Dummy interactions | Product-level goods trade |
|
| USITC 2003 | 10 | 1989-2001 | Tariff preference: Dummy interactions | Product-level goods trade |
|
| Time Series Models | |||||
| Pacheco-Lopez 2005 | 25 | 0 | 1970-2000 | Dummy | Country-level goods & services trade |
| CBO 2003 | 10 | 7 | 1969-2001 | Dummy & Tariff changes | Country-level non-oil goods trade |
| Garcés-Díaz 2001 | 0 | 0 | 1990-2000 | Dummy | Country-level goods trade |
So what does the evidence show? The first two studies apply the most recent techniques in general equilibrium models. Romalis (2007) looks at data through 1999 and finds 10% of U.S.-Mexico trade in that year can be attributed to NAFTA. Caliendo and Parro (2015) find stronger effects looking at data through 2005. They attribute 41% of U.S. exports to Mexico to NAFTA, and 26% of U.S. imports from Mexico. The big increase in their results over all the other studies derives from a model that fully incorporates intermediate goods in supply chains. With supply chains, a tariff applied to an early-stage import cascades through further production stages. Finally, a working paper by Zylkin (2016) applies the workhorse gravity model using recently developed techniques to improve identification power. His results attribute 19% of U.S. exports to Mexico through 2002 to NAFTA, and 4% of trade in the other direction.
To summarize across a range of estimates it is probably reasonable to say that NAFTA may be responsible for roughly 25% of current U.S.-Mexico trade. This ballpark estimate gives higher-quality studies more weight and considers that the point estimates provide an upper bound to current causal attribution. One-quarter of trade is a substantial amount, but far less than most casual discussions of NAFTA would imply.
One important pattern emerges from 0 that deserves attention. If the studies found an effect of NAFTA, universally the effect was larger for U.S. exports than for Mexico’s exports. These results suggest that other factors like the exchange rate, economic growth, and pressure on firms to establish cross-border supply chains deserve the preponderance of credit for U.S.-Mexico trade growth, but especially for the growth of Mexico’s exports.
Regional Effects
How much did NAFTA contribute to the state-level patterns of overall and supply chain trade with Mexico described above? Two studies looked at the state-level patterns. However, it is important to caveat their results with the fact that they use relatively basic gravity model techniques and small samples.
Coughlin and Wall (2003) find that after controlling for state-level economic growth rates, among the six leading states NAFTA only boosted exports to Mexico from the border states of Arizona, California and Texas. However, Texas saw a surprisingly small impact; NAFTA boosted exports, but less than for the average state in the Southeast U.S. (Funk et al. 2006).
Some states lost business due to NAFTA, and some of the gains represented shifts between states. New York and Minnesota were found to have experienced substantial export declines due to NAFTA (Coughlin and Wall 2003). Electrical equipment and automotive industries saw export activity move from states like Louisiana to Texas and Tennessee (Funk et al. 2006). Electrical equipment and autos are the two industries that most strongly developed cross-border supply chains, both around the world and within NAFTA. Relatedly, Coughlin and Wall show that Texas received a disproportionate lift in exports to Canada, perhaps because of Texas’ role in the integration of the auto industry.
One interpretation of the state-level evidence might be that if studies cannot identify a clear, strong impact of NAFTA on Texas, it is no surprise that few of the early studies on aggregate trade found an impact. It appears most likely that the auto and electrical industries in Texas, other border states and Tennessee saw the most concentrated gains from NAFTA. These gains were not enormous, however, accounting for no more than a small fraction of overall trade growth during the NAFTA period.
Footnotes
1. Most studies that have attempted econometrically to control for other factors (exchange rates and economic growth in particular) have suffered from two weaknesses that bias the results in opposite directions. First, they mostly use trade data aggregated across all industries. Country-level data is well known to suffer from aggregation bias — that different responses among different industries or products may offset each other, falsely indicating no response at all.
Second, many studies suffer from what is known as omitted variable bias. NAFTA is measured as a simple year dummy, meaning the results reflect anything that affected the U.S.-Mexico relationship that were omitted as control variables. Studies that examined time series data (Pacheco-López 2005; CBO 2003; Garcés-Díaz 2001) suffer badly from these concerns, but the simple NAFTA measures for those that use panel data in gravity models (Montenegro and Soloaga 2006; Lederman, Maloney, and Serven 2005; Coughlin and Wall 2003; McDaniel and Agama 2003) also may pick up other factors. In these studies the devaluation of the peso in 1994 and strong growth in the US in the 1990s accounted for a much larger share of trade growth than NAFTA. For a much broader and less targeted set of estimates, see for instance Maria Cipollina and Luca Salvatici, “Reciprocal Trade Agreements in Gravity Models: A Meta-Analysis,” Review of International Economics 18, no. 1 (February 2010): 63–80
2. For a much broader and less targeted set of estimates, see for instance Maria Cipollina and Luca Salvatici, “Reciprocal Trade Agreements in Gravity Models: A Meta-Analysis,” Review of International Economics 18, no. 1 (February 2010): 63–80.
Russell A. Green, Ph.D., is the Will Clayton Fellow in International Economics at Rice University’s Baker Institute and an adjunct professor in the Economics Department, where he teaches financial markets, international finance and macroeconomics. Green’s current research focuses on monetary and exchange rate policy, financial market development in emerging market economies and Indian developmental challenges.