The Economics and Politics of Revoking NAFTA

Abstract

We provide a quantitative assessment of both the aggregate and the distributional effects of revoking NAFTA using a multi-country, multi-sector, multi-factor model of world production and trade with global input–output linkages. Revoking NAFTA would reduce US welfare by about 0.2%, and Canadian and Mexican welfare by about 2%. The distributional impacts of revoking NAFTA across workers in different sectors are an order of magnitude larger in all three countries, ranging from − 2.7 to 2.23% in the USA. We combine the quantitative results with information on the geographic distribution of sectoral employment, and compute average real wage changes in each US congressional district, Mexican state, and Canadian province. We then examine the political correlates of the economic effects. Congressional district-level real wage changes are negatively correlated with the Trump vote share in 2016: districts that voted more for Trump would on average experience greater real wage reductions if NAFTA is revoked.

Appendix: Solution Algorithm

To solve Eqs.  (8)–(15) start by guessing $\{{\hat{w}}_{jn},{\hat{r}}_{jn}\}$ and use the following algorithm.

• (i) Solve for ${\hat{p}}_{jn}$ using Eqs. (14) and (12):

  $$\begin{aligned} {\hat{p}}_{jn}= & {} \bigg (\sum _{m=1}^{N}\pi _{j,mn}({\hat{c}}_{jm}{\hat{\kappa }}_{j,mn})^{-\theta }\bigg )^{-\frac{1}{\theta }}\\ {\hat{p}}_{jn}= & {} \bigg [\sum _{m=1}^{N}\pi _{j,mn}^{j}\bigg (\big ({\hat{w}}_{jm}^{\alpha _{jm}}{\hat{r}}_{jm}^{1-\alpha _{jm}}\big )^{\beta _{jm}}\big (\prod _{i=1}^{J}({\hat{p}}_{im})^{\gamma _{ij,m}}\big )^{1-\beta _{im}}{\hat{\kappa }}_{j,mn}\bigg )^{-\theta }\bigg ]^{-\frac{1}{\theta }} \end{aligned}$$

  which can be solved iteratively. Then use ${\hat{p}}_{jn}$ to solve for ${\hat{c}}_{jn}$ and ${\hat{P}}_{n}$:

  $$\begin{aligned} {\hat{c}}_{jn}= & {} ({\hat{w}}_{jn}^{\alpha _{jn}}{\hat{r}}_{jn}^{1-\alpha _{jn}})^{\beta _{jn}}\big (\prod _{i=1}^{J}({\hat{p}}_{in})^{\gamma _{ij,n}}\big )^{1-\beta _{jn}}\\ {\hat{P}}_{n}= & {} \prod _{j=1}^{J}\big ({\hat{p}}_{jn}\big )^{\xi _{jn}} \end{aligned}$$

• (ii) Solve for ${\hat{\pi }}_{j,mn}$ using Eq. (11) and ${\hat{c}}_{jn}$:

  $$\begin{aligned} {\hat{\pi }}_{j,mn}=\frac{({\hat{c}}_{jm}{\hat{\kappa }}_{j,mn})^{-\theta }}{\sum _{m'=1}^{N}\pi _{j,m'n}({\hat{c}}_{jm'}{\hat{\kappa }}_{j,m'n})^{-\theta }} \end{aligned}$$

• (iii) Use Eqs. (8) and (9) to solve for ${\hat{Y}}_{jn}$ and ${\hat{Q}}_{jn}$:

  $$\begin{aligned} {\hat{p}}_{jn}{\hat{Y}}_{jn}= & {} \sum _{i=1}^{J}{\widehat{w}}{}_{in}SL_{in}+\sum _{i=1}^{J}{\widehat{r}}{}_{in}SK_{in}+\sum _{m\ne n}\sum _{i=1}^{J}\frac{\tau '_{i,mn}{\widehat{\pi }}{}_{i,mn}{\widehat{p}}{}_{in}{\widehat{Q}}{}_{in}}{1+\tau '{}_{i,mn}}\frac{\pi {}_{i,mn}p{}_{in}Q{}_{in}}{I_{n}}+{\widehat{D}}{}_{n}SD_{n} {\hat{p}}_{jn}{\hat{Q}}_{jn}(p_{jn}Q_{jn})= & {} {\hat{p}}_{jn}{\hat{Y}}_{jn}(p_{jn}Y_{jn})+\sum _{i=1}^{J}(1-\beta _{in})\gamma _{ji,n}\Big (\sum _{m=1}^{N}\frac{{\hat{\pi }}_{i,nm}\pi _{i,nm}{\hat{p}}_{im}{\hat{Q}}_{im}(p_{im}Q_{im})}{(1+\tau '_{i,nm})}\Big ) \end{aligned}$$

  This can be solved iteratively.

• (iv) Update the next guess for ${\hat{w}}_{jn}$, ${\hat{r}}_{jn}$ from the labor market clearing condition

  $$\begin{aligned} {\widehat{w}}{}_{jn}={\widehat{r}}{}_{jn}= & {} \frac{\sum _{m=1}^{N}\frac{{\widehat{\pi }}{}_{j,nm}{\widehat{p}}{}_{jm}{\widehat{Q}}{}_{jm}\pi {}_{j,nm}p{}_{jm}Q{}_{jm}}{1+\tau '_{j,nm}}}{\sum _{m=1}^{N}\frac{\pi _{j,nm}p_{jm}Q_{jm}}{1+\tau _{j,nm}}}. \end{aligned}$$

  the solution is defined up to a numeraire, and in updating the ${\hat{w}}_{jn}$ and ${\hat{r}}_{jn}$’s, re-set a numeraire country’s ${\hat{w}}_{1}=1$ (where country 1, sector 1 is the numeraire). Then, the actual next guess to be returned to step 1 is:

  $$\begin{aligned} {\hat{w}}_{jn}^{\rm next}= & {} \frac{\hat{\tilde{w}}_{jn}^{\rm next}}{\hat{\tilde{w}}_{11}^{\rm next}} \end{aligned}$$

  (25)
  $$\begin{aligned} {\hat{r}}_{jn}^{\rm next}= & {} \frac{\hat{\tilde{r}}_{jn}^{\rm next}}{\hat{\tilde{w}}_{11}^{\rm next}} \end{aligned}$$

  (26)

See Figs. 9, 10, 11 and Tables 7, 8, 9, and 10.

Fig. 9

Real wage changes and 2016 trump vote share, $\theta =2.5$. Notes: This figure depicts the scatterplots of the average real wage change from revoking NAFTA and the 2016 Trump vote share by congressional district (left side) and state (right side), along the OLS fit. The boxes report the coefficient, robust standard error, and the $R^{2}$ of the bivariate regression. The model is solved under $\theta =2.5$

Fig. 10

Real wage changes and 2016 trump vote share, $\theta =8$. Notes: This figure depicts the scatterplots of the average real wage change from revoking NAFTA and the 2016 Trump vote share by congressional district (left side) and state (right side), along the OLS fit. The boxes report the coefficient, robust standard error, and the $R^{2}$ of the bivariate regression. The model is solved under $\theta =8$

Fig. 11

Real wage changes and the difference between 2016 trump vote share and the 2012 romney vote share. Notes: This figure depicts the scatterplots of the average real wage change from revoking NAFTA and the difference between the 2016 Trump vote share and the 2012 Romney vote share by congressional district (left side) and state (right side), along the OLS fit. The boxes report the coefficient, robust standard error, and the $R^{2}$ of the bivariate regression

Table 7 List of countries

Table 8 List of sectors

Table 9 Assumed changes in US Tariffs and NTB on Canada and Mexico if NAFTA is revoked

Table 10 Top and Bottom 10 US Districts (Tariff and NTB Baseline)