Congressional voting patterns on NAFTA: an empirical analysis - North American Free Trade Agreement

American Journal of Economics and Sociology, The, Oct, 1996 by Leo Kahane

Immigration. To test for the importance of this factor a variable (IMMIG) equal to the number of immigrants admitted in 1989 by state of intended origin from Mexico is included in the econometric model. The expected sign is positive.(15)

Other Predictors. Several other factors which are not explicitly covered by the themes discussed above are expected to have played an important role in shaping the vote on NAFTA.(16) The first is that of organized labor which strongly opposed NAFTA. Two variables are used to control for labor's effects on voting patterns on NAFTA. The first, UNION, is the percent of a states manufacturing employment that is unionized. All else equal, the greater the union membership, the less likely it is that a legislator will vote in favor of NAFTA. The second variable used to control for the influence of organized labor is the amount of total campaign funds received by a legislator from labor groups for their most recent election. It is expected that the greater the campaign funds received from labor, the more likely a legislator would vote against NAFTA, other things equal.

Previous research on trade policy has found that, in general, Democrats tend to be more protectionist than Republicans.(17) Further still, there may be significant ideological differences for Northern versus Southern Democrats. In order to control for these potential effects, two dichotomous variables are created. The variable NDEMO takes the value of 1 if the legislator is a Northern Democrat, 0 otherwise and is expected to have a negative coefficient. Similarly, SDEMO, takes the value of 1 if the legislator is a Southern Democrat, 0 otherwise. The expected sign on SDEMO is somewhat ambiguous since Southern Democrats are often considered to be more conservative and closer to Republican ideology than are Northern Democrats. On the other hand, other factors (e.g. party loyalty), would lead Southern Democrats to oppose NAFTA, all else equal. Previous research (e.g. Kahane, 1994) supports this latter view, and as such a negative coefficient is expected.

Lastly, a variable UNEMP equal to a state's unemployment rate is included in the regression model. States witnessing a high unemployment rate may develop an anti-foreigner stance as they (and perhaps their elected officials) search for reasons to explain their economic hardship. The expected sign for this coefficient is thus negative.

IV

The Empirical Model

The votes cast by legislators on NAFTA were used to create a dichotomous dependent variable VOTE which takes a value of 1 if the legislator voted in favor of NAFTA, 0 otherwise. A logit analysis is conducted for both the Senate and House vote on NAFTA using the variables described in section III above as predictors. In addition, two equations are estimated for both the Senate and House. The first estimation uses absolute measures for GAINERS, LOSERS, EXPORT, IMMIG and LCON. The second uses relative measures for these variables which are created by dividing GAINERS, LOSERS and IMMIG by state employment, EXPORT by total state value added from manufacturing and dividing LCON by total campaign contributions. As noted in Kahane (1994), arguments can be made for both specifications. For example, one argument in favor of using absolute numbers stems from the political economy literature on collective action. That is, a small group which stands to, say, lose from a particular policy may prove to be a very effective lobbying group if they are well organized and visible.(18) Deflating this groups lobbying ability by considering its size relative to total state employment may understate its true effectiveness. On the other hand, legislators may in fact be considering relative employment effects when choosing their position on NAFTA. Similar arguments apply to measures of immigration and labor campaign contributions. As such, both specifications are tested.