The Determinants of US Congressional Voting on the
Trade and Development Act of 2000
Baban Hasnat* and Charles Callahan, III** Department of Business Administration and Economics, State University of New York College at Brockport, U.S.A.
Abstract
The paper provides an empirical examination of the determinants of support for the Trade and Development Act of 2000 (TDA2000) in the United States Congress. We estimate a logistic regression model and control for both economic and political influences. We find that business political action committee contributions to lawmakers, the percentage of the Afri-can-American population in their constituency, the percentage of the Hispanic population in their constituency, and the skill level of the constituents had a significant positive influence on lawmakers voting in favor of TDA2000. Democratic party affiliation, import-competing in-dustries in the constituency, and labor union membership had a significant negative influence on the TDA2000 vote.
Key words: trade and development act of 2000; trade bill; trade policy; congressional voting JEL classification: D72; F13
1. Introduction
The beginning years of the 21st century have been an extraordinary time for trade policy for the United States. The US has signed into law the Trade and Devel-opment Act of 2000, the Free Trade Agreement with Jordan, and the Permanent Normal Trade Relations with China and is currently negotiating free trade agree-ments with several Middle Eastern countries.
A major trade bill signed into law in the US in recent years is the Trade and Development Act of 2000 (TDA2000), commonly known as the Africa-CBI trade bill. The TDA2000 grants over 70 African and Caribbean countries duty-free and quota-free access for certain goods, mostly textiles and apparels, to American
Received February 2, 2004, accepted April 5, 2004.
*Correspondence to: Department of Business Administration and Economics, State University of New York,
350 New Campus Drive, Brockport, NY 14420-2965, U.S.A. E-mail: bhasnat@brockport.edu.
**A United University Professions and State University of New York Professional Development Grant and a
Scholarly Incentive Grant from SUNY Brockport provided partial funding for this research to Charles Calla-han, III. The authors express their appreciation for research support to Steve Breslawski, Dena Levy, and Nathan Halat. We are grateful to three anonymous referees, the Managing Editor, and the Editor for their helpful comments and suggestions.
kets. It is hailed as “a new milestone” for American trade with these nations. A trade bill for Africa was under consideration for several years, but the lack of interests and partisan politics among lawmakers made it difficult to pass such a bill. The TDA2000 has its roots in the 1983 Caribbean Basin Initiative Act that was designed to promote development in the region by allowing certain imports to enter the US duty free. President Clinton was determined to pass it, considering it a landmark trade bill with the African and Caribbean nations. The US House of Representatives (House) voted 309-110 to pass the TDA2000 in May 2000.
Despite having a long historical relationship, trade between the US and Africa remains at a very low level. In 2000, imports from the African countries to the US totaled $27.6 billion, approximately 2.6% of total US imports. In comparison, im-ports from the Asian countries to the US totaled $457.7 billion during the same year. The US exported only $10.9 billion of goods and services to Africa. The corre-sponding figures for the Caribbean countries are even smaller.
The vote on the TDA2000 generated heated debates in public forums and aca-demic circles. Major corporations and trade associations spent thousands of dollars to pass or amend it, while labor unions put up a serious challenge. Corporations, such as Fruit of the Loom, The Limited, and Chiquita Brands Internationals, with business interests in the beneficiary countries made large political donations to con-gressional candidates and committees. These corporations were successful in insert-ing provisions into the TDA2000 in their favor or in preventinsert-ing their businesses be-ing hurt by it. As expected, labor unions—particularly the Union of Needletrades and Industrial and Textile Employees—lobbied against TDA2000.
The supporters of the TDA2000 argued that the bill would strengthen US trade and investment relations with the African and Caribbean countries, protect US intel-lectual property rights, promote peace and democracy, and improve labor standards in the beneficiary countries. The critics of the TDA2000 saw it as more for boosting and guaranteeing American investment in Africa rather than African trade with the US. They claimed that the bill imposes unrealistically high labor standards on the African countries, does not write off debt as demanded by many African-American leaders, and makes the beneficiary countries dependent on failed International Monetary Fund stabilization policies. Nelson Mandela criticized the bill because it would impose conditions on the beneficiary countries’ freedom to trade with coun-tries such as Cuba, Libya, or Iran. Many lawmakers from textile-intensive states were against the TDA2000; they asserted that the bill would lead to the elimination of jobs, particularly in their districts.
This paper provides an empirical examination of the determinants of support for the TDA2000 in the US Congress. The Bush administration is currently negoti-ating free trade agreements with several countries in the Middle East. It is important to know what motivated lawmakers to vote for or against the TDA2000. In particular, we examine whether the support for TDA2000 in the Congress is linked to constitu-ency factors, partisanship, and legislators’ ideology.
2. A Brief Survey of the Recent Empirical Literature
Baldwin’s (1985) pioneering study finds that party affiliation, union contribu-tions, and the importance of import-sensitive industry in a constituency affect a lawmaker voting on a trade bill. Using Baldwin’s approach, Coughlin (1985), Tonsi and Tower (1987), and Allen and Hopkins (1997) find comparable results. Coughlin (1985) finds that the ideology of lawmakers is not an important factor, while Nollen and Iglarsh (1990) find it to be very important.
Concerning the North America Free Trade Agreement (NAFTA) vote, Cony-beare and Zinkula (1996), Kang and Greene (1999), Thorbecke (1997), and Kahane (1996) find that legislators responded to labor interests by voting against the bill. Ideological position had little effect on the NAFTA vote in the Kang and Greene (1999) and Kruger (1996) analyses but had a significant impact in the Kahane (1996) and Wink et al. (1996) analyses. Baldwin and Magee (2000) find that Political Ac-tion Committee (PAC) contribuAc-tions from business, higher ratings from the Cham-ber of Commerce, higher percentages of workers in export-oriented industries to that in import-competing industries in members’ districts, higher proportions of Hispan-ics in House members’ districts, and greater proportions of employment of workers in furniture, fabricated metals, and electronic equipment industries increased the likelihood of House members voting for NAFTA. On the other hand, their study finds that PAC contributions from labor, higher ratings from the American Conser-vative Union and the AFL-CIO, higher proportions of union workers in House members’ districts, and greater proportions of employment of workers in the chemi-cal and industrial machinery industries decreased the likelihood of House members voting for NAFTA. Holian et al. (1997) find that the probability of voting for NAFTA by House members was not negatively affected by an increase in the per-centage of the minority population in their districts—a finding that was contrary to their expectation. They find, however, that business PAC contributions positively impacted the NAFTA vote, while organized labor PAC contributions negatively impacted the NAFTA vote. Wink et al. (1996) find that the African-American con-stituency variable was significant and negative; however, the Latino variable was positive and significant when legislators were in danger of not being reelected. 3. The Model and the Data
The empirical literature on voting models shows that the appropriate statistical technique to use is either probit or logistic regression (Adkisson and Daniel, 2001, Allen and Hopkins, 1997, Conybere and Zinkula, 1997, Kahane, 1996, Kang and Greene, 1999, Thorbecke, 1997). We choose to use the logistic model because it can be readily extended to more than one predictor variable and inference procedures are more easily carried out than with the probit model (Fox, 1997, Neter et al., 1996). Kmenta (1986, p. 555) notes that one significant advantage of logit over probit is logit’s close approximation to the cumulative normal function. The model takes the
following form: Pr(Yi = 1) = Pi = eL/(1 + eL), where L is a linear combination of the
predictor variables, i.e., L = ß0 + ß1X1 +… + ßkXk (Menard, 2002).
The dependent variable (AFRICA) takes on the value of 1 if a representative voted for the trade bill and 0 otherwise.Geographical constituents’ interests are rep-resented by the unemployment rate, the skill level of the population, im-port-competing sectors, and the importance of labor unions (Thorbecke, 1997). Con-stituents who experience a high level of joblessness are more likely to put pressure on legislators to vote against liberalized trade legislation. Therefore, we would ex-pect the coefficient on the unemployment rate variable (UNEM) to be negative. A higher percentage of college-educated constituents (COLLEGE) leads to a higher level of skills in the congressional district and thus a greater likelihood of a positive vote on a trade liberalization bill (Kang and Greene, 1999). The greater the percent-age of workers who are employed in the textile and apparel industries (TEXAPR), those industries that are import-competing industries, the greater is the likelihood of a negative vote on a trade bill (Allen and Hopkins, 1997). Since organized labor has generally opposed liberalized trade bills, it is hypothesized that the higher the per-centage of workers who are covered by collective bargaining (COLBAR) in a repre-sentative’s constituency, the greater the likelihood of a vote against the TDA2000 (Baldwin and Magee, 2000).
Electoral constituents’ interests and the philosophy of the lawmaker are repre-sented by political action contributions, party affiliation, the ideology of the House member, and the proportions of African-American (PBLACK) or Hispanic (PHISPA) constituents in the House member’s district. Members of the Congressional Black Caucus (CBC) are from districts that are predominantly black. The TDA2000 may be supported by CBC House members if the perception exists that African-American constituents firmly believe that trade with Africa is of utmost importance for devel-opment and other reasons. Those representatives who had a high percentage of His-panics in their districts voted in favor of NAFTA (Baldwin and Magee, 2000). Whether this pattern of voting in favor of trade bills continues is a matter to be ex-plored.
Democrats are more inclined to vote for trade protection than are Republicans (McArthur and Marks, 1988, Allen and Hopkins, 1997, Krueger, 1996) and organ-ized labor has tended to primarily support Democrats in congressional races (CQ
Weekly Report, 2000). Party affiliation (DEMO) is represented by a binary variable
with 1 for members of the Democratic Party and 0 otherwise. The party affiliation variable is expected to have a negative sign. Both business and organized labor tend to support congressional races through the use of PAC monies. Recent studies by Beaulieu (2002) and Baldwin and Magee (2000) show that organized labor’s PAC contributions tend to decrease the probability of voting for trade liberalizing bills while business PAC contributions tend to increase the probability of voting for trade liberalizing bills. Their results are consistent with the implications of the Stol-per-Samuelson theorem that predicts that trade liberalization policies benefit busi-ness owners and hurt laborers. PACBUS and PACLBR measure the percentage of total contributions received by individual lawmakers from business and organized
labor PACs, respectively. The expected sign of PACBUS is positive and that of PACLBR is negative (Beaulieu, 2002, Baldwin and Magee, 2000, Kang and Greene, 1999, Uslaner, 1998, Holian et al., 1997).
A legislator’s ideology may have an impact upon voting behavior. Conserva-tives in general tend to be supporters of “laissez faire” and free trade, while liberals are seen as those who favor government regulation and protectionism (Baldwin and Magee, 2000, Wink et al., 1996). However, some conservatives oppose specific trade liberalizing bills on the grounds of national sovereignty and religious freedoms (Doran, 1994). Given that the literature has not reached a consensus on the use of legislator’s ideology as a determinant of voting behavior (see Stratmann, 1992, for example), we use it as a control variable. The American Conservative Union (ACONU), which measures the conservatism of a representative on a variety of is-sues, is used as a measure of ideology. The sign of the ACONU variable is indeter-minate since conservatives were not in agreement on NAFTA or TAD2000 (CQ
Weekly Report, 2000).
Data for the study come from a number of sources: the dependent variable, the voting results, from the Congressional Quarterly, May 2000; UNEM and COLBAR from the Employment and Earnings, May 2000; COLLEGE, DEMO, ACONU, PBLACK, and PHISPA from the Almanac of American Politics 2000; PACBUS and PACLBR from the Center of Responsive Politics; and TEXAPR from the
Depart-ment of Commerce Website. UNEM, COLBAR, and TEXAPR are measured at the
state level, while COLLEGE, PBLACK, PHISPA, PACBUS, and PACLBR are measured at the district level. Given that our data set is comprised of both state and district level observations, there is a possibility of measurement error (Allen and Hopkins, 1997, Conybeare and Zinkula, 1996, Tosini and Tower, 1987). Neverthe-less, we use state and district level data for several reasons. First, unemployment data can be found on counties; however, district boundaries and county boundaries are not the same in too many cases. Second, location of economic activity and where constituents reside and vote are not the same in many instances. Overall, our study, like many others, is dictated by data availability.
4. Results
Table 1 presents means and standard deviations of the variables and Table 2 presents the correlation matrix. Since the correlation matrix shows that there is a high degree of correlation between PACLBR and DEMO, PACLBR and ACURAT, PACLBR and PACBUS, and DEMO and ACURAT, we ran three regressions and report the results in Table 3.
As expected, the African American variable, PBLACK is positive and highly significant in all of the models. Thus, a one percent increase in the African-American population in a House member’s district leads to a 0.64 percent to 0.70 percent increase in the probability of voting in favor of TDA2000. The Hispanic variable, PHISPA, is also positive and significant (at least at the 5 percent level) in all of the models. Thus, a one percent increase in the percentage of
Hispanics in a legislator’s district increases the likelihood of a House member voting for TDA2000 by 0.37 percent to 0.45 percent. These findings are contrary to the hypothesis advanced by Holin et al. (1997) that legislators with a high concentration of minorities in their constituency tend to oppose free trade bills.
Table 1. Means and Standard Deviations (n = 415)
Variable Mean Standard Deviation
AFRICA 0.7422 0.4380 PBLACK 11.7410 15.4335 UNEM 4.2361 0.8290 PHISPA 8.7918 14.0505 COLBAR 14.0617 6.2504 COLLEGE 45.2217 10.9985 TEXAPR 1.1748 1.3260 PACBUS 67.3713 25.6948 PACLBR 21.9959 21.9675 DEMO 0.4892 0.5005 ACONU 50.2651 39.0971
Table 2. Correlation Matrix AFRICA PBLACK UNEM PHISPA COLBAR
COL-LEGE TEXAPR PACBUS PACLBR DEMO
PBLACK 0.037 UNEM 0.044 0.021 PHISPA 0.109 -0.034 0.351 COLBAR -0.093 -0.109 0.262 -0.051 COLLEGE 0.162 -0.243 0.053 -0.045 0.087 TEXAPR -0.156 0.268 0.013 -0.089 -0.354 -0.152 PACBUS 0.292 -0.181 -0.071 -0.073 -0.248 0.023 0.104 PACLBR -0.303 0.294 0.085 0.122 0.307 -0.057 -0.078 -0.755 DEMO -0.261 0.309 0.072 0.160 0.166 -0.088 -0.035 -0.513 0.734 ACONU 0.182 -0.307 -0.076 -0.199 -0.291 -0.009 0.081 0.542 -0.734 -0.891
The sign of the COLLEGE variable is positive and significant at the 1 percent level in all models. Thus, a one percent increase in the skill level of a district in-creases the likelihood of a House member voting for the TDA2000 by 0.69 to 0.82 percent. The result is consistent with Kang and Greene (1999). The collective bar-gaining coverage variable (COLBAR) is negative but is only marginally significant in Model 1. Our result is broadly similar to Wink et al. (1996), who find that union constituency strength did not influence the vote on NAFTA. The unemployment variable (UNEM) has the wrong sign but is not significant. This result is consistent with the result found by Allen and Hopkins (1997) but is not in conformity with the hypothesis that districts with high unemployment rates tend to have representatives that vote for protectionist legislation because of concerns of joblessness of their constituents (Kahane, 1996). It should be noted that the unemployment rate was 3.87
percent (which is below the full employment level) at the time of the vote and, con-sequently, does not appear to be an important factor in the voting decision.
Table 3. Logistic Estimates
(Dependent variable is vote on the Trade and Development Act of 2000, Yes = 1 and No = 0)
Notes: absolute t ratios are in parentheses, and marginal effects evaluated at the mean are in brackets. The marginal effects can be calculated as follows: Pi/Xi = Pi(1−Pi )ß, where Pi and (1−Pi) are the probabilities that the dependent variable takes the value 1 and 0, respectively, and ß is the estimated coefficient. * denotes significance at the 10% level, ** at the 5% level, and *** at the 1% level (all two-tail tests).
The likelihood of House members voting for TDA2000 decreased as the proportion of workers employed in the textile and apparel industries increased.
Model 1 Model 2 Model 3
Constant 0.1377 (0.0978) [0.0212] -1.5168 (1.4169) [-0.2336] -2.1367** (1.8215) [-0.3356] PBLACK 0.0418*** (3.9878) [0.0064] 0.4510*** (4.3709) [0.0069] 0.0446*** (4.2878) [0.0070] UNEM 0.1683 (0.9469) [0.0259] 0.1249 (0.7126) [0.0192] 0.1240 (0.7060) [0.0195] PHISPA 0.0240** (2.1272) [0.0037] 0.0291*** (2.6730) [0.0045] 0.0282** (2.5515) [0.0044] COLBAR -0.0507* (1.7825) [-0.0078] -0.0348 (1.2979) [-0.0054] -0.0253 (0.9489) [-0.0040] COLLEGE 0.0447*** (3.2985) [0.0069] 0.0511*** (3.8725) [0.0079] 0.0525*** (3.9464) [0.0082] TEXAPR -0.5140*** (4.7518) [-0.0792] -0.5060*** (4.6938) [-0.0779] -0.4873*** (4.5788) [-0.0765] PACBUS 0.0197*** (2.7049) [0.0030] 0.0186*** (2.5950) [0.0029] 0.0169** (2.4039) [0.0026] PACLBR -0.0173 (1.6071) [-0.0027] -0.0158 (1.4943) [-0.0024] -0.0264*** (2.5853) [-0.0041] DEMO -1.9289*** (2.9592) [-0.2973] -0.9731*** (2.5959) [-0.1499] ACONU -0.0162* (1.8093) [-0.0025] 0.0051 (0.9875) [0.0008] n 415 415 415
Log Likelihood Function -182.106 -183.79 -186.661 Model χ2 109.531*** 106.164*** 100.422***
% Correctly Predicted 81.8 81.4 80.7
Specifically, a one percent increase in the proportion of workers employed in the textile and apparel industries (TEXAPR) decreased the likelihood of a House member voting for the TDA2000 by 8 percent.
We expected that political action contributions would have an impact on the voting behavior of House members. Indeed, PACBUS is positive and significant. As business PAC contributions increased as a percentage of overall PAC contributions, the likelihood increased that the House member would vote for the TDA2000. More specifically, a one percent increase in business PAC contributions increased the likelihood of a House member voting for TDA2000 by 0.3 percent. These results are supported by Kang and Greene (1999) and Allen and Hopkins (1997). That is, it appears that business contributors have an impact on votes or that business con-tributors buy access to have influence on future legislation. Organized labor PAC contributions (PACLBR), while having a negative sign as expected, is significant in Model 3 only (DEMO is dropped in the model). This should not be surprising since the correlation between DEMO and PACLBR is 0.734.
The variable for party affiliation (DEMO) is negative and significant at the 1 percent level in Models 1 and 2. We hypothesized that the sign would be negative because Democrats are thought to be the party of labor and because Democrats, rela-tive to Republicans, tend to vote consistently for protectionist trade policies (Tosini and Tower, 1988, Kahane, 1996, Allen and Hopkins, 1997).
The ACONU variable yields mixed results. It has a negative sign and is only marginally significant in Model 1 but has a positive sign and not significant in Model 3. So it appears that personal ideology was not an important factor in the vote on TDA2000. Caution must be exercised, however, because “free trade votes often make strange bedfellows and with diverse opponents, conventional ratings may be poor measures of ideology” (Kang and Greene, 1999).
5. Summary
The analysis of House of Representatives voting on the Trade and Develop-ment Act of 2000 contributes to our understanding of trade policy towards Africa and the Caribbean nations. The voting behavior of House members was positively and significantly related to the percentage of the African-American population in their constituency, the percentage of the Hispanic population in their constituency, the skill level of their constituents, and to business PAC contributions. These same House members’ voting behavior was negatively and significantly influenced by Democratic partisanship, import-competing industries (textile and apparel sectors), and the proportion of union members covered by collective bargaining contracts in a House member’s state. It appears that future trade bills will be voted favorably upon if (1) they are seen to benefit the perceptions of all race/ethnic groups, (2) the educa-tion/skill level of Americans increases over time, and (3) business lobbyists continue to have influence with Congressional members.
References
Allen, S. D. and A. S. Hopkins, (1997), “The Textile, Apparel, and Footwear Act of 1990: Determinants of Congressional Voting,” Public Finance Review, Sep-tember, 542-552.
Baldwin, R. E., (1985), The Political Economy of US Trade Policy, Cambridge, Mass: MIT Press.
Baldwin, R. E., (1996), “The Political Economy of Trade,” in The Political Economy
of Trade Policy, R. C. Freenstran and G. M. Crossman eds., Cambridge, Mass:
MIT Press.
Baldwin, R. E. and C. S. Magee, (2000), “Congressional Trade Votes: From NAFTA Approval to Fast-Track Defeat,” Washington, D.C.: Institute for International Economics.
Barone, M. and G. Ujifusa, (1999), “The Almanac of American Politics, 2000,” Washington, D.C.: National Journal Group, Inc.
Beaulieu, E., (2002), “The Stopler-Samuelson Theorem Faces Congress,” Review of
International Economics, 10, 343-360.
Congressional Quarterly, 2000. CQ Weekly Report. Various issues.
Conybeare, J. and M. Zinkula, (1996), “Who Voted Against the NAFTA? Trade Un-ions Versus Free Trade,” The World Economy, 1, 1-12.
Coughlin, C., (1985), “Domestic Content Legislation: House Voting and the Eco-nomics of Regulation,” Economic Enquiry, 23, 437-448.
Doran, C. F., (1994), “Trade and Party: Flip-Flops and Causal Linkages,” in The
NAFTA Puzzle: Political Parties and Trade in North America, C. F. Doran and
G. P. Marchildon eds., Boulder, Colorado: Westview Press.
Holian, D. B., T. B. Krebs, and M. H. Walsh, (1997), “Constituency Opinion, Ross Perot, and Roll-Call Behavior in the U. S. House: The Case of NAFTA,”
Leg-islative Studies Quarterly, 22, 369-392.
Kahane, L. H., (1996), “Congressional Voting Patterns on NAFTA: An Empirical Analysis,” American Journal of Economics and Sociology, October, 395-409. Kang, I.-B. and K. Greene, (1999), “A Political Economic Analysis of Congressional
Voting Patterns on NAFTA,” Public Choice, 98, 385-397.
Kmenta, J., (1986), Elements of Econometrics, New York, New York: MacMillan. Knack, S., (1993), “The Voter Participation Effects of Selecting Jurors from
Regis-tration List,” Journal of Law & Economics, April, 99-114.
Krueger, A. B., (1996), “Observations on International Labor Standards and Trade,”
National Bureau of Economic Research, June, Working Paper 5632.
Menard, S., (2002), Applied Logistic Regression Analysis, 2nd edition, Thousand Oaks, California: Sage Publications.
McArthur, J. and S. V. Marks, (1988), “A Constituent Interest VS. Legislator Ideol-ogy: The Role of Political Opportunity Cost,” Economic Inquiry, July, 461-470. Nollen, S. D. and H. J. Iglarsh, (1990), “Explanations of Protectionism in
Interna-tional Trade Votes,” Public Choice, 66, 137-153.
Stratmann, T., (1992), “The Effects of Logrolling on Congressional Voting,”
Tosini, S. C. and E. Tower, (1987), “The Textile Bill of 1985: The Determinants of Congressional Voting Patterns,” Public Choice, 54, 19-25.
Thorbecke, W., (1997), “Explaining House Voting on the North American Free Trade Agreement,” Public Choice, 92, 231-242.
Wink, K. A., C. D. Livingston, and J. C. Garand, (1996), “Dispositions, Constituen-cies, and Cross-Pressures: Modeling Roll-Call Voting on the North American Free Trade Agreement in the U.S. House,” Political Research Quarterly, 49, 749-770.
Uslaner, E. M., (1998), “Let the Chits Fall Where They May? Executive and Con-stituency Influences on Congressional Voting on NAFTA,” Legislative Studies
