The Passing of the Act

The Hart-Celler Act of 1965, named after its sponsors Senator Philip Hart (D-MI) and Representative Emmanuel Celler (D-NY), was a set of amendments to the Immigration Act of 1952, and, in order to appreciate its relevance, it must be compared with the pre-existing legislation. The Immigration Act of 1952, known as the McCarran-Walter Act after its sponsors Senator Patrick A. McCarran (R-NV) and Representative Francis E. Walter (D-PA), had largely reaffirmed the status quo in immigration policy. In particular, it retained country-of-origin quotas in proportion to the ancestry of the U.S. population in the 1920 census. As a result, 70.4 percent of the total numerical annual quota of 154,547 was allocated to three countries, the United Kingdom (42.3 percent), Germany (16.7 percent), and Ireland (11.5 percent), with a further 10.5 percent allocated to the countries of northwestern Europe. Unused allocations could not be transferred from one origin country to another. One small relaxation was the introduction of a quota of just 2,000 in total for immigrants from the so-called Asia-Pacific Triangle, who had been completely excluded by earlier Acts.Footnote ¹ The Act also introduced a hierarchy of preferences for immigrants admitted under the quota. The first preference was for immigrants admitted based on employment (50 percent of the total), followed by three preferences for different classes of relatives. As in the previous legislation, there was no numerical limit on immediate relatives of U.S. citizens. There was no quota or preference system for immigrants from the Western Hemisphere, who were admitted only under a labor certification system or as immediate relatives of U.S. citizens or permanent residents.Footnote ²

The Hart-Celler Act was a radical break with the past, largely because it abolished the country-of-origin quotas. These were replaced with an overall numerical limit of 170,000 per annum for the Eastern Hemisphere, which now included the countries formerly comprising the Asia-Pacific Triangle, with a maximum of 20,000 for any one country. Thus, for the first time since the early 1920s, each source country had equal numerical access. Another major change was the introduction of a numerical limit of 120,000 for the Western Hemisphere, initially without a system of preferences or a cap for individual countries and still subject to labor certification.Footnote ³ As in the earlier legislation, the admission of spouses and unmarried minor children was exempt from the quota, and this was extended to include parents of U.S. citizens. The Hart-Celler Act also reorganized the system of preferences that was applied to the Eastern Hemisphere. It established preferences for four classes of relatives, including one (the 5th preference) for brothers and sisters, and one preference for refugees (the 7th). There were two employment-based preferences, one for professional and scientific workers and one for skilled and unskilled manual occupations. The employment preferences were demoted in the order (to the 3rd and 6th preferences) and they amounted to only 20 percent of the total. This represents a substantial shift in the composition of visas away from employment toward family reunion. There was also a change in the procedure for gaining labor certification. Under the McCarran-Walter Act, the burden of proof lay with the Secretary of Labor to show cause to deny entry. Now the burden of proof lay with applicants (and for manual workers with their prospective employers) to show no adverse labor market consequences. This was seen as a tougher labor test that would most affect prospective immigrants from the Western Hemisphere (Schwartz Reference Schwartz1966, pp. 99–102; Keely Reference Keely1971, p. 160).

Although there had previously been a number of proposals for abolishing or reforming the country-of-origin quota system,Footnote ⁴ the immediate antecedent of the 1965 Act lies with proposals put forward in 1963 by President Kennedy (see Kennedy Reference Kennedy1964). These included the abolition of the national origins system (and the Asia-Pacific Triangle quota) to be replaced with an individual country cap of 10 percent of the overall quota. Under the proposed preference system, up to 50 percent of the quota was allocated to highly skilled or trained workers, with subsequent preferences for occupations in demand, non-immediate relatives, and refugees. Furthermore, there was to be no numerical cap on the Western Hemisphere. Although Kennedy’s bill did not reach the floor of the House or the Senate, essentially the same proposals were put forward after his death by President Johnson.Footnote ⁵ The legislative history of the Act is far from straightforward, and what emerged at the end of the process differed significantly from the administration’s original bill. The two most important differences were, first, that the balance of preferences was shifted radically away from employment toward family reunion. And second, the version that was finally passed imposed a cap of 120,000 on immigration from the Western Hemisphere.

The bill, HR 2580, was introduced by House Judiciary Committee Chairman Celler in January 1965, and it was sent to be considered by a subcommittee of the Judiciary Committee.Footnote ⁶ During the deliberations, several alternative bills were put forward, and some of their provisions were incorporated into HR 2580. The most important was that introduced by subcommittee chair Michael Feighan (D-OH), which amended the administration bill, introducing a system of seven preferences and increasing the share allocated to relatives of American citizens.Footnote ⁷ On the floor of the House, an amendment to place a numerical limit on the Western Hemisphere, proposed by Clark MacGregor (R-MN), was narrowly defeated, with a majority of Republicans voting in favor and a majority of Democrats against. The House proceeded to vote on the unamended bill, which was approved by a large majority of both parties (Table 1).

Table 1

ROLL CALL VOTES ON HR 2580

Notes: Including paired votes. The dates of the votes are, respectively, 25 August, 22 September, and 30 September 1965.

Source: Lewis et al. (Reference Lewis, Keith Poole, Boche, Rudkin and Sonnet2021).

In the Senate, the Judiciary Committee’s Subcommittee on Immigration and Naturalization heard a wide range of testimony. Eventually, the House version of the bill was approved, subject to a number of amendments, the most important of which was a numerical limit of 120,000 on immigration from the Western Hemisphere.Footnote ⁸ After some further negotiations in the Senate Judiciary Committee, the bill was reported out and approved by the Senate on a vote of 76 to 18. The conference report (essentially the Senate’s version of the bill) was then passed by the House with a slightly larger majority than previously, despite the addition of the Western Hemisphere cap. The Act was signed into law by President Johnson on 3 October 1965 and it came fully into effect on 1 July 1968.

The Political Economy of Immigration Reform

Ever since its introduction, the country-of-origin quota system had been criticized as discriminatory. After 1945, it was increasingly seen as particularly harmful in foreign relations, where the United States was concerned with forging closer strategic ties with Cold War allies and with newly independent developing countries. The fact that the immigration quotas discriminated strongly against some of the nation’s key (and potential) allies was seen as a major obstacle (FitzGerald and Cook-Martín Reference FitzGerald and Cook-Martín2014, ch. 3). This was one of the reasons that successive presidents had sought to abolish the quota system: Truman in 1952, Eisenhower in 1956 and 1960, and then Kennedy in 1963.Footnote ⁹ In domestic politics, the anti-communist scares of McCarthyism that formed the background to the McCarran-Walter Act had receded by the mid-1960s. And the landslide victory of the Democrats in 1964 created the conditions for liberal reforms, not least through changes to Congressional rules and by sidelining some of the leading supporters of the national origins system (Stern 1975, pp. 60–63). Thus, influential historians of immigration policy have emphasized external causes and threats, distinct political coalitions, and the role of institutional forces and professional experts (Tichenor Reference Tichenor2002, pp. 212–18; Daniels 2004, pp. 130–36).

The rise of the Democrats, linked with generational change, may also have been important. With a majority of Democrats in the House of 295 to only 140 Republicans, Democrats from the Eastern and Northern States were less dependent on their colleagues from the South. And of the 48 Democrats who replaced Republicans in 1965, 47 voted in favor of the final passage of HR 2580. In 1965, three-fifths of the House of Representatives were aged under 55 and were therefore members of what Putnam called the “civic generation” (Putnam Reference Putnam2000, p. 254). The latter was also important in the membership of the key subcommittees, both in the House and the Senate (Stern 1975, pp. 63, 66–71; Tichenor Reference Tichenor2002, ch. 7). In the Senate, the subcommittee that considered the 1965 bill was chaired by Edward Kennedy (D-MA), and it included Hiram Fong (R-HI), the first Asian-American to be elected to the Senate, and Philip Hart, who had long campaigned against the quotas and who had introduced the 1963 Administration Bill to the Senate. In the House, the relevant Judiciary subcommittee (# 1) was chaired by Michael Feighan (D-OH), who became a convert to reform and who replaced Francis Walter, who had died in 1963.

As noted earlier, the 1965 Act has often been seen as part of the program of radical reforms undertaken in the Kennedy-Johnson era. Among these were reforms in foreign policy and international trade and domestic reforms that included the “war on poverty” and, above all, the Civil Rights Act of 1964 and the Voting Rights Act of 1965. As Attorney General Robert Kennedy put it in 1964, “Everywhere else in our national life, we have eliminated discrimination based on one’s place of birth. Yet this system is still the foundation of our immigration law” (Statement to Subcommittee # 1 of the House Judiciary Committee, 22 July 1964). However the process that brought about the immigration reform is often seen to be very different from that of civil rights. As one historian put it:

This account suggests that, while immigration may have ridden on the coattails of the civil rights movement in Congress, it meant little to the average voter. The combined AFL-CIO endorsed the bill, whereas in 1952, the AFL supported maintaining the national origins quotas. But it is far from clear that this reflected the views of its grassroots members. While there were a number of lobby groups that favored reform, such as the National Committee for Immigration Reform, these were not broad-based. Others include religious organizations, notably Jewish and Catholic, but also some Protestants, such as the Lutherans. There were also a number of organizations based on ancestry, notably Italian, Greek, Japanese, and Polish. Some, like the American Committee for Italian Immigration and the Order of the Sons of Italy in America, testified before the immigration subcommittee of the Senate Judiciary Committee and campaigned more widely (Stern 1975, pp. 201–24). But it is unclear how far this was reflected in popular support for immigration reform.

Some observers see the divisions over civil rights and those over immigration as arising from a common groundswell of anti-discrimination opinion (Chin Reference Chin1996; DeLaet Reference DeLaet2000; Chin and Spencer Reference Chin and Spencer2015). For civil rights reform, it has been convincingly argued that grassroots organization and local activism pressed the demand for change into the public consciousness and only then into the agendas of legislators (Lee Reference Lee2002). Nothing like this happened in the case of immigration reform, and much less attention has been paid to assessing public opinion on immigration and its potential influence in Congress. Nevertheless, according to DeLaet (Reference DeLaet2000, p. 24):

While the civil rights movement may have provided the initial support for immigration reform, it took on a separate identity as the debate progressed.Footnote ¹⁰ As Marinari (Reference Marinari2014, p. 24) puts it:

However, such arguments have yet to be quantitatively assessed.

The 1965 Gallup Poll

We can obtain unique insight into the climate of public opinion from a range of questions asked in a Gallup poll, which was taken on 24–29 June, 1965, at the time when the immigration bill was being discussed and two months prior to the initial vote on the bill. A representative sample of 3,536 persons of voting age was polled in 38 states plus Washington, DC. The excluded states are Connecticut, New Hampshire, Delaware, North Dakota, South Dakota, Arizona, Idaho, Nevada, New Mexico, Utah, Alaska, and Hawaii. Thus, the western states are somewhat underrepresented. In the remaining states, the numbers polled were broadly in proportion to their populations. Further details of the dataset are provided in Online Appendix 1. The results are summarized in Table 2. As the first panel shows, there was very little popular support for an increase in immigration. One-third of respondents wanted to see immigration decrease, while almost two-fifths wanted to see no change. However, the second panel reveals rather more support for abolishing the national origins system, with just over half in favor and less than a third opposed. Thus, the desire for no increase in the numbers was somewhat mitigated by the support for reforming the country-of-origin quota system. This is all the more impressive given that respondents were explicitly told what the existing system entailed.

Table 2

RESULTS OF GALLUP POLL 713, TAKEN 24–29 JUNE 1965

Source: Gallup Poll 713, supplied by the Roper Center on Public Opinion Research.

The remaining questions relate to the importance of different immigrant characteristics and these can be seen as mediating overall policy preferences (the full correlation matrix is reported in Online Appendix 5, Table A5.1). Most respondents expressed support for bringing refugees into the immigration system and this is positively correlated with favoring abolishing the quotas but negatively with favoring decreasing immigration. Over 70 percent thought that occupational skills were a very important criterion, and this is positively correlated with favoring abolition of the quotas and less strongly with wishing to decrease immigration, while the opposite holds for the view that having relatives in the United States was very important. Nearly one-third thought that country of birth was important, and, not surprisingly, this is positively correlated with preferences for decreasing immigration and negatively with favoring abolition of the quotas. Overall, preferences on desirable immigrant characteristics correspond with those on policy. It is also worth noting that there is no significant correlation between responses to the two policy-related questions, and so these are very distinct elements of public opinion.

Explaining Attitudes to Immigration

There is now a large literature on the relationship between attitudes toward immigration and a range of personal and contextual characteristics (for surveys, see Ceobanu and Escandell Reference Ceobanu and Escandell2010; Hainmueller and Hopkins Reference Hainmueller and Hopkins2014). The strongest and most consistent finding is that more educated individuals tend to have attitudes that are less anti-immigration, either because they face less labor market competition from unskilled immigrants (Scheve and Slaughter Reference Scheve and Slaughter2001; Mayda Reference Mayda2006; Facchini and Mayda 2012) or because those with more education tend to hold more liberal views, are more tolerant toward minorities, and are more positive about ethnic and cultural diversity (Hainmueller and Hiscox Reference Hainmueller and Hiscox2007, 2010). Other variables often included are income and labor market participation, which are linked to the presence or absence of labor market threats but may also relate to the potential fiscal impact of immigration (Hanson, Scheve, and Slaughter 2007; Facchini and Mayda 2009).

Personal characteristics such as age, sex, and ethnic minority or immigrant status often matter, with younger people, females, and minorities often found to be more pro-immigration. Much of the literature, particularly in political science, has emphasized “sociotropic” concerns over individual self-interest. This stems from identification with certain groups or communities, which may run along the lines of class, ethnicity, region or locality, cultural traits, or political affiliation (Citrin et al. Reference Citrin, Green, Muste and Wong1997; Sides and Citrin Reference Sides and Citrin2007; Rustenbach Reference Rustenbach2010; Manevska and Achterberg Reference Manevska and Achterberg2013). Such sociotropic concerns may manifest as racial intolerance and opposition to immigrants by color, creed, or country of origin (Dustmann and Preston Reference Dustmann and Preston2007; Card, Dustmann, and Preston Reference Card, Dustmann and Preston2012). A variety of individual and contextual variables have been used to capture these motivations, although the conclusions drawn from the results are often a matter of interpretation.

We model the responses to the 1965 questionnaire in the light of the existing literature, but subject to the constraints of the variables that are available. We capture the civic generation with a dummy for the younger cohort, specifically those born after 1910, and who reached adulthood in the Great Depression, WWII, and early postwar periods (the minimum age in the data is 20). We also include a dummy for college education (differences among groups with less than a college education were never significant). The income variable is annual household income and is the midpoint of each of the 11 categories. Dummies are included for being Protestant (relative to all others but mainly Catholic), male, and white. A dummy for urban takes the value 1 if the individual was living in an urban area with a population greater than 10,000 (using 50,000 as the cutoff makes little difference). Finally, we include a dummy for those who profess to be Democrats and one for those living in the South (the 11 states that seceded from the Union prior to the Civil War).

Table 3 provides the results of probit regressions, where the coefficients are reported as marginal probabilities and the standard errors are clustered at the state level. We lose 104 observations, mainly due to lack of information on income. In Column (1), the dependent variable is to favor the abolition of the country-of-origin quota system, relative to “oppose” or “no opinion.” The highly educated tended to favor abolition, while those residing in the South tended not to. In Column (2), the dependent variable is support for immigration to be decreased. The highly educated were less likely to want immigration decreased, although the effect is not strong. Protestants and whites were significantly more likely to favor a decrease in immigration, while the other explanatory variables, including the South dummy, are insignificant. Overall, the results paint a rather traditional picture of opposition to immigration. In the responses to both of these policy-related questions, it is notable that the coefficients on the dummies for civic generation, household income, and Democrat are small and insignificant. Thus, while the growth in college education may have shifted attitudes toward reform, the idea that more liberal public attitudes toward immigration were driven by the coming of age of the civic generation, postwar economic prosperity, or the decisive shift toward the Democrats receives little support—at least in the cross-section.

Table 3

EXPLAINING PUBLIC OPINION ON IMMIGRATION

Notes: The coefficients are marginal probabilities from probit regressions; standard errors in parentheses clustered at the state level. *** p<0.01, ** p<0.05, * p<0.1.

Source: Authors’ calculations, see text.

Turning to the questions on the criteria for admitting immigrants, Column (3) shows that Protestants and whites tended to think that country of origin was an important criterion. While this echoes the desire to decrease immigration in Column (2), it does not translate very strongly into resistance against abolishing the quota system, as shown in Column (1). Column (4) shows that those with higher education and higher incomes were more likely to think that skills were very important. This is contrary to what would be expected if immigrants were seen as a labor market threat, but it could be consistent with concerns about the welfare burden of low-skilled immigrants. However, the coefficients on these two variables are negative in the equation for the importance of family in Column (5). As the question relates specifically to support from relatives, this suggests that the welfare motive might not be the highest concern. It is worth noting also that this is the one question for which the dummy for the civic generation is significantly negative, suggesting perhaps a weakening of the emphasis on family support mechanisms. Finally, attitudes toward providing a formal channel for the admission of refugees (Column (6)) were strongest among those with higher education and incomes—the mirror image of attitudes toward the importance of relatives. Interestingly, support for a refugee channel was also strong among whites, urban residents, Democrats, and those residing in the South.Footnote ¹¹

The two responses most directly related to policy are “abolish the quota system” and “decrease immigration.” These tend to give opposite signs, although the coefficients are not always significant. Broadly speaking, they suggest greater support for reform among the more educated and among non-Protestants and non-whites. As with civil rights reforms, opposition to abolishing the quotas was strongest in the South. But perhaps the most striking finding is the lack of significance for party allegiance, which suggests that partisanship was not a major influence on attitudes toward immigration. It remains to be seen how, if at all, such mixed messages conformed with congressional votes.

Explaining Votes in Congress

Among the wide range of quantitative studies analyzing congressional votes, a number have focused on immigration bills. Some of these have examined each vote individually (Gimpel and Edwards Reference Gimpel and Edwards1999; Gonzalez and Kamdar Reference Gonzalez and Kamdar2000; Fetzer Reference Fetzer2006), while others pool different immigration bills (Milner and Tingley Reference Milner and Tingley2009; Facchini and Steinhardt Reference Facchini and Steinhardt2011; Conconi et al. 2020). The focus is chiefly on skills and regional and ethnic composition differences across congressional districts. For earlier votes, Goldin (Reference Goldin, Goldin and Gary1994) for the literacy test and Alston and Ferrie (Reference Alston and Ferrie1999) for the Bracero Program focus on shifting coalitions of regional interests. Both Goldin (Reference Goldin, Goldin and Gary1994) for congressional votes in 1913 and 1915 and Tabellini (Reference Tabellini2020) for votes on the 1924 Act find that votes for restriction were negatively related to the foreign-born share of population and positively to the increase in foreign-born over the preceding decade.Footnote ¹² Only Gimpel and Edwards (Reference Gimpel and Edwards1999) have formally analyzed votes on HR 2580, the bill that became the Hart-Celler Act. In their study, they linked the votes in the House of Representatives with characteristics of both the member of Congress and the district that he/she represented. The main findings were that a pro-reform vote was less likely for representatives from southern states and more likely from districts with higher shares of foreign-born and lower shares of rural population (Gimpel and Edwards Reference Gimpel and Edwards1999, pp. 106–9, 146). However, it remains unclear how far this reflects public opinion, as distinct from other characteristics of the district or of the congressional representative.

Although a number of studies have considered the links between public opinion and congressional voting, little is known about how they relate to immigration policy specifically (see Monroe Reference Monroe1998; Burstein Reference Burstein2003; Shapiro Reference Shapiro2011). One reason is that nationwide polls usually contain too few observations to provide a reliable indicator at the state or district level. One way to overcome this limitation is to use the two-stage MRP method developed by political scientists (Park, Gelman, and Bafumi Reference Park, Gelman and Bafumi2004; Lax and Phillips Reference Lax and Phillips2009a; Kastellec, Lax, and Phillips Reference Kastellec, Lax and Phillips2010). In the first stage, we estimate a multi-level model for individual responses using a range of individual-level socioeconomic and geographical predictors. The resulting estimates for each socioeconomic/geographic cell are then weighted (post-stratified) by the proportions of each type in the state population using census data. This method has been found to produce accurate and robust predictions (Lax and Phillips 2009b). We use it to estimate at the state level the two dimensions of public opinion that are closely related to reform: the proportion favoring decreased immigration and the proportion wishing to abolish the system of quotas. For the choice of variables in the first stage, we rely mainly on the work of Butz and Kehrberg (Reference Butz and Kehrberg2016), who used MRP to estimate anti-immigration preferences in U.S. states in the 2000s. As individual predictors, we use gender, race, age, education, and interactions of race and gender, and age and education. At the state level, we include the religious composition of the population, the proportion of foreign-born, the proportional change in the foreign-born population, and the 1964 presidential vote for Johnson. In the second stage, we use data from the 1960 Census for post-stratification. Further details of this methodology are provided in Online Appendix 2.

We examine two roll call votes in the House of Representatives and one in the Senate. These are taken from Keith Poole’s “Voteview” compilation (Poole and Rosenthal Reference Poole and Rosenthal1997, 2007), which, in addition to the votes cast, includes the representative’s name, party affiliation, and congressional district. From the same source, we also include DW-NOMINATE, a continuous variable that places the representative on a spectrum between conservative and liberal based on past voting behavior (Lewis et al. Reference Lewis, Keith Poole, Boche, Rudkin and Sonnet2021). This reflects the individual’s revealed ideology, which differs within parties, notably between the northern and southern Democrats. Other individual characteristics matched to the name of the representative are taken from Swift et al. (Reference Swift, Brookshire, Canon, Fink, Hibbing, Humes, Malbin and Martis2009). In addition to these, we created a dummy variable for whether the member of Congress had an immigration (foreign-born or foreign parentage) or ethnic minority background using information from myheritage.com. We also include as explanatory variables a dummy for the South, district-level variables for the share of urban population and median family income, and the share of Western Hemisphere immigrants at the state level. Further details on all these variables can be found in Online Appendix 1.

The First House Vote on HR 2580

We first examine the original House vote on HR 2580, which opened the way for progress toward final approval of the bill. Table 4 presents marginal probabilities from probit regressions with standard errors clustered by state.Footnote ¹³ The dependent variable is the representative’s vote, where yea = 1 and nay = 0, including paired and announced votes. We drop states that were not represented in the public opinion survey and cases where no vote was cast, leaving 400 observations.Footnote ¹⁴ The first column of Table 4 reports the results of a parsimonious specification, accounting for only the two public opinion shares generated by MRP. We expect representatives of areas where public opinion is more in favor of abolishing the national origin quotas to be more likely to support the 1965 Act, as this was one of its key provisions. At the same time, as the impact of the reform on the overall number of immigrants was expected to be negligible, we do not necessarily expect representatives of areas where public opinion is more in favor of reducing the number of immigrants to be more likely to oppose it. In Column (1), public opinion on abolishing the quotas is positively and strongly correlated with support for the bill, while, in contrast, public opinion on decreasing immigration takes a small and insignificant coefficient.Footnote ¹⁵ The second column additionally controls for district-level characteristics that are not included in the MRP calculation. Support for the bill was much stronger among representatives of districts outside the South and among those representing more urban constituencies. Accounting for these variables, the coefficient on abolish quotas falls in size by more than half, but it is still significant at the 1 percent level, while opinion on decreasing immigration remains insignificant.

Table 4

HOUSE VOTES ON THE ORIGINAL PASSAGE OF THE 1965 IMMIGRATION BILL

(Marginal effects; dependent variable: Yea = 1, Nay = 0)

Notes: The coefficients are marginal probabilities from probit regressions; robust standard errors in parentheses are clustered at the state level; *** p<0.01, ** p<0.05, * p<0.1.

Source: Authors’ calculations, see text.

The specification reported in the third column of Table 4 accounts for individual characteristics of the representative. Being from the “civic generation” is positively associated with voting in favor of the bill, as is a background in business, which may reflect an employer’s perspective. Having held previous office at the state level is associated with voting against it, which could be consistent with a greater attachment to the pre-existing immigration policy. Not surprisingly, those with an immigrant or ethnic background were more likely to be in favor. Party affiliation is insignificant, in contrast with later periods when voting on immigration policy often followed party lines (Facchini and Steinhardt Reference Facchini and Steinhardt2011; Mayda, Peri, and Steingress Reference Mayda, Peri and Steingress2022). In Column (4), party affiliation is replaced by a more precise measure of the representative’s left-right political orientation using the DW-NOMINATE score (increasing from left to right on a scale of –1 to +1). The coefficient is strongly negative, indicating that right-leaning representatives were much more likely to vote against the bill. Not surprisingly, the effect of this modification is to reduce in size the coefficient for the South, reflecting the fact that southern Democrats were more conservative.

Overall, the results indicate that voting in the House on the original version of the bill was highly consistent with state-level opinion on abolishing the country-of-origin quota system but not with opinion on restricting immigration. Adding district and representative characteristics attenuates the coefficient, as would be expected if local opinion was reflected in the characteristics of the elected representative. These results are robust if we exclude from the analysis individuals reporting no opinion, if we deploy alternative econometric specifications (logit and linear probability models), or if we bootstrap standard errors rather than clustering them at the state level (see Online Appendix 5, Tables A5.5 and A5.6).

Table 5

SENATE VOTES ON THE PASSAGE OF THE 1965 IMMIGRATION BILL

(Marginal effects; dependent variable: Yea = 1, Nay = 0)

Notes: The coefficients are marginal probabilities from probit regressions; robust standard errors in parentheses are clustered at the state level; *** p<0.01, ** p<0.05, * p<0.1.

Source: Authors’ calculations, see text.

Subsequent Votes on HR 2580

The patterns in subsequent roll call votes in the Senate and the House are likely to differ from those of the original House vote for at least three reasons. One is that these votes were on a significantly amended version of the bill, as noted previously. Another is that the process of negotiation and persuasion could have induced some members of Congress to change their initial votes (Binder and Lee Reference Binder and Lee2013). And third, there are structural differences between votes in the House and in the Senate, not least in the types of constituencies that they represent. There was a single roll call vote in the Senate on HR 2580, which took place after lengthy deliberations in committee. The Senate Judiciary Subcommittee on Immigration and Naturalization heard a wide range of testimony and engaged in an intensive internal debate (Hutchinson Reference Hutchinson1981, p. 376). As is often noted, in this debate, members of the subcommittee and its witnesses produced some of the most widely quoted statements about the likely results of the proposed reform. The version of the bill that was reported out contained a number of modifications and compromises, the most important of which was the cap on Western Hemisphere immigration. In the event, the amended bill was passed by a comfortable majority.

Table 5 presents the results of the roll call vote in the Senate. As before, they include announced and paired votes, but as we omit senators who did not vote and states for which we lack opinion data, we have just 76 observations. Because the public opinion variables are measured at the state level, they exactly match the Senate constituencies. In the first column, the coefficient on “abolish quotas” is large and significantly positive, while that on “decrease immigration” is not, results that are in line with those for the House reported in Table 4. Similarly, senators from the South tended to vote against the bill, while those representing more urban states tended to support it. In Columns (3) and (4), the coefficients on the senators’ characteristics are less robust, as might be expected given the limited number of observations. Unfortunately, we are unable to control for migrant/ethnic backgrounds, as all senators with migrant/ethnic backgrounds supported the bill. Neither party affiliation nor the DW-NOMINATE score is significant, suggesting less partisanship over the Senate version of the bill. Even in the presence of controls for state and senator, the coefficient on abolish quotas remains positive and significant at the 1 percent level, while that on “decrease immigration” remains insignificant.

In Table 6, we turn to the vote in the House on the conference report, which, as previously noted, was essentially the Senate’s version of the bill. As fewer members cast a vote, the number of observations drops to 383. Similar to the result for the original passage of the bill, when public opinion is included with no other controls, the coefficient on “abolish quotas” is strongly positive, while that on “decrease immigration” is insignificant. However, in Columns (2) to (4), the coefficient on decrease immigration becomes significantly positive, contrary to what might be expected. This suggests that the revised, and more restrictive, version of the bill, and the reassurances that came with it, brought on board members from more anti-immigration districts. In Columns (2) to (4), the negative coefficients on the South and positive coefficients on the share urban are similar to those reported in Table 4. Individual member characteristics give less significant coefficients—in particular, that on ethnic/immigrant background, discussed later. It is notable that the coefficient on the DW-NOMINATE score declines in size and significance, consistent with the result for the Senate, suggesting decreased partisanship. As in Table 4, the coefficients on “abolish quotas” remain positive and significant at the 1 percent level in all four specifications, although the coefficients are reduced in magnitude by about one-third.

Table 6

HOUSE VOTES ON THE CONFERENCE REPORT OF THE 1965 IMMIGRATION BILL

(Marginal effects; dependent variable: Yea = 1, Nay = 0)

Notes: The coefficients are marginal probabilities from probit regressions; robust standard errors in parentheses are clustered at the state level; *** p<0.01, ** p<0.05, * p<0.1.

Source: Authors’ calculations, see text.

The results for the vote in the Senate and for the House vote on the conference report differ in some respects from those on the earlier version of the bill. But in all three votes, the coefficient on abolishing country-of-origin quotas remains large and significant even when controlling for key characteristics of the state and of the individual representative. This suggests a clear consistency between local attitudes toward a less discriminatory immigration policy and congressional votes, which is not simply mediated through the characteristics of the representative and the state. In contrast, there is little evidence that congressional votes conformed with public opinion on decreasing the number of immigrants, which is often seen as one of the key connections between public opinion and immigration policy. Indeed, for the House vote on the final version of the bill that became the Hart-Celler Act, the association between votes and anti-immigration opinion becomes significantly positive. In the next section, we explore two questions raised by these results. First, to what extent did support for abolishing the quotas follow directly from attitudes toward civil rights reform? And second, how, if at all, can we reconcile the passing of the Hart-Celler Act with local opinion on the scale of immigration?

The Influence of the Civil Rights Movement

As noted previously, some observers see the immigration reform as driven by the same popular imperatives that underlie the Civil Rights Act of 1964 and the Voting Rights Act of 1965. Not surprisingly, racial issues consistently ranked as one of the most salient political issues in the mid-1960s. One indicator of the importance of these concerns is illustrated in Figure 1, where we plot the proportion of respondents in the American National Election Studies (ANES) who listed “racial problems” as the most important issue (among nine alternatives) that the government in Washington should take care of. This question was asked in 1960 and then every two years from 1964 until 2000. As Figure 1 shows, the salience of the racial issue reached a peak in the mid-1960s, which was not matched before or after. But even though both civil rights reform and immigration reform fall loosely under the anti-discrimination umbrella, it is not clear how closely opinions on these issues were linked.

Figure 1

SALIENCE OF RACIAL ISSUES IN ANES POLLS

Notes: Responses to question VCF0875: “What are the most important problems the government in Washington should try to take care of?” The graph shows the proportion of respondents listing “racial problems” as the most important issue.

Source: ANES data from the American National Electoral Studies, Time Series Cumulative Data File at https://electionstudies.org/data-center/anes-time-series-cumulative-data-file/.

In the 1964 election campaign, immigration as an issue was massively overshadowed by civil rights (and also by the Vietnam War). Nevertheless, the Immigration Act, which followed in the wake of the landmark civil rights acts, is seen as embodying the same fundamental principle of anti-discrimination (Chin Reference Chin1996; Chin and Spencer Reference Chin and Spencer2015). However, it is far from clear that precisely the same sentiments that underpinned the civil rights revolution applied equally to immigration or that they applied with the same force across the nation (Graham Reference Graham and John2001). For one thing, most African Americans, who were the focus of the Civil Rights Act, were in the South whereas most immigrants were in the northern states. Indeed, the correlation between the share of African Americans and the share of foreign-born is –0.46 across states and –0.18 across congressional districts. Here we examine the links between the Civil Rights Act and the Hart-Celler Act, first by examining the correspondence across districts of votes on these two pieces of legislation, and second by examining directly the correlation across states between public opinion on civil rights and House votes in favor of immigration reform.

How do House votes on immigration by congressional district correlate with votes on the Civil Rights Act? Table 7 reports the correspondence by district between the first House vote on what became the Civil Rights Act of 1964 and the two House votes on immigration we have considered so far that took place in 1965. Not surprisingly, there is a fairly strong correspondence between these votes, and the value of Cramér’s V is 0.49 for the first House vote and 0.44 for votes on the conference report. But there are some differences. Of the 43 districts whose representatives voted against civil rights reform but in favor of immigration reform in the first House vote, 12 were districts that switched from Republican to Democrat in the 1964 election. And of the 13 who switched from yea on civil rights to nay on immigration reform, two switched from Democrat to Republican. While the changing composition of the House may account for some of the differences between the initial votes on civil rights and on immigration, there were also differences between the two House votes on immigration reform, where a total of 59 representatives changed their votes—22 of those who supported the bill in the first vote abstained in the second, while another 27 shifted from nay to yea. This somewhat weakened the correspondence across districts between votes for civil rights and immigration reform.Footnote ¹⁶

Table 7

VOTING BY CONGRESSIONAL DISTRICT ON CIVIL RIGHTS AND ON IMMIGRATION

Notes: To compare votes between the 88th and 89th Congresses, it was necessary to match districts. Where representatives were elected at large in the 88th Congress, they were matched to the 89th Congress by name. In the case of the eight seats in Alabama, all elected at large in the 88th Congress but not in the 89th, only three were the same members, but as all eight voted against the Civil Rights Bill, it does not affect the comparison.

Source: Calculated from the “Voteview” database at https://voteview.com/data.

The correspondence between congressional votes may simply reflect political opinion “within the beltway” rather than any spatial consistency between these issues in grassroots public opinion. However, we may shed some light on this issue by examining the correlations across states between support for civil rights and opinion on abolishing the immigration quotas. In late 1964, two Gallup Polls surveyed opinions about the Civil Rights Act, which had been passed just a few months earlier. These polls asked: “As you know, a civil rights law was recently passed by Congress and signed by the President. In general, do you approve or disapprove of this law?” For comparison with opinion on abolishing the quotas, we combine the results of these two polls and apply MRP using the same variables that were applied to opinion on immigration.Footnote ¹⁷ Figure 2 shows the scatterplot of state average predictions for the proportion of respondents who, in 1965, wished to abolish the country-of-origin quotas and the proportion who, a year earlier, stated their approval of the Civil Rights Acts. This is for the 34 states covered in all surveys. There is a clear positive correlation (0.48), but nevertheless, there are some differences. Not surprisingly, approval of the Civil Rights Act was lower among southern states, but also in Montana and Wyoming, while the proportion wishing to abolish the quotas varies more widely. Among other states, there is considerable variation in the responses to both questions.

Figure 2

STATE-LEVEL OPINION ON ABOLISH QUOTAS AND APPROVAL OF CIVIL RIGHTS LEGISLATION (PREDICTED VALUES) Sources: MRP predictions of the proportion of survey respondents who approved of the 1964 Civil Rights Act and the proportion who supported the abolition of the country-of-origin immigration quotas.

Even though opinion on abolishing the quotas does not move in lockstep with attitudes on civil rights, the latter may still have been linked with congressional votes as, suggested by the correspondence between votes on civil rights and on immigration in Table 7. In order to assess this hypothesis, we use the MRP prediction on approval of civil rights legislation in regressions for the House votes on HR 2580. As the regressions reported in Table 8 are estimated over the 34 states for which we have opinions on both immigration and civil rights, there are slightly fewer observations than in the comparable regressions in Tables 4 and 6. The first column is otherwise similar to Column (4) of Table 4, except that “civil rights approve” is used in place of “abolish quotas.” The positive and significant coefficient is consistent with the idea that the first vote in the House rode on the coattails of civil rights reform. When both “civil rights approve” and “abolish quotas” are included in the regression (Column (2)), the coefficient on the former decreases in size but remains significant. The results for the conference report in Columns (3) and (4) of Table 8 provide a sharp contrast. Here, the coefficient on “civil rights approve” is small and insignificant, whether or not “abolish quotas” isalso included. So while the first round of voting in the House was consistent with grassroots attitudes on civil rights, that impetus seems to have faded by the time of the conference report more than a month later.

Table 8

HOUSE VOTES ON THE 1965 IMMIGRATION BILL AND OPINION ON CIVIL RIGHTS LEGISLATION

(Marginal effects; dependent variable: Yea = 1, Nay = 0)

Notes: The coefficients are marginal probabilities from probit regressions; robust standard errors in parentheses are clustered at the state level; *** p<0.01, ** p<0.05, * p<0.1. Columns (2) and (4) differ from the preceding columns only by the addition of opinion on “Abolish quotas.”

Source: Authors’ calculations, see text.

As mentioned earlier, 59 representatives changed their votes, which largely reflected negotiations in the Senate and the debate on the floor of the House. Gimpel and Edwards note that “the legislation lost the support of many minority members of Congress—Blacks, Hispanics and Asians—as roughly half of them voted against the legislation” (1999, p. 108). Table 9 provides details of those who switched their votes, and these are listed individually in Online Appendix 4. Of the 32 who switched their vote in a positive direction (rows 1 and 2), only three had immigrant or ethnic backgrounds. Of the 27 who switched their vote in a negative direction (rows 3, 4, and 5), 11 had immigrant or ethnic backgrounds. Regression results presented in Online Appendix 4, Table A4.2, confirm that the bill lost support from those with immigrant/ethnic backgrounds. They also show that representatives of districts where opinion more strongly favored abolishing the quotas were less likely to switch away from supporting the bill, whereas those from districts with higher shares of immigrants from the Western Hemisphere were more likely to do so. A plausible explanation for this finding is represented by the cap placed on Western Hemisphere immigration, a change that also helped to swing some of those with the strongest anti-immigration constituencies behind the Act. Finally, Democrats and those having held previous political office were less likely to switch in either direction.

Table 9

HOUSE MEMBERS WHO SWITCHED THEIR VOTES BETWEEN THE ORIGINAL AND FINAL BILLS

Sources: From the “Voteview” database at https://voteview.com/data and author calculations.

Immigration Reform and Anti-Immigration Pressures

While the votes of members of Congress clearly mirrored the opinions of their constituents on abolishing the country-of-origin quotas, equally intriguing is the lack of correspondence with anti-immigration attitudes. Our results indicate that local opinion on whether immigration should be decreased is not associated with votes on the initial House version of the bill or with the votes in the Senate. On the other hand, local anti-immigration sentiment was positively associated with representatives’ voting behavior on the final version of the bill, but only in the presence of the full set of controls. Observers have often commented on the apparently anomalous fact that a reform that is credited with leading to an unprecedented increase in immigration met with so little resistance in Congress. In just two decades, total immigration almost doubled, from an annual average of 290,000 in 1961–5 to 573,000 in 1981–5. Many would agree with Daniels (Reference Daniels2002, p. 341) who observed that “it is doubtful if any drafter or supporter of the 1965 Act envisaged this result” (see also Hatton Reference Hatton2015). Here we look at three issues. First, was the public salience of immigrant numbers uniquely low in the mid-1960s? Second, to what extent did legislators anticipate that the reform would increase immigration? And third, what contribution did the 1965 Act actually make to the subsequent dramatic increase in immigration?

After four decades of low immigration, salience of the total volume of immigration as a pressing policy issue in public opinion was low in the 1960s, and hence it was less of a political imperative. Although we do not have a direct measure, salience can be inferred from the percentage of respondents to the question on whether immigration should be increased/decreased who gave no opinion or answered “don’t know,” where higher percentages reflect low salience. Figure 3 reports this (inverse) measure of salience for various years from 1955 to 2016.Footnote ¹⁸ In 1965, salience was lower than at any other time, and by a substantial margin. This is also supported by the low level of coverage of immigration in prominent news outlets in the 1960s (Simon Reference Simon1985, pp. 134–160). In light of muted opposition to immigration and its low salience, perhaps it is not so surprising that commentators have seen immigration reform as stemming largely from “inside the beltway” and largely decoupled from public opinion.

Figure 3

SALIENCE OF IMMIGRATION AS REFLECTED BY “NO OPINION” ON INCREASE/DECREASE IMMIGRATION

Sources: Gallup polls reported by Lynch and Simon (Reference Lynch and Simon2003 p. 47); ANES data from the American National Electoral Studies, Time Series Cumulative Data File at https://electionstudies.org/data-center/anes-time-series-cumulative-data-file/.

Yet, as has often been noted, the leading proponents of the bill went to some lengths to reassure their colleagues that it would bring about only modest increases in immigration. In the Senate Judiciary Subcommittee on 10 February 1965, Edward Kennedy famously stated that “The bill will not flood our cities with immigrants. It will not upset the ethnic mix of our society. It will not relax the standards of admission. It will not cause American workers to lose their jobs.” Attorney General Nicholas Katzenbach assured the committee that the bill would not increase or accelerate immigrant arrivals by more than a small fraction. Speaking in the House debate, Emmanuel Celler argued that:

Such statements have often been cited in retrospect with considerable irony, but they do seem to have been believed at the time (Reimers Reference Reimers1992, pp. 76, 81, 92).Footnote ¹⁹ In contrast, dissenting voices were very much in the minority (Graham Reference Graham2004, p. 90).

From the perspective of 1965, was it reasonable to believe that there would be only a modest increase in the numbers? Immediately before the 1965 Act came into force, the Eastern Hemisphere quota stood at 158,561 per annum, and so the increase in the numerical limit to 170,000 represents an increase of just 7 percent. But under the McCarran-Walter Act, the severely skewed country-of-origin quotas meant that the number admitted subject to the numerical limit had been only two-thirds of the overall quota; in 1961–5, it averaged just over 98,000. At first sight, it would seem reasonable to anticipate that once the Act became effective the quota would be filled, which would imply an increase of 73 percent. Reassurances such as that of Emmanuel Celler, while often interpreted as focusing principally on the ethnic mix of future immigration, also imply that the numerical limit would not be filled, at least not in the near future. But the numerical cap was filled almost immediately. How far this could or should have been anticipated by members of Congress is a moot point.Footnote ²⁰

Perhaps more difficult to anticipate was the growth in the number of immigrants not subject to numerical limitation. Those admitted from the Eastern Hemisphere but not subject to the numerical limit increased from around 54,000 in 1961–5 to more than 207,000 in 1981–5, so the ratio of those exempt from the numerical limit to those subject to it increased from 0.55 in 1961–5 to 1.37 in 1981–5. An important reason for this is that the family reunion multiplier was much larger for immigrants from poor countries than for those from richer countries (Hatton Reference Hatton2015). While this might seem self-evident in retrospect, it might not have been so clear at the time, as past immigrants had been overwhelmingly from countries for which the multiplier was low. Another important factor is that a significant number of the original immigrants who, once settled, eventually brought in their immediate relatives (both within and outside the numerical limit) were admitted as refugees outside the Act, rather than as primary immigrants under the Act.Footnote ²¹

Much of the debate focused on immigration from the Western Hemisphere, which had not previously been subject to numerical limitation. These immigrants were admitted through a process of labor certification and could bring immediate family members, but those with non-immigrant status present in the United States were not permitted to adjust to permanent resident status (and therefore could not sponsor other family members). In 1961–5, an average of 137,000 were admitted per annum (only 2 percent of whom were spouses and children). With the ending in 1964 of the Bracero Program, which in its last five years admitted an average of 254,000 temporary laborers (92 percent from Mexico), a substantial increase in the number of Western Hemisphere immigrants might have been expected. The introduction of a numerical limit of 120,000 in the final version of the Act helps to explain why those representing districts with a higher proportion of population wishing to see immigration decrease voted in favor of it. Indeed, in the absence of a system of preferences and with a (tighter) procedure for labor certification and the restrictions on adjustment remaining in place, the 1965 Act could be seen as highly restrictive on immigration from the Western Hemisphere.Footnote ²² As Massey and Pren (Reference Massey and Pren2012, p. 2) put it, “the 1965 legislation in no way can be invoked to account for the [subsequent] rise in immigration from Latin America.”

So perhaps the massive increase in migration from Mexico could not have been easily foreseen. While the dramatic rise in undocumented migration across the southern border has attracted most of the attention, there was also a substantial increase in legal immigration from the Western Hemisphere, which rose from an average of 137,000 in 1961–4 to 210,000 in 1981–4. To a large extent, the rise in Western Hemisphere immigration up to the 1980s and beyond was not due to the Hart-Celler Act itself but to subsequent Acts, which were beyond the purview of the legislators of 1965. In particular, the 1976 Amendments introduced into the Western Hemisphere the same system of preferences (weighted toward family reunion) that were applied to the Eastern Hemisphere and, importantly, also conferred the right to adjust to permanent residency.Footnote ²³ Later, the Immigration Reform and Control Act of 1986 legalized 2.7 million undocumented immigrants, of whom more than 90 percent were from the Western Hemisphere.

To summarize, three factors help to explain why anti-immigration sentiment failed to galvanize more resistance in Congress to an Act, which, according to later accounts, opened the door to a massive increase in immigration. The first is that by the mid-1960s, after decades of low immigration, the salience in public opinion of the total volume of immigration was at an all-time low. The second is that members of Congress may have underestimated the effects of the Act on the likely number of immigrants, especially the number not subject to numerical limitation. And third, the 1965 Act itself accounts for only a part of the increase in immigration that is often attributed to it.