Puzzle Data and Methodology Results Conclusion
Where are the Winners:The Determinants of US Immigration Policy, 1795-2008
Margaret E. PetersStanford University
November 13, 2009
1
Puzzle Data and Methodology Results Conclusion
The Forms of Globalization
19th Century 20th Century
Trade Closed Open
Immigration Open Closed
Puzzle 2
Puzzle Data and Methodology Results Conclusion
Economic Openness in the US
Source: Immigration flows: U.S. Citizenship and Immigration Services (2009) .Tariffs: Historical Statistics of the United States – Millennial Edition (Carter etal. 2006).
Puzzle 3
Puzzle Data and Methodology Results Conclusion
Traditional Explanations for International Economic Policy
• Importance of domestic interest groups
• Apply models of trade policy formation to immigration policy
Interest Groups
Trade Immigration
Producers Producers
Labor Labor
Nationalists Nativists
Puzzle 4
Puzzle Data and Methodology Results Conclusion
Traditional Explanations for International Economic Policy
• Importance of domestic interest groups
• Apply models of trade policy formation to immigration policy
Interest Groups
Trade Immigration
Producers Producers
Labor Labor
Nationalists Nativists
Puzzle 4
Puzzle Data and Methodology Results Conclusion
The US Case
• Hard Case
• Assume that party will predict voting
• Importance in the world economy
• Precursor to EU?
• 50 small “countries”
Puzzle 5
Puzzle Data and Methodology Results Conclusion
Overview of the Argument
• Use a formal model based on Grossman and Helpman’s (1994)model of trade protection and apply to immigration
• Rely on menu auction problems more generally
• Intuitively, model generates prediction that as demands ofpro-immigrant groups decrease anti-immigrant groups becomemore influential
• Examine when pro-immigrant groups demands for opennesschange
Puzzle 6
Puzzle Data and Methodology Results Conclusion
Overview of the Argument
• Use a formal model based on Grossman and Helpman’s (1994)model of trade protection and apply to immigration
• Rely on menu auction problems more generally
• Intuitively, model generates prediction that as demands ofpro-immigrant groups decrease anti-immigrant groups becomemore influential
• Examine when pro-immigrant groups demands for opennesschange
Puzzle 6
Puzzle Data and Methodology Results Conclusion
Overview of the Argument
• Use a formal model based on Grossman and Helpman’s (1994)model of trade protection and apply to immigration
• Rely on menu auction problems more generally
• Intuitively, model generates prediction that as demands ofpro-immigrant groups decrease anti-immigrant groups becomemore influential
• Examine when pro-immigrant groups demands for opennesschange
Puzzle 6
Puzzle Data and Methodology Results Conclusion
Producer Support for Immigration
• Producers support for immigration varies inversely with
• Labor productivity• Ability to produce overseas
• As these factors increase, support for immigration byproducers decrease
• Anti-immigrant groups have more influence over policy
Puzzle 7
Puzzle Data and Methodology Results Conclusion
Producer Support for Immigration
• Producers support for immigration varies inversely with
• Labor productivity• Ability to produce overseas
• As these factors increase, support for immigration byproducers decrease
• Anti-immigrant groups have more influence over policy
Puzzle 7
Puzzle Data and Methodology Results Conclusion
Producer Support for Immigration
• Producers support for immigration varies inversely with
• Labor productivity• Ability to produce overseas
• As these factors increase, support for immigration byproducers decrease
• Anti-immigrant groups have more influence over policy
Puzzle 7
Puzzle Data and Methodology Results Conclusion
Roll Call Votes in the Senate
• Use senators’ vote on immigration in a non-traditional way
• Best historical data on support for immigration
• Uses state level variation Importance of State Level Variation
• Hard test
• Assume that party will often perfectly predict voting behavior
Data and Methodology 8
Puzzle Data and Methodology Results Conclusion
Roll Call Votes in the Senate
• Use senators’ vote on immigration in a non-traditional way
• Best historical data on support for immigration
• Uses state level variation Importance of State Level Variation
• Hard test
• Assume that party will often perfectly predict voting behavior
Data and Methodology 8
Puzzle Data and Methodology Results Conclusion
Histogram of Votes per Year
020
40
60
Vote
s p
er
Year
1800 1850 1900 1950 2000Year
Histogram of Votes per Year
Data and Methodology 9
Puzzle Data and Methodology Results Conclusion
Percent of Expansive Votes Given Votes Reaching the Floor
Data and Methodology 10
Puzzle Data and Methodology Results Conclusion
Can Voting be Predicted by Party Alone?
Voting by Party
Not Cohesive Perfectly Cohesive
Number 1248 143
Percent 90% 10%
Voting by State
Senators Vote Differently Senators Vote the Same
Number 9169 23993
Percent 28% 72%
Data and Methodology 11
Puzzle Data and Methodology Results Conclusion
Positions of the Two Parties
Coefficents from logit regression on each vote regressed on an indicator forRepublicans.
Data and Methodology 12
Puzzle Data and Methodology Results Conclusion
Positions of the Regions
Coefficents from logit regression on each vote with party as control variable;the Northeast is the omitted category.
Data and Methodology 13
Puzzle Data and Methodology Results Conclusion
Observable Implications of My Argument and AlternativeHypotheses
Concept Variable Expected Sign
Productivity Capital in Agriculture –
Productivity Value produced per worker –
Ability to offshore Workers/ Establishment –
Rise of Labor Union participation rates –
Rise of Nativism Percent Foreign Born –
Results 14
Puzzle Data and Methodology Results Conclusion
Specification for Multivariate Analysis
• Agriculture:
voteit = α + β1acapoveragdpit + β2numestabit + β3perforeignbornit +β4agriwageit + β5perunionit + beta6perunemploymentit +β7cashtransferspercapitait + β8republicanit + εit
• Manufacturing:
voteit = α + β1workersperestabit + β2numestabit + β3perforeignbornit +β4manuwageit + β5perunionit + beta6perunemploymentit +β7cashtransferspercapitait + β8republicanit +β9valueproducedperworkerit + εit
Results 15
Puzzle Data and Methodology Results Conclusion
Multivariate Analysis : My Argument
Concept Variable Expected Sign Coefficient: Panel Coefficient: Logit
Productivity Capital in Agriculture – -0.01** 5%/ 60%
Productivity Log value produced per worker – 0.1*** 35%/ 61%
Ability to offshore Workers/ Establishment – 5.06e-05 2%/ 42%
Reference Republicans – 0.33*** 4%/ 62%
*** p<0.01, ** p<0.05, * p<0.1Panel Regression: Binomial regression on imputated data, 1950-2008Logit Regression: Percent of all coefficients with the hypothesized sign andstatistically significant from logit regressions on each vote/ percent of significantcoefficents with the hypothesized sign, 1950-2008
Full Results
Results 16
Puzzle Data and Methodology Results Conclusion
Alternative Hypotheses
Concept Variable Expected Sign Coefficient: Panel Coefficient: Logit
Rise of Labor Union participation rates - 1.42*** 7%/ 46%
Rise of Nativism Percent Foreign Born - -1.81*** 7%/ 36%
Reference Republicans – 0.33*** 4%/ 62%
*** p<0.01, ** p<0.05, * p<0.1Panel Regression: Binomial regression on imputated data, 1950-2008Logit Regression: Percent of all coefficients with the hypothesized sign andstatistically significant from logit regressions on each vote, 1950-2008Coefficents from manufacturing specification - results do not change significantlybetween specifications.
Results 17
Puzzle Data and Methodology Results Conclusion
Support for Alternative Hypotheses
• Coefficient on unions
• Only examining modern period• Hypothesis: Unions have begun representing more foreign born
workers?• Hypothesis: Rise of public sector unions?
• Coefficent on nativism is not robustly significant
• Hypothesis: Is it nativism or the position of the median voter?
Results 18
Puzzle Data and Methodology Results Conclusion
Support for Alternative Hypotheses
• Coefficient on unions
• Only examining modern period• Hypothesis: Unions have begun representing more foreign born
workers?• Hypothesis: Rise of public sector unions?
• Coefficent on nativism is not robustly significant
• Hypothesis: Is it nativism or the position of the median voter?
Results 18
Puzzle Data and Methodology Results Conclusion
Robustness Checks
• Results are robust to different specification of the dependentvariable
• Results are robust to choosing different subsets of votes
• Key legislation• Final passage
• Robust to several different models
• OLS with Robust SE (DV: proportion expansive votes)• OLS with senator fixed effects (DV: proportion expansive
votes)
• With clustered se by congress (DV: proportion expansivevotes)
• With clustered se by state (DV: proportion expansive votes)
• Tobit with senator random effects (DV: proportion expansivevotes)
• Binomial with Robust SE (DV: number of expansivevotes)
Results 19
Puzzle Data and Methodology Results Conclusion
Review of the Argument on Immigration Policy
• Key to understanding policy formation is to understand whenproducers are willing to organize
• Producers have outside options available
• Mechanization• Oppotunities to move production overseas
• Led to less incentive to lobby for open immigration
• Opened up policy space for anti-immigrant groups
Conclusion 20
Puzzle Data and Methodology Results Conclusion
Review of the Argument on Immigration Policy
• Key to understanding policy formation is to understand whenproducers are willing to organize
• Producers have outside options available
• Mechanization• Oppotunities to move production overseas
• Led to less incentive to lobby for open immigration
• Opened up policy space for anti-immigrant groups
Conclusion 20
Puzzle Data and Methodology Results Conclusion
Review of the Argument on Immigration Policy
• Key to understanding policy formation is to understand whenproducers are willing to organize
• Producers have outside options available
• Mechanization• Oppotunities to move production overseas
• Led to less incentive to lobby for open immigration
• Opened up policy space for anti-immigrant groups
Conclusion 20
Puzzle Data and Methodology Results Conclusion
Review of Results
• Producer interests matter
• Preferences over immigration change in a predictable way
• Labor
• Preferences may change with composition of unions
• Nativism
• Not clear if measures nativism or median voter
Conclusion 21
Puzzle Data and Methodology Results Conclusion
Immigration to Globalization
• Immigration a product of producers’ preferences
• Producers’ preferences change with ability to produce overseas
• Overseas production has expanded due to choice of trade andcapital mobility policies
• Reduced the size of the open immigration coalition
• Trade openness ⇒ Closed immigration
Conclusion 22
Puzzle Data and Methodology Results Conclusion
Thank you
Conclusion 23
Appendix
Appendix
Appendix 24
Appendix
Coefficient on Republicans
Coefficents from logit regression on each vote regressed on an indicator forRepublicans.
Appendix 25
Appendix
Productivity: Coefficient on Capital in Agriculture
Coefficents from logit regression on each vote with party as control variable.Years/ states missing data are dropped from the regressions.
Appendix 26
Appendix
Productivity: Coefficients on Value Produced per Worker
Coefficents from logit regression on each vote with party as control variable.Years/ states missing data are dropped from the regressions.
Appendix 27
Appendix
Ability to Offshore: Coefficents on Workers PerEstablishment
Coefficents from logit regression on each vote with party as control variable.Years/ states missing data are dropped from the regressions.
Appendix 28
Appendix
Rise of Labor: Coefficients on Unions
Coefficents from logit regression on each vote with party as control variable.Years/ states missing data are dropped from the regressions.
Appendix 29
Appendix
Rise of Nativism: Coefficients on Percent Foreign Born
Coefficents from logit regression on each vote with party as control variable.Years/ states missing data are dropped from the regressions.
Appendix 30
Appendix
Native and Foreign Born by Industry
0
5
10
15
20
25
30
35
40
1910 1920 1940 1950 1960 1970 1980 1990
Perc
ent o
f Men
in a
Giv
en In
dust
ry
Year
Distribution of Men by Industry
Native men in agriculture
Native Men in Manufacturing (durable)
Native Men in Manufacturing (non-durable)
Foreign Born Men in Agriculture
Foreign Born Men In Manufacturing (Durable)
Foreign Born Men in Manufacturing (Nondurable)
Appendix 31
Appendix
Agricultural Productivity by Region
Agricultural Productivity by Region Capital in Agriculture/ Production in Agriculture
Year
Capi
tal i
n A
gric
ultu
re/P
rodu
ctio
n in
Agr
icul
ture
00.
250.
50.
751
1850 1885 1916 1947 1978 2009
NortheastIndus MidwestAgri MidwestSouthMT WestPacificBorder States
Appendix 32
Appendix
Value Produced per Worker by Region
Appendix 33
Appendix
Agricultural Wages by Region
Appendix 34
Appendix
Manufacturing Wages by Region
Appendix 35
Appendix
Workers per Establishment by Region
Appendix 36
Appendix
Number of Establishments by Region
Appendix 37
Appendix
Welfare Spending by Region
Appendix 38
Appendix
Why Governments Let Firms Leave
Back to presentationAppendix 39
Appendix
Regression Results
Back to presentation
Appendix 40
Appendix
Why State Level Variation is Important
0
20
40
60
80
100
1820
1833
1846
1859
1872
1885
1898
1911
1924
1937
1950
1963
1976
1989
Immigra'on, Trade Openness, and Produc'vity
Average Tariffs (% Value)
Capital in Agriculture ($10 million)
Immigrants/ GDP ($10 million)
Value Added per Worker ($1000)
Back to presentation
Appendix 41
Appendix
Coefficient on Republicans - Procedural and Final Votes
Coefficents from logit regression on each vote regressed on an indicator forRepublicans.
Appendix 42