Political Persecution or Economic Deprivation? A Time-Series Analysis of Haitian Exodus, 1990–2004

Abstract

This study addresses the factors that lead individuals to flee their homes in search of refuge. Many argue that individuals abandon their homes in favor of an uncertain life elsewhere because of economic hardship, while others argue that threats to their lives, physical person, and liberty cause them to flee. This study engages the debate by analyzing flight patterns over time from Haiti to the United States as a function of economic and security factors. Which factors have the largest influence on Haitian-U.S. migratory patterns? Our results show that both economics and security play a role. However, our analyses are able to distinguish between the effects of different individual economic and security indicators on Haitian-U.S. migration.

Keywords:

• Haiti
• persecution
• political violence
• refugees
• US foreign policy

The authors thank Hongri Jiang and Andrew Reeves for their research assistance. Finally, we wish to thank some individuals who provided useful comments and suggestions: Greg Miller, Glenn Palmer, Jaroslav Tir, Samantha Meek, Clare Hatfield, and anonymous reviewers. Several sources of support made this study possible. The American Political Science Association's Small Research Grant Program and the Chappell Faculty-Student Research Fellowship awarded by the Charles Center at William & Mary facilitated the completion of the study. Grants awarded by the National Science Foundation (SES 0516545 & 0214287) supported generation of the domestic and foreign conflict-cooperation data used in the study.

Notes

∗∗∗ = 0.01 level

∗∗ = 0.05 level

∗ = 0.10 level (one tail tests).

^a These tests were run without robust standard errors (given the limitations of the test).

¹We filed a written request to obtain the U.S. Coast Guard's logs.

²See http://www.uscg.mil/hq/g-cp/comrel/factfile/, accessed 9/5/05.

³See http://www.uscg.mil/hq/g-cp/comrel/factfile/, accessed 9/5/05.

⁴Week-level refugee/migration data is not available.

⁵See Shellman and Stewart (2007) for more information on coding rules and procedures.

⁶Such projects include: Cooperation and Peace Data Bank (COPDAB), World Events Interaction Survey (WEIS), Integrated Data for Events Analysis (IDEA), Protocol for the Assessment of Nonviolent Direct Action (PANDA), Intranational Political Interactions Project.

⁷See http://web.ku.edu/keds/ for information on the KEDS and TABARI projects.

⁸TABARI recognizes pronouns and dereferences them. It also recognizes conjunctions and converts passive voice to active voice (Schrodt, 1998).

⁹These particular data are coded from Associated Press reports available from Lexis-Nexis.

¹⁰See “World Event/Interaction Survey (WEIS) Project, 1966–1978,” ICPSR Study No. 5211.

¹¹KEDS has introduced new codes in addition to those used by McClelland and the WEIS project. Most of these are borrowed from the Protocol for the Assessment of Nonviolent Direct Action (PANDA) project. The KEDS project investigators assigned weights to the new codes that are comparable to the Goldstein weights, and we used those weights in tandem with the Goldstein weights to create the scaled event data series analyzed in this study. See http://web.ku.edu/keds/data.html for WEIS codes and adaptations PANDA.

¹²We also experimented with separating out the military from the government.

¹³See http://laborsta.ilo.org/.

¹⁴However, the data range from similar starting and ending values and have similar means.

¹⁵It would not be surprising to find that there is a level shift, since it is clear from looking at the series that there is a clear downward shift in the series though the means are similar and share a similar range.

¹⁶If we had merged the level in first and then taken the first difference, this would not be the case as several observations would be zero since the monthly value did not change from week 3 to week 4.

¹⁷To illustrate, a change from 0 to 5 and a change from 50 to 55 both result in a five unit change, while 0 and 50 and 5 and 55 are very different levels.

¹⁸The leadership and groups remain consistent from 1997–2004 so we feel that using the existing dictionaries rather than creating new ones does not pose great threat to the data's reliability and validity.

¹⁹See Moore and Shellman (2004a) for more on the use of a ZINB on these types of data.

²⁰The test basically compares the log-likelihood values from restricted and unrestricted models.

²¹Note that the Haiti CPI dummy variable is negative and statistically significant. This means that the level of the series changes when the new CPI series begins. This is not surprising, given the large downward shift in the actual CPI series. The negative coefficient on the dummy variable accounts for this change in the series. We then checked to see if there was a different slope prediction for the second series by inter-acting the dummy variable with ΔHAITICPI. A significant coefficient would indicate that the second part of the series has different effect on the dependent variable, which would be disturbing. However, the results confirmed no such relationship, as the interaction term was not statistically significant.

²²For an analysis of the relationship between GNP and wages, see the U.S. Government Import Administration's report on the topic at: http://ia.ita.gov/wages/98wages/98wages.htm.

²³ http://www.refintl.org/content/article/detail/3897.

²⁴See Schmeidl (1997, 1998, 2000), Davenport, Moore and Poe (2003), Moore and Shellman (2004a), and Neumayer (2005a).

²⁵See Shellman (2004a) for a review of the economics and political science literatures on temporal aggregation.