III. Methodology

When Human Rights Watch commenced research for our 2007 report "Forced Apart," we sent a Freedom of Information Act (FOIA) request together with Boston College Law School on March 15, 2006 to Immigration and Customs Enforcement to answer basic questions about the legal status of those deported for crimes (for example, how many were green card holders, how many had other immigration statuses in the US, and how many were undocumented), the nature and seriousness of the criminal convictions forming the basis for deportations (for example, how many convicted of shoplifting, how many of homicide), and the family relationships of those deported (for example, how many had US citizen or lawfully present spouses and children).

Human Rights Watch delayed publication of "Forced Apart" for one year while we waited to receive a response to our FOIA request.Unfortunately, that response did not come in time, and we had to publish our initial findings without the requested data. The history of ICE's non-responsiveness to our repeated requests (which can be viewed in the appendix to "Forced Apart")[14] suggests at best a lack of commitment to transparency and the goals of the FOIA legislation; at worst it suggests deliberate stonewalling.

After two-and-a-half years of administrative wrangling, including an assertion by the agency that providing a response to our request would cause statistical reporting by the agency to "virtually grind to a halt,"[15] and with the assistance of pro bono counsel, we ultimately amended our request and finally received a response on August 13, 2008.[16]

Upon receipt of the data, we began analysis. The data were first imported into statistics software, Statistical Package for the Social Sciences (SPSS) version 14.0, which was used for all statistical analysis. Data were then organized and cleaned, removing empty variable sets resulting from redactions by ICE (these redacted variables included individual identifiers such as names and identification numbers). Included in the dataset are the individual's nationality; the country to which he or she was deported; the date of deportation; the individual's immigration status; and crime codes indicating up to five crimes that the individual was arrested for or convicted of (we do not know conclusively from these data whether every non-citizen was convicted of the crime listed by ICE), and which formed the basis for the deportation of the individual from the United States. Unfortunately, despite our request for additional information, the dataset did not include information about the marital status or next of kin of the deported individual. The agency claimed that with regard to these data, "ICE does not track this information and therefore, has no records responsive to this portion of your request."[17]

We grouped and coded crime data based on descriptions and categories from the National Crime Information Center (NCIC) codebook. We assigned each crime to one of six categories by cross-referencing with the NCIC codebook and ranked them from most to least serious: "offenses involving violence against persons," "non-violent general offenses with potential to cause harm," "non-violent drug offenses," "non-violent general offenses," "non-violent immigration offenses," and "non-violent theft offenses." When individuals in the dataset were deported for more than one crime, the deportee's most serious crime was used for analysis.

The immigration status of each individual in the dataset was decoded using an immigration code database provided by ICE. We then grouped each of the immigration statuses into three general categories: "illegally present," "legally present" (of which there are four sub-categories), and "unknown."

Variables were created to allow for grouping of individuals. Cases were grouped based on the types of crimes committed, the nature of these crimes (violent versus non-violent), and the individual's immigration status. The main statistical analysis was conducted by running basic descriptive statistics, namely frequencies. Cross-tabulations were also conducted to compare groups. Examples of these cross-tabulations include examining the types of crimes for which individuals holding specific immigration statuses were deported. In specific cases illustrating data deficiencies, regression analysis was used to compare groups of deportees.