Illegal Immigration Outcomes on the U.S. Southern Border

By Bryan W. Roberts

The state of U.S. border security has been a contentious public issue for many decades. It has not been possible to establish a coherent definition of border security, let alone measure and evaluate it. In recent years, my research colleagues and I have led an effort to define what core border security measures are and to estimate their actual values. This article reviews the key results of that effort with respect to illegal immigration across the U.S. land border with Mexico, which has been perhaps the single most controversial border security issue. Although we now have a set of credible estimates that can inform policy making and public debate, the U.S. government continues to face major challenges in reporting these measures and establishing credibility with the American public. This article concludes by offering some constructive recommendations to the government on how it could do better.

What Do Americans Believe about Illegal Immigration and Border Security?

Although many experts and analysts believe that illegal immigration has fallen significantly in recent years, a majority of the American public does not. Table 1 summarizes the results of a 2015 poll that asked if illegal immigration has risen, fallen, or stayed the same over the past few years. Most people believed that it has increased, and this is true regardless of party or political affiliations as well as socio-demographic characteristics. Polls conducted over the last 15 years show that large majorities of Americans believe that U.S. borders are not “secure” and that the government could be doing more to increase border security with respect to illegal immigration.1

Given these widespread perceptions, it is not at all surprising that border security has remained an intense focal point of concern and controversy. This has been driven at least in part by a complete failure of the U.S. government to publish credible border security measures that the public trusts and tell them what they really want to know about the state of border security.

Border Security Measures

American taxpayers pay for border security in order to enforce U.S. laws at the border. With respect to illegal immigration, those laws require preventing entry of those without lawful permission to enter. Therefore, the core strategic measure of border enforcement success is the number of unauthorized migrants who escape detection and successfully enter the United States illegally.2 I will refer to this measure as the number of successful illegal entries.

Measures also should be reported on law enforcement outcomes that play important roles in determining how many migrants attempt illegal entry, and the degree to which they are deterred. One obvious measure is the probability of apprehension — the chance (on average) that someone attempting illegal entry is caught. Another measure is the rate of at-the-border deterrence — the chance that someone who has been caught and returned to their home country gives up and goes home rather than tries again to enter illegally. These variables affect a migrant’s initial decision to come to the border in the first place in order to attempt illegal entry. People who know there is a very high chance of being caught, and a very high chance of giving up after being caught, will be less likely to make an initial illegal entry attempt.

There is a strong tendency in the border security debate to describe the border as either “secure” or “insecure.” This is misguided, because the state of border security is never a black-or-white situation. If a border were perfectly sealed, then the number of successful illegal entries would be zero, and the probability of apprehension would equal 100 percent. If the border was perfectly open, then all entries would by definition be legal, and the probability of apprehension would be irrelevant, because there would be no border enforcement. Because these two states never actually apply in the real world, it makes more sense to estimate the values of the three core measures highlighted above and try to understand what they imply about the state of border security. These estimates are critically important for both informing public debate and helping government agencies understand how their policies and programs have affected outcomes and how they might need to be changed.

Estimates of Border Security Measures for the U.S. Southern Border

There are three basic approaches that have been used to try to estimate the above border security measures (see Whitley 2012):

Known Flow-Based Estimates. This approach measures successful illegal entries by observations of law enforcement agents. It relies on evidence obtained through cameras and other sensors as well as tracking detections by agents in the field (“sign-cutting”). The Border Patrol refers to this estimate as “gotaways.” It also estimates “turnbacks” — those who enter illegally but then decide to return to Mexico before being apprehended. “Known flow” data comprises gotaways, turnbacks, and apprehensions. Using this data, the Border Patrol calculates the interdiction effectiveness rate (IER), which is the ratio of apprehensions plus turnbacks to the sum of apprehensions, turnbacks, and gotaways. The IER is intended to function as a proxy for the probability of apprehension.

Survey Based Estimates. This approach is based on collecting data from migrants about their past illegal trips across the border. The probability of apprehension can be calculated from the number of reported apprehensions. The probability of at-border deterrence can be calculated from the number reporting they were ultimately unsuccessful in entering. The number of successful illegal entries can then be estimated by combining the probability of apprehension estimate with the number of apprehensions in a simple mathematical formula. Data from the Mexican Migration Project survey has been used to make these estimates since the early 1990s.3

Analytically Based Estimates. This approach is based upon the repeat trials model (RTM) of the illegal entry process. Under the RTM, a migrant attempts illegal entry, is either caught or not caught, is returned to his or her home country after being subjected to any consequence if caught, and decides after his or her return to either continue attempting illegal entry or give up. If the migrant decides to continue, another round of this process takes place. It can be shown mathematically that under this process, the probability of apprehension is a simple function of the probability of at-border deterrence and the recidivist ratio, which is the ratio of apprehensions after the first one to all apprehensions. The first application of the RTM to estimate the probability of apprehension and number of successful illegal entries was published in 1990, and the approach was applied again using better data in a 2006 study.4

Until recently, the U.S. government did not publicly report estimates for any key border security measure. In 2012, the U.S. Department of Homeland Security (DHS) permitted publication of known-flow data, and it currently publishes values for the IER in its annual performance report. Values for gotaways have never been officially published by DHS, although values for the period 2006-2013 have been released through other channels.5 Publishing measures that are based on known-flow data is an important first step. However, known-flow data are fundamentally flawed for estimating border security measures.

In January 2015, my colleague John Whitley of the Institute for Defense Analyses (IDA) and I were asked by DHS Secretary Jeh Johnson to carry out a research project to produce credible border security estimates. Over the course of 2015, our IDA research team developed estimates by using a migrant survey to calculate the probability of at-border deterrence and the RTM approach to estimate the probability of apprehension and number of successful illegal entries.6 We were given full access to all internal administrative DHS data related to immigration enforcement, which enabled us to develop estimates of the key measures for illegal entry between ports on the southern border, at ports on the southern border, and in the maritime domain.

In order to develop estimates for illegal entry between and at ports on the southern border, we implemented an analytical approach that begins with an important distinction between “traditional” migrants, who seek to evade U.S. law enforcement authorities when illegally entering the United States, and asylum seekers, who enter illegally but then present themselves to law enforcement in order to claim asylum. Although the number of asylum seekers was low prior to 2011, their numbers have risen dramatically since then. Our approach is implemented through the following steps:

• First, we separated apprehensions of asylum seekers, who are not returned to their home country in any set timeframe and are not part of the repeat-trials population, from apprehensions of traditional migrants, who are assumed to have been attempting to evade Border Patrol or the port enforcement authority — the Office of Field Operations (OFO);

• Second, we estimated the probability of at-border deterrence for Mexican nationals returned to Mexico using migrant survey data;

• Third, we estimated the probability of apprehension for traditional Mexican nationals using estimates of the probability of at-border deterrence and the recidivist ratio for this group;

• Fourth, we estimated the number of successful illegal entries of traditional Mexican nationals using the estimate of the probability of apprehension and the number of apprehensions;

• Fifth, we estimated the number of successful illegal entries of traditional non-Mexican nationals using the estimate of the probability of apprehension for traditional Mexican nationals. We thus assume that the probabilities of apprehension for Mexican nationals and non-Mexican nationals are not systematically different.

• Finally, we report the number of apprehensions of asylum seekers. It may be the case that some migrants who successfully evaded Border Patrol would have claimed asylum had they been caught. We did not try to estimate the number of these illegal entries since the number of successful illegal entries by this group is unlikely to be very large.

Our estimates related to illegal entry at ports are the first plausible estimates that have ever been made. Migrants attempt illegal entry at ports by hiding inside cars or trucks, using counterfeit documents, or using someone else’s legitimate document. We applied the RTM approach to at-port entry and found that it could provide plausible estimates.7 We were also able to develop estimates for the maritime domain, which involves illegal entry by sea from Haiti, the Dominican Republic, and Cuba.

Table 2 presents the key results of our research. From 2005 to 2015, successful illegal entries by traditional migrants have fallen by roughly 90 percent, from almost 2 million to 200,000. The large majority of these illegal entries took place between ports. There has been a dramatic change in the at-border deterrence rate. In 2005, only 11 percent of Mexican nationals caught and returned to Mexico gave up attempting illegal entry, but in 2015, 60 percent gave up. This rise has arguably been due primarily to the Border Patrol imposing consequences on those they apprehend. The between-port probability of apprehension rose from roughly 40 to 55 percent, a smaller increase that reflects the fact that Border Patrol faces an adaptive adversary — smuggling organizations that seek to prevent this probability from rising as border enforcement intensifies.

Table 2 also shows that the number of apprehensions of asylum seekers has risen dramatically since 2011, reaching 140,000 in 2015, which is close to the estimated number of successful illegal entries of traditional migrants. Clearly, there has been a major change in the composition of the migrant flows coming to the southern border, a change that has important implications for illegal immigration enforcement policies and programs.

>Our estimates are subject to various limitations that are documented in Institute for Defense Analyses (2016). Continued research and refinements will improve their accuracy, but they are the highest quality estimates that have been made for key border security outcomes to date.8 In particular, they much more accurately represent true illegal immigration outcomes on the southern border than the known-flow measures DHS has recently pursued. Gotaways systematically undercount the number of true successful illegal entries, because there will always be entries that are never observed. The IER is intended to represent the probability of apprehension, but it will always systematically overstate it, because the IER illegitimately includes apprehensions of people not trying to evade Border Patrol (e.g., asylum seekers) and turnbacks. Figure 1 shows the RTM-based estimates of between-port successful illegal entries and the probability of apprehension with known-flow gotaways and the IER. As would be expected, the number of gotaways is significantly below the number of RTM successful illegal entries, and the IER is significantly above the RTM probability of apprehension. The reasons for these differences are well understood, and known-flow-based measures should be avoided for purposes of border security reporting.

Why Has Illegal Migration across the Southern Border Changed?

Our estimates clearly establish that successful illegal entries on the southern border have fallen dramatically over the past decade. Why have they fallen? Illegal migration flows are influenced by many factors, including economic conditions in the destination country (e.g., the United States), economic conditions in the home country (e.g., Mexico), enforcement against illegal migration, the ease of migrating legally, demographic change in the home country, and other factors (e.g., criminal activity in Mexico.) According to analysis of data from the Mexican Migration Project migrant survey, prominent academic researchers have long argued that border enforcement does not deter initial illegal migrations from Mexico to the United States, but it does cause these migrants to not return to Mexico, so that intensified border enforcement ironically increases the resident illegal immigrant population (see Massey, Durand, and Pren 2016).

Over the past decade, several things have happened that would be expected to have significant impacts on the decision of a Mexican national to illegally migrate to the United States or not.

• First, the United States experienced the Great Recession and then a recovery.

• Second, the number of potential migrants from Mexico was falling due to demographic change.

• Third, the Mexican economy may have been improving.

• Fourth, the U.S. government carried out a border enforcement buildup that involved a large increase in the number of Border Patrol agents as well as deployment of infrastructure and technology.

This buildup included a major change in the consequences applied to Mexican nationals who are caught attempting illegal entry. Prior to the mid-2000s, almost all of those apprehended were subjected to “voluntary return,” which simply meant a short bus ride back to Mexico. After 2005, the U.S. government began to apply a range of consequences, including expedited removal and reinstatement of removal (which curtailed the ability to migrate legally in the future), misdemeanor and felony prosecutions for illegal entry, and consequences designed to disrupt the relationship of a migrant with their smuggler. The dramatic rise in the at-border deterrence rate suggests that intensifying consequences had a major impact on migrant decisionmaking.

My research colleagues and I have tried to identify the impact of each of these variables on the decision to illegally migrate from Mexico using data and estimation techniques that differ from previous research. We use quarterly data for the period 2002-15 from the nationally representative Mexican Encuesta Nacional de Occupación y Empleo (ENOE) household survey, which captures when an individual in a household migrates to a foreign country. We restrict our ENOE sample to a group that accounts for the large majority of illegal and legal migrations and to control for attrition bias.9

We develop region-specific economic and enforcement measures that permit exploiting geographic variation to identify the impacts of these variables on the decision to migrate of individuals observed in the ENOE survey. Economic explanatory variables include a U.S. unemployment rate variable and an expected Mexican wage variable that reflects the probability of being employed and wage earned if employed. U.S. enforcement explanatory variables include a variable that captures the intensity of border enforcement that a person in the ENOE survey would be likely to encounter if attempting illegal entry and the likelihood of being prosecuted for illegal entry if caught. These variables are based on the historical crossing locations that an individual from a given state in Mexico is likely to choose. The impacts of enforcement and consequences are thus identified by variations in the intensity of enforcement and consequences along the border. In addition to economic and enforcement variables, our regression analysis includes a variable that indicates the likelihood that a person observed in the ENOE survey would migrate illegally as opposed to legally, as well as a range of control variables.

Regression results are presented in Table 3. Border enforcement intensity significantly lowers the likelihood that someone with a high propensity to migrate illegally will choose to do so. Although misdemeanor prosecutions have no significant impact on the decision to migrate, felony prosecutions have a powerful one. A higher expected income in Mexico significantly lowers the odds of migration, and a higher U.S. unemployment rate also lowers it.10

We used the regression results in Table 3 to predict historical levels of aggregate illegal emigration from Mexico during 2002-15 of this particular group. Demographic change in Mexico is incorporated into these predictions because of the use of ENOE survey weights. We then simulated what would have happened if explanatory variables remained constant at their 2005 levels. We first held enforcement variables (agents and consequences) constant. We then additionally held the economic variables constant. Figure 2 shows the results of these simulations.

The historical-level values suggest there has been no rebound in illegal migration from Mexico since 2010. The downturn that began in 2007 was initially driven almost entirely by the U.S. Great Recession, and the enforcement buildup had limited impact. Starting in 2010, however, the U.S. economy began to recover, and the impact of the enforcement buildup began to grow. By 2015, the enforcement variables accounted for all of the difference between the historical level and the counterfactual simulations. This means that had the enforcement buildup not occurred, illegal migration would have increased to levels consistent with recovery in the U.S. labor market. Border enforcement has now broken the link between the U.S. business cycle and illegal Mexican immigration flows by impacting the decision of potential illegal Mexican migrants and deterring them from coming to the border.

Building Credible Security on the Southern Border

The U.S. government faces two basic challenges in building credibility on border security with respect to mass illegal immigration. First, it must carry out actions that improve border security. Second, it must credibly report on border security outcomes and determine whether its actions have impacted those outcomes. The Washington political establishment has been reluctant to do either since illegal entries started rising in the late 1960s. For many decades, an establishment consensus has held sway that treats border enforcement with respect to mass illegal immigration as a nuisance responsibility that must be carried out but has little or no impact on these flows, which are believed to be affected only by economic and demographic trends. As a result, there are more political and institutional rewards in Washington for focusing on facilitation of cross-­border travel and trade flows, which supports the economy and minimizes complaints from a traveling and trading public.

However, the public has insisted for decades that enforcement be taken with the same level of seriousness as facilitation. After the 1986 Immigration Reform and Control Act failed to curtail illegal immigration, the Clinton administration carried out an initial border enforcement buildup in the 1990s. Although there was some initial promising discussion of how to measure border security outcomes, this was quickly dropped.11 The Bush administration inaugurated a second major enforcement intensification after failed immigration reform efforts in 2006 and 2007. After the Obama administration took office in 2009, DHS Secretary Janet Napolitano did not roll back this intensification, but she chose to pursue a misguided approach to measuring border security outcomes that ended in failure. This put the government in a difficult situation when comprehensive immigration reform was again attempted in 2013. As could have been predicted from earlier immigration reform failures, border security was the subject of intense dispute and disagreement, and the lack of any credible evidence on border security outcomes inevitably undermined the possibility that the 2013 reform could succeed.

A new administration has come into office that has made achieving southern border security one of its core goals. The new administration and Congress will need to determine what constitutes desirable border security outcomes, and what policies and measures are needed to achieve them. I would offer the following recommendations:

The Washington political establishment needs to stop treating border enforcement against illegal immigration as a nuisance issue. Three failed immigration reform attempts over the course of a decade did not fundamentally change this attitude. It remains to be seen if the 2016 election will. The research results presented here also make it clear that enforcement does impact migrant decisions. It is no longer tenable to oppose border enforcement because it doesn’t work.

The White House must demand meaningful border security measurement. Unless the new White House requires the executive branch to report objective and rigorous measures, opposition from lower levels in the federal bureaucracy will likely prevent progress. Border security measurement needs the same status and protection from political pressures that are enjoyed by programs producing economic estimates such as inflation, income, and unemployment.

The Department of Homeland Security and its components must stop opposing objective measurement of meaningful border security outcomes. DHS has never reported meaningful border security measures. Instead it continues to report measures that either show that DHS components have achieved “mission success” or that convey little or no meaningful information about the state of border security. DHS currently reports a misleading proxy for the probability of apprehension between ports of entry, and more importantly, it has never officially reported any estimate of successful illegal entries — not even Border Patrol’s estimate of gotaways. DHS has never reported any measure related to illegal inflow and probability of apprehension for ports of entry. The new Secretary of DHS will face the choice of fundamentally transforming DHS’s culture so that it focuses and reports on meaningful enforcement outcomes, or giving in to constituencies that seek to distort or suppress quality estimates. If public perceptions of border security are to change, DHS’s culture must change, and transparency must prevail.

The U.S. government must make its administrative data and estimate methodologies available for external review and research. Producing credible estimates will not be enough to establish credibility if external researchers cannot replicate these estimates and evaluate the methodology used to generate them. We on the IDA research team know that the quality of our estimates is high — but why should anyone believe us? Belief comes through a scientific process of review and critique that is open to all researchers. Government agencies routinely share highly sensitive data with researchers in ways that protect the data from misuse. The sharing of DHS administrative data with external researchers is long overdue and should be a top priority of the new administration.


The challenges that will likely continue to impede progress are well illustrated by the fate of our research project in 2016. In his annual “State of Homeland Security” speech in February 2016, Secretary Johnson revealed publicly that he was supporting the research reviewed in this article: “We are working with outside, non-partisan experts on a project called BORDERSTAT, to develop a clear and comprehensive set of outcome metrics for measuring border security, apprehension rates, and inflow rates.”12 Later on in this address, the secretary stated: “In terms of border security, as I mention in my remarks, I want to get to a better place in how we measure border security. We have an initiative which I hope to finish before we leave office.”13

Although this project had been completed by November 2015 and could have been released to Congress and the broader public well before he left office, neither happened. The news media became aware of the project in September 2016, and several media outlets ran stories on it. Congress then sent a letter to DHS in October 2016 demanding that it be provided with the study, but DHS declined to respond. As of today, the study has still not been authorized for dissemination to either Congress or the public.

This lack of progress is made all the more puzzling by the fact that our results show a dramatic fall in successful illegal entries on the southern border and the major role of border enforcement in bringing that about and sustaining that decline. The government needs to take charge of this story, tell it honestly and properly, and start establishing greater credibility.

Two tests will take place this year that will indicate whether progress can be made. First, the National Defense Authorization Act passed in 2016 requires that DHS provide to Congress estimates of all of the measures discussed in this artilce.14 Second, the executive order on border security and immigration enforcement improvements that President Trump issued in January requires that DHS produce a comprehensive study of southern border security by July 2017.15 This study provides another opportunity for DHS to report proper measures on southern border security. One can only hope that the executive branch will pass these tests successfully.


Bipartisan Policy Center (2015) “Measuring the Metrics: Grading the Government on Immigration Enforcement.” Available at

Chang, J. (2006) “CBP Apprehensions at the Border.” Homeland Security Studies and Analysis Institute study. Not publicly disseminated.

CNN/ORC (2015) “International Poll” (July 22-25). Available at

Espenshade, T. (1990) “Undocumented Migration to the United States: Evidence from a Repeated Trials Model.” In F. D. Bean, B. Edmonston, and J. S. Passel (eds.) Undocumented Migration to the United States: IRCA and the Experience of the 1980s, 159-81. Santa Monica, Calif.: Rand Corporation.

Government Accountability Office (2012) “Border Patrol: Key Elements of New Strategic Plan Not Yet in Place to Inform Border Security Status and Resource Needs.” GAO-13-25. Available at

Institute for Defense Analyses (2016) “Assessing Southern Border Security.” IDA Paper NS P-5304. Not yet publicly disseminated, but available at

Massey, D.; Durand, J.; and Pren, K. (2016) “Why Border Enforcement Backfired.” American Journal of Sociology 121 (5): 1557-1600.

Roberts, B.; Alden, E.; and Whitley, J. (2013) Managing Illegal Immigration to the United States: How Effective Is Enforcement? Washington: Council on Foreign Relations.

Roberts, B. (2016). “Illegal Outcomes on the U.S. Southern Border: What Are They, and Why Have They Changed?” Presentation at the Cato Institute’s conference on “Immigration Economics” (September 2).

Whitley, J. (2012) “Five Methods for Measuring Unobserved Events: A Case Study of Federal Law Enforcement.” IBM Center for the Business of Government, Improving Performance Series. Available at


1 Qunnipiac University polling results are available at

2 For more extensive discussions of what constitutes a good set of border security measures, see Roberts, Alden, and Whitley (2013), Bipartisan Policy Center (2015), and Institute for Defense Analyses (2016).

3 See Massey, Durand, and Pren (2016) for the most up-to-date estimates based on MMP data.

4 See Espenshade (1990) for the first application of RTM. Border Patrol began collecting fingerprints from each person apprehended in 2000, which has enabled a much more precise calculation of the recidivist ratio. Chang (2006) used this data to estimate the probability of apprehension and successful illegal entries during 2001-05. This study has never been publicly disseminated by the U.S. government. It is important to note that both Espenshade and Chang assumed that the probability of at-border deterrence equaled zero.

5 Known-flow data for 2006-11 were published in Government Accountability Office (2012), and data for 2012-13 were released to the public through FOIA requests.

6 Unlike previous RTM studies, we do not assume the probability of at-border deterrence to be equal to zero. We use data from the EMIF migrant survey implemented at the border by the Mexican research institute COLEF to estimate the probability of at-border deterrence, and we validate the EMIF-based estimates though econometric analysis of apprehension records. See Institute for Defense Analyses (2016) for in-depth discussion.

7 The port enforcement authority, CBP’s Office of Field Operations, runs the Compliance Examination (COMPEX) program, whose data should permit estimation of successful illegal entries through ports of entry. This program’s data suggest that in recent years, the number of such entries is zero, which is not plausible. Although we had been hoping to use COMPEX data to develop at-port estimates, we realized that these data are useless for that purpose, and we turned to the RTM approach, which proved successful at yielding more plausible estimates.

8 The report was reviewed by several external academic researchers, who found no significant material weaknesses in our methodology.

9 We include only Mexican males between 16 and 50 years of age with less than a post-secondary education. Attrition bias is present because the ENOE survey follows a given household cohort for five quarters: we use data for the second quarter only of each household cohort to control for this bias.

10 Although the latter coefficient is not significant at the 10 percent level.

11 See Bipartisan Policy Center (2015) for more extensive discussion of the history of border security measures.

12 The speech is available on the DHS website at

13 See discussion at 1:00:49 in video posted here:

14 See Subtitle G, section 1092, pp.S.2943 — 430-37. Available at

15 See Section 4(d) of the executive order. Available at

Bryan W. Roberts is a Research Economist formerly employed by the Department of Homeland Security. The analysis and opinions provided in this article are the author’s alone and do not necessarily reflect those of the Department of Homeland Security or any other organization or individual. The author would like to thank Jeh Johnson, Gregory Pejic, and others at DHS who advanced Secretary Johnson’s Unity of Effort initiative; Michael Chertoff, Michael Fisher, John Whitley and the research team members at the Institute of Defense Analyses; Edward Alden, Theresa Brown, Gordon Hanson, Scott Borger, and the many scholars and analysts who have worked on analysis of border and immigration issues over many years.