Everyone knew that separating children was about deterrence, using them as pawns in a grownups’ game to send a message that those seeking to migrate to the United States should stay out. White House Chief of Staff John Kelly said family separation “could be a tough deterrent,” and then corrected himself. “Would be a tough deterrent,” he added. Some officials had tried to avoid admitting this effect of the new policy was its actual purpose, but the Justice Department made clear that this is exactly what the policy is about in the main brief it submitted to a federal court on Thursday.

Seeking relief from a consent decree limiting the detention of children and families, the Acting Assistant Attorney General laid out the administration’s justification explicitly: “this Court’s construction of the Flores Settlement Agreement eliminates the practical availability of family detention across the nation, thus creating a powerful incentive for aliens to enter this country with children in violation of our criminal and immigration laws and without a valid claim to be admitted to the United States.” This was not just an aside, but rather a main part of the administration’s argument. The government presented some statistics and an affidavit of an ICE official to buttress these claims.

The use of immigrant detention to deter is neither new nor limited to one political party. It dates back at least to the Carter administration’s treatment of Haitian migrants. It was inherent in Caribbean interdiction policies inaugurated by Ronald Reagan and continued by subsequent administrations, both Democratic and Republican. And at times the goal of deterrence was even made explicit, for example, by the George W. Bush administration in Attorney General Ashcroft’s policy of detaining Haitian immigrants who made it onshore. Detention must be mandatory, Ashcroft wrote, because “there is a substantial prospect that the release of such aliens into the United States would come to the attention of others in Haiti and encourage future surges in illegal migration by sea.”

It is important to separate three issues: (1) whether it is moral to use a particular penalty to deter; (2) whether, as an empirical matter, it is effective to use a particular penalty to try to deter; and (3) whether it is legal to use a particular penalty to deter. We address only the first two here. But we should note that, just a few years ago, a federal judge held that the Obama administration’s decision to detain migrant families in order to deter others raised grave due process concerns and violated the Immigration and Nationality Act.

Ethical redlines

There are some “ethical redlines” when it comes to deterrence policies in general. Most Americans support the state’s killing people to deter. That’s capital punishment. But even those who argue that the death penalty is morally required reject the idea that it would be permissible for the government to knowingly execute innocent people in order to deter other homicides. Or take a gruesome example: state-sponsored acts of sexual violence could deter crimes, but that’s not even a utterable proposition because it breaks all manner of moral standards. In short, government strategies that would maximize deterrence are off limits when they run afoul of ethical redlines.

The Trump immigration policy violates ethical redlines in obvious and less-than-obvious ways. The policy transgresses one ethical redline by deliberately inflicting mental suffering on children to deter their parents from migrating to the United States. That’s an obvious ethical flaw in the administration’s efforts in this policy space.

What’s not so obvious is that if the Trump policy actually worked to deter migration, the policy would end up punishing the most legitimate asylum seekers, the people fleeing the most horror. That’s because they would be the ones least likely to be deterred by detention, criminal prosecution, or family separation, and thus the ones most likely to still come to the United States and end up with the pain and suffering of Trump’s gruesome detention policy. The policy would be using them–the most innocent and most deserving–to deter others.

Political science professor Jonathan Hiskey explained this effect of using immigration detention as a deterrent in a court declaration in 2015. He stated, “Detention of those asylum seeking families who still continue to migrate across Mexico and into the United States despite this knowledge of increased risk would then represent a choice to detain those individuals who likely have the strongest reasons to flee their homes because of violence and persecution.”

Empirics of Deterrence

It is important to take a hard look at the theoretical and empirical evidence when it comes to claims about deterrence. Existing social science research provides important theoretical reasons why we should be skeptical of claims that immigration detention significantly deters prospective migrants. Among those reasons are the strong “push factors” that propel migrants to travel North–particularly families in recent years. As social science professor Cecilia Menjivar explained in a declaration in 2015: “it is my conclusion that any perception of lax border enforcement or detention policies does not meaningfully contribute to the migration of families from Central America to the United States. Rather, the primary reason individuals and family units migrate from Central America is because of the conditions in their home countries. They are leaving countries that currently occupy the first, fourth and fifth place—Honduras, El Salvador and Guatemala, respectively—worldwide in homicide rates.” Many of the migrants who decide to make the journey know that they will risk much worse consequences than detention. “Compared to the others[sic] expected risks—such as rape or death—detention is actually less serious and thus less likely to function as a significant deterrent,” professor Menjivar stated.

Similarly, political science professor Nestor Rodriguez stated in a declaration in 2014: “My decades of research on Central American migration further allow me to conclude that rumors regarding lenient immigration detention policies in the United States are not a significant factor motivating current Central American immigration to the United States. A very small percentage of Central Americans may be prompted to undertake migration as a result of information suggesting that they will be released from detention promptly after arrival in the United States. However, this small minority of migrants would be made up of individuals who are predisposed to migrate without authorization under any circumstances. These individuals would very likely undertake the journey at some point regardless of any policies on immigration detention in the United States.”

Theory aside, in Thursday’s court filings the government made a very specific empirical claim about deterrence and family detention. The government argued that the July 2015 court decision applying Flores to family detention (it previously had applied only to unaccompanied minors) spurred many more families to come to the United States by restricting the use of family detention. Seeking relief from the Flores agreement, Justice Department lawyers argued: “Following the July 2015 ruling, there was a 3 to 5-fold increase in the number of illegal family border crossings. This surge is not a mere coincidence, it is the direct result of the message sent to those seeking illegal entry: we will not detain and deport you.”

Such bold and categorical claims about causation are like fingers on a chalkboard to social scientists. “It is extremely difficult to establish a causal effect between detention and deterrence . . . especially [for] an outcome as complex as migratory patterns,” explained professor Menjivar. This is only more true when the causal effect is claimed not on the basis of a carefully designed study, but instead on the basis of simple correlational statistics relating to a government policy change.

But there’s no need to delve into debates about causal identification strategies in order to see that the government’s claimed “evidence” just does not stack up. The government’s correlational statistics fail on their own terms. Here is how the government describes the claimed relationship between the Flores decision and migration patterns:

After a significant reduction in family units crossing the border in FY 2015 when the Government was holding families together, see ECF 184-1 at 8 ¶ 17, family crossings away from legal ports of entry nearly doubled in FY 2016, as measured by apprehensions. . . . The month-to-month figures show the sharp rise in family border crossings during 2015—from a figure in the range of 1,600 to 4,000 before this Court’s July 2015 decision, to a figure ranging from 5,000 to nearly 9,000 in the months after the decision.

At first blush, these numbers do sound like there is at least a correlation between Flores and families’ migration decisions. But the apparent relationship is based on the selective use of only a small slice of apprehensions data. The chart below provides a more complete view of monthly family apprehensions over the last several years. The chart is based on the very same Customs and Border Patrol data that the government used in its Thursday court filing. On the chart, the July 2015 Flores decision–the decision the government claims caused a huge surge in family migrations–is represented by the vertical dashed line.

As is plain from the plot, border apprehensions began rising months before the decision. The decision was simply not an inflection point. Forget causation: there’s not even a correlational relationship between Flores and family migration, as the government asserted in its brief.

Moreover, if the 2015 Flores decision really had changed the incentives for families, you would expect crossings for families and unaccompanied minors to respond differently after the decision. After all, unaccompanied minors had always been covered by the Flores settlement. The fight in 2015 was over whether the settlement also covered families. If the government wants to claim that the court’s decision applying Flores to families was the “treatment” that caused a migration surge by families, then unaccompanied minors are the “control group,” unaffected by the court decision.

So do we see sharply different responses by these two different groups? Not whatsoever. In the second chart below, we overlay apprehensions of unaccompanied minors on the apprehensions data for families. The result is striking. The pattern of apprehensions for these two groups track each other almost perfectly over time. This is devastating evidence against the government’s contention that rising rates of family apprehensions in the second half of 2015 were caused by the court’s July decision in Flores.

Beyond these statistics, the only other evidence the government submitted in support of its empirical deterrence claim was an “expert” declaration written in 2015 by Tae Johnson, then-assistant director of ICE’s Custody Management Division. But that declaration contains no real evidence either: it is based solely on citations to a few news articles and some second-hand information he heard from ICE agents who talked with some unidentified migrants they had caught. That’s it.

What’s more, Mr. Johnson’s declaration has already been discredited. It is the same declaration the Obama administration relied upon in 2015 when it tried to get around the Flores settlement. An expert who reviewed the declaration concluded that it “offer[s] no valid or reliable empirical justification for the detention practices,” and another expert also concluded the declaration’s “opinions regarding detention and its deterrent effects unreliable and invalid.” The federal court adjudicating these expert opinions decided against the government.

The Trump administration is trying to go down the road of using a combination of zero tolerance criminal prosecutions, family separation, and indefinite family detention to deter migration into the United States. Given the extreme ethical issues the administration’s approach raises, there is a heavy burden on the government to prove that these forms of deterrence even work. The government’s selective and flawed presentation of empirical evidence so far undermines its credibility. But even if the government did not have the burden, an objective analysis of the available evidence shows that the government’s empirical claims do not withstand scrutiny. All this effort on the part of the administration to strike terror into the minds of migrant families over the past several weeks, and zero evidence that it will even reach the actual goal of deterrence. That mistaken reasoning is often what happens when one crosses ethical redlines in making policy.