When the Roberts Court agreed to take up a case challenging the application of disparate impact theory—as in, the idea that a policy’s impact can be discriminatory even if its intentions were not discriminatory—to one of the most successful pieces of civil rights legislation, the 1968 Fair Housing Act, civil rights advocates were concerned. We were concerned the Roberts Court would hamstring the Fair Housing Act, just as it did two years earlier to another cornerstone piece of civil rights legislation, the Voting Rights Act. Much to advocates’ surprise, including mine, that did not happen. And for that, we can thank Justice Anthony Kennedy.

Conservative legal advocates have had their sights set on gutting a key component of anti-discrimination law since 1971 when, in Griggs v. Duke Power Co., the Supreme Court first ruled plaintiffs don’t always need proof of discriminatory intent to win their civil rights case: Sometimes it is enough if the impact of a particular policy or decision is discriminatory. This idea, disparate impact theory, has become an important tool in getting at the more insidious and less immediately apparent forms of discrimination. But it is a tool that has been used inconsistently in civil rights litigation, in part because of another Supreme Court decision that said proof of intentional discrimination was required under the 14th Amendment. Every federal appeals court to consider the issue of whether or not the Fair Housing Act provides for disparate impact claims has ruled in favor of disparate impact theory thus far. Still, civil rights advocates were concerned the FHA would be yet another casualty in conservatives’ continuous erosion of civil rights jurisprudence.

Justice Anthony Kennedy’s role in moving the Supreme Court to a likely decision in support of marriage equality is well-documented. But on Thursday, it was his role in shaping the court’s jurisprudence on race that was apparent. Texas Department of Housing and Community Affairs (TDHCA) v. The Inclusive Communities Project involved a challenge to disparate impact theory and its specific application under the FHA. Inclusive Communities Project, a Dallas-area nonprofit that promotes racially and economically diverse communities, had discovered that between 1995 and 2009, the TDHCA had been allocating almost all affordable-housing tax credits to developments in poorer neighborhoods of color, while denying credits to those in wealthier white neighborhoods. This allocation, ICP argued, had the effect of preserving racial segregation by preventing low-income residents, largely people of color, from moving into white communities.

ICP sued, claiming the result of TDHCA’s policy was racially discriminatory, even if the housing agency’s intent in doling out the subsidies was not. To support its claim, ICP presented statistical evidence of the disparate impact of the housing agency’s policy. TDHCA countered that its policy was not discriminatory and that FHA did not recognize the kind of disparate impact claims made by ICP.

It was a bold claim to make, especially since every federal appeals court to consider the issue had ruled that yes, the FHA does in fact recognize disparate impact claims. But much like the cases against Obamacare, opponents made the claim with the hopes they could get the Roberts Court to bite and take the case. Which it did.

One reason conservative legal scholars pushed so hard to get a disparate impact case before the Roberts Court is many believed they had a conservative majority on their side, especially under the leadership of the chief justice. As Ian Millhiser notes at ThinkProgress, Roberts spent his time in the Reagan administration honing his social conservative bona fides on the very issue of undermining disparate impact claims. And on Thursday Roberts, predictably, joined in the opinion that would have dramatically weakened the FHA by eliminating the ability of plaintiffs to bring disparate impact claims. But Kennedy did not. And it’s that split—one that goes back as far as 2007 between the two conservative jurists—that made the difference in this week’s case.

In 2007 the Roberts Court took up a key school desegregation case, Parents Involved in Community Schools v. Seattle School District No. 1. The case involved the desegregation plans of Seattle, Washington, and Louisville, Kentucky, which both voluntarily used individualized racial classifications of students to achieve diversity in their districts, meaning the districts were specific in determining which students of which race attended in order to achieve what the districts argued was a compelling interest in a diverse student body. The Court ultimately ruled against the desegregation plans in 4-1-4 decision, with Kennedy acting as the key swing vote. He sided with the conservatives in ruling that the plans did not meet constitutional muster because they resulted in determinations of who would attend made expressly on the basis of race. But he also agreed with the liberals that ending racial discrimination is a compelling state interest, and that it is not as easy to “end discrimination on the basis of race by ending discrimination on the basis of race,” as the chief justice said in his Parents Involved decision.

Roberts’ quip about ending discrimination by ending discrimination at the close of Parents Involved would come to define the chief justice’s role in shaping the Court’s race jurisprudence as much as Kennedy’s split in Parents Involved between its liberal and conservative coalitions would. Parents Involved demonstrated the “post-racial” ideology the chief justice would articulate later in Shelby County v. Holder in his majority opinion dismantling the Voting Rights Act. But it also showed that on the issue of remedying the lasting effects of racial discrimination in this country, much work remained.

Since Parents Involved, Kennedy has authored two other important decisions on the issue of racial discrimination: Fischer v. University of Texas and Schuette v. Coalition to Defend Affirmative Action. Both cases dealt with the use of race in college admissions policies, and both demonstrate Kennedy grappling with the question of race and how, if at all, the courts can address the ongoing problem of race discrimination in all its manifestations. Neither Fischer nor Schuette can be considered wins for progressives—after all, both decisions made it significantly more difficult to consider race in college admissions. But they are good examples, like Thursday’s FHA decision, on when Kennedy acted like a levee to stop Roberts’ post-racial floodwaters from completely overwhelming civil rights jurisprudence. Like Parents United and ICP, both fundamentally affirmed the truth that discrimination in 2015 in many ways looks different than discrimination in 1965, and that the law is going to have to grow up as a result.

“Recognition of disparate impact liability under the FHA also plays a role in uncovering discriminatory intent: It permits plaintiffs to counteract unconscious prejudices and disguised animus that escape easy classification as disparate treatment,” Kennedy wrote in ICP.

That one sentence, I think, explains the difference between Kennedy and his conservative allies on the Court and why Kennedy ultimately sided with the liberals here. Perhaps it comes from his evolution on the issue of LGBTQ rights, but on Thursday, Kennedy made it clear that he believes at least in some degree in “unconscious prejudice and disguised animus” while his conservative colleagues do not.

But it is also clear that Kennedy is not completely comfortable with the idea of decision makers explicitly using race in their process of trying to remediate the effects of discrimination, as evidenced in both his affirmative vote in Shelby County v. Holder to gut the Voting Rights Act and his opinion in ICP. In his majority decision in ICP, Kennedy cautions that disparate impact theory also poses “special dangers” that lead individuals and the courts to improperly consider race, which results in such things as racial quotas. Kennedy’s opinion takes care to emphasize that only policies that are “artificial, arbitrary, and unnecessary barriers” to racial equality will violate disparate impact analysis, and to warn future plaintiffs that a disparate impact claim relying on a statistical disparity like the one in ICP will fail if the plaintiff cannot point to a specific policy causing that disparity. In other words, if a plaintiff sees racial imbalance in something like the makeup of the tenants of a particular apartment building but can’t identify a single cause, their claim under the FHA will likely fail.

So just how big a win is Thursday’s Fair Housing Act decision? Well, thanks to disparate impact claims under the FHA, the U.S. Department of Justice secured millions in damages from banks that charged higher rates to Black and Latino borrowers with similar credit histories as white borrowers. And disparate impact theory has been an important weapon in the fight against discriminatory renting practices as well as providing key protections for domestic abuse survivors. By confirming disparate impact theory as a legitimate claim under the Fair Housing Act, the Roberts Court also protected other civil rights statutes from similar attacks while shoring up disparate impact analysis as an important tool in fighting housing discrimination moving forward.

“If you look at the cases involving lending abuses coming out of the foreclosure crisis you see a very large number of cases involving what looks like pretty clear racial discrimination in setting the terms of loans, but race discrimination is very difficult to prove. If it weren’t for disparate impact, the Department of Justice wouldn’t have been able to get those very large settlements,” said Samuel Bagenstos, professor of law at the University of Michigan Law School and former Department of Justice attorney. Bagenstos was part of a group of Department of Justice housing attorneys who submitted an amicus brief in support of disparate impact theory under the FHA.

But Bagenstos noted that while the case was definitely a win for civil rights advocates, it did not represent a sweeping expansion of civil rights jurisprudence. “The Court was clear that a disparate impact claim can’t be made out on a statistical disparity alone and that it can’t be used to adopt a kind of quota system.”

Nor does Thursday’s win necessarily represent the end of challenges to disparate impact theory. “There’s this reminder from Justice Kennedy, not surprisingly, that courts should be careful to avoid requiring things that would themselves violate the 14th Amendment,” said Bagenstos.

In other words, the question of just how much race can factor into a decision or policy without running afoul of the 14th Amendment, which expressly prohibits different treatment on account of race, remains unanswered for now.

I’m usually pretty hard on Justice Kennedy because, frankly, his decisions in the reproductive rights area are abysmal. And even when his opinions are good, like in U.S. v. Windsor with the defeat of the Defense of Marriage Act, his reasoning is so muddled it’s hard to find the kind of forceful prose attorneys and journalists like to quote in articles and legal briefs. But on ICP Kennedy not only got it right, he did so by further distancing himself from a conservative ideology that insists it is colorblind while advancing policies that serve only to reinforce the lasting effects of racial discrimination in this country. For that, Justice Kennedy deserves both praise and credit from progressives.