Reverse Discrimination Case

[jillio]

I know. Jillio says that the test is invalid already based on the simple fact that race was not evenly distributed. In theory, a black person is not different from a white person, so the results should have the same demographics as the test takers. About 25% of the test takers were black, so 25% of the promotions should have gone to black people. She says that there should be no reason why it shouldn't be 25% of the black people getting the promotion, therefore the test is invalid. I'm saying that the lack of a higher quality education due to parents' low income could be a reason.

(BTW, I did the math.... in order to represent the sample of the test takers, only 4 black people had to be promoted.... And if you take account the standard deviation, 0 is feasible...)

Thank you!!! Someone who finally gets it! Except that the standard deviation would not enter into a zero promotion for one segment of the population because the SD would be the same for all races in a valid test.

 

[jillio]

Quality not quantity.

Again, thank you. Someone who has a grasp of the concepts is sooooo refreshing!

 

[jillio]

An even distribution of what? Scores? Nothing was said about the distribution of scores on the exam. The only "distribution" that was considered was race. But please, explain to me again how I'm the one that insists on focusing on race as the main issue. If you really can't think of any other factors that might affect someone's score other than race, I would appreciate if you stop telling me that I'm the one focused on it. I'm on the side of people who are advocating not judging them based on race.



It really doesn't. Just because you make a statement doesn't make it true. I have backed up my statements with quotes and evidence from the situation. You back them up with "clearly you don't know what you're talking about". If you'd like to explain how the fact that over twice as many non-black firefighters also weren't eligible for promotion supports bias, I'd love to hear it. But an actual explanation requires more than just you "saying it's true".



The "disparities" only included racial distinctions. They did not suspect flaws, they suspected a lawsuit from the black firefighters. Again, please actually read the articles before trying to make a worthwhile statement.

The bolded section of your post is also very clearly and easily shown to be simply false. Read the articles first. It would help your case enormously.



Strangely enough, I rest my case based on your own quote. Your "knowledge" gives you no advantage over anyone else in this thread unless, as I have said before, you know this test, how it was evaluated, administered, interpreted, and anything specifically about firefighting. You do not have training in the area that is being discussed, please stop claiming a superficial superiority over anyone else in this thread. All it does is make you appear vain, childish, and stubborn.

not knowing what a normal distribution is, nor how it functions in regard to test scores is simply evidence that you are attempting to argue points for which you absolutely no understanding.:roll: Anytime there is such disparity in scores, the distribution is not normal. That is a given.

My knowledge obviously gives me a distinct advantage over you when it comes to the topic of reliability, validity, testing assessments, and normal distributions. I not only have the knowledge, I have the practical experience with such. You do not. The only vanity being exhibited is your belief that you can successfully argue regarding a concept you know nothing about.

 

[jillio]

Well clearly your math must be wrong. That can be argued by all the room for interpretation there is in all of this. :roll:



Also, to argue that since 25% are black that 25% of the promotions should be black people just lumps everyone together based on one thing: race. And I don't think that judging people based solely on their race is right or fair. But that must be because I can't see past the race issue....

Again, you just don't get it. What Daredevel is explaining is exactly that which takes away the consideration of race.

 

[Jiro]

not knowing what a normal distribution is, nor how it functions in regard to test scores is simply evidence that you are attempting to argue points for which you absolutely no understanding.:roll: Anytime there is such disparity in scores, the distribution is not normal. That is a given.

My knowledge obviously gives me a distinct advantage over you when it comes to the topic of reliability, validity, testing assessments, and normal distributions. I not only have the knowledge, I have the practical experience with such. You do not. The only vanity being exhibited is your belief that you can successfully argue regarding a concept you know nothing about.

so I see that you're asserting your "qualification." Since you think the test is faulty... so exactly what questions make it racial (non-cultural friendly)? Any example?

 

[Reba]

I truly do not see why it is so difficult to understand that if a test has a disparate impact on minorities, it is not a valid assessment for that population. It is a well known fact, and is not difficult to understand at all.

Not necessarily true.

A test can be an excellent assessment of a person's ability for a position, and still have a negative impact on those who don't get high grades, regardless of race.

Did that "population" get lower grades on the test just because they were from a minority race? Has anyone actually researched and proven that? Did each black test taker get the same questions wrong? Did the white test takers get all those same questions right? Did anyone get any feedback as to the study habits of all test takers? Did any of the black test takers themselves complain that the questions were biased?

Yes, there can sometimes be cultural bias in test questions but no one has proven that was situation in the New Haven case.

 

[jillio]

so I see that you're asserting your "qualification." Since you think the test is faulty... so exactly what questions make it racial (non-cultural friendly)? Any example?

The way the scores are distributed indicate that there is a problem with several types of validity, Jiro. I do not have the test in front of me, and therefore cannot give you specific questions. You already know that. However, I do know that when scores are distributed in the way that these test scores were distributed, the disparity indicates that there is a problem with cross cultural validity. I doubt that cross cultural is the only area of validity that is affected, however, given what the Board for Industrial and Organizational Psychology has stated.

 

[jillio]

Not necessarily true.

A test can be an excellent assessment of a person's ability for a position, and still have a negative impact on those who don't get high grades, regardless of race.

That is not the same thing being discussed here. When race is not a factor in the lower scores, then the test is most likely valid for the whole population who took it. In this case, the results point to a different conclusion.

Did that "population" get lower grades on the test just because they were from a minority race? Has anyone actually researched and proven that? Did each black test taker get the same questions wrong? Did the white test takers get all those same questions right? Did anyone get any feedback as to the study habits of all test takers? Did any of the black test takers themselves complain that the questions were biased?

Again, it has been determined that there was questionable validity on the test that created disparity organized by race in the results. The questions you have asked are relatively unimportant when deciding validity of an instrument. Validity is checked against a population. A test that consistently produces lower or higher scores for a particular population shows concerns about the validity of that instrument when used for that population. For instance, if I gave an assessment on which people who were white, older, ASL terps consistently scored in the low ranges, I would question the validity of that assessment for white, older, ASL terps.

Yes, there can sometimes be cultural bias in test questions but no one has proven that was situation in the New Haven case.

Evidently, the information provided by the Board of Industrial and Organizational Psychology has convinced the courts that it is more than likely.

 

[lsfoster]

not knowing what a normal distribution is, nor how it functions in regard to test scores is simply evidence that you are attempting to argue points for which you absolutely no understanding.:roll: Anytime there is such disparity in scores, the distribution is not normal. That is a given.

My knowledge obviously gives me a distinct advantage over you when it comes to the topic of reliability, validity, testing assessments, and normal distributions. I not only have the knowledge, I have the practical experience with such. You do not. The only vanity being exhibited is your belief that you can successfully argue regarding a concept you know nothing about.

Since again you have resorted to arguments that are no better than "You don't know what you're talking about because I say you don't, and I do because I say I do", I will ask you again to please stop assuming things about anyone in this thread. You have no idea what experience or knowledge I have about any of these topics.

You also have not explained or responded to any of the points I brought up. You have shown no evidence of what the actual distribution of scores was in relation to any factor but race, nor have you explained how the distribution based on race rules out any other possibility other than the test being invalid.

The only thing I have seen your "knowledge" do is lead you to stubbornly and childishly ignore any other points made which you have no response to, and to grossly misunderstand the actual issue being discussed in an attempt to somehow throw in everyone's face how much you "know". If this is the best you can do to respond to any of the multiple comments which have shown that you are wrong, making illogical assumptions, and interpreting events incorrectly, then I'd rather just leave it at the comments I've already made, and maybe when you feel like you can actually respond to or argue any of them, this discussion can move forward.

 

[jillio]

Since again you have resorted to arguments that are no better than "You don't know what you're talking about because I say you don't, and I do because I say I do", I will ask you again to please stop assuming things about anyone in this thread. You have no idea what experience or knowledge I have about any of these topics.

I have not had to assume. Your posts have indicated such.

You also have not explained or responded to any of the points I brought up. You have shown no evidence of what the actual distribution of scores was in relation to any factor but race, nor have you explained how the distribution based on race rules out any other possibility other than the test being invalid.

That entire paragraph shows that you have no idea what Daredevel and I are talking about when we discuss distributions. That is not an assumption. That is something indicated through your post. And the distribution itself indicates where the likely problem lies.

The only thing I have seen your "knowledge" do is lead you to stubbornly and childishly ignore any other points made which you have no response to, and to grossly misunderstand the actual issue being discussed in an attempt to somehow throw in everyone's face how much you "know". If this is the best you can do to respond to any of the multiple comments which have shown that you are wrong, making illogical assumptions, and interpreting events incorrectly, then I'd rather just leave it at the comments I've already made, and maybe when you feel like you can actually respond to or argue any of them, this discussion can move forward.

I am not ignoring other possibilities. I am saying the the other possibilities that you have brought up are not feasable. You are "stubbornly and childishly" arguing a point that you do not understand. There are many things you can argue regarding this story...however validity, reliability, statistical analysis, and assessment don't appear to in your repetoire. I am not wrong, nor am I making illogical assumptions. Nothing you have stated indicates that at all. It only indicates that you do not understand the basics of the concepts being explained.

I have responded to any and all other possibilities that you have presented. I have thoroughly explained why your assumptions and your train of thought is off. It has absolutely nothing to do with my attempting to prove how much I know. It is simply to help you learn that which you do not know. However, you refuse to do that for nothing more than sake of argument. What I have presented is well known and well accepted fact in the area of assessment and testing. You have simply refused to accept it. But that's okay. No skin off my nose at all.

 

[Daredevel7]

If you have 100 people, 30 of them are black, 20 of them are hispanic, and the rest is white. And you want to pick 10 of the best, assuming race isn't a factor, you should get an average of 5 whites, 3 blacks, and 2 hispanics. Give or take a few. If you do this every year, you will NOT get those exact numbers. That's where the standard deviation comes in. Sometimes you won't get hispanics at all (hence standard deviation CAN cause zero promotions for a specific race), sometimes there will be more blacks than whites, and so on.

I didn't see in the article if this is a consistent result (no blacks being promoted every year). There ARE some black people with high positions (not sure if they were hired using this very test).

Personally, I think the low income upbringing of minorities is the biggest factor in this disparity.

Interesting note: Out of the 77 test takers, 19 were blacks. That's 25% of the test takers, but they make up for 32% of the force. Shouldn't the nominal distribution say that blacks should make up for 32% of the test takers?

 

[jillio]

If you have 100 people, 30 of them are black, 20 of them are hispanic, and the rest is white. And you want to pick 10 of the best, assuming race isn't a factor, you should get an average of 5 whites, 3 blacks, and 2 hispanics. Give or take a few. If you do this every year, you will NOT get those exact numbers. That's where the standard deviation comes in. Sometimes you won't get hispanics at all (hence standard deviation CAN cause zero promotions for a specific race), sometimes there will be more blacks than whites, and so on.

I didn't see in the article if this is a consistent result (no blacks being promoted every year). There ARE some black people with high positions (not sure if they were hired using this very test).

Personally, I think the low income upbringing of minorities is the biggest factor in this disparity.

Interesting note: Out of the 77 test takers, 19 were blacks. That's 25% of the test takers, but they make up for 32% of the force. Shouldn't the nominal distribution say that blacks should make up for 32% of the test takers?

I see where you are coming from. I was referring to standard deviation of scores. I think you are referring more to the range of the standard deviation.

If I am not mistaken, this is the first year this particular test was used.

You can never exclude SES when you are looking at disparity in test scores.

It would depend on the given year. It is likely that even though they make up 32% of the force, 7% did not have the seniority necessary to test for eligibility for promotion.

 

[Reba]

According to the original post:

1. "The promotion exams were closely focused on firefighting methods, knowledge and skills. The first part had 200 multiple-choice questions and counted for 60 percent of the final score. Candidates returned another day to take an oral exam in which they described responses to various scenarios, which counted for 40 percent."

The multiple choice section of the exam was not 100 percent of the grade--it was 60 percent. It did cover firefighting methods, knowledge and skills. Candidates did have an oral interview portion of the exam.

"The exams were designed by a professional testing firm that followed federal guidelines for mitigating disparate racial outcomes, the plaintiffs say."

"The city says it found evidence that the tests were potentially flawed."
Who is "the city"? What kind of expert is "the city"? Also, their "evidence" showed the tests were potentially flawed, not actually flawed.

The "flaws" that the city claims were filed in a brief after the city had decided to throw out the test. At the time that the city threw out the test they didn't have any proof of bias. The only reason they tossed the test was because they didn't want white guys to take all the promotions.


2. What about bias?

"Tinney, a black lieutenant who has been a firefighter for 14 years, was seeking a promotion to captain when he took the exam.

He says both the test and his fire department have hidden biases against minorities: The department is historically white, with the first blacks joining in 1957, and jobs, relationships, knowledge and choice assignments are passed on from friend to friend and generation to generation.

"I just call it 'the network,"' Tinney says."

Tinney makes a general statement that there are "hidden" biases in the test. Has anyone investigated and proven that charge? Does he have any specific examples pertaining to this particular test?

 

[Reba]

From the Wall Street Journal:

"...Frank Ricci, the lead plaintiff, had trusted a test of merit. He had been a firefighter for 11 years and was determined to become a lieutenant. All applicants were given three months to prepare for the exam and provided with a detailed reading list. Mr. Ricci is dyslexic, so he paid an acquaintance more than $1,000 to read textbooks onto audiotapes, made flashcards, took practice tests, worked with a study group and participated in mock interviews. He gave up a second job in order to study long hours. His work paid off: He came in sixth among the 77 candidates who took the exam."

"The city set aside the results, although the test had been designed by an experienced Illinois company, Industrial/Organizational Solutions, which routinely scrubbed its assessments for any possible racial bias to protect the agencies from potential civil rights complaints."

"A representative of the black firefighters association told the New Haven civil service board that the tests were irrelevant, since they measure only the "ability to read and retain."

But the days of bucket brigades fighting fires in log cabins are long gone. As any firefighter would tell you, to deal with fires in today's technologically complex environment requires at least some understanding of mathematics, structural engineering, electricity and chemistry."

New Haven's Racial Test - WSJ.com

 

[Reba]

In case anyone is interested in the case before the Supreme Court:

http://www.abanet.org/publiced/preview/briefs/pdfs/07-08/07-1428_Petitioner.pdf

 

[Hwy99]

Discrimination and immunity don't mix.

 

[jillio]

According to the original post:

1. "The promotion exams were closely focused on firefighting methods, knowledge and skills. The first part had 200 multiple-choice questions and counted for 60 percent of the final score. Candidates returned another day to take an oral exam in which they described responses to various scenarios, which counted for 40 percent."

The multiple choice section of the exam was not 100 percent of the grade--it was 60 percent. It did cover firefighting methods, knowledge and skills. Candidates did have an oral interview portion of the exam.

"The exams were designed by a professional testing firm that followed federal guidelines for mitigating disparate racial outcomes, the plaintiffs say."

"The city says it found evidence that the tests were potentially flawed."
Who is "the city"? What kind of expert is "the city"? Also, their "evidence" showed the tests were potentially flawed, not actually flawed.

The "flaws" that the city claims were filed in a brief after the city had decided to throw out the test. At the time that the city threw out the test they didn't have any proof of bias. The only reason they tossed the test was because they didn't want white guys to take all the promotions.


2. What about bias?

"Tinney, a black lieutenant who has been a firefighter for 14 years, was seeking a promotion to captain when he took the exam.

He says both the test and his fire department have hidden biases against minorities: The department is historically white, with the first blacks joining in 1957, and jobs, relationships, knowledge and choice assignments are passed on from friend to friend and generation to generation.

"I just call it 'the network,"' Tinney says."

Tinney makes a general statement that there are "hidden" biases in the test. Has anyone investigated and proven that charge? Does he have any specific examples pertaining to this particular test?

And the Society of Industrial and Organizational Psychology found that there were definate problems with the test.

 

[jillio]

From the Wall Street Journal:

"...Frank Ricci, the lead plaintiff, had trusted a test of merit. He had been a firefighter for 11 years and was determined to become a lieutenant. All applicants were given three months to prepare for the exam and provided with a detailed reading list. Mr. Ricci is dyslexic, so he paid an acquaintance more than $1,000 to read textbooks onto audiotapes, made flashcards, took practice tests, worked with a study group and participated in mock interviews. He gave up a second job in order to study long hours. His work paid off: He came in sixth among the 77 candidates who took the exam."

"The city set aside the results, although the test had been designed by an experienced Illinois company, Industrial/Organizational Solutions, which routinely scrubbed its assessments for any possible racial bias to protect the agencies from potential civil rights complaints."

"A representative of the black firefighters association told the New Haven civil service board that the tests were irrelevant, since they measure only the "ability to read and retain."

But the days of bucket brigades fighting fires in log cabins are long gone. As any firefighter would tell you, to deal with fires in today's technologically complex environment requires at least some understanding of mathematics, structural engineering, electricity and chemistry."

New Haven's Racial Test - WSJ.com

The test was not intended to test the ability to put out a fire. It was to test eligibility for a promotion to a leadership position.

 

[jillio]

In case anyone is interested in the case before the Supreme Court:

http://www.abanet.org/publiced/preview/briefs/pdfs/07-08/07-1428_Petitioner.pdf

Yep. This is the same case that has been dismissed by two lower courts.

 

[Reba]

"STATEMENT OF THE CASE
In 2003, New Haven sought to fill command
vacancies in its fire department. Petitioners qualified
for promotion under a race-blind, merit-selection
process mandated by local law but were denied the
positions by the city because of their race and racial
disparities in the examinations’ outcome.1 City officials
refused to promote the successful candidates
and left the vacancies unfilled, intending to repeat
the competition with the aim of awarding a higher
proportion of the promotions to minority candidates."

"II. RESPONDENTS HAVE REPEATEDLY VIOLATED
THE LAW TO AVOID MERIT SELECTION.
The Connecticut Supreme Court has consistently
mandated strict compliance with civil service laws,
citing the public need for the most able workforce free
of favoritism, corruption, and the spoils system that
these laws were designed to eradicate. See, e.g., Kelly
v. City of New Haven, 881 A.2d 978, 1000-1004 (Conn.
2005) (finding that New Haven’s post hoc manipulation
of promotion test scores to evade the rule-ofthree
violated the law and undermined the purpose of
ensuring selection of the most qualified by limiting
discretion and hence patronage, race discrimination,
and corruption). The administration of New Haven’s
Mayor DeStefano has drawn multiple, stern rebukes
from state judges for “blatant lawlessness,” Henry v.
Civil Serv. Comm’n, 2001 WL 862658, *1 (Conn. Super.
July 3, 2001), in employing “charade,” id., at *2, and
“subterfuge,” Kelly, 881 A.2d, at 1003, to subvert the
law and stood accused in multiple suits of repeatedly
and intentionally discriminating against whites and
manipulating exam results for political gain. See Henry,
2001 WL 862658, *1-*3; Bombalicki v. Pastore, 2001
WL 267617, *2-*3 (Conn. Super. Feb. 28, 2001), aff ’d,
804 A.2d 856 (Conn. App. 2002); Hurley v. City of New
Haven, 2006 WL 1609974, *1 (Conn. Super. May 23,
2006); see also Pet.App. 756a-761a, 935a-937a. In
reaction, respondents unsuccessfully sought voter
approval to eliminate the rule-of-three by charter
amendment. Kelly, 881 A.2d, at 1001, n.41; see also
CA2 J.A. 1683-1684; Pet.App. 756a-761a. Only after
suffering serial setbacks in the state courts and at the
ballot box did respondents for the first time adopt the
justification of “voluntary compliance” with federal
law when they refused to grant the promotions
earned through the 2003 examinations."

"III. RESPONDENTS STROVE TO ENSURE THE 2003
EXAMS WERE JOB-RELATED AND THE PROMOTION
PROCESS WAS RACE-NEUTRAL.
Respondents engaged Industrial/Organizational
Solutions, Inc. (IOS), a professional testing firm with
experience in public safety, to develop promotional
examinations that would identify those with the
knowledge, skills, and abilities (KSAs) needed to
perform the command responsibilities of captains and
lieutenants. Pet.App. 308a-328a. The NHFD is a
multi-disciplined emergency-service agency in a port
city with major transportation networks. Lieutenants
and captains must have a sophisticated level of KSAs
and must possess considerable scientific and tactical
knowledge, leadership skills, and good judgment.
Pet.App. 348a-352a, 361a-366a. Apart from fire
science, they must be well versed in building and
high-rise construction, structural collapse, tactical
response protocols for fire and non-fire-related catastrophes,
confined-space and high-angle rescue, use of
sophisticated equipment, and other subjects. They
report directly to a battalion chief and indirectly to
the chief. They must be able to train, discipline, and
lead first responders. At the direction of the state, the
NHFD also responds to medical emergencies and
must provide pre-hospital medical care. Ibid.; see also
Pltffs.’ Exhs. 46, 47, R. Doc. #120.
The qualifying process included a written job knowledge
examination followed by a comprehensive
structured oral assessment of applicants’ skills
and abilities to command others in emergencies.
Pet.App. 1077a-1164a (containing the exams). The
job-knowledge exam accounted for 60% and the oral
assessment for 40% of the total score. Pet.App. 312a,
483a. The cutoff composite score was calibrated to
equate with minimal competence. Pet.App. 330a-
331a.
IOS composed and validated the exams based on
EEOC-recommended practices. Pet.App. 316a-317a,
329a-335a.3 Aware that New Haven, like other cities,
routinely experiences racial disparities in outcomes of
qualifying exams, IOS went to great lengths in collaboration
with city officials to mitigate that impact
to the greatest extent possible without compromising
the integrity of the exams. It engaged in a painstaking
process of job analyses, employing questionnaires,
interviews, and ride-along exercises with incumbents
to identify the importance and frequency of essential
job tasks. There was a deliberate overrepresentation
of minority incumbents in this process. Pet.App.
150a-154a, 262a-264a, 337a-343a, 597a-650a."

"IOS identified professional texts and other source
material in collaboration with NHFD Chief Grant
and Assistant Chief Dumas, who is black. Pet.App.
625a-627a, 817a-818a, 847a-848a. The written tests
went through several drafts and were pared down to
100 questions after painstaking analysis and crosschecking
of each item against source material.
Pet.App. 309a-310a, 623a. The city went beyond the
norm by allowing a three-month study period and
providing highly particularized syllabi that allowed
candidates to focus on specific chapters of each text
from which test questions were drawn. Pet.App.
346a-374a, 799a-800a. Promotional examinations are
infrequent opportunities for career advancement, and
petitioners bore significant expense and personal
sacrifice during the three-month study period.4
To mitigate any potential adverse impact, exams
were written below a tenth-grade reading level.
Further aiming to stem adverse impact, all candidates
could proceed to the oral assessment phase
irrespective of their performance in the job-knowledge
examination. Pet.App. 160a-161a, 267a-268a, 338a,
666a.
4 For example, Frank Ricci, to overcome dyslexia, paid to
convert study texts to audio recordings; Gregory Boivin resigned
from part-time jobs; Benjamin Vargas and his wife both took
leave from their second and primary jobs; and Christopher
Parker studied in his wife’s hospital room as they awaited the
delivery of their son. Pet.App. 375a-378a, 392a-398a, 402a-409a,
413a-419a."

"Unlike the written exam, which measures job
knowledge, the oral phase is designed to test skills
and abilities (though some measure of parity is
expected). Pet.App. 667a, 708a-709a. City officials
and IOS went to similar extraordinary lengths to
devise a content-valid, comprehensive, structured
oral assessment process for candidates to be rated by
panels of fire-service professionals. Seeking the most
knowledgeable assessors, IOS searched nationwide
and consulted minority organizations for officers in
the rank of captain or above. It assembled a pool of
thirty assessors that included battalion chiefs, assistant
chiefs, and chiefs, in which minorities were
overrepresented—all nine active three-member
panels included only one nonminority member (even
though most candidates were white). Pet.App. 162a-
165a, 268a-270a, 344a-345a, 654a-661a.
Panelists reviewed keyed responses and performed
mock rehearsals and exercises designed to
calibrate ratings. They were trained to engage in
consensus rating, a measure designed to ensure
consistency and prevent one assessor from skewing
results by atypical scoring. Candidates were allowed
to organize their thoughts on paper before articulating
responses to various incident scenarios and were
measured for their tactical knowledge and skills,
leadership ability, and sound judgment in life-anddeath
situations. The process was monitored by IOS
experts, and post-assessment review showed the
panel ratings were sound, consistent, and indicative
of a high level of reliability. Pet.App. 164a-168a,
269a-271a, 656a-657a, 661a-663a. Ninety applicants
took the lieutenant’s written exam, of whom seventyseven
chose to proceed to the oral assessment.
Pet.App. 168a, 271a, 428a-432a. Forty-one applicants
took the captain’s exam. Pet.App. 433a-436a. Review
of candidate feedback questionnaires showed that,
overall, candidates believed the examinations were
fair and job-related. Pet.App. 338a, 651a-652a. The
exams fairly and validly tested candidates’ relative
levels of KSAs. Pet.App. 179a, 197a-199a, 277a, 289a,
329a-339a, 603a-606a, 633a-634a, 1023a-1025a."

"IV. AFTER LEARNING THE RACE OF CANDIDATES
WHO WOULD HAVE BEEN IMMEDIATELY PROMOTED,
RESPONDENTS SCUTTLED THE PROMOTIONS.
Candidates were race-coded. The scoring results
revealed racial disparities in pass rates and levels of
KSAs for those who did pass that were similar to
adverse impact ratios seen in previous exams. Pet.App.
423a-427a, 950a-957a. Candidates of all races passed
both exams and were eligible for promotion. Pet.App.
428a-436a. Most of the petitioners were among the
top scorers and eligible for immediate promotion.
Pet.App. 24a-25a, 390a, 437a-438a. Because of the
small number of immediate vacancies, however, applying
the rule-of-three meant, according to respondents,
that no African-Americans could be immediately
promoted and the new lieutenants “w[ould] all be
white,” though two Hispanics would be among the
eight new Captains. Pet.App. 439a-445a, 475a-476a."