"A Bully Pulpit Approach to Elections in the Early Biden Administration"

I have this piece at the Illinois Law Review Online entitled A Bully Pulpit Approach to Elections in the Early Biden Administration. It begins:

2020 yielded innovation to administer an election during a pandemic, high voter participation, divisive social media maelstroms, and extraordinary doubt cast by the losing presidential candidate on the results, all culminating in a riot at the Capitol during the counting of electoral votes. Election disputes looked bad in 2016,1 but they appear to have become even worse. Election law has attracted President Joe Biden’s attention, but much less in legislative reforms and much more in the bully pulpit to advance his agenda.

2020 employment figures show big swings at some law schools

I offered my overall assessment of the state of employment for the Class of 2020, as recently released. Some schools saw significant improvement in placement in full-time, long-term, bar passage-required jobs. Many schools are the same ones that saw significant improvements in bar passage rates or are in “diploma privilege” jurisdictions, which is perhaps no surprise.

	BPR 2020	BPR 2019	Delta
Arkansas	75.9%	59.6%	16.3
Chapman	69.7%	53.8%	15.9
Montana	87.0%	73.2%	13.7
Gonzaga	81.7%	68.2%	13.5
Ave Maria	52.6%	39.4%	13.2
Willamette	81.5%	69.0%	12.5
Dayton	81.4%	69.0%	12.4
Detroit Mercy	80.5%	68.2%	12.3
Southwestern	57.1%	46.0%	11.1
South Carolina	81.6%	70.7%	10.9
South Dakota	80.4%	70.0%	10.4
Utah	84.8%	75.3%	9.5
Florida A&M	49.1%	40.3%	8.8
Elon	73.1%	65.0%	8.1
San Francisco	47.6%	39.5%	8.1
Seattle	70.3%	62.6%	7.7
Wayne State	77.5%	69.9%	7.5
St. Thomas (Florida)	65.2%	58.0%	7.3
Oklahoma City	72.8%	65.9%	6.9
Idaho	75.5%	68.8%	6.8

Other schools—especially those in states hardest hit by the pandemic like New York, or those schools that also saw a significant drop-off in bar passage rates, did not fare as well:

	BPR 2020	BPR 2019	Delta
CUNY	46.3%	75.0%	-28.7
North Carolina Central	23.7%	50.4%	-26.7
Touro	52.2%	75.2%	-23.0
Ohio Northern	62.0%	80.9%	-18.9
Oklahoma	70.5%	89.1%	-18.6
UMass-Dartmouth	42.4%	60.9%	-18.5
NYLS	56.6%	73.5%	-17.0
Widener-Commonwealth	62.6%	78.9%	-16.3
Pace	63.5%	79.5%	-15.9
Quinnipiac	46.1%	61.1%	-15.0
Pepperdine	61.6%	75.8%	-14.2
Southern Illinois	48.5%	61.9%	-13.4
Florida International	70.1%	83.2%	-13.1
Mitchell|Hamline	49.0%	61.8%	-12.9
Faulkner	50.0%	62.7%	-12.7
Widener-Delaware	48.1%	60.6%	-12.5
Baylor	79.2%	91.3%	-12.1
Western New England	39.8%	51.9%	-12.1
Liberty	68.4%	80.4%	-12.0
Mercer	73.3%	85.3%	-11.9

2020 bar exam saw major swings in pass rates at some institutions

Given all the variance we saw in bar exam administration in 2020, it’s perhaps no surprise to see some significant swings in how individual schools performed on the bar exam. I looked at one metric: a school’s modal bar exam jurisdiction, and how it over- or under-performed relative to the statewide average. That’s the USNWR metric (even though a state’s modal jurisdiction might change year over year). California significantly lowered its cut score, for instance; other states had a version of “diploma privilege,” which assuredly altered what bar exam test-takers looked like. Schools that always do extremely well are not going to be in the top of any list of annual year-over-year changes. And some gain by gaining; others gain by staying still while others perform worse.

I pulled this data from the ABA spreadsheets. The ABA did not include some data for some schools in diploma privilege-heavy jurisdictions, so those schools are omitted. Here are the ten schools that did the best year-over-year in terms of their performance compared to the statewide average:

	2019 pass pct	2019 state pct	2020 pass pct	2020 state pct	Delta
San Francisco	38.7%	69.1%	76.1%	77.9%	28.6
Roger Williams	68.9%	82.1%	70.0%	60.0%	23.16
Northern Kentucky	62.5%	79.5%	85.4%	80.7%	21.75
Southwestern	56.2%	69.1%	82.7%	77.9%	17.84
District of Columbia	52.0%	72.1%	70.6%	75.0%	15.66
Hofstra	60.2%	84.1%	76.6%	85.9%	14.61
Georgetown	90.8%	84.1%	96.2%	75.0%	14.53
Chapman	59.3%	69.1%	80.1%	77.9%	12.13
Oklahoma City	69.5%	84.2%	83.7%	86.6%	11.77
Tulane	76.3%	74.1%	88.9%	75.1%	11.59

Maybe unsurprisingly, a few California schools saw dramatic improvement as the cut score was lowered. Others like Georgetown appeared to find success as the bar rate suffered elsewhere.

Here are the ten schools that saw the biggest declines in year-over-year performance compared to the statewide average:

	2019 pass pct	2019 state pct	2020 pass pct	2020 state pct	Delta
Atlanta's John Marshall	63.1%	65.4%	58.3%	76.1%	-15.6
Chicago-Kent	81.1%	74.1%	74.3%	81.0%	-13.8
Western State	56.7%	69.1%	51.7%	77.9%	-13.7
Belmont	97.1%	81.0%	82.4%	80.0%	-13.7
Ohio Northern	95.0%	79.5%	86.4%	84.3%	-13.4
Tulsa	91.8%	84.2%	81.0%	86.6%	-13.3
Faulkner	81.8%	86.3%	62.2%	79.6%	-13
Pepperdine	80.3%	69.1%	76.1%	77.9%	-13
CUNY	74.5%	84.1%	65.4%	85.9%	-10.9
Lewis & Clark	84.3%	80.9%	79.4%	86.8%	-10.7

The timing of congressional election contests in Iowa

On the heels of Rita Hart’s decision to withdraw her congressional election contest in Iowa’s Second Congressional District dispute, the final formal ratification of the election Congresswoman Mariannette Miller-Meeks, I want to return to one point about the nature of Iowa’s contest court.

I’ve argued that failure to exhaust state remedies is a reason for Congress to reject an election contest. True, as some of the briefing has pointed out, (a) it is not a reason under the Federal Contested Elections Act for a prompt dismissal before discovery, as the Act contemplates four specific categories for early dismissal; and (b) it is not a requirement, as Congress can, of course, alter its precedents as it sees fit. But it has, in the past, required exhaustion of post-Election Day mechanisms in state court as a basis for rejecting a complaint.

That’s for a couple of good reasons, in my view. The first is the development of a contemporaneous fact record. Indeed, consider one remark by the Hart campaign’s attorneys to Congress during the contest: “With no written rules of procedure or official records of recount board votes, descriptions of these ad hoc processes are subject to the limitations of human memory.” Of course, a process that plays out months later in Congress looks very different from a process that plays out days after the recount concludes in state court.

The second is an interpretation of state law. Many of the allegations in the Hart contest turned on definitions of state law—what kinds of unsealed envelopes can or cannot be counted, when “commissioner” refers to the commissioner of a voter’s county or a commissioner of any county, and so on. They were nuanced questions of state law. And for the most part, Congress defers to state law—although in exceptional circumstances, as the Hart contest notes, “the Committee should therefore exercise its discretion to depart from Iowa law.” But starting with state law is best understood as starting with what a state’s courts interpret the law as—not starting with members of Congress construing it.

To do so—to address challenges closely in time to ensure there’s a fresh record, and to ensure state courts get a first look at questions of state law—the Iowa contest court has a fixed jurisdiction. It is a creation solely of statute, and the Court cannot exist beyond the jurisdiction authorized by the legislature.

It also has no authority to engage in a second recount—the recount process is performed by county boards, created statute. Instead, here are the grounds for a contest:

2. Grounds for contesting an election under this chapter are:

a. Misconduct, fraud or corruption on the part of any election official or of any board of canvassers of sufficient magnitude to change the result of the election.
b. That the incumbent was not eligible to the office in question at the time of election.
c. That prior to the election the incumbent had been duly convicted of a felony, as defined in section 701.7, and that the judgment had not been reversed, annulled, or set aside, nor the incumbent pardoned or restored to the rights of citizenship by the governor under chapter 914, at the time of the election.
d. That the incumbent has given or offered to any elector, or any precinct election official or canvasser of the election, any bribe or reward in money, property, or thing of value, for the purpose of procuring the incumbent’s election.
e. That illegal votes have been received or legal votes rejected at the polls, sufficient to change the result of the election.
f. Any error in any board of canvassers in counting the votes, or in declaring the result of the election, if the error would affect the result.
g. That the public measure or office was not authorized or required by state law to appear on the ballot at the election being contested.
h. Any other cause or allegation which, if sustained, would show that a person other than the incumbent was the person duly elected to the office in question, or would show the outcome of the election on the public measure in question was contrary to the result declared by the board of canvassers.

The only recanvass right is as follows:

The parties to any contested election shall have the right, in open session of the court or tribunal trying the contest, and in the presence of the officer having them in custody, to have the ballots opened, and all errors of the precinct election officials in counting or refusing to count ballots corrected by such court or tribunal.

All these things require specificity. There is no freestanding right to a second recount. In a way, that entirely makes sense. The recount is over. The contest court is about specific legal challenges. For instance, the recount board could not count ballots that were not counted on Election Day. The recount board does not determine whether there was fraud. And so on.

One could, I suppose, construed any “error” in “counting” so broadly that the whole district needs to be recounted. But given the extensive and specific rules elsewhere about how to engage in a recount, and the absences of such mechanics in the contest provision, it seems to be an inappropriate remedy.

Some commentary in recent weeks, however, lamenting Ms. Hart’s loss has misconstrued, I think, the nature of the contest court.

Consider this take from John Deeth of Johnson County:

After the recount that left her six votes short was canvassed on November 30, Hart had an extremely short timeline, just over a week to request and complete a state level challenge to the election. Team Hart decided, and they weren't wrong, that a week was insufficient time. The challenge process is designed for small city councils or maybe a legislative district, not for a quarter of the state.

That’s demonstrably false—the challenge process was specifically designed for statewide presidential elections, and later included all congressional contests, both Senate and House. It was specifically designed for large elections. It was enacted shortly after the federal Electoral Count Act of 1887 was enacted, designed for presidential elections. It was updated in 1970 to allow for congressional election contests—again, specifically for congressional contests.

Consider, too, this take from Laura Belin over at Bleeding Heartland:

For the sake of optics, Hart should have attempted to contest the election in Iowa before going to Congress. But her legal team was correct about the time constraints. The Iowa contest court could never have thoroughly considered the issues at hand, much less order the full recount the situation warranted, by the December 8 deadline. (Some states allow much more time for federal election contests; Norm Coleman’s contest of Minnesota’s 2008 U.S. Senate race took months to resolve.)

This, too, inaccurately describes the law. The contest court is not a court in Iowa to offer a second recount (or a third count). It is a limited venue for particular grievances.

I also want to offer some contrast about why the Iowa contest court can move dramatically faster than the Minnesota contest court, specifically in the context of Coleman-Franken 2008:

• Coleman-Franken looked at 3 million ballots, not the 400,000 in Hart—Miller-Meeks; admittedly, a statewide contest in Iowa would also involve far more ballots, but not in this case.

• The election in 2008 was not certified until January 5; in Iowa, the congressional race was certified November 30.

• To that end, the 2008 election had litigation about 12,000 rejected absentee ballots that delayed the initial certification.

• The Minnesota Supreme Court took two weeks to select a trial panel; by statute, the Iowa Supreme Court has two days to select a panel.

• After the Minnesota trial court issued its decision, there is a right of appeal to the Minnesota Supreme Court; in contrast, there is no appeal from the Iowa contest court.

• To that end, there was a ten-week delay between the trial court’s decision appealed to the Minnesota Supreme Court.

All that is to say, if you delay certification, delay the selection of the contest court, and allow a right of appeal, of course the process is going to take dramatically longer. By my count, Iowa’s law offers at least 17 weeks of increased efficiency over Minnesota’s at a bare minimum—by constraining the certification deadline, fixing the time to compose the court, and eliminating a second level of judicial review, Iowa’s contest court procedure offers far more efficiencies.

Now, it’s entirely plausible, I think, to consider extending the timeline a couple of weeks, for congressional elections (not presidential elections). And with some light modifications, the recount boards can still meet their end-of-November deadline with some added personnel to assist counting; instead, it would be to give the contest court an extra couple of weeks to decide any issues. All this could easily be wrapped up by January 3—to allow a state to have full representation in Congress and to avoid needless delays and vacancies that occurred in Minnesota.

Overall legal employment for the Class of 2020 declines slightly, with large law firm and public interest placement growing

Pandemic, lockdowns, delayed bar exams—there were many challenges facing the Class of 2020, whose graduations were moved online and whose job opportunities became all the more perilous. The trends were slightly negative, which is maybe impressive given those challenges. Below are figures for the ABA-disclosed data (excluding Puerto Rico’s three law schools). These are ten-month figures from March 15, 2021 for the Class of 2020.

	Graduates	FTLT BPR	Placement	FTLT JDA
Class of 2012	45,751	25,503	55.7%	4,218
Class of 2013	46,112	25,787	55.9%	4,550
Class of 2014	43,195	25,348	58.7%	4,774
Class of 2015	40,205	23,895	59.4%	4,416
Class of 2016	36,654	22,874	62.4%	3,948
Class of 2017	34,428	23,078	67.0%	3,121
Class of 2018	33,633	23,314	69.3%	3,123
Class of 2019	33,462	24,409	72.9%	2,799
Class of 2020	33,926	24,006	70.8%	2,514

The placement is still quite good. There was a decline in just 400 bar passage-required jobs year-over year, and the graduating class size increased for the first time in several years. Those yielded a drop to 70.8%—still better than the Class of 2018. But there continues a notable decline in J.D. advantage jobs, which have dropped nearly in half in 6 years, to 2514.

We can see some of the year-over-year categories, too.

FTLT	Class of 2019	Class of 2020	Net	Delta
Solo	236	260	24	10.2%
2-10	4,761	4,948	187	3.9%
11-25	1,769	1,755	-14	-0.8%
26-50	1,075	1,010	-65	-6.0%
51-100	864	856	-8	-0.9%
101-205	1,059	1,001	-58	-5.5%
251-500	1,044	1,030	-14	-1.3%
501+	4,976	5,073	97	1.9%
Business/Industry	2,801	2,546	-255	-9.1%
Government	3,656	3,189	-467	-12.8%
Public Interest	2,146	2,284	138	6.4%
Federal Clerk	1,197	1,126	-71	-5.9%
State Clerk	2,135	1,938	-197	-9.2%
Academia/Education	296	269	-27	-9.1%

Last year’s sharp uptick in public interest placement was not an outlier. Public interest job placement is up over 50% in two years. Last year’s eye-popping number rose further. It is likely not an understatement to say that law students are increasingly oriented toward public interest, and that there are ample funding opportunities in public interest work to sustain these graduates.

Additionally, extremely large law firm placement continues to boom. Placement is up more than 1000 graduates in the last several years, breaking 5000.

Despite a bevy of new federal judges confirmed to the bench, federal clerkship placement slid, a suggestion, perhaps, that federal judges continue to look toward clerks with experience. The drop in state clerkship and government clerkship placements might be related to the pandemic, but we’ll see if those rebound next year.

Diploma privilege, July bar exam administration, and law school employment outcomes

We saw a lot of variance in how the bar exam was administered in 2020. That assuredly affected employment outcomes for the Class of 2020, specifically the 10-month employment figures publicly released this week. (The underlying data is here.) Employment figures dropped. But the drop was not evenly distributed. And I think we can learn some things from decisions to introduce a version of “diploma privilege,” and decisions to maintain a July administration of the bar exam—I think. Maybe.

Most graduates of most law schools take the bar in the state where the law school is. There are obviously huge outliers (Yale, among others). And most of the “emergency” diploma privilege jurisdictions—Louisiana, Oregon, Utah, and Washington—particularly favored in-state law schools. (Please note, I exclude the District of Columbia, because its “diploma privilege” is really a lengthy supervised practice requirement.) It’s only four jurisdictions with only about 5.1% of all law school graduates. Yes, some of them take other bar exams, and law school graduates in other states may take advantage of diploma privilege here. But we could compare this cohort to graduates of law schools from other states—the 46 others (there are no law schools in Alaska, and I excluded the law schools in Puerto Rico). (Numbers may not evenly match due to rounding.)

	2020 BPR	2019 BPR	Delta	2020 JDA	2019 JDA	Delta
Emergency diploma privilege (4)	69.9%	68.2%	1.7	7.6%	9.5%	-2.0
Others (46)	70.8%	73.2%	-2.4	7.4%	8.3%	-0.9

The raw numbers are below:

	2020 BPR	2020 JDA	2020 Grads	2019 BPR	2019 JDA	2019 Grad
Emergency diploma privilege	1,221	132	1,747	1,143	160	1,677
Others	22,785	2,382	32,179	23,244	2,631	31,745

Bar passage-required (“BPR”) job outcomes (“full-time, long-term,” for this and all other categories) rose 1.7 points between 2019 and 2020 among graduates from schools in these four states, compared to a 2.4-point decline in the rest of the country. The correlation is of limited value, of course, but I think it’s a good place to start considering the effect the bar exam might have on legal employment.

A related component is to compare J.D.-advantage (or “JDA”) positions—i.e., positions for which passing the bar is not a prerequisite. In the four “diploma privilege” jurisdictions, J.D.-advantage position placement declined 2 points, whereas in the rest of the country it declined 0.9 points. (Yes, those JDA figures are small….)

This inverse relationship between BPR and JDA suggests, I think, that diploma privilege did not inherently improve overall job placement at a school; instead, it may have shifted graduates from less desirable positions into more desirable ones—or, more importantly, shifted graduates from non-practice positions into the practice of law. To that end, diploma privilege does exactly what it’s designed to do (if this correlation is sufficient to suggest diploma privilege is doing something…).

But not all four of these states saw equal changes to job placement. Louisiana saw a 0.9-point decline in bar passage-required job placement, while Oregon saw it rise 0.5, Utah rise 7.9, and Washington rise 5.3. Is there something about the West that saw a better job market than the East? Hold that thought while I address another….

States had significantly different timing for their bar exam administration. About 1/4 of graduates came from 22 states that offered a July administration of the bar exam compared to states that did not. (Not all graduates, of course, took the July test—many may have deferred, or taken a later offering.) Some postponed a week to August, like Indiana; others canceled it entirely, like Delaware. Some had it in July just like normal, others offered a July exam and an additional fall exam. Did jurisdictions that had a July bar exam look any different in employment outcomes? I pulled out the four emergency diploma privilege jurisdictions (but I did keep Wisconsin).

	2020 BPR	2019 BPR	Delta	2020 JDA	2019 JDA	Delta
July 2020 bar offered (22)	72.8%	73.6%	-0.8	7.3%	8.5%	-1.2
No July 2020 bar offered (24)	70.1%	73.1%	-3.0	7.4%	8.2%	-0.8

And the raw numbers are below.

	2020 BPR	2020 JDA	2020 Grads	2019 BPR	2019 JDA	2019 Grad
July 2020 bar offered	6,015	604	8,265	6,035	698	8,198
No July 2020 bar offered	16,770	1,778	23,914	17,209	1,933	23,547

These results show that employment looked better in states with a July bar exam. Placement in bar passage-required jobs declined from 73.6% to 72.8%, a 0.8-point drop in these states. Placement in other states, however, declined from 73.1% to 70.1%, nearly a 3-point drop. And we see the opposite trends in J.D.-advantage jobs again, too—J.D. advantage jobs declined 1.2 points in July bar exam states, but only 0.8 points in other states.

There’s a geographic divide, however, between Eastern and Western states. Western states (16 in total, about 1/4 of all grads) saw a decline of just 0.9 points in bar passage-required placement, while Eastern states (34, 3/4) saw it fall 2.7 points.

Break those down further into states with a July bar exam or without, and the divide becomes starker still—exacerbated, of course, by New York, which did not have a July bar exam, has the largest potential employment market, and was hit hardest by the pandemic.

	2020 BPR	2019 BPR	Delta	2020 JDA	2019 JDA	Delta
West, July 2020 bar offered (12)	74.0%	73.0%	1.0	7.1%	8.4%	-1.2
East, July 2020 bar offered (12)	72.3%	73.4%	-1.1	7.4%	8.7%	-1.3
West no July 2020 bar offered (6)	68.2%	70.0%	-1.8	6.6%	8.3%	-1.7
East, no July 2020 bar offered (32)	70.6%	73.8%	-3.2	7.7%	8.3%	-0.5

Once more, the raw numbers:

	2020 BPR	2020 JDA	2020 Grads	2019 BPR	2019 JDA	2019 Grad
West, July 2020 bar offered	2,022	195	2,734	2,044	234	2,800
East, July 2020 bar offered	4,613	470	6,384	4,550	537	6,201
West no July 2020 bar offered	4,144	401	6,072	4,278	508	6,110
East, no July 2020 bar offered	13,227	1,448	18,736	13,515	1,512	18,311

It’s only a small cache of data that only reflects a sliver of some of the variables at play in the employment outcomes. But it does appear that diploma privilege or sticking with a July bar exam administration had a positive effect on employment. Of course, we can run back around the correlation-causation fights: were those jurisdictions with July bar exams least affected by the pandemic, in which case legal employment was able to hire more, as opposed to anything about the timing of the bar exams. But it’s some data worth considering when examining the costs of changes to the bar exam.

For the second year in a row, Alabama's admissions standards (partially) trump Yale's

For the second year in a row, Alabama has reported that its incoming 1L class has a 75th percentile undergraduate GPA of (exactly) 4.0. That means at least 25% of its incoming class has a 4.0 GPA or higher. That trumps all schools, including Yale, which this year at a 75th percentile UGPA of 3.99.

And Alabama also reports this year a 50th percentile undergraduate GPA of 3.94. That’s tops in the nation, tied with Yale. (Harvard’s is 3.88, for comparison.)

Above-4.0 GPAs are not uncommon, because LSAC calculates an A+ as a 4.33. It all depends on one’s undergraduate program and the frequency of A+s, I suppose.

Alabama enrolled 127 1Ls last year, so that’s about 32 students with a 4.0 or higher, and about 64 with a 3.94 or higher. The total, Alabama reports, includes 11 students without an LSAT score—that cohort instead relies on UGPA and ACT scores (as some schools have done in the past consistent with ABA regulations), and that cohort of 11 had a mean UGPA of 4.04. The threshold for participation in that program is a 3.90 UGPA (and it’s extending a “streamlined” program with a 3.90 UGPA requirement to a number of Alabama schools this fall).

"The Diamonds Hidden in H.R. 1's Massive Mine"

Over at RealClearPolitics, I have this piece, "The Diamonds Hidden in H.R. 1’s Massive Mine.” It begins:

At a whopping 886 pages, H.R. 1, the For the People Act of 2021, has stirred plenty of controversy. It passed the House along almost perfectly partisan lines: 220 Democrats supported it; 209 Republicans and one Democrat opposed it. The Senate is considering a similar bill.

But within those 886 pages are at least a few provisions that can generate some consensus. Most are rolled over from previous failed bills in Congress, and if they were standalone measures, perhaps they could garner supermajority bipartisan support.

New USNWR metric favors $0 loans over $1 loans for graduating law students

On the heels of the new USNWR law school rankings, with its many pre-release errors, we also have a new metric: student debt. It’s actually two components: 3% weight to average law school debt incurred upon graduation among those incurring debt; and 2% weight to the percentage of students who took out loans.

The only problem? It artificially boosts schools with relatively high debt loads and relatively high percentages of students who do not take out loans. Call it the “independently wealthy law student enrollment” bonus. (Of course, it could also be disproportionate full tuition plus living expense scholarships, but that seems less likely.)

When I started tracking law school affordability years ago, I considered students who took out no debt as $0, weighted in the average of law school debt. I included the caveats as I’ve continued to do. But it doesn’t make much sense to treat these two cohorts differently. Some take out loans; others don’t. The average is the average. Include some caveats and weigh all the students together.

USNWR thinks otherwise.

It separately ranks average indebtedness and students who take out loans. This results in a skewing of results against schools with relatively low debt but a relatively high percentage of students who take out loans; and in favor of schools with relatively high debt but a relatively low percentage of students who take out loans. The latter is the independently wealthy law student enrollment bonus.

This is exacerbated by the tremendous compression for law schools around the percentage of students incurring debt. The difference between 74% and 75%—really, a rounding error, is 0.044 “scaled and weighted” points. That’s the same as the difference an extra $2300 in average tuition indebtedness, which converts to about 0.044 “scaled and weighted” points. Even though the average is 3% and the percentage incurring debt is 2%, the disparity is stark.

UPDATE: I should add, if one were to weight them at 2% and 3%, the figures would be much smaller, more like 0.003. I was doing a 60%-40% allocation using the larger figures to compare these figures to one another, not as components of the overall rankings. But the relative figures still hold, whichever “scaled and weighted” points one wants to use.

How stark? Assume a graduating class with 150 students. Let’s assume, for the moment, they have the average debt of $108,000 (about the median) and 74% indebtedness (about the median).

To move from 75% indebted to 74% indebted is 2 students. Pick off the two students with the least debt and tell them they qualify for some student loan forgiveness program—really, the two students least in need of it. That could bring you down 0.044 points.

In contrast, among the remaining 111 students indebted, they bear $11,988,000 in cumulative debt—nearly $12 million even. To bring the average debt loan down $2300—an equivalent amount of scaled and weighted points—you’d need to trim a whopping $255,300 from the cumulative debt load to bring the average down. (Admittedly, a school can only “game” this so much before the average indebtedness begins to rise substantially among the remaining students.)

It’s a poor metric, in my judgment, when the tradeoffs are placed against each other like that. I ran some figures to examine how the methodology would play out if we had a 5% weight for the single metric of average indebtedness, factoring in students who did not incur any debt among the overall debt figures. I then compared that metric to a 3%-2% weight separating the schools. (The raw figures are here.)

Here’s my rough calculation of the ten schools that benefited most, and the ten schools that benefited least, from the decision to separate the columns rather than to incorporate together. Remember, these aren’t sorted by lowest-to-highest debt loads. Instead, they’re sorted by those who benefited the most from the decision to separate the two metrics rather than put them together.

USNWR debt metric winners (over metric that averages all students together)

Southwestern Law School $190,184 - 82% ($155,950)

California Western School of Law $164,918 - 88% ($145,127)

St. Thomas University $161,701 - 86% ($139,062)

Nova Southeastern University (Broad) $155,193 - 87% ($135,017)

Columbia University $190,141 - 66% ($125,493)

Golden Gate University $151,854 - 83% ($126,038)

American University (Washington) $159,723 - 76% ($121,389)

Harvard University $170,866 - 70% ($119,606)

University of San Francisco $156,460 - 77% ($120,474)

Florida Coastal School of Law $145,245 - 86% ($124,911)

While some schools have percentage incurring debt above the median of 74%, even 85% indebtedness, when weighted, only presses schools so high because of compression. When factoring into the overall debt figures, however, it keeps these schools will above the overall median.

USNWR debt metric losers (over metric that averages all students together)

University of South Dakota $53,253 - 80% ($42,602)

Cleveland State University (Cleveland-Marshall) $69,727 - 90% ($62,754)

Florida A&M University $61,500 - 81% ($49,815)

Ohio Northern University (Pettit) $71,134 - 88% ($62,598)

University of Nebraska--Lincoln $63,027 - 78% ($49,161)

Rutgers University $62,210 - 75% ($46,658)

University of North Dakota $67,281 - 78% ($52,479)

University of Arkansas--Fayetteville $68,877 - 79% ($54,413)

Texas Tech University $56,898 - 72% ($40,967)

University of Utah (Quinney) $76,344 - 85% ($64,892)

These schools—unsurprisingly, public schools with low tuition and low indebtedness—suffered in the rankings because a 90% indebtedness looks absolutely terrible compared to the 74% median; in contrast, it would seem that these students are, on the whole, dramatically better off than some of the ones that benefited from a separate ranking.

One notable outlier is the University of Tulsa, with around $90,000 indebtedness, well below the median, but reporting 100% indebtedness, putting it at the very top and giving it an extremely poor score on this metric. But when balanced against the actual median of about $80,000, it performs only slightly above the median and would benefit significantly.

It’s worth noting that some schools that most benefited had above-median (against, about 74%) percentages, and some schools that benefited least had below-median percentages. But the size of the debt was spread far more significantly from top to bottom, which skewed the overall results.

In short, if schools want to manipulate this poor metric from USNWR, the solution is to buy off the least debt-ridden students before graduation to reduce their loans to $0—or admit far more independently wealthy students, or provide more full-tuition scholarships at the expense of partial tuition scholarships. Those, in my judgment, are the wrong incentives.