Thursday, May 14, 2015

Further Thoughts on the July 2014 Bar Results -- A Response to Erica Moeser

Late last fall, Erica Moeser responded to a letter from Dean Kathryn Rand of the University of North Dakota (on behalf of a large number of law school deans), reiterating that the NCBE had double-checked its scoring of the MBE on the July 2014 bar examination and could find no errors in its calculations.  Erica Moeser also took to the pages of the December 2014 issue of The Bar Examiner to further validate her conclusion that the historic drop in the mean MBE scaled score is attributable solely to the fact that the class that sat for the July 2014 bar exam was “less able” than the class that sat for the July 2013 bar exam.  In January, Dean Stephen Ferruolo of the University of San Diego also wrote to Erica Moeser requesting the release of more information on which to assess the July 2014 bar examination results in comparison with previous years’ results.  In February, Erica Moeser responded to Dean Ferruolo’s request by declining to provide more detailed information and reiterating her belief that the July 2014 scores “represent the first phase of results reflecting the dramatic and continuing downturn in law school applications.”

In an earlier blog posting, I explained why Erica Moeser is partly right (that the Class of 2014 could be understood to be slightly less able than the Class of 2013), but also explained why the decline in “quality” of the Class of 2014 does not explain the historic drop in mean MBE scaled score.  The decline in “quality” between the Class of 2013 and the Class of 2014 was modest, not historic, and would suggest that the decline in the mean MBE scaled score also should have been modest, rather than historic.  Similar declines in “quality” in the 2000s resulted in only modest declines in the MBE, suggesting that more was going on with the July 2014 exam. 

Others have written about these issues as well.  In January, Vikram Amar had a thoughtful reflection on Moeser’s statements and in recent weeks Debby Merritt has written a series of posts -- here, here, and here -- indicating in some detail why she believes, as I do, that the ExamSoft debacle in July could have impacted the MBE scaled scores in jurisdictions that used ExamSoft as well as in other jurisdictions.

I write now to take issue with four statements from Erica Moeser – three from her President’s Page in the December 2014 issue of the Bar Examiner and one from her letter responding to Dean Kathryn Rand.  I remain unpersuaded that the historic decline in the mean MBE scaled score is solely attributable to a decline in quality of the class that sat for the July 2014 bar examination and remain baffled that the NCBE refuses to acknowledge the possibility that issues with test administration may have exacerbated the decline in the performance on the July 2014 MBE.

Item One – Differential Declines in MBE Scores

In her December article, Moeser stated: 

I then looked to two areas for further corrobo­ration. The first was internal to NCBE. Among the things I learned was that whereas the scores of those we know to be retaking the MBE dropped by 1.7 points, the score drop for those we believe to be first-time takers dropped by 2.7 points. (19% of July 2014 test takers were repeaters, and 65% were believed to be first-time takers. The remaining 16% could not be tracked because they tested in jurisdictions that col­lect inadequate data on the MBE answer sheets.) The decline for retakers was not atypical; however, the decline for first-time takers was without precedent dur­ing the previous 10 years. (Emphasis in original.)

Moeser starts by referencing data that is not publicly available to support her cause.  This is unfortunate, because it makes it really hard to understand and critique the data.  Nevertheless, there are some inferences we can take from what she does disclose and some questions we can ask.  Moeser asserts that the 19% of MBE “retakers” saw an MBE drop of 1.7 points compared with MBE “retakers” in July 2013, while the 65% believed to be first-time takers saw a drop of 2.7 points compared with first-time takers in July 2013.  It would have been helpful here if Erica Moeser would have released publicly the declines among MBE retakers in the previous 10 years and the declines among first-time takers in the previous 10 years so that patterns could be assessed, particularly in relation to the changes in class composition for each of those years.  Without that information available it is hard to do much more with Moeser’s assertion.  (I find it odd that she would reference this point without providing the underlying data.) 

Nonetheless, this assertion raises other questions.  First, the overall decline in the mean MBE scaled score was 2.8 points. Moeser notes that 19% of takers (MBE retakers) had an average drop of 1.7 points, while 65% of takers (first-time takers) had an average drop of 2.7 points.  Unless there is something I am missing here, that should mean the remaining 16% of test-takers had to have an average decline of 4.51 points!  (This 16% of test-takers represents those who Moeser notes could not be tracked as first-time takers or MBE retakers “because they tested in jurisdictions that collect inadequate data on the MBE answer sheets.”) (Here is the equation --- 2.8 = (.19*1.7)+(.65*2.7)+(.16*x).  Solve for X. This translates to 2.8 = .323+1.755+.16x.  This translates to .722 = .16x and then .722/.16 = X.  X then equals 4.51.)  It would have helped, again, if Moeser had indicated which jurisdictions had these even larger declines in mean MBE scaled scores, as we could then look at the composition of graduates taking the bar in those jurisdictions to see if there was an unusual decline in entering class statistics in 2011 at the law schools from which most bar takers in those states graduated.

Item Two – The MPRE

In the December article, Moeser also stated:

I also looked at what the results from the Multistate Professional Responsibility Examination (MPRE), separately administered three times each year, might tell me. The decline in MPRE performance supports what we saw in the July 2014 MBE numbers. In 2012, 66,499 candidates generated a mean score of 97.57 (on a 50–150 scale). In 2013, 62,674 candidates generated a mean score of 95.65. In 2014, a total of 60,546 candi­dates generated a mean score of 93.57. Because many MPRE test takers are still enrolled in law school when they test, these scores can be seen as presaging MBE performance in 2014 and 2015.

At first blush, this looks like a pretty compelling argument, but Moeser’s selectiveness in looking at the data is troubling, and her failure to discuss whether the MPRE and MBE are meaningfully comparable test-taking experiences also is troubling.  Essentially, Moeser is making the following assertion – because the mean MPRE scaled score declined by 1.92 points between 2012 and 2013, we should have expected a large decline in the mean MBE scaled score in July 2014 (and because the mean MPRE scaled score declined another 2.08 points between 2013 and 2014, we should expect another large decline in the mean MBE scaled score in July 2015).

But the “relationship” between changes in the mean MPRE scaled score and changes in the mean MBE scaled score over the last decade does not support this assertion. If one looks at a decade’s worth of data, rather than data just for the last couple of years, the picture looks significantly more complicated, and suggests the collective performance on the MPRE may not tell us much at all about likely collective performance on the MBE in the following year. 

MPRE Year

Mean MPRE Score

Change

MBE Year

July Mean MBE Scaled Score

Change

2004

99.1

 

2005

141.6

 

2005

98.7

-0.4

2006

143.3

+1.7

2006

98

-0.7

2007

143.7

+0.4

2007

98.6

+0.6

2008

145.6

+1.9

2008

97.6

-1.0

2009

144.5

-1.1

2009

97.4

-0.2

2010

143.6

-.9

2010

96.8

-0.6

2011

143.8

+0.2

2011

95.7

-1.1

2012

143.4

-0.4

2012

97.6

+1.9

2013

144.3

+0.9

2013

95.6

-2.0

2014

141.5

-2.8

2014

93.6

-2.0

2015

????

????

The data Moeser cites from the last two years conveniently makes her point, but it consists of a very small sample size.  The data over the last decade looks much more random.  In three of the nine years, the change is not in the same direction (MPRE 2005, 2006, 2010, MBE 2006, 2007, 2011).  In the six years where the change is in the same direction, there are two years in which the MBE change is significantly larger than the MPRE change (MPRE 2007, 2009, MBE 2008, 2010) and there are two years in which the MBE change is significantly smaller than the MBE change (MPRE 2011, 2012, MBE 2012, 2013).  In only two of the nine years, do the changes in the MPRE and MBE roughly approximate each other (MPRE 2008, 2013, MBE 2009, 2014).   Nonetheless, this remains a very small sample and more analysis of data over a longer period might be helpful to better understand how/whether changes in mean MPRE scores inform meaningfully changes in mean MBE scores the following year.  At this point, I think the predictive value seems marginal given the wide range of changes on a year-over-year basis.

Item Three – Mean LSAT Scores

In the December article, Moeser further stated:

Specifically, I looked at what happened to the overall mean LSAT score as reported by the Law School Admission Council for the first-year matricu­lants between 2010 (the class of 2013) and 2011 (the class of 2014). The reported mean dropped a modest amount for those completing the first year (from 157.7 to 157.4). What is unknown is the extent to which the effect of a change to reporting LSAT scores (from the average of all scores to the highest score earned) has offset what would otherwise have been a greater drop. (LSAC Research Reports indicate that roughly 30% of LSAT takers are repeaters and that this num­ber has increased in recent years.

This assertion is misguided for purposes of this comparison, a point Vikram Amar made in his post.  If we were comparing the first-year matriculants in 2009 with the first-year matriculants in 2010, the question of the change in reporting from average LSAT score to highest LSAT score would have mattered.  But the 2010 matriculants were the first class for which the mean was reported based on highest LSAT score and the 2011 matriculants were the second class for which the mean was reported based on highest LSAT score.  Thus, there is no “unknown” here.  The reported mean LSAT dropped only a modest amount between the matriculants in 2010 and the matriculants in 2011.  Nonetheless, the mean MBE scaled score in July 2014 decreased by an historic 2.8 points from the mean MBE scaled score in July 2013. 

Item Four – Administration Issues

In her letter to Dean Kathryn Rand, Moeser stated:  "To the extent that the statement you attached referenced both administration and scoring of the July 2014, bar examination, note that NCBE does not administer the exam; jurisdictions do."

This response suggests not only that the NCBE is not responsible for administering the bar examinations in the many different jurisdictions, but implicitly suggests that issues with administration could not have contributed to the historic decline in the mean MBE scaled score. 

Were there issues with administration?  Yes.   Could they have contributed to the historic decline in the mean MBE scaled score?  Yes.

Debby Merritt’s recent posts discuss the administration issues and the potential consequences of the administration issues in some detail.  In over forty states that used ExamSoft to administer the bar examination, the MBE came on Wednesday, after the essay portion of the exam on Tuesday.  But because of an ExamSoft technical problem, tens of thousands of test-takers, who were initially informed by their respective state board of bar examiners that they would FAIL THE EXAM if their essay answers were not uploaded in a timely manner, spent most of Tuesday night dealing with the profound stress of not being able to upload their exam answers and not being able to contact anyone at the board of bar examiners (who were not answering phones) or at ExamSoft (due to the flood of calls and emails from anxious, frustrated, stressed out exam takers) to figure out what was going on and what they should do. 

Given that this “administration” issue caused untold stress and anxiety for thousands of test-takers, who spent Tuesday night completely anxious and stressed out trying repeatedly and unsuccessfully to upload their essay answers, should it be a surprise that they might have underperformed somewhat on the MBE on Wednesday?  (If you want a sense of the stress and anxiety, check the twitter feed for the evening of Tuesday, July 29, 2014)

The responses from the boards of bar examiners to this issue with administration of the bar examination were far from uniform.  Different jurisdictions granted extensions at different times of the night on Tuesday, July 29, or on Wednesday, July 30, with some granting short extensions and some granting longer extensions.  Thus, in states that gave notice of an extension out earlier on Tuesday, July 29, test-takers may have had less stress and anxiety, while in those states that didn’t give notice of an extension out until later (or for which the extension was relatively short), or where there may not have been any communication regarding extensions of the submission deadline, test takers likely experienced more stress and anxiety.  (It would be worth studying exactly when each jurisdiction gave notice of an extension and whether there is any correlation between timing of notice of the extension and the relative performance of bar takers in those states.)

The NCBE’s unwillingness to acknowledge any issues with administration of the bar examination is all the more surprising at a time when the NCBE is pushing for adoption of the Uniform Bar Examination.  On its webpage, the NCBE states: “[The UBE] is uniformly administered, graded, and scored by user jurisdictions and results in a portable score that can be transferred to other UBE jurisdictions.” (Emphasis added.)  This simply was not true in July 2014.  The Uniform Bar Examination was administered under different exam conditions across jurisdictions.  First, three of the states administering the Uniform Bar Examination in July 2014 did not use ExamSoft – Arizona, Nebraska and Wyoming -- and therefore, bar takers in those states had a vastly different “exam administration” experience than bar takers in ExamSoft jurisdictions.  Across ExamSoft jurisdictions, different approaches to extensions also meant different administration experiences. Given the significance of consistent administration for the purpose of equating performance on a standardized exam like the bar exam, that the NCBE allows such varied approaches to administering a supposedly “uniform” exam strikes me as very problematic.

Many questions remain unanswered, largely because adequate information has not been made available on which to assess the various factors that might have contributed to the historic decline in the mean MBE scaled score.  With the release of February bar results and the NCBE’s publication of the 2014 statistical report, some additional information is now available to put the results of July 2014 in context.  In my next blog posting regarding the July 2014 bar results, I will delve into some of those statistics to see what they tell us.

(Edited as of May 20 to correct the 2013 MPRE and 2014 MBE change and corresponding discussion.)

http://lawprofessors.typepad.com/legalwhiteboard/2015/05/further-thoughts-on-the-july-2014-bar-results-a-response-to-erica-moeser.html

Current events, Data on legal education, Data on the profession | Permalink

Comments

Post a comment