Over the years, the vast majority of law schools have
participated in LSAT validity studies that examine the relationship between students’ LSAT scores
and their first-year grades in law school. The studies show that
LSAT scores help to predict which students will do well in law
school. Moreover, a combination of students’ scores and
undergraduate grade-point averages (GPAs) gives a better
prediction than either the LSAT or the GPA alone.
Correlation is stated as a coefficient for which 1.00 indicates
an exact positive correspondence between candidates’ test
scores and subsequent law school performance. A coefficient
of zero would indicate nothing more than a coincidental
relationship between test scores and subsequent performance. The closer to 1.00 the correlation coefficient is, the
greater the test’s predictive validity. In other words, the closer
to 1.00 the correlation coefficient is, the less chance there will
be of candidates with high LSAT scores failing in their studies
or candidates with low test scores performing at the top of
their law school class.
The correlation between LSAT scores and first-year law school
grades varies from one law school to another (as does the
correlation between GPA and first-year law school grades).
During 2014, validity studies were conducted for 178 law
schools. Correlations between LSAT scores and first-year law
school grades ranged from .19 to .56 (median is .38). The
correlations between UGPA and first-year law school grades ranged
from .06 to .43 (median is .26). However, correlations between
LSAT scores combined with undergraduate grade-point averages and first-year law school grades ranged from .31 to
.64 (median is .48).
The LSAT, like any admission test, is not a perfect predictor
of law school performance. The predictive power of an
admission test is limited by many factors, such as the
complexity of the skills the test is designed to measure and
the unmeasurable factors that can affect students’
performances (e.g., motivation, physical and mental health,
or work and family responsibilities). The LSAT is a strong
predictor of first-year law school grades and compares very
favorably with admission tests used in other graduate and
professional fields of study.
Test Score Accuracy—Reliability and Standard Error
Reliability is a measure of how consistently a test measures
the skills being assessed. The higher the reliability coefficient for
a test, the more certain we can be that test takers would get
very similar scores if they took the test again.
LSAC reports an internal consistency measure of reliability for
every test form. Reliability can vary from 0.00 to 1.00, and a
test with no measurement error would have a reliability
coefficient of 1.00 (never attained in practice). Reliability
coefficients for past LSAT forms have ranged from .90 to .95,
indicating a high degree of consistency for these tests. LSAC
expects the reliability of the LSAT to continue to fall within
the same range.
LSAC also reports the amount of measurement error
associated with each test form, a concept known as the
standard error of measurement (SEM). The SEM, which is
usually about 2.6 points, indicates how close a test taker’s
observed score is likely to be to his or her true score. True
scores are theoretical scores that would be obtained from
perfectly reliable tests with no measurement error—scores
never known in practice.
Score bands, or ranges of scores that contain a test taker’s
true score a certain percentage of the time, can be derived
using the SEM. LSAT score bands are constructed by adding
and subtracting the (rounded) SEM to and from an actual
LSAT score (e.g., the LSAT score, plus or minus 3 points).
Scores near 120 or 180 have asymmetrical bands. Score bands
constructed in this manner will contain an individual’s true
score approximately 68 percent of the time.
Measurement error also must be taken into account when
comparing the LSAT scores of two test takers. It is likely that small
differences in scores are due to measurement error
than to meaningful differences in ability. The standard error of
score differences provides some guidance as to the importance of differences between two scores. The standard error
of score differences is approximately 1.4 times larger than the
standard error of measurement for the individual scores.
Thus, a test score should be regarded as a useful but
approximate measure of a test taker’s abilities as measured by
the test, not as an exact determination of his or her abilities.
LSAC encourages law schools to examine the range of scores
within the interval that probably contains the test taker’s true
score (e.g., the test taker’s score band) rather than solely interpret the reported score alone.
Adjustments for Variation in Test Difficulty
All test forms of the LSAT reported on the same score scale
are designed to measure the same abilities, but one test form
may be slightly easier or more difficult than another. The
scores from different test forms are made comparable
through a statistical procedure known as equating. As a result
of equating, a given scaled score earned on different test
forms reflects the same level of ability.