|
Measure
3.2.2: Random Jury Selection Procedures
This
measure determines whether a court is using random selection to select
prospective jurors from the juror source list(s). Data are obtained by
comparing actual prospective juror panels with those that would be expected
if random selection was used.
Planning/Preparation.
Although courts may say that all names are considered for selection, some
statutes, rules, or jury plans specify that strata be observed. In such
cases, courts draw names to represent each strata equally or to represent
each strata according to some ratio. For instance, some jurisdictions draw
by district strata so that the number of names selected from each district
is in proportion to the population of the district as compared with the
population of the whole jurisdiction. If the source list equally represented
each district, a random selection would equally represent each district, to
within a small margin of error. These stratified selections are intended to
overcome any unequal representation in the source list or lists. However,
before applying such techniques, courts should ensure that they are allowed
by some authority.
Measures
of randomness can be very complex.11 For this
measurement, it is recommended that courts compare several observations with
expected values. Although deviations from expectations are in some cases
proof of randomness, persistent patterns of nonexpected results should
require investigation. For instance:
-
A
panel of 30 prospective jurors, all male, is expected to occur in every
billion panels. One occurrence is reason for great amazement; two
occurrences should provoke great concern.
-
Although
the alphabet has never been shown to produce a bias, a group of prospective
jurors in alphabetical order, or representing only a portion of the
alphabet, raises questions of inclusiveness or discretion.
-
A
potential jury pool consisting of more than one individual with the same
last name or the same address can be expected to occur occasionally but
should be checked if occurring regularly.
-
The
same people often are called for jury service year after year or several
times within the same year. Repeat selections are expected. If 10 percent of
the list is selected each year, 1 percent will be selected in 2 successive
years and .1 percent will be selected 3 successive years. Values greater
than this need to be investigated.
Data
Collection.
This measure is conducted by examining the list of persons reporting for
jury service. These persons may be the entire pool of prospective jurors or,
if persons are brought to the court in panels, a number of panels could be
examined. Several hundred names should be adequate for these examinations.
If
suspicious patterns are found, persons reporting in at other court or jury
terms should be examined. If the patterns persist, problems clearly exist.
If the patterns are related to the date of service, problems likewise may
exist. Patterns to examine are:
-
Alphabetical
distribution. Half of the last names should be grouped A through K.
Deviations of more than a few percent should be investigated to examine the
alphabetical distribution of the source list or lists.
-
Alphabetical
inclusiveness. The last names of those serving should represent the
entire alphabet. Omissions of the top or bottom of the alphabet should be
examined because such omissions would indicate that the whole list was not
used. Panels of persons whose last names contain only a portion of the
alphabet are probably being called in via a recording that identified
individuals to report by last name. This practice should be replaced with
one that uses random numbers to select individuals.
-
Geographical
distribution. To the extent possible, the panels or pool of prospective
jurors should represent the entire jurisdiction. Lack of representation for
a distinct area of the jurisdiction could indicate that a geographical
listing such as the voter list is being used sequentially rather than by a
random selection from the entire list.
Data
Analysis and Report Preparation.
Nonrandom results are usually the result of the use rather than the generation
of random numbers. The problem is in how these numbers are used to select
names. If nonrandom results are discovered, detailed discussions with those
making selections (i.e., data processing or court staff) are needed. Factors
to examine include:
-
Are
the same key factors used for each selection?12
If a random start/fixed interval method is used, the start number must be
randomly selected in the range from one to the interval number. (If 100
names are desired from a list of 1,000 names, the interval is 10. If
"2" is randomly selected, the names at 2, 12, 22, 32, etc., in the
order are selected.)
-
If
a computer random number generator is used, are the input numbers or seeds
changed each time the program is run?
-
Are
names held out or passed over due to permanent exemptions or prior service?
If these names represent more than a few percent of the source list, this
could be the cause of the problem.
-
Are
the lists or files thought to be random actually sequential lists or files
by alphabet or geography? Voter registration lists are often geographically
separated by precinct, ward, or district. Lists ordered by voter
registration number may have an age order with older citizens having lower
voter numbers.
-
Do
the selected names represent the same list? If a printout of the voters list
or merged lists contains the same number of pages of "A’s" and
"W’s," the selected names should have equal numbers of
"A’s" and "W’s." The same list could be counted by
ZIP Code, and the distribution of those selected should match the
distribution of the source list. That is, if 10 percent of the names on the
source list has ZIP Code 22180, about 10 percent of those selected should
have that ZIP Code.
The
lack of proper numbers for certain demographic groups (e.g., young or black)
probably is due to the shortcomings of the source list rather than a problem
of randomness. This lack of representation is the topic of the next measure.
____________________________________
11
D.J. Knuth, The Art of Computer Programming, Semi-Numerical Algorithms,
vol. 2, 2d ed. (Reading, MA: Addison-Wesley, 1981).
12 National Center for State Courts, A
Supplement to the Methodology Manual for Jury Systems: Relationships to the
Standards Relating to Juror Use and Management (Williamsburg, VA,
1987), pp. 10-15.
Go
to Measure 3.2.3
Go
to Standard 3.2
Go
to Table of Contents |
