College and Research Libraries


Demand-Adjusted Shelf 
Availability Parameters: 

A Second Look 

Philip Schwarz 
The application of Kantor's demand-adjusted shelf availability model to a medium-sized aca-
demic library is described. The model can be applied in a working library environment with 
relative ease. The data indicate that there are significant differences in shelf availability when 
the data are sorted, adjusted, and analyzed by last circulation date, acquisition date, and im-
print date. There is also a significant difference between the results of data gathered during 
periods of low and high use. 

his research report is the sec-
ond in a series. The papers 
describe the use of scientific 
management techniques to 

evaluate and describe various library op-
erations and services, in order that mana-
gerial decisions may be made on a rational 
rather than an intuitional basis. The pre-
vious publication in this series was: Philip 
Schwarz and Linda Olson, "Examination 
of Potential Management Decisions Based 
upon a Core Collection Derived from Last 
Circulation Date Data" (Menomonie, 
Wis. Research Report no.1. U.S., Educa-
tional Resources Information Center, 
ERIC Document ED 214 496, Aug. 1982). 

BACKGROUND 

In an article appearing in the May 1981 
Journal of Academic Librarianship, Paul 
Kantor described a simple theoretical 
model for determining shelf availability 
for library materials. 1 The purpose of this 
paper is to apply this model to a working 
library environment and in the process ex-
amine several additional considerations 
not discussed in the Kantor article . These 
include: (1) gathering data on the time re-

quired to apply the model developed by 
Kantor to a working library environment; 
(2) determining demand-adjusted shelf 
availability for a medium-sized university 
library; (3) determining whether there is 
any significant difference between the 
results of data expostulated in what 
Kantor describes as a naive fashion, and 
adjusted data sorted by last circulation 
date, acquisition date, and imprint date; 
and (4) determine the degree of difference 
between stack availability as recorded 
during the initial weeks of the semester 
when demand for materials is low as com-
pared to the latter weeks of a semester 
when material is in heavy demand. 

Historically, two approaches have been 
used to determine shelf availability in li-
braries. Shelf availability, as used in this 
paper, is the probability that a patron go-
ing to the shelf will find the item he is look-
ing for. One approach to this problem is 
the collection of data based on expressed 
demand and described in papers by Buck-
land, Kantor, and others .2 Using this tech-
nique, demand as expressed by users is 
measured by actually surveying library 
users. The user is handed a form or a sur-

Ph ilip Schwarz is special assista nt for automation development, University of Wi sconsin- Stout, Men omonie, 
Wisconsin . 

Th e author wishes to thank Dr. Paul Ka ntor for commenting on the data . 

210 



vey worker accompanies the patron 
around the library and determines the 
number of items found and the number of 
items not found. For those items not 
found, data are gathered to determine the 
reasons why they are not found. Using 
this technique, one can identify the poten-
tial impact of the various ways a user can 
be frustrated in his search for library mate-
rials. Sources of frustration in order of log-
ical occurrence are: (1) collection develop-
ment failure-the library has never 
acquired the item desired by the patron; 
(2) the patron does not have the necessary 
skills to use the catalog successfully; (3) 
the item is checked out; (4) the item is 
missing from its appropriate location on 
the shelf; (5) and lastly, the item is on the 
shelf in its proper location but for some 
reason the patron cannot locate it. The 
overall document availability is the prod-
uct of all of these factors. Although pro-
viding a wide range of useful manage-
ment data, this technique requires 
considerable effort to administer. 

A simpler technique was introduced by 
Kaske and elaborated on by Altman and 
de Prospo. 3 This technique utilizes a small 
sample drawn from the shelflist . Items in 
the sample are checked against the stacks 
and circulation records to determine the 
percentage of items not found. This ap-
proach provides less management infor-
mation than the first technique described. 
One can only determine if the item is in 
circulation or if it is missing from its 
proper shelf location. It does not provide 
information regarding the adequacy of 
collection development policies, patron 
skills in using the catalog, or patron skills 
in locating materials in the collection. In 
addition, as Kantor points out, the data 
which this technique provides on circula-
tion interference and on "other" factors 
are subject to inherent bias because of the 
failure to adjust for the fact that not all ma-
terials are equally in demand . 4 The impor-
tance of this fact could have a significant 
impact upon the findings when using this 
technique. For example, libraries with 
very old and large collections are likely to 
find that the collection extends far beyond 
the interest of the current users. As a 
result, data gathered using this technique 

Shelf Availability Parameters 211 

are likely to overestimate the probability 
that an item, in the relatively small subset 
of materials currently in demand, will be 
found on the shelf. The items in high de-
mand are precisely the ones that are likely 
to be in circulation or not available for cir-
culation for some reason. It is this issue 
that this paper is intended to address. 

METHODOLOGY 

Several points are worth noting in con-
nection with this study. The author was 
able to conduct the comparative analysis 
of data involving last circulation date, ac-
quisition date, and imprint date because 
the library utilized a circulation system 
that retained information regarding item 
circulation activity. It is also worth noting 
that all library users are limited to a 
twenty-eight-day circulation period. This 
may be important if other libraries intend 
to compare their findings with data pre-
sented in this study. 

The first phase of the study involved the 
selection of a random sample of 504 items 
drawn from a total population of 141,000. 
The random sample was created using a 
standard computer random number gen-
erator program. The numbers, once gen-
erated, were sorted into numerical se-
quence to facilitate matching against the 
numerical sequence of the shelflist 
drawers. Once this was completed, the 
survey worker went to the shelflist to 
gather the sample. The survey worker 
opened the appropriate drawer and laid a 
ruler alongside the cards. A second set of 
random numbers was used to select the 
card or cards in each drawer correspond-
ing to the number of samples to be drawn 
from the drawer. For example, if two sam-
ples were to be drawn from a drawer and 
the random number table indicated they 
should be drawn from one and ten inches, 
slips were inserted in the shelflist at these 
points. The call number, imprint date, and 
the date of acquisition for each sample 
were recorded on the data collection form 
shown in figure 1. If the card happened to 
be for an item with multiple volumes, a 
second random number table was used to 
select the volume number to be recorded 
on the data collection form. 

The second phase of the study involved 



212 College & Research Libraries July 1983 

1=STACI<S 
2=CIRC 
3=0THER 

SAI"1PLE 
NUMBER 

LAST 
IMPRINT ACQUISITION CIRCULATION 

CALL NUI"1BER DATE DATE DATE CODE 

1 AC 8 .D56 1968 1968 02-26-68 09-26-68 
2 AS 4 .U83L3 1957 07-27-81 12 - 18-81 2 
3 B 851 .B7 1969 1967 07-05-73 07-05-73 3 
4 
'S 

FIGURE 1 
Data Collection Form 

checking on the status of each item in the 
sample. Kantor suggests starting with the 
circulation records. 5 However, the author 
suggests that it would be more efficient to 
begin in the stacks. A quick calculation of 
the number of items in circulation indi-
cated that only a small percentage of the 
collection (approximately 6 percent) 
would likely be in circulation at any one 
time. As a result, the survey worker first 
went to the stacks and searched for each 
item listed on the data collection form. If 
the item was located, the last circulation 
date and the disposition code (in this case 
a 1 for stacks) was recorded in the appro-
priate columns on the data collection 
form. If the item had not circulated, the ac-
quisition date was duplicated in the last 
circulation date column on the data collec-
tion form. 

When the survey worker neared com-
pletion of his shift, he stopped searching 
in the stacks and went to the circulation 
desk to search the circulation records for 
items not located in the stack search. If the 
item was located in the circulation rec-
ords, indicating that it was not available 
for circulation, the previous (next to last) 
circulation date was recorded in the last 
circulation date column on the date collec-
tion form. If this date was not available, 
the date of acquisition was recorded in the 
last circulation date column. A code (2) 
was recorded in the disposition column on 
the data collection form indicating the 
item was in circulation. 

In cases where an item was not located, 
a 0 was recorded in the disposition 
column. Items not located were later 
searched for on several occasions. If 

found, the last circulation date was re-
corded in the appropriate column on the 
data collection form. If the last circulation 
date was not available, the date of acquisi-
tion was duplicated in the last circulation 
date column. Items falling into the 
''other'' category were either misshelved, 
in transit, in use within the library, miss-
ing, or incorrectly processed. No attempt 
was made to quantify this information al-
though it would be relatively easy to do 
and would be required if the library felt 
this category could be a significant factor 
in the availability analysis. 

The final step, prior to the data analysis, 
involved keypunching the sample num-
ber, imprint date, acquisition date, and 
last circulation date onto IBM cards for 
later sorting and tallying. The subsequent 
analysis of the data is discussed in the fol-
lowing section. 

ANALYSIS OF DATA 

One of the objectives of the study was to 
determine the personnel requirements for 
conducting a study of the type described 
in this paper. Table 1 provides this infor-
mation. As can be seen, the two most 
time-consuming aspects of a study of this 
type involve the selection of the sample 
from the shelflist and the time required to 
search for the material. There would be a 
slight savings in time if the survey worker 
had not opened each item to determine 
the last circulation date, however, the to-
tal savings in time would have been less 
than two hours. Information regarding 
the time required for data analysis was not 
included. It should be noted that it is not a 
time-consuming activity. An IBM Per-



Shelf Availability Parameters 213 

TABLE 1 
PERSONNEL REQUIREMENTS FOR CONDUCTING THE AVAILABILITY 

ANALYSIS 

Activit 

Developing strategy 
Selectin~ sample from the shelflist 
Developmg programming requirements 
Searchmg for the material 
Total hours 

sonal Computer and the VisiCalc software 
were utilized to produce the statistical 'ta-
bles. Development of the master tables 
and calculation of the data presented in 
the tables required only a few hours. 

A second objective of the study was to 
determine if there was any significant dif-
ference between the results of data expos-
tulated in what Kantor describes as naive 
fashion and adjusted data sorted by last 
circulation date, acquisition date, and im-
print date. The results of this analysis are 
shown in tables 2 through 5 and discussed 
below. 

Table 2 displays what Kantor refers to as 
naive data, that is, unadjusted data which 
assume that demand is distributed uni-
formly over the collection. The data 
shown in this table were gathered during 
a period of high circulation. The data are 
sorted by last circulation date and divided 
into three equal categories labeled young, 
middle, and elder. Within each category, 
material is identified as being in circula-
tion, on the shelf, or ''other.'' The young 
category represents material with a last 
circulation date within the last eighteen 
months. The middle category represents 
material with a last circulation date of be-
tween nineteen and seventy-two months. 
The elder, and final, category represents 
material that has not circulated within the 
last seventy-three months. 

The total number of items in the young 
category is found to be 168, of which 24 are 
circulating, 132 are on the shelves, and 12 
cannot be accounted for. There are 168 
items in the middle category of which 1 is 
circulating, 154 are on the shelves, and 13 
cannot be accounted for. The elder cate-
gory also contains 168 items of which none 
are circulating, 151 are on the shelves, and 
17 cannot be accounted for. It is clear from 
the table that most of the items circulating 
fall into the young category. The impor-

Time 

1.00 
6.25 
1.00 
9.15 

17.40 

tance of this fact will be examined in 
greater detail in the discussion of the next 
table. 

As table 2 also indicates, the circulation 
dysfunction, that is the failure rate result-
ing from an item being in circulation, for 
the naive analysis is 4. 9 percent. This was 
calculated by dividing the number of 
items in circulation (25) by the total num-
ber of items in the sample (504). The 
''other'' category dysfunction, that is the 
failure rate resulting from not being able to 
account for an item, is 8.7 percent. This 
was calculated by dividing the total num-
ber of items in the ''other'' category ( 42) 
by the difference between the total num-
ber of items in the sample (504) less the to-
tal number in circulation (25). The total 
stack dysfunction, that is the failure rate 
resulting from not being able to locate an 
item in the stacks, is 13.2 percent and is 
determined by adding the total number of 
items in circulation (25) to the total num-
ber of items in the ''other'' category ( 42) 
and dividing the resulting total by the total 
number of items in the sample (504). The 
stack availability is found by subtracting 
the stack dysfunction (13.2 percent) from 
100 percent which, in this case, is 86.7 per-
cent. That is, a patron going to the stacks 
looking for an item found in the public cat-
alog, could potentially find it on the shelf 
86.7 percent of the time. 

As mentioned earlier, this analysis as-
sumes that demand is distributed uni-
formly throughout the collection. An ear-
lier study by the author clearly showed 
that demand was not so distributed. 6 In 
fact, 84 percent of the circulation was be-
ing met by 33 percent of the collection. 
This fact is reflected in the data shown in 
table 2, where all but one item in circula-
tion falls into the young category. Kantor 
suggests that it is possible to adjust the 
data to account for the uneven distribu-



214 College & Research Libraries 

tion of demand. This adjustment is based 
upon a formula that relates the demand 
for an item to the chance that it is in circu-
lation. A more complete discussion of the 
formula and its derivation can be found in 
Kantor. 7 

Table 3 shows how the data can be ad-
justed to account for current demand. The 
raw or naive data are again shown on the 
left of the table. A weighting factor or ad-
justment factor is recorded in a new 
column following the raw data. This factor 
is derived by dividing the number of items 
in circulation for the row by the number of 
items on the shelf and in the II other'' 
column. This calculation is designed to 
correct for the fact that the demand for an 
item actually exceeds its circulation be-
cause patrons will be looking for it even 
though it is not available. Using the young 
grouping as an example, the number of 
items in circulation (24) is divided by the 
total of the number of items on the shelf 
(132) plus the number in the II other'' cate-
gory (12). The result of this calculation is a 
weight factor for the row of .1666667. This 
calculation is carried out for each grouping 
and row in the table. Once these calcula-
tions have been completed, the raw data 
are multiplied by the weight factor to pro-
vide the adjusted data. Continuing with 
the example of the young circulation cate-
gory used previously, we find the number 
of items in circulation (24) is multiplied by 
the weight factor ( .1666667) to find the ad-
justed data (4.00000). This calculation is 
repeated for each category in each of the 
adjusted rows. 

The same calculations that were carried 
out in table 2 to find circulation availabil-
ity, "other" availability, and stack avail-
ability are repeated using the adjusted 
data. In this case the circulation dysfunc-
tion is calculated by dividing the adjusted 
circulation total (4.005988) by the adjusted 
total of the items in the sample (4.005988 
+ 22.92216 + 2.077844 = 29.005992). The 
"other" dysfunction (7 .1 percent) is calcu-
lated by dividing the adjusted "other" 
category (2.077844) by the difference be-
tween the adjusted total number of items 
in the sample (29.005992) less the total 
number in circulation (4.005988). The total 
stack dysfunction (20.9 percent) is calcu-
lated by dividing the adjusted total of the 

July 1983 

in circulation and "other" columns 
(4.005988 + 2.077844) by the adjusted total 
of the items in the sample (29.005992). The 
stack availability (79 percent) is calculated 
by subtracting the stack dysfunction (20.9 
percent) from 100 percent. 

The same procedures described above 
were used to carry out the analysis in ta-
bles 4 and 5. The only difference was in the 
method of sorting the data. In the case of 
table 4, data were sorted by acquisition 
date and in the case of table 5, by imprint 
date. In each of these cases the circulation 
was more evenly distributed throughout 
the three categories although, in each 
case, almost one-half of the circulation 
was in the young category. 

A third objective of the study was to de-
termine the degree of difference between 
data gathered during a period when de-
mand was heavy and during a period 
when demand for library materials was 
light. The latter is shown in tables 6 
through 9. As can be seen and as might be 
expected, the data differ significantly be-
tween the two periods . 

CONCLUSIONS 

The time required to conduct a simple 
availability analysis of the type described 
in this paper is minimal. The majority of 
the time will be spent in selecting the sam-
ple and searching for the material. This 
approach would appear to be more effi-
cient than analysis based upon expressed 
demand, although it clearly provides less 
management information. 

It would also appear that the technique 
described in this paper is a good manage-
ment tool for monitoring library perfor-
mance in two areas: (1) circulation dys-
function and (2) library housekeeping 
dysfunction. It will provide the library 
manager with quantitative data to mea-
sure against previously established con-
trol standards. If, for example, availability 
falls outside acceptable limits, the man-
ager can take the necessary corrective 
action. In addition to acting as a flag to the 
manager that corrective action is required, 
it also will tell the manager if the corrective 
action has been successful. This is perhaps 
just as important as the aforementioned 
function. 

In the case of circulation dysfunction, 



Shelf Availability Parameters 215 

TABLE 2 
NAIVE RESULTS OF A SHELFLIST STUDY OF ITEM 

AVAILABILITY WITH DATA SORTED BY LAST 
CIRCULATION DATE 

Group 

Young 
Mid die 
Elder 
Totals 

In Cir-
culation 

24 
1 
0 

25 

Circulation availability in percent 
Circulation dysfunction in percent 
Other availability in percent 
Other dysfunction in percent 
Stack availability in percent 
Stack dysfunction in percent 

NOVEMBER 

Status 
On 

Shelf 

132 
154 
151 
437 

TABLE 3 

Other 

12 
13 
17 
42 

Total 

168 
168 
168 
504 

95.03968 
4.960317 
91 .23173 
8.768267 
86.70635 
13.29365 

WEIGHTED RESULTS OF A SHELFLIST STUDY OF ITEM AVAILABILITY WITH DATA 
SORTED BY LAST CIRCULATION DATE 

Raw (Naive) Data 
In Cir- On 

Group culation Shelf 

Youn~ 24 132 
Midde 1 154 
Elder 0 151 
Totals 25 437 

Circulation availability in percent: 
Circulation dysfunction in percent: 
Other availability in percent: 
Other dysfunction in percent:· 
Stack availability in percent: 
Stack dysfunction in percent: 

Other 

12 
13 
17 
42 

NOVEMBER 

Weighting 
Factor 

.1666667 

.0059880 
0 

TABLE 4 

Adjusted Data 
InCir- On 

culation Shelf 

4 .000000 
.0059880 

0 
4.005988 

22 .00000 
.9221557 

0 
22 .92216 

Other 

2.000000 
.0778443 

0 
2.077844 

86.18910 
13.81090 
92.83650 
7.163501 
79.02560 
20.97440 

WEIGHTED RESULTS OF A SHELFLIST STUDY OF ITEM AVAILABILITY WITH DATA 
SORTED BY ACQUISITION DATE 

Raw (Naive) Data 
lnCir- On 

Group culation Shelf 

Youn~ 13 139 
Midde 8 148 
Elder 4 150 
Totals 25 437 

Circulation availability in percent : 
Circulation dysfunction in percent: 
Other availability in percent: 
Other dysfunction in percent: 
Stack availability in percent: 
Stack dysfunction in percent: 

Other 

15 
14 
13 
42 

NOVEMBER 

Weighting 
Factor 

.0844156 

.0493827 

.0245399 

Adjusted Data 
InCir- On 

culation Shelf 

1.097403 
.3950617 
.0981595 

1.590624 

11 .73377 
7.308642 
3.680982 

22.72339 

Other 

1.266234 
.6913580 
.3190184 

2.276610 

94.01810 
5 .981897 
91.43830 
8 .561703 
85.45640 
14.54360 



216 College & Research Libraries July 1983 

TABLES 
WEIGHTED RESULTS OF A SHELFLIST STUDY OF ITEM AVAILABILITY WITH DATA 

SORTED BY IMPRINT DATE ATE 

Raw (Naive) Data 
ln Cir- On 

Group culation Shelf 

Youn~ 14 138 
Midde 7 150 
Elder 4 149 
Totals 25 437 

Circulation availability in percent: 
Circulation dysfunction in percent: 
Other availability in percent: 
Other dysfunction in percent: 
Stack availability in percent: 
Stack dysfunction in percent: 

Other 

19 
9 

14 
42 

NOVEMBER 

Weighting 
Factor 

.0891720 

.0440252 

.0245399 

TABLE6 

Adjusted Data 
lnCir- On 

culation Shelf 

1.248408 
.3081761 
.0981595 

1.654743 

12.30573 
6 .603774 
3.656442 

22.56595 

NAIVE RESULTS OF A SHELFLIST STUDY OF ITEM 
AVAILABILITY WITH DATA SORTED BY LAST 

CIRCULATION DATE 
SEPTEMBER 

Status 
InCir- On 

GrouE culation Shelf Other 

Youn~ 1 152 14 
Midde 2 166 4 
Elder 5 156 4 
Totals 8 474 22 

Circulation availability in percent 
Circulation dysfunction in percent 
Other availability in percent 
Other dysfunction in percent 
Stack availability in percent 
Stack dysfunction in percent 

TABLE7 

Other 

1.694268 
.3962264 
.3435583 

2.434052 

93.79194 
6.208063 
90.86822 
9.131779 
84.66016 
15.33984 

Total 

168 
168 
168 
504 

95 .03968 
4.960317 
95.56452 
4.435484 
94.04762 
5.952381 

WEIGHTED RESULTS OF A SHELFLIST STUDY OF ITEM AVAILABILITY WITH DATA 
SORTED BY LAST CIRCULATION DATE 

Raw (Naive) Data 
lnCir- On 

GrouE culation Shelf 

Youn~ 1 152 
Midde 2 166 
Elder 5 156 
Totals 8 474 

Circulation availability in percent: 
Circulation dysfunction in percent: 
Other availability in percent: 
Other dysfunction in percent: 
Stack availability in percent: 
Stack dysfunction in percent: 

Other 

14 
4 
4 

22 

SEPTEMBER 

Weighting 
Factor 

.0060241 

.0117647 

.03125 

Adjusted Data 
InCir- On 

culation Shelf 

.0060241 

.0235294 

.15625 

.1858035 

.9156627 
1.952941 
4.875 
7 .743604 

Other 

.0843373 

.0470588 

.125 

.2563962 

97.73017 
2 .269826 
96.86779 
3 .132205 
94.59797 
5.402031 



Shelf Availability Parameters 217 

TABLE 8 
WEIGHTED RESULTS OF A SHELFLIST STUDY OF ITEM AVAILABILITY WITH DATA 

SORTED BY ACQUISITION DATE 
SEPTEMBER 

Raw (Naive) Data 
InCir- On 

Group culation Shelf Other 

Youn~ 1 160 6 
Midde 5 159 2 
Elder 2 155 10 
Totals 8 474 18 

Circulation availability in percent: 
Circulation dysfunction in percent: 
Other availability in percent: 
Other dysfunction in percent: 
Stack availability in percent: 
Stack dysfunction in percent: 

the corrective action might involve a re-
duction in loan periods. 8 The reduction 
could encompass all materials or it could 
be selective, applying ohly to those mate-
rials in high demand. A second approach 
might be the purchase of more duplicate 
volumes for high-demand items. A third 
option would be to use a combination of 
the two approaches. 9 

In the case of library housekeeping dys-
function, the manager may want to assign 
more personnel to shelving or shelf read-
ing. The results might also indicate a need 
to replace missing materials in a more 
timely fashion. The technique described 
in this paper would appear particularly 
useful if conducted on a periodic basis so 
that comparative data would be available 
over time. Data collected in this manner 
might be a useful tool in assisting the li-
brary manager in securing additional 
staffing or maintaining the same levels of 
staffing by providing specific data to li-
brary and university administrators re-
garding the impact of staff reductions, in 
selected areas, on patron success in locat-
ing materials in the library. 

It is also clear from the analysis that the 
adjustment for uneven distribution of de-
mand can make a significant difference in 
the findings regarding availability if the 
data are sorted by last circulation date. In a 
working environment, it would appear 
that the analysis of data when sorted by 
acquisition date and imprint date more 
closely approximates that employed in the 
naive analysis. The key question is which 
analysis (naive, last circulation date, ac-

Weighting 
Factor 

.0060241 

.0310559 

.0121212 

Adjusted Data 
lnCir- On 

culation Shelf 

.0060241 

.1552795 

.0242424 

.1855460 

.9638554 
4.937888 
1.878788 
7.780531 

Other 

.0361446 

.0621118 

.1212121 

.2194685 

97.73325 
2.266752 
97.31883 
2.681171 
94.91576 
5.084240 

quisition date, or imprint date) most 
closely approximates a true picture of 
stack availability? Table 10 offers a com-
parison of the availability analysis for the 
four approaches. Since the problem is to 
correct for the effects of actual use, the au-
thor would hypothesize that adjusted 
data sorted by last circulation date would 
provide the best approximation of actual 
availability. It would be useful if a library 
were to conduct an expressed demand 
analysis in conjunction with a study simi-
lar to that described in this paper. Such a 
study would help answer this question. 

It would also appear likely that manage-
ment could easily make several erroneous 
assumptions if the results of naive data 
analysis were used. Two problems are evi-
dent: (1) the availability is overstated and 
(2) the major cause of the dysfunction is 
incorrectly identified. As an example, in 
the latter situation, if a manager wished to 
increase stack availability based upon the 
data presented in table 2, naive data, he or 
she would most likely commit the library 
resources to reducing the number of items 
in the II other'' category. This would be 
the logical assumption because this cate-
gory represents a larger proportion of the 
dysfunction than does the circulation cate-
gory. However, when the manager exam-
ines the adjusted data sorted by last circu-
lation date presented in table 3, it becomes 
clear that the major source of stack dys-
function and, hence, user frustration is 
the fact that materials are in circulation. 

As shown in the data, there is a signifi-
cant difference between the results of data 



218 College & Research Libraries July 1983 

TABLE9 
WEIGHTED RESULTS OF A SHELFLIST STUDY OF ITEM AVAILABILITY WITH DATA 

SORTED BY IMPRINT DATE 

Raw (Naive) Data 
InCir- On 

Group culation Shelf 

Youn~ 0 159 
Midde 4 158 
Elder 4 157 
Totals 8 474 

Other 

8 
4 

10 
22 

SEPTEMBER 

Weighting 
Factor 

0 
.0246914 
.0239521 

Adjusted Data 
InCir- On 

culation Shelf 

0 
.0987654 
.0958084 
.1945738 

0 
3.901235 
3.760479 
7.661714 

Other 

0 
.0987654 
.2395210 
.3382864 

Circulation availability in percent: 
Circulation dysfunction in percent: 
Other availability in percent: 

97.62558 
2.374423 
95.87182 
4.128176 
93.49740 
6.502598 

Other dysfunction in percent: 
Stack availability in percent: 
Stack dysfunction in percent: 

TABLE 10 
SUMMARY TABLE SHOWING ITEM AVAILABILITY DURING A HIGH CIRCULATION 

PERIOD 

Last 
Circula- Acquisi- Imprint 

Naive tion Date tion Date Date 

Circulation availability 95.03 86.18 94.01 93.79 
Circulation dysfunction 4 .96 13.81 5.98 6.20 
Other availability 91 .23 92.83 91.43 90.96 
Other dysfunction 8.76 7.16 8.56 9.13 
Stack availability 86.70 79.02 85.45 84 .66 
Stack dysfunction 13.29 20.97 14.54 15.33 

TABLE 11 
WEIGHTED RESULTS OF A SHELFLIST STUDY OF ITEM AVAILABILITY WITH DATA 

SORTED BY LAST CIRCULATION DATE 
NOVEMBER 

Raw (Naive) Data Adjusted Data 
InCir- On Weighting InCir- On 

GrouE culation Shelf Other Factor culation Shelf 

Youn~ 24 139 12 .1666667 4.000000 22 .00000 
Midde 1 154 13 .0059880 .0059880 .9221557 
Elder 0 151 17 0 0 0 
Totals 25 437 42 4.005988 22.92216 

Circulation availability in percent: 
Circulation dysfunction in percent: 
Other availability in percent: 
Other dysfunction in percent: 
Stack availability in percent : 
Stack dysfunction in percent: 

Other 

2 .000000 
.0778443 

0 0 
2.077844 

Error Analxsis 
Factor Factor 
One Two 

.0373333 .0010468 

.0015999 2.547£-5 
0 

.0010722 

86.18910 
13.81090 
91.68862 
8.311377 
79 .02560 
20.97440 

Error Squared: 1.716E-6 
Error: .0013098 



gathered during periods of low use and 
high use. The question naturally arises 
during which period should the study be 
conducted? Since most of the demand and 
hence the circulation and library house-
keeping dysfunction will occur during pe-
riods of high demand, the author feels this 
would be the most appropriate time to 
conduct the study. 

ERROR ANALYSIS 

Kantor provides the following formula 
as a means of establishing a rough esti-
mate of the standard error. 10 

E"2 = W(1 - W)AIN"2 • ((2 - W)W-
(1 - P))"2 

The standard error is calculated for each 
row within each group. In the formula, W 
= the weight factor for the row, A = the 
row sum for the raw data, N = the total 
number of circulations for all groups, and 
P = the adjusted circulation availability. 

Shelf Availability Parameters 219 

Table 11 illustrates the use of the stan-
dard error calculation. Using the row for 
the young group as an example, the data 
show that W = .1666667, A = 168 (24 + 
132 + 12), N = 25, and P = .8618910. In 
the table factor one is calculated as fol-
lows: W*(1 - W)* AI(N*N) and recorded 
in the next to the last column for the row as 
.0373333. The product for the last column 
is calculated using the formula ((2 - W)* 
W- (1 - p))"2 times the product from the 
previous column and results in a product 
of .0010468. This process is repeated for 
each row. As can clearly be seen, the cal-
culation for the first row has the greatest 
impact upon the error calculations. 

The complete error formula utilizes the 
products shown in the last column for 
each row. E"2 = (P1 + P2 + P3)/(N*N). In 
the case of table 10, this is E"2 = (.0010468 
+ .00002547 + 0)/(25*25) orE = 1716E -
6. 

REFERENCES 

1. P . B. Kantor, "Demand-Adjusted Shelf Availability Parameters," Journal of Academic Librarianship 
7:78-82 (May 1981) . 

2. M . K. Buckland and others, Book Availability and the Library User (New York: Pergamon, 1978); J. A. 
Urquhart and J. L. Schofield, ''Measuring Readers Failure at the Shelf,'' Journal of Documentation 
27:273-86 (Dec. 1971); P. B. Kantor, "Availability Analysis," Journal of the American Society for Infor-
mation Science 27:311-19 (Sept. 1976); T. Saracevic; W. Shaw, Jr.; and P. B. Kantor, "Causes and 
Dynamics of User Satisfaction in an Academic Library," College & Research Libraries 36:7-18 (Jan. 
1977); J. B. Whitlatch and K. Kieffer, "Service at San Jose State University: A Survey of Document 
Availability," Journal of Academic Librarianship4 :196-99 (Sept. 1978); E. Palais, "Availability Analy-
sis Report," in Association of Reserach Libraries, Office of Management Studies, User Suroeys and 
Evaluation of Library Seroices (Washington, D.C.: ACRL, 1981), p .73-82; R. Smith and W. Granade, 
''Undergraduate Library Availability Study 1975-1977, '' in Association of Research Libraries, Of-
fice of Management Studies, User Suroeys and Evaluation of Library Seroices (Washington, D.C.: 
ACRL, 1981), p.83-90. 

3. N . K. Kaske, "Effectiveness of Library Operations : A Management Information Systems Ap-
proach and Application" (Library Science dissertation, University of Oklahoma, 1973); E. Altman 
and others, A Data Gathering and Instructional Manual for Performance Measures in Public Libraries 
(Chicago: Celadon Press, 1976). 

4. Kantor, "I?emand-Adjusted Shelf Availability Parameters," p.78-82. 
5. Ibid., p .80 . 
6. P . Schwarz and L. Olson, Examination of Potential Management Decisions Based upon a Core Collection 

Derived from Last Circulation Date Data (ERIC Document 214 496). 
7. Kantor, "Demand-Adjusted Shelf Availability Parameters," p.82. 
8. Buckland, Book Availability. 
9. E. Palais, "Increasing the Effectiveness of a University Library: Evaluation of an Added Copy 

Project," in Options for the 80s Program of th e Association of College and Research Libraries 2nd Annual 
Conference, October 1-4, 1981 (Minneapolis, Minn.: ACRL, 1981). 

10. Kantor, " Demand-Adjusted Shelf Availability Parameters," p.82.