Audiovisual Materials in U.S. Public Libraries

 

Douglas Galbi

June 9, 2007

 

 

** Rough Analysis Ė Comments and Suggestions Welcomed **

Version 1.1

 

 

Public libraries have long held items other than printed texts.† Audiovisual materials have included vinyl records, audio CDís, film (8mm, 16mm), video cassette tapes (Betamax and VHS), and video DVDís.† Audiovisual materials as a percent of the number of books (volumes) in librariesí collections have increased from about 3.5% in 1987 to 9.5% in 2004, with the share of videos growing much faster than that of audios.† Estimates suggest that audiovisual materials have a much higher circulation rate than books.† These estimates indicate that videos account for about 20% of libraries circulation in 2004.† Audio and visual materials together (audiovisuals) account for about 35% of circulation.† †While these estimates are crude, they are consistent with much narrower survey data.† The relative popularity of audio and video is also consistent with a variety of other evidence from the communications industry.

 

I. Share of Audiovisual Materials over Time

 

Table 1

Audio and Video Materials

as Percent of Book Volumes in U.S. Public Libraries

 

Year

Audio

Video

Audio and Video

Reporting Libs.

1987

2.9%

0.6%

3.5%

2,580

1988

3.2%

0.5%

3.8%

6,313

1989

2.8%

0.6%

3.4%

7,644

1990

2.9%

0.7%

3.6%

7,930

1991

3.2%

0.9%

4.1%

8,614

1992

3.2%

1.1%

4.3%

8,570

1993

3.2%

1.3%

4.5%

8,552

1994

3.4%

1.4%

4.8%

8,575

1995

3.6%

1.7%

5.3%

8,393

1996

3.5%

1.9%

5.4%

8,480

1997

3.7%

2.1%

5.8%

8,628

1998

3.7%

2.5%

6.2%

8,676

1999

3.9%

2.7%

6.6%

8,777

2000

4.2%

3.0%

7.2%

8,816

2001

4.4%

3.4%

7.8%

8,693

2002

4.6%

3.8%

8.4%

8,884

2003

4.6%

4.2%

8.8%

8,924

2004

4.9%

4.6%

9.5%

8,942

 

Table 1 Sources and Notes

 

All the figures are calculated based on the ratios for individual libraries reporting audio, video, and book holdings.† The individual library data are weighted by the population of the librariesí serving areas.† Libraries reporting an audiovisuals share greater than 50% are excluded as data errors or highly specialized libraries.

 

Calculated from Robert E. Molyneuxís Longitudinal Public Library Data.† This dataset matches libraries across public library censuses from the U.S. National Center for Education Statistics.† I am grateful to Dr. Molyneux for compiling public data, adding considerable value to it by matching libraries and resolving a wide variety of data issues, and making the resulting dataset publicly available.

 

 

II. Circulation of Audiovisual Materials

 

The library statistics do not separate circulation of books from circulation of audiovisual materials.† However, variation in circulation and the share of audiovisual materials provides information on the circulation of audiovisual materials.† I analyze the correlation between circulation and audiovisual share across libraries in 2002.

 

The behavior considered is library usersí borrowing of library items.† The model is that a userís borrowing is linearly related to the logarithm of number of books in a library for which the user is part of the population served and to the logarithm of the ratio of videos to books in the library.† Many factors affect usersí borrowing behavior. I focus on a simple model of collection size and collection composition to provide a simple description of correlations between statistics of interest.† That description is meant to support rough inferences about user behavior, not structural estimates of borrowing behavior.

 

A libraryís circulation statistics are interpreted as aggregates of borrowing behavior in the population that the library serves.† Thus I analyze circulation per capita using population-weighted statistics.† That weighting is significant because populations that libraries serve span a wide range: the ratio of the 95íth percentile to the 5íth percentile in the population served distribution is about 200.

 

Table 2 describes the distribution of the relevant statistics from the 2002 census of public libraries.

 

Table 2

U.S. Public Library Statistics, 2002

(all statistics population weighted)

 

 

adult circ per capita

children's circ per capita

books

in lib

Videos

(% of books)

pop. quartile 1

2.2

1.2

103,913

2.4%

pop. mean

4.3

2.4

1,310,626

3.9%

pop. quartile 3

5.6

3.1

1,299,166

4.9%

 

Table 2 Sources and Notes:

 

These data are directly from the NCES 2002 data file.†† Because the model uses logarithms of video share, libraries that report zero videos (240) have been excluded.† Libraries reporting greater than 50% video share, libraries not reporting book volume, and libraries reporting zero adult or zero childrenís circulation have also been excluded.† These statistics† encompass 8885 of 9141 census records.

 

 

 

Table 3

Circulation Model Coefficients for Video Share

(for logarithms of variables)

 

 

adult circ. coefficients

children's circ. Coefficients

estimation type

total books

video/

books

total books

video/

books

least-squares

0.33

1.09

0.12

0.59

pop-weighted

0.08

1.53

-0.02

0.83

quantile

0.41

0.91

0.14

0.55

robust

0.42

0.83

0.18

0.49

 

 

Estimated model coefficients indicate that the video/book ratio significantly affects circulation.† The model includes a constant (not reported above).† Standard errors on the coefficient estimates were typically .02 to .05.† Different estimation procedures produced much greater variation in coefficient estimates.†

 

In a model of individual usersí borrowing behavior, the large range in library size produces large heteroskedasticity in unmodeled variation.† The population-weighted regression accounts for this heteroskedasticity.† Pure heteroskedasticity affects only the efficiency of estimates, not coefficient values.†† However, different types of libraries and different degrees of geographic association among libraries produce complex correlations in the unmodeled variation.† I suspect that these correlations along with the heteroskedasticity generated different coefficient estimates with different estimation procedures.†

 

One way of interpreting the size of the coefficients is to calculate the magnitude of change in circulation associated with the interquartile range of the variables in the model.† Under all the estimation procedures, the effect for the video/book share was roughly as large or larger than the effect for total book collection size (see Table 4).† This table is based on the least-square estimates in Table 3, which have intermediate values among those of the estimation procedures. †The population-weighted estimates, however, most directly fit my simple model for the data.† Those estimates produced the largest effects.

 

Table 4

Magnitude of Interquartile Effects

(for least-square estimates in Table 3)

 

 

effect on adult circ.

effect on children's circ.

 

books in lib

videos/

books

books in lib

videos/

books

effect q1-q3

0.37

0.35

0.13

0.19

% mean circ.

8%

8%

5%

8%

 

 

Another way of interpreting the size of the coefficients is to use them to evaluate changes in circulation and video/book ratios over time.† Circulation per capita has risen along with rises in the video/book ratio (see Tables 1 and 5).† Based on the (unweighted) least-squares estimates in Table 3, the rise in the video/book ratio from 1992 to 2004 accounts for 15% and 14% of adultís and childrenís items circulation in 2004.†

 

The similarity in effect on adultís and childrenís circulation may indicate a common pattern of behavior in choosing between books and videos.† This is similar to time spent with media, where both children and adults spend much more time watching television than reading books.† Alternatively, the similarity between adultís and childrenís circulation effects may indicate that the estimates are largely driven by common, unmodeled differences in library characteristics.† Suppose, for example, libraries that engage in a variety of efforts to better serve users needs and encourage borrowing have a higher video/book share.† Then the video/book share effect is a library effect, rather than a user behavior effect.† The unmodeled library effect probably exists and partly accounts for the different estimates across the different estimate procedures (different implicit models).† At the same time, the gross similarities of the estimates and other evidence on artifacts of different sensory forms also suggests a significant user behavior effect.[1]

 

 

 

Table 5

Circulation Per Capita

 

year

childrenís items

adultís items

total

circ.

1987

 

 

4.1

1988

 

 

4.7

1989

 

 

5.4

1990

 

 

5.6

1991

 

 

5.8

1992

2.2

4.0

6.1

1993

2.1

3.9

6.1

1994

2.2

3.9

6.2

1995

2.2

4.0

6.2

1996

2.2

4.1

6.3

1997

2.3

4.1

6.4

1998

2.3

4.0

6.4

1999

2.3

4.0

6.2

2000

2.3

4.0

6.3

2001

2.3

4.0

6.4

2002

2.4

4.3

6.8

2003

2.5

4.4

6.9

2004

2.5

4.5

7.0

 

Notes and Sources for Table 5

 

Statistics refer to circulation divided by the population in the libraryís service area.† The circulation per capital for each library is weighted by service area population.†† The source is as described in the notes for Table 1.

 

The model can also be used to estimate the total contribution of videos to circulation.† The video/book share at 4.6% in 2004 and the middling coefficient estimates (least-squares in Table 5) imply that videos contributed about 20% to both adult and childrenís circulation.† Because these estimates push the model across a wider range of collection compositions, these estimates are less reliable.†

 

Modeling audio collection shares in the same way as video shares produces similar but smaller results.† Given that the estimates are crude and simple descriptions are more helpful, I report results for audio and video combined (audiovideo materials).†† Table 6 gives the audiovisual model coefficients.† The overall estimated effects for audiovisual collection shares, using the middling least-square coefficient estimates, are about 35% for adultís and childrenís circulation.

 

 

Table 6

Circulation Model Coefficients for Audiovisual Share

(for logarithms of variables)

 

 

adult circ. coefficients

children's circ. Coefficients

estimation type

total books

av/

books

total books

av/

books

Least-squares

0.18

1.45

0.04

0.77

pop-weighted

0.00

1.82

0.00

0.88

quantile

0.29

1.24

0.07

0.73

robust

0.31

1.10

0.11

0.65

 

 

The significant share of audiovisuals in circulation is consistent with survey evidence. Sarah Ann Long, a former president of the American Library Association and currently director of NSLS, a library consortium in the northern suburban region of Illinois, recently noted public library users' interest in audiovisual materials:

In 2001, the NSLS conducted an informal survey of member public libraries and found that in a few libraries, loans of AV materials were about 40 percent of all loans. The same survey was just repeated and the numbers have grown. Many libraries now report that AV borrowing is in the 40 percent range. The Gail Borden Public Library in Elgin said that almost 57 percent of their loans were for AV materials and the Glencoe Public Library reported that AV accounted for 63 percent of all items borrowed.[2]

National library census statistics provide indirect evidence for a similar pattern of borrowing behavior.

 



[1] One might try to disentangle these effects through panel estimation techniques.† That would make a good thesis project for a graduate student in economics with a concentration in econometrics.† But a better practical approach to the problem would be to survey libraries about audiovisual circulation.

[2] Sarah Ann Long, ďPublic preference shows changing face of libraries,Ē The Daily Herald, March 17, 2007, available at http://www.sarahlong.org/ourlibraries/read/index.php?articleID=300