INFORMATION

ECONOMICS
AND POLICY
Information Economics and Policy 17 (2005) 175–198
www.elsevier.com/locate/econbase

Valuing the benefits of public libraries

Svanhild Aabø *

Faculty of Journalism, Library and Information Science, Oslo University College, P.O. Box 4,
St. Olavs Plass, N-0130 Oslo, Norway
Received 6 April 2004; received in revised form 15 April 2004; accepted 11 May 2004
Available online 20 July 2004

Abstract

Constraints on public budgets oblige libraries to document their value. This paper presents
a contingent valuation study eliciting how a random sample of Norwegian citizens values
public libraries, applying two recently developed elicitation approaches. Possible and actual
protest bids are differentiated and a split sample used, eliciting both willingness to pay (WTP)
and to accept (WTA). An overwhelming majority perceives they have property rights to a local
library, justifying the application of WTA. Estimates of WTA among non-protesters are
higher than estimates of WTP but only by a factor of about 4. Several measures of average
valuation are derived, all being higher than average costs.
Ó 2004 Elsevier B.V. All rights reserved.

JEL classification: H31; H41; H70

Keywords: Public goods; Public libraries; Contingent valuation

1. Introduction

By law, all Norwegian municipalities shall have a public library (Public Library
Act, 1985, Section 4). This article in the Public Library Act is subject to current
political discussions spurred by growing pressure on public budgets. In 2002, the
Ministry of Cultural and Church Affairs suggested an amendment of this article,
which was interpreted by many in the library profession and local authorities as a
step to loosen the municipality’s obligation to have a public library. Due to massive

*
Tel.: +47-2245-2694; fax: +47-2245-2605.
E-mail address: [email protected] (S. Aabø).

0167-6245/$ - see front matter Ó 2004 Elsevier B.V. All rights reserved.
doi:10.1016/j.infoecopol.2004.05.003
176 S. Aabø / Information Economics and Policy 17 (2005) 175–198

protests in the hearing the original proposal was withdrawn and only a minor
amendment was passed by the Parliament (Public Library Act 1985, 2003, Section 4).
The public libraries’ purpose is ‘‘to promote enlightenment, education and other
cultural activity by the dissemination of information and by making books and other
suitable material available free of charge to all those who live in Norway’’ (Section
1). This foundation, ‘the free-of-charge principle’, is also challenged and cause
controversy in the Norwegian society because many municipalities today face severe
budget restrictions. There is also discussion whether or not to enhance local au-
tonomy at the expense of national laws regulating local public services, thereby
possibly deviating from the equity principle and accepting greater differences in the
level of public services across municipalities. Public libraries are a topic in this
debate.
It is well known that a free market is likely to fail as a mechanism for allocating
public goods, including library services (Aabø, 1988; Kingma, 1996), but they can be
valued using methods for nonmarket valuation. An objective of this study is to make
such an assessment of public libraries in Norway. Two main approaches to valuation
of nonmarket goods exist in the literature, namely revealed and stated preferences
(Mitchell and Carson, 1989). Of these only stated preference (SP) methods were
considered useful here, since they are capable of capturing nonuse values in addition
to use values. Recent research has shown that the value people attach to public li-
braries has a variety of origins, including appreciation of other people’s use, dis-
semination of knowledge and culture, upholding of the literary heritage and reducing
of informational gaps (Aslib, 1995; Audunson, 2001). Contingent valuation (CV) is
by far the most widely used SP method. Within cultural economics there is a growing
CV literature (Noonan, 2003), including studies of theatres (Bille Hansen, 1997),
museums (Martin, 1994; Santagata and Signorello, 2000), cultural heritage (Ben-
hamou, 1996; Navrud and Ready, 2002), the arts (Throsby and Withers, 1983),
national television programming (Papandrea, 1999), and libraries (Harless and Al-
len, 1999; Holt et al., 1999).
This study appears to be the first CV study valuing public libraries at a national
level, in Norway or internationally. 1 The aim is to determine whether public li-
braries in Norway are, overall, worth their price as viewed from the population’s
perspective.
The study also has methodological aspects. Two elicitation formats are used re-
cently developed in environmental economics, which is at the fore in CV techniques.
To our knowledge, this is the first comparative test of these elicitation methods
valuing the same good. We stress the detecting of response uncertainty and error and
develop a procedure for identifying and differentiating possible and actual protest

1
National studies of public libraries and their value using other methods than CV have been conducted,
for instance in Britain (Morris et al., 2001), Australia (Mercer, 1995), and USA (D’Elia, 1993). CV was
used in cost-benefit analyses of several urban public libraries in USA (Holt et al., 1999; Holt and Elliott,
2003) and is proposed included in a follow-up study measuring the benefits and economic impacts from
public libraries in Florida (McClure et al., 2001, 7, p. 10).
 S. Aabø / Information Economics and Policy 17 (2005) 175–198 177

bids. A further important methodological feature is to apply both willingness to pay

(WTP) and willingness to accept (WTA), each used on half of the sample. Re-
searchers are cautious about applying WTA in CV studies due to possibilities of
implausibly high stated values. In our case however the property rights issue, as
discussed below, appears to make WTA imperative. Overall, this study gives clues to
whether CV, developed in economics, can contribute to the theoretical and meth-
odological arsenal of library and information science. 2
The rest of the paper is organized as follows. Section 2 presents the survey,
Sections 3 and 4 the WTP and WTA results, respectively, Section 5 protest bid
treatments, and Section 6 discussion and conclusions.

2. The survey

The survey was administered by a professional opinion company, ACNielsen

Norway AS, as part of their bimonthly omnibus survey (January 2000) that collects
data from a national random sample of private households. 999 persons over the age
of 15 were interviewed in their homes as representatives of their households.
Our CV study aims at measuring the total benefits accruing to the Norwegian
population from public libraries, at today’s activity and service levels. All 433 mu-
nicipalities of Norway have public libraries, comprising 1108 units. In 1997 they
owned 21.5 mill. document items and 22 mill. items were borrowed during that year.
Children on average borrowed 9.3 and adults 3.2 books, and each inhabitant on
average visited a public library 4.5 times during the year. Total operating costs were
803 mill. NOK, which is 420 NOK per household (Norske folkebibliotek, 1998). 3
Norwegian public libraries are widely used; 52% of all citizens visit a public library
during a year and, except for cinemas, this is the highest percentage of use of any
cultural institution. Only 6% of the population aged 9–79 has never visited a public
library (Norsk kulturbarometer, 1997). The citizens’ satisfaction with the public li-
brary service is also high 4 and we infer that the good is familiar to the citizens.
The first part of the questionnaire used a top-down design, starting with a de-
scription of overall municipal services, moving down to cultural goods, and then to
public libraries. The respondents were told how the total costs in an average Nor-
wegian municipality are allocated among main budget posts. 5 The intention was to
put libraries in a context and remind respondents of other local public goods that
compete for scarce private and public resources.

2
For a theoretical discussion of whether economic models for valuing nonmarket goods can fruitfully be
applied on public libraries, see Aabø and Audunson (2002).
3
100 NOK was
US$ 8.5 at the time of the survey.
4
In 2003, public libraries were ranked as number one of 53 public services, together with kindergartens
and physiotherapy (TNS Gallup, 2003).
5
Education 25%, hospitals 22%, health care 16%, social security and welfare 12%, business 6%, ordinary
public services 6%, recreation, culture and religion 4%, and other purposes 9% (Statistics Norway, 1998).
178 S. Aabø / Information Economics and Policy 17 (2005) 175–198

Only at this point focus turned to public libraries and specifically to the respon-
dent’s local library. 6 A main concern is how each individual citizen values public
library benefits. Public libraries can be seen as a mixed good. Some of their benefits
have individual property rights (photocopies of journal articles), others have col-
lective property rights (browsing facilities), and others again property rights that can
be considered as both individual and collective (book loan) (Aabø, 1998). Private
and mixed goods are primarily distinguished by the difference between individual
and collective property rights (Mitchell and Carson, 1989). To elicit the issue of
perceived property rights to public library benefits we posed the question: ‘‘Do you
think you have a right to have access to a public library in the municipality where
you live?’’ The answers were almost unanimous, 94% saying ‘‘yes’’, a much higher
fraction than those who stated to be library users, only 60%.

2.1. WTP versus WTA

CV implies that respondents are asked to state their values of a change in the
provision of a nonmarket good, in the form of WTP for an improvement or mini-
mum compensation (WTA) to accept a change to the worse. In theory, WTP and
WTA shall differ only by small amounts for nonessential goods with low budget
shares. Empirical WTA estimates are however often considerably higher than WTP
estimates for the same good (Hanemann, 1991; MacDonald and Bowker, 1994). Due
to this observation, and especially since the recommendation from the ‘‘NOAA-
panel’’ to use WTP as a conservative choice (Arrow et al., 1993), when the natural
setting calls for estimating WTA it is instead customary to estimate WTP.
We conducted two pilot studies to test the scenario plausibility and use of the
WTA format. One hypothesis concerning the WTP/WTA discrepancy is the exis-
tence of an ‘‘endowment effect’’, that is the supposition that individuals value losses
more that commensurate gains. MacDonald and Bowker (1994) studied the presence
of the endowment effect in an economic analysis of localized air pollution. Their
findings show that perceived property rights may well be an important factor of the
divergence. If this is the case, the WTP/WTA choice becomes a crucial factor in
estimating potential welfare impacts of a policy alternative. If citizens appear to have
an inherent right to the valued good, ‘‘the common practice of substituting WTP
measures for WTA will understate welfare impacts, increasing the probability of
rejecting a proposal that would provide a potential Pareto improvement’’ (Mac-
Donald and Bowker, 1994, p. 547). In Norway, all citizens have a legally based right
to a local public library and, as seen in this study, acknowledge this right.
Another hypothesis is that the divergence between WTP and WTA values also
depends on a substitution effect. Hanemann (1991) showed that the smaller the
substitution effect the greater the difference between WTP/WTA. Public libraries

6
The local library is here defined as the public library in the municipality where the citizen lives and it
typically includes a main library and one or more branches. The local public libraries are a municipal
responsibility.
 S. Aabø / Information Economics and Policy 17 (2005) 175–198 179

provide a wide range of services, some having high substitution ability (book loans)
and others very low (‘‘Reach out’’-services). Library and information studies show
that respondents appreciate social benefits of libraries (Holt et al., 1999; Audunson,
2001). Nonuse values thus motivated are shown to constitute a substantial share of
public libraries’ total value (Aabø and Strand, 2004). Library services with high
nonuse value have small substitution effect and may be one of the explanations of a
possible high difference between estimated WTP and WTA values.
Boyle and Bergstrom (1999, p. 192) discuss rational explanations of the WTP–
WTA disparity and comment on the NOAA Panel’s recommendation of always
using WTP in CV studies: ‘‘Because of the theoretical relevance of WTA under
certain property-rights structures, it seems inconsistent simultaneously to advocate
the use of CV and exclude applications to WTA’’. For the choice situation in our
study, depicting the public library to be closed down if the local taxes are not raised,
WTA is arguably the theoretically correct welfare measure. In this situation, the
individuals suffer the loss of not being able to use the local library to which a solid
majority acknowledge their property rights. Due to this fact we found strong ar-
guments to elicit also WTA.

2.2. The scenario design

The scenario description starts by referring to the Norwegian Public Library Act
and its purpose statement (Section 1) and continues to describe the choice situation:
‘‘It is well known that the economic situation in most of the municipali-
ties is deteriorating. This can imply that some public services have to be
reduced or closed down, unless the municipality’s revenues are increased.
Assume the Public Library Act amended, so that the municipalities them-
selves could decide whether or not they wanted a public library. Imagine
that the council administration was considering closing down the library.
An option would then be to use the public library in a neighbouring mu-
nicipality or to buy all books, reference manuals, information services,
etc., needed by yourself and your household. Library services to schools
and adult training courses and to various groups in the local community,
such as the ‘‘Reach out’’-service to elderly in institutions, kindergartens,
etc., will cease to exist.
Another alternative is maintaining the library services, if the municipal-
ity’s revenues are sufficiently increased through additional local taxes’’. 7
The WTA scenarios have the same framing but here the choice is between
either (1) closing down the local library to use the saved budget funds on other

7
This scenario description was used in the two WTP subsamples.
180 S. Aabø / Information Economics and Policy 17 (2005) 175–198

municipality tasks benefiting the household (e.g., education, health), or (2) main-
taining the library and also other municipality tasks on today’s level of activity.
The scenario descriptions are constructed to elicit the total value of the local
public library holistically (Randall, 1991). They do not intend to describe all attri-
butes or value components but point to some well-known public library services
(book lending, provision of reference literature, and information services) and some
activities that are less known (services targeted at specific groups and outreach
programmes).
Our study is based on a national population sample with respondents from a
variety of municipalities. The public libraries in the municipalities vary regarding
range and quality of the library service due to the size of their population, allocated
resources, and efficiency. By our scenario descriptions each respondent values the
public library in her municipality, e.g., the public library services she is aware of and
puts weight to, thus reflecting the present level of both the library’s activities and its
information of them to the citizens.

2.3. Elicitation formats

Previous research has shown that the choice of elicitation method can significantly
influence estimates of mean and median WTP in CV studies. We used two different
elicitation formats, multiple bounded discrete choice (MBDC) and dissonance
minimizing (DM), to seek to correct for elicitation method effects (Blamey et al.,
1999; Welsh and Poe, 1998).
The DM format was developed by Blamey et al. (1999) with the objective to re-
duce overestimation of values due to ‘‘yea-saying’’. 8 When answering a valuation
question respondents may have two objectives, revealing their (i) preferences and (ii)
attitudes for the good. The dominant need is often to express attitudes. Respondents,
who think the program at issue should proceed but have WTP less than the bid
amount, may still respond ‘‘yes’’ to a standard dichotomous choice (DC) question in
order to express their attitudes towards the program. The DM format is a discrete
choice technique designed to minimize such yea-saying by including additional re-
sponse categories that permit respondents to express support for the good without
having to commit money, thus making it possible to separate between demonstration
of attitudes and stated WTP.
The theoretical reasoning behind this elicitation format indicates that it can be
helpful when pilot studies show that there are respondents who have objections to
aspects of the scenario. Our pilot studies indicated that the group of protest voters
included both real zero bidders and respondents with positive valuation. In the main
survey, the valuation question in the DM format allowed respondents to choose
between multiple statements, including an option to say ‘‘no’’ to the bid but still
express support for libraries. Table 1 shows the wording of the first valuation

8
Blamey et al. (1999, p. 126) define yea-saying as ‘‘the tendency to subordinate outcome-based or ‘true’
economic preferences in favor of expressive motivations when responding to CVM questions’’.
 S. Aabø / Information Economics and Policy 17 (2005) 175–198 181

Table 1
The first valuation question, Q1, in elicitation format DM-WTP with distribution of answers to six re-
sponse options
Do you support maintaining the local public library services and are willing to pay . . . NOKa in
additional annual local taxes to prevent closing down of the library? Which one of these statements best
expresses your answer?
Frequences Percent
01. I support maintaining the local public library services and am 89 36
willing to pay . . . NOK in local tax increase
02. I support maintaining the local public library services and am 25 10
willing to pay an additional local tax, but it is worth less than . . .
NOK to me
03. I support maintaining the local public library services but 98 39
disagree that it demands additional local taxes
04. I do not support maintaining the local public library services 7 3
even if it does not cost me anything
05. Do not know 15 6
06. Will not answer 7 3
Missing 9 4
Total 250 101
a
The bids varied randomly between 100, 300, 500 and 1000 NOK. 100 NOK was
US$8.5 at the time
of the survey.

question, Q1. It presents six response options and option 03 aims to catch possible
protest voters. They were posed follow-up questions especially designed to prompt
them to state their true preferences, see Box 1.

Box 1. Follow-up valuation question in the elicitation format DM-WTP posed to

possible protest bidders, i.e., respondents who answered the 03-alternative in Q1, see
Table 1.

Do you fully agree, partly agree, partly disagree or fully disagree with the
statements below, or do not you know:
(1) I would be willing to pay . . . NOK to maintain the local library service if I
was convinced that the municipality is unable to pay the costs within their
budget.
(2) I would be willing to pay . . . NOK to maintain the local library service if I
was convinced that the municipality is unable to pay the costs without having to
reduce services in the health and educational sectors.
(3) I cannot afford to pay anything to maintain the local library service.

Our second elicitation format is MBDC, developed by Welsh and Poe (1998). It
contains two dimensions, first a number of dollar amounts in ascending order, and
secondly, a scale of certainty levels (‘‘definitely no’’, ‘‘probably no’’, ‘‘not sure’’,
182 S. Aabø / Information Economics and Policy 17 (2005) 175–198

‘‘probably yes’’, ‘‘definitely yes’’). This format combines elements from both the
payment card (PC) and DC approaches. Like the PC format, respondents are pre-
sented with an ordered sequence of dollar amounts but, rather than just to circle a
single value or interval, they are given a ‘‘polychotomous choice’’ and asked to
choose a level of voting certainty at each of the dollar thresholds.
The two formats were both adapted to scenario descriptions for WTP and
WTA, respectively, thereby developing a four-cell design, see Box 2. The respon-
dents were randomly assigned to these four subsamples and each respondent be-
longed to only one. The first valuation question, Q1, had a referendum form and
was posed in either MBDC or DM formats. For all four subsamples this question
was immediately followed by an open-ended (OE) valuation question, Q2. This
design yields four independent value estimates and provides the possibility to
compare answers between (i) the MBDC and DM formats, (ii) Q1 and Q2, and (iii)
WTA and WTP.

Box 2. The four subsamples.

DM-WTP (n ¼ 250) DM-WTA (n ¼ 241)

MBDC-WTP (n ¼ 257) MBDC-WTA (n ¼ 251)

Our valuation estimates are based on several different model variants. Q1 is, in the
DM format, estimated using a logit model. In the MBDC format, we use two dif-
ferent principles for estimation. The first assumes that the respondents state their
‘‘true’’ preferences at one certainty level, namely ‘‘probably yes’’ given to be 95%
certain. The second assumes that the true preferences are derived as an arithmetical
average of a combination of several certainty levels, weighted or unweighted.

3. WTP results

3.1. WTP estimates of the first valuation question, Q1, in the MBDC format

Respondents were presented a payment card with eight amounts (100, 300,
500, 700, 1000, 2000, 5000, 10,000 NOK) and asked to assign a certainty level to
each amount, using a scale from ‘‘definitely yes’’ to ‘‘definitely no’’. This tech-
nique permits estimation of mean and median values for each certainty level. We
wanted to arrive at a single estimate for each respondent, which would express as
plausible as possible her WTP response to Q1. As one of two methods applied we
used the level ‘‘probably yes’’, which may be agreed to yield a conservative es-
timate. The highest amount to which the respondent stated ‘‘probably yes’’ was
interpreted as the lower limit of her valuation of the local library and the lowest
amount on the next certainty level, ‘‘not sure’’, as her upper limit. The single
 S. Aabø / Information Economics and Policy 17 (2005) 175–198 183

estimate of each respondent’s valuation, EP, is determined as the average mean

value of the upper and lower limit.
Our second technique to calculate WTP was to combine, in different ways, all of
the certainty levels except the ‘‘not sure’’ option. Three different estimates for each
respondent were calculated as weighted or unweighted sums of these certainty levels.
In EW1 the weights 0.1 and 0.4 were assigned to the levels ‘‘definitely’’ and
‘‘probably’’, respectively, in EW2 the weights 0.3 and 0.2, and in EU the four cer-
tainty levels were given equal weights. The variation between these three estimates is
small and we therefore, in Table 2, present only EW1 together with EP, the estimate
using only one certainty level. For both calculation techniques median and mean
WTP and a 95% confidence interval for the means are presented. Observe that EP is
cleary the lowest estimate. The difference between the two elicitation methods is
relatively large, up to a ratio 1:1.5.
The size of the estimates is influenced by the treatment of protest bids. We treat
the protest bids in three ways, as seen in Table 2: (1) exclude them from the sample,
(2) impute them using a missing value (MV) analysis and (3) include their stated
value, i.e., 0. Statistical equivalence of the estimates according to treatment of
protest bids is evaluated with the following hypothesis test:

i; j ¼ EP; EW1; EW2; EU; i 6¼ j;

H0 : X
ik ¼ X
jk ;
k ¼ 1; 2; 3;

where X is the estimate and k indexes protest bids treatments. Of the 12 possible
pairwise WTP hypothesis tests only three were rejected at the 5% level, while the

Table 2
Valuation question Q1 in elicitation format MBDC with two WTP estimates of mean and median in NOK,
including a 95% confidence interval for mean
WTP
Median Mean 95% Confidence N
interval for meana
Lower Upper Absolute numbers
bound bound and percent
EP, one certainty level: ‘‘Probably sure’’ 400 980 750 1200 240; 93%
Exclusive protest bids
EP 400 940 715 1170 257; 100%
Protest bids given MV value
EP 400 950 720 1170 250; 97%
Inclusive protest bids, i.e., 0 NOK
EW1, four certainty levels 500 1500 1210 1780 240; 93%
Exclusive protest bids
EW1 500 1480 1170 1795 257; 100%
Protest bids given MV value
EW1 500 1440 1160 1710 250; 97%
Inclusive protest bids, i.e., 0 NOK
a
The confidence interval is found in a one-way analysis of variance.
184 S. Aabø / Information Economics and Policy 17 (2005) 175–198

Table 3
Regression analyses of log-transformed WTP to Q1, EP with zero bids excluded (column 2) and included
(column 3)
Independent variables Log-transformed posi- Log-transformed WTP includ-
tive WTP (B-coefficient) ing zero bids (B-coefficient)
Library user (0 ¼ no, 1 ¼ yes) 0.290 (0.160)a 0.759 (0.305)b
Household income (log-transformed) 0.284 (0.101)c 0.187 (0.196)
High education (1 ¼ university or college, )0.035 (0.160) )0.065 (0.308)
0 ¼ lower)
Urbanity ¼ town (0 ¼ city and country )0.376 (0.176)b )0.471 (0.327)
side, 1 ¼ town. city ¼ ref. cat.)
Urbanity ¼ countryside (0 ¼ city and )0.793 (0.197)c )0.571 (0.377)
town, 1 ¼ countryside)
Cultural activity (scale 0–10, where 10 is 0.060 (0.034)a 0.118 (0.067)a
the highest level)
Distance to library (kilometers from )0.170 (0.082)b )0.433 (0.144)c
dwelling, log-transformed)
Constant 2.759 (1.252)b 3.240 (2.428)
Adjusted R2 0.175 0.129
Standard errors are indicated in parentheses.
a
Denotes 10% significance level.
b
Denotes 5% significance level.
c
Denotes 1% significance level.

other nine were not rejected. Generally, thus, protest bids treatments (1) and (2) give
estimates that do not differ significantly.
The respondents’ WTP varied considerably, 26 gave zero bids and 214 positive
bids varying from 50 to 10,000 NOK. To explain first the variance of the positive
bids in this subsample, we included several explanatory variables in a linear multi-
variate regression analysis, as shown in Table 3. 9 We used a log-transformation of
EP with protest bids excluded as the dependent variable. Column 2 shows that the
background variables library user and cultural activity have positive effects on WTP.
Household income elasticity is highly significant and positive but small, not un-
common either in CV studies or studies exploring library use. Elasticity of distance to
the local library is negative, and rural residents and town folks have less WTP than
city dwellers. High education is not significant. When respondents who gave zero bids
are included in the analysis, see column 3, we find that the variables library user and
distance have enhanced both effect and significance, and cultural activity has stronger
effect. The remaining four explanatory factors are not statistically significant.

9
Three demographic variables available in the survey, sex, age, and family size, were found to have
almost no effect on predicted valuation, either in the regression analysis of this subsample or of the other
three subsamples.
 S. Aabø / Information Economics and Policy 17 (2005) 175–198 185

3.2. WTP estimates of Q1 in the DM format

We now turn to the valuation procedure of Q1 in the DM format. The estimations

here used the standard logistic distribution function, as follows (in the semivariate
specification):

1
F ðX ; bÞ ¼ ;
1þ eðaþbX Þ

where F is the distributive function, X is the WTP, and a and b are coefficients to be
estimated. The dependent variable in the logistic regression is a dichotomous vari-
able representing the respondent’s probability to answer ‘‘yes’’ or ‘‘no’’ to the bid.
This variable differs according to identification and treatment of protest bids and
logit coefficients for four DM models are given in Table 4. The estimated mean and
median vary between 616 and 1220 NOK in the models. The regression coefficient
for bid is significant at the 1% level in DM3 and DM4. When we include explanatory
variables in the DM4 model, we see that only two explanatory factors in addition to
bid are significant at the 10% level, cultural activity has a positive and distance a
negative effect. None of the other independent variables are significant at the 10%
level in this subsample, but the B-coefficients and their signs accord with expectation
and the results in the MBDC subsample. An exception is library user which, contrary
to general expectations, is negative but its B-coefficient is far from significant.
Household income is here dichotomized and the B-coefficient is positive but not
significant.
An assessment of the DM-models according to treatment of protest bids, price
sensitivity, model significance, percent of correct predictions, and significance of the
logit coefficients (Blamey et al., 1999) showed that DM4 was best suited for library
valuation.

3.3. WTP estimates to Q2

The second valuation question, Q2, was open-ended and posed immediately after
Q1. The interviewer introduced Q2 by informing the respondents that their possible
WTP levels were no longer limited by the amounts on the payment card (MBDC
subsample) or the amount in the dichotomous choice (DM subsample). They were
rather asked to consider freely what was the maximum amount they and their
household were willing to pay, as an increase in annual local taxes to maintain the
library services in their municipality. The interviewer was instructed to differentiate
nonpositive responses into protest bids (not accepting the question) and real zero
bids (willing to pay nothing). We present three Q2 estimates in Table 5, varying by
the treatment of protest bids. Note that the difference in means between the sub-
samples MBDC and DM is small and clearly less than in Q1.
Table 6 describes linear and log-linear OLS relationships between WTP and
background variables, in model 1 explaining the variance of the positive bids and in
model 2 including the zero bids as well. The explanatory power of the models
 186
Table 4
Logit coefficients for the four WTP DM models
DM1 DM2 DM3 DM4 without DM4 (inclusive
zero bids zero bids)

S. Aabø / Information Economics and Policy 17 (2005) 175–198

Constant )0.018 (0.234) 1.807 (0.389)c )0.869 (0.243)c 2.661 (0.440)c 1.207 (0.534)b
Bid )0.001 (0.000) )0.002 (0.001)a )0.001 (0.000)c )0.002 (0.001)c )0.003 (0.001)c
Library user (0 ¼ no, 1 ¼ yes) )0.068 (0.439)
Household income (0 < average, 1 P average) 0.428 (0.387)
High education (1 ¼ university/college, 0 ¼ lower) )0.109 (0.440)
Urbanity ¼ town (0 ¼ city and countryside) )0.406 (0.423)
Urbanity ¼ countryside (0 ¼ city and town) )0.603 (0.479)
Cultural activity (scale 0–10, where 10 is the highest level) 0.376 (0.110)c
Distance to library (kilometers from dwelling, )0.480 (0.222)b
log-transformed)
Correct predictions (%) 52.8 73.6 60.6 82.8 74.2
Estimated median C  904 C  870 C 1 1016
Estimated mean C  904 C  869 C 1 1016
C  616 C  935 C  1220 C 1 741
C  684 C  980 C  791 C 1 1047
N 216 121 216 145 (1901 ) 178
Standard errors are indicated in parentheses.
C  ¼ a=b, infinite < C  < infinite.
max
C  ¼ 1=b  ln½ð1 þ ea Þ=ð1 þ eabðB Þ Þ0 6 C  < Bmax .
 a 
C ¼ 1=b  lnð1 þ e Þ; 0 6 C < infinite.
a
10% Significance level.
b
5% Significance level.
c
1% Significance level.
1
The zero bids (45 respondents) are included in these estimates based on a total of 190 respondents.
 S. Aabø / Information Economics and Policy 17 (2005) 175–198 187

Table 5
Mean and median in NOK of Q2 with differing treatment of protest bids for the two WTP subsamples
Q2 – WTP in NOK
Mean Median
MBDC DM MBDC DM
Protest bids excluded 425 (55) 350 (55) 300 100
Protest bids given MV value 370 (75) 370 (50) 300 200
Protest bids included ¼ 0 NOK 375 (50) 325 (50) 200 100
Standard errors are indicated in parentheses.

increased when we added two variables, the first registering whether the respondent
had used the library as a child, and the second whether the municipality library had a
professionally qualified chief librarian. Viewing the four analyses together, observe
that cultural activity is the only explanatory factor that is statistical significant in all
of them, showing an
10% positive effect per step up on the 10-point scale.
Household income elasticity is positive and significant at the 5% level in three of the
analyses. Distance elasticity is negative and significant in both log-linear analyses.
The explanatory factor library use is split into occasional use (1–3 library visits per
year) and frequent use (4 and more visits per year), the first significant in the log-
linear analysis of model 2 and the latter in both. Using the library as a child and a
professionally qualified chief librarian both have positive effect, significant in the
second log-linear analysis. Explained variance in the four models is 8–25% (adjusted
R2 varying between 0.076 and 0.245), somewhat lower but in line with earlier findings
in the library and information literature. 10 In most CV studies, explanatory power is
10–15%. More important than a high R2 are significant explanatory factors with the
expected signs.

4. WTA results

For the two WTA subsamples we used the same valuation procedures as for the
WTP subsamples. Starting with the MBDC results, we present estimates based on
the two estimation techniques in Table 7, noting that the estimates are strikingly
equal. The estimate differences according to treatment of protest bids follow the
same pattern in the WTA estimates, as they did for WTP. Respondents’ individual
WTA varied extensively, 35 respondents demanded no compensation and 152 de-
manded positive amounts varying from 100 to 25,000 NOK. To explain some of this
variation, we regressed the log-transformed EP first for the positive demands and

10
Zweizig and Dervin (1977) analysed variance in library use in the population and identified 10 factors
that explained 33% of the variance. D’Elia’s (1980) analysis explained 29% of library use and 36% of use
frequencies. Audunson (1995) replicated D’Elia’s analysis on Norwegian data and found that the model
then explained 19% of the variance.
 188
Table 6
Explanatory factors’ impact on WTP to Q2 for respondents with positive bids and for respondents with positive and zero bids
Independent variables Model 1 Model 2 Model 1 Model 2

S. Aabø / Information Economics and Policy 17 (2005) 175–198

Linear OLS, positive Linear OLS, Log-linear OLS, Log-linear OLS,
bids positive and positive bids positive and
zero bids zero bids
Cultural activity (scale 0–10, where 10 is the highest activity level) 70.73 (29.56)b 60.24 (24.39)b 0.099 (0.026)c 0.135 (0.062)b
Household income (log-transformed) 195.10 (85.55)b 148.91 (71.07)b 0.164 (0.076)b 0.161 (0.179)
High education (1 ¼ university or college, 0 ¼ lower) 241.90 (138.71)a 192.29 (112.00)a 0.160 (0.122) 0.143 (0.282)
Urbanity (0 ¼ town and countryside, 1 ¼ city) )1.52 (139.79) )14.082 (109.09) )0.001 (0.123) )0.120 (0.275)
Distance to library (kilometers from dwelling, log-transformed) 86.96 (77.33) 41.81 (54.47) )0.138 (0.068)b )0.293 (0.137)b
Subsample (0 ¼ MBDC, 1 ¼ DM) )80.14 (119.16) )74.95 (95.28) )0.247 (0.105)b )0.696 (0.240)c
Debrief, WTP (0 ¼ payment vehicle unfair, 1 ¼ fair) )6.86 (122.06) 99.75 (102.98) 0.082 (0.108) 1.437 (0.260)c
Library use, 1–3 visits (1 ¼ 1–3 Visits per year, dummy1. )155.95 (155.95) )73.92 (127.54) )0.159 (0.148) 0.657 (0.322)b
Ref.cat. ¼ nonuse)
Library use, 4-visits (1 ¼ 4 or more visits per year, dummy2) 233.06 (160.88) 234.67 (127.86)a 0.050 (0.142) 0.872 (0.322)c
Library user as a child (0 ¼ no, 1 ¼ yes) )66.91 (137.02) 7.64 (105.14) 0.143 (0.121) 0.623 (0.265)b
Professionally qualified chief librarian (0 ¼ no, 1 ¼ yes) 219.06 (207.33) 177.32 (148.62) 0.237 (0.183) 0.693 (0.375)a
Constant )2261.90 (1159.40)a )1749.97 (951.74)a 3.954 (1.023)c 2.785 (2.400)
Adjusted R2 0.076 0.085 0.151 0.245
Model significance 0.001 0.000 0.000 0.000
NOK per respondent in linear relationships.
Standard errors are indicated in parentheses.
a
10% Significance level.
b
5% Significance level.
c
1% Significance level.
 S. Aabø / Information Economics and Policy 17 (2005) 175–198
Table 7
Valuation question Q1 in the elicitation format MBDC with four WTA estimates of mean and median in NOK, including a 95% confidence interval for mean
WTA
Median Mean 95% Confidence interval for meana N
Lower bound Upper bound Absolute numbers and percent
EP, based on one certainty level: ‘‘Probably sure’’ 760 2090 1620 2560 186; 74%
Exclusive protest bids
EP based on one certainty level: ‘‘Probably sure’’ 1000 2420 2040 2800 251; 100%
Protest bids given MV value
EP based on one certainty level: ‘‘Probably sure’’ 1000 3530 2980 4080 229; 91%
Inclusive protest bids, i.e., 0 NOK

EW 1, based on four certainty levels 700 1950 1500 2400 180; 72%
Exclusive protest bids
EW1, based on four certainty levels 1100 2430 2060 2800 251; 100%
Protest bids given MV value
EW1, based on four certainty levels 1100 3460 2900 4000 223; 89%
Inclusive protest bids, i.e., 0 NOK
a
The confidence interval is found in a one-way analysis of variance.

189
190 S. Aabø / Information Economics and Policy 17 (2005) 175–198

secondly including also the non-compensated demands, in a way analogous to the

MBDC-WTP subsample and with the same independent variables. In both regres-
sion analyses, library user is significant at the 5% level, showing a clear interde-
pendence between library use and the size of compensation demands. Among
respondents with positive compensating demands, high education is significant, in-
dicating that people with university or college education have 40–50% higher WTA
than people with lower education. When non-demands are included, the B-coefficient
for high education is still positive, but smaller and no longer significant. Living in
a town compared to a city and distance both have negative effects and are now
significant.
In the DM-WTA subsample, respondents were posed a valuation question with
three answer options and asked to choose the alternative that best expressed their
opinion, see Table 8. In contrast to the DM-WTP subsample, respondents who chose
option B (which can include possible protests) were not posed a follow-up question
designed to help differentiating these answers. The most striking result here is that
only 3% of the subsample support closing down the local public library, although the
money saved would be transferred to other municipality tasks benefiting their
households. This result seems to reflect the fact that the overwhelming majority of
respondents feel they have property rights to public libraries.
Response option B in Table 8 was chosen by half of the subsample. Although it
can contain protest voters, it will normally consist of respondents with a positive
valuation of the local library, some of who may have a very high valuation of it. It
seems reasonable to assume that the B responses imply at least as high average
valuation as the A responses. After testing two models based on different interpre-

Table 8
The first valuation question, Q1, in elicitation format DM-WTA with distribution of answers to the three
response options A–C
Imagine that the municipality council considers two alternatives:
1. To close down the local library and use the saved budget funds to increase the efforts on other
municipality tasks that will benefit your household.
2. To maintain the local library and also other municipality tasks on today’s level of activity.
Which one of the statements A–C best expresses your answer?

Frequencies Percent
A. I support maintaining the local public library if the alternative is to 94 39
close down the library and transfer . . . NOK in saved budget funds to
other municipality tasks that will benefit my household
B. I support maintaining the local library if the alternative is to close 123 51
down the library, independent of the amount of saved budget funds
that then can be transferred to other municipality tasks that will benefit
my and other households
C. I support closing down the local library if it involves that . . . NOK is 8 3
transferred to other municipality tasks that will benefit my household
Missing 16 7
Total 241 100
 S. Aabø / Information Economics and Policy 17 (2005) 175–198 191

Table 9
Mean and median in NOK of Q2 with differing treatment of protest bids for the two WTA subsamples
Q2 – WTA in NOK
Mean Median
MBDC DM MBDC DM
Protest bids excluded 1500 (250) 850 (220) 400 300
Protest bids given MV value 2000 (190) 1050 (120) 500 600
Protest bids (10,000 NOK) inclusive 4350 (300) 5700 (310) 1000 10,000
Standard errors in parentheses.

tations of the responses to options B, we chose to use a model where A and B re-
sponses are both interpreted as rejecting the compensation offer for closing down the
local library. The amount they were asked to consider is interpreted as the lower
limit of their positive valuation of the library. The C option implies values lower than
the bids. They are here interpreted as real zero bids, which is very conservative.
In logit regression analyses, the bid coefficient was not significant at the 10%
level when only bid was included as independent variable, and this is to be expected
since so few respondents accepted the bid offered as compensation for closing down
the local library. When the explanatory factor library user was included as well,
both coefficients were significant at the 10% level, showing that a library user
was more likely not to accept the compensating bid for closure of the library than
a nonuser and that, for low bids, fewer respondents accepted the compensating
offers.
The second, open-ended valuation question, Q2, was posed immediately after Q1,
analogous to the WTP subsamples. Respondents should consider the minimum
amount per household that must be transferred to other municipality activities
benefiting their household, for them to support the proposition to close down the
local library service. The interviewer was instructed to differentiate between protest
voters (not accepting the question or any money amount) and those who were in-
different and did not demand any compensation (library value ¼ 0 NOK). WTA
estimates to Q2 are shown in Table 9, differing according to protest bids treatment.
For the two estimates where protest bids are excluded or imputed using a MV
analysis, the highest difference is the ratio 1:2.1 for mean and 1:1.3 for median values.
The third estimate includes protest bids as 10,000 NOK and therefore displays very
high figures and differs significantly from the other two, as expected.

5. Protest bids

Dealing with protest bids is important in many CV studies, and particularly

critical for the WTA part of our study. In our four-cell design, the DM-WTP format
was developed to deal specifically with protest bids. Response option 03 in Table 1
was formulated with the intention of catching possible protest bids and bids
192 S. Aabø / Information Economics and Policy 17 (2005) 175–198

motivated by considerations of possible inefficiency of public library expenditure. 98

respondents, nearly 40% of the subsample, voted for this option. They were posed
follow-up questions, see Box 1, making it possible to differentiate this group of
possible protest bids into: (i) zero bids, 39%, (ii) positive bids, 32%, (iii) real protest
bids, 19%, and (iv) do not know-answers, 10%. This procedure indicated that only
20% of the possible protest bids in this subsample were real protest bids. For the
three remaining subsamples, however, follow-up questions to make this distinction
were not posed. As a means to differentiate between possible and real protest bids we
considered the answers from these respondents to the second valuation question,
Q2. 11
An overview for all four subsamples of the response distribution to the answer
categories missing values, possible protest bids, real protest bids, zero bids, and
positive bids is given in Table 10, both for the raw data and the data after treatment
of protest bids in Q1 and for Q2. Looking first at the Q1 data, observe that the
treatment significantly reduced the number of protest bids in all three subsamples
where this procedure was used, but clearly most in DM-WTP, from nearly 40% to
8%. This indicates that the procedure of utilizing information from Q2 to differen-
tiate possible protest bids in the other two subsamples does not unduly reduce the
number of protest bids. In MBDC-WTP, the raw data protest bids constitute 15%
and after treatment only 5%. The share of real protest bids thus is low in both WTP
formats. In the WTA subsamples, this treatment of protest bids was applied only to
MBDC-WTA. Here the raw data’s share of protest bids is nearly 40%. After
treatment the share is still substantial, over 20%. The interpretation of DM-WTA
implied that the Q1 raw data were used directly without treatment and no protest
bids were identified.
The share of positive bids increased after treatment of the raw data except for
MBDC-WTP, which had a high share of positive bids at the outset, more than 80%.
In subsample DM-WTP, the positive bids make up almost 60%. The share of zero
bids lies between 4% and 18%, highest for DM-WTP. Considering the responses to
Q1 as a whole, our two WTP elicitation formats render a share of both protest and
zero bids that are lower than in many other CV studies. For WTA the share of
protest bids is high in MBDC-WTA but here the share of positive compensation
demands is high as well, 60%.
Turning to Q2, observe that the share of protest bids is significantly higher, with
exception of DM-WTP. The share of protest bids in the WTP formats is still low, 7%
for DM-WTP and 12% for MBDC-WTP, whereas it is disturbingly high in the WTA
formats, 53% and 33%, respectively. The same pattern is seen in the proportion of
positive amounts, where
70% of the WTP respondents state positive bids. Of the
WTA repondents less than a half, 43%, state positive bids, 47% in MBDC-WTA and

11
In the MBDC format, two groups were identified as possible protest bids: respondents who (i) did not
accept the highest amount, 10,000 NOK, as compensation for closure of the local library and (ii) gave
inconsistent answers by saying definitely or probably ‘‘yes’’ to accepting a compensation which was lower
than one to which they said ‘‘no’’. Vice versa for WTP.
 S. Aabø / Information Economics and Policy 17 (2005) 175–198
Table 10
Distribution in percent of response categories missing values, possible protest bids, real protest bids, zero bids and positive bids of all four subsamples to Q1a
and Q2
DM-WTP MBDC-WTP DM-WTA MBDC-WTA
Response category Q1 Q1 Q2 Q1 Q1 Q2 Q1 Q2 Q1 Q1 Q2
Raw data After Raw data After Raw data Raw data After
treatment treatment (without treatment
treatment)
MVb 12 16 2 3 3 0 7 0 4 6 2
Possible protest bids 39 15 0 38
Real protest bids 8 7 5 12 53 21 33
Zero bids 13 18 26 10 16 4 18 14 19
Positive bids 36 58 66 83 83 72 90 29 58 60 47

Total 100 100 101 101 101 100 101 100 100 101 101
a
For Q1 the raw data as well as the data after protest bid treatment are given.
b
Missing values (MV) include response categories ‘‘Do not know’’ and ‘‘Will not answer’’.

193
194 S. Aabø / Information Economics and Policy 17 (2005) 175–198

Table 11
Mean and median in NOK, respectively, to Q1 and Q2 for all subsamples
Q1 Q2
Mean
MBDC-WTP DM-WTP MBDC-WTP DM-WTP
1500–980 (N ¼ 238) 675 (N ¼ 190) 425 (N ¼ 227) 350 (N ¼ 228)
MBDC-WTA DM-WTA MBDC-WTA DM-WTA
2000–2100 (N ¼ 184) 1720 (N ¼ 225) 1500 (N ¼ 168) 850 (N ¼ 113)
Median
MBDC-WTP DM-WTP MBDC-WTP DM-WTP
500–400 675 300 100
MBDC-WTA DM-WTA MBDC-WTA DM-WTA
700–760 1720 400 300
Protest bids are excluded.

only 29% in DM-WTA. The proportion of zero bids differs little between the
subsamples, 16% and 26% in the WTP formats, and 18% and 19% in the WTA
formats. The large difference between Q1 and Q2 in DM-WTA, where more than
half of the subsample states protest bids in Q2, is striking. The high shares of protest
bids in the WTA subsamples, and especially in DM-WTA, are conspicuous.
The WTA protest voters were examined further to find out whether they had
common characteristics. They were regressed against two explanatory variables
based on debriefing information. 12 The first was not significantly different from zero
but the second was highly significant. In addition, a protest WTA voter is charac-
terized by being a library user with high education who uses few other cultural ac-
tivities in the municipality. These explanatory factors validate that among the protest
voters there are, expectantly, respondents with a high valuation of public libraries.

6. Discussion and conclusions

An overview in the four-cell design of mean and median to both Q1 and Q2 for
the whole sample is presented in Table 11. Observe that elicitation effects are present,
both between the formats MBDC and DM and between our two variants of MBDC.
There is a systematic tendency for amounts stated to Q1 to be considerably higher
than to Q2. The WTP–WTA disparity, on the other hand, is small compared to
many other studies applying these two approaches, despite the fact that our re-
spondents expressed exceptionally strong property rights to public libraries. The
high share of protest bids in the DM-WTA subsample may be a source of uncer-
tainty. Another source of possible errors in the WTA estimates is that the

12
The first debrief variable was the interviewer’s evaluation of whether or not the respondent had
problems with the valuation questions, and the second was the respondent’s agreement or disagreement
with the statement ‘‘We must retain the local library regardless of the sum of budget funds saved by closing it
down’’.
 S. Aabø / Information Economics and Policy 17 (2005) 175–198 195

compensation for closing down the local library in our scenario is in the form of (a
money-equivalent value of) other local public goods.
An objective of this study is to measure the total benefits to the Norwegian cit-
izens of public libraries at today’s service levels, and thus determine whether public
libraries are ‘‘worth their price’’ as seen from the population’s perspective. Based on
the WTP estimates, we are able to ascertain a probable minimum estimate. The lower
bound for their average public library valuation is
400 NOK per household, which
is close to the average library cost per household. To establish an upper bound is
more difficult. The high share of WTA protest bids, especially in the DM-WTA
subsample, is problematic. There is no way to define a plausible upper bound for the
compensating demands of WTA protest voters, and this complicates interpretation
of these answers. The upper bound for citizens’ valuation of public libraries, how-
ever, clearly appears to lie considerably higher than the lower valuation limit – a
conservative estimate is 2000 NOK, based on the WTA estimates in Table 11. It is
reasonable to assume that the population’s ‘‘true’’ value is considerably higher than
the lower bound, in particular since there seems to be no a priori justification to
reject WTA in our case. An essential argument to attach importance to WTA esti-
mates is the property rights question, and a solid conclusion of our study is that an
overwhelming majority of the population perceives they have such property rights.
Our CV study appears to be the first CV study for valuation of public libraries at
the national level, in Norway or internationally. It explores the social value of public
libraries by eliciting this value among a random sample of the citizens. Based on
empirical data we conclude that Norwegian public libraries are, overall, worth their
price as viewed from the population’s perspective. At the national level, their benefits
decidedly outweigh their costs.
CV is used for policy-making purposes and the official acceptance of the method
has advanced this use. Therefore, it is important to make sure that there is close
correspondence between what we want to measure and what we really have mea-
sured. In our CV study, it is the respondents’ stated valuation amounts of the overall
value of the public library in their municipality that are measured. Some effects of the
public library service are probably not captured by the general public, for instance
long-term effects such as the library’s impact on community development, cumula-
tive results concerning social inclusiveness and citizenship, and effects of literacy on
employment opportunities. It is shown (Aabø and Strand, 2004) that many of the
respondents in this CV study value social and cultural aspects of the public library
service. These aspects have long-term effects and seem to constitute about 35–40% of
the stated total value, underscoring that the general public considers long-term ef-
fects as well as short term effects. It is however improbable that the respondents are
fully aware of all types of long-term effects of the public library service.
There may also be claimed to exist types of value yielded by public libraries, that
are not captured by CV or other methods of economic assessment but that never-
theless may be relevant for public library policy. Throsby (2003, pp. 279–282) points
to the term ‘‘cultural value’’. It has recently been used as a supplement to ‘‘economic
value’’ attempting to capture the worth of a good assessed in cultural terms that
cannot be expressed according to any quantitative or qualitative scale or in monetary
196 S. Aabø / Information Economics and Policy 17 (2005) 175–198

terms. CV and other WTP studies ‘‘will tend systematically to undervalue a cultural
good to the extent that there exist significant positive elements in the good’s value
that are incapable of expression as individual WTP’’ (p. 280).
In view this, our estimate of the social value of the Norwegian public libraries
based on assessments by a population sample may be an underestimation, since the
libraries have long-term effects that the general public is not aware of and since they
may yield ‘‘cultural value’’ not captured in stated valuation estimates. For assessing
the wide range of long-term impacts of public library services, experts and politicians
possess an information basis that the general public does not hold. In attempting to
arrive at an estimate of the full value of public libraries in Norway, the results of our
CV study may need to be supplemented. An option is to supplement the population’s
valuation with valuation from experts and politicians. Experts may also be able to
express the ‘‘cultural value’’ of public libraries.
Decisions of local public libraries are in Norway taken at both the national (the
Public Library Act), and at the municipal (the funding) level. Our sample is repre-
sentative for the population implying that we can draw conclusions at the national
level. The estimate from our study of the population’s valuation of public libraries
seems to lie within the range of 400–2000 NOK per household and, arguably, closer
to the upper bound. This in case implies that, as an average over all households, the
benefits from public libraries are greater than the costs of producing such library
services. A possible interpretation of this conclusion is that there is popular support
for a higher average level of public library funding. 13
Obviously, however, our sample represents only a small fraction of all 433 Nor-
wegian municipalities. Considering individual municipalities, both average valua-
tions and average library costs vary across them. Typically, and as expected,
valuation seems to be higher in larger municipalities with many cultural activities
and short average distances to the public library, which in addition is more likely to
provide a high service level and be headed by a professional chief librarian, when
compared to small municipalities where there are typically fewer cultural activities,
the public library is of lower quality, and average distances to the public library are
greater. From our national study conclusions cannot be drawn for each of the dif-
ferent municipalities’ public libraries. As a consequence, we cannot from our study
establish that all public libraries in Norway have positive net value; indeed, this is
unlikely.
This points to a need for more knowledge about the benefit–cost relationships for
public libraries, at the municipal level. An interesting avenue for future research
would be to attempt to find a relationship between public library services and dif-
ferent characteristics of the municipalities, making it possible to suggest a demand
and supply function for library services at the municipal level. Such an analysis may
make it possible to derive some limits for whether the public library service can

13
Compared with the other Scandinavian countries, the costs used on Norwegian public libraries per
inhabitant are the lowest. This situation has been subject to media discussion.
 S. Aabø / Information Economics and Policy 17 (2005) 175–198 197

provide net value, e.g., a minimum number of inhabitants in the municipality and a
minimum amount of library funding per inhabitant.

Acknowledgements

The author thank the Norwegian Research Council for financial support and Jon
Strand, professor of economics at the University of Oslo, and St
ale Navrud at the
Agricultural University of Norway for valuable suggestions and comments.

References

Aabø, S., 1988. Bibliotek i samfunnsøkonomisk perspektiv (Statens bibliotekhøgskole, Oslo). Available
from <http://www-bib.hive.no/tekster/ekstern/aaboe/index.html> (last accessed April 2, 2004).
Aabø, S., 1998. Citizens’ assessment of public libraries. Nordisk Kulturpolitisk Tidskrift 3, 60–82.
Aabø, S., Audunson, R.A., 2002. Rational choice and valuation of public libraries: can economic models
for valuating nonmarket goods be applied on public libraries. Journal of Librarianship and
Information Science 34, 4–14.
Aabø, S., Strand, J., 2004. Public library valuation, nonuse values, and altruistic motivations. Library and
Information Science Research 26 (3) (2004) (in press).
Arrow, K.J. et al., 1993. Report of NOAA panel on contingent valuation. Federal Register 58,
4601–4614.
Aslib, 1995. Review of the public library service in England and Wales. Final report, Department of
National Heritage, London.
Audunson, R.A., 1995. Utvikling og testing av en kausalmodell for folkebibliotekbruk. Norsk tidsskrift
for bibliotekforskning 2, 59–78.
Audunson, R.A., 2001. Folkebibliotekenes rolle i en digital framtid. In: Audunson, R.A., Windfeld Lund,
N. (Eds.), Det siviliserte informasjonssamfunn. Fagbokforlaget, Bergen.
Benhamou, F., 1996. Is increased public spending for the preservation of historic monuments inevitable?
The French case. Journal of Cultural Economics 20, 115–131.
Bille Hansen, T., 1997. The willingness-to-pay for the Royal Theatre in Copenhagen as a public good.
Journal of Cultural Economics 21, 1–28.
Blamey, R.K., Bennett, J.W., Morrison, M.D., 1999. Yea-Saying in contingent valuation surveys. Land
Economics 75, 126–141.
Boyle, K.J., Bergstrom, J.C., 1999. Doubts, doubts, and doubters: the genesis of a new research agenda?
In: Bateman, I.J., Willis, K.N. (Eds.), Valuing Environmental Preferences: Theory and Practice of the
Contingent Valuation Method in the US, EU, and Developing Countries. Oxford University Press,
Oxford.
D’Elia, G., 1980. The development and testing of a conceptual model of public library user behavior.
Library Quarterly 50, 410–430.
D’Elia, G., 1993. The roles of the public library in society: the results of a national survey. Urban Libraries
Council, Evanston, IL.
Hanemann, W.M., 1991. Willingness to pay and willingness to accept: how much can they differ?
American Economic Review 81, 635–647.
Harless, D.W., Allen, F.R., 1999. Using the contingent valuation method to measure patron benefits of
reference desk service in an academic library. College and Research Libraries 60, 56–69.
Holt, G.E., Elliott, D., Moore, A., 1999. Placing a value on public library services. Public Libraries 38, 98–
108.
Holt, G.E., Elliott, D., 2003. Measuring outcomes: applying cost-benefit analysis to middle-sized and
smaller public libraries. Library Trends 51, 424–440.
198 S. Aabø / Information Economics and Policy 17 (2005) 175–198

Kingma, B.R., 1996. The economics of information: a guide to economic and cost-benefit analysis for
information professionals. Libraries Unlimited, Englewood, Colo.
MacDonald, H.F., Bowker, J.M., 1994. The endowment effect and WTA: a quasi-experimental test.
Journal of Agricultural and Applied Economics 26, 545–551.
Martin, F., 1994. Determining the size of museum subsidies. Journal of Cultural Economics 18, 255–270.
McClure, C.R., Fraser, B.T., Nelson, T.W., Robbins, J.B., 2001. Economic benefits and impacts from
public libraries in the State of Florida. Information Use Management and Policy Institute, School of
Information Studies, Tallahassee, FL. Available from <http://dlis.dos.state.fl.us/bld/finalreport/> (last
accessed April 2, 2004).
Mercer, C., 1995. Navigating the Economy of Knowledge: A National Survey of Users and Non-users of
State and Public Libraries. Griffith University, Brisbane.
Mitchell, R.C., Carson, R.T., 1989. Using Surveys to Value Public Goods: The Contingent Valuation
Method. Resources For the Future, Washington DC.
Morris, A., Hawkins, M., Sumsion, J., 2001. The economic value of public libraries. Resource: The
Council for Museums, Archives and Libraries, London.
Navrud, S., Ready, R.C. (Eds.), 2002. Valuing Cultural Heritage. Elgar, Cheltenham.
Noonan, D., 2003. Contingent valuation and cultural resources: a meta-analytic review of the literature.
Journal of Cultural Economics 27, 159–176.
Norsk kulturbarometer 1997, 1998, Ukens statistikk nr 22, Statistisk sentralbyr a, Oslo.
Norske folkebibliotek, 1998. Med bibliotekstatistikken 1997, Statens bibliotektilsyn, Oslo.
Papandrea, F., 1999. Willingness to pay for domestic television programming. Journal of Cultural
Economics 23, 149–166.
Public Library Act 1985. [2003], Act No. 108 of 20th December: Including latest amendments as of 4th
July 2003. Availble from <http://www.lovdata.no/all/hl-19851220-108.html#4> (last access April 2,
2004).
Randall, A., 1991. Total and nonuse values. In: Braden, J.B., Kolstad, C.D. (Eds.), Measuring the
Demand for Environmental Quality. North-Holland, Amsterdam, pp. 303–321.
Santagata, W., Signorello, G., 2000. Contingent valuation and cultural policy design: the case of Napoli
Musei Aperti. Journal of Cultural Economics 24, 181–204.
Statistics Norway, 1998. Statistisk arbok 1998, Statistisk sentralbyr
a, Oslo.
Throsby, D., 2003. Determining the value of cultural goods: how much (or how little) does contingent
valuation tell us? Journal of Cultural Economics 27, 275–285.
Throsby, C.D., Withers, G.A., 1983. Measuring the demand for the arts as a public good: theory and
empirical results. In: Hendon, W.S., Shanahan, J.L. (Eds.), Economics of Cultural Decisions. Abt.
Books, Cambridge, pp. 177–191.
TNS Gallup, 2003. Befolkningens vurdering av folkebibliotekene. ABM-utvikling, Oslo.
Welsh, M.P., Poe, G.L., 1998. Elicitation effects in contingent valuation: comparisons to a multiple
bounded discrete choice approach. Journal of Environmental Economics and Management 36, 170–
185.
Zweizig, D., Dervin, B., 1977. Public library use, users, uses: advances in knowledge of the characteristics
and needs of the adult clientele of American public libraries. Advances in Librarianship 7, 231–255.