public preferences regarding rabies-prevention policies in the uk
TRANSCRIPT
Public preferences regarding rabies-prevention
policies in the UK
Mark Cox*, Edward B Barbier, Piran C.L. White,Geraldine A. Newton-Cross, Lorraine Kinsella,
Heather J. Kennedy
Environment Department, University of York, Heslington, York YO10 5DD, UK
Accepted 19 April 1999
Abstract
The current 6-month quarantine system for all cats and dogs entering the UK has kept the UK
rabies-free since 1922. However, pressure is mounting for a change to a system of vaccination,
microchip identification and serological testing. In response to the increasing controversy
surrounding the quarantine system, the UK government recently set up an independent review panel
to assess the alternatives. This paper quantifies public preferences for the current policy and three
alternative rabies-prevention measures. A survey was used not only to assess the overall preferences
for rabies-prevention policies but also to assess the importance of policy attributes and socio-
economic characteristics in determining policy preferences. We interviewed a sample of pet-owners
in North Yorkshire. The results showed that the existing system was the single most-preferred
policy option. However, a large proportion of the sample preferred the vaccination-based policies. A
logistic-regression model and ordered probit models were used to find that safety and animal
welfare were the most-important factors determining policy preferences. The respondents'
awareness of the rabies-policy review, a desire to take a pet abroad, the amount of foreign travel,
occupation and previous experience of quarantine were all important factors in policy choice.
Socio-economic characteristics such as income, pets owned and the number of children were not
significant determinants of policy preference. # 1999 Elsevier Science B.V. All rights reserved.
Keywords: Rabies; Public preferences; Regional survey UK; Quarantine; Vaccination; Logistic regression;
Ordered probit
Preventive Veterinary Medicine 41 (1999) 257±270
* Corresponding author. Tel.: +44-1904-43-2999; fax: +44-1904-43-2998E-mail address: [email protected] (M. Cox)
0167-5877/99/$ ± see front matter # 1999 Elsevier Science B.V. All rights reserved.
PII: S 0 1 6 7 - 5 8 7 7 ( 9 9 ) 0 0 0 5 5 - 0
1. Introduction
In the 18th and 19th centuries, rabies was widespread throughout the UK (Waterhouse,
1971). The disease was eliminated by measures taken to control dogs, including
muzzling, elimination of stray dogs and the quarantine of imported dogs and cats (King
and Turner, 1993). Since 1922, the UK has been rabies-free (with the exception of two
dogs that became rabid after release from quarantine in 1969 and 1970) (Waterhouse,
1971).
The 6-month quarantine system has undoubtedly been effective. However, advances in
vaccination technology (BVA, 1997; Scott, 1997), a greater understanding of the potential
for rabies spread and control in the UK (Harris and Rayner, 1986; Smith and Harris,
1991; White et al., 1995) and the replacement of quarantine with vaccination in other
European countries (BVA, 1997; MAFF, 1997), has led the UK to re-consider its rabies-
prevention policy. Maintaining the current animal-quarantine policy has become a
controversial public issue; in response, the UK Government recently set up an
independent review panel to assess the alternatives (MAFF, 1997).
Much of the debate over alternatives to quarantine concerns the public perception of
potential changes in risk of a rabid animal entering the UK (BVA, 1997). A Royal Society
report stressed the importance of incorporating public perceptions of risk into policy-
making (The Royal Society, 1993). The quantification of public preferences is also more
frequently entering scientific work in the field of environmental protection, health and
conservation via the application of methods such as contingent valuation (Carson et al.,
1992; Whittington et al., 1993; Kerr and Cullen, 1995; White et al., 1997).
This paper summarises the results of a survey designed to quantify public preferences
for rabies-prevention policies and assess the importance of policy attributes such as
safety, animal welfare and cost. The analysis considers the overall preferences as well as
the importance of policy attributes and socio-economic characteristics of the respondents
in determining policy preference.
2. Materials and methods
2.1. Questionnaire design
We employed the `stated preference' method of survey design widely used in
environmental valuation, market research and transportation research. Respondents to
such surveys are often asked either to indicate their most-preferred choice, to rank options
from their most- to least-preferred (Beggs et al., 1981; Smith and Desvousges, 1986:
Lareau and Rae, 1989) or to rate options on a scale defined by the researcher (Mackenzie,
1993). The use of ratings and rankings is often associated with conjoint analysis (Greene
and Srinivasan, 1990). Ratings are preferred to rankings because respondents are more
familiar with providing ratings and can indicate preference intensity, order and express
indifference in their ratings (Mackenzie, 1993).
The policies the respondents were asked to rate represent the options under
consideration by the governmental review panel and proposed in the discussion
258 M. Cox et al. / Preventive Veterinary Medicine 41 (1999) 257±270
document `Quarantine for Pets' (MAFF, 1997). There were two quarantine-based
policies. The first was the existing 6-month quarantine system for all pets entering the UK
(`̀ total quarantine''). The second was a 1-month quarantine period, which would be
followed by five months restricted home movement for animals travelling within the EU
(`̀ partial quarantine''). Under this policy, animals from outside the EU would still be
required to undergo a 6-month quarantine. There were also two vaccination-based
policies. The first would allow animals to enter the UK from EU destinations based on a
system of vaccination, microchip identification and serological testing (`̀ partial
vaccination''). Under this policy, animals from outside the EU would still be required
to undergo a 6-month quarantine. The second vaccination-based policy would allow
animals (regardless of country of origin) to enter the UK under the same regulations as
partial quarantine, and would be backed up by compulsory vaccination of all pets in the
UK (`̀ total vaccination'').
The questionnaire was split into three sections. In the first section, the respon-
dents were provided with introductory information on rabies-prevention policies.
Respondents were then asked about: (1) their awareness of the governmental policy
review; (2) pet ownership; and (3) travel behaviour. In the second section, each
respondent was provided with an in-depth description of the policies and given two
tables (the information provided in these two tables is summarised in Table 1). The first
table gave a brief description of each policy. This table was given to the respondent,
whilst the interviewer read out the background information to the policies. The
respondent was then given the second table, which provided them with five extra pieces
of quantitative information: (1) the period of isolation the dog or cat has to undergo; (2)
the cost to the owner of a travelling pet; (3) the annual cost to the owner of a non-
travelling pet; (4) a relative safety rating; and (5) an animal welfare rating. Then the
respondents were asked to rate each policy on a scale of 1 (least desirable) to 10 (most
desirable) and to indicate which of the policy attributes they thought was the most
important in influencing their policy preference. The final section of the questionnaire
gathered information about the number of children, combined household income and
occupation.
2.2. Survey implementation
The fieldwork was carried out over a 3-month period from January to March 1997. A
pilot survey was conducted on a sample of 58 staff at the University of York. Initial
statistical analysis of the results led to no changes to final survey design. Face-to-face
interviews were used and the survey took, on an average, 10 min (median�10, 50th
percentile�10 and 90th percentile�11) to complete during both the pilot stage and final
surveying.
The surveying was carried out by a group of four researchers who were supervised
within a group consisting of the interviewers and two others. The interviewers were
required to read out the same information to each respondent and provide the tables at the
same time. Each interviewer conducted equal numbers of interviews (N�65). Weekly
meetings were held during the surveying and analysis periods to check questionnaires for
omissions or errors and deal with any problems as they arose.
M. Cox et al. / Preventive Veterinary Medicine 41 (1999) 257±270 259
Table 1A brief summary of the qualitative and quantitative information about the rabies-prevention policies provided to each respondent to the rabies questionnaire
Policy Isolation
(months)a
Travel
costbAnnual
cost
Safety
ratingc
Welfare
ratingd
Total quarantine: a 6-month quarantine for all cats and dogs entering the UK 6 EU, 6 others 1200 0 5 1
Partial quarantine: a 1-month quarantine followed by 5-months restricted
home movement for cats and dogs from EU countries. Cats and dogs from
other countries undergo 6-months quarantine
1 EU, 6 others 200 0 3 2
Partial vaccination: serological testing, vaccination and microchip identification
for cats and dogs from EU countries (no quarantine). Cats and dogs from
other countries still undergo 6-months quarantine
0 EU, 6 others 150 0 4 3
Total vaccination: serological testing, vaccination and microchip identification
for all cats and dogs entering the UK regardless of country of origin.
Compulsory vaccination of all cats and dogs resident in the UK
0 EU, 0 others 120 35 3 5
a Required period of isolation for all cats and dogs entering the UK. The period of isolation is dependent on country of origin. This is split by EU or all other countries.b Cost estimates from local veterinary practices and Morgan (1995).c A relative rating based on estimates of maximum risk of varying periods of quarantine compared to vaccination (Corrin and MacDiarmid, 1997) (1�the least safe;5�the most safe).d A relative rating based on the period of isolation that cats and dogs have to undergo on entering the UK (1�the poorest treatment of travelling animals; 5�the best-possible treatment).
26
0M
.C
ox
eta
l./Preven
tiveV
eterina
ryM
edicin
e41
(1999)
257±270
Only dog and cat owners were chosen for two reasons. The first reason was the
fact that the Ministry of Agriculture, Fisheries and Food stated that `̀ the cost of any
new system which might replace quarantine should be borne by pet owners, as is the
case with quarantine at the moment'' (MAFF, 1997). The second reason was that dog
and cat owners are the most likely to be affected by any change in rabies prevention
policy.
Veterinary practices around North Yorkshire were contacted to ask if they would co-
operate in administering the final survey. This required them to let one of our team
interview people in the waiting room of their practice. In the final survey, 260 dog and cat
owners were interviewed at 20 veterinary practices. The survey worked well in the field
and there was no cases of respondents refusing to complete the questionnaire or
apparently not understanding the information provided. The survey sample was small but
recent survey research has shown that providing carefully designed and implemented
surveys can be smaller and therefore cheaper with no adverse effect on the results
obtained (Harrison and Lesley, 1996). In addition, the sample size compares well with the
existing studies used as a guide for the design of our survey. Beggs et al. (1981) surveyed
193 individuals and Lareau and Rae (1989) surveyed 140 individuals.
2.3. Statistical methods
Two modelling tools (common to survey analysis) were used to evaluate the
respondents' choices between the two quarantine-based policies and the two vaccination-
based policies and the actual ratings provided to each policy. Analyses were performed
using LIMDEP 7 (Econometric Software, 1997).
Logistic regression was used to model the probability of an individual choosing a
vaccination-based policy instead of a quarantine-based policy as a function of each
individual's indicated policy attribute and their socio-economic characteristics. The
dependent variable in this model was created from the ratings. It was set equal to 1 if the
individual provided their highest rating to one of the two vaccination policies and equal to
0 otherwise.
The other modelling approach concentrated on separately analysing the actual ratings
provided to each policy. The logistic model fails to account for the ordinal nature of
ratings and ordinary regression would treat the difference between the levels as the same
when in fact they are only a ranking. The ordered probit model is a useful technique for
modelling discrete data when there are more than two possible outcomes and it has come
into fairly wide use as a framework for analysing ratings from survey responses (Zavoina
and McElvey, 1975; Greene, 1997). The ordered probit model exploits the fact that the
dependent-variable outcomes have a natural ordering (in other words, the responses can
be ordered in some qualitative fashion). The major advantage of the ordered probit model
is that by exploiting such a feature, the resulting model is relatively easy to estimate. The
down-side is that the behavioural model may be considered too restrictive.
The structure of the model can be understood by considering a sample of data (Yi, Xi) of
size n drawn independently from a population, where the dependent variable has M
possible outcomes, which have some natural ordering (that is, m�1 is in some sense
`̀ better'' than m). The observed values are assumed to derive from some unobservable
M. Cox et al. / Preventive Veterinary Medicine 41 (1999) 257±270 261
latent variable Y�i , where
Y�i � X0i� � Ui:
Xi represents the observable individual specific factors for some (k�1)-parameter vector
� and stochastic-disturbance term Ui. The M outcomes for observed variable Yi are
assumed to be related to the latent variable Y�i in the following manner:
Yi � 0 if Y�i � 0;Yi � 1 if 0 < Y�i � �1;Yi � 2 if �1 < Y�i � �2;� � �Yi � M if �mÿ1 � Y�:
This is a form of censoring. The �'s (�1<�2<� � �<�mÿ1) are unknown parameters to be
estimated with �. The respondents' intensity of feeling (their ratings) depend on certain
measurable factors (X) and certain unobservable factors Ui. Then, the conditional
probability of observing the mth category Yi�m can be written as
Pr Yi � m j Xi� � � Pr �mÿ1 < Y�i � �m
ÿ � � Pr �mÿ1 < X0i� � Ui � �m
ÿ �:
By rearranging to isolate Ui, we have
Pr Yi � m j Xi� � � Pr �mÿ1 ÿ X0i� < Ui � �m ÿ X0i�ÿ �
for m � 1; . . . ;M:
To evaluate the conditional probability, we simply specify a distribution for the
disturbance term Ui. The ordered probit assumes that Ui�N(0,1) (see Greene, 1997 or
Zavoina and McElvey, 1975). The conditional probabilities can be derived as
Pr Yi � m j Xi� � � � �m ÿ X0i�ÿ �ÿ � �mÿ1 ÿ X0i�
ÿ �:
These probabilities can be evaluated for any combination of parameters. In order to
isolate those parameters that best suit a sample of data, we exploit maximum likelihood
techniques. This will yield the following probabilities:
Pr Yi � 0� � � � ÿX0i�ÿ �
;Pr Yi � 1� � � � �1 ÿ X0i�
ÿ �ÿ � ÿX0i�ÿ �
;Pr Yi � 2� � � � �2 ÿ X0i�
ÿ �ÿ � �1 ÿ X0i�ÿ �
;� � �Pr Yi � M� � � � �mÿ1 ÿ X0i�
ÿ �:
A positive coefficient in the ordered probit model implies that as the value of the
explanatory variable increases, then so does the probability of the individual providing a
higher rating to the policy in question. The opposite is the case for a negative coefficient.
In our case, there were 10 possible outcomes or rating positions (the dependent variable
in these models takes the value of the rating provided by the individual). Each policy was
modelled separately using ordered probit techniques to analyse the important factors in
driving policy preference for each policy separately.
A forward-stepwise procedure was used to develop best-fit models for both the logistic
and ordered probit models. All of the explanatory variables, described in Table 2, were
262 M. Cox et al. / Preventive Veterinary Medicine 41 (1999) 257±270
tested in the model specifications. Each factor was included if it had a significant
p-value (<0.1) and improved the fit of model, which was tested using the likelihood-ratio
test.
3. Results
3.1. Policy preferences
The results of the questionnaire revealed that total quarantine was the single most-
preferred policy option (N�87 provided their highest rating to total quarantine).
However, combining the preferences for the vaccination-based policies suggests that most
respondents were in favour of a change to vaccination (N�150 provided their highest
rating to partial or total vaccination). The summary statistics for the ratings can be seen in
Table 3 (Table 2 provides a list of variables and descriptions). Partial vaccination
received the highest median rating followed by total vaccination and total quarantine
together. Fig. 1 shows that a large proportion of respondents appeared to be either
providing a rating of 10 (23%) or a rating of 1 (15%) to total quarantine and a similar
pattern was obtained for total vaccination (13% of respondents rated 10 and 19% rated 1).
Table 2The names, descriptions and values of the explanatory variables used in the logistic and ordered probit analysis
Variable name Description and values
AWARENESS Respondents awareness of the rabies policy review (Yes�1, No�0)
BREEDER Respondent comes from a household which has or does breed animals (Yes�1,
No�0)
CATS Number of cats that the respondent owns (range 1±6)
CHILDREN Number of children (under 16) in the household
DOGS Number of dogs that the respondent owns (range 1±7)
DOG AND CAT OWNER Respondent owns both dogs and cats (Yes�1, No�0)
FARMER Respondent comes from a farming household (Yes�1, No�0)
INCOME The household income category (£) (range 1±6). (1) <10 000, (2) 10 000±20 000,
(3) 20 000±30 000, (4) 30 000±40 000, (5) 40 000±50 000, (6) >50 000
ISOLATION Respondent indicated the period of isolation their animal has to undergo as the
most-important policy attribute (Yes�1, No�0)
LIKE TO TAKE Respondent would like to take his/her pet abroad (Yes�1, No�0)
RABIES-FREE Number of trips to countries on the WHO rabies-free list
RABIES-PRES Number of trips to countries not on the WHO rabies-free list
SAFETY Respondent indicated safety as the most-important policy attribute (Yes�1,
No�0)
TAKEN PET Respondent previously took a pet abroad and/or brought a pet back to the UK
through quarantine (Yes�1, No�0)
TRAVEL COST Respondent indicated travel cost as the most-important policy attribute (Yes�1,
No�0)
TRIPS The number of trips the respondent takes abroad in an average year (range 0.5±
20)
WELFARE Respondent indicated animal welfare as the most-important policy attribute
(Yes�1, No�0)
M. Cox et al. / Preventive Veterinary Medicine 41 (1999) 257±270 263
However, 50% of respondents rated partial vaccination in the 7±9 and most respondents
rated partial quarantine at 7 or below (87%).
3.2. Choice between quarantine and vaccination
Those individuals indicating animal welfare and travel cost as important policy
attributes preferred vaccination (Table 4). In addition, those individuals who would like
to take their pet abroad and those already travelling frequently to rabies-free destinations
preferred vaccination.
3.3. Policy ratings
The results of ordered probit models are presented in Table 5. The results show those
respondents indicating safety as the most important policy attribute rated total quarantine
highly (shown by the positive coefficient on SAFETY in model 1, Table 5). However, that
was only the case for total quarantine, because those concerned about safety did not rate
partial quarantine highly (illustrated by the negative coefficient on SAFETY in model 2,
Table 3The summary statistics for the nominal, ordinal and continuous variables involved in the logistic and orderedprobit analysis
Variable % Yes Min Percentiles Max N
25th 50th 75th 90th
AWARENESS 62 260
BREEDER 3 260
CATS 0 0 1 1 2 6 260
CHILDREN 0 0 0 1 1 2 260
DOGS 0 0.5 1 2 2.5 7 260
DOG AND CAT OWNER 19 260
FARMER 10 260
INCOME 1 2 3 4 5 6 239
ISOLATION 21 260
LIKE TO TAKE 38 260
RABIES-FREE 0 0 0 1 1.5 5 260
RABIES-PRES 0 0 1 1 2 7 260
RATING TQa 1 3 6 9 10 10 260
RATING PQb 1 3 5 7 8 10 260
RATING PVc 1 5 7 8 10 10 260
RATING TVd 1 2 5 8 10 10 260
SAFETY 48 260
TAKEN PET 3 260
TRAVEL COST 4 260
TRIPS 0 0.5 1 2 3 20 260
WELFARE 24 260
a Ratings given to total quarantine policy.b Ratings given to partial quarantine policy.c Ratings given to partial vaccination policy.d Ratings given to total vaccination policy.
264 M. Cox et al. / Preventive Veterinary Medicine 41 (1999) 257±270
Table 5). Safety was not a significant factor in the determination of ratings for partial and
total vaccination.
Respondents indicating animal welfare as their most-important policy attribute
generally rated total quarantine and partial quarantine low (shown by the negative
coefficients on WELFARE in models 1 and 2) and partial vaccination and total
vaccination highly (shown by the positive coefficients on Welfare in models 3 and 4).
Awareness of the Government's decision to review the current quarantine policy
generally led to lower ratings for partial vaccination (negative coefficient on
AWARENESS in model 3) and was not a significant determinant of preferences for
Fig. 1. A stacked histogram showing the frequency of ratings (1�least desirable to 10�most desirable) given to
each policy.
Table 4Summary of the logistic regression model analysing the factors determining the choice of the pet-owning surveyrespondents between quarantine-based and vaccination-based policies regarding rabies control (the dependentvariable takes the value 1 if the respondent provided the highest rating to a vaccination-based policy and is equalto 0 otherwise)
Variable b Se(b) p-Value Odds ratio 95% Confidence bounds for OR
Lower Upper
CONSTANT ÿ1.01 0.22 <0.01 0.3 ÿ0.1 0.7
TRAVEL COST 1.87 0.84 0.03 2.6 0.9 4.2
WELFARE 2.89 0.56 <0.01 8.1 7.0 9.2
LIKE TO TAKE 1.88 0.34 0.10 2.7 2.0 3.4
RABIES-FREE 0.34 0.21 <0.01 0.4 ÿ0.0 0.8
N�260; L(��0)�ÿ177.13; likelihood ratio test �2�49.13.
M. Cox et al. / Preventive Veterinary Medicine 41 (1999) 257±270 265
the other policies. Respondents expressing a desire to take their pet abroad rated partial
vaccination and total vaccination highly (positive coefficients on LIKE TO TAKE in
models 3 and 4) and rated total quarantine low (negative coefficient on LIKE TO TAKE
in model 1). LIKE TO TAKE was not significant in the ordered probit model for partial
quarantine (model 2).
Frequent travellers out of the UK rated total quarantine and partial quarantine low
(negative coefficient on TRIPS in models 1 and 2) and partial vaccination highly (positive
coefficient on TRIPS in model 3) and TRIPS was not significant in the determination of
preferences for total vaccination. However, the significance of RABIES-PRES in model 4
suggests that those travelling to destinations with endemic rabies did not rate total
vaccination highly. Farmers generally did not rate partial vaccination highly (negative
coefficient on FARMER in model 3). Those who had previously put a pet in quarantine
were more likely to rate total vaccination highly (positive coefficient model 4). Finally,
those respondents with dogs gave low ratings to partial quarantine. However, rather than
implying an actual trend between the number of dogs owned and ratings of partial
Table 5Summary of the ordered probit models analysing the factors determining the ratings given to the totalquarantine, partial quarantine, partial vaccination and total vaccination policies regarding rabies prevention bythe pet-owning survey respondents (the dependent variable takes the value of the rating given to each policy,respectively, by the respondent)
Variable b Se(b) p-Value 95% Confidence bounds for b
Model 1: total quarantine (N�260, �2�123.48)
CONSTANT 1.88 0.18 <0.01 1.5 2.2
LIKE TO TAKE ÿ0.79 0.16 <0.01 ÿ0.5 ÿ1.1
SAFETY 0.39 0.16 0.01 0.1 0.7
TRIPS ÿ0.09 0.05 0.05 ÿ0.2 0.0
WELFARE ÿ1.12 0.22 <0.01 ÿ1.6 ÿ0.7
Model 2: partial quarantine (N�260, �2�18.49)
CONSTANT 1.83 0.18 <0.01 1.5 2.2
DOGS ÿ0.09 0.02 0.01 ÿ0.1 ÿ0.0
SAFETY ÿ0.49 0.15 0.01 ÿ0.8 ÿ0.2
TRIPS ÿ0.12 0.05 0.03 ÿ0.2 ÿ0.0
WELFARE ÿ0.38 0.19 0.06 ÿ0.8 ÿ0.0
Model 3: partial vaccination (N�260, �2�42.49)
CONSTANT 1.62 0.17 <0.01 1.3 1.9
AWARENESS ÿ0.22 0.14 0.10 ÿ0.5 0.1
FARMER ÿ0.50 0.19 0.01 ÿ0.9 ÿ0.1
LIKE TO TAKE 0.57 0.15 <0.01 0.3 0.9
TRIPS 0.08 0.04 0.07 0.0 0.2
WELFARE 0.34 0.18 0.06 ÿ0.0 0.7
Model 4: total vaccination (N�260, �2�69.03)
CONSTANT 0.67 0.12 <0.01 0.4 0.9
LIKE TO TAKE 0.49 0.14 0.00 0.2 0.8
RABIES-PRES ÿ0.10 0.05 0.04 ÿ0.2 ÿ0.0
TAKEN PET 0.90 0.41 0.03 0.1 1.7
WELFARE 1.02 0.16 <0.01 0.7 1.3
266 M. Cox et al. / Preventive Veterinary Medicine 41 (1999) 257±270
quarantine, this was probably due to the lack of people across the whole sample rating
partial quarantine highly.
The models only report the significant factors determining policy ratings for each
policy separately, therefore factors important in deciding preference for one policy might
not affect another. None of the models picked up significant relationships between the
period of isolation, pet ownership characteristics, the number of children in a household,
a dog or cat breeding household or income and policy preference.
4. Discussion
This survey has provided an insight into the preferences of a sample of North Yorkshire
(UK) dog and cat owners towards the existing rabies-prevention policy in the UK and
three alternative policies that are currently being considered by the UK government. Only
cat and dog owners were surveyed due to their direct involvement in this issue. However,
a limitation of the survey was that non-pet-owners' preferences were not included. Rabies
prevention is a public issue that concerns all the population and future work should look
towards incorporating non-pet-owners' preferences.
The survey was conducted on a relatively small sample in only one area of the UK
(North Yorkshire). Therefore, the results should only be extrapolated to the UK with
caution. Survey implementation involved visiting only those veterinary practices in the
sample area that responded positively to our initial enquiries for co-operation. This might
have led to some bias in the results, but any such biases were unavoidable and cannot be
easily quantified.
The survey design followed a well-established technique common to many uses of
surveys. Ratings provide an easy method for respondents to express their preference on
complex issues. In fact, no respondents failed or refused to complete the questionnaire
correctly (N�21 refused to reveal an estimate of yearly household income). The analysis
used logistic regression and ordered probit models, which identified the most statistically
significant factors from those collected by the survey which determined policy
preferences. When modelling preferences, there will always be unobservable influences.
In this case, it could be individual characteristics or policy attributes, but it is impossible
to determine what these are without further surveying. Both of the modelling approaches
contain a stochastic component (error term), which effectively assumes the part of the
unobservable effects on preference but any important factors not included in the analysis
remain a fault in survey design.
The sample shows that most respondents preferred the existing rabies prevention policy
(N�87). However, it appears that the length of time an animal spends in quarantine was
seen to be more important than the system itself, which is borne out by the fact that so
few people preferred partial quarantine (N�15). We suggest that partial quarantine was
the least-favoured policy due to doubts over the enforceability of home restrictions and
the increased risk of only one month in kennels. Despite the overall preference for total
quarantine as a single policy, the vaccination-based policies were preferred overall
(N�150) implying that the majority of pet owners are in favour of a change to
vaccination. Fig. 1 shows the extreme ratings given to total quarantine which suggest that
M. Cox et al. / Preventive Veterinary Medicine 41 (1999) 257±270 267
respondents were either very satisfied or extremely dissatisfied with the existing policy.
The same motivations were evident in the ratings of total vaccination.
The ordered-probit analysis showed that respondents who were aware of the
governmental policy review rated both partial quarantine and partial vaccination low. It
can only be hypothesised that they preferred total vaccination (as AWARENESS was not
significant in the ordered probit analysis of total vaccination). However, perhaps due to
the adverse publicity that quarantine receives, the positive reports about vaccination and
the growing influence of pressure groups who favour change (RSPCA, 1998), it might be
expected that those who were aware of the issue would favour vaccination. The ordered
probit analysis for partial vaccination showed that farmers opposed a change to a
vaccination-based policy. This could be because of the potentially catastrophic economic
effects on their livelihood should rabies enter the UK wildlife population (Smith and
Baer, 1988).
Many popular European holiday and working destinations for UK citizens are rabies-
free suggesting that many would prefer the freedom to travel with their pet and feel safe
to do so. However, much of Europe despite declining incidence of the disease (WHO,
1994) is still not rabies-free. Therefore, those travelling to rabies-present destinations
close to the UK may be unsure of the risks and prefer to leave their pets behind.
Consequently, they favour the safety of quarantine. Much of Africa, India, Southeast
Asia, Central and South America still has a high incidence of rabies in urban animals and
wildlife (presenting a much-higher risk). Therefore, those people who have previously
and frequently do travel to these rabies-present destinations may be even more risk-averse
and therefore have a strong preference against vaccination and in particular total
vaccination.
The UK Government has already gone part of the way to accepting vaccination by
allowing animals for trade to enter the UK based on vaccination, blood testing and
certification under the Balai directive. Vaccination (if administered correctly) will rarely
fail and many authors suggest that with the development of antibody tests, it can be
equally as safe as quarantine (Eng and Fishbein, 1990; Corrin and MacDiarmid, 1997;
Scott, 1997). However, our analysis shows that public support for a quarantine policy
with a proven record of safety is still strong. Unless the safety of vaccination can be
assured, many pet-owners appear willing to forego the freedom and animal-welfare
benefits that would come from a change to a vaccination-based policy.
Should rabies become endemic, control can cost large amounts of public money as
shown by the US (Rupprecht et al., 1996) and worldwide experience with rabies
prevention and control (Meltzer and Rupprecht, 1998a, b). The maintenance of the
current rabies-free status of the UK is therefore a very-important public issue. Public-
preference assessment should therefore play a role in the policy-making process along
with the scientific, veterinary and economic information crucial to rabies control. Levine
and Levine (1997) show that public preferences can play an integral role in developing
disease-control policies. The recent literature yields few specific studies that attempt to
quantify and use public preferences to guide disease control policy making. However,
public preference studies have been used recently in related issues, such as understanding
the public's responses and behaviour to disease management programmes such as
tuberculosis (Liefooghe et al., 1997), malaria (Gessler et al., 1995) and schistosomiasis
268 M. Cox et al. / Preventive Veterinary Medicine 41 (1999) 257±270
(Jemaneh et al., 1996). They have also been used to help understand the perception of
disease and its severity (McGull et al., 1997). In conclusion, this study provides not only
important results for rabies prevention policies but also an incentive to carry out further
work in the area of public perception of risk of animal diseases (transmissible to humans)
and their control.
Acknowledgements
Mark Cox and Geraldine Newton-Cross are grateful to the Natural Environment
Research Council for financial support.
References
Beggs, S., Cardell, S., Hausman, J., 1981. Assessing the potential demand for electric cars. J. Econometrics 16,
1±19.
British Veterinary Association, 1997. Rabies and quarantine: report of the BVA working group. Vet Rec. 19, 79±
81.
Carson, R.T., Mitchell, R.C., Hanemann, W.M., Kopp, R.J., Presser, S., Ruud, P.A., 1992. Contingent valuation
study of lost passive use values resulting from the Exxon Valdez oil spill. Attorney General of the State of
Alaska, Anchorage 1992.
Corrin, K., MacDiarmid, S., 1997. The risks of introducing rabies through importation of dogs. Surveillance 24,
22±25.
Econometric Software, 1997. LIMDEP Version 7.0 Users Manual. Econometric Software.
Eng, T.R., Fishbein, D.B., 1990. Epidemiological factors, clinical findings and vaccination status of rabies in
cats and dogs in the United States in 1988. JAVMA 197, 201±209.
Gessler, M.C., Msuya, D.E., Nkunga, M.H.H., Schar, A., Heinrich, M., Tammer, M., 1995. Traditional healers in
Tanzania: the perception of malaria and its causes. J. Ethnopharmacol. 48(3), 119±130.
Greene, W.H., 1997. Econometric Analysis, 3rd ed. Prentice-Hall, London.
Greene, P.E., Srinivasan, V., 1990. Conjoint analysis in marketing: new developments with implications for
research and practice. J. Mkt. 3, 19.
Harris, S., Rayner, J.M.V., 1986. Models for predicting urban fox (Vulpes vulpes) numbers in British cities and
their application for rabies control. J. Animal Ecology 55, 593±603.
Harrison, G.W., Lesley, J.C., 1996. Must contingent valuation surveys cost so much? J. Environ. Econ. Manage.
31, 79±95.
Jemaneh, L., Shewankena, F., Teolla, S., 1996. The use of questionnaires for the identification of high risk areas
for urinary schistosomiasis: the Ethiopian experience. Ethiop. Med. J. 34(2), 93±105.
Kerr, G.N., Cullen, R., 1995. Public preferences and efficient allocation of a possum-control budget. J. Environ.
Manage. 43, 1±15.
King, A.A., Turner, G.S., 1993. Rabies: a review. J. Comput. Path. 108, 1±39.
Lareau, T.J., Rae, D.A., 1989. Valuing WTP for diesel odour reductions: an application of contingent ranking
technique. SECJA 55, 728±742.
Levine, M.M., Levine, O.S., 1997. Influence of disease burden, public perception and other factors on new
vaccine development, implementation and continued use. LANCA 350(9088), 1386±1392.
Liefooghe, R., Baliddawa, J.B., Kirupto, E.M., Vermeire, C., DeMunynck, A.O., 1997. From their own
perspective. A Kenyan community's perception of tuberculosis. Trop. Med. 2(8), 809±821.
Mackenzie, J., 1993. A comparison of contingent preference models. Am. J. Agr. Econ. 75, 953±1603.
McGull, M.W., Kreindel, S.M., DeMaria, A., Rupprecht, C., 1997. Knowledge and attitudes of residents in two
areas of Massachussetts about rabies and an oral vaccination programme in wildlife. JAVMA 211(3), 305±
309.
M. Cox et al. / Preventive Veterinary Medicine 41 (1999) 257±270 269
Meltzer, M.I., Rupprecht, C.E., 1998a. A review of the economics of the prevention and control of rabies. Part 1:
Global impact and rabies in humans. Pharmacoeconomics 14(4), 365±383.
Meltzer, M.I., Rupprecht, C.E., 1998b. A review of the economics of the prevention and control of rabies. Part 2:
Rabies on dogs, livestock and wildlife. Pharmacoeconomics 14(5), 481±498.
Ministry of Agriculture, Fisheries and Food, 1997. Quarantine for Pets: A Discussion Document. MAFF
Publications, London.
Morgan, D.R., 1995. The BMA guide to rabies. Radcliffe Medical Press, Oxford.
RSPCA, 1998. The case for change. A report jointly published by the RSPCA, Passport for pets and vets in
support of change.
Rupprecht, C.E., Smith, J.S., Krebs, J., Niezgoda, M., Childs, J.E., 1996. Current issues in rabies prevention in
the United States: health dilemmas, public coffers, private interests. Public Health Rep. 111, 400±407.
Scott, G.R., 1997. Dissemination of rabies. Vet. Rec. 25, 452.
Smith, J.S., Baer, G.M., 1988. Epizootiology of rabies: the Americas. In: Campbell, J.B., Charlton, K.M. (Eds.),
Rabies. Kluwer Academic Publishers, Boston, pp. 267±300.
Smith, V.K., Desvousges, W.H., 1986. Measuring Water Quality Benefits. Norwell, Kulwer-Nijhoff, Mass.
Smith, G.C., Harris, S., 1991. Rabies in urban foxes (Vulpes vulpes) in Britain: the use of spatial stochastic
simulation model to examine the pattern of spread and evaluate the efficiency of different control regimes.
Phil. Trans. R. Soc. London B 334, 459±479.
The Royal Society, 1993. Risk: analysis, perception and management. Report of a Royal Society study group.
The Royal Society, London.
Waterhouse, R., 1971. Report of the committee inquiry on rabies: final report. Her Majesty's Stationary Office
(HMSO), London.
White, P.C.L., Harris, S., Smith, G.C., 1995. Fox contact behaviour and rabies spread: a model for the estimation
of contact probabilities between urban foxes at different population densities and its implications for rabies
control in Britain. J. Appl. Ecol. 32, 693±706.
White, P.C.L., Gregory, K.W., Lindley, P.J., Richards, G., 1997. Economic values of threatened mammals in
Britain: a case study of the otter Lutra lutra and the water vole Arvicola terrestris. Biol. Cons. 82, 345±354.
Whittington, D., Lauria, D.T., Wright, A.M., Choe, K., Hughes, J.A., Swarna, V., 1993. Household demand for
improved sanitation services in Kumasi, Ghana: a contingent valuation study. Water Resources Res. 29,
1539±1560.
WHO, 1994. World survey of rabies 28: for the year 1992. Veterinary Public Health Unit, Division of
Communicable Diseases, World Health Organisation, Geneva, Switzerland.
Zavoina, R., McElvey, W., 1975. A statistical model for the analysis of ordinal level dependent variables. J.
Math. Socio., Summer 103.
270 M. Cox et al. / Preventive Veterinary Medicine 41 (1999) 257±270