Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
Summary
Using econometric evidence, this article confirms that distribution of medicines online is split into two market
segments of very diverse quality, and identifies the factors that drive quality and quality assurance in this activity.
Unlike fraudulent, ‘rogue,’ websites, which offer scant guarantees and usually sell just a few medicines without
prescription, online pharmacies offering insurance coverage and linked to conventional pharmacies typically sell a
whole range of drugs, require third-party medical prescriptions and provide abundant information to patients. It is
shown that, where online pharmacies are allowed to act legally, market forces enhance quality, as private insurers
require professional standards, and specialized third parties make a business of certifying them. Furthermore, older
online pharmacies and those running conventional operations offer higher quality, probably because of reputational
investments. Overall, this evidence supports licensing online pharmacies, especially considering that prohibiting
them is ineffective against fraudulent sites. Copyright # 2003 John Wiley & Sons, Ltd.
*Correspondence to: Department of Economics and Business; Universitat Pompeu Fabra; Trias Fargas, 25, Barcelona 08005,
Spain. E-mail: benito.arrunada@upf.edu
Different regulations and weak enforcement The simultaneous presence of licensing and
ineffective enforcement leads online distribu-
In the USA, online pharmacies have been allowed
tion of medicines to split into two market
to operate on essentially the same basis as ordinary segments, with very diverse levels of quality: one
‘brick and mortar’ pharmacies [1]. This freedom,
segment of high-quality online pharmacies and
together with greater development of the Internet
another of rogue sites. Their standard processes
and more fragmented and market-driven health can be described by reference to their extreme
care and insurance, has allowed online distribution
types.
of medicines to become relatively more developed
High-quality online pharmacies are licensed and
there.
comply with regulations like any other pharmacy,
In contrast, only three European countries
from which they differ only because they serve
(Denmark, Netherlands and the UK) allow
their customers online. Typically, each user first
distance selling of medicines. According to the e-
opens an account, so must send in the details of his
commerce directive,a the pioneer online pharma-
credit card and medical insurance. After registra-
cies established in these countries are supposedly
tion, each patient is assigned to a specific team of
allowed to sell to buyers from other countries of
pharmacists; then, either the patient or his doctor
the EU. However, national authorities are increas- sends in a valid prescription by conventional mail,
ingly constraining such practice. A prominent
fax or, increasingly, by certified e-mail. The
example can be seen in the Doc.Morris case in
pharmacy then sends the medicines by mail from
Germany, still under litigation.b Paradoxically, a central warehouse or from a conventional
the European Union strives to facilitate e-com-
pharmacy close to the patient’s home from which
merce by all means [7, 8] except, in the case of
the patient may also collect them personally.
e-pharmacy, for lifting the regulatory constraints
Medicines usually take 2 or 3 days to arrive
that make it impossible in most of its member
for which the patient does not incur additional
countries.
mail costs, or can be delivered in 24 hours at a
Given the difficulties for effective enforcement,
special cost for urgent delivery. Most of these
prohibition of online pharmacies has not stopped
pharmacies also have permanent channels of
illegal trading, however. Enforcement is seriously
communication so that customers can consult
hindered in e-commerce because it breaks the
their pharmacists either by a free phone service
territorial correspondence between enforcers and or by e-mail [9, 10].
the regulated. Providers are difficult to control
Low-quality sites, on the contrary, substitute
when their presence in the territory is confined to
independent medical prescriptions by online
the computer screen of users. The end result is that
questionnaires of doubtful effectiveness. Some
prohibition hinders the existence of reliable online
operators even advertise the fact that no pre-
operations and not only protects conventional
scription is required. (‘No prescription? No
pharmacies but may also indirectly benefit rogue
doctor? No problem’). Those that use question-
sites, as an increasing number of customers use
naires ask patients for basic personal details,
them to buy legal medicines that could be provided
such as their age and whether they are following
by reliable online sellers, if only they were allowed
any treatment or have any allergies, or whether
to operate. they have a family or personal background of
certain illnesses, such as heart disease and
cancer. They then provide a text in which they
explain that, in order to make the purchase,
Different standards of practice the customer must be over 18, they recom-
mend patients to consult their doctors about
Two factors, therefore, determine the odd struc- the use of the drugs and, finally, they ask
ture of this market with respect to quality of the patients to accept liability for the con-
service. Firstly, some countries have allowed sequences of the purchase and use of the
licensed online pharmacies to be created. Sec- medicines. Most of these sites state that the
ondly, enforcement of prohibitions to sell medi- questionnaires are checked by a doctor who
cines online from unlicensed sites is weak in all decides on whether or not the medicine is
countries, due to the novelty and difficulties of appropriate and whether the order should go
controlling e-commerce. through or not.
Copyright # 2003 John Wiley & Sons, Ltd. Health Econ. 13: 329–344 (2004)
Quality Safeguards and Regulation of Online Pharmacies 331
Copyright # 2003 John Wiley & Sons, Ltd. Health Econ. 13: 329–344 (2004)
332 B. Arrunada
examined a clinical questionnaire. The simulated tion and therefore increases its intensity. For
purchaser was a woman aged 69, with difficulty in example, it may be less costly for many patients
experiencing orgasm, obesity, coronary insuffi- to ask for advice by e-mail than in person and it
ciency and high blood pressure and who was may be more convenient for the pharmacist to use
taking captopril, pravachol, athenolol and ery- the same method because it enables him or her to
thromycin. She requested 66 pills from 10 of the 11 schedule his time better. This, together with a
pharmacies requiring a questionnaire, 8 of which wider use of computer databases, may increase the
were based in the United States. Three of these amount and quality of professional advice instead
companies sold the pills, in spite of the serious of reducing it.
contraindications, and seven rejected the order.
The operation of these rogue sites, seemingly
selling drugs of dubious quality without effective
control, has subjected the sector to heavy criticism.
Experienced regulators of online pharmacies have Analytical framework
a more nuanced view, however. First, it is clear
that the main problems of e-commerce medicine Quality assurance and testable hypotheses
sales come from the poor-quality service provided
by fraudulent online operators, who use the new The economics of quality assurance is based on
sales channel to carry out illegal transactions, distinguishing a set of specific problems and
selling unapproved, contaminated, false or expired safeguards. Problems are defined in terms of the
drugs [9]. Furthermore, not only are such practices type of information asymmetry extant between the
far from being general, but their health effects are parties and their unique or repetitive interaction
unclear. Even the FDA accepts that there are very [17, pp. 106–26]. Safeguards are mainly based on
few known cases of real damage being caused by the threat of losing quasi-rents linked to previous
the sale of medicines on line [9]. It seems that such investments in real or reputational assets [18, 19].
rogue sites do not sell narcotics, stimulants, Most medicines present a mix of ‘experience’
depressants and anabolic steroids that are con- attributes [20], whose quality is only known after
trolled substances regulated in the US by the Drug consumption (pain relief, delivery time), and
Enforcement Administration.f To make things ‘credence’ attributes [21], for which quality is
even more ambiguous, allowing users to get round never known (hidden effects). This, together with
regulatory obstacles is not always inefficient. This the need of expertise for the evaluation of
may well be the case when regulators get into the attributes, generates substantial scope for infor-
habit of delaying authorization of new medicines mation asymmetry. This asymmetry is the main
as a silent way of saving pharmacy costs. reason behind normative arguments for licensing
What seems clear, in any case, is that the risks and other conventional regulation. Licensing not
associated with rogue sites are not present when only controls quality directly but it also defers
dealing with high-quality online pharmacies. These compensation, by creating a stream of quasi-rents
may even provide better quality and control than that helps prevent the provision of low-quality
conventional European pharmacies. First, because services to uninformed users.
online pharmacies enjoy more powerful incentives In the case of online pharmacies, identical
to comply due to their larger scale, the role of mechanisms are in place for pharmacies which
reputation (when compared to the typical Eur- are licensed by a public agency or whose sales are
opean pharmacy owned by an individual) may covered by insurers. Both of these decisions,
take on greater importance, and policing their licensing and providing insurance coverage, are
behaviour is easier. Second, their technology also made by third parties with an interest in avoiding
allows them to provide a service of high quality. low quality. Both also create quasi-rents to the
The main alleged defect of online pharmacies in extent that losing a license or the coverage of
this respect is the lack of direct contact between insurers would cause substantial losses.
patient and pharmacist. Personal contact certainly In addition, two other drivers of quality may be
diminishes, but this does not necessarily imply a present as a consequence of pure market forces.
reduction in clinical activity. Personal consultation First, some online pharmacies put great amounts
is always restricted for reasons of cost whereas the of reputational capital in play, the value of which
online pharmacy reduces the cost of communica- would be destroyed if they were caught cheating.
Copyright # 2003 John Wiley & Sons, Ltd. Health Econ. 13: 329–344 (2004)
Quality Safeguards and Regulation of Online Pharmacies 333
This is clearer for those which also run conven- prescription (FrgnSales); if it sells or not other
tional operations and cover their online operations products in addition to medicines (OtherPrd); the
under the same reputational ‘umbrella’. Second, number of different products it sells (Products);
older online pharmacies can be expected to and if it sells or not generic drugs (Generics). In
provide greater quality because of reputational addition, the cost of express delivery service
and learning effects. On the one hand, given the (UrgntFee) is also considered. All these variables
presence of experience attributes in the purchase of (with the exception of the urgency fee, because of
medicines, reputation will tend to accumulate with the lower number of observations available) were
each instance that an online pharmacy delivers on used to build a global index of service quality
its promises so, the longer a pharmacy is in (QualIndex) using principal components analysis,
business, the better known and more valuable its as reported in Table 2.
trade name will be.g On the other hand, substantial Second, the information supplied to users is
learning economies can be expected in a new proxied through five dummy variables. Three of
activity like this, and these economies will also be them represent the different kinds of information
greater in older operators. channels, consisting of a phone line or e-mail
In summary, the following four hypotheses will address to deal with problems related to purchases
be tested in this article: and deliveries (PhoneLine and E-mail variables
respectively), or a service that allows buyers to
1. Pharmacy type: Two types of online pharmacies consult about the use of medicines (AskPharm),
deliver widely different quality. plus the total number of information channels
2. Insurance coverage: Quality is higher for phar- available (InfoIndex). Two other dummies con-
macies with insurance coverage. sider the provision of pricing information,
3. Reputation extension: Online quality is posi- either through lists of prices (PriceLis) and/or
tively related to reputational capital. price comparisons with competitive pharmacies
4. Age effect: Quality is positively related to the (PriceCom). An index of information quality
age of the pharmacy. (Info-Index) has also been built using principal
components analysis (weights are given in Table 3).
Third, external certification of quality is mea-
Data and methodology sured considering the availability of each different
accreditation via specific dummies (VIPPS, HON
A specific sample was built to test these hypoth- and VeriSign) that take value one when the
eses. It was drawn up by compiling the most pharmacy holds the corresponding certification,
outstanding characteristics of all online pharma- zero otherwise. The total number of certifications
cies listed by the Yahoo search engine.h The held by each site (Certifcs) is also measured,
information was obtained by direct visits and adding five other certificates that are less common
examination of their web sites during the months to these three.j
of February and March 2001. Visits were repeated The first hypothesis – on pharmacy type – will
for consistency in June 2002 and changes observed be tested using the variables in the previous sets of
were not statistically significant. attributes, while the three quality factors that
Business practices of online pharmacies were define hypotheses two to four will be proxied as
measured based on 17 attributes, each represented follows. In the second hypothesis, licensing and
by a dependent variable, and three additional insurance will be measured by third-party insur-
indices built from them. Table 1 defines the ance coverage, InsurCov, a dummy that takes the
variables considered and summarizes the descrip- value of one if insurers cover purchases of
tive statistics. Attributes considered are of three medicines in this pharmacy, and zero otherwise.
kinds: service quality, information and third-party This assumes that third-party insurers would not
certification. be willing to pay for purchases in unlicensed
First, the quality of services provided takes into pharmacies. This is somehow stricter than mere
account the following: if the pharmacy requires an licensing, but gets rid of potential difficulties
independent prescription (PrescRqd), without re- caused by the use of variable standards of rigor
lying fully on an internal questionnaire or not in licensing decisions, mainly where these decisions
requiring any prescription at all;i if it is willing or are made by regional or state authorities. The
not to sell in foreign markets products requiring a assumption is also supported by the evolution of
Copyright # 2003 John Wiley & Sons, Ltd. Health Econ. 13: 329–344 (2004)
334 B. Arrunada
Dependent variables
Quality of service
QualIndex Scores of principal 104 0.00 1.846 1:315 4.014
components analysis
of standardized PrescRqd,
FrgnSales, OtherPrd,
Products and Genericsb
PrescRqd Requires the buyer to 113 0.301 0.461 1 0
present a prescription,
without relying fully on an
internal questionnaire
FrgnSales Sells in foreign markets 104 0.702 0.460 1 0
products requiring
prescription
OtherPrd Sells other products in 113 0.142 0.350 1 0
addition to medicines
Products Number of products soldb 113 34.699 45.030 100 1
Generics Sells generic medicines 113 0.142 0.350 1 0
UrgntFee Additional cost of express 65 17.246 4.131 26 7
or overnight delivery
service (in US dollars)b
Information to users
InfoIndex Scores of principal compo- 113 0.000 1.258 3:123 3.549
nents analysis of
standardized PhoneLine,
E-mail, AskPharm,
PriceLis and PriceComb
PhoneLine Offers a telephone line to 113 0.522 0.502 1 0
help in problems related to
purchases and deliveries
E-mail Offers an E-mail address for 113 0.991 0.094 1 0
buyers to contact about
purchases and deliveries
AskPharm Offers a service for buyers to 113 0.345 0.478 1 0
consult about the use of
medicines, doses, etc.
PriceLis Presents lists of its prices 113 0.956 0.207 1 0
PriceCom Offers price comparisons 113 0.062 0.242 1 0
with competitive
pharmacies
Copyright # 2003 John Wiley & Sons, Ltd. Health Econ. 13: 329–344 (2004)
Quality Safeguards and Regulation of Online Pharmacies 335
Table 1 (continued)
Variable Explanationa Observa- Mean Standard Maximum Minimum
names tions deviation
Third party quality certification
Certifcs Sum of certifications 113 0.283 0.661 3 0
(including VIPPS, HON,
VeriSign and others; could
take integer values zero to
eight)b,c
VIPPS Holds the VIPPS quality 113 0.088 0.285 1 0
certification, created by the
National Association of
Boards of Pharmacy
HON Follows the code of conduct 113 0.053 0.225 1 0
of the foundation Health on
the Net
VeriSign Site registered as a VeriSign 113 0.080 0.272 1 0
Secure Site Program
a
All variables are dummies that take values one when true, zero otherwise, unless those marked.
b
Not a dummy variable.
c
In addition to VIPPS, HON and VeriSign, five other certifications are represented in the sample: NCPA, Hi-Ethics, TRUSTe,
BBBOnLine and ePublicEye.
online pharmacies in the US, in whose success paying special attention to the existence of one-
insurers played an essential role [26,12]. In the way causations by dummy variables (structural
third hypothesis, which tests whether service zeroes). Thirdly, a system of structural equations
quality increases with reputation, reputation will will be estimated to check that the results are not
be proxied by RealAssets, another binary variable substantially affected by potential endogeneity of
that takes value one when the firm also runs a any of the independent variables.
conventional pharmacy operation, and zero if it
operates only through the Internet. This variable,
therefore, captures reputational spillovers from or
to ‘brick and mortar’ activities.k In addition, the Results
existence of such spillovers will also be examined
by checking the prediction that pure online The diversity of online pharmacies
pharmacies rely more on third-party certification.
Finally, the joint effect of accumulated reputa- The descriptive statistics presented in Table 1
tional investments and learning economies, posited already show the diversity of business practices
in the fourth hypothesis, will be measured by the followed in the sale of medicines through the
time the seller has been operating online, measured Internet. Insurers cover purchases in 19.47% of all
in years (PharmAge). the online pharmacies in the sample. A minority
The empirical analysis will be based on several (13.27%) also run conventional operations. On
methods. Firstly, a cluster analysis will be used to average, they had been open for slightly more than
identify well-separated groups in the sample. These 1 year (13.7 months), with a maximum of 7 years.
groups will then be tested by comparing their A doctor’s prescription is required by 30.09% of
means in order to confirm the typological hypoth- them. Only a few (14.16%) sell other products in
esis. Secondly, quality attributes of online phar- addition to drugs but most of them (70.19%) are
macies will be regressed on three independent happy to sell in foreign markets. Around a third
variables (InsurCov, RealAssets and PharmAge) (31.86%) carry a full range of drugs,l in contrast to
which are provisionally considered as exogenous, another third (33.63%) who carry only one
Copyright # 2003 John Wiley & Sons, Ltd. Health Econ. 13: 329–344 (2004)
336 B. Arrunada
Table 2. Principal components analysis used to build the service quality index
Component Eigenvalue Difference Proportion Cumulative
1 3.40920 2.66557 0.6818 0.6818
2 0.74364 0.32318 0.1487 0.8306
3 0.42046 0.07426 0.0841 0.9147
4 0.34620 0.26570 0.0692 0.9839
5 0.08050 0.0161 1.0000
Eigenvectors
Variable 1 2 3 4 5
PrescRqd 0.51322 0.06843 0.07530 0.37672 0.76442
FrgnSale 0.43557 0.31256 0.71138 0.43878 0.11829
OtherPrd 0.40149 0.56575 0.66609 0.27334 0.01859
Products 0.49684 0.09502 0.00484 0.58785 0.63130
Generics 0.37262 0.75400 0.21111 0.49525 0.05280
Scoring Coefficients
Variable 1
PrescRqd 0.51322
FrgnSale 0.43557
OtherPrd 0.40149
Products 0.49684
Generics 0.37262
Note: Score based on unrotated principal components, 4 scorings not used.
Copyright # 2003 John Wiley & Sons, Ltd. Health Econ. 13: 329–344 (2004)
Quality Safeguards and Regulation of Online Pharmacies 337
Eigenvectors
Variable 1 2 3 4 5
PhoneLine 0.62425 0.07059 0.05812 0.35167 0.69157
E-mail 0.18742 0.49982 0.81065 0.23052 0.06907
AskPharm 0.59815 0.20443 0.22856 0.32514 0.66518
PriceLis 0.45076 0.16375 0.24090 0.84340 0.02553
PriceCom 0.11924 0.82255 0.47875 0.07846 0.27172
Scoring Coefficients
Variable 1
PhoneLine 0.62425
E-mail 0.18742
AskPharm 0.59815
PriceLis 0.45076
PriceCom 0.11924
Note: Score based on unrotated principal components, 4 scorings not used.
95.6
71.7
Dissimilarity measure
47.8
23.9
0
G1 G2 G3 G4 G5 G6 G7 G8 G9 G10 G11 G12 G13 G14 G15 G16 G17 G18 G19 G20
Copyright # 2003 John Wiley & Sons, Ltd. Health Econ. 13: 329–344 (2004)
338 B. Arrunada
Table 4. Characteristics and comparison of means of the two main clusters of online pharmacies
Group of low quality pharmacies Group of high quality pharmacies Mean comparison
Variable Number Mean Std. Dev. Number Mean Std. Dev. t test P > jtj
InsurCov 77 0 0 36 0.6111111 0.4944132 7.4152 0.0000
RealAssets 77 0 0 36 0.4166667 0.5 5.000 0.0000
PharmAge 77 0.8571429 0.8226127 36 1.75 1.679711 3.0243 0.0042
QualIndex 70 1.171907 0.3741161 34 2.412749 1.198849 17.0367 0.0000
PrescRqd 77 0.012987 0.1139606 36 0.9166667 0.280306 18.6367 0.0000
FrgnSales 70 0.9 0.3021661 34 0.2941176 0.4624973 6.9519 0.0000
OtherPrd 77 0 0 36 0.4444444 0.5039526 5.2915 0.0000
Products 77 4.168831 4.894687 36 100 0 171.8016 0.0000
Generics 77 0 0 36 0.4444444 0.5039526 5.2915 0.0000
UrgntFee 44 18.70455 2.833188 21 14.19048 4.781412 4.0039 0.0004
InfoIndex 77 0.5123301 0.8651883 36 1.095817 1.273691 6.8706 0.0000
PhoneLine 77 0.3766234 0.4877165 36 0.8333333 0.3779645 5.4365 0.0000
Email 77 0.987013 0.1139606 36 1 0 1.0000 0.3205
AskPharm 77 0.1428571 0.3522217 36 0.7777778 0.421637 7.8454 0.0000
PriceCom 77 0.0649351 0.2480271 36 0.0555556 0.2323107 0.1957 0.8454
PriceLis 77 0.987013 0.1139606 36 0.8888889 0.3187276 1.7943 0.0804
Certifcs 77 0.025974 0.1601008 36 0.8333333 0.9411239 5.1127 0.0000
VIPPS 77 0 0 36 0.2777778 0.4542568 3.6690 0.0008
HON 77 0 0 36 0.1666667 0.3779645 2.6458 0.0121
VeriSign 77 0.012987 0.1139606 36 0.2222222 0.421637 2.9279 0.0058
Note: Means are compared assuming unequal variances and using Welch formula [27].
Table 5. Dependent variables which are fully determined by some independent variables (‘structural zeroes’)
Dependent variables Independent dummy variables
Variable name Observations Values InsurCov ¼ RealAssets ¼
¼0 ¼1 ¼0 ¼1
PrescReq 113 0 79 0 79 0
1 12 22a 19 15a
PhoneLine 113 0 54 0 52 2
1 37 22a 46 13
E-mail 113 0 1 0 1 0
1 90 22a 97 15a
AskPharm 113 0 74 0 72 2
1 17 22a 26 13
VIPPS 113 0 91a 12 90 13
1 0 10 8 2
HON 113 0 91a 16 93 14
1 0 6 5 1
Observations 113 91 22 98 15
a
According to the sample, the value of the independent variable fully determines the dependent variable.
one of the values of some of the dependent macies in which purchases can be charged to
variables (PrescReq, PhoneLine, E-mail, Ask- insurers require an independent prescription,
pharm, VIPPS, HON) is fully determined by the provide an ‘Ask the Pharmacist’ information
InsurCov variable. In particular, all online phar- service and receive queries via e-mail. Similarly,
Copyright # 2003 John Wiley & Sons, Ltd. Health Econ. 13: 329–344 (2004)
Table 6. Determinants of quality attributes of online pharmaciesa
Dependent variables
Service Quality dimensions included Express Information Main specific Price information Total certificates Main specific
Quality in online pharmacies’ delivery index information channels provided online number of certifications
Index service quality index
Independent
variables
InsurCov 2.798nnn +b,nnnn 4.408nnn 6.814nnn 60.749nnn 0.874 4.958nnn 1.710nnn +c,nnnn +c,nnnn 0.004 2.006 4.801nnn +d,nnnn +d,e,nnnn 2.780nnn
(0.263) (1.110) (2.298) (7.324) (0.685) (1.199) (0.249) (1.301) (1.350) (0.861) (0.906)
RealAssets 1.782nnn +b,nnnn 2.871nnn 0.842 44.477nnn 2.170nnn 1.430 0.838nnn 1.070 2.336nn 0.320 2.016n 1.196 1.715n 1.567 0.182
(0.294) (0.963) (0.977) (8.313) (0.701) (2.032) (0.249) (0.907) (0.932) (1.251) (1.120) (0.845) (1.004) (1.223) (0.889)
PharmAge 0.311nnn 0.761nn 0.352 1.236nn 6.477nnn 0.311 0.334 0.052 0.192 0.230 0.718 1.160 0.383n 0.015 0.075 0.387
(0.078) (0.333) (0.233) (0.488) (2.180) (0.208) (0.357) (0.074) (0.190) (0.216) (0.574) (0.905) (0.229) (0.367) (0.366) (0.269)
Constant 1.145nnn 3.616nnn 2.586nnn 8.250nnn 9.573nnn 3.018nnn 17.871nnn 0.503nnn 0.641nn 1.953nnn 2.173nnn 3.465nnn 0.440 0.621 4.331nnn
(0.132) (0.707) (0.490) (2.826) (3.638) (0.520) (0.653) (0.124) (0.291) (0.390) (0.522) (0.828) (0.927) (0.939) (0.882)
Observations 104 85 104 113 113 113 65 113 91 91 113 113 113 22 22 113
R2 0.74 0.64 0.22 0.46
Pseudo-R2 0.15 0.50 0.71 0.22 0.02 0.10 0.05 0.24 0.42 0.11 0.09 0.29
a
Logit regressions except OLS in models (1), (5), (7) and (8) and ordered logit (13).
b
The variables InsurCov and RealAssets are dropped due to estimability caused by structural zeroes. As a consequence, only the 85 observations with InsurCov ¼ 0 and
RealAssets ¼ 0 are considered.
c
The variable InsurCov dropped due to estimability – only the 91 observations with InsurCov ¼ 0 are considered.
d
The variable InsurCov dropped due to estimability – only the 22 observations with InsurCov ¼ 1 are considered.
e
Standard errors in parentheses.
n,nn,nnn
Statistically significant at the 99%, 95% and 90% confidence level, respectively. nnnn Structural zeroes.
339
Table 7. Basic attributes of services provided by online pharmacies (estimation using interactive age variables)
Dependent variables
Service quality Information provided Third party certification
QualIndex QualIndex InfoIndex InfoIndex Certifcs Certifcs Certifcs
(1) (2) (6) (7) (3) (4) (5)
Independent variables
InsurCov 2.796nnn 2.745nnn 1.988nnn 1.751nnn 7.963nnn 8.095nnn 7.978nnn
(0.423) (0.261) (0.398) (0.242) (1.565) (1.560) (1.523)
RealAssets 2.385nnn 2.386nnn 1.269nnn 0.839nnn 0.082 2.080nn
(0.431) (0.429) (0.415) (0.282) (2.032) (0.933)
PharmAge 0.381nnn 0.374nnn 0.141 0.948nnn 0.908nnn 0.948nnn
(0.096) (0.084) (0.091) (0.300) (0.289) (0.299)
InsurCovnPharmAge 0.029 0.189 1.315nnn 1.375nnn 1.319nnn
(0.191) (0.184) (0.471) (0.468) (0.458)
RealAssetsnPharmAge 0.407n 0.402n 0.327 1.398 1.353nn
(0.214) (0.210) (0.209) (1.259) (0.587)
VIPPS and HON certificates are found only in Internet certificates seem to act as a substitute
pharmacies in which drugs can be charged to for extending pharmacies’ reputation from the
insurers.m As a consequence, logit models for conventional to the online segments of the market.
variables with such one-way causation are esti- (This effect on third-party certification, which is
mated by dropping the variables with zeroes and not statistically significant for the full sample,
excluding the observations fully determined by the becomes significantly negative when considering
dropped variables, in order to avoid biasing the the interactions with age, as shown in Table 7).
remaining coefficients. This affects models (2), (9), Thirdly, the age of the online operations is
(10), (14) and (15) in Table 6, in which the absence strongly related to most of the dependent vari-
of a coefficient, instead of indicating a lack of ables. Older online pharmacies are more likely to
statistical significance, means quite the opposite: a require a prescription, sell products other than
certain relation between the corresponding depen- medicines, carry a full line of medicines, sell
dent and independent variables. generics and be certified. Specific analysis using
There is also a positive connection between most interactive variables shows that the effect of age on
of these quality factors and the fact that firms also quality of service is concentrated amongst phar-
operate conventional pharmacies. What’s more, all macies operating only online (Table 7). Similarly,
online pharmacies running conventional opera- third party certification is more likely for older
tions also require prescriptions, as described in sites without insurance coverage or conventional
Table 5, causing another case of structural zeroes. operations. This is consistent with the idea that
They are also less likely to sell prescription drugs pure online players increase their quality with time
in foreign markets, and they stock more products but conventional firms diversifying online or
and generics, and provide better information to linked to insurers start their operations by
users. As predicted, they are, however, less likely providing a high-quality service.
to be certified, probably because they use the Coefficients in all models are not only statisti-
reputation of their conventional operations to cally significant but also economically significant.
distinguish them effectively from rogue sites. For instance, insurance coverage and running
Copyright # 2003 John Wiley & Sons, Ltd. Health Econ. 13: 329–344 (2004)
Quality Safeguards and Regulation of Online Pharmacies 341
conventional operations increase the quality index their insured parties in a specific pharmacy or
level in half and one-third, respectively, of this grant them a quality certificate.
index’s full range. The effect of pharmacies’ age is To check for the possible effects of endogeneity,
also substantial, with only 5 and 9 years being a system of structural equations was estimated
necessary for age to produce a similar impact to using three-stage regression. The equations con-
that of insurance and running conventional sider the three independent variables, the service
operations, respectively. quality index and the certification and information
indices used in the previous econometric analysis.
The equations are:
QualIndex ¼ b1 þ b2 InsurCov þ b3 RealAssets
Endogeneity of quality determinants þ b4 PharmAge þ e1
The previous section assumed that all the inde-
Certifcs ¼ b5 þ b6 InsurCov þ b7 RealAssets
pendent variables were exogenous. This allowed us
to analyse service quality, information and certi- þ b8 QualIndex þ e2
fication attributes at the level of specific variables.
The analysis, however, could be affected by the InsurCov ¼ b9 þ b10 QualIndex þ b11 Certifcs
endogenous nature of two of the independent þ b12 InfoIndex þ e3
variables – insurance coverage and certification.
When pharmacies decide on their strategy and The results, presented in Table 8, confirm the
quality, they are likely to take into considera- analysis in the previous section – in particular, the
tion both insurance coverage and insurers’ and positive effect of insurance coverage, running
certifiers’ requirements. However, insurers and conventional operations and pharmacies’ age on
certifiers also consider pharmacies’ quality when service quality. Also, the propensity of pharmacies
deciding if they are going to cover purchases by with insurance coverage and without conventional
Copyright # 2003 John Wiley & Sons, Ltd. Health Econ. 13: 329–344 (2004)
342 B. Arrunada
Copyright # 2003 John Wiley & Sons, Ltd. Health Econ. 13: 329–344 (2004)
Quality Safeguards and Regulation of Online Pharmacies 343
purchases, gathering data on the actual prices and prescription history with or without drug name
quality provided. The integration of both ap- display – you choose!; manage your household’s
proaches, however, would be a promising avenue account; personalized health topics and tools – based
for future research. By combining information on on your expressed interests; time-saving non-pre-
scription shopping lists’’ (http://www.merck-medco.
quality assurance and performance, such integra- com/medco/index.jsp).
tion would make it possible to test whether quality d. CVS, an US leader in both conventional and online
assurance explains not only the quality offered but pharmaceutical retailing, estimates that half its
also actual quality. It would also allow the net prescription sales contain some element of urgency
benefits for consumers to be gauged with greater and will never be sold online (‘‘Batalla de precios en
precision. Internet,’’ Diario M!edico, January 13, 2000).
e. Of up to 799 per cent as with ranitidin: 60 tablets
of Zantac cost 98.70 dollars in October 2000 in
Drugstore.com, a certified pharmacy, whereas it sold
Acknowledgements the same amount of generic ranitidin for 10.98 dollars
(accessed October 15, 2000).
The author thanks Merck Company Foundation, the f. ‘‘Prescription for Trouble’’ (Consumer Reports,
philanthropic branch of Merck & Co. Inc., Whitehouse February 2001, 66(2), 19–22).
Station, New Jersey, United States, for the uncondi- g. See [22–24] for a similar interpretation of age as a
tional support received during preparation of this study. proxy of reputation in the service sector
He also thanks Veneta Andonova, Joan Dur!an, Jos!e h. In http://dir.yahoo.com/Business and Economy/Sho-
Manuel Freire, Maia Gu. ell, Albert Oriol, Vicente Ortu! n, pping and Services/Health/Pharmacies/. Six further
Amadeo Petitbo! , Albert Satorra, St!ephane Saussier, pharmacies found using Altavista were added. No
Andreu Segura, Xos!e H. V!azquez-Vicente and Ernesto others were found using other search engines. All the
Villanueva for comments and Petar Balachev and pharmacies that sold more than one product were
Andr!es Jariod for research assistance. Usual disclaimers taken into account. For those offering a single
apply. product, a sample of 39 observations was taken,
giving a sample error of 1 per cent. Internet sites that
only offered information or sold over-the-counter
Notes products were excluded
i. Giving three values to this variable for distinguishing
a. See [2]. Three EU directives are involved in defining these two possibilities did not make any significant
the regulatory environment of European online difference.
pharmacies. In principle, Directive 2000/31/EC on j. See the following sites for information: http://
electronic commerce enables providers to serve www.nabp.net/vipps/intro.asp, http://www.hon.ch/
consumers in all EU member States, and Directive HONcode/Conduct.html and http://www.verisign.
97/7/EC on distance selling makes it possible to com (visited September 4, 2001). To obtain a VIPPS
deliver prescription medicines by mail. However, they certification, which is the most important, an online
may be superseded by national laws which, with the pharmacy must meet 17 criteria about patient
excuse of protecting public health and consumers, confidentiality, prescription security, quality assur-
may in fact be protecting local and conventional ance, and patient–pharmacist consultation [25].
providers. The stringent regime set by Directive 95/ k. Lack of data on alternative reputational investments,
46/EC on data protection might also unnecessarily such as advertising in launching the online operation,
constrain the efficient use of consumers’ databases, an is not likely to be a major problem in this case, given
important comparative advantage of online pharma- the relative values of the assets involved.
cies [3]. A more general analysis of the comparative l. For these, a value of 100 was taken. The signs and
prospects faced by online pharmacies in the US and statistical significance of the parameters and the
the EU is given in [4]. model did not change when other assumptions (35,
b. See [5, 6] on this case, and [3] for a description of the 1000 and 10 000) were introduced.
business strategies of three other pioneer European m. In this sample, it seems sensible to interpret the lack
online pharmacies: the British Pharmacy2U, Allcur- of observations with certain characteristics (zeroes in
es.com and the Swiss Pharmaworld.com. Table 5) as being ‘fixed zeroes’ (i.e. ‘structural’),
c. For example, the services offered by the CVS-Merck- which are impossible to observe, instead of as
Medco on-line pharmacy included the following in ‘random’ (‘sampling’) zeroes, which might have
September 2000: ‘‘Coverage and pricing comparisons simply remained unobserved in the sample [28, pp.
for brand-name and generic medications for both 135–140]. The reason lies in the additional adverse
mail service and retail; point-and-click refills and selection and moral hazard that would arise when
renewals; 12 months of mail service and retail insuring medicine expenditures without requiring
Copyright # 2003 John Wiley & Sons, Ltd. Health Econ. 13: 329–344 (2004)
344 B. Arrunada
that prescriptions be provided by independent third 11. Zbar JD. Pharmacies surge online. Advertising Age
parties. 2000; 71(14): S8–S9.
n. Time is needed also for designing the services with 12. Scanlon B. Vital signs. Inter@ctive Week 2001;
the highest value, as shown by some surveys of 8(11): 26–29.
online shoppers [29], probably because of the 13. Gambardella A, Orsenigo L, Pammolli F. Global
substantial change implicit in e-commerce. It is competitiveness in pharmaceuticals: a European
hardly surprising, then, that developing effective perspective. Report Prepared for the Directorate
safeguards also takes time. General Enterprise of the European Commission,
o. See details of these activities of the FDA in http:// November 2000 (http://pharmacos.eudra.org/F2/
www.fda.gov/oc/buyonline/ (visited September 4, pharmacos/docs/Doc2000/nov/comprep nov2000.
2001). pdf, visited June 8, 2001).
14. Smith MD, Bailey J, Brynjolfsson E. Understanding
digital markets: review and assessment. In Under-
standing the Digital Economy: Data, Tools, and
References Research, Brynjolfsson E, Kahin B (eds). MIT
Press: Cambridge, 2000; 276–288.
1. Goetz J, Lund D. What the law allows. Pharma 15. Armstrong K, Schwartz JS, Asch DA. Direct sale of
Exec 2000; 20(8): 76–84. Sildenafil (Viagra) to consumers over the internet.
2. Bordoni L (rapporteur). Issue Paper. Working New Engl J Med 1999; 341(18): 1389–1392.
Group Manufacturing ‘‘Consumer Goods Industry. 16. Eysenbach G. Online prescribing of Sildanefil
Conference on The Economy in Europe, Brussels, ðViagra1 Þ, on the world wide web. J Med Internet
1–2 March, 2001. Res 1999, 1(2); e10 (http://www.jmir.org/1999/2/
3. Twibell D. Europe takes to the net. Pharma Exec e10/, visited September 1, 2001).
2000; 20(8): 116–122. 17. Tirole J. The Theory of Industrial Organization. MIT
4. Twibell D. E-pharmacies: strategic implications for Press: Cambridge, 1988.
the healthcare industry. Datamonitor (Reuters Busi- 18. Klein B, Leffler K. The role of market forces in
ness Insights, London), June 2000. assuring contractual performance. J Polit Econ
5. Weber W. German controversy over internet 1981; 89: 615–641.
pharmacy. Lancet 2000; 356(9245): 1912. 19. Shapiro C. Premiums for high quality products as
6. Zwick S. M!edecines sans Fronti"eres: A Dutch returns to reputations. Q J Econ 1983; 98: 659–679.
dotcom that fills prescriptions for its neighbors 20. Nelson P. Information and consumer behavior.
challenges Germany’s strict regulations. Time Eur- J Polit Econ 1970; 78: 311–329.
ope, 2001; June 8 (http://www.time.com/time/eur- 21. Darby MR, Karni E. Free competition and the
ope/biz/column/0,9868,101418,00.html, visited June optimal amount of fraud. J Law Econ 1973; 16:
8, 2001). 67–88.
7. European Commission. E-Europe: an information 22. Lafontaine F. Agency theory and franchising: some
society for all. Communication on a Commission empirical results. Rand J Econ 1992; 23(2): 263–283.
Initiative for the Special European Council of 23. Thompson RS. The franchise life cycle and the
Lisbon, 23 and 24 March 2000, COM(1999) 687 Penrose effect. J Econ Behav Organ 1994; 24(2):
final, December 1999 (http://europa.eu.int/informa- 207–218.
tion society/eeurope/news library/pdf files/initiative 24. Arrun* ada B, Garicano L, V!azquez L. Contractual
en.pdf visited June 8, 2001). allocation of decision rights and incentives: the case
8. European Commission. E-Europe: an informa- of automobile distribution. J Law Econ Organ 2001;
tion society for all. Objective 3}Stimulate the 17(1): 256–283.
use of Internet: accelerating e-commerce, 2001 25. Fisher SE. Online health care contends with fraud.
(http://europa.eu.int/information society/eeurope/ Info world 2000; 22(27): 33.
index en.htm, visited June 8, 2001). 26. Woldt J. Internet transforming the pharmacy
9. Henkel J. Buying Drugs Online: It’s convenient business. Chain Drug Rev 1999; 21(17): 1 and 86.
and private, but beware of ‘Rogue Sites’. FDA 27. Welch BL. The generalization of student’s problem
Consumer, January–February 2000, revised on when several different population variances are
June 2000 (http://www.fda.gov/fdac/features/2000/ involved. Biometrika 1947; 34: 28–35.
100 online.html, visited September 4, 2001). 28. Hosmer DW, Lemeshow S. Applied Logistic Regres-
10. Heinrich J. Internet pharmacies}adding disclosure sion (2nd edn.) Wiley: New York, 2000.
requirements would aid state and federal oversight. 29. Yang Z, Peterson RT, Huang L. Taking the pulse of
Report to Congressional Requesters, FDCH Gov- internet pharmacies. Marketing Health Services
ernment Account Reports, 2000; October 19. 2001; 5–10.
Copyright # 2003 John Wiley & Sons, Ltd. Health Econ. 13: 329–344 (2004)