4. LOS SEGUROS PRIVADOS Y EL VIH/SIDA
4.4. ASPECTOS DETERMINANTES DEL ÁMBITO CONTRACTUAL
4.4.2. La declaración de asegurabilidad
(the web) as a means of collecting large amounts o f data is. increasing dramatically not only
among survey professionals and large organisations but also in other fields o f management
study. However, management accounting research using the web survey method is very scant
despite this mode of data collection becoming established in other fields o f management
study, e.g. marketing. One recent example o f a management accounting research using a web
survey is Lowe and Locke’s (2005) study of UK academics perception o f the ranking o f peer
reviewed accounting journals. The scarcity o f studies using the web survey method may be
because most management accounting researchers have little expertise in the use o f the web
technology for research purposes. Although there are similarities between the mail and web
survey methods, the construction and successful deployment o f web surveys demand different
skills sets from those required by the traditional mail survey, i.e. computer programming
ability and webpage design skills. With the growing number o f internet users and ongoing
advances in computing devices and computer programming, the use o f the web to conduct
surveys and collect large amounts o f data is likely' to become prevalent in both .academia and
industry. This presents huge opportunities as well as. significant challenges.
19 The 8 journals included in Van der Stede et al. (2005) study are 4 0 5 , CAR, J A E , JAR J M A R , M A R , TAR, and
Chapter 4 Research D esign & M ethodology 106
Web surveys are very attractive compared to the traditional mail or telephone surveys because
it is relatively much cheaper to deploy (no interviewer, stationery or postage required). It also
provides instant access to a wider audience irrespective of their geographical location; and by
the same token achieves faster response rates. In addition, it offers the facility for the direct
recording and automatic encoding of survey responses in a format readable by most statistical
packages like Excel, SPSS and SAS. Moreover, this is achieved while the survey is still in
progress enabling survey responses and other information to be captured in real time and in
the desired format. Thus it reduces considerably, and in some cases completely eliminates, the
data encoding and entry time required by the traditional survey methods. Furthermore this
eliminates potential data errors introduced by such manual data encoding and data entry. For
instance, it avoids missing responses and ambiguous ticks. It is therefore not inconceivable
that in the near future web surveys may dominate mail surveys as the method for large data
collection in management accounting research.
4.3.3 Issues in C onstructing a W eb Survey
Web surveys are similar to mail surveys in many respects and thus are not immune to some of
the issues that are generally identified with the survey methodology; rather they raise new
dimensions to them. Salant & Dillman (1994) identified the main sources o f errors that affect
the ‘accuracy’ of survey results as coverage error, sampling error, measurement error, and
non-response error. Coverage error results when the known probability o f including all the
units o f a population in the sample is not greater than zero. This can be a great concern for
web surveys particularly when the target population may not all have access to the internet.
On the other hand, sampling error results when only a subset o f the survey population is
surveyed and this is completely unavoidable except if the entire population of interest is
surveyed. Related to this, non-response error results when a significant number of non-
respondents from the survey sample are different from the respondents in a manner that
Chapter 4 Research D esign & M ethodology 107
they would have provided a significantly different distribution o f answers to the issues of interest than that obtained from those who responded.
Finally, measurement error results from inaccurate responses to questions. This may be due to
poorly worded questions and/or the survey method. However, Salant and Dillman (1994) note
that measurement error is more critical for questions dealing with sensitive or abstract ideas.
Sensitive questions were more likely to result in measurement error in face to face or
telephone interviews compared to other survey methods because of the personal contact
between respondents and the researcher. On the other hand, abstract idea questions were more
critical with mail surveys and by extension web surveys because the researcher is not
physically present to clarify respondents’ misunderstandings. This may be more acute in mail
surveys than in web surveys as web surveys are not as limited in the format and presentation
o f additional information as the mail surveys. For example, further information explaining a
concept can be provided by way of pop-up widows that are activated by the respondent only
when more clarification is required, and so it does not obstruct the main survey if not needed.
Thus more interactivity can be built into web surveys than is possible with mail surveys.
In this study, coverage and measurement error are not much o f an issue by the fact that the
target sample was pre-identified and the questionnaire sufficiently developed in prior
discussions with the target organisation using the Microsoft Word version o f the survey.20
Moreover, staff in the case organisation all had internet access, and worked with computers
and emails regularly in their day to day work. Several pilot tests of the word version o f the
survey had also been carried out both within and without the target organisation. However, in
designing the survey web pages several other factors were of particular concern, some relating
to non-response error. The factors included ease o f access and navigation o f the survey,
identification o f individual responses and issues o f confidentiality, and the integrity o f the
20 A whole region o f ABC LLP was identified as the sample, but thereafter it is a population as the whole region was included in the survey.
Chapter 4 Research D esign & M ethodology 108
data collected. Therefore, the main aim in the design of the web survey was to create web
pages that were accessible only to the target sample and that would enhance the likelihood of
the target sample responding, and doing so in the manner intended. Another objective was to
integrate the web pages with a database such that responses to each web page were
automatically captured on the database, and to ensure that the data were not ‘corrupted’ by
uninvited responses from people straying onto the survey on the web. Therefore, secure but
accessible and easy to navigate web pages needed to be constructed such that the chances of
responding to the survey did not depend on a number factors such as: the computer skills of
the respondents; the transmission limitations, the computer equipment or internet browser
used by respondents. Thus, the web pages were generally constructed within the perceived
limits o f what respondents’ computers, browsers, and transmission lines can deal with. During
the web survey construction, tests were run using different computers and screen sizes,
different operating systems, different browsers and from three different locations using
different connection speeds.
4.3.4 Description of the W eb Survey
The web survey pages started with a welcome screen that introduced the survey and thanked
respondents for agreeing to participate in the survey. Screen shots o f the web pages are
presented in Appendix 3. The ease o f completing the survey and the anticipated completion
time were also emphasised, while providing instructions on how to progress to the next page
and the rest o f the survey. The page after the welcome screen required participants to login
into the survey using a pre-assigned username and password. The required username and
password was earlier communicated to the target sample in the cover email invitation sent out
on behalf o f their regional Chairman. Prior to the cover email, an earlier mail had been sent
out on behalf o f the regional Chairman intimating staff o f the survey. In the cover email, the
web link to the survey was included such that a mouse click on the link opened up the survey
welcome screen page on their internet browser. Since anonymity was agreed beforehand with