This discussion describes the statistical procedures used to prepare the following sections. It
covers the methods employed for calculating national and firm size dependent differences.
The survey questionnaire was designed by CRITO; the survey itself was conducted by IDC
during the period of February 18, 2002 to April 5, 2002. The survey was conducted in ten
countries (Brazil, Denmark, China, Germany, France, Japan, Mexico, Singapore, Taiwan, and
United States) with altogether 2,100 establishments. An establishment is defined as the physical
location of a firm. The sampling was a stratified random sample classified by size (large firms
with 250 or more employees, and small firms with between 25 and 249 employees) and by
industry (manufacturing, wholesale/retail distribution and banking/insurance). In Germany, 202
establishments were investigated, subdivided into 68 from the manufacturing industry, 66 from
the wholesale/retail industry and 68 from the banking and insurance industry. A total of 102
interviewed establishments belong to the class of small and medium sized enterprises, 100 to the
class of large establishments. The survey included only establishments which used the Internet to
buy, sell or support products or services.
In general, two different methods are used to analyze the data at hand. To analyze the relative
efficiency of e commerce users (both SMEs and large firms), a data envelopment analysis (DEA)
is used. DEA (Charnes et al., 1978) may be used to compare multi input with multi output data
to analyze the efficient combinations of implemented IT infrastructure and resulting impacts on
the output side.
The object of interest in a DEA model is the decision making unit (DMU) which is similar to a
firm in this case. For analyzing the relative efficiency of e commerce deploying SMEs, a Data
Envelopment Analysis (DEA) was used (Charnes, Cooper & Rhodes, 1978). Since most SMEs
cannot determine the benefits they derive from implementing innovative technologies in
monetary units, the survey asked for the set of e commerce technologies adopted on the one
hand, and the individually perceived efficiency or satisfaction on the other hand.
The model used in the DEA analysis is the BCC model (Banker, Charnes and Cooper) which
offers a differentiation between technical efficiency and scale efficiency (Golany & Roll 1989, p.
249) and evaluates solutions for non increasing, decreasing, and variable returns of scale. The
object of interest in a DEA model is the decision making unit (DMU) which is similar to a firm.
A DMU is a flexible unit responsible for the input and output variables. DEA only compares
each DMU with the `most efficient' DMUs in the sample (Bala, 2003). Efficient combinations of
input and output relations or efficient DMUs of a sample form the so called `efficient frontier
line.' In an n dimensional space the efficient frontier is equivalent to an imaginary umbrella over
the sample, covering the efficient DMUs and all theoretically possible combinations of efficient,
virtual DMUs. The DEA model calculates the relative position inside the data sample for each
DMU, based on its set of inputs and set of outputs (Parsons, 1992). Using a linear programming
procedure for the frontier analysis of inputs and outputs, DEA accordingly evaluates the best