As this thesis comprises of three separate studies within the methodological framework of multi-method research, this paragraph first discusses the findings for each detailed research question from paragraph 1.2 that motivated the respective study, before discussing the overall findings.
The first study dealt the effects of one category of market information, product quality information, in the form of the following research question:
(DRQ1): How does reduced product quality information affect market process, market outcome and market performance in an auction?
This question was investigated in a case study at a large Dutch flower auction. The auction had introduced a new auction system in which participants lacked certain informational cues for product quality, such as flower color and flower freshness, which led to greater product quality uncertainty. Statistical analysis of transaction data before and after the introduction of this new auction system revealed that the increased product quality uncertainty led bidders to bid about 2% lower on average. Screen auctioning as such is not a fully electronic market as bidders still physically gather in the auction hall. However, it does represent an intermediate step towards such an electronic market as in both cases electronic product representation is involved. This study shows that one of the consequences of switching to an electronic market may be reduced availability of product quality information, which in turn has a negative impact on prices.
The second study dealt with the effects of another category of market information, market state information, as well as the consequences of reduced search and switching costs, by investigating the following research question:
(DRQ2): How do reduced search and switching costs and reduced market state information affect market process, market outcome and market performance in an auction?
This was studied in another case study at the same flower auction as the first research question. The auction allowed bidders to bid electronically from their home or office instead of being physically present in the auction hall (the KOA system). The electronic bidders bid in the exact same auctions at the same time as the bidders in the auction hall, so this allows for a detailed comparison between online bidders and offline bidders. While still not being a fully electronic market, KOA is a step beyond screen auctioning towards a fully electronic market, because now (in addition to electronic product representation) there are also electronic bidders present in the market.
Electronic bidding brings a reduction in search and switching costs for the electronic bidders, but at the same time they lacked information about the state of the market such as the number of bidders present in the auction hall. A statistical comparison of electronic bidders and bidders in the auction hall revealed that electronic bidders paid lower prices on average, as predicted by Bakos (1991). Subsequent investigation showed that this reduced price hypothesis only held for the electronic bidders who had their offices on the auction hall complex (internal bidders). Electronic bidders that had their offices elsewhere (external bidders) paid the same prices as the bidders in the auction hall. One explanation for this somewhat surprising finding is that there is a selection bias at work that cannot be controlled for: internal and external bidders may differ in their bidding strategies or their general purchasing strategies (orderbook buyers vs. cherry-pickers). Another explanation is that the internal bidders, through the security camera system that they had access to, had non-transaction information about the state of the market that the external bidders lacked. In particular, the external bidders lacked accurate information about the number of bidders present, which increased uncertainty may have led them to pay a premium in their bids compared to the internal bidders. More generally speaking, reduced market state information increases bidder uncertainty regarding the optimal bid level, given his preferences.
This type of uncertainty may cause bidders to make conservative assumptions regarding these unknown variables, with conservative in the sense that it minimizes their chance of not receiving the product they want to acquire. This results in higher prices. In addition to showing that reduced transaction costs in electronic markets do indeed lead to lower prices, this second study highlighted the importance of market state information. A change in ICT may change the market state information available to participants in an electronic market and a preliminary hypothesis is that a reduction of market state information leads to higher prices.
When moving from a market towards a fully electronic market, the physical components of the market processes gradually disappear as they are substituted by informational components. This implies that especially for electronic markets, the emphasis on the information exchange processes in the market becomes stronger and stronger, simply because that is all there is in an electronic market. This focus on information exchange processes in the first two studies led to the introduction of the concept information architecture of the market in the third study: the information architecture describes what type of information is, or when and how it becomes available, to whom during the market process.
Whereas the first two studies dealt with the role of ICT in automating existing market processes, in the third study a new type of ICT-enabled auction mechanism was developed, called the multidimensional auction. In a multidimensional auctions, bidders bid not only on price, but on multiple dimensions such as quality, delivery time and warranty. This is particularly relevant in a procurement setting in which the specifics of the traded object are not as fixed as in a traditional sales auction. Coupled with the concept of information architecture, this led to the third research question that investigated a third category of market information, market process information:
(DRQ3): What is the influence of a change in the information architecture of market process information on market process, outcome and performance in an electronic multidimensional auction?
This question was investigated in two laboratory experiments in which student subjects bid in a series of Web-based, multi-round multidimensional auctions under different information architectures (restricted and unrestricted) as well as
varying numbers of rounds (2 and 4 rounds). The results from the experiments showed that both market performance measures, winner efficiency and Pareto optimality, increased if a less restrictive information architecture was employed or if the auction last more rounds. Both of these findings are in line with the hypothesized effects. A somewhat surprising finding was that there was a strong interaction effect between the two experimental manipulations: the unrestricted information architecture significantly increases efficiency and optimality only in the 2-round case. Similarly, increasing the number of rounds yields only significant efficiency and optimality improvements under a restricted information architecture. It seems to be that there are strongly diminishing marginal returns to information. A tentative explanation for this could be that individual bidding behavior follows the belief-adjustment model (Hogarth and Einhorn 1992). The consequence is that information, despite being credible and correct, may not always be effective in changing bidding behavior. Bidders’ belief that the current highest bid is correct may simply be too strong to be swayed by any information. This implies that market performance will then remain at a sub-optimal level (Koppius and van Heck, 2001). In other words, there is a phenomenon of information saturation at work in the market: beyond a certain point, more information does not improve market performance any further.
A more general point of this particular study is that it emphasizes the process nature of markets. For instance, in much of the auction literature the length of the auction process is treated as irrelevant (but see Engelbrecht-Wiggans (1988) for an example of how it could be incorporated) and the information available to bidders is exogenous to the auction process itself (see Bergemann and Pesendorfer (2001) for a recent exception though). The experiments reported here show that, contrary to current theory, both these process-related factors have a significant influence on the auction outcome.
The synthesis of the three studies is provided in the overall research question of this dissertation, which was:
(ORQ) How does ICT influence the information architecture of the market and how does this affect electronic market process, electronic market outcome and electronic market performance?
As described above, the three empirical studies each highlight different aspects of the answer to this question. A more general and comprehensive answer to the overall research question is given in chapter 6, where the results in chapters 3-5 are theoretically generalized into a conceptual model of electronic markets. Focusing on the information exchange processes among traders in a market, a model is constructed that incorporates ICT, information architecture, market processes and market outcome. The constructs are linked through several propositions, eventually relating ICT and market outcome through the following five steps:
1. A change in ICT changes the information architecture of the market. (P6.5)
2. The information architecture determines which elements of the market information set are (or become) available to traders during the market process. (P6.4)
3. The market information set, as determined by the information architecture, affects the information exchange processes among traders (P6.3)
4. The information exchange processes among traders result in a set of transactions, i.e. market outcome, based on the market rules regarding allocation and transaction validity. (P6.2)
5. Market process and market outcome determine market performance, based on the specific performance criteria. (P6.1)
Although the model as a whole would need a comprehensive empirical test before we can accept it, each of the links among the constructs in the model has been empirically established in one or more of the studies. Furthermore, as the first check on the validity of the model, the three empirical studies were reframed in terms of the general model. That analysis suggests that the model has a degree of usefulness and parsimony that warrants further application in future studies, which lends initial credence to the model as a theory of electronic markets.