Global Comparison Aggregation Services
Hongwei Zhu, Stuart E. Madnick, Michael D. Siegel MIT Sloan School of Management 30 Wadsworth Street Cambridge, MA 02142 {mrzhu, smadnick, msiegel}@mit.edu Abstract Web aggregation has been available regionally for several years, but this service has not been offered globally. As an example, using multiple regional comparison aggregators, we analyze the global prices for a Sony camcorder, which differ by more than three times. We further explain that lack of global comparison aggregation services partially contribute to such huge price dispersion. We also discuss difficulties encountered in the manual integration of global web sources. Motivated by this example, we propose a context mediation architecture for global aggregation to address semantic disparities of global information sources. Global aggregation services can bring efficiency to the global market and can be useful for market research and other business uses. Keywords Web Aggregation, Context, Semantic Integration
Fig. 1. Prices for DCR-IP5 in Sweden Among the five vendors, 18,082 Swedish Krona (SEK) is the lowest total price. Is this the best deal, or is there a substantially better deal, on a global basis? Without a global aggregator, this can only be found out manually even with help of other comparison aggregators available in other countries. Our manual exercise found one vendor in the U.S. who ships the product to worldwide destinations at a total price of $1,099.99 ($999.99 plus $100 international shipping charge, which includes warranty valid anywhere in the world), as shown in Figure 2.
1. Introduction With its increasing connectivity and capability, the World Wide Web is becoming the platform for global ebusiness. The global connectivity of the Web has not been exploited fully by existing service oriented ebusiness applications. For example, most of todays shopbots still only offer regional comparison shopping services, where regional (as opposed to global) information sources are used [1]. Comparison shopbots are also known as comparison aggregators for their capability of transparently aggregating information from multiple web sources [2]. What if comparison aggregation service is offered on a global basis? Imagine for the moment you are from Sweden and interested in buying a pocket sized digital camcorder. After some research on the Web you decide to buy a SONY DCR-IP5, which records video in MPEG format for easy editing on computers and weighs only 12 ounces (i.e., 336 grams). So you launch your favorite comparison aggregator to find the best deals and it returns information as shown in Figure 1.
Fig. 2. An Offer for DCR-IP5 from the U.S. Between 18,082 SEK and $1,099.99, where would you buy? A seemingly simple question once you figure out that 1 US dollar is about 10 SEK. The Swedish offer is
64% more expensive than the U.S. offer. However, is this a special case? Again, a global aggregator will be helpful to provide an answer. In the next section, we present a case study on worldwide price dispersion for the Sony camcorder. We explain various reasons why such dispersion exists and argue how global comparison aggregation can help connect global information sources, thereby bring efficiency to the global market. In section 3, we discuss deficiencies of existing comparison aggregation and technological challenges to providing global aggregation services. In section 4, we propose a scalable architecture that is promising for those challenges. We conclude with our insights of global comparison services in section 5. 2. Case Study Global Price Dispersion for Sony DCR-IP5 One of the expectations of the European Union (E.U.) is to have an efficient integrated market with small price differences among member countries. A recent survey in the E.U. [3] shows that in the fresh food market high price countries are often two times more expensive than countries with minimum prices; even in the consumer electronics market, one country could be over 50% more expensive than another for a particular product. Data for that study was collected by three consultants who sampled various products in different stores. Since comparison aggregation is a great tool for collecting price information, it has been used in a number of price dispersion studies in the U.S. for products such as books, CDs, and consumer electronics. Inter-store price differences were found to be 25-40% [4-6]. Although price dispersion still exists among online stores, overall online prices are lower than physical stores; for books and CDs, online prices were found to be 9-16% lower [4]. We could only find one study on price dispersion in the global online market [7], which showed that a U.S. buyer could save 42% for a particular textbook by purchasing it from the U.K. instead of from the U.S. As there have been few studies on global price dispersion of the online market, we conducted an empirical study on the SONY digital camcorder in section 1: MICROMV DCR-IP5, which was introduced into the consumer electronics market in early 2002. Market prices for such a new product are extremely volatile; we took a snapshot of global prices by collecting data within 24 hours between March 8 and 9, 2002. We used a number of regional comparison aggregators to retrieve the prices for the product. These aggregators

include BizRate, mySimon, Dealtime, Shopper, PriceRunner, PriceGrabber, Kelkoo, and Kakaku. We report our analysis on the unique vendor/price basis within a country. That is, if multiple aggregators in a country report on the same vendor, we treat them as one observation if the prices are the same or within $1 difference. If a vendor has its online and physical stores as two entities, we treat them as two different observations even though both may charge the same price. All prices are listing prices not including shipping charges. 2.1 Worldwide Price Dispersion We collected 172 observations covering US, Japan, and nine European countries. Figure 3 shows the histogram of prices. It is obvious that prices are highly dispersed. Most prices are within the range of $1000-2000 and they are nearly evenly distributed in this range. Prices outside this range exist at both ends.

Frequency

2400 More
Min 783 Max 2254 Median 1569 Mean 1524 Standard 335 Deviation

Price ($)

Fig. 3. Price Histogram (N=128) Figure 4 shows the price distribution for all 13 countries, with the number of observations at the bottom. This is a box plot with each box representing 50% of price observations (i.e., the 25% and 75% quartiles) and the line within the box being the median. Lines stemming out of boxes cover all the other prices except for the extremes marked as solid circles.

Price of DCR-IP5 ($)

Brazil France Spain Netherlands Norway Japan Mexico UK Germany Italy Denmark Sweden
Fig. 4. Price distribution in different countries
Clearly, prices are different between countries. US and Japan have the lowest price levels. Most of continental European countries, except for Italy, have medium high prices. Italy and northern European countries have the highest price levels in our observation. Comparing with the international book price study [7], which shows that the U.K. has lower book prices, here we find that the U.K. has higher prices for this camcorder than the U.S. Lets look at US prices in more detail, shown in Figure 5. These 53 unique price observations do not include SonyStyle US, Sonys online store in the U.S., and major consumer electronics vendors like BestBuy and CircuitCity, which offer the product at the same official price: $1299.99. We can see from the figure most prices are at or below this price level. The average price is $1203, which is 7.7% below the official price. More importantly, U.S. average price is 26.3% lower than the worldwide average.

30 Frequency 25 Frequency 1400 More Price ($) 1 0% Cumulative % 100% Cumulative % 80% 60% 40% 20%
aggregators ever make to the top 50 most visited sites in the U.S. measured by Jupiter Media Metrix. In domestic e-business, it is possible that the three assumptions are met to some degree. In the context of global e-business, even the basic assumptions could be violated. Although in terms of features the camcorder is nearly homogenous worldwide, other factors exist to result in heterogeneity. The product may be assembled in different plants that have different cost structures (e.g., plant in Malaysia vs. in Japan). Manufacturers often use different labeling to segment the market, e.g., different languages for product manuals in different regions. Warranty and other post sales services are often divided into regions. Further, search costs are much higher due to lack of services that provide worldwide price information. We gave a hypocritical situation in the motivational example, but in reality chances are the Swedish buyer does not know any price information in the U.S. These factors and the lack of a global comparison tool contribute to the worldwide price dispersion phenomenon. The following summarizes various explanations: Manufacturers have heterogeneous production costs around the world. Vendors have different pricing strategies, e.g., some may offer specials in certain parts of the world to promote sales. Buyers involve different search costs and have different preferences, e.g., buyers are not aware of price differences and weigh other factors more than price. Fluctuation of exchange rate causes price differences among countries. Manufacturer price control via market segmentation and other means of price discrimination, e.g., introducing product at different times. Although price dispersion will not completely disappear, price transparency resulting from comparison aggregation should help mitigate dispersion and lower overall prices. This effect has been observed in the online market, e.g. average online prices are 7.7% lower than official price of the Sony camcorder and for books and CDs online prices are 9-16% lower than prices in physical stores [4]. Further, the U.S. average price for the camcorder is 26.3% lower than the worldwide average and the adoption rate of comparison aggregator in the U.S. is among the highest. Arguably, regional aggregation has helped increase competition and lower
Fig. 5. Price Distribution in the U.S. (N=52) Explanations for Price Dispersion 2.2 Textbook economic theory predicts that under perfect competition (e.g., Bertrand competition) commodity prices converge to one price, the so-called Law of One Price. But real world markets have shown no evidence to support this. The price dispersion phenomenon has been explained as a violation of one of the Bertrand assumptions: product homogeneity, zero search costs, or perfectly informed consumers [4]. In our case study, we looked at prices for one single product. Although it does have two models for video output (i.e., PAL and NTSC), this distinction is marginally important because its MPEG recording format allows for easy processing on a PC, which does not use the video output. In addition, many TV sets support dual video standards. So this product can be regarded as homogeneous. Regional aggregators can help lower search costs, which should lead to convergence of prices [8]. Whether all consumers are perfectly informed about price distribution is in question. Although comparison aggregation has gained some popularity, none of the popular comparison

the overall price level in the U.S. Global aggregation can potentially bring this efficiency to the global market, generating greater consumer benefits. Next, we will examine the deficiencies of existing web aggregation services and identify technological challenges to advance from regional aggregation to global aggregation. 3. Technology Challenges to Global Aggregation 3.1 Deficiencies of Current Aggregation Services Most of existing regional comparison aggregation is primarily implemented using web wrappers to extract information from web sources. This technology enables transparent aggregation even among non-cooperative sources, but conflicting implementation goals often deteriorate the quality of aggregated data. In addition, extraction tools do not address data semantics issues that are critical to service quality. System responsiveness is often implemented at the price of compromising information timeliness, i.e., to achieve fast response to user request, many aggregators cache extracted data in their systems, resulting in out dated information. A daily update of the cache is not sufficient to avoid compromises on data timeliness because online prices change frequently due to low menu cost [9] and dynamic pricing strategies. Within a 2-hour window we observed a more than 2% decrease in average prices of the camcorder reported by one U.S. comparison aggregator [1]. Erroneous information can significantly impair the quality of comparison aggregation services. Figure 6 shows an example where a vendor has updated its price from $62 to $77.15 while the aggregator still reports the old price. If an automated purchasing agent decides to buy from the first vendor and makes the deal without consistency check, it will end up with overpaying 24%. This tension will become more outstanding in global aggregation because of increasing number of sources.
Text based search, as opposed to semantics based search, can cause problems as well. In one occasion, an aggregator mistook a $2 accessory of the camcorder and reported it as the price for the camcorder (although occasionally vendors make mistakes like $1 laptops; but this is not the case here). Data semantics issues become more severe in global aggregation given the diverse contexts of sources, which we will explore next. 3.2 Data Semantics of Global Web Sources The diversity in the origination and destination of information causes enormous complexity in making aggregated information meaningful and understandable. We have seen in the motivational example that currency needs to be converted using retail conversion rate to make sensible comparison for the Swedish user. Other issues exist. Lets illustrate these issues with an example of information about laptop computers from multiple sources of Sony, summarized in Table 1. We will ignore language difference in the following discussion. Table 1. Information about a SONY Laptop Computer from Multiple Sources U.S. U.K. U.K. (in German) 2.76 lbs 1.26 kg 1,26 kg Weight NA NA Thickness 1.09 $2,029 plus 1,699.00 GBP 1.699,00 GBP Price $25 shipping incl. VAT inkl. MwSt. First to note is that not all information is available at a single source. In this case the thickness information is not immediately available from U.K. sources (it is buried in a PDF document). If an aggregator takes the information from the U.S. source and directly reports to its German users, 1.09 probably would not be helpful to users who are familiar with metric systems for measurement. In addition to different units being used (lbs vs. kilograms, inches vs. millimeters, US dollars vs. British Pounds, etc.) there are other representational differences, such as symbols for thousands separator and decimal point. These differences have to be detected and reconciled for the users. There is a more complicated problem in the data shown in Table 1. The last row shows pricing information for the product. Aside from representational differences, we note that the components going into price are quite different. Price, however simple as it appears, is in fact a complicated concept that has different meanings from different perspectives.

Fig. 6. Error due to Lack of Timeliness: Price Reported by Aggregator (top) vs. the Source (bottom)
How much an item costs for someone to acquire is often different from how much it is listed for because of other costs that are associated with this transaction, including
taxes, duties if it involves international trade, shipping and handling, etc. An accurate calculation for price in the sense of cost to acquire could be very complicated in the context of global e-business. Calculation of VAT alone requires lots of additional information because VAT varies depending on the type of product, origination, destination, and special treaties between regions. The variations range from 0 to 25% of the listing price in European countries. The information listed in Table 1 is a hybrid of the two concepts for price with some missing components. This makes aggregation and meaningful comparison difficult. McCarthy and Buvac [10] illustrated this problem with an example of different prices of the same GE aircraft engine perceived by different organizations, such as the U.S. Air Force and U.S. Navy depending on whether the price includes spare parts, warranty, etc. Another problem not explicitly shown in Table 1 is how the aggregators identify the same product from different regions. In the process of manually composing the Table, we noticed that the model numbers are different between laptops in the U.S. and those in Europe. We recognize their similarity (in this case identity except for the model numbers) by examining the configurations (e.g., CPU speed, hard disk capacity, weights, etc.). The fact that manufacturers often market the same product with different names in different regions makes it difficult for the aggregator to recognize their identity. This problem is best described from the following Camera example from Focuscamera.com:. a USA Minolta Maxxum is a Minolta Dynax overseas, the USA Canon EOS Rebel 2000 is an EOS 300 overseas, Pentax IQ Zooms are Pentax Espios overseas, etc. Conversely, when models with different features are named the same or slightly different in different regions, aggregators sometimes cannot recognize the distinction. In the Sony DCR-IP5 case study we found that some vendors label the product as DCR-IP5E to indicate that it is an international model compatible with the PAL standards rather than the NTSC standards in the U.S. What makes it worse is that most vendors use DCR-IP5 for both the NTSC model and the PAL model. Although this does not cause big problems because of its MPEG recording format, for other types of products this could be an issue. The preceding discussions can be summarized into three categories of issues: Representation how do we represent things Composition what are the components for the thing

Recognition referring to

what is the thing we are really
Next, we will propose an architecture that aims to address these issues so that users will get accurate, consistent, and meaningful aggregated information. 4. Architecture and Prototype for Global Aggregation Services 4.1 Context Mediation Architecture The adoption of XML data standards and the emergence of Web services show promising signs for mitigating the tension between timeliness and responsiveness of global aggregation. But given the large scale and diversity of global aggregation, we recognize that heterogeneity of sources will continue to exist. We propose a context mediation architecture (see Figure 7), which is based on the theories and techniques of context [10], mediated architecture [11], and the Context Interchange (COIN) project [12,13].
Context Query Result COIN Mediator Context Query Result Receivers Context Mediator Function Ontology Library Web Wrapper

Sweden

Query compiler

Mediated query

Context Elevation axioms ODBC SOAP

Optimizer

Executioner

Data Access Data Sources

Fig. 7. A Context Mediation Architecture for Global Aggregation Each online vendor or a regional aggregator is a data source, which can be accessed through the data access layer that implements various mechanisms to accommodate source heterogeneity. Both sources and receivers (i.e., users) have their contexts, which should be captured in logically distributed context knowledge bases. A common ontology or a set of aligned ontologies can be created by the aggregator. The mapping between data elements and ontologies is provided by elevation axioms. Contexts, ontology, and elevation axioms together address those three types of semantics issues. Conversion functions are used to translate values between contexts. The core part of a global aggregator is the COIN mediator, which resolves context conflicts between data sources and receivers. With this architecture, a global aggregation user can specify what currency to use for price (representation) and whether the price includes or excludes taxes and

shipping handling (composition) about the specific product (recognition) offered by global vendors. Scalability is achieved by the abstraction of context and the modular design. 4.2 Prototype of Global Aggregation A prototype has been developed to demonstrate the feasibility of the proposed architecture. We use a handful of regional aggregators as the sources. Using Cameleon web wrapper [14], we can impose a relational data model on these web sources and query them using SQL. For illustration purposes, our sources contain seller and price information for a single product the SONY DCR-IP5 camcorder. In the prototype, we focus on the issues of domestic and international taxation, shipping charges, and currency conversions that need to be addressed in global comparison services. Different situations of sources and receivers regarding these issues are represented as different contexts, samples of which are given in Table 2. They are axiomatized and recorded in the context knowledge base of the system. In addition, conversion functions are added to provide automated conversion service between contexts. Table 2. Contexts for Price in Global Compassion Shipping+ Currency Tax Included, 19.5% Domestic: 15 France Euro Intl: 80 Included, 25% Domestic: 20 Sweden Krona Intl: 800 Pound Included, 17.5% Domestic: 10 UK Intl: 35 USD Not included Domestic: 50 US Intl: 100 Exclude Exclude US, Base USD If domestic Include US, Cost USD vendor, no tax; domestic or otherwise, add intl shipping 3% import tax accordingly Include 25% tax intl shipping Sweden, Krona regardless accordingly Cost + : Assume vendors only distinguish between domestic and interchange shipping charges. This is being refined to use online shipping inquiry services to calculate shipping costs by supplying products dimensions and weight. Source context Receiver context A domain ontology, as shown in Figure 8, captures the common concepts (in rounded boxes) and their relationships pertaining to contexts illustrated earlier. For example, a seller is a specialization of an organization, which has a location of type country. A modifier is a special attribute whose value
is specified in the context knowledge base. For example, price has a type modifier to indicate if it is base price, price with tax included, or total cost in a particular context.

domesticTax basic importTax organization location country fromCur basic currencyType toCur
type seller soldBy product prodPrice price buyer location
exchangeRate exchangeDate date
internationalShipping domesticShipping Inheritance Attribute Modifier monetaryValue priceDate currency

attr mod

Fig. 8. Domain Ontology for Global Aggregator A mapping between data from each source and the concepts in the ontology is provided by a set of elevation axioms to relate semantics to the data. All axioms and functions are supplied to a recent implantation of COIN mediation system [15], which can take user queries in SQL, automatically detect and reconcile context conflicts, and execute mediated queries to return results in the context of the user. We will give an example below to show how the system can help users such as the Swedish buyer mentioned in the beginning to do global comparison shopping. The Swedish buyer is interested to know the total cost of the camcorder from worldwide vendors. His context has been recorded with the system. Now he can issue a query to compare prices of vendors all over the world using a predefined SQL, compare_all:
Select union Select union Select union Select union. seller, price from kelkoofrance //French source seller, price from pricerunnersweden //Swedish source seller, price from pricerunneruk //UK source seller, price from cnetshopper //US source

//etc.

As we illustrated in the sample contexts, differences exist between the sources and the receiver. The COIN mediator automatically detects these differences and reconciles them by calling conversion functions. This process generates mediated queries that perform all the necessary conversions from source context to receiver
context. Some of the conversions we expect the system to automatically generate are given in Table 3. Table 3. Appropriate Conversions for Reconciliation of Context Differences Source Conversion France Deduct 19.5% French tax, add 25% Swedish tax, add Euros to Krona Sweden Add 20 Krona domestic shipping US Add 25% Swedish tax, add $100 international shipping, convert USD to Krona UK Deduct 17.5% UK tax, add 25% Swedish tax, add 35 The input SQL query is translated into a DATALOG query for the abductive reasoning engine to generate mediated queries in DATALOG, which in turn are translated into optimized SQL queries to be executed in parallel by the executioner [15]. The following gives the final mediated query automatically generated by the system to answer the users initial query; we hope that readers can examine this and be convinced that all anticipated conversions are indeed performed by the following query. Note that olsen is an auxiliary online source that provides current and historical currency exchange rates; the system uses current date (i.e., date when the query is issued).

//French source. Deduct 19.6% French tax; add 25% Swedish tax; //add select kelkoofrance.seller, ((((kelkoofrance.price/1.196)+((kelkoofrance.pr ice/1.196)*0.25))+80)*olsen.rate) from (select seller, price from kelkoofrance) kelkoofrance, //find exchange rate using auxiliary source (select 'EUR','SEK',rate,'11/01/02' from olsen where exchanged='EUR' and expressed='SEK' and date='11/01/02') olsen union //Swedish source. Add 20 Krona domestic shipping select pricerunnersweden.seller, (pricerunnersweden.price+20) from (select seller, price from pricerunnersweden) pricerunnersweden union //UK source. Deduct 17.5% UK tax; add 25% Swedish tax; //add 35 intl shipping; convert Pounds to Krona select pricerunneruk.seller, ((((pricerunneruk.price/1.175)+((pricerunneruk. price/1.175)*0.25))+35)*olsen.rate) from (select seller, price from pricerunneruk) pricerunneruk, //find exchange rate using auxiliary source (select 'GBP','SEK',rate,'11/01/02' from olsen where exchanged='GBP' and expressed='SEK' and date='11/01/02') olsen union
//US source. Add 25% Swedish tax; add $100 intl shipping; // convert USD to Krona select cnetshopper.seller, (((cnetshopper.price+(cnetshopper.price*0.25))+ 100)*olsen.rate) from (select seller, price from cnetshopper) cnetshopper, //find exchange rate using auxiliary source (select 'USD','SEK' rate '11/01/02' from olsen where exchanged='USD' and expressed='SEK' and date='11/01/02') olsen union.
( & % $ # ! '# "
An excerpt of the results is shown in Table 4 (reformatted from prototype output). All prices have been translated into the context of the Swedish user, who can easily compare them on the same basis. Finding the best deal globally is now as simple as clicking the predefined query with the help of this prototype of global comparison aggregation services. Table 4. Excerpt of Results in Users Context Source Seller Price (i.e. total cost in Krona) Sweden Foto & Elektronik AB 15815 Expert Citybutiken/Konserthuset 16015 Click ontime 23470 US Bridgeviewphoto.com 10255 PC-Video Online 10594 Circuit City 14933 4.3 Extensions to Prototype and Related Issues With this context-mediated architecture, a global aggregator can compare worldwide prices in a meaningful way for various users. This prototype successfully resolves representation and composition semantic conflicts. Recognition can be addressed by using a mapping of product codes to identify the exact product that may be labeled differently in various parts of the world. Alternatively, the system can use a formal ontology for products, which may become available on the Semantic Web in the future. Our future research on mediation using multiple ontologies and the development of the Semantic Web will help find alternative solutions. The prototype can be readily extended to serve a broad audience by adding axioms for new sources and receivers. Clearly, technologies used here can enable full-scale implementation of global aggregation services, which will significantly increase the efficiencies of global e-business. Opportunities for

T 6 FP S 7 @ FP R 7PI 6 FE C A@ 3 442F 424Q42GGH44GD446 B4B 4120
aggregation services are abundant. Readers interested in how aggregated information can be used to enhance values are referred to [16] for a thorough account. Data reuse plays an important role in the success of global aggregation services. The proposed COIN architecture provides solutions to technical challenges to reusing data from multiple web sources. Other obstacles still exist. Policy issues regarding data reuse are discussed in [17]. 5. Conclusions Despite the global presence of comparison aggregation, most of the services are offered regionally, not globally. Lack of global information can result in inefficiency in the global market. Our price dispersion case study shows that the worldwide prices for DCR-IP5, a Sony digital camcorder, can differ by nearly three times. A global aggregator can close the information gap and bring efficiency to the global market. With this motivation, we propose a context mediation architecture to address data semantics issues for global aggregation. A prototype global aggregator has been developed to validate the architecture. The technologies used here show promising signs for building scalable platforms of global comparison aggregation services. These new services will benefit a variety of users. They will certainly help consumers find the best deals around the world; they can also assist researchers and policy makers to systematically collect market data with low cost (recall that the E.U. price dispersion survey mentioned in section 2 relied on three consultants who visited stores to manually collect retail prices); manufacturers can also use the services to find out the actual retail prices of their products around world, with which they can better assess demand and set appropriate wholesale and suggested retail prices. The emergence and the wide usage of global aggregation services will make the web the truly efficient platform for e-business. Acknowledgement The study has been supported, in part, by BSCH, Fleet Bank, Merrill Lynch, MITRE Corporation, the Singapore-MIT Alliance, and Suruga Bank References [1] Zhu, H. (2002) A Technology and Policy Analysis for Global E-Business, MIT Masters Thesis. [2] Madnick, S.E., Siegel, M.D. (2002) Seizing the Opportunity: Exploiting Web Aggregation, MISQ Executive, 1(1), March 2002, 1-12. [3] EU Economic Reform (2001) Price Dispersion in the Internal Market [4] Brynjolfsson, E., Smith, M.D. (2000) Frictionless Commerce? A comparison of the Internet and

Conventional Retailers, Management Science, 46(4), 563-585. [5] Clay, K., Krishnan, R., Wolff, E. (2001) Prices and Price Dispersion on the Web: Evidence from the Online Book Industry, Journal of Industrial Economics, 49(4), 521-539. [6] Baye, M.R., Morgan, J., Scholten, P. (2001) Price Dispersion in the Small and in the Large: Evidence from an Internet Price Comparison Site, Working Paper, Indiana University. [7] Clay, K., Tay, C.H. (2001) Cross-Country Price Differentials in the Online Textbook Market, Working Paper, Carnegie Mellon University. [8] Bakos, J.Y. (1997) Reducing Buyer Search Costs: Implications for Electronic Market Places, Management Science, 43(12), 1613-1630. [9] Bailey, J.P. (1998) Intermediation and Electronic Markets: Aggregation and Pricing in Internet Commerce, Ph.D. dissertation, MIT. [10] McCarthy, J. and S. Buvac (1994) Formalizing Context (Expanded Notes), Stanford University. [11] Wiederhold, G. (1992). Mediators in the Architecture of Future Information Systems. Computer, 25(3), 38-49. [12] Goh, C.H., Bressan, S., Madnick, S., Siegel, S. (1999) Context Interchange: New Features and Formalisms for the Intelligent Integration of Information, ACM Transactions on Information Systems, 17(3), 270-293 [13] Madnick, S.E. (1999) Metadata Jones and the Tower of Babel: The Challenge of Large-Scale Semantic Heterogeneity, Proceedings of the 1999 IEEE Meta-Data Conference, 1-13, April 6-7, 1999. [14] Firat, A., Madnick, S., Siegel, M. (2000) The Cameleon Web Wrapper Engine, Proceedings of the Workshop on Technologies for E-Services, September 14-15, 2000, Cairo, Egypt. [15] Alatovic, T. (2002) Capabilities Aware Planner/Optimizer/Executioner for Context Interchange Project, MIT Masters Thesis. [16] Madnick, S.E., Siegel, M.D. (2002) Seizing the Opportunity: Exploiting Web Aggregation, MISQ Executive, 1(1), March 2002, 1-12. [17] Zhu, H., Madnick, S.E., Siegel, M.D. (2002) The Interplay of Web Aggregation and Regulations, Proceedings of 3rd International Conference on Law and Technology, November 6-7, Cambridge, USA.

A TECHNOLOGY AND POLICY ANALYSIS FOR GLOBAL E-BUSINESS
by Hongwei Zhu B.S. Thermal Eng., Env. Eng., Tsinghua University, 1991 M.S. Env. Eng., Tsinghua University, 1994 M.S. Env. Eng., University of Cincinnati, 1998
Submitted to the Engineering Systems Division in Partial Fulfillment of the Requirements for the Degree of Master of Science in Technology and Policy at the Massachusetts Institute of Technology June 2002 @2002 Massachusetts Institute of Technology All rights reserved.
Signature of Author Technology and Policy Program, Engineering Systems Division May 17, 2002
Certified by Stuart E. Madnick John Norris Maguire Professor of Information Technology Professor of Engineering Systems Thesis Supervisor
Accepted by Daniel Professor of Aeronautics and Astronautics and Engineering Director, Technology and Policy Chairman, Committee for Graduate Hastings Systems Program Students
by Hongwei Zhu Submitted to the Engineering Systems Division on May 17, 2002 in Partial Fulfillment of the Requirements for the Degree of Master of Science in Technology and Policy

Abstract

We introduce an e-business analytical framework that focuses on transaction flows, including information, physical goods, and services. Within this framework, global e-business involves transaction flows that cross both organizational and national boundaries. Many challenging technology and policy issues arise from this trans-boundary characteristic of global e-business. These issues are analyzed using web aggregation as an example global e-business application. We start the analysis by introducing web aggregation services and their enabling technologies. Our survey of current status of web aggregation indicates that most services are still operated regionally despite their global presence. Although benefits of web aggregation have been realized in regions with extensive use of information aggregation, little is done at the global level. Our case study on worldwide price distribution of a nearly homogeneous consumer electronics product indicates great potential for global aggregation to bring information and efficiency to the global market. In addition to lack of global integration, we identified other deficiencies of web aggregation. Technological challenges and possible solutions to overcoming these deficiencies are discussed. However, having technological capability for trans-boundary information flow does not solve all problems in global aggregation. National policies often prohibit such flow into nations that have different policies, especially in database and privacy protection areas. We analyze these policy issues and propose future research on international policy harmonization.
Thesis Supervisor: Stuart E. Madnick Title: John Norris Maguire Professor of Information Technology & Professor of Engineering Systems

ACKNOWLEDGEMENT

I would to thank my supervisor Professor Stuart Madnick, whose ideas have inspired this research. In addition to his intellectual advices, Professor Madnick has been my mentor and supporter from whom I get courage and wisdom that will go a long way with my career and beyond. I also would like to thank Dr. Michael Siegel for his insightful advices and challenging questions, which have been my continuous sources for improvement. My thanks also go to Professor Nazli Choucri and Ms. Sharon Gillett for their advice, support and helpful discussions. I would like to extend my appreciation to Sydney Miller and Jean Marie De Jordy from TPP for various help they offered over the past two years. Thank you Yubettys Baez and Claudia ForeroSloan for making the office areas a nice work place. Special thanks go to my wife, who sacrificed in supporting my study. And my parents, who I owe too much to express in words; you may never be able to read my thesis, but you understand better than anybody the impact of barriers to trans-border people flow because you are even disallowed to enter a country that claims to respect freedom for my commencement at MIT. Finally, I would like to acknowledge financial support for the research from Banco Santander Central Hispano, MIT Center for eBusiness, Fleet Bank, Instituto de Empresa, MITRE Corporation, and Suruga Bank.

Intra-organization Transaction flow: * Information

Physical goods

Intra-nation (domestic)
Figure 3. Categories of E-Business

Inter-nation

The lower left quadrant represents e-business where the transaction flows stay within the organizational and national boundary. Issues involved here are domestic and internal. As we move up to the upper left quadrant, transaction flows cross organizational boundaries. Issues of dealing with external organizations arise. Similarly, moving right to the lower right quadrant results in flows that cross national boundaries but within the same organization. Here international differences may affect such flows. And finally the upper right quadrant represents global e-business that has both organizational and national trans-border flows. Had the Swedish buyer bought the product from the U.S. vendor, the transaction would include trans-border flows and constitute global e-business. Each type of e-business has a set of unique issues that deserve detailed analysis. Take domestic internal e-business for example. Ownership of data and incentive systems within an organization may affect the flow of information (Van Alstyne, et al., 1995). As we move away from domestic internal e-business into other types of e-business, new issues arise and complexity increases. This framework captures these characteristics of e-business and allows us to analyze technologies and policies that support, and sometimes limit, such transaction flows. Instead of discussing each type of e-business, we will focus our analysis on global e-business using Web aggregation as an example. The comparison shopping engine used by the hypothetical Swedish user is an example of regional comparison aggregation. Lack of a global aggregation tool makes getting useful information from global sources a difficult task. In the example, the Swedish user has to know the existence of multiple regional comparison aggregators and use them to perform global comparisons by manually integrating results. Global aggregation will automate this process, help the user "think globally", and bring many benefits to the global market. The rest of the thesis will analyze various technology and policy issues for global aggregation. Section 2 describes web aggregation, including types of aggregation, enabling technologies, current status. Section 3 presents worldwide price dispersion phenomenon, which to some extent indicates the inefficiency of the global market. Section 4 discusses the benefits of aggregation and global aggregation. We devote section 5 to the analysis of deficiencies, technology and
policy challenges to global aggregation. Finally we conclude in Section 6 and discuss future research.

- 13 -

Relationship aggregation is a service that collects information from multiple relationship accounts on the user's behalf. These accounts often include various rewards program accounts, financial accounts, online bills, etc. Figure 5 gives an example of financial account aggregation.

Tell a friend@

Sign OutO

Update all accounts here

You have I bills due! [Last Login:13 MAR 200216:3420]

13 MAR 2002

000000
Figure 5. Example of Relationship Aggregation (screenshot from www.ewise.com.au).
2.2 Current Status of Web Aggregation
Comparison and relationship aggregation services emerged in the U.S. a few years ago and have been widespread around the world. Comparison aggregation has been more widely adopted than relationship aggregation, but we began to see a quick spread of relationship aggregation since 2001. Despite their worldwide presence, aggregation services are primarily operated regionally with little trans-border information flow. 2.2.1 ComparisonAggregation We surveyed comparison aggregators in the U.S., Europe, and Asia and listed a few examples in Table 1. Few of them qualify as a global aggregator according to our framework. AddAll.com, a book comparison aggregator, does qualify to be a global aggregator in that it queries bookstores from a handful countries and displays prices in more than a dozen currencies. Aggregators that have an international presence often have regional operations and aggregate information from regional sources. For example, most information sources of PriceGrabber are from North America, while PriceRunner and Kelkoo are more Europe oriented. Even though mySimon and Dealtime have a presence in both North America and Europe, there is little overlap of information sources between the two regions. More interestingly, little overlap is found in European aggregators unless a vendor has presence in multiple European countries. This indicates that information flow for comparison aggregation seldom crosses national boundaries, which qualifies their services to be regional, rather than global, according to the framework we described earlier.

- 14 -

Table 1. Comparison Aggregators and Their International Presence.
Aggregator mySimon DealTime' PriceGrabber PriceRunner Kelkoo
Country U.S., France, Germany, U.K. U.S., U.K. U.S., Mexico, Brazil Denmark, France, Sweden, U.K. Denmark, France, The Netherlands, Italy, Norway, Spain, Sweden, U.K.
Capabilities are also different among aggregators, and sometimes among different regional sites of the same aggregator. Most of the aggregators do not provide easy search for the same item across different countries, i.e., a user has to repeat similar search steps for each country. Only PriceGrabber allows users to see the prices for the same item in a different country by clicking the country link. We should point out that PriceGrabber aggregates one set of information sources in the U.S., then presents them to Mexican and Brazilian users with estimated country specific tax and shipping costs. This is essentially not a true cross-country search. Most of the international aggregators maintain similar user interfaces across their sites, however, mySimon does have different ones for each country. On the U.S. site, one can search by keywords across all categories, but this is not offered on sites outside the U.S. Figure 6 shows the search interface for its French and U.K. sites, with one having no keyword search at all and the other having search by model name after you have navigated to a particular category. These differences indicate the disconnected nature of current international aggregators and pose difficulties in achieving global aggregation.

Engine

SpecFile #Relation=oanda #Source=http: //www.onda.com/convert/classic #Method=POST #Param="value"="1" "date"=#date# "datagefmt"="us" "exch"=#exchanged# "expr"=#expressed# #Attribute=Rate#String #Begin=<font color="#bcOal7"> #Pattern=>(.?) (\d+) (\.) (\d+)< #End=</ font>
XML Output IV<?xml version='1.0'?> <DOCUMENT> <ELEMENT> 1.14745 </RATE> </REE> </DOCUMENT>

"Tj

Figure 7. Screen-Scraping Technology.
The upper left of Figure 7 shows how a user would normally see a web page via a browser. To its right is an excerpt of the underlying HTML file that generates this display. The Cameloen screen-scraping engine uses a set of declarative rules predefined in a spec file shown at the lower right to sift through the HTML file and find specific information element of interest. With a relational interface, Cameleon can accept SQL queries and return results in either a table format or as an XML document. Figure 7 shows that with such a tool, an online historical currency conversion web site becomes a relational database from the user or software client's point of

- 17 -

view. That is, the web site is defined as a database with a simple relation called oanda with four attributes: rate, date, expressed, and exchanged. Variations and improvement can be made to the technology. These include the development of efficient parallel query processing, database connectivity middleware, automated spec file generation, etc. For example, Parker's Edge in Australia developed a graphical tool to allow non-programmers to pick and choose the content to "scrape" (Celent Communciations, 2002). According mySimon, they developed a technology called Virtual Learning Agent that can be trained by mimicking human behavior to extract certain content. In addition, aggregators, especially comparison aggregators that do not need login credentials from individual users, also use information discovery technologies (i.e., web crawlers) to find online information. This simple technique becomes a powerful information aggregation tool when it is implemented with parallel processing capability to simultaneously submit and process queries to multiple sources. Note that technically as long as an aggregator has spec files that match the sources (and is granted access to the sources by its users in the case of relationship aggregation), the tool can collect useful information without any special arrangement with the information providers. With these capabilities, information such as prices or account balances from any online vendors or service providers can be aggregated into a consolidated view. While screen-scraping is flexible and effective even with non-cooperative sources, it is errorprone and can be costly to maintain. All spec files have to be kept updated. The overall computation model is not efficient, which hinders the scalability of the technology. These drawbacks can be overcome with standards-based information sharing, which we will discuss next. 2.3.2 Standards-basedData Exchange Traditionally organizations that engage in information interchange agree to a common coding standard and often require dedicated connections and special applications, such as Electronic Data Interchange (EDI). The implementation and operation of such data interchange technology are expensive. The emergence of Extensible Markup Language (XML) a few years ago made it possible to develop standards that are easier to implement. The financial industry has developed a number of XML based standards, such as Open Financial Exchange (OFX), Interactive Financial Exchange (IFX), Quicken Interchange Format (QIF), and many others. Participating organizations can publish their data using one or multiple standards via their web services. Aggregators can retrieve such data more efficiently and accurately. Although standards-based data exchange technology solves some of the problems of screenscraping, it has its own drawbacks. First, there are too many of them 4 , which makes it difficult for organizations to follow or very costly to support multiple standards. Second, it requires agreement to a standard by multiple organizations, which requires some coordination efforts.

See, for example, Mary Knox from Gartner Group, "Commentary: XML Standards in the Financial Industry", November 8, 2001, news.com.com/2009-1001-275607.html; and John Rymer's "Why 90% of XML Standards Will Fail" at http://www.zdnet.com/filters/printerfriendly/0,6061,2689765-84,00.html.
Sources that do not join the agreement will continue to require screen-scraping technique for aggregation. 2.3.3 Technology Adoption by Aggregators Most comparison aggregators use automated discovery and extraction tools to collect online information, sometimes with manual assistance. PriceRunner's FAQ section explains how a typical aggregator collects information: "Some aspects of the information-gatheringare automatic thanks to our 'price agent technology' (methodical data collection), while other aspects demand manual work. Our software runs daily to crawl the web updatingprices. We have an army of Pricerunnersmanually collecting data everyday." Although none of the comparison aggregators we studied disclose how "extraction" is done or if they use "screen-scraping", the techniques they use follow the general principle we described earlier. Some aggregators seem to largely rely on manual data collection. Here is an excerpt from PriceScan's FAQ: "We obtain pricing informationfrom magazine ads, vendor catalogs and vendor web sites. Some vendors email orfax us daily with price changes and others have no direct contact with us." Some provide mechanisms for vendors to directly input their prices, e.g., Shopper.com has such practice (Baye et al., 2001). For relationship aggregators an automated discovery tool is less critical because they cannot aggregate until users provide their login credentials for the organizations with which they have relationship accounts. Initially, most financial account aggregators used screen-scraping data extraction technology. With the emergence of XML based standards, about two thirds of financial accounts are aggregated using screen-scraping, with the rest using one of the industry standards (Celent Communications, 2002). To gain back some of the control over data, the financial industry is in favor of standards-based aggregation and recently has proposed a framework for next generation aggregation based on FAST: Financial Agent Secured Transactions. Relational aggregation requires an initial configuration that collects login credentials to multiple accounts. Most aggregators store such information centrally on their servers. An Australian company, eWise (www.ewise.com.au), developed a technology that allows users to store their login information locally on their PC or other devices, similar to desktop aggregation software such as Quicken and Microsoft Money. This distributed solution is promising for aggregation in regions that require a secure login device to access account information.

- 25 -

Table 3. Price Dispersion of Representative Countries.

Price Range

(US $)

Mean Price

Range/Min 0.686 0.280 0.233 0.202 0.278 0.348 1.878 1.878 and Brazil.

Coefficient

of Variation
US [949, 1600] 1203 UK [429, 1829] 1606 [1565, 1929] 1677 France Germany [1529, 1929] 1734 [1750, 2237] 1975 Sweden Denmark [1671, 2254] 1959 Europe [783, 2254] 1758 Worldwide [783, 2254] 1633 Note: Worldwide statistics excludes Mexico
0.092 0.060 0.078 0.073 0.077 0.134 0.132 0.220
Even though U.S. has the lowest price level, it has the second highest dispersion in term of coefficient of variation, next to Denmark. This value is very well in line with the results of four million observations of 1000 consumer electronics prices in the U.S. (Baye et al., 2001). Note the last row of the table is the worldwide dispersion measurements, which is not the total of each column. Their large values indicate substantial worldwide price dispersion. Let's look at US prices in more detail, shown in Figure 11. These 53 unique price observations do not include SonyStyle US, Sony's online store in the U.S., and major consumer electronics vendors like BestBuy and CircuitCity, which offer the produt at the same "official" price: $1299.99. We can see from the figure most prices are at or below this price level. The average price is $1203; if we exclude one extreme observation, the average price drops to $1196. That is, the average price is around $1200, which is 7.7% below the "official" price. This is close to the range of 9-16% lower prices for books and CDs on the Internet estimated by Brynjolfsson and Smith (2000). More importantly, U.S. average price is 26.3% lower than the worldwide agerage.

=-requency

-Cumulative %

100% 80%

40%75 20% i """ I men More 24 0%
Figure 11. Price Distribution in the U.S. (N=53).
3.3 Explanations for Price Dispersion
Textbook economic theory predicts that under perfect competition (e.g., Bertrand competition) commodity prices converge to one price, the so-called Law of One Price. But real world markets have shown no evidence to support this. The price dispersion phenomenon has been explained as a violation of one of the Bertrand assumptions: product homogeneity, zero search costs, or perfectly informed consumers (Brynjolfsson and Smith, 2000). Other theories are being developed, e.g., Rauh shows that when buyers and sellers make small mistakes in their beliefs about prices a temporary equilibrium exists to result in price dispersion. In our case study, we looked at prices for one single product. Although it does have two models for video output (i.e., PAL and NTSC), this distinction is marginally important because its MPEG recording format allows for easy processing on a PC, which does not use the video output. In addition, many TV sets support dual video standards. So this product can be regarded as homogeneous. Regional aggregators can help lower search costs, which should lead to convergence of prices (Bakos, 1997). Whether all consumers are perfectly informed about price distribution is in question. Although comparison aggregation has gained some popularity, none of the popular comparison aggregators ever make to the top 50 most visited sites in the U.S. measured by Jupiter Media Metrix. In domestic e-business, it is possible that the three assumptions are met to some degree. So Brynjoflsson and Smith (2000) turn their explanations to heterogeneity in other vendor characteristics that are not directly observable and suggest trust in vendor brands may play a major role for price dispersion, which they further explored (Smith and Brynjoflsson, 2001). In the context of global e-business, trust certainly plays an important role (e.g., would the Swedish buyer trust the U.S. vendor even though it can potentially save him nearly 40%?). Even the basic assumptions could be violated in the setting of global e-business. Although in terms of features the camcorder is nearly homogenous worldwide, other factors exist to result in heterogeneity. The product may be assembled in different plants that have different cost structures (e.g., plant in Malaysia vs. plant in Japan). Manufacturers often use different labeling to segment the market, e.g., using different languages for product manuals in different regions. Warranty and other post sales services are often divided into regions, e.g., camcorder sold by the U.S. vendor to the Swedish buyer may not be covered by manufacturer's warranty in Sweden if it is a U.S. only model6. We already know that even in domestic e-business, not all buyers use comparison aggregators to reduce search costs. In global e-business, search costs are much higher because few buyers search for worldwide prices. Even worse, most aggregators are regional, making global price comparison a difficult task. We gave a hypocritical situation in the motivational example, but in reality chances are the Swedish buyer does not know any price information in the U.S.

NewProduct

(2) 1+ a where p, and x1 are post-introduction price and quantity; p* is a virtual price that effectively set demand to zero, therefore p*xo is zero; a is price elasticity for demand, and $ is the percentage of price decrease. Equation 1 is for consumer surplus of the new product effect; equation 2 is for the price reduction effect from increased competition.

- 29 -

4.2 Benefits of Global Comparison Aggregation Global comparison aggregation will be an effective tool for global price dispersion studies. As we discussed earlier, without an aggregation tool, we would have to rely on aggregate data or time consuming and labor intensive market surveys to obtain price data. These methods are poor in terms of level of specificity (e.g., price of a particular product) and timeliness (e.g., current price rather than price of two days ago). It was an aggregation tool that made it possible to collect four million observations on 1000 products over several months of time. Using aggregation I literally collected global prices of the case study within hours. Most of the time was spent on merging the data from multiple regional aggregators to facilitate comparative analysis. Had we have a global aggregation tool, this process would have been automated and instead of having only snapshots, we can collect data over time to facilitate global longitudinal price analysis. Timeliness is important. New products are introduced to the market on roughly an annual basis due to the effect of Moore's law or marketing strategies. Prices of a certain product therefore change frequently. Over time, some vendors may start carrying a product and likely some vendors may cease to carry that product. This dynamic nature of market is obvious in the online auction market. Timeliness of aggregation allows us to study the dynamics of the market. Table 4 summarizes the changes of vendors and their prices within two hours. Most of the vendors lowered their prices (except for one increased by $20) as reflected in the lower average price of the later sample; one vendor was not listed and four new vendors were reported in the second snapshot. Note the experiment was done on March 3, 2002, one week before the case study. The average price reported by the same aggregator dropped to $1199 at the time of the case study.

7 See, for example, October 2001 report from Celent, "Finding the ROI from Online Retail Banking Applications";
July 2001 report from Forrest Research, "Account Aggregation: The Elusive ROI"; and June 2001 report from Datamonitor, "Financial Services Aggregation 2001: Drivers, Trends and ROI". 8 From an article by Mark Bruno, "Questions about Aggregation", US-Banker.com, October 2001. 9 Datamonitor December 2001 report, "Account Aggregation: The Right Way Forward?" '0 UNISYS July 30, 2001 news release citing Datamonitor's June 2001 report, "Unisys Teams up with Corillian to Bring New Consumer Service to UK Financial Service Sector". 1 According to Catherine Allen, CEO of BITS, who presented "BITS Perspective on Account Aggregation" at the 2 nd Account Aggregation Conference in April 2001. 12 Tiffany Bass Bukow, founder of MSMoney.com, "Get Your Financial Life Organized with Less Effort", www.msmoney.com/2001/11/30/billpay.htm. In contrast, a millionaire spends 20 hours a month. 13 See, for example, ftp://ftp.bls.gov/pub/special.requests/lf/aat39.txt.

- 32 -

Financial Services Roundtable (2000) shows that consumers of the FSR 90 institutions can save $17 billion and 320 million hours per year. Comparing with this study, our estimation is very conservative, yet it still reflects enormous consumer benefits. As an example of global relationship aggregation, Maxmiles already provides services for international travelers to aggregate their rewards program accounts from different countries. As the world economy gets more integrated, more people will have relationship accounts beyond frequent flyer programs that are located outside of their home country, therefore, we will see increasing needs for global relationship aggregation. We need to evaluate this great opportunity and its benefits as it emerges and grows. We have shown the Swedish buyer in the motivating example could save substantially using global price information. Using the economics technique introduced in the beginning of the section and a set of assumptions, we further gave a rough estimate of consumer benefits that global aggregation could potentially generate. Next, we are going to analyze technology and policy challenges for global aggregation services to realize these benefits.

- 33 -

CHALLENGES FOR GLOBAL AGGREGATION
There are various deficiencies of existing regional aggregation services, most of which are technology related. Moving from regional aggregation to global aggregation, more technological challenges arise from increased complexity. National and international policy issues will also arise in cross-country information flows. We will analyze these challenging issues and briefly discuss other trade barriers related to other types of trans-border transaction flows.

- 36 -

Legendary Independent Bookstore
Be one of the 1st to review this store

$62.00

Tax: check Aste

Ship: $2.50

check site

Huge selection of books

-- UP TO 50% OFFI

6 Reviews

$70.45
Tax: check site Ship: $1.99

I SES75FLLER.5 I

[TECH BESTSELLERS

C-OMPUt TER

USED BOOKS

EABOUT THIS BOOK

Synopses & Reviews More Books by Trevor Hastie Find Related Books

P sopel

Figure 12. Example of Errors due to Lack of Timeliness: Price Reported by Aggregator (top) vs. the Source.
Some times the errors are easy to recognize. See Figure 13, $2 is way too low to be a price for a digital camcorder. The aggregator might have mistaken the price for an accessory of the camcorder and mixed it with prices for the camcorder.
sort by: Store Name I sort by: Pige Delivery Score:

FocusCamera com

Store Rating: 6.
Delver 77% On scre Time $2.00

02/24/2002

No Special Offers WSee it at

FocusCamera.com

Sony Sony DCR IP Micro MV Diital Camcorder NTSC linternational Model sore fr |SAVE BIG the RIGHT WAY - Gomez image

Digital Net Shop

Store Rating- N$
Deivery Score: n/a $1079.99

''2127 /2002

No Special Offers
See See site for more information store for image
DCR-1P5 DIGITAL MICROMV CAMCORDER

* See it at

Figure 13. Example of Obvious Errors.

- 37 -

Figure 14 shows another type of easy-to-spot error; one of the entries has an incorrect model DCR-IP7 is a model that is more advanced than DCR-IP5, even though we requested comparison for the DCR-IP5 model (though it is nice to get a better model with a lower price).
Sony DCR-IPS: ComDare Prices
- Be one of the first to review this product Format Type: micro my - LCD Display - See more product features

c xanin

22 Reviews
DCR-IP7E.MICROMV CAMCORDER,1OX OPTICAL/120X DIGITAL ZOOM.CARL ZEISS.2,5 INCH LCD SCREENCVF,I LINK.

Tri-State

Computers

LJI "' Sa :pe

ial dav B m

13VIDEO 133 Reviews

SONY DCRIP5 microMV DIGITAL CAMCORDER, DIGITAL VIDEO CAMCORDER
$1599.99 m c CameraClub.com
Figure 14. Another Example of Obvious Errors.
Sometimes errors are not easy to identify. See Figure 15. Are there any problems with the displayed information? Occasionally aggregators do have duplicates in their listing, but this is not the problem here. Although there are two entries for each vendor, the prices are not the same. Following the links we found that both vendor's websites reported the second price. For the first vendor it could be an error (e.g., not being able to uniquely identify a vendor during an update so that outdated information is not overridden with the latest); for the second, however, the first price was a special offer to consumers who are willing to give away their email addresses.

special treaties between regions. The variations range from 0 to 25% of the listing price in European countries (see Appendix 1). The information listed in Table 5 is a hybrid of the two concepts for price with some missing components. This makes aggregation and meaningful comparison difficult. McCarthy and Buvac (1994) illustrated this problem with an example of different prices of the same GE aircraft engine perceived by different organizations, such as the U.S. Air Force and U.S. Navy depending on whether the price includes spare parts, warranty, etc. Another problem not explicitly shown in Table 5 is how the aggregators identify the same product from different regions. In the process of manually composing the Table, we noticed that the model numbers are different between laptops in the U.S. and those in Europe. We recognize their similarity (in this case identity except for the model numbers) by examining the configurations (e.g., CPU speed, hard disk capacity, weights, etc.). The fact that manufacturers often market the same product with different names in different regions makes it difficult for the aggregator to recognize their identity. This problem is best described from the following Camera example from Focuscamera.com: ". a USA Minolta Maxxum is a Minolta Dynax overseas, the USA Canon EOS Rebel 2000 is an EOS 300 overseas, PentaxIQ Zooms are Pentax Espios overseas, etc." Conversely, when models with different features are named the same or slightly different in different regions, aggregators sometimes cannot recognize the distinction. In the Sony DCR-IP5 case study we found that some vendors label the product as DCR-IP5E to indicate that it is an international model compatible with the PAL standards rather than the NTSC standards in the U.S. What makes it worse is that most vendors use DCR-IP5 for both the NTSC model and the PAL model. Although this does not cause big problems because of its MPEG recording format, for other types of products this could be an issue. An aggregator should not compare apples with oranges. The preceding discussions can be summarized into three issues, with increasing complexities in the order shown below: 1. Representation - how do we represent things 2. Composition - what are the components for the thing 3. Recognition - what is the thing we are really referring to How can we solve these problems so that users will get accurate, consistent, and meaningful aggregated information? Approaches of standardization and tight coupling through software engineering are extremely difficult to implement in the case of global aggregation where sources are highly autonomous due to the fact they are owned by many organizations located in different regions. The mediated architecture (Wiederhold, 1992), logic based context interchange approach (McCarthy and Buvac, 1994; Madnick, 1999; Goh et al., 1999), and the use of ontologies (Fensel, 2001) provide a promising solution. The mediated architecture achieves scalability by adding an abstract knowledge layer on top of data the data layer. The context interchange approach captures the afore mentioned three issues

injunction because of too little evidence of harm. Details can be found at http://www.gigalaw.com/library/ticketmaster-tickets-2000-03-27.html. 15 News release of BH1B on February 9, 2001. Full text can be found at www.bhb.co.uk/press-release.asp?id=255.
the way in which data is extracted may differ by region. These complicate the implementation of global aggregation and add frictions to international data flow. 4.2.2 PrivacyProtection Since aggregation is still a fairly new phenomenon and little aggregation data flow is transborder, the differences in database protection have not caused major litigations so far. But things are different in the area of privacy protection, where the differences already have caused troubles for trans-border personal data flow. For example, American Airlines majority-owned Sabre reservation system is prohibited from transferring personal data of Swedish passengers according to the ruling of the Swedish Data Inspection Board in 1995. Even though some of the information that the Sabre system would like to collect (e.g., health condition for special care or religious belief to avoid serving religiously prohibited meal) is a crucial part of the business, this ruling was held after two unsuccessful appeals and it is pending for final hearing before Sweden's Supreme Administrative Court. More recently in 1999, Microsoft had to pay $60,000 to settle charges brought by Spain because Microsoft's disclosure regarding Windows registration information is not clear to Spanish consumers (Kemp, 2001). What went wrong? The differences in privacy protection policies have caused a "data embargo" situation, where data is prohibited to flow into a country whose privacy protection is regarded inadequate. This guideline is set in article 25 of the E.U. Data Privacy Directive, which took effect in 1998. Lack of privacy protection (in terms of level of protection, not number of privacy laws) in the U.S. has resulted in the data embargo to Sabre and the litigation against Microsoft. What are the differences? Bennett (1997), who suggested using divergence instead of difference to emphasize the process rather than the status, proposed three aspects of the issue: policy content (the principals and goals), policy instruments (the implementation), and policy outcomes (effects on individual and organizational behavior). Nations are converging on the contents of privacy policy to a set of standards, i.e., the First Principles of information privacy that includes data quality, processing transparency, treatment of sensitive data (e.g., religious belief), and enforcement mechanisms (Reidenberg, 2000). However, divergence exists in policy implementations. The E.U. takes a centralized approach by establishing an independent agency in each member state to oversee privacy. But the U.S. has a highly decentralized system (Glancy, 2000) and largely relies on industry self-regulation. Since there is no effective measure for the outcome of privacy policies, the principle of independent oversight agency may be used as a test of "adequacy" (Bennett, 1997). This seemingly trivial implementation divergence, in fact, has deep root in national differences in the normative roles of the state, the market and the individual, e.g., liberal market-base governance vs. social-protective rights-based governance (Reidenberg, 2000). These deep-rooted differences will stay and require other mechanisms to reconcile. As a temporary resolution, the U.S. and the E.U. reached the "Safe Harbor" agreement in 2000. But its legitimacy, scope (i.e., only industries under the jurisdiction of the FTC and DOT), and enforceability (e.g., FTC is charged to protect U.S. companies and citizens, not the Europeans)

Staff Working Paper released on 2/13/2002. relationship aggregation user needs to sign an agreement, often by clicking an "I Agree" button when signing up for aggregation services, which grants the service provider the right of collecting personal information on the user's behalf.

17 Commission

Another type of flow, which we did not include in the framework, is people flow. Although the electronic connectivity greatly enhances interactions between trading partners, which can potentially reduce the needs for traveling, face-to-face meetings cannot be completely replaced due to complexities of some transactions and other factors, such as trust. Enormous obstacles for mobility of people still exist. For example, one can easily buy an airline ticket online with the lowest fare to a foreign country, but the person may not be able to use the ticket because of visa problems. This often happens to travelers from developing countries to the developed world, where visa denials occur for various reasons. The analysis of this issue is out the scope of this study, largely due to its complexity.
6 CONCLUSIONS AND FUTURE RESEARCH 6.1 Conclusions
We introduced a framework to categorize e-business according to organizational and national trans-border transaction flows. This framework captures a wide range of e-business practices and their related issues. We have focused our analysis on technology and policy issues for global e-business that has both types of trans-border flows using global aggregation as an example. Despite the global presence of comparison aggregation and the worldwide spread of relationship aggregation, most of the services are offered regionally, not globally. Lack of global information can result in inefficiency in the global market. Our price dispersion case study shows that the worldwide prices for DCR-LP5, a Sony digital camcorder, can differ by nearly three times. A global aggregator can close the information gap and bring benefits to researchers, manufacturers, and consumers. We discussed these benefits and showed how to estimate consumer surplus from a more competitive market. The major benefits of relationship aggregation are twofold: time savings and potentials for optimization and other value-added services. While this service is still in its early stage, it has been steadily growing within the U.S. and started to spread to the rest of the world. As the world economy gets more tightly integrated, more people will have business relationships with international service providers, creating great opportunity for global relationship aggregation. While we see a promising future for global aggregation, we also realize the existence of many challenges and obstacles. We identified various deficiencies of existing aggregation services and suggested solutions that involve technologies such as the mediated architecture, use of ontology and Context Interchange techniques. The capabilities of these technologies combined will be demonstrated in the proposed global aggregator prototype discussed shortly for future research. Another set of obstacles for global e-business come from policies that govern the transaction flows. Directly related to global aggregation are information policies for international data flow. We focused on policies of property rights in databases and privacy protection. We speculated that impact on global aggregation would be limited, but uncertainties and international differences pose frictions to trans-border information flow that is crucial to global aggregation.