Pricing differences between dotcoms and multi-channel retailers in the online video market.

ABSTRACT

Focusing on a homogenous product (videotapes), we use a unique data set with a total of 4800 price observations to compare the pricing behavior between online branches of six traditional retailers and six online-only retailers. We find that posted prices by the pure Internet players are significantly lower than posted prices by the multi-channels online, 6.42% on average. However, it is only 3% lower on average in the full price sense (including shipping costs) and such differences do not seem statistically significant. Further, price changes by both types are few but adjustment magnitudes are large, indicating that both types of online retailers do not change their prices frequently although many claimed that menu cost might be as small as negligible for the online market. Price dispersion seems rather large by both types, around 30%, and statistic evidence shows that it is significantly lower among the dot corns than among the multi-channels online in the sense of posted prices, but the contrary in the sense of full prices. The empirical evidence suggests that the online videotape market is far from perfect competition. Market power and offline pricing behavior influence pricing efficiency in the Internet.

1. INTRODUCTION

With the rapid growth of e-commerce, more and more conventional retailers have started selling online. It is interesting to see how these conventional retailers compete with online-only retailers on the Web. Online retailing promises the potentials of low barrier of entry, easy access to information, and low transaction costs. These features imply that online retailing has the potential of realizing the economic ideal for the effective-competition market: low search costs, fierce price competition, low margins, and weak market power. However, if conventional retailers can successfully translate their market power and branding to online markets, we can rationally expect that conventional retailers will charge higher prices than online-only retailers when they go into online markets, if this hypothesis can be verified by empirical evidence, it would contribute to our understanding of efficiency in the Internet market.

Empirical studies on online market efficiency so far do not answer this question. Clay et al (2002) compared prices of books sold by thirteen online and two physical bookstores. They found that average prices in online and physical stores were similar after controlling for book characteristics, in a more comprehensive study, Brynjolfsson and Smith (2000) examined prices of books and CDs sold through Internet and conventional channels in 1998 and 1999. They found that online prices were 9-16 percent lower than that in conventional stores, and the price dispersion was lower among online retailers than that among conventional stores, after weighting the prices by proxies for market share.

Morton, Zettelmeyer, and Risso (2000) compared prices of cars sold in online and conventional channels. They found that on average, online consumers paid 2% less than do offline consumers. Both Bailey (1998a) and Brynjolfsson and Smith (2000) found that online menu costs were lower. Based on comparing online retailing and traditional retailing, these empirical results lend support to the hypothesis that online markets are more pricing efficient than offline markets.

This study takes an approach different from the earlier ones. We explore the different pricing behavior between online-only retailers and online branches of multi-channel retailers. In particular, we seek to contrast the pricing of multi-channel retailers with those of online-only retailers and derive implications. As far as we know, this is the first and only study on the online videotapes market from such a perspective.

Since both types of retailers will presumably be exposed to the same set of shopping pressures, one might anticipate that prices would tend to converge. However, Clay, Krishnan, and Wolff (2001) found that prices in the online book market did not converge over their sample period. They partially attributed price dispersion to the fact that stores had succeeded in differentiating themselves although they were selling a commodity product. They argued that high priced online stores seemed to use the Web primarily as a way to advertise, rather than as a vehicle for sales.

To explore different pricing behaviors between online-only retailers (thereafter referenced as DotComs) and the online branches of multi-channel retailers (thereafter OBMCRs), we investigate prices of videotapes sold on the Web by both DotComs and OBMCRs. We will test: (a) whether online-only retailers offer prices lower than online branches of multi-channel retailers who sell mainly through traditional channels; (b) whether the two types of online retailers adjust their prices frequently and at the similar frequency or magnitude; and (c) whether the dispersion of prices exhibited by online-only retailers is smaller than that by the online branches of multi-channel retailers. Section 2 discusses our hypotheses on online pricing behavior by online-only and multi-channel retailers. Section 3 describes data collection methodology. Results of our empirical analysis are discussed in Section 4. Section 5 concludes.

2. HYPOTHESES

Although marginal cost pricing may never prevail in the Internet markets, the lower online search costs will lead to lower prices and lower price dispersion for homogeneous goods (see, e.g., Bakos, 1997, 1998). As mentioned earlier, empirical studies have shown that online retailers are more efficient than land-based retailers and will charge lower prices (see e.g., Brynjolfsson and Smith, 2000; and Morton, Zettelmeyer, and Risso, 2000). Since multi-channel retailers may coordinate prices across their different channels in order to prevent destructive competition among them, they may charge higher prices on the Web than their online-only competitors, although it is not necessary for a multi-channel retailer to charge the same prices online and offline. Thus we make the following hypothesis.

[H.sub.1]: Online-only retailers charge prices lower than do online branches of multi-channel retailers.

Frequent Changes in prices permit retailing system to adjust more rapidly to variations in supply costs and demand, and therefore enhances market efficiency. Bailey (1998a, b) found that online menu costs were lower than that in traditional stores. Brynjolfsson and Smith (2000) concluded "lnternet retailers' price adjustments over time are up to 100 times smaller than conventional retailers' price adjustments--presumably reflecting lower menu costs in Internet channels". These results lend support to the hypothesis that online retailers change their prices more frequently and the online market is more efficient. But even if online menu costs are very low, online branches of multi-channel retailers may not change their prices frequently, because they have to integrate their online and offline retailing operations and hence their online pricing behavior must be part of their integrated strategies of doing business in both the online and offline markets. Therefore, our second hypothesis is

[H.sub.2]: Online-only retailers change their prices more frequently than do online branches of multi-channel retailers.

According to search theory, the low online search cost would result in low price dispersion. Stigler (1961) analyzed price dispersion across firms for a homogenous good. He showed that higher search cost in a market with imperfectly informed customers would result in a greater dispersion of prices, and greater price dispersion would result in more searches. Following Stigler's seminal paper, both theoretical and empirical studies in the literature have attempted to explain the existence of price dispersion. For example, Nelson (1970) extended Stigler's model by assuming that consumer behavior is affected by both price and quality. Pratt, Wise and Zeckhauser (1979) theoretically showed several cases where even small search costs may lead to substantial price dispersion. Recently, Sorensen (2000) found empirical evidence that price dispersion could be substantial, even in a small local market for prescription drugs.

Empirical studies on online prices have showed that there exist great price dispersions in online markets. Clemons, Hann, and Hitt (1998) found that the difference in prices across online travel agents was as large as 20 %, even after controlling for observable product heterogeneity. Clay, Krishnan, and Wolff (2001) investigated price dispersion in the online book market. They found that unweighted price and standard deviation were stale or rising over their sample period. Brynjolfsson and Smith (2000) found price dispersion was not lower in Internet markets for books and CDs as compared to conventional markets. However, after weighting these prices by proxies for market share, they reversed their conclusion to the statement that the dispersion was lower in the Web than in conventional stores. They attribute this phenomenon to the dominance of certain heavily branded retailers, in addition to several other factors such as asymmetric information, search costs, and retailer heterogeneity. Morton, Zettelmeyer, and Risso (2000) investigated the prices of cars sold in California and found that price dispersion declines with online sales.

As multi-channel retailers' pricing behavior would inevitably have influence upon demand in their conventional stores, given the early empirical findings that price dispersion for traditional retailers may be greater than that for online-only retailers, we can expect that when traditional retailers go to online markets, they may charge prices quite differently among them, as what they do in brick-and-mortar stores. Thus, we make the following hypothesis.

[H.sub.3]: Price dispersion across the online branches of multi-channel retailers is wider than that across online-only retailers.

3. DATA

We investigate the online videotape market for testing our hypotheses. This market is chosen for several reasons. First, as far as we know, there does not exist such study on the online videotape market yet. Second, this market is one of the most successful ones that have migrated online and enjoyed considerable growth and sales. Third, the fact that the videotapes are relatively homogeneous makes data collection tractable and price comparison meaningful.

Two types of retailers are selected: those that conduct commerce only through the Internet and those that sell also through traditional channels. The main criterion that a retailer must meet is that it sells a general selection of titles and selling prices are posted on its Web sites.

Following the 100 Hot Shopping Sites by Web21.com ("The most rigorous and widely recognized web ranking service", Brynjolfsson and Smith 2000), the rating by bizrate.com and the PowerRanking for Movies by Forrester Research, six top DotComs that sell a general selection of videotapes were selected. The four of six top OBMCRs were selected according to the ranking in Darnay and Piwowarski (1999) on the conventional stores in the category of record and prerecorded tapes. Since other stores in the category are either small or have no online branch, we picked up two top OBMCRs (Tower and Djangos) according to the rating by bizrate.com. The market share of these retailers together is dominant thus their pricing behavior is certainly representative in the online videotape market. We did not include the more specialized retailers in specific entertainment niche to minimize selection bias.

Next, a selection of titles for comparison must be made. A total of 50 titles were examined. Half of them (25 titles) were selected as an even mix of the top bestsellers among the retailers when the study was initiated, while the rest were chosen randomly. The reason for such a combination is that if all the titles were selected from a specific bestseller list such as the Amazon's VHS top-sellers, which contains only popular titles by a specific retailer, the results may be biased as these titles are likely to be loss leaders selected by the respective retailer. However, if all the titles were selected randomly, one major trend of pricing behavior would be overlooked because competition in the bestsellers' niche is crucial for any market structural analysis. We refer to the first category of titles as "popular" and the second as "random" from now on. Further, during the data collection process, we took extreme care to make sure that the version and other features are exactly the same for the same title.

Then, the frequency of temporal data collection must be decided. The interval between data collection should not be so long as to miss price changes that might occur. But if the interval chosen were too short, the data collection would be unnecessarily costly. Thus we collected data once every five days in this study. We started our data collection on July 12, 2000 and stopped after August 16, 2000. In total, we have 4800 price observations.

We carried out the observations by accessing the Web sites of the selected retailers and recording the prices of the selected titles, both for the DotComs and OBMCRs. Note that the prices posted on the Internet by offline retailers need not necessarily concur with prices found in the physical stores of these offline retailers.

Table 3-1 also includes the standard shipping costs by each retailer for the normal delivery within the United States. Since the shipping cost structure varies with different retailers, we have calculated the per item shipping cost based on their shipping cost tariff table for various baskets of typical purchases (from one up to eight items). The last column summarizes the average per item shipping cost of these basket purchases from each of these retailers.

Further, Table 3-2 summarizes the mean and median per item shipping costs by the two types of retailer, with the OBMCRs being significantly lower in this aspect (p-value of t-test being 0.008). We have also run the Wilcoxon test, which is a non-parametric test that does not depend on any distribution assumption, on the individual retailer' average per item shipping cost (the last column in Table 3-1). The result is also highly significant (p<0.0206), indicating the robustness of this phenomenon - We have calculated average shipping costs of various baskets and the results are qualitatively similar. For instance, if we include only three items as the typical shopping basket, the mean shipping cost is $2.91 for DotComs and $2.27 for OBMCRs. The corresponding p-values are 0.027 (t-test) and 0.013 (Wilcoxon test). - Note that this finding is contrary to what we observed in the other online retail markets (book, CD and DVD; see Tang and Lu 2000, Tang and Xing 2001) that exhibit lower DotCom shipping costs on average though not significant statistically. This unique feature adds certain complication to our data analysis, since one major finding is still that the DotComs price on average lower than the OBMCRs in the online video market. In next section, we will address this issue by measuring the "full price" which includes the shipping cost in some explicit way.

4. EMPIRICAL RESULTS

4.1. Price Levels

We took several parametric and non-parametric statistical tests to analyze the relative levels of prices between these two types of online retailers. Table 4-1 summarizes the mean prices of our data sample at the most aggregate level.

Table 4-1 shows that posted dollar prices charged are clearly lower by DotComs than by OBMCRs, on average of $0.90 or 6.42%. We also calculated the percentage of the posted dollar price by each retailer for each title at each date, relative to the list price of each title. The percentage price level is more comparable across titles because it shows clearly how much discount each retailer gives to each title compared to the regular list price for each title. It seems that the DotComs sampled in our survey gave 21.87% discount on average compared to 16.62% by the OBMCRs, indicating lower average price level in this aspect as well. Further, we have added the respective retailer's average per item shipping cost (see the last column of Table 3-1) into the posted dollar price by that retailer for each title at each date. This new price indicates how much a random shopper actually pays for each item from that retailer to have the item delivered to her doorstep by standard delivery procedure. Although the difference is only one half compared to the case without adding the shipping costs, the full prices including shipping costs are still clearly lower by DotComs than by OBMCRs, on average of $0.47 or 3%. In the rest of this paper, we will use "posted-price" as the abbreviation for the posted dollar prices, "percentage-price" for the percentage of the posted dollar prices relative to the list prices for each title and "full-price" for the prices including shipping costs as above defined.

To control for the serial correlation problem, we have run statistical tests for each data set day by day (all data and test details are available upon request). For both the posted price and percentage price, t-tests clearly reject the null hypothesis that the mean DotCom price is equal to the mean OBMCR price in favor of the alternative hypothesis that it is lower by DotCom than by OBMCR (p<0.03), for every case in the all-titles and random-titles' category. For the popular titles, all t-tests are also weakly significant (p<0.1 including two below 0.05) for the posted prices while highly significant (p<0.01) for the percentage prices. However, the results seem rather mixed for the full prices. For all titles and the random-titles' category, all t-tests are weakly significant (p<0.1) except the first two dates of July 12 and 17. But for the popular titles, none of the t-tests is significant in any conventional sense, with all p-values above 0.15. This result may indicate that the market for popular titles is more competitive. It seems that the full-price differences mainly come from the random titles. That is, for the random online video shopper, it may save money to buy from the DotComs for the random titles but not quite so for the popular titles.

Note that the power-efficiency of T-test decreases when sample sizes decrease, and this problem may become more severe with smaller sub-samples because of its distribution assumption. Thus, in order to control for the possible bias associated with distribution assumptions, we also ran the median test (see Sheskin, 1996: 232-233) that is a non-parametric test that does not require any assumption on sampling distribution properties, to ensure that this finding is robust. We have run the median test for each data set day by day. For both the posted price and percentage price, all p-values are highly significant (p< 0.01) for every case in the all-titles and random-titles' categories and marginally significant (p<0.07) in the popular-titles' category. For the full price, however, none of them is significant in any conventional sense, with all p-values above 0.25. It seems that the full prices for most titles of either category scatter around the median values without any clear separation between DotComs and OBMCRs. This finding further shadows the weak results from t-tests on the full-prices. Pricing differences in the full sense (including shipping costs) do not seem significant between the two types of retailers.

In addition to the tests on mean and median prices, we run a third test on the pricing differences between these two types of retailers: To compare the lowest prices found among all the DotComs in our sample for each title with the lowest prices found among all the OBMCRs sampled. We find that the minimum posted-price among the DotComs is $0.53 lower on average than the minimum posted-price charged by OBMCRs. Further, Table 4-2 shows that during this period the lowest posted-price is found among the DotComs 87% of the time (with the difference being over half a dollar during 49% of the time and over one dollar during 33% of the time). Note that Table 4-2 summarizes both the posted-price and percentage-price cases that coincide in this situation. For each day, if we form the null hypothesis that minimum prices are the same for both DotComs and OBMCRs, that is, half the time the lowest prices should be found by either type of retailers, this statistic should follow a binomial distribution. Then we can run the binomial test against the alternative hypothesis that it is more likely to find minimum prices by the DotComs in each day. All results clearly reject the null hypothesis (p<0.001), in favor of the alternative hypothesis.

However, the situation with the full price seems very different again. The minimum full-price among the DotComs is only $0.31 lower on average than that among the OBMCRs and during this period the lowest full-price is found among the DotComs only 49.75% of the time. Further, for each day, none of the binomial tests is significant in any conventional sense, with all p-values above 0.20. Again, pricing differences in the full sense (including shipping costs) between the two types of retailers do not seem significant.

For the individual retailers, we have calculated their mean prices (both posted and full) across the period of our study, summarized in Table 4-3. Wilcoxon test fails to reject the null hypothesis that the individual DotCom's mean prices are generally lower than the OBMCR ones, at any conventional level (all p-values above 0.20). It is also interesting to note that, for the online video market, Buy.corn does not seem to be the lowest pricing retailer as the company usually likes to claim. TheTop5.com beats it in both the posted and full prices, on average. Even Tower, a traditional retailer, ranks as the second lowest in full price, although Buy.com is still the second lowest in posted price, on average.

Overall, it seems clear from these possible tests of differences (T-test on means, median test, binomial test on minimum prices and Wilcoxon test on individual retailer's mean prices) that the DotComs generally posted lower prices than the OBMCRs. This supports hypothesis [H.sub.1]. But for a random online shopper who is more concerned with the full prices with shipping costs included, such differences become statistically insignificant.

4.2 Price Changes

We measure the price changes by subtracting the previous date data from the current date data (except the first data set), for each title by each retailer. Note that price changes in both the posted-price and full-price sense coincide in this situation. In total, we find 114 price changes out of 4200 observations. In other words, there was no price change in 97.29% of time during this period. More detailed categorizations are summarized in Table 4-4.

It can be seen that there are much fewer cases in price increases by DotComs (9, maximum increase of $4.51) than by OBMCRs (55, maximum of $4.0) and also in price decreases (6 by DotComs, maximum of $-5.0 while 44 by OBMCRs, maximum of $-4.0). Further, most price changes are at least half a dollar in magnitude, 100 cases among the 114 price changes or over 87.7%. In fact, even the price changes of at least $0.99 are almost 81% out of all the price changes (92 cases), and more than 34% of the price changes are over $2.00 (39 cases). It is interesting to notice that the one-cent-price-change strategy documented in Brynjolfsson and Smith (2000) now seems adopted only by Tower Records (11 cases of $-0.01 on July 22, one on August 16, plus two cases of $+0.01 on August 11), an OBMCR rather than a DotCom. All other online video retailers surveyed in our study change their posted prices rather sharply, if they do.

We have also measured the price changes in another way, by subtracting the July 12 data from the August 16 data for each title by each retailer, that is, the price change across the whole period during which all data were collected. Similar patterns as above can be observed. In total, we find 66 price changes out of 600 observations, but no price change remains the norm in 89% of time across the five weeks' time span (more detailed categorization is summarized in Table 4-5). However, the difference between the retailer types seems rather clear now, in the sense that the mean price change is $0.179 for the OBMCRs and $-0.005 for the DotComs. Note that the OBMCRs have increased their posted selling prices by almost 20 cents on average, after a mere five weeks, while the DotComs' pricing level is basically unchanged. The general hypothesis that competition will drive the online price level lower over time does not seem well supported by our data.

More specifically, there are much fewer cases in price increases by DotComs (6, maximum increase of $2.0) than by OBMCRs (34, maximum of $4.0) and also in price decreases (3 by DotComs, maximum of $-5.0 while 23 by OBMCRs, maximum of $-3.11). Further, most price changes are at least half a dollar in magnitude, 56 cases among the 66 price changes or almost 85%. In fact, even the price changes of at least $0.99 are 81.8% out of all the price changes (54 cases), and almost 35% of the price changes are over $2.00 (23 cases). Note again that both the DotComs and OBMCRs (except Tower Records that has adopted the one-cent-price-change strategy) change their posted prices rather sharply, but the price changes by DotComs (15) are much fewer than the price changes by OBMCRs (99 cases) across the same time span of five weeks. Across these tests, the empirical evidence does not support hypothesis [H.sub.2].

4.3 Price Dispersion

Following Sorensen (1998) and Bryjolfsson and Smith (2000), we use both absolute and relative measures to analyze price dispersion by the DotComs and OBMCRs.

Absolute price dispersion refers to the range of the dollar-, percentage- or full-price across our sampled DotComs and OBMCRs, that is, the highest price minus the lowest price for a particular title at each date for the DotComs or the OBMCRs. The absolute dispersion statistics for our data show a substantial range of prices available by both the DotComs and the OBMCRs for the same video title in the same time period, with the OBMCR case being larger. The range of posted-prices across the DotComs averages $4.37 (see Table 4-6), which corresponds to an average percentage-price range of 26.81%, and is $5.29 or 31.75% across the OBMCRs, almost 5% higher. This pattern is also clearly exhibited when we further examine the title categories of popular versus random. Besides, both for the DotComs and for the OBMCRs, the posted and percentage price ranges are even larger with the popular titles while slightly smaller with the random ones, up and down around 1-1.5%. On the other hand, the full-price range means seem quite close between the DotComs ($4.89) and OBMCRs ($4.96), with both being almost close to five dollars (note again that the average full-price is only around fifteen dollars).

To control for serial correlation, we have run the T-test for each data set day by day and furthermore for both the popular titles and the random titles in each date data. For the posted- and percentage-price ranges, the dominant majority of results are highly significant (p<0.02), clearly in favor of the alternative hypothesis that the price range is significantly smaller by DotComs than by OBMCRs. All exceptions occur in the random titles' category, particularly on July 27 and August 1 with p-values between 0.15 to 0.20 (other p-values are weakly significant below 0.10 except the posted-price case for August 16 at 0.107). However, for the full-price range, the results seem quite different, with none of them being clearly significant in favor of the above alternative hypothesis, although six cases (July 22-August 16) in the random titles' category are weakly significant (p-values slightly lower than 0.10). Further, for the random titles, three cases (July 22 and August 11, 16) are weakly significant (p<0.10) and two (July 27 and August 1) are even highly significant (p at 0.032), in favor of the opposite alternative hypothesis that the price range is significantly larger by DotComs than by OBMCRs. It seems that higher average shipping costs by the DotComs qualitatively change the whole picture in the full-price sense. Similar results are also obtained when we apply the median tests to control for possible distribution bias, at even clearer significance levels.

For case of the standard deviation (calculated according to the standard formula), basically similar patterns as above are also exhibited. Particularly, all test results in the popular titles are clearly significant (below 0.05) for the posted and percentage prices, in favor of the alternative hypothesis that the price standard deviation is significantly smaller by DotComs than by OBMCRs. For the random titles, there are four exceptions in the percentage price tests, in favor of the opposite alternative hypothesis (p<0.03). For the full-price case, all mean DotCom standard deviations are at least as large as the mean OBMCR ones, except for the popular titles on July 22 and 27 (also almost equal). Further, none of the test results for the popular titles is significant in any conventional sense (all p-values being between 0.40 and 0.50), but six out of eight dates are highly significant (p<0.01) for the random titles with the other two being weakly significant (p<0.09). This pattern seems quite robust, from the price range to standard deviation, indicating that the DotComs actually exhibit more pricing divergence than the OBMCRs in the full price sense. This evidence is unique in the online retail market surveys we have conducted so far, contrary to the cases of books, CDs and DVDs.

For the relative dispersion in prices across OBMCRs and DotComs, we compare measures of the price range and the standard deviation by counting the number of titles where a particular measure is larger by the DotComs than by the OBMCRs, for each date. The results are summarized in Table 4-7. Similar to Section 4-1, we can run the binomial test against the alternative hypothesis that it is more likely to find lower price dispersions by the DotComs for each day. For the posted-price range, all results clearly reject the null hypothesis (p<0.01) in favor of the alternative hypothesis. However, none of the tests is significant in any conventional sense for the full-price range measure, and also for the posted-price standard deviation except July 12 (p-value at 0.033). For the full-price standard deviation, all results clearly reject the null hypothesis (p<0.01) in favor of the opposite alternative hypothesis that it is more likely to find lower ones by the OBMCRs. The unique pattern in the online video retail market that the DotComs actually exhibit more pricing divergence than the OBMCRs in the full price sense seems quite robust. The empirical evidence for posted prices supports hypothesis [H.sub.3], but this hypothesis cannot be supported by our empirical results from full prices.

5. CONCLUSION

This study takes a methodology that is unique in the literature. Instead of comparing online prices with prices offered in conventional markets, we investigate how the online branches of conventional retailers compete with their dot.corn counterparts. The purpose of doing so is to test the hypothesis that the online branches' pricing behavior should be part of the conventional retailers' integrated strategies of exercising their market power both in the online and offline markets.

Our findings partly support such a hypothesis. First, we find that the online branches of conventional video retailers tend to sell their products more expensively--about 6.4% in posted prices and 3% in full prices including shipping costs--than their dot.corn rivals. However, unlike the case in posted prices, such differences are not statistically significant in the full price sense. Second, our evidence shows that online price changes are neither as frequent nor as small as expected. Third, the price dispersion is generally lower within the dot.coms than within the multi-channels online in their posted prices, but not in their full prices including shipping costs. This pattern is unique for the online video retail market, contrary to the observations in the online markets of books, CDs and DVDs.

The online pricing behavior of a conventional videotape retailer seems indeed influenced by its market power in the conventional market, since it has to treat the products sold online as close substitutes for the products in its conventional branches. It would therefore be more cautious about cutting prices in online competition. An online-only retailer, on the other hand, does not have such constraints and therefore could be more aggressive in the online price competition.

In the online marketplace, market power and market dominance may become even more phenomenal since the four drivers in the conventional retailing business (namely, profit margin, volumes and strategic assets, brand, and location) not only persist but are also merged into three. In the short run, the emergence of "B2C" (business-to-consumer) online market may lower entry barriers relative to the conventional market. In the longer run, however, as the key drivers and strategic assets gradually clear the competitive landscape, the online market is likely to succumb to the market power of the dominant players. As all the key drivers (optimized volumes, margins, brand, and access) move the online marketplace towards long term equilibrium, conventional retailers' online branches tend to exhibit price change and price dispersion patterns closer and closer to their online-only rivals and vice versa for the successful online-only retailers.

TABLE 3-1. RETAILERS AND PER ITEM SHIPPING COSTS

                  Shipping Rate Number of Items Per Order

                  Per        Per    1      2      3      4      5
OBMCRs            shipment   item

Borders           3.00       0.95   3.95   2.45   1.95   1.70   1.55
Musicland         1.99       1.50   3.49   2.50   2.16   2.00   1.90
Trans World       2.99       1.00   3.99   2.50   2.00   1.75   1.60
National Record   ~          ~      2.15   1.58   1.05   0.94   0.75
Tower             ~          ~      2.95   1.98   1.32   1.24   0.99
Djangos           1.00       0.99   1.99   1.49   1.32   1.24   1.19

DotComs

Amazon            3.49       0.99   4.48   2.74   2.15   1.86   1.69
Bigstar           2.89       1.09   3.98   2.54   2.05   1.81   1.67
Buy.com           3.00       0.95   3.95   2.45   1.95   1.70   1.55
800.com           3.95       0.95   4.90   2.93   2.27   1.94   1.74
TheTop5.com       1.80       0.99   2.79   1.89   1.59   1.44   1.35
Cdworld           ~          ~      4.95   2.60   2.18   1.98   1.85

                  Shipping Rate Number of Items Per Order

                  6      7      8      average
OBMCRs                                 (per item)

Borders           1.45   1.38   1.33   1.97
Musicland         1.83   1.78   1.56   2.15
Trans World       1.50   1.43   1.37   2.02
National Record   0.75   0.64   0.56   1.05
Tower             0.83   0.71   0.62   1.33
Djangos           1.16   1.13   1.12   1.33

DotComs

Amazon            1.57   1.49   1.43   2.18
Bigstar           1.57   1.50   1.45   2.07
Buy.com           1.45   1.38   1.33   1.97
800.com           1.61   1.51   1.44   2.29
TheTop5.com       1.29   1.25   1.22   1.60
Cdworld           1.74   1.66   1.61   2.32

TABLE 3-2. ANALYSIS OF PER ITEM SHIPPING COSTS

DotCom   Mean     2.07
         Median   1.72

OBMCR    Mean     1.64
         Median   1.50

TABLE 4-1. RETAILER TYPE AND MEAN PRICES

Type                    DotCom   OBMCR

Posted-Price Mean       $13.12   $14.02
Percentage-Price Mean   78.13%   83.38%
Full-Price Mean         $15.19   $15.66

TABLE 4-2. RETAILER TYPE AND MINIMUM PRICES

               Min DotCom <   Min DotCom =   Min DotCom >
               Min OBMCR      Min OBMCR      Min OBMCR

Posted-Price   87%            0.0%           13%
Full-Price     49.75%         0.0%           50.25%

TABLE 4-3. MEAN PRICES OF INDIVIDUAL RETAILERS

DotCom         Amazon    Bigstar     Buy.com      800.com     TheTop5

Posted-Price   15.11     14.52       11.41        13.73       11.16
Full-Price     17.29     16.59       13.38        16.02       12.76

OBMCR          Borders   Musicland   Transworld   NatRecord   Tower

Posted-Price   14.03     13.95       13.20        16.85       11.84
Full-Price     16.00     16.10       15.22        17.90       13.17

DotCom         CDworld

Posted-Price   12.77
Full-Price     15.09

OBMCR          Djangos

Posted-Price   14.26
Full-Price     15.59

TABLE 4-4. RETAILER TYPE AND FIVE-DAY PRICE CHANGES

Price increase   Total   =0.01    [greater than      [greater than
                                  or eqaul to] 0.5   or equal to] 0.99

OBMCRs           55      2        53                 52
DotComs          9       0        9                  5

Price decrease   Total   =-0.01   [less than or      [less than or
                                  equal to] -0.5     equal to] -0.99

OBMCRs           44      12       32                 31
DotComs          6       0        6                  4

Price increase   greater than
                 or equal to] 2.0

OBMCRs           25
DotComs          2

Price decrease   [less than or
                 equal to] -2.0

OBMCRs           9
DotComs          3

TABLE 4-5. RETAILER TYPE AND FIVE-WEEK PRICE CHANGES

Price increase   Total   =0.01    [greater than      [greater than
                                  or equal to] 0.5   or equal to] 0.99

OBMCRs           34      0        34                 34
DotComs          6       0        6                  4

Price decrease   Total   =-0.01   [less than or      [less than or
                                  equal to] -0.5     equal to] -0.99

OBMCRs           23      10       13                 13
DotComs          3       0        3                  3

Price increase   [greater than
                 or equal to] 2.0

OBMCRs           18
DotComs          1

Price decrease   [less than or
                 equal to] -2.0

OBMCRs           2
DotComs          2

TABLE 4-6. ABSOLUTE PRICE DISPERSION

1.1.1.1   1.1.1.3          1.1.1.4          1.1.1.5
          Posted           Full             Percent
1.1.1.2   Price            Price            age
e                                           Price

a         DotCom   OBMCR   DotCom   OBMCR   DotCom   OBMCR
n

Range     4.37     5.29    4.89     4.96    26.81%   31.75%

STD       1.77     1.88    1.94     11.8    10.94%   11.21%

TABLE 4-7. PROPORTION OF PRICE DISPERSION

DotCom < OBMCR   Range    Standard Deviation

Posted-Price     78%      52.25%
Full-Price       53.75%   28%

REFERENCES

Bailey, J. P., Intermediation and Electronic Markets: Aggregation and Pricing in Internet Commerce. Ph.D. thesis, Technology, Management and Policy, MIT, 1998a.

Bailey, J. P., "Electronic Commerce: Prices and Consumer Issues for Three Products: Books, Compact Discs, and Software", Organisation for Economic Co-Operation and Development, OCDE/GD (98) 4. 1998b.

Bakos, Y., "Reducing Buyer Search Costs: Implications for Electronic Marketplaces", Management Science Vol. 43 (December), 1997, 1676-92.

Bakos, Y. "The Emerging Role of Electronic Marketplaces on the Internet", Communications of the ACM, Vol. 8, 1998, 35-42.

Brynjolfsson, E., and Smith, M.D., "Frictionless commerce? A Comparison of Internet and Conventional Retailers", Management Science, Vol. 46 (April), 2000, 563-85.

Clay, K.; Krishnan, R.; and Wolff, E., "Prices and Price Disperson on the Web: Evidence from the Online Book Industry", Journal of Industrial Economics, Vol. 49(4), 2001, 521-540.

Clay, K.; Krishnan, R.; Wolff, E.; and Fernandes, D., "Retail Strategies on the Web: Price and Non-Price Competition in the Online Book Industry", Journal of Industrial Economics, Vol. 50(3), 2002, 351-367.

Clemons, E. K.; Hann, I.-H.; and Hitt, L. M., "The Nature of Competition in Electronic Markets: An Empirical Investigation of Online Travel Agent Offerings", Working Paper, 1998. The Wharton School of the University of Pennsylvania.

Darnay, A.J., and Piwowarski, J. (eds.). Wholesale and Retail Trade USA: Industry Analyses, Statistics, and Leading Companies. Detroit: Gale. Second edition, 1999.

Morton, F.S; Zettelmeyer, F; and Risso, J.S., "Internet Car Retailing", Mimeo, 2000, Yale University, New Haven, CT.

Nelson, P., "Information and Consumer Behavior", Journal of Political Economy, Vol. 78(2), 1970, 311-329.

Sheskin, D. J., Handbook of Parametric and Nonparametric Statistical Procedures. New York: CRC Press, 1996.

Smith, M. D.; Bailey, J. P.; and Brynjolfsson, E., "Understanding Digital Markets: Review and Assessment", in Erik Brynjolfsson and Brian Kahin, eds. Understanding the Digital Economy, Boston: MIT Press, 2000.

Sorensen, A.T., "Equilibrium Price Dispersion in Retail Market for Prescription Drugs", Journal of Political Economy, Vol. 108(4), 2000, 833-850.

Stigler, G., "The Economics of Information", Journal of Political Economy, Vol. 69(3), 1961, 213-225.

Tang, F.-F., and Lu, D., "Pricing Patterns in the Online CD Market: An Empirical Study", Electronic Markets, Vol. 11(3), 2000, 171-185.

Tang, F.-F. and Xing, X., "Will the Growth of Multi-Channel Retailing Diminish the Pricing Efficiency of the Web?", Journal of Retailing, Vol. 77, 2001, 319-333.

Fang-Fang Tang, Chinese University of Hong Kong

Xiaolin Xing, National University of Singapore

Author Profiles

Dr. Fang-Fang Tang received his PhD at Bonn University (Germany) in 1996. Currently he is an associate professor of marketing at Chinese University of Hong Kong.

Dr. Xiaolin Xing received his PhD at University of Pittsburgh in 1992. Currently he is an associate professor of economics at National University of Singapore.

The work described in this paper was supported in part by a grant from the Research Grants Council of the Hong Kong Special Administrative Region, China (Project No. CUHK4310/02H) and, in part by NUS Research Grant No. R-122-000-050-112.