Lecture 10: Search, Part 2

Key Takeaways

okay uh welcome back so if you remember uh last class I was talking about uh search Theory uh and discussed you know two main things you know one is that uh search cost can uh complement uh differentiation and raise prices in markets uh and the other is that in in Bertrand like markets search cost can also lead to price dispersion uh you know in most of this class we've had um pattern of two lectures you know one where I do the theory and then one where I talk about the empirics uh on search it's a big topic we decided to do three so I did um uh I did the theory last time today I'm going to talk about uh reduced form evidence on empirical evidence on search and then we're going to have a third lecture on search where tobas is going to talk about the more structural work and uh that will come later um so anyway uh empirical work on search um you know why do again why do we do empirical work in IO you know often generally trying to do two different things or maybe a combination of two things one is we're trying to get evidence on the theories whether they're applicable whether they're you know what might be correct or incorrect or all theories are correct but you know like what is about the theories we might want to modify to have a fit application and the other thing we're trying to do is just take the theories and apply them to sort of particular empirical examples that we care about and estimate parameters of Interest know how big or small effects are um so I'm going to talk about four papers today the first of these is of the you know just gaining insights in the theory um literature generally when we're trying to gain insights in the theory we're often looking at unusual uh situations because we want to look at markets that are particularly simple where the theory is particularly Stark try to understand this uh its predictions whether they apply in particular you know Allen in his paper is talking about um trying to understand you know do we see the kinds of price dispersion predicted by stylik models of uh of of heterogeneous consumer search and you know is it true that sort of like you know VI frequent violations we see of the quote law of One Price are because of consumer search costs leading to heterogeneity uh and so in this paper you know what Allan was trying to do is think an application where he could argue that the brand assumptions fit very very well and examine in a model where betran assumptions fit very very well do we see price dispersion and can we provide evidence that that price dispersion seems to be due to consumer search costs um you'll see that you know this is a paper written when Allen was a PhD student at MIT very much it's a paper that has sort of minimal data self-collected that anyone could write um so his idea was that we could get evid a good example of think about uh ber Tran like competition is prescription drugs one of the nice things about prescription drugs is that you know a bottle of amoxicillin tablets bought from Pharmacy a and a bottle of amoxicillin tablets bought from Pharmacy B should be completely identical products people should not people should know that these things are exactly the same product that we just buying it from different stores so we could think of people as having betrand like preferences one of the other nice things about prescription drugs is you're always struggling in empirical papers how do I get the number of degrees of freedom to actually estimate something great thing about prescription drugs there are hundreds of different prescription drugs the hundreds of different prescription drugs give them hundreds of different um observations um so what Allan did for this paper is um if you think about the you know Northeastern United States so we're in Massachusetts which looks something like this um we've got Connecticut here Rhode Island New York State uh Allan noticed that um New York uh in the late 1990s had a law that pharmacies had to post all of their uh retail prices for the 200 most popular prescription drugs on a poster that was in displayed in large format in the stores therefore what one could do is go to a pharmacy in New York and very quickly find out what prices they were charging for uh every different drug uh in the 1990s people did not have phones in their pockets with high resolution cameras um so you could not go there and just take a picture of the price photo so what Allan did is walk go to New York with a clipboard and just copy down off these price posters what prices the pharmacies were charging for all the drugs and in particular he does this in two small towns uh Middletown and Newberg New York and these happen to located where if you drive from Boston uh along the Mass Pike and then get on 84 and then go to New York these are sort of two of the first towns you come to uh in New York state one thing that was nice about them is he wanted to get all the prices in a market um Middletown and Newberg are both sort of small towns they have 10 or 12 pharmacies in them and then there are no other pharmacies within like you know 15 minute drive or 20 minute drive away and so you could think of anyone who lives in middl toown as shopping having the options be buying prescription drugs all the pharmacies in Middletown anyone Newberg they can buy the things in Newberg okay uh and I think I've said all these things oh I guess another thing about how the world was different um uh another thing about how the world was different in late 1990s is in the late 1990s most people did not have prescription drug insurance and so at the time the prices that thees charged for prescription drug was what people customers would be paying customers would have the sort of same incentive to shop for drugs they would have for any other product obviously today many more people have prescription drug Insurance the prices of pharmacies would not be what they're paying but at the time that this thing had this Advantage um so anyway here's an example of the study design so you can see that this is not the actual poster but there was a poster that was very much like this I think it was roughly 150 drugs but then many of them had generic as well as branded versions and so you were getting rough over 200 prescription drugs uh and so there this big poster it would list acail 10 milligram tablets acpr 20 milligram tablets and so on and then it would have the prices there and so Allan could just go into thees copy down the prices listed on this poster and and get an example of what they cost um and you know different drugs have different characteristics and so what he's going to be able to do is not only say how much dispersion is there within a drug but then how much dispersion is there across drugs uh sorry how does the dispersion vary across drugs with drug characteristics um one thing that's also nice about this application is you could say one reason why drug prices should be different acrosses is because uh their drugs are bundled with Pharmacy amenities some Pharmacy has better customer service is cleaner night e quicker to get in and out of people would pay extra to go to that Pharmacy uh but the idea is having 200 different two 200 different drugs you can sort of put in Pharmacy fixed effects and say you know the pharmacy may be nicer but there's not something like a this Pharmacy is a nicer place to get amoxicillin and that Pharmacy is a nicer place to get sexin you know Pharmacy fix effects Auto account for most the amenities and let them control for that okay you know not much sophistication of the analysis so I won't go through how it's really done but let me just sort of talk about some of the basic facts so first basic fact there's substantial price dispersion uh the difference between the highest and lowest price for drug across the drug stores town is $13 for the average drug um some drugs have less price dispersion 10th percentile range is $5 uh 95th percentile range is $25 so substantial price dispersion you know obviously it's not enormous but $13 is about what what you can think of um second finding is that the price dispersion doesn't appear to be due to Pharmacy fixed effects uh he argues this in a couple ways first is just this simple table uh where he puts for every pharmace in these towns how often is their price in the bottom third of the distribution here in Middletown they're 10 so um how often is their price one of the three lowest how often is it in the middle how often is in the three highest and for many of the pharmacies like the ones I've highlighted eer immediate and Kmart there's a substantial number of drugs where they're the low one of the lowest prices a substantial number of drugs where they're one of the highest price pharmacies there are a couple exceptions Right Aid seems to be a terrible place to buy drugs if you live in Middletown it's almost always one of the highest three prices uh also almost one of the highest prices in in Newberg um Price Chopper and Walmart seem to have lower prices than the others but um if you think about how important are the pharmacy fixed effects he has a simple table where you sort of uh regress the prices of uh drugs in a pharmacy on drug fixed effects that has an R square of 0.907 you include drug and Pharmacy effects as an R square of 938 so what that's saying is you know the pharmacy fixed effects or the amenities are accounting for about onethird of the price dispersion which leaves you know two-thirds of the remaining price dispersion so 2third of the $13 appears to be pure idiosyncratic uh variation across pharmacies not correlated with Pharmacy fixed effects and then Allan also argues in the paper that even that may be an under estimate of how much of his Pharmacy fixed effects because he looks at the expense of pharmacies like goes to the right AIDS uh and goes to the Price Chopper and his his his informal evidence of how friendly is the pharmacist how clean is the place how long is the line he would rather go to the Price Chopper than go to the Right Aid so in some sense this sort of high prices at Right Aid may just be part also of the idiosyncratic price dispersion not part of the farm not part of the sort of bundling with amenities okay um main thing that Allan does to then argue that this price dispersion is related to search costs is to sort of you you want to think about well let's go back to the theory if you remember what did I teach about price dispersion if you have a market and you have uh you know if we have very low search cost and this is price equals cost um what you would see is with very low search costs we get a distribution that's fairly tightly distributed around we get a distribution of prices that's fairly tightly distributed around costs because if anyone sees a high price they're going to be tempted to go back and search again that gives us a search a cost distribution that's low prices are close to cost and then not very dispersed because if prices were very dispersed anyone seeing a high price would just go search again and try to get a low price so we get very little price dispersion when we have high search costs um what you get instead is uh more dispersed distribution that may look something like this where a few firms set low prices many more firms set high prices even the consumers who see the high prices don't want to search again so we get two different things we get a higher average price and we get more dispersion whereas here we get a lower average price and less dispersion um so that's what Allen would like to say is can we observe drugs that have lower search costs versus higher search costs and find out do the ones that are have lower search costs have lower and less dispersed prices okay um it's hard to think you know well it's you can think of reasons why some drugs may have higher or lower search costs so for instance drugs that treat conditions that are acute and very painful or make it difficult for you to shop could have high search costs you know drugs that treat sexually transmitted diseases might have high search costs because you're embarrassed about the condition you don't want to go tell every Pharmacy in the town call them up say know what would your cost be for buying this prescription um but what Allan comes up with is this idea of frequently purchased drugs so you have some drugs like antibiotics that you're typically buying and you're going to use once you have other drugs like ant like high blood pressure medications high blood pressure medications almost everyone who's taking them is going to take them every buy them every month for the rest of their lives and so if you think about shopping for an antibiotic versus shopping for high blood pressure medication it pays to shop around for the high blood pressure medication because you're going to be able to amortize those search costs over the rest of your life whereas the antibiotics you're just going to buy this thing once you're never going to buy it again and the benefit of search is smaller um so what Allan does is think about frequently purchased drugs as drugs that have effectively lower search costs uh and examine whether the range in prices and the level of prices are lower for drugs that are purchased more frequently um the range is what's done here so he has four different models uh four different measures of the range of prices the absolute the range standard deviation residual range after you're controlling for some Pharmacy fixed effects and other things residual standard deviation in all four cases he shows the more frequently purchased drugs do have less price dispersion than the frequently purchased drugs okay he also shows do I have a table for this I do not uh also shows um that the average markups for these drugs relative to sort of published wholesale prices are lower for the frequently purchased drugs and then a third facts I said we expect for idiosyncratic reasons some drugs to have lower or higher search costs apart from this sort of purchase frequency and what he shows is that drugs that have unexpectedly High average markups also have unexpectedly High dispersion suggesting that there's also unobserved variation in search costs and that unobserved variation in search cost is driving some of the price dispersion in price levels anyway so I you know I I think it's a you it's a very nice paper it's sort of providing very simple evidence saying that these theories we have like stals model of price dispersion do appear to be applicable and do appear that like the price dispersion they predict is there and it appears that that price dispersion is search cost related Stango and Zinman are sort of on the opposite side of the sort of you know second primary second motivation trying to look at a question that we really care about and understand you know does price disp do sort of models of price dispers help us understand what's going on and how big are the magnitudes of price dispersion in this application uh particular thing they're talking about is credit card debt um you know credit card debt is something that uh you know many people feel is unfortunate there are many people in the United States who pay much more on credit card interest and have gotten themselves into substantial trouble um there was sort of a set of papers in the 1990s set off by this this AER paper by Larry oabel raising what what he called the credit card puzzle and the question is why is it the credit card interest rates are so high given that there are a million banks in the United States why is it that this competition between this huge number of banks does not uh dissipate profits in the credit card industry and how do we get profits that are so high for the firms the other thing that Asel raised was sort of actually uh non-dispersion intuition which is that also reported that not only are credit card rates very high but every big Bank charges very high rates on their credit card so there's not price dispersion he would have thought you know he sort of say why is it that all these banks are charging very high interest rates why doesn't one of them undercut the others and try to compete with them bring interest rates down uh ASA also reported that rates were fairly un insensitive to interest rates so his data covers times when you know there are times when interest rates uh on mortgage loans are 10% times when those rates are 2% when interest rates on on Treasury bills and mortgages are changing credit card rates seem to be fairly insensitive and that was another puzzle raised is you know why is it these rates are not why is it they're so fixed and there's so little dispersion they don't vary over time or across Banks okay um so Stango and Zinman uh are actually going to argue that that uh basic facts that aabel had us think about were wrong um that it's not the case it what's it is true that rates across Banks tend to be fairly similar uh but what's not true is that rates across individuals are true and so in particular there's this hidden thing that osel was missing which was that all of these banks have high average interest rates but all these banks are also offering many different interest rates to their different clients some of their clients are paying low interest rates some of their clients are paying high interest rates and so uh what sanguin Zim trying to think about is you know you know document that there's substantial price dispersion and think about search cost as a potential explanation for why it is that we see these high interest rates um so anyway some basic facts in their paper and I should say where does their data come from so you know this is not the sort of amateur data collection we see in Sorenson's paper um this is the data collection where you make connections with some firm uh that has access to very high quality data so in particular there's a company called light spe and there are a few companies like this what light speeded does is it has a panel of consumer a panel of consumers who basically open their lives to light speed in exchange for some type of payment so in particular uh there 4,312 uh customers in their data what these customers do is they give light speed access to their credit card statements so every month Lightspeed gets their complete credit card statement everything they bought um how much what they paid what how much interest they paid did they pay a late fee is B basically it's like they've got the credit card statement for 4,312 people um they also have information on the people in their panel so when people join their panel they complete a survey that gives them a lot of demographic information they also know the consumer's credit scores they've got this from I don't know Experian or some other credit reporting agency uh so they have some measure of the um some measure of the uh creditworthiness that people do think would be a driver of the heterogene and interest rates being paid um so let me start from some basic facts um first fact is there's tremendous heterogeneity in interest rates across um consumers if you look at the inter quartile range Just Between the 75th percentile and the 25th percentile uh it's 800 basis points or you know eight percentage points per year um this is omitting anyone who has teaser rates anyone who pays in full so these are people who are paying interest their credit cards every month there's just tremendous heterogenity there um of course there's going to be heterogen and credit card rates because some people are bad credit risks some people are good credit risks you would think if you were running a credit card company you would give lower interest rates people were going to pay you back a higher interest rate than people who you think might default um what they find is that default risk explains about 40% of the interest rate uh variation um other factors like offsetting rewards demographic Graphics explain very very little so roughly you know 60% of this Gap is due to um what appears to be pure idiosyncratic heterogenity people who are similar situated equally likely to pay the bank back some are paying much more than others uh and a third fact I'll talk about later is that there's also substantial within consumer variation offers receive they have a separate second data set where they have people open like they know all of the credit card mailers that are going to particular individuals tools uh people in their data I don't know if this applies to you get multiple credit card offers per month they look within one month what credit card offers people get and there's substantial variation in the offers if you open them up okay so let me just sort of uh some summary statistics you know this is a case where you know I think part of the interest in this paper is it just this light speed data gives us this view of what is it that consumers are doing and and has sort shocking statistics about problems people have so um they divide the data into the quartiles of the average balance that people have on their credit card you know you were probably told that it's sort of good practice to pay off your credit card bills every month and simply not pay any interest uh and what they find is that the 25th percentile balance is $499 which means you know the fraction of people in their data who pay off their bill every month is less than 25% so it's close to 25% but most people have a balance on their credit cards and keep run a balance on their credit cards every month and never pay it off okay uh what's the upper quartile the upper quartile is people with credit card balances between 4586 and $62,000 uh people in this quarti on average have a res revolving balance of $111,000 in their credit cards they're paying $2,000 in interest and obviously you know these are the people that people really worry about about in people who are making big Financial mistakes and there there something we can do to sort of help people avoid getting stuck in the situation where you're spending uh $2,000 um a month on interest you may also notice that the people who were spending $2,000 a month on interest what's their income you know a quarter of them have incomes below $45,000 so being income below $445,000 paying 2,000 a month on your credit card these are people who've made very big mistakes you know some of them have incomes of above a 100,000 uh but it's still substantial um also sort of noteworthy is some sense the some of the u-shaped relationships here it's actually people in the bottom quartile and people in the top quartile who have the best credit scores and who are charging the most every month some people in the second quartile are you know people in second quartile are spending less per month have lower credit reports lower credit scores these people are probably constrained in how much they can borrow but it's the people at the top actually have fairly good credit scores and the you know credit card companies enjoy being able to sort of charge them $2,000 a month in interest um okay so this fact I told you about there's tremendous dispersion in credit card interest rates um you know here's the basic uh raw information on on what people are paying and you can see it's not based on your revolving balance so people in the lowest quartile there people paying between 12 they're people paying 12% and people paying 26% so this is the 10th percentile this is the 90th percentile people in the upper quartile people paying 11% people paying 26% it seems to be very similar Acro across those four groups in in at for every sort of level of revolving balance there are people paying very low interest rates uh people paying very very high interest rates um you know as I said you know second fact I have is are people paying very different interest rates because they have very different credit scores uh what we see here is the blue dashed line is the distribution of interest rates people are pay being paid relative to the average rate so zero is here so they're a number of people paying interest rates 5% below average there a number of people paying 10% below average also many people paying 0 to 10% above average not many people pay more than 10 or 12% above average but the blue is a distribution of interest rates people are being pay are paying relative to the average the red is what you get if you say okay now I'm going to control for your credit score and ask once I've controlled for your credit score what are you paying relative the average for someone with your credit score and other attributes and what they find is that you know this explains some of the variation the red distribution is uh more tightly uh distributed than the blue distribution but you know that's the it's explaining perhaps 40% of the variation so this some of this Tales go away but we still have very substantial differences in interest rates paid by people who appear to be uh similar on all attributes so then the next thing the paper does is try to sort of say can we sort of uh can we tie this to search intensity and we find that people who are sear or with lower search cost people who are more likely to search are the ones paying uh interest rates down here and the people who are less have higher search costs end up getting stuck and paying high interest rates up there okay um you know it's not as clean as Allen's paper where Allen has this sort of exogenous characteristic of a drug that shifts at search costs um but what they have as a measure of search costs is just a self answer to a a survey question uh they ask consumers um you know so they have not only have access to the panel data they had the ability to ask questions of consumers in the light speed panel data and they asked those consumers How likely are you to look at a credit card offer you get in the mail and you might think that that would be a measure of search cost people who have low search cost people who like opening credit card envelopes and look thinking about them are people with low search cost people who dislike opening their credit card credit card offers to get in the mail or people with high search costs uh is that going to explain things when you think about that there's this obvious indogen concern you ask people How likely are you to open a credit card uh offer you get in the mail well if you know you have a bad credit card and you're paying a high interest rate you may be more likely to say yes I'm going to open up things because you're looking for a new credit card whereas if you already have a really good credit card you may say no I don't open them up and you're not opening them up not because you have high search costs but because you know you already have a very good deal um so what they do is you know the um first thing they do is they just run this OS regression and they regress the APR that you end up paying on your self-reported search uh intensity and we get a small negative coefficient saying people who search more actively are paying slightly lower rates um but then they sort of it has that endogenity so they run this IV regression um the IV regression says that we can think about gender and marital status as instruments for search intensity because um the offers that you get in the mail should not differ across gender across marital status once you control for uh credit score and other factors because it's illegal to discriminate uh against women or against married or unmarried people um therefore those things should relate to your search intensity but be unrelated to the set of raw set of offers that you get uh in the mail for credit cards when they put that IV in now they get a very large negative coefficient saying people who search more intensively um are paying lower interest rates I mean it's it's a you know it's a very large gap between the OS and the IV estimates and then the sort of standard error really blows up relative to this so you sort of Wonder you know you again it it sort of it's it's only a slightly significant result why is that you know is is the reverse causality really so large that it causes this but anyway that that's sort of the evidence they provide that it does seem to be the dispersion that people are paying is related to some characteristic that that does their search intensity so and you also see at the end of this lecture I'm going to come back to sort of another paper that sort of further examin this question but I I think it does you know clearly reframe the debate relative to what you had thought from oabel and other early papers to say that you know there is just substantial price dispersion in credit cards Search cost seem be a part of why it is that it's so hard for firms to sort of compete in this market is if you sort of send out lots of credit card offers and people don't open them people don't consider buying them uh that may make it sort of one of the big factors that's obstacles to sort of eliminating the sort of high markups that we see okay so moving on I was G to spend you know much longer talking about a paper of mine um you know talk about in part because I know it better I you know it's one of my favorite of my papers um papers mine with Sarah um you know this is a paper where we're you know we're both trying to we are trying to uh it's more talking about the theory paper where we are trying to understand uh what causes uh price dispersions and markups um I I see this we saw it at the time as sort of a paper that was looking forward um you know the question was that sort of in the early 2000s search engines price search engines had recently been invented um there were few markets like airplane flights hotels rental cars where many people use price search engines most other markets people don't still use price search engines uh but you know our question was what's going to happen in the world if price search engines become much better you know first of all like you know how are the like the airplane flights and the hotel and Rental Car Market's going to adjust when flights get better when search Technologies get better and what's going to happen to the rest of the economy when search Technologies get better I mean I think somehow this still seems topical 20 years later you know 20 years later when we see these amazing AI advances and chat robots and whatever you Wonder am I going be able to sort of are we going to get to a point where I can just type into Google what's the best place for me to buy a Mox ayin and have a Mox and have Google say you're living I I can see that you live in Newton Massachusetts for you to buy a MOX ayin the best pharmacy to go to is this one in this one that's like on your drive home and it costs this much there and this much somewhere else are we going to get some point where I can sort of say I want to buy a tennis racket and then Google just tells me you know you know given what I know about you this would be the best tennis racket for you to to buy and this store has it at this price this is where you should buy it and just get it shipped directly to your house um so you know if who knows whether we will ever get to that point but know question I think Still Remains how do we expect markets to change when search cost when internet Technologies improve search costs go down are we going to see sort of a collapse in price dispersion where formerly High dispersed prices become very low tyght prices and you know we mentioned the paper this potential berran Paradox you know real retailers have substantial fixed costs you know the fixed cost of running a store is something like 20% you know so if you if you can't earn 20% markups over your wholesale cost on Goods you're going to go out of business Even If you're sort of Walmart you're super efficient you need to have 20% markups of your wholesale cost to sort of cover all the fixed costs of having a physical store and things like that and so the question is you know how is retail and other businesses going to survive when the internet makes price search more efficient and I guess you know one of the answers we're going to say in this paper is that you know models of search traditionally thought of consumer search as a one-sided problem where these consumers have these search costs the search costs are fixed uh and and then they search and that fixed search cost ter the markups thing we're going to think about is actually you know sort of two-sided you have sort of the search engine is trying to make search more efficient the consumers are trying to make their search more efficient and at the same time their sort of uh stores are on the other side side trying to make search less efficient and so we can think of search cost as not being sort of exogenous or not being just a result of Technology but being a outcome of a two equilibrium of a two-sided game where some people are trying to make search more efficient other people are trying to make Search Le less efficient and the sort of balance between search and what we call obviation ends up determining price levels and markets okay so as I said you know this is a paper where we're focused on this theory of how would obfuscation work we go to a particular narrow corner of the internet where everything is much simpler and it makes it much simpler to do the analysis than it would if we sort of look at a more popular product and so particular we look at um a price search engine called pricewatch.com uh I think it still exists although it's it's long past its Heyday but in the early 2000s pricewatch was this simple database based search engine that people who were sort of sophisticated would use to find things that they needed so for instance you know if you were to go to the Computer Guys in in the MIT economics department and say I have a computer I'm getting sort of low memory errors could I just buy some extra memory and plug it into the you know plug some extra memory modules into this part of my circuit board and increase the memory capacity of my computer um people who worked in it offices might use pricewatch.com to find an inexpensive place to buy memory that you could add to your computer computer what did it look like and again this is you know think 2000 the world was much more Primitive Place the people had lower bandwidth on their on their home internet um things just had Simple Text based design so this was price watch very proud that it was established in 1995 and you can see the pixelated Graphics there um but anyway you'd go to pricewatch.com and it would sort of say these are the things you can buy through our site you can buy computers PCS with or without an operating system system you can buy CPUs you can buy motherboards um we're going to be buying memory um so RAM memory versus flashcards we're going be buying Ram so anyway I go to this page I click on Ram uh when I click on RAM it brings up this page which shows me all the different kinds of ram I can buy um again you need to be somewhat sophisticated I have do you need DDR memory do you need notebook ddr2 memory these are all the memory modules you can buy they're going to have various descriptions so the five 52 megabytes 4 gbt 2 gbes you know that's the sort of storage capacity of the memory module this uh 4, 3500 3200 those are sort of uh describing the speed with which the memory com uh uh communicates with the mo motherboard the DDR ddr2 those are other aspects of the thing you know basically you need to know what type of memory you you need for your computer like my computer needs DDR pc3200 memory it is a choice then do I buy uh 4 gbt 2 gab 1 gbyte based on how much I want to spend how much memory I want to add but it's the site is designed for people who are sophisticated in knowing you know what memory do I need for my motherboard um this is now getting to actual data so at you know this was taken a bit later when you can see people buying 4 megabyte and 2 megabyte or gigabyte modules uh back in 1999 with people would buy is either 128 megabytes of memory or 256 megabytes of memory so I clicked on one particular type of memory 128 megabyte modules that follow the PC 100 would be the sort of uh you know 100 would be the speed of the which it communicates with your computer so you know you need pc1 100 memory you want by 128 megabytes you click on this link and it brings up this page this page just has a set of offers for where you could buy the memory from I can buy it from computer craft Inc from connect computers from First Choice memory this retailer is located in Florida this one's in California and so on and then the price of which they'll sell me the memory 68 69 70 72 74 74 74 75 and then how much they'll charge me for shipping you know it's usually uh the time price watch had a cap you couldn't charge more than $111 and so many of the firms are some number that's sort of slightly less than $11 um again it was a primitive period where if I wanted to buy the memory I could then click on computercraft inc's website uh it would take me not to it's not like this would sort of add the memory to my cart I would click on that button I would typically get to the computercraft site and then I'd have to search around the computer craft site to find where is their 128 megabyte pc100 memory can I find it at that price of $68 that price watch is telling me that they offer at um price watch was not scraping these numbers the retailers were having to enter these numbers into a database that price watch so the if you're one of these retailers you would get up in the morning decide what your price is upload the new price the price watch change it on your website and that's what you would charge um so basic observations on price watch is you know there was a tremendous amount of competition on price watch um prices were very low you know at this time when this memory module cost $68 if I gone to the Best Buy to buy the same thing I might have paid $100 so markups were very low on price watch um but there was some dispersion and it was clearly not at all the frictionless ideal of I click there I just everyone's in bertran competition I get exactly what I want uh two experiences were common when you used it first as I said you didn't go directly to the landing page even when you did go directly to the page where you clicked on you would find these sort of page that would be annoying and timec consuming to buy what you wanted to buy so this was an example from a bit later where I clicked on a a offer from tough shop.com to buy some me memory module That was supposed to cost $538 I clicked on a link saying buy this module for $53 81 and they give me this Mo page that tells me and there's actually more than this page this page saying price with all the selected options is $90.30 and so you know like why is it that it's $90. 36 well they've done is they've had this page it's got all these other things that you can click on to add to the cost of your module and some of them are just pure annoyances like they've somehow they pre-checked for me bonus buy 10 pack of hand thumb screws for my case so you know if I had hand thumb screws to open up the the computer case in which I wanted to open to change my memory why I would need new thumb screws I can't imagine but for 495 I can get those 10 thumb screws I just have to go through read that and say no I don't want that uncheck that box and it takes 4.95 off the price um others of these are things that it's less clear that I should uncheck so this is um you know take advantage of these special offers memory upgrade cast 2.5 upgrade improve performance and helps things they're charging me 635 from that um this is another example six layer stability more layer better design uh not in perfect English 837 but then they also have these others that are sort of more concerning uh preest standard preest avoid costly uh return agreements 697 so they're telling me is that if I want them to test the product they're shipping it to me uh before they ship it to me and make sure it works I have to pay $7 to get them to test the product they're going to sell me and make sure that it works and if I don't ask them to test it for me and make sure that it works they're going to charge me to ship it back to them uh and so you know it may be that you look at this offer and realize this offer is not what I thought this offer was I should think of this offer as being $7 higher than this offer was um and you know in some sense I'm going to have to just spend a long time reading the fine print of tough shops offers where I trust do I want to buy this product do I not want to buy this and it's just going to take me a while to shop okay um second thing that was common was you know offers that were not designed to just delay and annoy you but ones what was clearly aimed to get you to shift away from buying something you bought into something else so if you remember um you know remember back to sort of when I talked about competitive price dispersion I talked about add-on pricing that is if you know if you have two firms selling two products models of price dispersion work totally well with sorry models of of price discrimination work totally well with differentiated products in multiple firms and you can have two firms competing against each other they advertise all their prices they sell low quality Goods at low prices high quality Goods at high prices and that can be rational if um the consumers for the low quality goods are more price sensitive across firms and so you have an equilibrium where you it's like the competition on two line model for lowquality goods firms are very undifferentiated for the high quality goods are differentiated these goods are sold at a low markup these goods are sold at a high markup um and one thing that I said when I when I talk about that is one way in which firms in this model can raise prices is you can think of another game where you SE advertise your lowquality product and then you have an unadvertised price for a high higher quality product uh in that game where you only show people the price for a high quality product once they show up at the store uh I gave this this model that said markups are going to be even higher in equilibrium and the reason is that firms are going to sort of charge the EXP poost Monopoly price for the upgrades uh that caus a you know that sort of caus a wedge between when you sort of force the prices of the low and high quality Goods further apart um there's this adverse selection effect where you don't want to get sort of you know you don't want to sell the cheap skates who are going to sort of buy your rental car for 1995 a day and not get the insurance and not get the car seat and because there's this sort of adverse selection effect where it's the cheap skates who don't buy the insurance and the rental and the car seats um you then want to dump the you know you you go from sort of wanting to undercut the Rival to get consumers to wanting to overcut the Rival to dump the cheap skates on them markups for both products can end up go markups for the sort of average markups go up in equilibrium um we see what looks like sort of an add-on strategy here so you've clicked on this product I want to buy I clicked on I want to buy OEM 512 megabyte memory they actually put the check mark in the correct box this is what I said I wanted to buy but then they tell me that I can spend $15 more get this product or I can spend $25 more and get this product which I'm going to think of as add-ons um what do they do they tell me a lot about why the better products are better um so for instance this one is cast 3 latency this is cast 2.5 latency this is cast 2.5 latency this one has a four layer board these each have six layer boards um these have uh OEM D Ram downgrade trips chips whereas these have industry standard chips whereas these have even better brand name chips Samsung Micron or Major Brands um you know these have a restocking fee these have no restocking fee satisfaction and compatibility fully guaranteed um Fact one thing that you know for fun I sort of clicked on some of these things you know this one has verified compatibility with your memory configurator I I clicked on this and entered a zillion different computers every consumer said this seem to be button that just says can't verify compatibility so no matter what you typed in in some sense it would just sort of say always is not compatible so clearly you know this page is designed to get us to buy this instead of buy that and just telling you all the reasons you know these words like OEM downgrade chips I don't know what that means but I know it's bad it's not trying to teach us about the products it's not trying to tell me why is cast 2.5 latency better than cast 3 latency or why is a six layer board better than a four layer board should you care about that but it does seem like it's trying to convince us to buy this product and pay $5 extra instead of buying that product so you know we're going to argue in this paper is that you know if you really had berran competition for memory modules and they were being sold at cost the retailers couldn't cover their fixed cost Market wouldn't exist the market does exist why is that well our thought is that firms are engaged have you know that while price watch has sort of invested in technology to reduce search costs The Firm are sort of fighting back with obus to try to raise search costs and use those higher search costs to keep the uh increased markups we can think of two different ways you could do that you know one is simplest theory of obfuscation is you know we know that you get higher markups and more dispersion when s s is higher than when s is lower so maybe what firms are doing is just making s higher you if I make you spend five minutes clicking reading through unclicking the thumb screws unclicking other things I spent 5 minutes clicking through doing all the things where I finally found out the price I'm like do I really want to go back and go through another five minutes with the next website and find out what their real price is so by just making shopping more timec consuming I've raised the S raising the S is going to raise the equilibrium price distribution and then the second thought would be this could be what I mentioned before this add-on pricing situation where we know that sort of equilibrium prices add-on pricing are higher than equilibrium prices when all good prices are advertised maybe what firms are trying to do is shift the equil shift the game change the game from a game in which firms advertise all prices to a games in which firms advertise um base goods and then invent add-ons and you know that model the larger is the add-on as a fraction of the price the higher are the equilibrium markups going to be so firms are in some sense inventing add-ons or inventing inferior products that the add-ons could be even larger and that can be a way to raise equilibrium prices and you know what we're going to try to do in this paper is sort of look at demand and look what demand looks like and see you know can this sort of model of add-on pricing explain the equilibrium markups that we end up with okay so um paper we're going to look at four different products um four different products are8 megabyte PC 100 memory modules 128 megab pc133 modules and 256 megabyte versions um we're going to treat these four categories differently and just basically do it as an opportunity to have sort of four separate products we'll analyze each one separately and just get some reinforcement of the conclusions we would get from one U but we do spend most of the time I'll Focus today on 120 megabyte pc- 100 memory modules that was the product that was most popular and stayed alive for longest over the uh for the you know one-year data collection so what is our data collection so you know within each of these categories um there's one retailer that we have uh we got internal data from and then we have data external data from all the other retailers from pricewatch the one retailer that we had the special data connection with that retailer sold three prices three products in each category we'll call them the lowquality product the medium quality product the high quality product you can roughly think of the lowquality product as something that looks like this the medium quality product looks like this the high quality product looks like this um something to know about them is in this market um the medium quality products really didn't cost much more than the low quality products often like if these cost $60 wholesale these would cost $61 wholesale uh the retailer we got the data from said in fact sometimes he would even sell someone the $60 product and then Shi the $61 product just because dealing with the hassles and the customer you know customer disappointment when the prod when they got the lowquality product then they broke it was broken they return it it wasn't worth his hassles and you sometimes sell them a better product than they bought just because these really didn't cost much more than these ones um but anyway but he had the three products available um the high quality products that had the name brand chips in them did cost substantially more um when you're thinking about this though note that high quality versus low quality it's really a bundle so it's not case that you can compare my medium quality medium quality product from firm a and medium quality product from firm B Because quality is this many many dimensional things what's the restocking fee policy what's the shipping policy what's the how many layers do the board have what's the cast latency you know these are uh not something that's comparable across retailers so we can really sort of every different retailer may hav