To Groupon or Not to Groupon: The Profitability of Deep Discounts

Benjamin Edelman Sonia Jaffe Scott Duke Kominers

Working Paper
11-063

Copyright 2010 by Benjamin Edelman, Sonia Jaffe, and Scott Duke Kominers Working papers are in draft form. This working paper is distributed for purposes of comment and discussion only. It may not be reproduced without permission of the copyright holder. Copies of working papers are available from the author.

To Groupon or Not to Groupon: The Protability of Deep Discounts

Benjamin Edelman Sonia Jae Scott Duke Kominers December 17, 2010

Abstract We examine the protability and implications of online discount vouchers, a new marketing tool that oers consumers large discounts when they prepay for participating merchants goods and services. Within a model of repeat experience good purchase, we examine two mechanisms by which a discount voucher service can benet aliated merchants: price discrimination and advertising. For vouchers to provide successful price discrimination, the valuations of consumers who have access to vouchers must systematically dier fromand be lower thanthose of consumers who do not have access to vouchers. Oering vouchers is more protable for merchants which are patient or relatively unknown, and for merchants with low marginal costs. Extensions to our model accommodate the possibilities of multiple voucher purchases and merchant price re-optimization. Keywords: voucher discounts, Groupon, experience goods, repeat purchase.

The authors appreciate the helpful comments and suggestions of Peter Coles, Clayton Featherstone, Alvin Roth, and participants in the Harvard Workshop on Research in Behavior in Games and Markets. Kominers gratefully acknowledges the support of a National Science Foundation Graduate Research Fellowship, a Yahoo! Key Scientic Challenges Program Fellowship, and a Terence M. Considine Fellowship. Harvard Business School; bedelman@hbs.edu. Department of Economics, Harvard University; sjae@fas.harvard.edu. Department of Economics, Harvard University, and Harvard Business School; skominers@hbs.edu.

Introduction

A variety of web sites now sell discount vouchers for services as diverse as restaurants, skydiving, and museum visits. To consumers, discount vouchers promise substantial savingsoften 50% or more. To merchants, discount vouchers oer opportunities for price discrimination as well as exposure to new customers and online buzz. Best known among voucher vendors is Chicago-based Groupon, a two-year-old startup already touting a ten-digit valuation and, purportedly, recently rejecting a $6 billion acquisition oer from Google (Surowiecki (2010)). Hundreds of websites oer discount schemes similar to that of Groupon.1 The rise of discount vouchers presents many intriguing questions: Who is liable if a merchant goes bankrupt after issuing vouchers but before performing its service? What happens if a merchant simply refuses to provide the promised service? Since vouchers entail prepayment of funds by consumers, do buyers enjoy the consumer protections many states provide for gift certicates (such as delayed expiration and the right to a cash refund when value is substantially used)? Must consumers using vouchers remit tax on merchants ordinary menu prices, or is tax due only on the voucher-adjusted prices consumers actually pay? What prevents consumers from printing multiple copies of a discount voucher and redeeming those copies repeatedly? To merchants considering whether to oer discount vouchers, the most important question is the basic economics of the oer: Can providing large voucher discounts actually be protable? Voucher discounts are worthwhile if they predominantly attract new customers who regularly return, paying full price on future visits. But if vouchers prompt many long-time customers to use discounts, oering vouchers could reduce prots. For most merchants, the eects of oering vouchers lie between these extremes: vouchers bring in some new customers, but also provide discounts to some regular customers. In this paper, we oer a model to explore how consumer demographics and oer details interact to shape the protability of voucher discounts. We illustrate two mechanisms by which a discount voucher service can benet aliated merchants. First, discount vouchers can facilitate price discrimination, allowing merchants to oer distinct prices to dierent consumer populations. In order for voucher oers to yield protable price discrimination, the consumers who are oered the voucher discounts must be more price-sensitive (with regards to participating merchants goods or services) than the population as a whole. Second, discount vouchers can benet merchants through advertising, by informing consumers of a merchants existence. For these advertising eects to be important, a merchant must begin with suciently low recognition among prospective consumers. The remainder of this paper is organized as follows. We review the related literature in Section 2. We present our model of voucher discounts in Section 3, exploring price discrimination and advertising eects. In Section 4, we extend our model to consider the possibility of consumers purchasing multiple vouchers and of merchants adjusting prices
1 Seeing these many sites, several companies now oer voucher aggregation. Yipit, one such company, tracked over 200 dierent discount voucher services as of December, 2010.

in anticipation of voucher usage. Finally, in Section 5, we discuss implications of our results for merchants and voucher services.

Related Literature

The recent proliferation of voucher discount services has garnered substantial press: a multitude of newspaper articles and blog posts, and even a short feature in The New Yorker (Surowiecki (2010)). However, voucher discounts have received little attention in the academic literature. To our knowledge, the only prior academic study of online voucher discounts is the survey work of Dholakia (2010). In that work, Dholakia (2010) polls businesses that offered Groupon discounts. Echoing sentiments expressed in the popular press,2 Dholakia (2010) nds mixed empirical results: some business owners speak glowingly of Groupon, while others regret their voucher promotions. Unlike the survey approach of Dholakia (2010), we seek to understand voucher discount economics on a theoretical level. Our results indicate that voucher discounts are naturally good ts for certain types of merchants, and poor ts for others; these theoretical observations can help us interpret the range of reactions to Groupon and similar services. Although there is little academic work on voucher discounts, a well-established literature explores the advertising and pricing of experience goods, i.e. goods for which some characteristics cannot be observed prior to consumption (Nelson (1970, 1974)). The parsimonious framework of Bils (1989), upon which we base our model, studies how prices of experience goods respond to shifts in demand. Bils (1989) assumes that consumers know their conditional valuations for a rms goods, but do not know whether that rms goods t until they have tried them.3 Analyzing overlapping consumer generations, Bils (1989) measures the tradeo between attracting more rst-time consumers and extracting surplus from returning consumers. Meanwhile, much of the work on experience goods concerns issues of information asymmetry: if a merchants quality is unknown to consumers but known to the merchant, then advertising (Nelson (1974); Milgrom and Roberts (1986)), introductory oers (Shapiro (1983); Milgrom and Roberts (1986); Bagwell (1990)), or high initial pricing (Bagwell and Riordan (1991); Judd and Riordan (1994)) can provide signals of quality. Of this literature, the closest to our subject is the work on introductory oers. Voucher discounts, a form of discounted initial pricing, may encourage consumers to try experience goods they otherwise would have ignored. However, we identify this eect in a setting without asymmetric information regarding merchant qualityconsumer heterogeneity, not information asymmetries, drives our main results.4 Additionally, our work diers from the classical
For example, Overly (2010) reports on Washington merchants mixed reactions to the LivingSocial voucher service. 3 Firms know the distribution of consumer valuations and the (common) probability of t. 4 Of course, our treatment of advertising includes a very coarse informational asymmetry: some consumers are simply not aware of the merchants existence. However, conditional upon learning of the merchant, consumers in our model receive more information than the merchant does about their valua2

literature on the advertisement of experience goods, as advertising in our setting serves the purpose of awareness, rather than signaling.5 A substantial literature has observed that selective discounting provides opportunities for price discrimination. In the settings of Varian (1980), Jeuland and Narasimhan (1985), and Narasimhan (1988), for example, merchants engage in sale pricing in order to attract larger market segments.6 Similar results have been found to motivate the use of cents-o coupons, certicates which promise discounts on repeat (rather than initial) purchases (Cremer (1984); Narasimhan (1984)). We bring the insights of the literature on saledriven price discrimination to bear on voucher discountinga new sale technology. Like the price-theoretic literature which precedes our work, we nd that price discrimination depends crucially upon the presence of signicant consumer heterogeneity. Although our work is theoretical, perhaps most central for an empirical comparison of our work to its antecedents is the observation that the prior theoretical literature, including the articles discussed above, has considered only marginal pricing decisions. The previous work on experience goods and price discrimination does not consider deep discounts of the magnitudes now oered by voucher services.

Model

Oering a voucher through Groupon has two potential advantages: price discrimination and advertising. We present a simple model in which a continuum of consumers have the opportunity to buy products from a single rm. The consumers are drawn from two populations, one of which can be targeted by voucher discount oers. First, in Section 3.1, we consider the case in which all consumers are aware of the rm and vouchers serve only to facilitate price discrimination. Then, in Section 3.2, we introduce advertising eects. We present comparative statics in Section 3.3. Our model has two periods, and the rm ex ante commits to a price p for both periods. The rm and consumers share a common discount factor . Following the setup of Bils (1989), consumers share a common probability r that the rms product is a t. Conditional on t, the valuation of a consumer i for the rms oering is vi . A consumer i purchases in the rst period if either the single-period value, rvi p, or
tions. This is in sharp contrast to much of the previous work on experience goods, in which merchants can in principle exploit private quality information in order to lead consumers to purchase undesirable (or undesirably costly) products (e.g., Shapiro (1983); Bagwell (1987)). 5 In the classical theory of experience goods, advertising serves a burning money role. Merchants with high-quality products can aord to advertise more than those with low-quality products can, as consumers recognize this fact in equilibrium and ock to merchants who advertise heavily (Nelson (1974); Milgrom and Roberts (1986)). In our model, advertising instead serves to inform consumers of a merchants existence; these announcements are a central feature of the service voucher vendors promise. 6 In other models, heterogeneity in consumer search costs (e.g., Salop and Stiglitz (1977)) or reservation values (e.g., Sobel (1984)) motivate sales. Bergemann and Vlimki (2006) study the pricing paths of a a mass-market and niche experience goods, nding that initial sales are essential in niche markets to guarantee trac from new buyers.

the expected discounted future value, rvi p + r(vi p), is positive, i.e. if max{rvi p, rvi p + r(vi p)} 0. For > 0, there is an informational value to visiting in the rst period: if a consumer learns that the rms product is a t, then the consumer knows to return. As a result, all consumers with values at least v(p) 1 + r p>p r + r

purchase in the rst period. To consider the eects of oering discounts to a subset of consumers, we assume there are two distinct consumer populations. Proportion of consumers have valuations drawn from a distribution with cumulative distribution function G, while proportion 1 have valuations drawn from a distribution with cumulative distribution function F . We denote by V supp(F ) supp(G) the set of possible consumer valuations. We assume that G(v) F (v) for all v V , i.e. that the valuations of consumers in the G population are systematically lower than those of consumers in the F population. The rm faces demand (1 G(v(p))) + (1 )(1 F (v(p))) in the rst period, and fraction r of those consumers return in the second period. The rm maximizes prots given by (p) (1 + r)((1 G(v(p))) + (1 )(1 F (v(p))))(p c), where c is the rms marginal cost. The rst-order condition of the rms optimization problem is (1 G(v )) + (1 )(1 F (v )) 1 + r ((p c)(g(v ) + (1 )f (v )) = 0 r + r (1)

where p is the optimal price and v v(p ). We assume that the distribution of consumers is such that prots are single-peaked, so that p is uniquely dened.

3.1

Discount Vouchers

After setting its optimal price p , the rm is given the opportunity to oer a discount voucher.7 Only fraction of consumers in the G population have access to the discount voucher system. (Consumers in the F population cannot buy vouchers.) Through the voucher system, these consumers have the opportunity to purchase from the rm at a discounted price p in the rst period. If > r, then everyone who purchases in the rst period will return if the rm is a t; the minimal valuation of consumers who purchase is
For now, we assume the rm did not consider the possibility of a voucher when setting its price. In Section 4.2, we consider the possibility of re-optimization.
7

 + r p > v(p). r + r If < r, then some consumers i with vi < p will purchase in the rst period but will not return (even if the rm is a t). Those will be the consumers i for whom v (p) rvi p > 0 and vi < p.

The voucher service retains proportion 1 of the voucher sales price, so the rms revenue from each discount voucher is p . Dening v v (p ) and v p , we see that r oering discount vouchers yields the following change in prots: (p ) = ((1 + r)(p c)(G(v ) G()) (1 )p (1 G())) v v > r, (2) (p c)(G(v ) G() + r(G(v ) G(p )) (1 )p (1 G()) < r. v v

The rst term of (2) represents the increased prots from the gain in consumers G(v ) G(); the second term measures prots lost by lowering the rst-period price v from p to p . Note that aects the magnitude of the change in prots, but not the sign.8 The following result is therefore immediate. Proposition 1. If it is protable to oer the voucher discount to one consumer (randomly drawn from the G population), then it is more protable to oer the voucher discount to all such consumers. When = 1, introducing the discount voucher does not aect consumers purchase decisions (v = v ), as consumers are oered no actual discount.9 Meanwhile, if consumers who use discount vouchers have the same distribution of valuations as other consumers (F (v) = G(v) for all v), then the rst- and second-order conditions show that the change in prots is maximal at = 1, so if < 1 then vouchers are unprotable. Conversely, if the distributions of valuations are well behaved and G is to the left of F , then there is some < 1 such that oering a discounted price of p p can be protable. This is because the rm wants to oer a lower price to consumers from the G population; setting < 1 brings the rms price to the G population closer to the optimal price for that population. We formalize these observations in the following proposition. Proposition 2. If F (v) = G(v) for all v and min{, } < 1, then the introduction of a discount voucher yields a decrease in prot. By contrast, if F (v) < G(v)
8 9

v V,

By contrast, aects p and can therefore aect the sign of (p ). If < 1, then vouchers decrease rm prots when = 1 because the voucher service keeps a cut of revenue. But this case is degenerate: ifas we have assumed in this sectionthe only benet of vouchers is price discrimination, then certainly the rm will only use vouchers if they oer actual price discrimination opportunities.

then there is some < 1 such that oering a discount voucher with discounted price of p p is protable, so long as is suciently close to 1. The conditions of Proposition 2 are sucient for the result, but not necessary. Indeed, whenever F and G are such that the rms optimal price for consumers from the G population is lower than the rms optimal price for the combined distribution of consumers, there are some , < 1 for which the discount voucher is protable.

3.2

Advertising Eects

In addition to oering consumers discounts, vouchers may also serve to inform consumers about the existence and details of aliated rms. We suppose that fraction of consumers know about a participating rm, but fraction 1 do not. The uninformed consumers may have high valuations for the rms product (conditional on learning about the rm), but they cannot purchase without rst being informed of the rms existence. If the probability of knowing about a rm is constant across consumers (independent of valuations), then a rms prots absent an online voucher are given by (p ) (p ); the optimal price p is unchanged from that dened by (1). In that case, if a rm oers a voucher, then all consumers receiving the voucher learn of the rms existence. The implied change in prots is:

(p ) = (p ) + (1 ) (p c + r(p c))(1 G()) v > r, (3) (p c)(1 G()) + r(p c)(1 G(p )) < r. v

Because uninformed consumers do not purchase absent the voucher, no prots are foregone by oering those consumers discounted pricing. Thus, so long as the discount is not so big that losses in the rst period outweigh gains in the second period, the second term in (3) is positive. A sucient (but not necessary) condition for this to occur is that the rms per-voucher revenue is larger than its marginal cost: p c. (4)

This condition (4) is not sucient for voucher protability absent the advertising eect it does not guarantee that (p ) is positive. However with advertising, as long as condition (4) holds, vouchers are always protable for suciently unknown rms as there is always some > 0 such that the second term of (3) dominates the rst. Proposition 3. Suppose that p c. Then, there exists some > 0 such that oering vouchers is protable whenever < . Unlike in the pure price discrimination case, when vouchers provide advertising benets the eect of the dierence between the distributions of valuations on the protability of 7

vouchers is ambiguous. For a xed p , the additional prots from advertising are higher when the valuations drawn from G(v) are higher (smaller dierence between distributions). However, changing G(v) changes p , so that the net eect of changing G depends on the shapes of F and G.

3.3

Example Comparative Statics

To obtain closed-form comparative statics, we consider the case where valuations are uniformly distributed in each population, but the population means vary: 0 v < 0 0 v < F (v) = v v [0, 1] G(v) = v + v [, 1 ] 1 v > 1, 1 v > 1 . Here, parameterizes the dierence between the two distributions: it measures the amount that G(v) is shifted to the left relative to F (v). For F and G so dened, the optimal price p is given by: p = 1 2 c+ (1 + ) (1 )r . (1 + r)

In this expression, (1 )r = ((1 ) 1 + (1 ))r is the weighted average of 1+ the maximal valuations of the two consumer populations. The multiplicative factor 1+r represents the premium consumers are willing to pay because of the information value of purchasing in the rst period. For simplicity, we consider the case in which = 1 (the voucher vendor charges no fee), > r (the proportion of the price paid with the voucher is greater than the probability of t), and = 1 (there are no advertising eects). With these assumptions, the minimum dierence between the distributions necessary for a discount voucher to be protable, denoted , is given by = (1 )(r(1 + ) + c(1 + r)) . (1 + )r(2(1 + r(1 )) (1 + ))

The comparative statics of are intuitive. Patience: The change in rm prots increases in , the discount factor, while decreases in . This is consistent with the interpretation that discount vouchers let a rm accept losses (decreased prots) in the short run in exchange for increased prots in the future, a benet more valuable to patient rms. Costs: Increased marginal costs lower the change in prots and raise the threshold for protability. Since oering a discount voucher increases quantity sold and decreases per-unit price, it is more protable for rms with lower marginal costs.

 Population share: If many consumers come from the low-value population (i.e. if is high), then it is dicult for the voucher service to be protable for rms. When is high, voucher users valuations signicantly aect the base price p so that, for a given , price discrimination is less protable. Discounted price: Although the eect of on the change in prots is ambiguous, increasing unambiguously decreases the threshold . This is intuitive, as the less of a price decrease the voucher represents, the less heterogeneity across populations is required in order for the discounted voucher to be protable. Probability of t: Like with , the eect of a higher probability of t r on the change in prots is unclear, but increasing r unambiguously lowers .

4
4.1

Extensions
The Possibility that Consumers Might Purchase Multiple Vouchers

So far we have assumed that each consumer can purchase at most one voucher, a restriction which lets us model the discounted price p as being available only in the rst period. However, if consumers can buy multiple discount vouchers, then they can enjoy the discounted price in both periods. If all discount vouchers must be purchased in the rst period (as is the case with Groupon and similar voucher vendors), then a consumer i with valuation vi only nds it protable to buy a second voucher if the expected value of second-period consumption is greater than the present voucher cost, that is, if rvi p .
Consumers with valuations vi [ p, r p] buy a discount voucher and do not return. r Consumers with vi [ r p, p] only return if they can buy two vouchers (and the rm is a match). Consumers with vi > p return whenever the rms product is a match. The prots in the rst period are the same as when consumers only purchase a single discount voucher. However, if consumers are able to purchase multiple discount vouchers, then some consumers who would normally return at full price instead return with a second voucher. It follows that allowing consumers to purchase multiple discount vouchers is only protable if the rm seeks to oer starkly dierent prices to the two consumer populations.

Proposition 4. Suppose that there exist and such that it is protable, relative to the possibility of allowing each consumer only one voucher purchase, for a rm to allow consumers to purchase two vouchers. Then, there exist and with < such that it is protable to oer a single voucher at rates and , but allowing consumers to purchase an additional voucher at those rates decreases prots.

The prots from allowing the purchase of a second voucher are negative at = 0, so if they are positive for some and they must cross zero at intermediate values of , . Since the prots from allowing the second voucher are lower than the prots from initially oering a voucher, it will still be protable to oer a single voucher at those intermediate values.

4.2

The Possibility that the Firm Might Re-optimize Its Prices

If the rm can adjust its posted prices to account for the presence of discount vouchers, then the rm will re-optimize to set a new price p maximizing ( p c + r( c))(1 G(( ))) p v p + (1 + r) (1 )(1 G(v(p))) + (1 )(1 F (v(p))) . Pricing at p raises the prots from a discount voucher promotion and lowers the required minimum dierence between distributions required for voucher protability. The prospect of adjusting prices places new importance on : the sign of the change in prots now depends on because the proportion of consumers receiving the discounted price aects the rms re-optimization. Even with the opportunity of price re-optimization, discount vouchers are not profitable when the two populations are identical (F (v) = G(v) for all v). Under the assumptions of Section 3.3, for suciently large, is decreasing in . In this case the rm would prefer to set dierent prices for the two consumer populations, it can do this more eectively when more of the G consumers may acquire vouchers (i.e. when is large). When is low, the rm is limited in the extent to which it can raise the price faced by consumers from the F population because many G-population consumers face the same price.

5
5.1

Discussion
Implications for Merchants

Our results oer practical advice for rms considering whether to oer discount vouchers. Our discussion in Section 4.1 indicates that a rm might want to disallow purchase of multiple discount vouchers. But as argued by Friedman and Resnick (2001), rms face substantial practical diculties in implementing this restriction. What stops consumers from creating multiple accounts and purchasing one voucher through each such account? To date, it appears that few rms have adjusted prices in anticipation of many consumers using discount vouchers. However, we did notice a sharp price increases at one local restaurant that oers frequent discounts through discount voucher services. Visiting this restaurant, our experience has been that at least half of consumers come bearing discount vouchers. Vouchers have noticeably increased the restaurants trac, yet at prices (net of voucher services fees, even bearing in mind the price increase) that surely reduce the restaurants margins. We doubt the restaurant would have enjoyed a similar trac 10

increase had it simply lowered its prices 25%; clearly vouchers play an important role in making consumers aware of discounts. Yet with voucher services retaining as much as half the consumers prepayment, heavy reliance on discount voucher marketing comes at a signicant cost. Firms would do well to assess whether their voucher-using customers have already visited and previously paid full price. Firms should also measure how many voucherusing customers later come back without vouchers. In principle, credit card systems could track this information with little harm to customer privacy or data security. But in practice, most rms currently lack the tools or expertise to run such analyses.

5.2

The Future of Voucher Services

Our model reveals that a discount voucher service is more likely to be protable for afliated rms, all else equal, if customers using that service have valuations substantially dierent from (and in particular, below) those of other customers. But notice the diculty as the voucher service grows: As more consumers use vouchers, voucher-users necessarily come to resemble average consumers, so the use of a voucher comes to convey less information about a consumers valuation. Thus, as voucher services grow, their aliates will have to rely on voucher advertising rather than voucher price discrimination. Current voucher services prots and recent growth may therefore not be good predictors of those services future values. Meanwhile, we are struck by the large fees that leading discount voucher services currently charge to participating rms. Groupon charges 50% of voucher price: if a restaurant oers a $20 voucher for $40 of food, Groupon retains $10a large fee relative to Groupons marginal costs of voucher provision (Groupon (2010)). As our results in Section 3.1 indicate, such large fees may impede rms usage of discount vouchers. However, competition among discount voucher services may drive down these fees.10,11 Recently, Groupon has begun to unbundle some of its services. As of December, 2010, Groupon is charging just 10% to a rm which only wants its oer to appear following user searches, while continuing to require the 50% fee for Groupons traditional service of featuring a merchant in a daily email to all consumers in a given city (Groupon (2010)). On one view, the 10% fee might look like a bargain to rms10% is certainly far less than 50%. However, if Groupons low-priced service only gives a rm access to consumers who already know about that rm, then it provides no advertising opportunitiesonly price discrimination.

Small voucher services may charge fees as low as 10%. Yet small voucher services cannot provide substantial advertising benets. As a result, despite lower fees at small voucher services, many rms prefer the larger services. 11 Another issue of competition in the voucher market is the possibility that a well-positioned competitor (such as Google, Facebook, or Yelp) could develop a similar service and promote that service to its existing users.

10

11

References
Bagwell, K. (1987). Introductory price as a signal of cost in a model of repeat business. Review of Economic Studies 54 (3), 365384. Bagwell, K. (1990). Informational product dierentiation as a barrier to entry. International Journal of Industrial Organization 8 (2), 207223. Bagwell, K. and M. H. Riordan (1991). High and declining prices signal product quality. American Economic Review 81 (1), 224239. Bergemann, D. and J. Vlimki (2006). Dynamic pricing of new experience goods. Journal a a of Political Economy 114 (4), 713743. Bils, M. (1989). Pricing in a customer market. Quarterly Journal of Economics 104 (4), 699718. Cremer, J. (1984). On the economics of repeat buying. nomics 15 (3), 396403. RAND Journal of Eco-

Dholakia, U. M. (2010). How eective are Groupon promotions for businesses? Mimeo, Rice University. Friedman, E. and P. Resnick (2001). The social cost of cheap pseudonyms. Journal of Economics and Management Strategy 10 (2), 173199. Groupon (2010, December 15). Run your own deals with Groupon stores. http://www.groupon.com/merchants/welcome. Jeuland, A. P. and C. Narasimhan (1985). Dealing-temporary price cuts-by seller as a buyer discrimination mechanism. Journal of Business 58 (3), 295308. Judd, K. L. and M. H. Riordan (1994). Price and quality in a new product monopoly. Review of Economic Studies 61 (4), 773789. Milgrom, P. and J. Roberts (1986). Price and advertising signals of product quality. Journal of Political Economy 94 (4), 796821. Narasimhan, C. (1984). A price discrimination theory of coupons. Marketing Science 3 (2), 128147. Narasimhan, C. (1988). Competitive promotional strategies. Journal of Business 61 (4), 427449. Nelson, P. (1970). Information and consumer behavior. omy 78 (2), 311329. Journal of Political Econ-

Nelson, P. (1974). Advertising as information. Journal of Political Economy 82 (4), 729 754. 12

Overly, S. (2010, December 6). LivingSocial deals draw mixed merchant reactions. Washington Post. Salop, S. and J. Stiglitz (1977). Bargains and ripos: A model of monopolistically competitive price dispersion. Review of Economic Studies 44 (3), 493510. Shapiro, C. (1983). Optimal pricing of experience goods. Bell Journal of Economics 14 (2), 497507. Sobel, J. (1984). The timing of sales. Review of Economic Studies 51 (3), 353368. Surowiecki, J. (2010, December 20). Groupon clipping. The New Yorker , 46. Varian, H. R. (1980). A model of sales. American Economic Review 70 (4), 651659.

13