Second-degree price discrimination by Japanese newspapers

G Model JAPWOR 903 No. of Pages 12

Japan and the World Economy xxx (2017) xxx–xxx

Contents lists available at ScienceDirect

Japan and the World Economy journal homepage: www.elsevier.com/locate/jwe

Second-degree price discrimination by Japanese newspapers David Flath Faculty of Economics, Ritsumeikan University, Noji Higashi 1 chome, 1-1 Kusatsu Shiga 525-8577, Japan

A R T I C L E I N F O

Article history: Received 4 August 2017 Received in revised form 13 October 2017 Accepted 14 October 2017 Available online xxx JEL classifications: D4 L4 Keywords: Second-degree price discrimination Two-sided markets Newspapers Advertising

A B S T R A C T

In Japan, the newspaper publishers with the greatest daily circulation offer both morning and evening editions in most of their distribution areas, and many of them allow their customers to choose between morning-only and morning-and-evening subscriptions. Each such newspaper publisher, in setting the subscription prices and numbers of pages of content in its morning edition and evening edition, must take into account self-selection of demanders as to type of subscription. The subscription offerings thus amount to second-degree price discrimination, which is the main reason why it might be profitable to even publish an evening edition along with the morning one. Estimates here show that the Japanese newspaper publishers that do offer optional evening editions enlarge their profits by around eight percent of their morning-edition subscription revenue. Furthermore, these newspaper publishers increase consumer surplus by an amount equal to half what they add their own profit. The second-degree price discrimination increases social welfare. © 2017 Elsevier B.V. All rights reserved.

1. Introduction The main contributions here are the following. First is a new method of estimating demand for newspaper subscriptions from aggregate data on subscription prices and pages of content of evening editions and morning editions, using nonlinear least squares. Estimates of demand parameters for Japanese newspapers using this method with aggregate data resemble estimates for those same newspapers based on multinomial logit specification with control function and using micro-data in Flath (2016). The new method is valuable because it could perhaps be implemented in cases where the data do not support other methods of demand estimation. The second contribution is an economic model of profit-maximizing subscription prices and pages of content for a newspaper if offering its subscribers a choice between morningonly subscription and morning-and-evening subscription, or if offering only morning subscriptions. It is the same model that underlies the new estimating method, but it has some value itself as a simple theoretical framework for thinking about seconddegree price discrimination in an oligopoly, even if its parameters could not be estimated. Third are the actual estimates and their interpretation: Japanese newspapers that offer their subscribers a choice between morning-only and morning-and-evening subscriptions enlarge their profit by about eight percent of their morning edition subscription revenue, compared to the case if

E-mail address: fl[email protected] (D. Flath).

offering only morning edition, and increase the consumers’ surplus by about half the increase in their own profit. The three contributions together amount to a novel and tractable model for empirical analysis of second-degree price discrimination in an oligopoly. The model is applied to Japanese newspapers but could be extended to other cases of oligopolistic second-degree price discrimination such as the setting of classes of air fares. Welfare analysis of price discrimination requires such empirical modelling. The finding that second-degree price discrimination by Japanese newspapers is both profitable and enlarges consumer surplus is a useful caution against antitrust rules that might be concocted to prohibit such price discrimination. The model could also be used as a tool for management decision-making by Japanese newspapers, to set prices and amount of content that profitably track changes in demand and costs. In Japan, the newspaper publishers with the greatest daily circulation offer both morning and evening editions in most of their distribution areas. The national dailies (Yomiuri, Asahi, Mainichi, and Sankei) all do so and the bloc newspapers (Chunichi, Chugoku and Hokkaido Shimbun) do as well. Typically, these offer subscribers a choice between subscribing to the morning edition only, or subscribing to both morning and evening edition. In some locales away from urban centers, the national newspapers are only offered as morning editions. And a number of regional newspapers with smaller circulation (To-o Nippo, Oita Godo, Iwate Nippo, Yamagata Shimbun, Ryukyu Shimpo) offer both morning and evening editions together, but with no option of subscribing to the morning edition only. Shizuoka Shimbun purports to offer

https://doi.org/10.1016/j.japwor.2017.10.003 0922-1425/© 2017 Elsevier B.V. All rights reserved.

Please cite this article in press as: D. Flath, Second-degree price discrimination by Japanese newspapers, Japan World Econ. (2017), https://doi. org/10.1016/j.japwor.2017.10.003

G Model JAPWOR 903 No. of Pages 12

2

D. Flath / Japan and the World Economy xxx (2017) xxx–xxx

subscribers a choice between morning and evening editions or just evening edition, but almost none opt for just the evening edition. Each newspaper publisher that offers subscribers a choice between morning edition and morning-plus-evening edition, in setting its subscription prices and numbers of pages of content, knowingly influences the self-selection of demanders as to type of subscription. Though seldom remarked, the pricing and quality of morning-only subscriptions and morning-and-evening subscriptions to the same newspaper is an instance of Pigouvian seconddegree price discrimination. Other examples of such pricing abound. Transport fares offered in differing quality classes, beverage prices that differ per unit depending on the size of container, prices of electricity that vary depending on monthly number of kilowatts used, and so on, are obvious examples. A newspaper publisher that offers its morning subscribers an option to also subscribe to an evening edition will do so with an eye to attracting more low-valuing customers without sacrificing the profit it earns by servicing its higher valuing customers. The publisher would shift content pages from the morning edition to its newly offered evening edition, and at the same time lower the price of a morning-only subscription compared to the price it had set when not offering an evening edition. Whether the self-selection by demanders that this behavior induces can be profitable for the publisher depends on the demand it faces. For instance, if all demanders were identical, then it would not be profitable (as second-degree price discrimination) to offer morning subscribers the option of also subscribing to an evening edition. But subscribers are not identical to one another. Estimates here show that the Japanese newspaper publishers that do offer optional evening editions enlarge their profits by around eight percent of their morning-edition subscription revenue. Furthermore, these newspaper publishers increase consumer surplus by an amount equal to half what they add to their own profit. The second-degree price discrimination increases social welfare. All of this requires a general framework for thinking about newspaper pricing that takes into account that newspapers are platforms in two-sided markets. Newspaper publishers sell both subscriptions and advertising, and the demand for advertising depends on the number of subscribers. The new estimates on which my analysis partly rests exploits the variation in pricing and circulation of the differing editions offered by same newspaper publishers, to identify parameters of the demand for newspaper content. I estimate these parameters using nonlinear least squares and find that price elasticity of demand is less than 2 and elasticity of demand with respect to pages of content is less than 1. Producing these estimates solves a difficult problem in that many of the newspaper publishers set exactly the same prices as one another. With virtually no crosssection (or temporal) variation in prices it is still possible to infer price elasticity of demand based on an assumed relationship between individuals’ valuations of pages of content and their valuations of subscriptions. Furthermore, this new method of estimating elasticities of demand for newspaper subscriptions offers a useful cross-check of estimates based on multinomial logit specification of an indirect utility function. This paper contributes to the empirical literature on seconddegree price discrimination. It also contributes to the econometric analysis of platforms in two-sided markets, newspapers in particular. In a study that occupies a similar niche to this one, Angelucci et al. (2013) have identified the difference between subscription price and kiosk price of French newspapers as an example of second-degree price discrimination. They show empirically that following the 1968 introduction of television advertising in France the national newspapers reduced the difference between their subscription prices and kiosk prices, that is, reduced their price discrimination. Other empirical studies of second-degree price discrimination, that is to say, price discrimination relying on self-selection of demanders

when presented with a pricing schedule, include Cohen (2008) for paper towels, Busse and Rysman (2005) for yellow pages advertising, McManus (2007) for coffee drinks, and Shum and Crawford (2007) for cable television subscriptions. Other econometric analyses of newspaper pricing, not focused on price discrimination, include Argentesi and Filistrucchi (2007) for Italy, Van Cayseele and Vanormelingen (2009) for Belgium, and Fan (2013) for the United States. A novel contribution here is an original, partial-equilibrium model of the demands for morning edition and evening edition subscriptions offered by the same newspaper publisher. This model is used in determining how a newspaper publisher’s offering its morning subscribers the option of also subscribing to an evening edition affects its subscription prices, pages of content in each edition, and profit, and how it affects consumer surplus. This leads to the main empirical finding of this paper, already mentioned, that Japanese newspaper publishers who do offer evening editions enlarge their profits and also increase consumer surplus. The inferences in this paper presume that the choice of each newspaper publisher takes as given the prices and contents of rival newspaper publishers. I will not address the strategic interdependence among the choices of the various newspaper companies, for example, how one newspaper publisher offering an evening edition affects whether rival newspaper publishers would do so. Furthermore, in the framework of this paper, I will assume that demanders regard the content pages of any one newspaper publisher as homogeneous, but that each demander does differentiate among the pages offered by different publishers. This is a novel and tractable way of modelling second-degree price discrimination in an oligopoly, and may be applicable to industries other than newspaper publishing, for example airlines. 2. Basic framework I begin by describing the pricing and content choices of a singleedition newspaper, one that is offered in morning edition only, taking as given the prices and contents of rival newspapers. Potential subscribers to the newspapers are households. I will presume that each household subscribes to at most one newspaper—either morning-only or morning-and-evening. Each element of the household choice set is either: (i) a combined subscription to both morningand-evening editions of the same newspaper, (ii) a subscription to the morning-edition-only of a newspaper, or (iii) no subscription to any newspaper. This matches the facts of the Japanese newspaper industry fairly closely. Based on the March 2007 random-direct-dial telephone survey of 26,490 persons living throughout Japan conducted by Video Research, Ltd. (the 2007 edition of its annual JREAD survey), virtually all who subscribe to an evening edition also subscribe to the morning edition of the same newspaper. And more than 95 percent of the respondents subscribed to only one ordinary newspaper or none (I exclude from this the Nikkei Shimbun, which is a business daily, similar to the Wall Street journal—most of its subscribers also subscribe to an ordinary newspaper). Thus, each household has its own discrete next-best alternative to subscribing to any one particular single-edition newspaper. Each household’s reservation price of subscribing to the one particular newspaper edition is the highest price at which it would subscribe to that one and forego its next most-preferred alternative. In the model of this paper, the next-best alternative of each subscriber to morning-and-evening edition will turn out to be a morning-only subscription to the same newspaper, and the next-best alternative of each morning-only subscriber will be either a morning-only subscription to a different newspaper or no subscription. More general substitution patterns can be imagined, but this pattern follows from the primitive assumptions about preferences of my model. More about this in due course.

Please cite this article in press as: D. Flath, Second-degree price discrimination by Japanese newspapers, Japan World Econ. (2017), https://doi. org/10.1016/j.japwor.2017.10.003

G Model JAPWOR 903 No. of Pages 12

D. Flath / Japan and the World Economy xxx (2017) xxx–xxx

The choices of the newspaper publisher are constrained not only by the demand for subscriptions, but also by the demand to place ads in the newspaper. Let us suppose that the demand for a page of ads “a” in the particular newspaper depends on the price to place such an ad per subscriber “pa/s” and that the readers regard the ads indifferently. Here an ad is defined as a printed item supplied to all subscribers, the same as the subscription content k. The only difference is that the ad is paid for by the advertiser but the other content is paid for by the subscribers. To keep matters simple, posit a constant elasticity demand system facing the single-edition newspaper: u

s ¼ Aps js k

ð1Þ


p ja a¼B a ; s

ð2Þ

where js > u, ja > 1, 0 < u<1. Here, A > 0 and B > 0 are parameters, and s is the number of subscribers. Let the costs of newspaper production depend on number of ads, circulation, and amount of content. These costs include firstcopy costs, faa + fkk, and costs that depend on number of copies, c0 s þ cas þ cks Cost ¼ f a a þ f k k þ c0 s þ cas þ cks:

ca ¼

@Cost ¼ c0 þ ca þ ck @s

ð4Þ

@Cost ¼ f a þ cs @a

ð5Þ

@Cost ¼ f k þ cs: ck ¼ @k

ð6Þ

The newspaper chooses content, price of ads, and price of subscriptions to maximize total profit: max p ¼ ps s þ pa a  f a a  cas  f k k  cks  c0 s:

ps ;pa ;k

ð7Þ

The necessary conditions for maximum profit reduce to the following pricing rules1 : pa  ca 1 ¼ pa ja

ð9Þ

Profit-maximizing also requires that content k, the number of pages not including ads, fulfills the condition: ck k u ¼ : p s s js

ð10Þ

This expression resembles the Dorfman-Steiner condition describing optimal advertising expenditure. This is because, analytically, the model here is the same as the Dorfman-Steiner advertising model: Demand for subscriptions has constant elasticity (0 < u <1) with respect to content, just as in the Dorfman-Steiner model demand for the product has constant elasticity with respect to quantity of ads. Rearranging in a useful way, the pricing and content rules are:   1 1 ð11Þ pa ¼ ðf a þ csÞ 1 

ja

 
p a 1 1 1 ps ¼ c0 þ ca þ ck  a s js

ð12Þ

k¼

ups s : js ðf k þ csÞ

ð13Þ

These Eqs. (11)–(13) describe the Nash equilibrium choices of prices and content of the one newspaper publisher. The subscription price ps is set as though the newspaper were receiving a subsidy per subscriber equal to the marginal profit that arises indirectly from advertising with each copy sold (for given price of advertising pa):

@ðpa a  f a a  casÞ @a @ðpa a  f a a  casÞ @a þ ¼ ðpa  f a  csÞ  ca @a @s @s @s pa ja ¼ ðpa  f a  csÞ s p a ¼ a  ca: s

 ca

ð14Þ This marginal profit equals the marginal revenue that arises indirectly from advertising (=paa/s), minus the added-cost-percopy attributable to advertising (=ca). All of the exposition so far reprises a standard model of newspaper publisher pricing and content, which is the framework of many previous empirical studies of newspaper economics, including Argentesi and Filistrucchi (2007), Van Cayseele and Vanormelingen (2009), Fan (2013), and Flath (2016). In the following I will build on and extend this model to accommodate newspaper publishers that offer both morning-only and morningand-evening subscriptions.

ð8Þ 3. Subscription to morning-edition-only versus morning-andevening edition

1 The price-cost margin for subscriptions exemplifies the behavior of a multiproduct monopolist supplying two goods that are complements in demand. In

j

general, where the demands are as follows

ps  cs 1 p a ¼  a : ps js p s s

ð3Þ

Here, c0 is the unit cost of distribution net of advertisers’ payments to distributors for including inserts (substantial in Japan, almost as large as advertisers’ payments to newspapers for print ads), c is the cost per page of actually printing the newspaper (where k and a are the numbers of pages of content and of ads), fk is the first-copy cost of producing a page of content and fa is the firstcopy cost of producing a page of advertising. The incremental costs cs, ca, and ck of supplying subscriptions, ads and content are the following: cs ¼

3

j12

Q 1 ¼ Ap1 1 p2 Q2 ¼

j j Ap2 2 p1 21 ;

and marginal costs

1 þ are c1 and c2, the profit-maximizing pricing rules are: p1  c1 p1 ¼ j1    


 p2  c2 j21 p2 Q 2 ; and p2pc2 ¼ j1 þ p1pc1 jj12 pp1 QQ 12 : See for example 2 1 2 2 2 P2 p1 Q 1 j1 Tirole (1988, pp. 69–71). But here the cross-elasticities of demand are: @lna j12 ¼ ¼ ja js ð¼ “j1 j2 ” Þ; and j21 ¼ @ ln s ¼ 0. @ ln pa @lnps

Many of the large circulation daily newspapers in Japan have both morning and evening editions. Most of these offer their customers a choice between subscribing to the morning edition only, or subscribing to both editions. The evening editions of these papers generally have about one third as many pages of content as the morning editions and about one fifth the number of pages of ads. For detailed data on the subscription prices, pages of content and pages of ads of morning and evening editions of selected Japanese newspapers in 2007, refer to Table 1. The newspapers in this table include the various regional editions of the four national

Please cite this article in press as: D. Flath, Second-degree price discrimination by Japanese newspapers, Japan World Econ. (2017), https://doi. org/10.1016/j.japwor.2017.10.003

G Model JAPWOR 903 No. of Pages 12

4

D. Flath / Japan and the World Economy xxx (2017) xxx–xxx

Table 1 Morning and evening monthly subscription prices, circulation, pages of content and pages of ads; selected Japanese daily newspapers, 2007. Newspaper

THE ASAHI SHIMBUN (Hokkaido) THE ASAHI SHIMBUN (Nagoya) THE ASAHI SHIMBUN (Osaka) THE ASAHI SHIMBUN (Tokyo) THE ASAHI SHIMBUN (Seibu) MAINICHI SHIMBUN (Hokkaido) MAINICHI SHIMBUN (Nagoya) MAINICHI SHIMBUN (Osaka) MAINICHI SHIMBUN (Tokyo) MAINICHI SHIMBUN (Seibu) YOMIURI SHIMBUN (Hokkaido) YOMIURI SHIMBUN (Osaka) YOMIURI SHIMBUN (Tokyo) YOMIURI SHIMBUN (Seibu) YOMIURI SHIMBUN (Chubu) YOMIURI SHIMBUN (Hokuriku) SANKEI SHIMBUN (Osaka) SANKEI SHIMBUN (Tokyo) HOKKAIDO SHIMBUN THE KAHOKU SHIMPO THE TOKYO SHIMBUN THE CHUNICHI SHIMBUN KYOTO SHIMBUN KOBE SHIMBUN THE CHUGOKU SHIMBUN THE NISHI-NIPPON SHIMBUN

subscription price incl. tax (yen per month)

circulation

pm

pe

sm

3007 3007 3007 3007 3007 3007 3007 3007 3007 3007 3007 3007 3007 3007 3007 3007 2950 3007 3007 3007 2550 3000 3007 3007 3007 3007

918 918 918 918 918 918 918 918 918 918 918 918 918 918

155,750 424,246 2,336,911 4,358,660 791,140 72,617 176,101 1,419,552 1,640,998 664,559 232,992 2,547,583 6,068,547 914,687 160,333 108,299 1,224,957 975,652 1,204,151 501,356 585,508 2,774,585 510,000 561,881 717,794 854,655

975 918 918 700 925 918 918 918 918

pages of content per month

pages of ads per month

se

km

ke

am

ae

53,740 146,107 1,294,724 1,958,061 135,628 17,697 49,069 830,585 462,952 102,896 64,396 1,332,683 2,357,250 109,479

621 598 599 606 605 524 546 547 575 498 614 609 590 578 655 647 581 613 599 611 592 617 589 587 629 600

159 189 188 223 181 174 169 180 204 151 233 234 282 191

466 451 544 618 507 267 267 377 379 346 522 608 640 521 444 485 320 320 553 364 299 465 344 314 416 476

82 70 163 190 88 38 47 113 81 63 101 179 195 78

633,154 639,634 109,871 280,889 650,297 317,881 254,782 73,628 175,518

227 252 136 203 217 173 185 162 183

124 131 65 90 121 113 121 46 131

Sources: pages of ads, total pages: Dentsu Inc. Dentsuu koukoku nenkan ’07-‘08 (Dentsu advertising annual ’07-‘08), Table 5, p. 130. Subscription prices, circulation: Japan Audit Bureau of Circulation. shimbun hakkousha repooto (newspaper publisher report), 2007 (Jan to June average) and 2007 (July to December average).

dailies, include the regional “bloc” dailies, and include the local (one-prefecture) newspapers that offer their subscribers a choice between morning-only and morning-and-evening and that have circulation greater than 100,000 for each. I exclude the Nikkei Shimbun which is a business newspaper similar to the Wall Street Journal, not really a general news daily like the newspapers in the sample. A useful way of thinking about the pricing and content of morning and evening editions is with reference to the economics of second-degree price discrimination. Demanders, all confronted with the same nonlinear tariff schedule, self-select in a way that allows the monopolist to appropriate more of the gain from trade than would be possible with a linear tariff. In the case of newspaper subscription prices, the nonlinearity inheres in the fact that the tariff schedule is coarse. Subscribers must choose either morningonly or morning-and evening subscriptions. A linear tariff would entail each subscriber choosing a number of pages of content per day, at a set price per page. The newspaper subscription prices form a tariff schedule (price per month as a function of pages of content per month) with only two points; the tariff schedule is not an unbroken line and so it is nonlinear (even if the price per page of content is the same for morning-only subscriptions as for morning-and evening). This suggests two useful lines of inquiry. First, do newspaper publishers reduce the number of content pages of their morning editions to discourage dual subscribers from cancelling their evening subscriptions, in the same way that airlines reduce the quality of economy seating to discourage firstclass passengers from switching to economy? This is a phenomenon noted by the nineteenth century French economist Jules Dupuit (1804–1866) with respect to train fares. The Dupuit phenomenon is an example of monopoly waste in the sense that quality of the lower priced version of the product is reduced below the economically efficient level for the low-valuing customers, to conform to the incentive compatibility constraint of the higher valuing customers. This waste has been empirically verified and

estimated by McManus (2007) for coffee drink pricing. Shum and Crawford (2007) infer similar wasteful degradation in the quality of cable TV offerings. A second line of inquiry is whether comparisons of the demand for morning editions and for combined morning-and-evening editions reveal underlying parameters related to the elasticity of demand for newspapers with respect to prices and to pages of content. Within the simple framework set out above, I would suggest that these parameters are so revealed. To show why requires that we develop the micro-foundations of our demand system. 3.1. Micro-foundations of the constant-elasticity demand system In the basic framework of this paper, each household subscribes to at most one newspaper and has a reservation value of subscription to the particular newspaper under analysis, meaning the highest subscription price that would induce the household to subscribe to the particular newspaper—publisher and edition—and forego its own most preferred alternative. That reservation value reflects the prices, pages of content and characteristics of other newspapers—all of which I shall take as given, and which determine a unique preferred alternative for each household. The reservation value also reflects the personal preferences of the household and the pages of content and characteristics of the particular newspaper. Within this framework, I will assume that the characteristics of subscription other than number of pages of content per month inhere to the newspaper publisher. For example, these characteristics might reflect a political slant or the balance between coverage of local versus global news. Offering more pages of content will mean offering more pages with the same unchanging characteristics. Posit a population of households s ¼ 1; . . . ; n; each of which has a reservation marginal valuation of newspaper content per page (that is printed pages k not including advertising), for a particular

Please cite this article in press as: D. Flath, Second-degree price discrimination by Japanese newspapers, Japan World Econ. (2017), https://doi. org/10.1016/j.japwor.2017.10.003

G Model JAPWOR 903 No. of Pages 12

D. Flath / Japan and the World Economy xxx (2017) xxx–xxx

newspaper morning edition, as follows, taking as given the prices and contents of rival newspapers:

subscriber of the morning edition is

u 1  1 js u v ðsÞ ¼ Ajs s js k js ; js

ve ðsÞ

0

ð15Þ

kmZþke

¼

v0 ðkÞdk

km

u u1 B C ¼ Ajs s js @ðkm þ ke Þjs  km js A: 1

5

1

0

ð20Þ

where 0 < u < 1, and js > u. This means that for each household, marginal value of the one newspaper per page of content becomes less as the number of pages increases, but this relationship is inelastic   @lnv0 j u ¼ s <0 ð16Þ @lnk js > 1:

Denote by pe the price of subscription to the evening edition (so a joint subscription including both morning and evening editions is priced pm + pe). The demand for subscription to the evening edition (that is, se such that ve(se) = pe) is

It also means that the index s numbers households in rank order from the highest marginal valuer (s = 1) to the lowest (s ! 1):


 u u js se ¼ Ape js ðkm þ ke Þjs  km js :

@lnv0 1 ¼  < 0: @lns js

Notice that demand for evening subscriptions does not depend on the price of the morning edition, pm. A lower price of morningonly subscription would attract new morning-only subscribers who had before that subscribed only to the morning edition of a different newspaper or else subscribed to no newspaper, but would not induce any morning-and-evening subscribers to cancel their evening subscriptions. Increasing the pages of content in the morning edition would precipitate cancellations of evening

ð17Þ

The value to household s of the newspaper—its reservation subscription price—if the newspaper has km pages of content is Zkm vðsÞ

¼ 0

Zkm ¼ 0

v0 ðkÞdk 1 1 js  u u js js  js A s k dk js

ð18Þ

1 1 u  ¼ Ajs s js km js : And the demand for the newspaper if it has km pages of content is the number of households for which the value of subscription is greater than or equal to the subscription price pm. Because the index s numbers households in rank order from the highest valuer to the lowest, the number of households for which the value of the newspaper subscription is greater than or equal to its price equals the index identifying the household for which the value just equals price. That is, the demand for subscription to morning edition of the one newspaper if it has km pages of content is sm ¼ Apm js km u :

ð19Þ

Notice that this matches the Eq. (1) demand for subscriptions of the basic framework set out in the previous section. Based on Eq. (19) (same as Eq. (1)), as the newspaper publisher lowers its subscription price pm or increases its pages of content km, some households that, before then, had not subscribed to this particular newspaper, would now do so, each foregoing its own preferred alternative—a subscription to a different newspaper or no subscription to any newspaper. 3.2. Demand for subscriptions to an evening edition Next suppose that, again given the prices and contents of rival newspapers, subscribers to the morning edition of the one newspaper may also subscribe to an evening edition of the same newspaper. Within our framework (in which pages of content are the objects of valuation), the same parameter s that orders households in rank order from the highest valuer (s = 1) of subscription to the morning edition to the lowest valuer, also orders them from highest valuer to the lowest valuer of combined subscription to both morning and evening edition. This means that the next-best alternative of any subscribers to both the morningand-evening edition will be morning-only subscription to the same newspaper. Denote the numbers of pages of content in the two editions km and ke. The value of subscription to the evening edition to a



ð21Þ

subscriptions, @@ksme < 0.

The same parameters js and u appear in the demand for evening subscriptions, Eq. (21) and in the demand for morning subscriptions, Eq. (19). In fact, the price elasticity of demand js is the same for both. The elasticities of demand with respect to pages of content are different, though related to the same parameters. From Eqs. (19) and (21), the fraction of the subscribers to the morning edition who also subscribe to the evening edition is the following se ¼ sm



!js   ju s pm js km 1 : pe km þ ke

ð22Þ

Although this equation is nonlinear, it might be used as an equation for estimating the demand parameters. The basic idea is that variation in prices and pages of content of morning editions relative to evening editions of different newspapers will induce systematic variation in the fraction of morning subscribers who also subscribe to the evening edition—variation that reveals the demand elasticities js and u. This assumes that the different newspapers face similar demand elasticities. It also assumes that all consumers face the same price schedules, meaning that newspaper publishers or news dealers do not practice third-degree price discrimination—making different price offers to different customers. An anonymous referee has helpfully pointed out that in Japan this is not strictly true. Japanese newspaper publishers are specifically authorized to set the retail subscription prices charged by the independent news dealers through which they distribute newspapers.2 But by special directive of the Japan Fair Trade Commission, if a newspaper publisher does set a retail price, it must set the same price for every customer in every geographic region where that edition of the newspaper is sold.3 By the same directive, news dealers are enjoined against discounting below the prices set by the newspaper publishers. Virtually all Japanese newspaper publishers do stipulate prices for morning-and-evening subscriptions. And in

2 Newspaper publishers are specifically exempted from antimonopoly prohibitions against resale price maintenance (Antimonopoly, Section 23-2 (4)), by the 1953 amendments to the 1947 Antimonopoly Law. 3 “Specific Unfair Trade Practices in the Newspaper Business”, Fair Trade Commission Notification, no. 9, July 21, 1999, amending the similar earlier notification no. 14, 1964.

Please cite this article in press as: D. Flath, Second-degree price discrimination by Japanese newspapers, Japan World Econ. (2017), https://doi. org/10.1016/j.japwor.2017.10.003

G Model JAPWOR 903 No. of Pages 12

6

D. Flath / Japan and the World Economy xxx (2017) xxx–xxx

the regions where a publisher only offers a morning edition, it stipulates the morning-only subscription price. But in regions where a Japanese newspaper publisher does offer morning-andevening subscriptions, it generally avoids publicizing its price for a morning-only subscription. Roughly half of the subscribers to the newspapers in Table 1, in regions where a morning-and-evening subscription is available, actually subscribe to the morning edition only (based on the author’s calculations, using subscription data for cities, towns and villages from the Japan Audit Bureau of Circulation, 2007). An anonymous referee has pointed out, based on personal experience, that if a customer asks a news dealer to cancel an evening edition subscription and revert to a morningonly subscription, only a modest price reduction is offered— between 200 and 300 yen per month, compared to the 900 to 1000 yen per month difference between morning-and-evening subscription price and the morning-only subscription price where only a morning edition is offered. This is evidence of third-degree price discrimination. The news dealer—quite rationally—sets a higher morning-only subscription price for a morning-andevening subscriber who wishes to cancel the evening subscription than for a new subscriber who only wants the morning edition. (A customer having once subscribed to the morning-and-evening editions has revealed himself to be a relatively high valuer of a morning-only subscription, in accordance with the reasoning laid out in the previous paragraphs). I will assume that the preponderance of morning-only subscribers have never subscribed to an evening edition, and are paying the same subscription price set by the newspaper publisher in regions where only the morning edition is available. In other words, I will ignore the third-degree price discrimination in morning-only subscription prices that is indeed present. I think the third-degree price discrimination is infrequent, and in any case am unable to observe the discriminatory prices. With these caveats I proceed to empirical estimation of demand. 4. Estimates of demand function parameters I estimate the parameters js and u in Eq. (22) using nonlinear least squares. The software I used for these estimates is SAS (“proc nlin”), but Stata (“nl”) could also have been used. For details on nonlinear regression, refer to MacKinnon and Davidson (2004, Ch. 6). The data for the estimation are, first, those shown in Table 1, for which the unit of observation is the nation. The number of observations equal to 23 is simply the number of newspapers in the sample. The results of estimation using this dataset are in the lefthand column of Table 2. Based on the point estimates, the price elasticity of demand js is near 1.2, while elasticity of demand with respect to pages of content is around 0.5, values that seem plausible, but the standard errors are very large. Perhaps these estimates are biased because the national circulation includes morning subscribers in regions where the publisher does not offer evening subscriptions. To address this source of bias I compiled a second dataset that includes circulation of morning and evening editions of the same 23 newspapers compiled by summing the circulation only for cities, towns or villages for which the newspaper publisher offers all residents the option of subscribing to an evening edition. The total circulation of the 23 newspapers, calculated in this way, represents about 87 percent of the total evening circulation of the newspapers, and 66 percent of the morning circulation. The 13 percent of evening circulation so excluded is in cities and towns where some residents—but not all— if they wished, could subscribe to an evening edition. To make this determination I relied on details in Japan Audit Bureau of Circulation (2007). The source of the city and town circulation data is the Japan Audit Bureau of Circulation (from their website, behind a paywall).

Table 2 Nonlinear least squares regression. Estimating equation:

se sm

¼


js 
pm pe

km km þke

 u

js

js 1 þe

Parameter

National circulation for each edition of each newspaper, summed over all citiesa

National circulation for each edition of each newspaper, summed only over cities in which evening subscriptions are available to all residents

js

1.24 0.65 0.50 0.49 23 64.7

0.88 0.55 0.34 0.41 23 47.9

u

s.e. s.e.

n F a

National circulation, as shown in Table 1.

The results of estimation using this adjusted dataset are in the right-hand column of Table 2. Based on the point estimates, the price elasticity of demand js is near 0.88, while elasticity of demand with respect to pages of content is around 0.34, with standard errors only slightly smaller than for the estimates using the unadjusted national circulation data. These estimates are plausible. An ordinary monopoly will not set a price in the inelastic region of demand (negative marginal revenue) but a monopoly platform in a two-sided market might do so. For example, a newspaper might set a low subscription price to increase advertising revenue (and thereby add to profit) even though it means reduced subscription revenue. An alternative to the nonlinear estimation (as in Table 2), is to construct estimates of the demand for each newspaper from estimates of an indirect utility function using microdata. This I have done in another paper Flath (2016) that mainly focuses on the effects of resale price maintenance by Japanese newspapers. The estimates there are based on two alternative multinomial logit specifications of the indirect utility function. One of the specifications makes no adjustment for the possibility of omitted-variable bias. The other specification uses the control function approach of Petrin and Train (2009) to condition out variation in subscription price that is correlated with unobserved newspaper characteristics. In this specification, the indirect utility function includes a variable that is the error term from a “control function”, an OLS regression equation relating subscription price to factors related to the net-marginal cost of supplying subscriptions—whether the subscription is morning-and-evening, advertising revenue per subscriber and printed pages per month. The estimated elasticity parameters based on simulations using the indirect utility function estimates just described are shown in Table 3. These represent the arc elasticities of demand for the morning edition with respect to increases in the number of pages
 lnsm of morning content @@lnk ¼ u , and with respect to an increase in m
 lnsm ¼ js . The simuthe monthly morning subscription price @@lnp m

lations for computing these arc elasticities entail five percent increases in pages of morning content or five percent increase in price of the morning edition, for each newspaper individually. A comment on the simulations used for computing the arc elasticities shown in Table 3 may be useful. There are alternative simulations for computing elasticities but the ones I used seem the most reliable. The logic for computing arc elasticity with respect to pages of morning content is straightforward, but for morning subscription price matters are a bit subtle. Arc elasticities based on change in the morning subscription price holding morning-and-

Please cite this article in press as: D. Flath, Second-degree price discrimination by Japanese newspapers, Japan World Econ. (2017), https://doi. org/10.1016/j.japwor.2017.10.003

G Model JAPWOR 903 No. of Pages 12

D. Flath / Japan and the World Economy xxx (2017) xxx–xxx

7

Table 3 Arc elasticities of demand for the morning edition with respect to pages of content per month u and with respect to subscription price js, based on Flath (2016) estimates of indirect utility function. and values of g m, Dkkmm , Newspaper

0

km ðkm þke Þ

and

Dp

p1sm s1m

implied by these elasticities.

From indirect utility estimate; g m From indirect utility estimate; control function specification* (est w/o control function* (est w/ w) o)

D km

0

predicted km  ðkm þ ke Þ

km

Dp

p1sm s1m

u

js

u

js

est w

est w/o est w

est w/o est w

est w/o

est w

est w/o

THE ASAHI SHIMBUN MAINICHI SHIMBUN YOMIURI SHIMBUN SANKEI SHIMBUN HOKKAIDO SHIMBUN THE KAHOKU SHIMPO THE TOKYO SHIMBUN THE CHUNICHI S. KYOTO SHIMBUN KOBE SHIMBUN THE CHUGOKU S. THE NISHINIPPON S.

0.41 0.44 0.41 0.46 0.20 0.29 0.37 0.40 0.35 0.35 0.28 0.36

1.45 1.46 1.53 1.26 0.75 0.72 1.27 1.50 1.27 1.33 1.16 1.38

0.34 0.38 0.34 0.39 0.19 0.22 0.3 0.33 0.29 0.27 0.27 0.27

1.13 1.22 1.17 1.01 0.53 0.60 0.93 1.14 1.00 1.00 0.87 1.07

0.294 0.225 0.272 0.416 0.343 0.091 0.271 0.167 0.441 0.307 0.083 0.151

0.269 0.212 0.245 0.391 0.208 0.104 0.237 0.148 0.451 0.325 0.067 0.157

0.271 0.208 0.242 0.387 0.219 0.110 0.242 0.144 0.470 0.335 0.068 0.155

1.04 0.95 0.93 1.18 1.10 0.90 1.00 0.87 1.41 1.10 0.87 0.89

1.02 0.94 0.92 1.17 0.90 0.92 0.98 0.87 1.46 1.14 0.85 0.91

0.066 0.037 0.021 0.072 0.195 0.048 0.083 0.026 0.126 0.084 0.017 0.026

0.086 0.049 0.036 0.090 0.255 0.075 0.111 0.033 0.156 0.115 0.012 0.042

MEAN S.D.

0.36 0.07

1.26 0.27

0.30 0.06

0.97 0.22

0.237 0.218 0.106 0.100

0.248 0.238 0.119 0.116

1.02 0.16

1.01 0.17

0.067 0.088 0.052 0.067

0.281 0.211 0.251 0.388 0.361 0.095 0.258 0.146 0.453 0.308 0.081 0.138

*Source of these estimates: Flath (2016).

evening subscription price constant, rather than holding evening subscription price constant, is the correct choice here for two reasons. First, the effect of a change in price of the evening edition pe is difficult to interpret because of the apparent policy of some newspapers offering only morning-and-evening subscriptions in some localities. The model of this paper presumes that all subscribers have a choice between morning-only and morningand-evening. Thus, focus on the effect of a change in price of morning edition (rather than evening edition) is likely to produce a more precise measure of elasticity that corresponds to the elasticity of the model. Second, the model (in which value of subscription is the same as value of content pages) implies a nesting of choices. The consumer first selects a morning subscription from among those available, and then decides whether to also subscribe to the evening edition of that same newspaper. The response to a change in price of the morning edition fully captures substitution across newspapers. From the indirect utility function estimates as just described (and reported in Table 3), it seems that the elasticity of demand for morning edition with respect to pages of morning content is between 0.3 and 0.4, and that the elasticity with respect to morning subscription price averages is between 1.0 and 1.3, with elasticities generally a bit higher for the national newspapers than for the local ones. The estimated elasticities based on the specification with control function are about one-fourth higher than those based on the simpler specification, corresponding to the upper and lower ends of the ranges just stated. It is not obvious which specification is better, because the attempt to control for possible omittedvariable bias using the control function approach may itself introduce new bias. This is why a new method for computing demand elasticity, not subject to the same bias as the other methods, is valuable. The elasticities based on the indirect utility specification with control function are close to the nonlinear least squares estimates using national circulation data as reported in Table 2. The elasticities based on the indirect-utility specification without control function are close to the nonlinear least squares estimates using the circulation aggregated over cities with full evening coverage as reported in Table 2. Sharp conclusions are not apparent, but the range of possibilities is nevertheless sufficiently narrow for our next task. That task is to model the newspaper publisher’s choices of subscription prices and pages of content if offering both a morning edition and evening edition.

5. Pricing and content choices by newspaper publishers offering morning and evening editions Extending the basic framework to accommodate pricing and content choices by newspaper publishers offering both morning and evening editions requires some additional assumptions. First consider the demand for advertising. The incremental cost of supplying ads per subscriber is inversely related to number of subscribers. As detailed above, and true in fact, the evening subscribers are a subset of the morning subscribers. Why then would the newspaper even want to offer advertising in the evening edition? Or to put it a different way, if as seems likely given the higher incremental cost, the ad price per subscriber is greater for ads in the evening edition than in the morning edition, then what advertiser would place an ad in the evening edition? Perhaps advertisers who wish to specifically target evening subscribers would do so. The evening subscribers have a greater demand for newspaper content and so plausibly are more educated and have higher incomes, which might make them valuable targets for some advertisers. In the logic of the very general framework developed by Veiga and Weyl (2012), enabling targeted advertising could be a sufficient motivation for offering an evening edition in the first place, not mutually exclusive from the motivation of profiting from second-degree price discrimination of subscriptions. Let us assume then that the demand for ads in the morning edition is independent of the demand for ads in the evening edition. Now, given the number of pages of content in both morning and evening editions km and ke, and assuming the demands for ads in the two editions are independent of one another, the profit-maximizing pricing rules are edition-specific variants of the rules derived in the basic framework4 : pam  cam 1 ¼ pam ja m

ð23Þ

4 Perhaps the notation here is obvious enough from analogy with the singleedition case shown by Eqs. (3)–(10). So for example, the profit function for the dualedition newspaper becomes p ¼ psm sm  ðc ðam þ km Þ þ c0m Þsm þ ðpam  f am Þam  f km km þ pse se  ðc ðae þ ke Þ þ c0e Þse þ ðpae  f ae Þae  f ke ke ; and the incremental costs of morning-edition subscriptions and morning content are csm ¼ @@Cost ¼ c0m þ cðam þ km Þ, and ckm ¼ @@Cost ¼ f km þ csm , and so on. sm km

Please cite this article in press as: D. Flath, Second-degree price discrimination by Japanese newspapers, Japan World Econ. (2017), https://doi. org/10.1016/j.japwor.2017.10.003

G Model JAPWOR 903 No. of Pages 12

8

D. Flath / Japan and the World Economy xxx (2017) xxx–xxx

Fig. 1. The figure shows the demand for morning subscriptions and, in an unconventional way, also shows the demand for evening subscriptions, given the pages of content in each edition. A small change in the price of either type of subscription has no effect on the demand for the other. For the marginal subscriber to the morning edition, s1m , the price of an evening subscription is greater than willingness to pay. He does not subscribe to the evening edition and small changes in the price of an evening subscription will not change that —the dashed line will remain above the other lines even if it were

 1 1 1 shifted downward a little. To put it another way, p1se > v s1m ; km þ ke  v s1m ; km , even if the price of evening subscription is reduced.
 1 For the marginal subscriber to the evening edition, s1e , willingness to pay for the morning edition is already greater than the price, v s1e ; km > p1sm , so a small change in price of the morning subscription does not affect her choice—she already subscribes to the morning edition and will not cancel (which would also mean cancelling the evening edition subscription), even if the price is increased a bit.

p sm  csm 1 p am ¼  am psm js psm sm

ð24Þ

pae  cae 1 ¼ pae jae

ð25Þ

p se  cse 1 p ae ¼  ae pse js pse se

ð26Þ

Notice that in the framework here, the demands for subscriptions to morning edition and to evening edition are independent—the two editions are neither substitutes in demand nor complements in demand. This is evident from the two demand functions, Eqs. (19) and (21). A change in price psm does not affect the demand for evening subscriptions, and a change in pse does not affect the demand for morning subscriptions. The intuition behind this is represented in Fig. 1. The figure shows the demand for morning subscriptions and, in an unconventional way, also shows

the demand for evening subscriptions, given the pages of content in each edition. In the figure, a small change in the price of either type of subscription has no effect on the demand for the other. For the marginal subscriber to the morning edition, the price of an evening subscription is greater than willingness to pay. He does not subscribe to the evening edition and small changes in the price of an evening subscription will not change that —the dashed line will remain above the other lines even if it were shifted downward a

 1 1 1 little. To put it another way, p1se > v s1m ; km þ ke  v s1m ; km , even if the price of evening subscription is reduced. For the marginal subscriber to the evening edition, willingness to pay for the
 1 morning edition is already greater than the price, v s1e ; km > p1sm , so a small change in price of the morning subscription does not affect her choice—she already subscribes to the morning edition and will not cancel (which would also mean cancelling the evening edition subscription), even if the price is increased a bit. The rules for profit-maximizing content are a bit different than for prices, reflecting the interdependence in demand for subscriptions with respect to pages of content in each edition. The interdependence entails an incentive-compatibility constraint on

Please cite this article in press as: D. Flath, Second-degree price discrimination by Japanese newspapers, Japan World Econ. (2017), https://doi. org/10.1016/j.japwor.2017.10.003

G Model JAPWOR 903 No. of Pages 12

D. Flath / Japan and the World Economy xxx (2017) xxx–xxx

9

the newspaper publisher’s choices of pages of content. The newspaper publisher adjusts pages of content mindful that reducing the content of the evening edition or expanding the content of the morning edition will cause cancellations of evening subscriptions. The profit-maximizing content rules are:    ckm km u 1 pse se @lnse ¼ þ ð27Þ ; psm sm js js psm sm @lnkm

that

where

publisher to decrease pkm smm by the factor g m and increase pke see by the

ckm km u ¼ ; psm sm js and cke ke u ¼ : pse se js

ð34Þ

The incentive-compatibility constraint leads the newspaper c

0


B @lnse ¼ uB @ @lnkm

km km þke

u 1 js m  kmkþk C e C < 0; if 0 < u < 1;
u A j j



1




km km þke

ð28Þ

s

s


 @lnse cke ke @lnke ¼ ; pse se js

ð29Þ

where 0 B @ln se ¼u B @ @ln ke


1

ke km þke






km km þke

1 C u C  u A > u ; if 0 < j < 1: j

ð30Þ

The second term on the right-hand side of Eq. (27) reflects the incentive-compatibility constraint. The upshot is that the newspaper publisher reduces the content of the morning edition compared to what it would choose if not offering an evening edition. By how much? A precise answer requires some further exploration. First, by rearranging Eqs. (27)–(30), notice that   ckm km c ke u u pse se ¼ þ ð270Þ  ke ; psm sm js js psm sm psm sm

cke ke u u psm sm ¼ þ pse se js js pse se

ask, then, how large is the Dupuit effect, is the same as asking how much content a publisher that only offered a morning edition would shift to an evening edition, if it were to offer one. The content shifted from the old morning edition equals the Dupuit effect. The starting point in answering this question is to relate the parameter, g m, to the percentage reduction in pages of morning content if an evening edition is offered, Dkkmm . After some tedious calculations explained in the appendix, the relation turns out to be:


D cpkm ksm g m ¼
c smkmm km

 ¼

 

ckm km pse se

ð280Þ

Thus Eqs. (27) and (29) can be expressed as     ckm km u u pse se j c ke ¼ þ 1  s ke psm sm u pse se js js psm sm ¼

u ð1  g m Þ; js 

u u psm sm þ js js pse se u ¼ ð1 þ g e Þ: js





1

js ckm km u psm sm

    ckm f k =sm þ cðjs  1Þ f k =sm Dkm : u f k =sm þ c psm f k =sm þ c km

ð35Þ

Here, Dxx  x xx 1 , where superscript “1” denotes a choice given that an evening edition is offered, and superscript “0” denotes a choice 0

1

1

km 1þDkkm

.

m

ð2700Þ

 ð2900Þ

ð31Þ

Also, from Eqs. (2700 ) and (2900 ),

g m pse se ¼ : g e p sm s m

1 þ js

0

Furthermore, from Eqs. (270 ) and (290 ), ckm km þ cke ke u ¼ : psm sm þ pse se js

ps m s m

given that only a morning edition is offered. Thus, km ¼


and ¼

se

factor g e. This suggests that the incentive-compatibility constraint causes the newspaper publisher to shift pages of content from the evening edition to the morning edition. How large is this “Dupuit” effect? First, notice that the newspaper publisher’s choices of pages of morning content, km, and morning subscription price, psm , if not offering an evening edition, exactly match what the newspaper publisher would choose if it were to offer an evening edition but myopically ignore the incentive-compatibility constraint—for example, in the mistaken belief that its morning edition content

and 

c k

lnse had no effect on the demand for evening subscriptions, @@lnk ¼ 0. To m

s

s

k

sm

and

cke ke pse se

ð33Þ

The variables not expressed as differentials are as actually observed, and so presume that an evening edition is offered—For these variables, superscript “1” is to be understood though it is not written. In the equation, ckm km ¼ ðf k þ csm Þkm is the cost of morning content, where fk is the first-copy cost of a page of content and c is the cost-per-page-per-copy of actually printing the newspaper, estimated by Flath (2016) to be one-yen: c ¼ 1: The Table 3 reports estimates of the parameter, g m, constructed from the observed pages of content and subscription revenues, and using the elasticities u and js (of morning subscriptions with respect to pages of morning content and with respect to morning subscription price) estimated in Flath (2016), for each newspaper publisher.5 The table also reports the estimated percentage reduction in pages of morning content implied by these estimates

ð32Þ 5

Now, suppose that the newspaper publisher were to ignore the incentive compatibility constraint and choose content for its morning edition without regard to the cross-effect on demand for subscriptions to its evening edition. Then, as evident from Eqs. (2700 ) and (2900 ), the publisher would set prices and content such




In

pse se psm sm



making these calculations, 0

u 1 B @

km km þke




1



km km þke

km km þke

ju

js

I

used

from

Eq.

(28),

g m ¼

C A:

s

Please cite this article in press as: D. Flath, Second-degree price discrimination by Japanese newspapers, Japan World Econ. (2017), https://doi. org/10.1016/j.japwor.2017.10.003

G Model JAPWOR 903 No. of Pages 12

10

D. Flath / Japan and the World Economy xxx (2017) xxx–xxx

Fig. 2. The figure shows the effect on newspaper revenue and on the supply of each type of subscription, if the publisher of a single morning edition were to introduce an evening edition and offer its subscribers the choice between morning-only or morning-and-evening subscriptions. The newspaper publisher will switch content from the morning edition to the new, evening edition, and lower its morning edition subscription price. Some incumbent subscribers, s  s1e , will choose morning-and-evening subscriptions, while others, s1e < s  s0m , will choose morning-only subscriptions. The nowlower price and content of the morning edition will also attract some new subscribers, s0m < s  s1m .

of g m, if an evening edition is offered, based on Eq. (35), using the cost parameter estimate c ¼ 1, from Flath (2016), to impute fk/sm.6 The data in Table 3 show that if any of the newspaper publishers in the sample were to cease publication of its evening edition, it would expand the content pages of its morning edition. As shown in the table, on average across newspaper publishers, the pages in the morning edition are about one-fourth fewer than would be included if the evening edition were suspended, g m ffi Dkkmm ffi 1=4, meaning a reduction in morning content pages that is nearly equal to the number of content pages currently included in the evening edition: predicted km 
0

1

km Dkm

1þ k

 ffi k1m þ k1e . So the answer to the

m

question that motivated this computation (How large is any shorting of content in the morning edition that is merely to deter subscribers from cancelling their evening subscriptions?) is that it is equal to the entire content of the evening edition. The Dupuit effect is not an incidental consequence of offering an evening edition, but is the very reason for offering an evening edition.

Fig. 2 illustrates the effect on newspaper revenue from each type of subscription if the publisher of a single morning edition were to introduce an evening edition and offer its subscribers a choice between morning-only or morning-and-evening subscriptions. By breaking the morning edition into a smaller morning edition and a separate evening edition, the newspaper publisher can continue to extract payment for the evening content from those most willing to bear it, while perhaps adding to its profit by attracting enough new subscribers to a reduced-content-butlower-priced morning edition to offset the losses caused by lowervaluing incumbent customers declining to take the evening edition but continuing to subscribe to the (now lower-priced) morning edition. Ponder for a moment the choice by a Japanese newspaper publisher of whether to offer an evening edition at all. And based on our empirical findings so far, let us presume that if it does offer an evening edition, it will reduce the pages of content in its morning edition by exactly the number of pages of content in the 0 1 k k evening edition: Dkkmm ¼ m 1 m ¼

km

1

ke 1 . km

Based on Eqs. (23)-(26), the profit of each newspaper company 



From Eq. (27 ), ju 1  g m psm sm ¼ ckm km ¼ ðf k þ csm Þkm , which is equivalent s  p to ju 1  g m ksmm  c ¼ f k =sm . Based on this logic and c ¼ 1, Table 3 estimates s 
 p of Dkkmm presume that f k =sm ¼ max 0; ju 1  g m ksm:m  1 . s 6

00

is

p¼

psm sm

js

þ

pse se

js

 f a ðam þ ae Þ  f k ðkm þ ke Þ:

ð37Þ

Please cite this article in press as: D. Flath, Second-degree price discrimination by Japanese newspapers, Japan World Econ. (2017), https://doi. org/10.1016/j.japwor.2017.10.003

G Model JAPWOR 903 No. of Pages 12

D. Flath / Japan and the World Economy xxx (2017) xxx–xxx

The profitability of offering an evening edition is most likely to be positive if the fall in revenue from morning subscriptions (the net effect of lower subscription price and fewer pages of morning content) is less than the revenue from evening subscriptions. In terms of the model, and assuming, based on the empirical findings, that offering an evening edition means offering the same total number of pages of content as before, attracting the same total advertising as before, the added profit if offering an evening edition is

Dp ¼ p1  p0 ¼

1

js

ðp1sm s1m þ p1se s1e  p0sm s0m Þ > 0;

if p0sm s0m  p1sm s1m < p1se s1e :

:

ð38Þ

Notice that the condition p0sm s0m  p1sm s1m < p1se s1e in Eq. (38) is equivalent to p0sm s0m  p1sm s1m p1 s1 g < 1se 1e ¼ m ; 1 1 ge psm sm psm sm

ð39Þ

where the last equality is from Eq. (32). From this, the condition for profitability of offering an evening edition can be related to parameters. The percent loss in morning subscription revenue if offering an evening edition is   p0sm s0m  p1sm s1m Dpsm Dsm Dpsm Dpsm Dkm ¼ þ ¼ þ u  j s psm sm psm km psm p1sm s1m  1 1  1 km ke 1 1 ke þu 1 ¼ ð1  js Þc 1 1 1  psm k js km !m 1 1 km ke ¼ u  js c 1 : psm k1 m

ð40Þ The change in profit if offering an evening subscription becomes

Dp ¼ p1  p0 ¼

1

ðp1sm s1m þ p1se s1e  p0sm s0m Þ ! ! p1sm s1m p1se s1e p0sm s0m  p1sm s1m  ¼ p 1 s1 p1sm s1m js ! sm m ! ! 1 1 1 1 1 1 psm sm pse se km ke ¼  u  j c : s js p1sm s1m p1sm k1

js

ð41Þ

m

The right-most columns of Table 3 report values of pDp 1 s1 for each sm m

newspaper publisher, again based on estimates of js and u (with and without instrumental variables) from Flath (2016), and on c=one-yen-per-page-per-copy, also as estimated in Flath (2016). From the estimates in Table 3, the change in profit from offering an evening edition generally falls in the range of 7 or 9 percent of morning subscription revenue, a fairly substantial payoff from second-degree price discrimination, if true. The effects on consumer surplus are also of interest. The change in consumer surplus induced by the offering of an evening edition is

DCS ¼ CS1  CS0 ffi

 1
1 1 p s þ p1se s1e  p0sm s0m ; 2 js s m m

ð42Þ

where the approximation assumes linear demand functions with point elasticity of demand js in the neighborhood of the prevailing prices and quantities. The linear demand function actually makes more sense than constant-elasticity-demand if elasticity of demand is less than one (which would imply infinite consumer surplus if elasticity of demand was constant). The change in consumer surplus is half as great as the change in profit—so equal to about 4 percent of morning subscription revenue. Consider how the change in consumer surplus from offering an evening edition is distributed across households. Offering an evening edition means incurring unit costs of distribution, c0, twice

11

for each morning-and-evening edition subscriber, rather than once, as before, and garnering additional advertising revenue per evening subscriber, paseae . This will be reflected in the evening

1 . Let us subscription price: p1sm þ p1se ¼ p0sm þ c0e  paseae 1  j1 s

consider two cases. First, suppose that the added unit cost of distribution net of per-subscriber ad revenue is small, c0e  paseae ffi 0. Then, because the morning-and-evening subscribers obtain the same total pages of content as in the previous morningonly edition, and pay nearly the same total subscription price as they did before, p1sm þ p1se ffi p0sm , households that subscribe to the evening edition obtain about the same consumer surplus as before. It follows that households that, before, had subscribed to the morning edition and who choose not to subscribe to the evening edition, are, by revealed preference, better off than before. Their consumer surplus is increased, though they now obtain fewer pages of content and pay a lower subscription price than before. Similarly, the new subscribers to the now lower-priced morning edition are, by revealed preference, better off than before, with more consumer surplus. Offering an evening edition is therefore likely to increase total consumer surplus. It is only unprofitable to offer an evening edition—and harmful to consumer surplus—if, contrary to the above discussion, the added cost of supplying an evening edition net of per-subscriber ad revenue induces a substantial increase in subscription price for the new morning-and-evening subscribers, p1sm þ p1se  p0sm ¼

1 1  j1 c0e  paseae > 0. Consider then a second case, in which s

the added unit cost of distribution net of per-subscriber ad revenue if offering an evening edition is large, c0e  paseae > 0, and is reflected in the price of evening subscription. This will cause the morningand-evening subscribers to pay more than they did before an evening edition was offered, though they receive the same content as before, lowering their consumer surplus and also lowering the consumer surplus of some incumbent subscribers who choose morning-only subscriptions to avoid this added burden. Any new morning-only subscribers are still better off than before, by revealed preference. Based on the theoretical framework and empirical findings of this paper, the incentive of each Japanese newspaper publisher in deciding whether to offer an evening edition is aligned with maximum social welfare: If it is profitable to offer an evening edition then it increases total consumer surplus to do so, and if unprofitable it decreases total consumer surplus. 6. Conclusion Many of the large circulation Japanese newspapers offer demanders a choice between subscribing to the morning edition only, or to both the morning edition and evening edition. This is an example of price discrimination based on self-selection by demanders (second-degree price discrimination). The subscription price schedule amounts to a nonlinear tariff, gauged to maximize newspaper profit. The possibility of engaging in such profitable price discrimination is one reason even to offer an evening edition in the first place, which seems not to have been previously remarked. The possibility of targeting advertising on higher income subscribers who are more likely to subscribe to an evening edition along with the morning edition is another reason to offer an evening edition. Subscriber preference for more frequent news is yet another reason. As newspapers move to digital rather than print format, some analogue of the “evening edition” is likely to be profitable for all of the reasons that apply to the print version. Here I have argued that not only the subscription prices but also the number of pages of content in each edition are gauged to

Please cite this article in press as: D. Flath, Second-degree price discrimination by Japanese newspapers, Japan World Econ. (2017), https://doi. org/10.1016/j.japwor.2017.10.003

G Model JAPWOR 903 No. of Pages 12

12

D. Flath / Japan and the World Economy xxx (2017) xxx–xxx

maximize newspaper profit subject to the constraint that subscribers rationally maximize utility. The Japanese newspaper publishers that offer the subscribers to their morning editions the option of also subscribing to an evening edition seem to have added to their own profit by about eight percent of their morning subscription revenue and have increased consumer surplus by about half as much as that. The analysis supporting these claims is based on an algebraic model calibrated with econometric estimates of the demand elasticities and observed subscription revenues of each newspaper publisher. In this setting, a Japanese newspaper publisher that offers its morning subscribers the option of also subscribing to an evening edition will shift pages of content from the morning edition to the evening edition, and at the same time reduce the price of a morning subscription. In the calibrated framework of this paper, this second-degree price discrimination—for that is what it is—will, if profitable, benefit low-valuing subscribers more than it harms high-valuing subscribers, and so increase social welfare. It is worth asking whether the framework of this paper could be extended to other examples of second-degree price discrimination in oligopolies. The most ubiquitous such example is the pricing of airfares. Business-class seating is the analogue of morning-andevening newspaper subscription, and economy-class seating is the analogue of morning-only newspaper subscription. If an index of amenities attached to each class of seating could be constructed and used as the analogue of pages of content for each type of newspaper subscription, the model of this paper could be usefully applied to airlines. The direct analogue for airlines of pages of content for newspapers, is the space taken up by a seat—greater for business class than for economy class. The airfare analogue of price of evening subscription is the difference in fares between businessclass and economy-class seats. The substitution pattern that the model entails would mean that lowering the price of economyclass seats while also lowering that of business-class seats by the same amount, would attract more low-valuing passengers without causing any of the higher-valuing passengers to switch from business class to economy class. The model could be used to estimate elasticity of demand for seats with respect to price and with respect to amenities in the same way I used it to estimate demand parameters for newspaper subscriptions. Funding This research is supported by Japan Society for the Promotion of Science, Grant-in-Aid for Scientific Research (C), grant no. 22530227. Appendix A. Derivation of Eq. (35) By definition,


 D cpkm ks m D ckm km Dpsm Dsm g m ¼
c smkmm ¼   ; km

psm sm

ckm km

psm

sm

ðA1Þ

where ckm km ¼ ðf k þ csm Þkm ;

sm ¼ Apm js km u ;

ðA3Þ

and  
p a 1 1 1 psm ¼ c0 þ ca þ ckm  a : s js Differentiating Eqs. (A2)–(A4) yields the following.  D ckm km Dkm c Dsm ¼ þ ckm km km ðf k =sm þ cÞ sm

Ds m sm

Dpsm psm

¼u

¼

Dkm km

 js

Dpsm psm

  ckm 1 1 Dkm 1 psm km js

ðA4Þ

ðA5Þ

ðA6Þ

ðA7Þ

By substitution of Eqs. (A5)–(A7) into Eq. (A1), we obtain the result below.


D cpkm ks m g m ¼
c smkmm km

p

s

   sm m   ckm f k =sm þ cðjs  1Þ f k =sm Dkm : u ¼ 1 þ js f k =sm þ c psm f k =sm þ c km

ð35Þ

References Angelucci, Charles, Cage, Julia, de Nijs, Rpmain, 2013. Price Discrimination in a TwoSided Market: Theory and Evidence from the Newspaper Industry, NET, Institute, Working Paper #13-13. . Argentesi, E., Filistrucchi, L., 2007. Estimating market power in a two-Sided market: the case of newspapers. J. Appl. Econ. 22, 1247–1266. Busse, Meghan, Rysman, Marc, 2005. Competition and price discrimination in yellow pages advertising. Rand J. Econ. 36 (Summer (2)), 378–390. Cohen, Andrew, 2008. Package Size and Price Discrimination in the Paper Towel MarketInternational Journal of Industrial Organization. Int. J. Ind Organiz 26, 502–516. Fan, Ying, 2013. Ownership consolidation and product characteristics: a study of the US daily newspaper market. Am. Econ. Rev. 103 (5), 1598–1628. Flath, David, 2016. Resale price maintenance by japanese newspapers. Jpn. Econ. Rev. 67 (4), 441–473. Japan Audit Bureau of Circulation, 2007. Shinbun Hakko-sha Repoto, Hanki (Newspaper Publisher Report, Semiannnual). JABC, Tokyo (second half). MacKinnon, James G., Davidson, Russel, 2004. Econometric Theory and Methods. Oxford University Press. McManus, Brian, 2007. Nonlinear pricing in an oligopoly market: the case of specialty coffee. Rand J. Econ. 38, 512–532. Petrin, Amil, Train, Kenneth, 2009. A control function approach to endogeneity in consumer choice models. J. Market. Res. 46 (1), 3–13. Shum, Matthew, Crawford, Gregory S., 2007. Monopoly product degradation in cable television. J. Law Econ. 50 (February (1)), 181–209. Tirole, Jean, 1988. Theory of Industrial Organization. MIT Press. Van Cayseele, P., Vanormelingen, S., 2009. Prices and Network Effects in Two-Sided Markets: the Belgian Newspaper Industry. (unpublished manuscript). Veiga, André, Weyl, Glen, 2012. Multidimensional Product Design. (unpublished manuscript). Video Research Ltd, 2007. Japan Readers and Area Data (JREAD) Survey Conducted in March 2007. .

ðA2Þ

Please cite this article in press as: D. Flath, Second-degree price discrimination by Japanese newspapers, Japan World Econ. (2017), https://doi. org/10.1016/j.japwor.2017.10.003