THE IMPACT OF TECHNOLOGY ON THE PRODUCTION OF INFORMATION

Transcription

1 THE IMPACT OF TECHNOLOGY ON THE PRODUCTION OF INFORMATION Adt Mukherjee PhD Program Krannert Graduate School of Management, Purdue Unversty West Lafayette, IN Emal: Jungpl Hahn Assstant Professor of Management Krannert Graduate School of Management, Purdue Unversty West Lafayette, IN Emal: Last revsed: October 31, 2006 ABSTRACT When new technologes for producton and dssemnaton of nformaton emerge and are adopted, we see a dramatc ncrease n the quantty of nformaton that s made avalable for consumpton. In recent years, wth the wde adopton of dgtal meda, ths ncrease has been dubbed the nformaton overload, whch s perceved by some to be a boon and by others to be a curse. The mplcatons of ths nformaton overload on the qualty of nformaton that s avalable for consumpton s unclear. By quantfyng the changes n the qualty of nformaton that s beng produced due to the adopton of new technologes, we wll be able to dentfy those aspects of the technology that result n the ncrease n the qualty of nformaton avalable for consumpton. These nsghts can therefore be exploted to develop ncentve mechansms and polces to further ncrease the qualty of nformaton beng produced. In ths paper we develop a framework for the process of nformaton producton and use t as the bass for an economc model that captures a proft maxmzng entty s decson to produce nformaton. Ths economc model s used to study the mpact of a new technology on the nformaton that s produced n terms of both the qualty and the quantty of nformaton. Our results show that reducton n costs leads to a margnal ncrease n both the qualty and the quantty of nformaton produced by a proft maxmzng entty, whereas the reducton n space and tme constrants results n a sgnfcant ncrease n the quantty and a margnal decrease n the qualty of nformaton. We also fnd that as the producton costs decrease, the overall quantty of hgh qualty nformaton avalable for consumpton ncreases more rapdly than the overall quantty of low qualty nformaton and the reverse effect occurs when the space and tme constrants are decreased. Keywords: Informaton producton, nformaton overload, nformaton qualty, economc model, smulatons Acknowledgments: an earler verson of ths paper was presented at the MIS Research Workshop at Purdue Unversty and at the 2006 Bg Ten IS Research Symposum at the Unversty of Illnos at Urbana-Champagn. We thank the partcpants at these workshops for helpful feedback at the early stages n the development of ths research. 0

2 What nformaton consumes s rather obvous: t consumes the attenton of ts recpents. Hence a wealth of nformaton creates a poverty of attenton, and a need to allocate that attenton effcently among the overabundance of nformaton sources that mght consume t 1 Herbert A. Smon, Nobel Laureate 1. INTRODUCTION One of the most profound consequences of the dgtal revoluton s the exponental growth n the amount of nformaton that has become readly avalable and accessble to end consumers (Lyman and Varan 2003). One of the factors responsble for ths s the advances n how nformaton processng, storage and communcaton technologes have sgnfcantly changed the way nformaton s produced and dssemnated. For example, the tradtonal medum used by publshers of encyclopedas was numerous volumes of books. However, the emergence of CD ROM technology resulted n the replacement of the tradtonal medum (.e., prnted books) by dgtal meda such as the CD ROM due to a drastc reducton n the costs of producng as well as the prce for purchasng the nformaton goods 2, whle at the same tme smultaneously ncreasng the rchness of the content (wth audo and vdeo multmeda content n addton to text and pctures). The evoluton of the encyclopeda does not end here. Once the Internet was adopted as the prmary medum by leadng encyclopeda producers, new encyclopedas emerged that used the Internet as the prmary medum. Ths was followed by the creators to further create systems that use the new technology. For example, novel technologes such as the Wk have enabled the creaton of Wkpeda ( that s wrtten, edted and mantaned collaboratvely n real tme by anyone wth access to a computer, web browser and Internet connecton (Wkpeda Contrbutors 2006). 1 As quoted by Hal Varan (1995). 2 the costs of producng encyclopedas dropped from $ to $1.50 for the publsher and the prce dropped from $1500-$2200 to $50 for the consumer (Evans and Wurster 1997) 1 1

3 Ths new mode of content creaton s drastcally dfferent from tradtonal modes of creatng and publshng content for an encyclopeda. As a pont of comparson, the current verson of the Encyclopæda Brtannca was wrtten by around 4,000 authortatve contrbutors whereas the Wkpeda has over 1.37 mllon regstered user accounts. 3 Smlarly, the Brtannca Ultmate Reference Sute 2006 DVD, whch s a combnes artcles of the regular Encyclopæda Brtannca, the Brtannca Student Encyclopeda, the Brtannca Elementary Encyclopeda and the Brtannca Book of the Year, has just over 100,000 artcles, whereas Wkpeda currently has over 1.1 mllon artcles n the Englsh language edton alone. 4 Whle one may argue about the relatve qualty of the artcles n these dsparate encyclopedas 5, one thng s certan there s at least an order of magntude dfference n the number of artcles between the old and the new meda. Whle, some may argue that the growth n the amount of avalable nformaton s a beneft to socety (.e., the more the better perspectve), we also need to consder the negatve consequences of the ongong phenomenon, snce the negatve consequences may potentally outwegh the benefts.(schultze and Ledner 2002) Consder the scenaros llustrated n Fgure 1, whch depct changes n the dstrbuton of the quantty of nformaton n terms of the qualty of nformaton, when there s some (fxed) ncrease n the quantty of nformaton. The curve at T 0 and T 1 represents the dstrbuton of nformaton content before and after the adopton of new technology respectvely. The area between the two curves (T 0 and T 1 ) represents the overall ncrease n quantty of nformaton, and has the same magntude n all three graphs. In the worst case scenaro (Fgure 1a), the ncrease n the quantty of lower content s relatvely greater than the ncrease n the quantty of hgher qualty content, and consequently the sgnal-to-nose rato worsens snce fndng hgh qualty content amdst the greater quantty of low qualty content s more 3 Obvously, not all 1.37 mllon users are actve contrbutors. In a recent statstc, the Wkpeda notes approxmately 27,000 users who made at least fve edts, and 4,000 users who made over 100 edts, just durng the month of December The total number of Wkpeda artcles for all languages s approxmately 3.8 mllon. 5 Although the Wkpeda and the Encylopæda Brtannca are dffcult to compare drectly, a recent artcle n Nature suggested that at least for scence-related artcles the accuracy of artcles n Wkpeda are comparable to those n the Brtannca (Gles 2005). 2 2

4 dffcult. When there s a unform ncrease n the quantty for all levels of qualty (Fgure 1b), the sgnal to nose rato remans the same. The best case scenaro (Fgure 1c), n whch the ncrease n the quantty of hgh qualty nformaton s greater than that of low qualty nformaton, results n a better sgnal-tonose rato. qualty qualty qualty T1 T1 T1 T0 T0 T0 qualty qualty qualty (a) (b) (c) Fgure 1. Hypotheszed Change n Qualty of Informaton Supply Whle the above ssues are mportant unresolved research problems, the lterature on nformaton goods does not seem to tackle ths problem drectly. Most of the extant lterature studyng nformaton goods focuses on the ssues of prcng (e.g., Rggns 2002) and packagng of nformaton goods (e.g., Bakos and Brynjolfsson 1999, Geng et al. 2005) due to neglgble margnal costs, socal and legal ssues such as copyrght and pracy (e.g., Chen and Png 2003), ssues related to mult-channel management (e.g., Dewan et al. 2000) and the market structure of search engnes (e.g., Telang et al. 2004). Whle the extant lterature provdes valuable nsghts wth respect to our understandng of the economcs of nformaton goods, they fal to fully consder the mpact of new technologes n ther entrety. For example, whle such studes take nto account the change n the qualty of nformaton consumed due to new technology, they do not look at the mpact of technology on the quantty and qualty of the nformaton that s produced and avalable, or n the other words the sze and qualty of the pool of nformaton from whch users consume nformaton. 3 3

5 The purpose of ths paper s to study the mpact of new nformaton and communcaton technologes on the supply sde of the nformaton revoluton. More specfcally, we study how the qualty and quantty of nformaton that s made avalable for consumpton changes due to the adopton of a new technology n the producton of nformaton contents. We focus on analyzng the mpact of two aspects of a new technology, namely the reducton n the costs assocated wth producng nformaton and the reducton of storage and/or transmsson costs whch are manfested n the relaxaton of tme and space constrants htherto mposed by older technologes. 6 In order to understand how technology mpacts the producton of nformaton, we frst develop a framework to understand the role of technology n the producton of nformaton. We dentfy the unt of analyss (.e., an artcle) and the two levels of analyss (.e., the genre and the channel levels) for ths study and ther relatonshps wth each other. The channel s an entty that decdes what artcles t wll produce, and all the channels aggregated comprse the genre. The aggregate change n the qualty of nformaton produced by the channels due to the adopton of new technology wll therefore be reflected at the genre level also. Based on ths framework, we develop an economc model that determnes what nformaton gets produced and use ths model to derve theoretcal propostons wth respect to changes n the qualty of nformaton produced by ndvdual channels based on ther varous characterstcs. Whle the theoretcal analyses help us dentfy certan behavoral dynamcs at the channel level, these channel level mpacts cannot be smply aggregated to represent the overall mpacts at the genre level. In order to understand the genre level mpacts, we conduct further numercal analyses usng computatonal smulatons. 6 It may seem obvous that the reducton n the above mentoned costs leads to an ncrease n the quantty of nformaton beng created. But the change n the qualty of nformaton s more dffcult to predct. Whle the reducton n producton costs may allow the producton of content that was prevously too expensve to produce resultng n an ncrease n hgh qualty nformaton, the relaxaton or elmnaton of tme and space constrants allows the publcaton of all nformaton that s conceptualzed wthout a check on the qualty resultng n an ncrease n lower qualty nformaton. The mpact of new technologes on the qualty of nformaton avalable for consumpton s therefore qute unclear. 4 4

6 Ths paper s organzed as follows. We frst present the conceptual framework that llustrates the process of nformaton producton conceptualzaton, producton, publcaton and dstrbuton. Secton 3 develops an economc decson model of content producton to derve propostons that hghlght the mpact of new technology. Numercal results are presented to further gan nsghts n secton 4. Fnally, we conclude the paper wth dscussons and mplcatons. 2. THE PROCESS OF INFORMATION PRODUCTION Our llustraton of the evoluton of the encyclopeda n the ntroductory secton helps us conceptualze the lfe-cycle of nformaton producton. In ths secton we formalze a framework that llustrates the process of nformaton producton and the role of technology wthn the process. By dentfyng the aspects of the process that are drectly affected by the technology employed, we wll be able to observe how changes to the employed technology wll mpact the process and consequently the qualty and quantty of nformaton that s the outcome of the process. The process of nformaton producton can be formulated as a four step sequence (Fgure 2). Genre Genre Channel Meda Conceptualzaton Producton Flter Creaton Dstrbuton Fgure 2. Process Model of Informaton Producton We defne an artcle as the unt of nformaton that s taken through each step of the producton process. An artcle belongs to a genre, whch s a socally recognzed type of communcatve acton and s dentfed by ts purpose and shared characterstcs of form (Orlkowsk and Yates 1992, 1994). The 5 5

7 purpose of a genre s to defne the type of content (or the type of artcles) and the form of the genre defnes the typcal format and presentaton of the sad content. The scope of a genre can be relatve. Examples of broad genres nclude news, cnema, and fcton; whle examples of narrower genres nclude sports news, acton moves, and scence fcton. In the frst step of the producton process, an artcle s conceptualzed by the creators (or authors) of the artcle. At ths stage, the artcle need not be n the fnal form that s made avalable to the end consumers, but s rather more of an dea or ntal draft. The acceptable form of the artcle at ths stage s also determned by the genre. For nstance n the news genre, an artcle at ths stage may be an dea for an news story, whle for cnema or fcton, an artcle at ths stage would be a draft of a manuscrpt. After ths stage, the artcle s submtted to a channel whch s an entty (or a person or an organzaton) that decdes whch artcles are to be produced and takes the responsblty for the costs of producng artcles n the fnal form that s to be dstrbuted for consumpton. For example, wthn the genre of fnancal or market news, the Wall Street Journal would be a channel that publshes nformaton content for that genre. There are typcally multple channels wthn a genre 7. When an artcle s submtted to a channel, t goes through the second stage of the producton process namely the producton flter. The producton flter s the channel s mechansm that determnes whether t wll produce an artcle t receves. Ths decson s made based on the costs and the expected returns of the artcles n addton to other characterstcs of the channels. 8 For example, wthn the cnema genre, 20 th Century Fox (.e., the channel) receves many scrpts from playwrghts. The channel wll evaluate the scrpts to decde whch ones to actually produce. If the channel decdes to produce the artcle, the artcle proceeds to the thrd step of the producton process whch s the creaton stage. The conceptual form of the artcle s converted nto a form that s sutable for sale and/or consumpton and the channel ncurs all the costs assocated wth the 7 Ths s not to say that a channel only produces artcles wthn a partcular genre. A channel may ndeed publsh artcles from one or more genres. For example, most news sources publsh artcles n a wde rage of sub-genres of news, such as sports news, nternatonal news etc. However, for smplcty we lmt our analyss and dscusson to a sngle genre for the remander of ths paper. 8 Ths process s descrbed n further detal n the followng sectons. 6 6

8 creaton of the artcle. At ths stage the channel chooses the approprate dstrbuton medum or meda, whch s determned by the form dctated by the genre, as noted above. For the news genre, ths would ental the converson of an dea for a news story nto the actual news story n a form that wll be broadcasted. Smlarly, for the cnema genre t would ental the creaton of the fnal cut of the move. In the fnal step of the producton process, the artcle s replcated and dstrbuted to the end consumers usng the selected dstrbuton meda. The choce of medum s determned by the rchness of the meda that s requred by the genre and/or the channel (Daft and Lengel 1984, 1986). Dfferent meda use dfferent technologes and afford dfferences n format (.e., dgtal vs. analog), level of personalzaton, number of cues, whether the content s pushed from the channel to the consumer or pulled by the consumer from the channel etc. Further, these prmary characterstcs of meda may mpact mportant nformaton producton and consumpton characterstcs such as the ease of reproducton, ease of search, ablty to archve etc. All these characterzed also affect the creaton or dstrbuton processes. In the cnema genre, the meda that could be employed for dstrbuton nclude VHS, DVD, theatrcal releases and the Internet 9. When a new technology assocated wth a medum emerges, the genres and channels that employ the medum are affected. At the channel level, the creaton and dstrbuton stages are the ones that are drectly mpacted by the changes due to the adopton of the new technology. Reducton n costs, for example, wll lead to lower creaton and dstrbuton costs and consequently hgher profts (assumng demand does not change). The more subtle but equally mportant mpact s experenced at the flterng stage, where the reducton n costs enters the decson process of the channel and changes the publshng behavor of the channel n terms of whch artcles they decde to produce. The aggregated change n the 9 Ths framework s generally applcable to most genres, though n some cases, some of these steps mght overlap. For example, for certan genres, the conceptualzaton and flterng may occur smultaneously or the creaton and dstrbuton may be ndstngushable, as s the case wth most textual content avalable on the Internet. Furthermore, the conceptualzaton and creaton of content may occur smultaneously before the producton flter. In such cases, we can perceve the creaton step as the tme at whch the creator or authors are remunerated by the channel. 7 7

9 behavor of all channels wthn a genre demonstrates the genre level mpact of the new technology. In the followng secton we develop a smple economc model that represents a channel s producton flter and models the decson process of choosng artcles for producton. Ths model s frst used to dentfy the prmary channel level mpacts of adoptng a new technology. It s then used as the bass for the numercal analyss to dentfy further channel level and genre level mpacts. 3. AN ECONOMIC MODEL OF INFORMATION PRODUCTION Based on the conceptual framework of nformaton producton presented n the prevous secton, we develop an economc model of a channel s decson process. Ths model hghlghts a channel s ratonal behavor wth respect to nformaton producton. Each artcle 10 wthn a genre can by characterzed by ts nherent qualty and cost. Snce the genre determnes the type of artcles, t also defnes the qualty parameters that are pertnent to the artcles wthn the genre. We consder genres to be very narrow or restrctve defntons of artcle types, where the qualty of ndvdual artcles can be vertcally dfferentated wthn the genre. The qualty of an artcle can also be conceptualzed as the determnant of the demand of the artcle. Ths mples that hgher qualty artcles wll have a larger demand than lower qualty artcles. The cost of an artcle ncludes all costs assocated wth producng the frst copy (or nstance) of the content n ts fnal form. These costs may nclude but are not lmted to remuneraton for the authors or creators, the costs of the resources such as nfrastructure employed n the creaton as well as the cost of the meda that s used. The dstrbuton costs are unform for all artcles publshed by a channel and are ndependent of the characterstcs of the artcle. For example, prntng and/or shppng a 200 page book wll cost the same regardless of ts content qualty! 10 For expostonal clarty, we use the subscrpt j for the artcle and the subscrpt for the channel. 8 8

10 Each channel that publshes artcles n the genre s characterzed as havng an aspraton level for qualty ( q ˆ ), a capacty ( n ˆ ) and a budget ( c ˆ ). q ˆ represents the channel s qualty threshold the lowest qualty of artcles that the channel s wllng to publsh. The capacty n ˆ s the upper lmt of the number of artcles that the channel s wllng or able to publsh n each tme perod. Ths s determned by the tme and space constrants of the channel whch s genre-specfc, whch are largely determned by the nature of the meda that s used by the channels. For example, the number of pages of a magazne s relatvely fxed across ssues. The budget (or cost threshold) c ˆ represents the maxmum cost that channel s wllng to ncur per artcle. The budget can also be conceptualzed as cn ˆˆ for the channel at a gven pont of tme. 11 In each tme perod, each channel receves a set of n artcles (.e., the supply) from whch t wll select the n ˆ artcles to publsh. For smplcty, we can assume that the supply set for each channel s unque (.e., the same artcle s not submtted to more than one channel durng the same tme perod). A proft maxmzng channel wll only want to produce and publsh a subset of the supply set that wll maxmze ts expected proft gven ts capacty and qualty and cost thresholds. If all artcles produced by a channel are sold at prce p each, at a dstrbuton and replcaton cost of d each, the expected revenue from an artcle j can be expressed as f (q j )(p d ), where f (q j ) s the expected demand for the artcle j publshed by channel and can be assumed to be an ncreasng functon n q j based on our pror defnton of qualty. Therefore, the expected proft for artcle j can be expressed as π j = f (q j )(p d ) c j where c j s the cost for producng the artcle j. A channel s problem can formulated as the followng nteger program (see Fgure 3). 11 Conceptualzng the budget as the maxmum cost for each artcle or the maxmum cost for all n artcles dd not ˆ result n any change n the results. 9 9

11 Objectve Functon: n n * ( )( ) [ ] Max π = π x = f q p d c x j 1, n j j j j j= 1 j= 1 Constrants: 1. Qualty Constrant: ( q ˆ j q ) xj 0 2. Budget Constrant: n cx nc ˆˆ j j j= 1 3. Capacty Constrant: n j= 1 x j nˆ Decson Varable: x j = 1, f artcle j s chosen for publcaton by channel, x j = 0, otherwse Fgure 3. A Channel s Optmzaton Problem The channel wll select a subset of ts supply set of artcles that wll obtan the maxmum proft (Objectve Functon). The frst constrant represents the qualty constrant that wll allow the selecton of only those artcles whose qualty s greater than the mnmum allowable qualty level (.e., the qualty threshold q ˆ ). The second constrant represents the budget constrant that wll only allow the selecton of those artcles that cost less than the cost threshold cn ˆˆ. Fnally, the capacty constrant wll ensure the selecton of a maxmum of n ˆ artcles. The decson varable x j s set to 1 f the artcle s chosen for producton/publshng, and 0 otherwse. Ths problem can be reduced to a knapsack problem whch can be solved usng dynamc programmng (Wnston et al. 2003). The above optmzaton problem s used to understand the mpacts of technology on the qualty and quantty of artcles produced at both the channel and the genre level. However, before we can proceed wth the dscusson of these mpacts we have to make certan assumptons regardng the supply set of artcles to the channels. Snce the channel s optmzaton problem amounts to selectng the set of artcles to publsh, and ths set determnes the quantty and qualty of nformaton publshed, t s mportant to establsh the ntal 10 10

12 consderaton set from whch the channel makes ts choce. We need to dfferentate between the supply sets of dfferent channels, snce t s realstc to expect the authors or the creators of artcles to self select nto channels n a way that wll maxmze the probablty of ther artcle gettng selected whle at the same tme maxmzng her returns. Ths would mply that an author wll submt her artcles to channels that have qualty aspratons that are close to that of her artcle and to channels that wll offer the hghest remuneraton, or have a suffcently large budget to produce her artcle. If the supply set of artcles for all channels are dstrbuted unformly over the nterval [0,1] 12 (.e., there s one artcle j for each combnaton of (q j,c j ) such that q [ 0,1 ], c [ 0,1] j for each channel ), then all these channels wll select the same set of artcles for publshng, rrespectve of ther cost or, qualty constrants (see Fgure 4a). In ths stuaton, only those artcles wth the hghest qualty and lowest cost wll be selected by all channels. In order to dfferentate between two channels wth dfferent qualty thresholds ˆq 1 and ˆq 2 (n Fgure 4a), we lmt the qualty of the artcles n the supply set to a channel wth qualty threshold qˆ to [ qˆ - δ q, q ˆ + δ q ] (see Fgure 4b). However, even wth ths restrcton, two channels wth the same qualty threshold ˆq but dfferent cost thresholds ĉ and 1 ĉ 2 wll publsh the same subset of artcles (.e., rrespectve of ther cost thresholds). Therefore n order to dfferentate the supply set to channels wth the same qualty threshold and dfferent cost thresholds, we can make the followng assumpton of the relatonshp between the cost and the qualty of an artcle the costs c j of an artcle j wth qualty q j s lmted to the nterval [q j - δ c, q j + δ c ] (see Fgure 4c). In other words, to a certan extent, the cost of an artcle s correlated wth ts qualty. Snce authors are remunerated based upon the expected returns from ther artcles whch s determned by the qualty of ther artcles, hgher qualty artcles wll cost more than lower qualty artcles. Furthermore, the resources and nfrastructure related costs of artcles are made n order to mprove the qualty of the artcles. Under these assumptons we are able to dfferentate the supply of artcles to channels wth dfferent cost and qualty thresholds (see j 12 Wthout loss of generalty, we use the ntervals [0, 1] for qualty and costs, where 0 denotes very low qualty (nexpensve) and 1 denotes hgh qualty (costly)

13 Fgure 4c). We now examne the mpact of technology on the cost and qualty of artcles selected by channels usng the above setup. In order to solate the effects of adoptng a new technology, we keep the consumpton (or demand) and conceptualzaton (or supply) constant. (a) (b) (c) Fgure 4. Assumptons for the Supply of Artcles to a Channel 3.1. Reducton n Cost of Producton New technology s adopted by channels when t helps them mprove the effcency and reduces the costs of ther current processes as well as when t allows them to undertake processes that they were unable to do before due to varous lmtatons (effectvely reducng the costs from nfnty to a tangble amount). The adopton of a new technology under these crcumstances wll therefore reduce the costs ncurred by the channel and result n hgher profts. Other ndrect mplcatons nclude changes n the supply set due to reduced costs, whch lead to changes n the artcles that are selected by the channel. Consder a channel wth a supply set of n artcles. If the adopton of the new technology leads to a reducton of ε% of the costs of each artcle n the supply set,.e., ( ε ) c = 1 c j = 1.. n where c j s the j j cost of artcle j n the supply set of channel. Ths can also be nterpreted as an ncrease of ε% n the cost = + cˆ. threshold of the channel.e., cˆ ( 1 ε ) 12 12

14 Let us frst examne the mpact of the reduced costs on the quantty of artcles publshed by channels. If a channel publshes nˆ artcles under the orgnal costs, then the reduced costs do not have any mpact on the number of artcles that are publshed, as such a channel wll not be able to ncrease the number of artcles t publshes due ts capacty constrants. On the other hand, f a channel publshes less than nˆ artcles, the reducton n costs wll lead to an ncrease n the number of artcles publshed by the channel. Let n < nˆ be the number of artcles publshed by such a channel under orgnal costs, whch s equal to the number of artcles n the supply set that satsfy the cost and qualty constrants. When the costs are reduced, a larger number of artcles wll satsfy the cost constrant snce the cost threshold s effectvely ncreased. If ths ncrease n the number of artcles s channel wll select for producton s n = mn ( n +Δn, nˆ ). n Δn, then the number of artcles that the Δ s ncreasng n ε, snce the greater the reducton n the costs, the larger the new cost threshold wll be and consequently the greater the number of artcles that satsfy the cost and qualty constrants. Proposton 1: The reducton n costs wll result n an ncrease n the quantty of artcles produced by channels that do not publsh to ther full capacty. Ths ncrease s proportonal to the magntude of the reducton n costs. 13 The mpact of reduced costs on the qualty of artcles s also determned by the addtonal artcles that satsfy the cost and qualty constrants as a consequence of the effectve ncrease n the cost threshold. Consder a channel whch publshes nˆ artcles under the orgnal costs (Fgure 5a). The artcles n the supply set that satsfy the cost and qualty constrants are enclosed n the area ADCB (the feasble regon) 14. The nˆ artcles that maxmze the objectve functon are enclosed n the area BCDHF. When the costs are reduced, the channel wll produce nˆ artcles, the artcles that satsfy the cost and qualty 13 Formal proofs of the propostons are avalable upon request. 14 The supply set of the artcles n Fgure 5 and Fgure 6 are bounded by the four lnes c = ĉ (the cost threshold), q = ˆq (the qualty threshold) and lower bound of the costs of the artcles n the supply set c = q δ c and the upper bound of the qualty of the artcles n the supply set q = ˆq + δ q as descrbed n the assumptons above

15 constrants are enclosed n the area A BCD and the new subset of artcles that s selected for producton s enclosed n the area BCD H F. The reducton n costs leads to the ncluson of the addtonal artcles enclosed n the area AA D D, and consequently the new selecton replaces the artcles enclosed n the area FHXF n the orgnal selecton wth those enclosed n XH D D. From the fgure, we can clearly see that the qualty of artcles n XH D D s greater than the qualty of artcles n FHXF, and therefore the average qualty of the new selecton set s greater that the average qualty of the orgnal selecton set. As dscussed above, the channels that publsh fewer than nˆ artcles wth the orgnal costs experence an ncrease n the quantty of artcles they publsh when costs are reduced. The mpact on the qualty due to reduced costs s consequently determned by the qualty of the artcles that are added to those selected wth the orgnal costs. Consder a channel that publshes n < nˆ artcles under the orgnal costs and n +Δn nˆ artcles after the costs are reduced (Fgure 5b). The artcles n the supply set that satsfy the cost and qualty constrants are enclosed n the area AEB and A E B for the orgnal costs and the reduced costs respectvely. The addtonal Δn artcles that the channel selects for producton are enclosed n the area A AEE and the selecton set s enclosed n A E B. the average qualty of the artcles enclosed n A E B s clearly greater than the average qualty of the artcles of the orgnal selecton (enclosed n AEB) due to the ncrease n the number of hgher qualty artcles enclosed n A AEE. The ncrease n the average qualty of artcles s determned by the ncrease n the qualty of the addtonal artcles that satsfy the cost and qualty constrants (enclosed n the areas A D DA and AE EA n Fgure 5a and b respectvely). The upper lmt of the average qualty of the artcles publshed by a 1 channel s q + δ, whch occurs when the artcles that are selected for producton are unformly ˆ 2 q dstrbuted over the nterval q ˆ, ˆ q + δ q

16 Proposton 2: The reducton n producton costs wll result n an ncrease n the average qualty of the artcles publshed by all channels and ths ncrease s proportonal to the ncrease n the magntude of the reducton n costs. (a) Full Capacty (b) Not Full Capacty Fgure 5. Increase n Qualty of Artcles due to the Reducton n Producton Costs 3.2. Relaxaton of Capacty Constrants The capacty constrants of channels are prmarly determned by the nature of the medum they employ. The medum dctates that amount of storage (n terms of space and/or tme) that can be used by the channels at a gven ponts of tme. New technologes (such as dgtzaton of audo/vdeo) sgnfcantly reduce the costs assocated wth storage (as well as space and/or tme lmtatons) assocated wth tradtonal meda, consequently relaxng the capacty constrants for channels that adopt the new technology. When the capacty of a channel s ncreased from nˆ to n ˆ, the quantty of the artcles that t publshes wll ncrease f t orgnally publshes nˆ artcles and n n ˆ where n s the number of artcles n the supply set that satsfy the cost and qualty constrants. If n < n ˆ, t wll publsh all artcles that satsfy the cost and qualty constrants. Therefore, the number of artcles publshed by the channel s ( n n ˆ ) mn,, and consequently the ncrease n quantty of artcles publshed by a channel wll ncrease wth n ˆ as long as n ˆ n

17 Proposton 3: Relaxaton of capacty constrants wll lead to an ncrease n the quantty of artcles publshed by those channels that publsh at least nˆ artcles. Ths ncrease wll be bound by the new capacty n ˆ, and the number of artcles n the supply set that satsfy the cost and qualty constrants of the channel. Snce there s no change n the subset of artcles selected by channels that publsh less than nˆ artcles, the ncrease n capacty wll not result n any change n the average qualty of artcles produced by these channels. Furthermore, certan channels such as the one llustrated n Fgure 6a wll not experence a change n the average qualty of the artcles that they publsh. The artcles n the supply set that satsfy the cost and qualty constrants are enclosed n the area ABCD. The set of nˆ artcles that maxmze the objectve functon under the ntal capacty constrant s enclosed n the area FGCB, and the set of n ˆ artcles that maxmze the objectve functon when the capacty s ncreased n enclosed n the area F G CB. The addtonal nˆ nˆ artcles that are selected by the channel are enclosed n the area FGG F and are unformly dstrbuted over the entre qualty range q ˆ, qˆ + δ, and therefore the average q qualty of the channel does not change. However, for other channels, such as the one represented n Fgure 6b, the average qualty of the artcles publshed by the channel wll decrease. Under the ntal condtons, the artcles that satsfy the cost and qualty constrants are enclosed n the area AEB. The nˆ artcles that maxmze the objectve functon are enclosed n the area FHEB and when the capacty s ncreased, the n ˆ artcles that maxmze the objectve functon are enclosed n the area FH EB. The addtonal nˆ nˆ artcles that are selected by the channel are enclosed n the area FH HF. These artcles have lower qualty than those orgnally selected by the channel, and wll contrbute to lowerng the average qualty of the artcles publshed by the channel. Therefore the change n the average qualty of artcles produced by a channel s determned by the qualty of the new artcles that are selected for producton. As descrbed above, the qualty of the addtonal artcles s always less that the qualty of the artcles n the orgnal selecton set. The 16 16

18 magntude of the ncrease n capacty determnes the ncrease n the number of new artcles. As the number of new artcles ncreases, the qualty of these new artcles decreases, and consequently the average qualty of the artcles produced by the channel decreases. Proposton 4: The ncrease n the capacty of a channel wll results n the decrease of the average qualty of artcles produced by channels that publsh at least nˆ artcles. Ths decrease n the average qualty s drectly proportonal to the ncrease n capacty. (a) No change n Average Qualty (b) Decrease n Average Qualty Fgure 6. Impact of Relaxaton of Capacty Constrant on the Average Qualty of Publshed Artcles In the above dscusson we have been able to analytcally dentfy certan trends n the change n the behavor of channels when the costs of artcles are reduced and the channels capacty s ncreased. However, these trends cannot be quantfed or generalzed for all channels wthn a genre snce the magntude of the changes n the quantty and average qualty of artcles depend sgnfcantly on the cost and qualty thresholds of the channels as well as the nature of the quantty and average qualty of the artcles produced by the channels under the ntal condtons, and not lnear n nature. Therefore we employ numercal analyss n the followng sectons to gan further nsghts at the channel and genre level, n addton to verfyng our analytcal results

19 4. NUMERICAL ANALYSIS In addton to the above mathematcal analyss the followng numercal analyss usng smulatons s done for a number of reasons. Frstly, the analytc results presented above dentfy general trends n the behavor of channels, but these behavors are not unform across all the channels and cannot be generalzed and therefore are nsuffcent to predct genre level mpacts. Smulatons have been found to be partcularly relevant when the focal phenomenon s non-lnear (Davs et al. Forthcomng) as s the case wth the dfferent channels and ther mpact at the genre level. Secondly, smulaton studes have also been successful n dentfyng the relatonshp between mcro-behavor (at the channel level) and macrodynamcs (at the genre level) (Lom and Larsen 1996). We have been able to dentfy the rules that regulate the behavor at the channel level; however the effects of these rules are unobservable at the genre level snce the behavors are not unform. Usng smulatons, we wll be able to dentfy the mpact of the ndvdual channels on the genre as a whole (Repennng 2002). Thrdly, smulatons have also been used snce t s challengng to obtan the emprcal data at the varous levels of analyss (Davs et al. Forthcomng, Zott 2003). Qualty beng a very subjectve measure cannot be very accurately quantfed across a number of channels from real world data. Smlarly dentfyng the dfferent thresholds for ndvdual channels s also challengng to obtan. Furthermore, dfferent channels adopt technologes at dfferent ponts of tme, and t s dffcult to accurately dentfy the exact tme at whch dfferent channels do so and solate the mpact on the qualty and quantty of nformaton produced by that channel and extrapolate t to the genre level. Fnally, the smulatons allow us to relax some of the restrctve assumptons made n the mathematcally analyss, such as the contnuty of the dstrbuton of the supply of artcles and lnear relatonshp between the costs and the qualty of the artcles. Matlab 7.1 was used for conductng the smulatons and the followng numercal analyss. Snce the behavor of the channels vares based on ther cost and qualty thresholds as well as ther supply set of artcles, smulatons were run for 100 combnatons of cost and qualty thresholds as well as 1000 dfferent supply sets. We normalze all cost and qualty values over the nterval [0,1]. Each combnaton 18 18

20 of the cost and qualty thresholds s represented by a sngle channel wth cost threshold cˆ [ 0,1] qualty threshold qˆ [ 0,1]. The cost and qualty thresholds of the 100 channels are unformly dstrbuted over the [ 0,1] [ 0,1] space. Each of the 1000 supply sets for each of the 100 channels were generated n the followng manner. The supply set for each channel comprses of 100 artcles,.e. n = 100 = 1,..,100. We relax the assumpton of the contnuous and unform dstrbuton of artcles n the supply set and use normal dstrbutons for the artcles n the supply set. For a channel, wth qualty threshold ˆ q, the supply set comprses of a draw of n artcles from N( qˆ, σ q) ( ˆ ) q ~ N q, σ j = 1,..., n ). The cost for each artcle j n the supply set of channel s drawn from j q N( qj, σ c ) (.e., ( ) (.e., c ~ N q, σ j = 1,..., n ). The values of n, σ q and σ c are set to 100, 0.3 and 0.1 j j c respectvely for all the smulatons 15. and The optmzaton problem descrbed n Fgure 3 was executed for each of the 1000 supply sets for each of the 100 channels for the base case and the subsequent altered scenaros resultng from the adopton of the new technology. The output of the nteger program s the subset of the supply set artcles that are selected for producton by the channel. Ths output was used for the analyss of the dfferent scenaros descrbed below. In the base case, the capacty was set to 10,.e., ˆ supply set used were the orgnal data that was generated. and n =10 [ 1,100] Fgure 7 and 8 shows the mpact of reduced costs and ncrease capacty respectvely at the channel level. The top row represents the average qualty of the artcles publshed by the channels and the bottom row represents the quantty of artcles. The frst column represents the average qualty and 15 We purposefully set the varance of the qualtes of the artcles submtted to a channel to be smaller than the varance of the costs of producng these artcles. Ths setup smulates a settng where authors effectvely self-select nto ther rght channels (.e., an author wth a hgh qualty content wll submt her work to a hgh-qualty channel, whereas an author wth a lower qualty artcle wll know better than to submt hs work to a hgh-qualty channel) but the costs of producng the artcles may vary substantally (.e., artcles of the same qualty may vary substantally n costs)

21 quantty of the artcles publshed by the channels. The remanng columns n Fgure 7 and 8 represent the ncremental reducton n costs and ncremental ncreases n the capacty of channels respectvely. The mplcatons of these results are dscussed n detal below Reducton n Cost of Producton To smulate the reducton n producton costs, the costs of each artcle n each supply set for all the channels were reduced by a fxed percentage ε. ε was vared from 10% to 90% n order to observe the varaton n the qualty and quantty of the artcles due to the varaton n the magntude of the reduced costs. Fgure 7 shows the mpact at the channels when the costs are reduced by 30%, 60% and 90%. To solate the mpact of the reduced costs, the capacty was kept constant.e., ˆ n =10 [ 1,100]. Let us frst examne the mpact on the quantty of artcles publshed by the channels. We observe an ncrease n the quantty of artcles (2 nd row of Fgure 7) publshed by channels that publsh less than nˆ artcles n the regon representng channels wth hgh qualty thresholds and low cost thresholds. We also observe that the quantty of artcles publshed by theses channels ncreases as magntude of the reducton n costs ncreases (untl the capacty threshold s reached). Nearly 60% of the channels publshed to ther full capacty under orgnal condtons. As the costs are reduced by 60% and 90%, the percentage of channels whch publsh to ther full capacty ncreases to 70% and 95% respectvely. These results are consstent wth Proposton 1. The average qualty of the artcles (1 st row of Fgure 7) publshed by a channel s slghtly greater than the qualty threshold of the channels n the base case. The rate of ncrease n the average qualty of the artcles publshed wth respect to the magntude of the reducton n costs s postve. Ths result s consstent wth Proposton 2. We can also observe that the magntude of the change of the average qualty of the artcles publshed ncreases wth the qualty threshold of the channels as the magntude of the reducton n costs ncreases. Ths mples that channels wth hgh qualty thresholds are more senstve to the reducton n the costs f the reducton n costs s suffcently large. Ths result can also be 20 20

22 nterpreted to mply that for genres where the costs due to meda are the major component of the cost of the artcle, the adopton of new technology can sgnfcantly ncrease the average qualty of artcles publshed by channels wth hgh qualty thresholds (by up to 8%). Smlarly for genres where the meda related costs do not sgnfcantly contrbute to the cost of artcles, the adopton of new technology ncreases the average qualty of the artcles publshed by channels wth lower qualty thresholds more notably than that of channels wth hgher qualty thresholds. We can also observe that the ncrease n the average qualty of artcles produced by channels s more sgnfcant for those channels that publsh nˆ artcles. Consequently as the reducton of costs ncreases and the number of channels that publsh nˆ artcles ncreases, a greater number of channels experence an ncrease n the average qualty of artcles, and by larger magntudes. Fnally, we observe that the rate of ncrease n the average qualty of artcles publshed wth respect to the magntude of the reducton n the costs of the artcles s ncreasng. The ncrease n average qualty of artcles publshed by channels when costs are reduced from 60% to 90% s greater than the ncrease when costs of reduced from 30% to 60% and 0% to 30%

23 Fgure 7. Reducton n Producton Costs 22

24 4.2. Relaxaton of Capacty Constrants To smulate the ncrease n capacty, we vary the capacty threshold nˆ for all the channels, whle keepng the characterstcs of the supply set constant. Fgure 8 shows the mpact at the channel level when the capacty s ncreased from 10 to 20, 50 and 100. The qualty of artcles publshed by channels ncreases for only those channels that publsh n ˆ =10 artcles n the base case. The channels that publsh less than n ˆ =10 artcles n the base case are restrcted by ther cost and qualty constrants, and therefore are unable to ncrease the quantty or change the qualty of artcles they publsh. These results are consstent wth Propostons 3 and 4. We can also observe that the ncrease n the quantty of artcles publshed by the channels decreases as the capacty ncreases. Ths mples that the producton flter constrants restrct the quantty of artcles publshed by the channel, and therefore the quantty of artcles wll not ncrease ndefntely as the capacty s ncreased. The channels that experence the greatest ncrease n quantty of artcles they publsh le n the regon that represents channels wth hgh cost thresholds and low qualty thresholds. The ncrease n the quantty of the artcles publshed s drectly proportonal to the cost threshold of the channel (.e., hgher the cost threshold, the greater the ncrease n quantty of artcles publshed) and nversely proportonal to the qualty threshold (.e., lower the cost threshold, the greater the ncrease n quantty of artcles publshed). The average qualty of artcles publshed by channels decreases as the capacty of the channels s ncreased. The decrease n the average qualty of artcles publshed drectly corresponds to the ncrease n the quantty of artcles publshed by the channel. We can therefore see that as the quantty of artcles publshed by a channel ncreases, the average qualty of the artcles t publshes decreases and approaches ts qualty threshold. The decrease n average qualty s ncreasng wth ts cost threshold and decreasng wth the qualty threshold. 23

25 Fgure 8. Increased Capacty 24

26 4.3. Aggregate Effects at the Genre Level From Fgure 7 and 8 we can see that the reducton n costs and the ncrease n capacty lead to an ncrease n the quantty of artcles publshed by channels. However, the channels that are affected n both case are mutually exclusve,.e. a channel that ncreases the quantty of artcles they publsh due to reduced cost wll not ncrease the quantty of artcles they publsh due to ncreased capacty and vce versa. However, the average qualty of artcles publshed by a channel may ncrease due to reduced costs and decrease to ncreased capacty. Therefore, when costs are reduced and the capactes ncreased there s an overall ncrease n the quantty of artcles publshed by all the channels. The mpact on the average qualty of artcles publshed by channels dffers wth the magntude of the reducton n costs and the magntude n the ncrease n capacty, but s prmarly drven by the reducton n costs. At the channel level, we only examne the changes n the average qualty of the artcles publshed by the channels and not the qualty of the artcles that selected for producton. At the genre level we examne the dstrbuton of the quantty of artcles n terms of the qualty. Fgure 9 shows the mpact of the reducng costs and ncreasng the capacty at the genre level. The fgures show the dfference n the percentage of the artcles publshed for each qualty from the base case. As we progress down a column, we see an ncrease n the quantty of the hgher qualty artcles; therefore, the reducton n costs leads to an ncrease n the percentage of hgher qualty artcles. Ths ncrease s proportonal to the ncrease n the percentage reducton of the costs of the artcles. Ths represents the stuaton descrbed n Fgure 1c. On the other hand, f we move from left to rght across a row, we see an ncrease n the quantty of lower qualty artcles, thus representng the stuaton descrbed n Fgure 1b. When both the costs are reduced and the capacty s ncreased to smlar levels, we observe a unform ncrease n the quantty of the artcles of all qualty levels (see reduced cost = 90% and capacty = 100). 25

27 Fgure 9. Genre Level & Interacton Effects 26

28 5. DISCUSSION AND CONCLUSION Ths paper examnes the mpact of the adopton of a new technology or broadcast medum on the process of nformaton producton and the mpact on the qualty and quantty of nformaton whch s the outcome of ths process. In order to dentfy these mpacts, we develop a framework that explcates the role technology plays n the creaton of nformaton. Ths framework s then used to develop an economc model for nformaton producton for a ratonal proft maxmzng entty (a channel) that produces and dssemnates nformaton. From a gven supply of nformaton (or artcles), the channel selects those artcles that maxmze ts proft under certan capacty, budget and qualty constrants. Ths economc model s used to study two of the prmary mpacts of technology at the channel level and the genre level. These mpacts are the reducton n producton costs and the relaxaton of capacty constrants. At the channel level, we fnd that the reducton n producton costs leads to a margnal ncrease n the average qualty and quantty of nformaton produced, whle the relaxaton of the capacty constrants leads to an ncrease n the quantty and a margnal decrease n the average qualty of nformaton. However these mpacts are not unform across all channels and vary wth the characterstcs of the channel. At the genre level, the reducton of producton costs results n an ncrease n the total quantty of nformaton avalable for consumpton, and ths ncrease s greater for hgh qualty nformaton than low qualty nformaton. The relaxaton of the capacty constrants however, has the opposte effect and results n a much larger ncrease n the quantty of low qualty nformaton compared to hgh qualty nformaton. When the producton costs are reduced and the capacty constrants are relaxed, the ncrease n the quantty of lower qualty nformaton s stll found to be hgher than the ncrease n the quantty of hgher qualty nformaton, though the dfference between the two s lower than when the only the capacty constrants are relaxed Implcatons The mplcaton of the ncrease n the overall quantty of orgnal nformaton avalable for consumpton n each tme perod s a decrease n the sgnal to nose rato of nformaton. Ths ndcates 27 27