Analyss of Subscrpton Demand for Pay-TV Manabu Shshkura Researcher Insttute for Informaton and Communcatons Polcy 2-1-2 Kasumgasek, Chyoda-ku Tokyo 110-8926 Japan m-shshkura@soumu.go.jp Tel: 03-5253-5496 Fax: 03-5253-5497 Norhro Kasuga Assocate Professor Nagasak Unversty 4-2-1 Katafuch, Nagasak Nagasak 850-8506 Japan n-kasuga@net.nagasak-u.ac.jp Tel/Fax: 095-820-6396 Ako Tor Professor Yokohama Natonal Unversty 79-4 Tokwada, Hodogayaku Yokohama 240-8501 Japan ator@ynu.ac.jp Tel: 045-339-3739 Fax: 045-339-3707 Contact Author. Our conclusons, opnons, and other statements n ths paper are those of the authors and are not necessarly of our afflated organzatons. We are very grateful for useful comments from the semnar partcpants of The Japan Economc Polcy Assocaton. 1
Summary In ths paper, we wll conduct an analyss from an emprcal perspectve concernng broadcastng demand behavor and characterstcs. More concretely, we wll assume three optons (subscrpton patterns), namely, nonsubscrpton, cable TV subscrpton, and subscrpton to broadcastng servces va communcatons satelltes (CS), and we wll conduct an analyss of broadcastng servce demand behavor usng a dscrete choce model. Accordng to our result, prce elastcty s low for cable TV and CS subscrbers, whch s a contrast to the Unted States, where prce elastcty for pad broadcastng s hgh. It shows that factors other than prce contnue to have a strong nfluence on demand behavor (subscrpton choces) n the Japanese market for regular and pad broadcastng (the multchannel market) and that the market has not yet entered the phase of prce competton between cable TV and CS broadcastng. Wth regard to channel number elastcty, the fgures are hgh for basc channel servce and low for specal channel servces. Ths ndcates that basc servces are an mportant concern for vewers. Ths suggests that, n lne wth reports that the spread of cable TV n the Unted States was promoted by dffculty n recevng terrestral broadcasts, basc channel servces may be more mportant than specal channels as keys to the spread of pad broadcastng n Japan. Keywords: Subscrpton demand, Broadcastng servce, Pay-TV, Dscrete choce model, Prce/Channel elastcty (JEL Classfcatons: L51, L82, R22) 2
1. Introducton In order to evaluate the effects on the welfare of broadcastng servces resultng from the shft to multchannel broadcastng, t s necessary to clarfy the structure of demand for broadcastng servces n the context of ths swtch. It has also been noted that desrable market condtons for the maxmzaton of socal welfare, fundng approaches (publc, commercal, and pad broadcastng), levels and structures of competton (ncludng monopoly, olgopoly, and compettve markets) all depend greatly on demand-sde features, such as substtutablty among programs and channels. 1 Analyses of demand for broadcastng servces nclude ones that focus on the measurement of ther economc value and ones that focus on evaluaton of compettve condtons n the broadcastng market; n the Unted States the latter sort have been more commonly conducted as part of anttrust/competton polcy. In partcular, aganst the backdrop of the further ntensfcaton of competton wth the spread of satellte broadcastng and the entry of cable companes and regonal telephone companes nto the market, n recent years analyses have been drected at the cable TV market and DBS (drect broadcast satellte) broadcastng. 2 Keschnck and McCullough (1998) made the followng three ponts about the polcy utlty of demand analyses of broadcastng servces: () They offer grounds for judgment n dscusson of the poston of terrestral broadcastng as the only medum that allows non-affluent households to watch programs and of the mantenance of unversal servce and publc utlty. () They offer grounds for judgment concernng meda concentraton and busness effcency n connecton wth the FCC s (Federal Communcatons Commsson s) ownershp rules concernng regonal broadcasters, ncludng the queston of whether cable statons should be taken nto consderaton or not. () They offer grounds for judgment about polcy measures (nterventon) amed at mantanng local terrestral broadcasters such as the dea that mposng a must carry rule on cable TV and satellte broadcasters does not ensure the fnancal 1 Spence and Owen (1977). 2 Karkar et al. (2003) conducted an analyss of the factors behnd spread of pad broadcastng, takng DBS broadcastng nto account, showng that t progressed n areas where cable TV charges were hgh because of regulaton and dd not progress n areas where the entry of cable companes and regonal phone companes had created ntense competton. Also, Goolsebee and Petrn (2004), usng subscrber data n a dscrete choce model, estmated the own-prce elastcty and cross-prce elastcty of cable servce (basc and premum) aganst DBS broadcastng and terrestral broadcastng. They showed that demand for premum cable servce and for DBS broadcastng was much more elastc than that for basc cable servce, and that subscrbers saw DBS and premum cable as substtutable servces. 3
health of all (terrestral) broadcasters, but t does am for the survval of a basc number of broadcasters and about market neutralty (compettve envronment). 3 Gven the fact that the spread of cable TV has already created a multchannel broadcastng market n the Unted States, t may not be possble to apply arguments advanced there to today s Japan wthout modfcaton. But there s an undenable possblty that ponts lke the market envronment, busness effcency, meda concentraton, and the natonal mnmum may also become ssues here. In partcular, t seems qute plausble that ponts lke busness effcency and meda concentraton (tem above) and the promoton of a multchannel market and the mantenance of a natonal mnmum (tem above) wll fully mert consderaton n Japan. 4 In ths paper, we wll conduct an analyss from an emprcal perspectve concernng broadcastng demand behavor and characterstcs, ncludng the move toward multchannel broadcastng and subscrpton actvty n pad broadcastng, takng nto account the progress toward multchannel broadcastng. Secton 2 of ths chapter sums up the characterstcs of broadcastng servce demand. Secton 3 explans the methodology of our emprcal study. Secton 4 presents the results of our study. Secton 5 dscusses the mplcatons to be drawn from these results. 2. Broadcastng Servce Demand Behavor The relatonshp between the shft to multchannel broadcastng and broadcastng servce demand behavor has not been made entrely clear up to ths pont. In ths secton, pror to our emprcal analyss, we wll sum up ths behavor wth a focus on multchannel broadcastng. Broadcastng servce demand has two aspects: vewng decsons and the purchase of channel subscrptons. Purchasng channel subscrptons does not n tself produce any drect utlty; utlty only acheved through vewershp. The level of utlty rses wth the ncrease of vewng tme (volume of use), and margnal utlty tends to dmnsh; n these respects broadcastng servce demand s lke demand for ordnary goods and servces. If vewng tme s taken as a gven, then the level of utlty rses wth the vewng of programs that are close to the preferences of the vewers. Partcularly n 3 Turner Broadcastng System Inc. v. FCC, 117 Sup. Ct. 1174 (1997). 4 In some respects ths sort of consderaton has already begun, as n the revew of the prncple of nonconcentraton, but there s room for further dscusson of ssues lke the poston of pad broadcastng. 4
the case of broadcastng servces, whch the supplers offer unlaterally wthn short perods, there s an ssue of probablty nvolved n vewers beng able to see the sort of programs that they prefer, but the larger the number of avalable channels s, the better the chances are that they wll be able to fnd programs matchng ther preferences. So t would seem that ncreasng the number of channels avalable (.e., ncreasng dversty) should ncrease users utlty per vewng hour. And we can thnk of the utlty functon as a functon of the number of channels and the amount of vewng tme. As lmtng condtons, wth respect to the purchase of channel subscrptons, t would seem that the same sort of budgetary constrants would apply as n the case of the use of ordnary goods and servces, whle wth respect to vewng, nasmuch as ths s a tme-consumng lesure servce, tme constrants act as mportant lmtng condtons. So we may consder that broadcastng servces demand (vewng tme and channel purchases) behavor should be formulated n the shape of the double constrants of budget and tme. 5 Above we have assumed that channels and cost burdens can be freely chosen, but t s probably approprate to vew the choce as nvolvng the provson of channel servces that have been packaged to some degree. Partcularly n Japan s case, purchasng a TV set and payng recepton fees 6 allows one to use basc channel servces; selecton of channel servces to whch one s not yet subscrbed, along wth the acceptance of cost burdens, s conducted through dscrete choces to subscrbe to pad broadcastng servces based on comparson of the avalable channels and cost n the case of nonsubscrpton aganst the avalable channels and cost n the case of subscrpton. Based on ths perspectve, here we wll assume three optons (subscrpton patterns), namely, nonsubscrpton 7, cable TV subscrpton, and subscrpton to broadcastng servces va communcatons satelltes (CS), 8 and we wll conduct an analyss of broadcastng servce demand behavor usng a dscrete choce model. 9 5 For the relatonshp between a specfc broadcastng demand model and an emprcal model, see Shshkura et al. (2005). 6 These may be consdered a sort of fxed charge. 7 Ths s defned to nclude those households that use publc broadcastng (ncludng broadcastng satellte, or BS, servces and terrestral commercal broadcastng. 8 Note that these optons are not entrely mutually exclusve. 9 In the dscrete choce model, a consumer s assumed to acqure utlty from varous features (prce, qualty, content) of a good. It s further assumed that utlty Un can be expressed as the lnear V combnaton of n, a determnstc term explaned by the above features, and stochastc terms ε n, An wth actor n choosng among a set of alternatves n such a way as to maxmze hs or her utlty. The condtons for actor n to choose alternatve are expressed as follows: Un = Vn + εn > U jn = Vjn + ε j jn j A n Multple estmaton methods exst, dependng on what sort of dstrbuton s assumed for the above 5
3. Emprcal Choce Model of Broadcastng Servces Demand We assumed the followng lnear functon for determnstc term Vn (Model 1). V = α D + β Bch + β Ech + β Mch + ϕ p + δ D Y n 1 = CATV 1 2 3 1 1 = ANT n We use the number of channels n three categores of servce, namely, basc ( Bch ), extended basc ( Ech ), and specal ( Mch ), as the basc attrbute varable for the varous subscrpton states, and charges (amount pad, p ) as the varable ndcatng the cost burden. 10 D = CATV and D = ANT are dummy varables that are equal to 1 when subscrbed to cable TV and when not subscrbed to pad broadcastng, respectvely. We also nclude household ncome ( Y n ) as a household-specfc attrbute. 11 The data are non-aggregated data 12 from a questonnare survey. 13 See Table 1 for the descrptve statstcs for each of the varables. We also conducted an estmaton for a case where vewng tme for each category s ncluded (Model 2), namely, basc ( Bwatch ), extended basc ( Ewatch ), and specal ( Mwatch ), as the varable ndcatng vewng preferences. 14 V = α D + β Bch + β Ech + β Mch + γ Bwatch + γ Ewatch + γ Mwatch + ϕ p + δ D Y n 1 = CATV 1 2 3 1 2 3 1 1 = ANT n To sum up the condtons for the attrbute varables for each type of subscrpton status, those who choose nonsubscrpton have access only to the basc stochastc terms. 10 Basc servce refers to publc broadcastng channels provdng comprehensve programmng (ncludng educaton, news, and entertanment) and commercal channels. Extended basc refers to publc and commercal channels transmtted by broadcast satellte (dgtal and analog). Specal refers to channels provdng specal programs, such as flms and musc. 11 We also looked at other attrbutes, such as the number of household members and ther average age, but we dd not fnd any sgnfcant results, so we have omtted them here. 12 The average value s used for the attrbutes of alternatves that were not chosen, but snce the numbers of channels avalable to nonsubscrbers and to cable TV subscrbers vary from regon to regon, regonal averages were used as the opportunty number of vewable channels for nonsubscrbers and cable TV subscrbers. The average for the entre sample was used for CS subscrbers, because there are no regonal dfferences n ths servce. 13 The survey was conducted n March 2004 by the Insttute for Informaton and Communcatons Polcy (IICP) and the Foundaton for MultMeda Communcatons (FMMC), drected at households n Tokyo and surroundng prefectures and n the Hokurku regon. Note that the geographcal target was lmted to these two regons; also, pror adjustments were made n order to secure a certan number of subscrbers for samplng. Weghtng was appled n such a way as to match the natonwde spread of servces n settng the samplng shares for aggregaton purposes, but we have some reservatons about the estmaton results. For detals see Shshkura et al. (2005). 14 We also made estmates for cases that allowed for regonal varables, such as dummes for each prefecture and a regonal (Hokurku) dummy, as well as for household attrbutes, such as the number of household members, but we dd not fnd sgnfcant results for any of these, so we have omtted them here. 6
channels, 15 and ther cost burden conssts of payment of the recepton fees collected by publc broadcaster NHK (Japan Broadcastng Corporaton). Those who choose cable TV subscrpton have access to addtonal basc channels and also to specal channels, and ther cost burden conssts of the cable TV charges n addton to the NHK charges. 16 Those who choose CS servce have access to specal channels n addton to the basc channels avalable to nonsubscrbers, and ther cost burden conssts of the NHK charges plus the charges for the number of channels for whch they have contracted. The dfference between cable TV and CS servces s expressed n the numbers of addtonal basc and specal channels to whch they provde access and n the cost burden. The choce to subscrbe to one or the other s made n consderaton of the channels avalable to nonsubscrbers and the users preferences and cost burdens wth respect to the basc channels and specal channels. ------------------------------Insert Table 1 about here ------------------------------ 4. Emprcal Results Below we gve an overvew of the estmaton results for three logt models, namely, multnomal, nested, 17 and mxed, n our two dscrete choce models (See table 2). Wth regard to the sgn condton and statstcal sgnfcance of each varable, basc channel servce both satsfes the sgn condton and acheves an adequate level of sgnfcance, whle specal channel servce meets the sgn condton but does not have a hgh level of sgnfcance. The coeffcent of extended basc channel servce fals to meet the sgn condton and also has a low level of sgnfcance. Wth respect to level of charges, the sgn condton s met n all cases, and the level of sgnfcance s adequate overall. ------------------------------Insert Table 2 about here ------------------------------ It s dffcult to dentfy whch of the three methods of estmaton s most approprate. Pseudo R2 (McFadden s R2) s hgh n each model, and no major dfferences were found for the goodness of ft. Also, we conducted IIA (Independence from Irrelevant 15 The number of avalable basc channels vares, however, dependng on whether the household has access to BS servces, on the resdental envronment, and on the regon (number of terrestral channels). 16 The degree of freedom of choce, however, s not hgh, and the charges are set under fxed-rate plans. 17 For the nested structure we use (nonsubscrpton) vs. (cable TV vs. CS). In other words, we assume that households frst decde whether or not to subscrbe to pad broadcastng and, f they choose to subscrbe, they then choose between cable TV and CS. 7
Alternatves) tests to determne the approprateness of the assumpton that the dsturbances are ndependent and homoscedastc among the alternatves, but nether the Hausman test nor the lkelhood rato test for the model wthout the CS alternatve showed any clear tendency (Table 3). Even n the mxed logt wth relaxed constrants for dsperson of the parameters, the level of sgnfcance of the correlaton parameters was not very hgh. Though there were some dfferences n estmaton results among the methods, these were not great, so below we wll focus our consderaton manly on the results for the mxed logt model, whch has the least constrants and good sgnfcance of coeffcents. ------------------------------Insert Table 3 about here ------------------------------ In Table 4 we present our calculatons for the elastcty wth respect to charges and channels. For charges, all the values are less than 1, ndcatng nelastcty. For channels, basc servce shows an extremely hgh elastcty of more than 6, but specal channel servce shows extremely low elastcty of under 0.3. ------------------------------Insert Table 4 about here ------------------------------ 5. Implcatons Let us now sum up the mplcatons of the above emprcal results. Wth regard to prce elastcty, t may be natural that prce elastcty s low for nonsubscrbers wth no room for choce, but the fgures for cable TV and CS subscrbers are also on the low sde, whch s a contrast to the Unted States, where prce elastcty for pad broadcastng s hgh. 18 Prce elastcty s a useful ndcator of the progress of competton n a market, but we may take the above fndngs to show that factors other than prce contnue to have a strong nfluence on demand behavor (subscrpton choces) n the Japanese market for regular and pad broadcastng (the multchannel market) and that the market has not yet entered the phase of prce competton between cable TV and CS 18 See Goolsbee and Petrn (2004). Though t s not possble to compare lne by lne, because the methods of estmaton and the formulatons are dfferent, the elastcty values they fnd are n the range of 1.5 to above 2, whch are hgh by comparson wth the fgures for Japan even after allowng for such dfferences. 8
broadcastng. 19 Wth regard to channel number elastcty, the fgures are hgh for basc channel servce and low for specal channel servces. Though we should probably not consder concrete values to be certan, t does seem clear that there s a major dfference n elastcty between the two. 20 Ths s hghly sgnfcant when consderng cable and CS from the perspectves of ther spread and the competton between the two. It suggests the possblty that the broadcastng market, ncludng the pad broadcastng market, may be heavly mpacted n the perod to come dependng on how basc channel servces are (re)transmtted. It also suggests that n cases where there are dfferences n the numbers of basc channels that can be drectly receved because of geographcal or other condtons, basc channel transmsson may have an mpact on the competton between cable and CS for subscrbers. 21 If we look at vewng tmes, we fnd that even among households subscrbed to pad broadcastng, the basc channels (terrestral broadcasts) stll account for the overwhelmng majorty of hours; ths ndcates that basc servces are an mportant concern for vewers. Ths suggests that, n lne wth reports that the spread of cable TV n the Unted States was promoted by dffculty n recevng terrestral broadcasts, basc channel servces may be more mportant than specal channels as keys to the spread of pad broadcastng n Japan. 19 Own-prce elastcty for CS and cable s hgh by comparson wth that for nonsubscrpton, and the fgures for the two are close; t s thus possble to judge that cable and CS are n somewhat of a compettve relatonshp regardng servces and prces. 20 The dfference n elastcty between the two reflects the dearth of channels and the substtutablty of the content provded. Ths may be taken to reflect the fact that basc channel servces are provded generally on a comprehensve-programmng bass, consstng largely of orgnal content, whle much of the content of specal channel servces competes wth other servces, such as packaged vdeo, and the same programs are broadcast repeatedly. 21 In terms of polcy mplcatons ths may offer a certan perspectve concernng the proper shape of retransmsson for basc channel servces (the must carry rule). 9
[References] Goolsbee, A., and A. Petrn (2004) The consumer gans from drect broadcast satelltes and the competton wth cable TV, Econometrca Vol. 72 No. 2., pp. 351 381. Karkar, J., S. Brown, and A. Abramowtz (2003) Subscrptons for drect broadcast satellte and cable televson n the US: An emprcal analyss, Informaton Economcs and Polcy Vol. 15., pp. 1 15. Keschnck, R., and B. McCullough (1998) Why do people not subscrbe to cable televson? A revew of the evdence, Paper presented at the 28th TPRC (Telecommuncatons Polcy Research Conference). http://www.tprc.org/agenda98.htm. Shshkura, M., N. Kasuga and A. Tor (2005) The Shft to Multmeda and Multchannel Broadcastng and the Future of the Broadcastng Industry The analyss of subscrpton choce for pay-tv -, Paper presented at The Japan Economc Polcy Assocaton (n Japanese), May 28, (http://prof.mt.tama.hose. ac.jp/~kazukuro/jepa/newpage1.html). Spence, M. and B. Owen (1977), Televson Programmng, Monopolstc Competton, and Welfare, Quarterly Journal of Economcs, Vol. 91, No. 1, pp. 103-126. 10
Table 1. Basc Statstcs for the Varables Alternatve Expresson Nonsubscrbed Cable TV CS Number of respondents 139 221 153 Number Basc Bch 6.918 9.529 6.819 of Extended Ech 0.667 8.000 1.961 channels (average) Specal Mch 0.000 19.718 21.606 Charges (average monthly payments) p 1600.714 7471.639 5484.806 Vewng Basc Bwatch 699.990 650.761 577.862 tme Extended Ewatch 6.579 28.257 28.147 (average) Specal Mwatch 0.000 48.339 91.935 Average annual ncome 592.086 708.145 760.784 Y n 11
Table 2 Estmaton Results for Models 1 and 2 Model 1 Model 2 Multnomal logt Nested logt Mxed logt Multnomal logt Nested logt Mxed logt Varable Coeffcent SD Coeffcent SD Coeffcent SD Coeffcent SD Coeffcent SD Coeffcent SD Cable subscrpton constant -1.1354 ** 0.4798-1.1866 ** 0.5166-1.1358 *** 0.4324-0.9429 * 0.4984-1.1371 ** 0.5700-0.9438 ** 0.4360 Basc channels 1.8660 *** 0.1612 1.8799 *** 0.2233 1.8676 *** 0.1056 1.9816 *** 0.1739 2.1552 *** 0.2773 1.9824 *** 0.1166 Extended basc channels -0.0078 0.0718-0.0037 0.0735-0.0078 0.0514-0.0101 0.0753 0.0052 0.0835-0.0100 0.0523 Specal channels 0.0162 a 0.0124 0.0200 a 0.0147 0.0162 a 0.0126 0.0075 0.0129 0.0169 0.0163 0.0074 0.0133 Charges -0.0001 * 0.0001-0.0001 a 0.0001-0.0001 ** 0.0001-0.0002 ** 0.0001-0.0002 ** 0.0001-0.0002 *** 0.0001 Basc channel vewng hours - - - - - - 0.0000 0.0003 0.0000 0.0003 0.0000 0.0003 Extended basc channel vewng hours - - - - - - -0.0025 0.0024-0.0028 0.0026-0.0025 0.0021 Specal channel vewng hours - - - - - - 0.0046 *** 0.0011 0.0052 *** 0.0013 0.0046 *** 0.0009 Income (nonsubscrber households) -0.0005 a 0.0003-0.0007 a 0.0004-0.0005 a 0.0004-0.0006 a 0.0003-0.0009 * 0.0005-0.0006 a 0.0004 Free - - 0.9875 *** 0.1669 - - - - 0.8592 *** 0.1534 - - Pay - - 0.9683 *** 0.1668 - - - - 0.8311 *** 0.1518 - - δ (cable constant) - - - - 0.0079 0.2833 - - - - 0.0145 0.2869 δ BCH) - - - - 0.0373 0.0883 - - - - 0.0272 0.0937 δ ECH) - - - - 0.0011 0.0303 - - - - 0.0031 0.0309 δ P) - - - - 0.0011 0.0058 - - - - 0.0008 0.0057 δ MCH) - - - - 0.0000 0.0000 - - - - 0.0000 0.0000 δ WBCH) - - - - - - - - - - 0.0000 0.0002 δ WECH) - - - - - - - - - - 0.0002 0.0019 δ WMCH) - - - - - - - - - - 0.0001 0.0008 δ Income) - - - - 0.0000 0.0002 - - - - 0.0000 0.0002 ln Lkelhood -227.8573-227.6987-227.7971-218.2827-217.4627-218.2356 ln Lkelhood (zero coeffcent constrant) -563.5881-614.8215-563.5881-563.5881-614.8215-563.5881 McFadden's R2 0.5933 0.6267 0.5910 0.6093 0.6425 0.6059 Sample 513 513 513 513 513 513 ***: 1% sgnfcance **: 5% sgnfcance *: 10% sgnfcance a: 20% sgnfcance 12
Table 3 Goodness of Ft and IIA Test Results for Each Model Method IIA test Multnomal Nested Mxed Hausman Lkelhood Model 1 76.0% 76.2% 86.4% Dsmssed (5% level) Cannot be dsmssed Model 2 77.0% 76.8% 87.1% Cannot be dsmssed Cannot be dsmssed 13
Table 4. Prce and Channel Elastcty Prce Basc channel servce Specal channel servces Model 1 ε Own ε Cross ε Own ε Cross ε Own ε Cross Unsubscrbed -0.138 0.049 8.059-3.445 0.000 0.000 Cable TV -0.504 0.379 6.423-7.593 0.189-0.135 CS -0.450 0.192 7.660-3.859 0.243-0.108 Model 2 ε Own ε Cross ε Own ε Cross ε Own ε Cross Unsubscrbed -0.186 0.066 8.569-3.642 0.000 0.000 Cable TV -0.677 0.508 6.818-8.060 0.087-0.062 CS -0.603 0.259 8.115-4.111 0.111-0.050 14