*Pages 1--82 from  Microsoft Word - 21710.doc*
 FEDERAL  COMMUNICATIONS  COMMISSION 
 Media  Ownership  Working  Group 


 Consumer  Substitution 
 Among  Media 


 By  Joel  Waldfogel  September  2002 


 2002  -  3 
1
 1 
 Consumer  Substitution  among  Media 
 Joel  Waldfogel  The  Wharton  School 
 University  of  Pennsylvania 


 This  research  has  been  supported  by  the  FCC  and  has  been  conducted  by  the  author  in  his  capacity  as  a  consultant  to  the  Commission’s  Media  Ownership  Working  Group.  I  thank 
 Lisa  George  and  Felix  Oberholzer-  Gee  for  the  use  of  resources  developed  for  other  joint  work  on  related  projects,  and  I  received  helpful  comments  from  Paul  Gallant  and  Jon 
 Levy  on  an  earlier  draft.  The  work  with  Oberholzer-  Gee  was  supported  by  the  Ford  Foundation.  All  errors  are  my  own. 
2
 2 
 Consumer  Substitution  among  Media:  Executive  Summary 
 The  past  decade  has  seen  a  number  of  major  changes  in  the  availability  and  use  of 
 information  through  media  channels.  The  use  of  some  media  –  cable  television,  the 
 Internet,  and  weekly  newspapers  –  has  grown,  while  the  use  of  others  –  radio,  broadcast 
 television,  and  daily  newspapers  –  has  declined.  Our  question  is  whether  the  changes  in 
 availability  or  use  of  some  media  have  brought  about  changes  in  the  availability  or 
 consumers’  use  of  other  media,  or  whether  different  media  serve  as  substitutes  for  one 
 another  for  information  consumers. 
 This  question  is  important  because  FCC  media  ownership  policies  are  predicated 
 to  varying  degrees  on  the  extent  of  substitutability  of  media  for  various  purposes  -  news, 
 entertainment,  etc.  This  study  examines  the  extent  of  substitutability  across  media  for 
 both  of  these  purposes  using  a  variety  of  demand  and  supply  measures. 
 The  specific  tasks  of  this  study  are  twofold.  First,  we  survey  the  media  landscape 
 since  the  mid  1990s.  To  accomplish  this  we  have  assembled  aggregate  data  on  both  the 
 availability  of  media  outlets  and  their  use  over  time,  as  measured  by  listening,  viewing, 
 circulation,  etc.  across  US  local  media  markets.  We  then  turn  to  our  main  task, 
 documenting  the  extent  of  consumer  substitution  across  media.  In  Part  I  we  do  this  with 
 aggregate  data,  by  examining  the  relationship  between  changes  in  the  availability  and  use 
 across  various  media.  For  example,  has  broadcast  TV  use  declined  more  in  the  markets 
 where  Internet  use  has  increased  most  quickly? 
 In  Part  II,  we  reexamine  this  question  using  a  large  cross  sectional  data  based  with 
 information  on  individuals’  use  of  various  different  media.  These  data  allow  us  to 
 observe  whether  a  person  using  one  medium  also  uses  another,  for  news  and  other 
3
 3 
 purposes.  We  explore  substitution  with  these  data  by  asking,  for  example,  whether 
 individuals  using  daily  newspapers  more  use  television  less. 
 What  do  we  find?  Standing  back,  there  is  clearest  evidence  of  substitution 
 between  Internet  and  broadcast  TV,  both  overall  and  for  news;  between  daily  and  weekly 
 newspapers;  and  between  daily  newspapers  and  broadcast  TV  news.  There  is  also 
 evidence  of  substitution  between  cable  and  daily  newspapers,  both  overall  and  for  news 
 consumption;  between  radio  and  broadcast  TV  for  news  consumption;  and  between  the 
 Internet  and  daily  newspapers  for  news  consumption.  There  is  little  or  no  evidence  of 
 substitution  between  weekly  newspapers  and  broadcast  TV,  or  between  radio  and  either 
 Internet  or  cable.  There  is  also  some  indirect  evidence  of  substitution  in  the  greater  use 
 of  national  media  by  groups  less  targeted  by  local  media. 
 This  study  leads  to  several  conclusions.  First,  we  can  reject  the  view  that  various 
 media  are  entirely  distinct.  As  noted  above,  certain  media  appear  to  compete  with  each 
 other  for  consumers’  attention.  Second,  the  study  provides  evidence  of  substitution  by 
 consumers  between  and  among  certain  media  outlets.  It  cannot  completely  answer  the 
 question  of  whether  substitution  is  sufficiently  effective  that  all  media  should  be 
 considered  substitutes  for  news  and  information  purposes.  Existing  research,  however, 
 may  provide  useful  clues.  While  entertainment  provided  through  media  channels  is  an 
 end  in  itself,  news  makes  it  easier  for  citizens  to  know  what  is  at  stake  in  electoral 
 contests  and,  in  turn,  can  make  them  more  likely  to  vote.  If  substitution  were  complete, 
 then  the  decline  of  local  daily  newspapers  would  be  offset  by  increased  use  of  other 
 media.  The  civic  behaviors  affected  by  media  consumption  would  also  be  unaffected  by 
 changes  in  availability  or  use  of  any  particular  medium.  Yet,  existing  research  on  media 
4
 4 
 consumption  and  voting  –  reviewed  in  the  study  conclusion  –  suggests  that,  even  if 
 substitution  operates,  it  is  not  complete  in  this  sense. 
5
 5 
 Part  I:  Evidence  from  Aggregate  Data 
 The  past  decade  has  seen  a  number  of  major  changes  in  the  availability  and  use  of 
 information  through  media  channels.  The  use  of  some  media  has  grown,  for  example 
 cable  television,  the  Internet,  and  weekly  newspapers.  The  use  of  other  media,  including 
 radio,  broadcast  television,  and  daily  newspapers,  has  declined.  An  important  question 
 that  we  seek  to  address  in  this  study  is  whether  the  changes  in  availability  or  use  of  one 
 medium  bring  about  changes  in  the  availability  or  use  of  other  media.  In  short,  we  are 
 interested  in  whether  different  media  serve  as  substitutes  for  one  another.  This  is  relevant 
 to  FCC  policies  which  are  predicated  to  varying  degrees  on  the  extent  of  substitutability 
 across  media  for  various  purposes  –  news,  entertainment,  etc.  This  study  examines  the 
 extent  of  substitutability  across  media  both  for  both  of  these  purposes  using  a  variety  of 
 both  demand  and  supply  measures. 
 The  specific  tasks  of  this  study  are  twofold.  First,  we  will  survey  the  media 
 landscape  since  the  mid  1990s.  To  accomplish  this  we  have  assembled  aggregate  data  on 
 both  the  “supply”  of  media  outlets  and  the  “demand”  for  the  content  provided  through 
 these  outlets,  as  measured  by  listening,  viewing,  circulation,  etc.  across  US  local  media 
 markets.  We  begin  by  simply  documenting  levels  and  trends  within  each  medium.  Are 
 television  viewing,  radio  listening,  and  newspaper  circulation  rising  or  falling? 
 We  will  then  turn  to  our  second  broad  task,  to  document  the  extent  of  consumer 
 substitution  across  media.  We  will  address  this  question  with  two  different  sorts  of  data. 
 First,  in  Part  I  of  the  study  we  will  examine  aggregate  data  on  media  outlets  and  use  over 
 time.  In  Part  II  we  will  turn  to  a  large  individual  cross  sectional  data  set  on  media  use. 
6
 6 
 We  would  ideally  like  to  know  the  answers  to  a  number  of  questions:  First,  how 
 does  the  supply  of  outlets  in  each  medium  respond  to  supply  of  outlets  in  other  media? 
 Second,  how  does  the  demand  for  information  through  each  medium  respond  to  the 
 availability  of  outlets  in  other  media?  Finally,  how  does  the  use  of  each  outlet  relate  to 
 the  use  of  other  outlets? 
 These  are  difficult  questions  to  answer,  in  particular  to  draw  tight  causal 
 inferences.  It  would  strain  credulity,  for  example,  to  draw  substitution  inferences  from 
 the  cross  sectional  relationship  among  markets’  outlets  in  different  media.  For  example, 
 larger  markets  tend  to  have  more  outlets  in  all  media;  and  this  is  not  because  these  outlets 
 are  complements.  Rather,  it  is  because  larger  markets  make  more  outlets  viable  in  all 
 media.  An  alternative  strategy  for  drawing  substitution  inferences  is  to  use  the  time-series 
 variation  in  outlets  to  ask,  for  example,  whether  there  is  greater  decrease  in 
 television  viewing  in  markets  with  larger  increases  in  Internet  use. 
 This  strategy  has  substantial  strengths.  To  the  extent  that  outlets,  or  use,  of 
 different  media  in  each  market  are  driven  by  fixed  unobservable  factors,  examining 
 changes  allows  these  factors  to  be  “differenced  out.”  This,  in  turn,  makes  it  possible  to 
 infer  causal  relationships  among  the  use  or  availability  of  outlets  in  different  media.  At 
 the  same  time,  however,  this  strategy  does  not  overcome  all  obstacles.  If  consumers  in  a 
 place  experience  a  change  in  their  level  of  interest  in  information,  this  may  bring  about  a 
 change  in  their  use  of  multiple  media.  In  particular,  this  will  tend  to  bring  about  a 
 positive  relationship  among  the  changes  in  supply  and  demand  for  different  media,  as 
 well  as  in  their  levels.  This  sort  of  effects  makes  it  appear  that  different  media  are 
 complementary.  That  is,  this  induces  a  bias  away  from  detecting  substitution  among 
7
 7 
 media,  if  that  substitution  operates.  As  a  practical  matter,  then,  this  approach  is 
 conservative,  and  the  true  extent  of  substitution  may  be  greater  than  any  we  can 
 document. 
 Bearing  in  mind  the  limitations  of  the  approach,  the  relationships  are  still  quite 
 interesting  to  examine.  In  fact,  we  know  very  little  about  how  consumers  substitute 
 among  different  media.  Our  ignorance  is  due  to  two  factors.  First,  as  discussed  above, 
 inferring  causality  is  difficult.  Second,  the  question  has  received  little  study. 
 Researchers  and  policy  makers  have  devoted  significant  attention  to  whether  advertising 
 in  one  medium  is  a  substitute  for  advertising  in  another,  but  there  little  research  (to  my 
 knowledge)  on  whether  information  provided  through  one  medium  serves  as  a  substitute 
 for  information  provided  through  another.  1  The  question  is  important.  If  consumers 
 substitute  information  across  media,  then  the  market  for  information  may  extend  across 
 media,  raising  questions  about  regulation  of  outlets  within  media.  The  importance  of  the 
 question,  in  conjunction  with  our  general  ignorance  about  the  answer,  amply  justify  the 
 adoption  of  sensible  approaches  to  studying  the  question.  This  is  the  spirit  in  which  we 
 proceed. 


 II.  Theoretical  Background 
 Consumers  seek  information  (news  and  weather,  etc)  and  entertainment  from 
 media  products.  Suppose  there  are  five  relevant  media  (television,  radio,  newspapers, 
 etc.),  and  each  can  be  divided  into  news  and  entertainment  components.  Then  there  are 
 ten  products  to  examine.  The  usual  economic  way  to  conceive  of  this  problem  is  through 
 consumer  theory:  each  consumer’s  demand  for  each  of  the  ten  products  depends 
8
 8 
 (negatively)  on  the  price  of  the  own  product  and,  if  the  other  products  are  substitutes, 
 negatively  on  the  other  products’  prices.  While  this  characterization  is  very  helpful  for 
 organizing  ideas,  it  is  not  literally  useful  in  our  context.  Content  is  unpriced  (free)  on 
 radio,  broadcast  television,  and  (for  the  most  part)  on  the  Internet.  And  while  newspapers 
 have  subscriber  prices,  these  prices  vary  little  across  markets,  largely  because  newspapers 
 rely  primarily  on  advertising  rather  than  subscriptions  for  revenue. 
 In  this  context  the  more  useful  variation  is  in  availability  of  outlets  within  a 
 medium  rather  than  in  the  price  that  consumers  pay  for  content.  It  is  well  known  that 
 larger  markets  have  more  local  media  outlets  within  each  medium  than  small  markets 
 (see  Berry  and  Waldfogel,  1999a;  Waldfogel,  1999,  2001;  George  and  Waldfogel,  2000). 
 There  are  also  changes  over  time  in  the  numbers  of  outlets  (see  Berry  and  Waldfogel, 
 2001  for  evidence  about  radio  following  deregulation).  Conceptually,  the  change  from 
 having  no  outlets  in  a  medium  to  having  one  is  a  price  reduction  from  infinite  levels  to 
 some  (perhaps  unmeasured)  finite  amount.  Changes  among  positive  numbers  of  outlets, 
 from  say  one  to  two  in  a  medium  and  a  market,  are  similar  in  spirit.  Such  changes 
 enhance  the  attractiveness  of  each  consumer’s  most  attractive  alternative  and  in  that  sense 
 can  be  viewed,  metaphorically  at  least,  as  reducing  the  “price”  of  consumption  in  that 
 medium.  Given  our  context,  then,  we  propose  to  study  substitution  chiefly  by  examining 
 the  relationship  between  use  in  one  medium  and  availability  of  outlets  in  other  media.  If 
 two  media  are  substitutes,  then  greater  availability  of  outlets  in  one  medium  will  reduce 
 the  use  of  the  other  medium. 
 While  this  question  is  not  much  studied,  there  is  some  evidence  in  existing 
 research  on  aspects  of  substitution.  Berry  and  Waldfogel  (1999a)  provide  a  study 


 1  We  discuss  related  research  in  part  I,  section  II  below. 
9
 9 
 documenting  high  levels  of  substitutability  across  outlets  within  the  commercial  radio 
 medium.  Additional  outlets  draw  most  of  their  listeners  from  other  stations,  suggesting 
 that  stations  are  rather  close  substitutes  for  one  another.  Berry  and  Waldfogel  (1999b) 
 examine  the  extent  of  substitutability  of  public  and  commercial  radio  broadcasting  news, 
 classical  music,  and  jazz.  Although  that  study,  too,  is  within  medium,  it  is  relevant  here 
 both  because  it  documents  substitutability  and  also  because  its  approaches  are  adopted 
 here.  That  study  infers  substitutability  of  public  and  commercial  radio  stations  from  a) 
 the  relationship  between  public  and  commercial  radio  outlets  and  b)  the  relationship 
 between  commercial  listening  and  the  numbers  of  commercial  and  public  radio  stations  in 
 the  format.  The  study  documents  that  markets  with  more  commercial  classical  stations 
 have  greater  commercial  classical  tendencies,  while  commercial  classical  listening  is 
 lower  in  markets  with  more  public  classical  stations,  all  else  constant.  The  latter  result 
 suggests  that  listeners  view  public  and  commercial  classical  radio  stations  as  substitutes. 
 Other  existing  work  provides  evidence  of  substitutability  of  different  types  of 
 media  outlets.  George  and  Waldfogel  (2002)  document  that  readers  view  national 
 newspapers  (the  New  York  Times)  as  a  substitute  for  local  daily  newspapers.  In  markets 
 with  greater  increases  in  Times  circulation,  there  is  a  larger  decrease  in  local  paper 
 circulation  in  zipcodes  containing  educated  persons  targeted  by  the  Times.  This  indicates 
 that  national  papers  are  substitutes  for  local  ones. 
 Finally,  there  is  some  indirect  evidence  of  cross-  medium  substitution.  It  is  well 
 documented  in  a  variety  of  studies  that  there  are  more  local  media  outlets  (and 
 programming)  in  larger  markets.  Moreover,  isolated  groups  face  more  outlets  or 
 programming  of  interest  as  they  are  in  markets  with  more  persons  sharing  their 
10
 10 
 preferences  (see  Waldfogel,  1999,  2001;  Siegelman  and  Waldfogel,  2001;  George  and 
 Waldfogel,  2000).  In  light  of  this,  if  consumers  can  use  nonlocal  media  to  substitutes  for 
 the  relative  paucity  of  local  media,  then  we  would  see  greater  use  of  nonlocal  media 
 (such  as  the  Internet  or  cable  television)  in  markets  where  consumers  were  more  isolated. 
 Sinai  and  Waldfogel  (2001)  documents  that  in  spite  of  their  lower  tendency  to  connect  to 
 the  Internet  overall,  blacks  are  more  likely  to  connect  as  they  are  more  isolated  in  markets 
 with  proportionately  few  blacks.  This  suggests  that  isolated  information  consumers 
 substitute  nonlocal  media  for  local  ones. 
 Finally,  there  is  evidence  elsewhere  to  support  the  supposition  that  media  matter 
 for  civic  participation.  Oberholzer-  Gee  and  Waldfogel  (2001)  document  that  the 
 presence  of  group-  targeted  media  outlets  (radio  stations  and  weekly  newspapers)  affect 
 the  tendency  for  blacks  to  vote.  George  and  Waldfogel  (2002)  find  that  in  markets  with 
 greater  distribution  of  the  New  York  Times,  not  only  do  the  targeted  (educated) 
 consumers  purchase  local  papers  less.  They  also  come  to  vote  less  in  local  elections  as 
 Times  penetration  grows. 


 III.  Data 
 Our  basic  data  set  in  this  part  of  the  study  is  a  panel  with  repeated  observations  on 
 the  number  of  outlets  in  each  medium  in  each  market,  as  well  as  measures  of  the  use  of 
 each  medium.  Our  data  come  from  five  sources  (BIA’s  Media  Access  Pro,  Duncan’s 
 American  Radio,  the  Current  Population  Survey’s  Computer  and  Internet  Use 
 Supplements,  Burrelle’s  Media  Directory,  and  the  U.  S.  Census)  that  collect  data  in 
 different  ways.  Still,  we  ultimately  have  multiple  years  of  use  data,  by  market,  for 
11
 11 
 broadcast  television,  cable,  radio,  daily  newspapers,  weekly  newspapers,  and  the  Internet. 
 We  have  multiple  years  of  outlet  availability  data  for  broadcast  TV,  radio,  and  daily  and 
 weekly  newspapers.  We  describe  these  data  in  detail,  and  by  medium,  below. 


 1.  Television 
 The  television  data  are  from  BIA  Media  Access  Pro.  BIA  provides  supply  data, 
 the  total  number  of  stations  in  each  market,  by  type  (UHF  vs.  VHF,  network  vs. 
 independent  vs.  public).  BIA  also  provides  market  aggregate  television  viewing,  derived 
 from  Nielsen  data,  for  each  of  210  DMAs.  Aggregate  HUT  (households  using  television) 
 is  provided  by  year,  1994-  2000.  BIA  also  provides  daypart  specific  viewing  data  (based 
 both  on  HUT  and  total  viewing)  for  a  recent  window  of  more  frequent  observations 
 (quarterly,  November  1999  to  July  2001). 


 2.  Daily  and  Weekly  Newspapers 
 We  document  the  number  of  daily  newspapers  available  in  each  DMA  using 
 Burrelle’s  Media  Directory  for  1993  and  1999.  Burrelle’s  also  provides  a  measure  of 
 total  circulation  for  each  daily.  We  aggregate  across  products  in  each  DMA  to  arrive  at 
 total  circulation  for  dailies  and  weeklies  in  each  market. 


 3.  Radio 
 We  have  data  on  the  number  of  radio  stations,  and  their  programming  formats,  in 
 each  of  roughly  250  Arbitron  markets,  in  1993  and  1997.  These  data  are  derived  from 
 Duncan’s  American  Radio  (1994,  1998),  based  on  Arbitron  listening  data  as  well  as 
12
 12 
 Duncan’s  classification  of  programming  formats.  Notably  for  our  purpose,  we  can 
 distinguish  news  stations  (“  News”  and  “News/  Talk”)  from  other  sorts  of  stations. 
 Duncan  also  provides  AQH  listening  data,  both  for  individual  stations,  as  well  as 
 for  entire  markets.  We  employ  the  data  based  on  the  Arbitron  Spring  surveys  in  the 
 respective  year.  Arbitron  markets  do  not  map  neatly  into  DMAs,  so  we  have  undertaken 
 effort  to  create  a  mapping  that  allows  the  radio  data  to  be  linked  with  data  at  the  DMA 
 level. 


 4.  Internet  Use 
 In  1997,  1998,  and  2000  the  Current  Population  Survey  includes  questions  about 
 Internet  use.  We  focus  on  the  following  five  questions,  available  in  all  three  of  these 
 surveys:  1)  whether  the  individual  has  Internet  access  at  home,  2)  whether  the  individual 
 has  Internet  access  at  work,  3)  use  of  the  Internet  for  news,  4)  use  of  the  Internet  for 
 information,  and  5)  use  of  the  Internet  for  email. 
 For  this  medium  we  cannot  clearly  distinguish  availability  of  outlets  from  use  of 
 the  medium.  Thus,  we  really  have  only  a  “demand”  measure  for  this  medium. 
 Moreover,  the  geographic  identifier  in  these  data  is  the  MSA.  We  have  created  a 
 mapping  that  allows  CPS  data  to  be  linked  to  DMA  data. 


 5.  Cable  Takeup 
 BIA  reports  the  percent  of  households  with  cable,  by  year,  1994-  2000,  at  the 
 DMA  level  of  geography. 
13
 13 
 6.  Demographics 
 We  obtain  demographic  data  from  various  sources,  derived  from  the  1990  US 
 Census. 


 IV.  Trends  within  each  Medium 
 In  this  section  we  document  trends  in  media  supply  and  use  in  each  medium. 
 Following  that,  we  examine  the  comovements  in  media  supply  and  use  in  our  attempt  to 
 learn  about  substitution  among  media.  The  latter  exercise  requires  linking  data  on  outlets 
 and  use  of  various  media.  While  simple  in  principle,  this  task  is  quite  complicated  in 
 practice  because  the  data  on  various  media  are  collected  for  different  levels  of  geography. 
 For  example,  the  BIA  television  and  cable  data,  as  well  as  the  Burrelle’s  newspaper  data, 
 are  reported  by  DMA.  The  CPS  data  on  Internet  use  are  individual-  level  data,  and  the 
 individual  records  include  MSA  identifiers.  The  Arbitron  radio  data  reported  in 
 Duncan’s  American  Radio,  are  reported  by  Arbitron  metro  area.  These  are  similar,  but 
 not  identical,  to  MSAs.  We  have  had  to  construct  mappings  linking  MSAs  and  Arbitron 
 metro  areas  to  DMAs. 
 The  linked  multi-  media  dataset  covers  a  subset  of  geographic  areas  for  which 
 linking  is  feasible  (roughly  140  of  210  DMAs).  To  address  concerns  about  whether  this 
 sample  is  representative,  we  compare  the  trends  in  each  medium’s  use  in  the  linking 
 sample  against  the  measures  for  the  medium’s  full  sample. 


 1.  Television 
14
 14 
 BIA  reports  two  measures  of  the  number  of  stations  available  in  a  DMA,  the  total 
 number  of  stations  and  the  number  of  rated  stations.  Neither  variable  is  reported 
 historically.  BIA  also  provides  two  breakdowns  of  the  number  of  stations:  UHF/  VHF 
 and  network/  independent/  public.  The  sums  of  each  of  these  two  breakdowns  yield 
 approximately  the  number  of  rated  stations.  Even  though  they  do  not  yield  precisely  the 
 current  number  of  rated  stations,  we  still  view  the  change  in  the  total  number  of  UHF  and 
 VHF  stations,  and  the  change  in  (network  +  independent  +  public)  stations  over  time  as 
 reasonable  reflections  of  the  time  pattern  of  local  broadcast  TV  station  availability.  Table 
 1  reports  trends  in  the  average  numbers  of  stations  of  each  type  in  each  DMA.  The  top 
 panel  includes  the  full  sample  2  while  the  bottom  panel  includes  only  the  DMAs  in  the 
 linked  sample.  Trends  in  both  are  quite  similar:  the  total  number  of  stations  has 
 increased  slightly,  from  8.4  in  1995  to  9.7  in  2000  (network  +  indep.  +  public  measure  in 
 the  linked  sample).  UHF  stations  have  increased  more  quickly  than  VHF.  The  number 
 of  network-  affiliated  stations  has  increased  as  networks  outside  the  traditional  three 
 majors  have  increased  their  reach.  Almost  by  construction,  this  decreases  the  number  of 
 independents.  The  number  of  public  stations  has  held  steady.  These  trends  are  clearly 
 visible  in  both  samples. 
 Table  2  reports  patterns  of  television  and  cable  use.  The  percent  of  households 
 with  cable  has  increased  steadily,  from  63  percent  in  1994  to  68  percent  in  2000  (linked 
 sample).  We  have  an  overall  annual  measure  of  television  viewing  (excluding  cable) 
 since  1994.  The  measure  is  “households  using  television.”  We  also  have  quarterly 


 2  The  full  sample  includes  207  of  the  210  DMAs  in  the  US.  Three  DMAs  are  excluded  because  of  missing 
 data. 
15
 15 
 measures  of  television  viewing  by  daypart,  from  November  1999  to  July  2001.  See  Table 
 3. 
 Television  viewing  (not  including  cable  channels)  has  declined  steadily  over  the 
 same  period,  from  37.3  percent  in  1994  to  36.8  percent  in  2000.  The  quarterly  measures, 
 in  Table  3,  indicate  both  a  strong  seasonal  pattern  –  higher  use  in  November  and 
 February,  lower  in  May  and  July  –  and  the  same  secular  decline  as  the  annual  HUT 
 figures.  This  decline  appears  both  for  the  dayparts  associated  with  news  as  well  as  all 
 others.  Because  these  rating  figures  do  not  include  time  spent  watching  cable  networks, 
 they  do  not  capture  overall  viewing  behavior. 


 2.  Radio 
 As  with  television,  the  mid  1990s  saw  an  increase  in  the  number  of  outlets,  as 
 well  as  increase  in  the  number  of  distinct  programming  formats  on  the  air  across  230  US 
 markets.  There  were  an  average  of  21.7  stations  available  in  each  market  in  1993  and 
 24.8  in  1997.  Of  these,  15.0  broadcast  from  inside  the  metro  area  in  1993  and  16.6  in 
 1997. 
 Table  4  shows  the  distributions  of  programming  formats  in  1993  and  1997.  The 
 average  number  of  formats  available  in  each  market  grew  from  11.6  in  1993  to  15.0  in 
 1997  (see  Berry  and  Waldfogel,  2001).  AQH  listening  declined  steadily  from  16.8 
 percent  in  1993  to  15.7  percent  in  1998. 
 We  can  separately  document  the  evolution  of  commercial  news  radio  3  .  As  Table 
 5  shows,  the  average  number  of  inside  (total)  news  stations  was  0.46  (0.78)  in  1993  and 


 3  These  data  exclude  public  radio.  See  Berry  and  Waldfogel  (1999b)  for  information  about  public  and 
 commercial  radio  in  1993. 
16
 16 
 0.55  (0.94)  in  1997  (using  the  overall  sample).  4  Perhaps  more  important  than  how  many 
 is  the  question  of  inside  news  station  presence.  Among  the  230  markets,  the  number  of 
 markets  without  an  inside  news  station  declined  from  142  in  1993  to  135  in  1997. 
 Interestingly,  28  markets  went  from  having  no  inside  news  stations  in  1993  to  having  one 
 or  more  in  1997,  while  21  markets  went  from  having  one  in  1993  to  having  none  in  1997. 
 Table  5  also  shows  how  the  number  of  formats  available  in  each  market  increased 
 from  11.6  in  1993  to  15.0  in  1997,  consistent  with  Table  4. 


 3.  Newspapers 
 The  average  numbers  of  daily  newspapers  per  DMA  declined  from  8.05  in  1993 
 to  7.85  in  1999,  while  over  the  same  period  the  number  of  weekly  newspaper  outlets 
 grew  from  44.18  to  46.21.  Average  daily  circulation  fell  from  309,579  in  1993  to 
 295,950  in  1999.  Weekly  newspaper  circulation,  on  the  other  hand,  grew  from  531,693 
 in  1993  to  728,804  in  1999.  Similar  patterns  appear  in  the  linked  sample.  See  Table  6. 


 4.  Internet 
 We  lack  data  on  the  availability  of  the  Internet,  either  on  the  availability  of  ISPs 
 or,  more  important,  on  the  number  of  local  and  national  sites  of  interest  to  users.  On  the 
 first  question,  since  we  are  examining  metropolitan  areas,  we  assume  that  connection 
 services  are  available  everywhere  in  the  sample  by  1997  (see  Downes  and  Greenstein, 
 2002).  We  have  Internet  connection  data  for  1997,  1998,  and  2000.  The  Internet  was 
 virtually  nonexistent  prior  to  1994,  so  we  can  document  the  entire  history  1994-  2000. 


 4  By  “inside”  we  mean  stations  broadcasting  from  inside  a  metro  area. 
17
 17 
 As  Table  7  shows,  Internet  use  has  grown  rapidly,  to  15.1  percent  of  households 
 with  home  access  in  1997  to  30.3  percent  in  1998,  and  to  45.9  in  2000.  The  Table  also 
 shows  the  fraction  of  individuals  using  the  Internet  for  email,  news,  and  information,  the 
 latter  two  representing  ways  in  which  Internet  use  might  substitute  for  the  use  of 
 traditional  media.  Patterns  are  virtually  identical  in  the  linked  sample  of  markets. 


 5.  Summary 
 The  mid  1990s  has  been  a  period  with  a  shifting  media  landscape.  Use  of  radio, 
 daily  newspapers,  and  traditional  television  have  declined,  while  the  use  and  availability 
 of  other  media  have  increased.  Cable  television  use  has  grown,  as  has  the  availability 
 and  use  of  weekly  newspapers  and  the  Internet.  These  comovements  may  reflect 
 substitution,  but  the  separate  trends  in  each  media  cannot  answer  this.  The  question, 
 addressed  in  the  next  section,  is  whether  one  medium  grows  in  the  places  where  others 
 decline. 


 V.  Substitution  across  Media 
 1.  Availablity  of  Media  Outlets 
 It  is  well  documented  that  larger  markets  offer  more  local  media  products.  This 
 relationship  holds  for  radio,  television,  daily  newspapers,  and  even  to  some  extent  for  the 
 websites  on  the  Internet.  5  These  relationships  are  clearly  visible  in  our  data  as  well. 
 Figure  1  shows  the  relationship  between  the  numbers  of  media  outlets  in  a  DMA  and 
18
 18 
 population  for  radio,  television,  daily  newspapers,  and  weekly  newspapers.  The  figure, 
 on  a  log  scale  for  ease  of  interpretation,  clearly  shows  positive  relationships  between 
 market  size  and  the  number  of  local  outlets  in  each  medium.  In  this  figure  we  use  radio 
 and  television  outlets  in  1997  and  newspaper  outlets  in  1999.  The  cross  sectional 
 relationships  for  other  years  are  nearly  identical. 


 2.  Use  of  Media  and  Availability  of  Media  Outlets 
 Our  data  allow  us  to  examine  the  relationships  between  outlet  availability  and 
 medium  use  in  two  ways.  First,  we  can  examine  the  cross  sectional  relationships  directly. 
 Table  8  presents  regressions  of  our  use  measures  on  outlets  for  the  four  media  with  outlet 
 data  (radio,  television,  and  daily  and  weekly  newspapers).  Coefficients  are  positive  and 
 significant  for  all  media.  These  regressions  are  vulnerable  to  the  concern  that  the  number 
 of  outlets  is  endogenous,  or  that  more  outlets  enter  in  markets  with  greater  tastes  for  the 
 outlet  in  question.  There  are  two  obvious  solutions  to  this  problem,  instrumental 
 variables  (IV)  and  panel  data.  First,  the  IV  solution  to  this  problem  is  an  instrument 
 determining  the  tendency  for  outlets  to  enter  but  that  has  no  direct  effect  on  the  tendency 
 to  join  the  audience.  Market  size,  measured  by  population,  is  a  natural  choice  for  local 
 media  and  one  that  has  been  employed  elsewhere  in  media  entry  studies.  The  latter  half 
 of  Table  8  reports  regressions  of  use  on  available  for  the  four  media,  now  instrumenting 
 outlets  with  a  population  measure.  The  coefficients  of  interest  are  again  positive, 
 indicating  that  additional  outlets  within  a  medium  attract  persons  to  their  audiences. 


 5  See  Berry  and  Waldfogel  (1999b)  and  Waldfogel  (1999)  for  radio  evidence,  Waldfogel  (2001)  for 
 television  evidence,  George  and  Waldfogel  (2000)  for  evidence  on  daily  newspapers,  and  Sinai  and  Waldfogel  (2001)  for  evidence  about  the  Internet. 
19
 19 
 For  Internet  and  cable  we  have  no  direct  measure  of  local  outlets.  Still,  we  can 
 examine  the  relationship  between  use  and  market  size.  Table  9  shows  the  relationship 
 between  medium  use  and  market  size  for  all  of  the  media.  Interestingly,  while  the  local 
 media  have  positive  and  significant  relationships,  cable  and  Internet  use  have  essentially 
 no  relationship  with  market  size.  For  these  media,  which  do  not  originate  locally,  market 
 size  has  a  different  natural  interpretation.  Rather  than  proxying  the  amount  of  content 
 available  locally  within  the  medium,  market  size  here  measures  the  availability  of  content 
 in  other  media. 
 The  difference  between  the  relationship  between  medium  use  and  market  size  in 
 local  and  nonlocal  media  thus  provides  our  first  piece  of  evidence,  albeit  indirect,  about 
 substitution.  While  the  tendency  to  use  local  media  increases  in  market  size,  the 
 tendency  to  use  the  national  media,  relative  to  the  tendency  to  use  local  media,  is  larger 
 as  markets  are  smaller.  This  suggests  that  in  smaller  markets,  Internet  and  cable  serve  as 
 substitutes  for  newspapers,  local  television,  and  radio.  One  obtains  similar  but  even 
 stronger  relationships  if  one  examines  Internet  use  relative  to  television  use.  Define  
I_  TV 
 as  the  Internet  use  –  broadcast  TV  use  in  a  market.  Then  a  regression  of  
I_  TV  on  log 
 population  yields  a  negative  and  significant  coefficient. 
 We  can  also  approach  the  question  of  whether  outlets  affect  use  employing  the 
 within-  market  variation  in  the  number  of  outlets  over  time.  Table  10  presents  market 
 fixed  effects  regressions  of  use  on  outlets,  including  year  dummies  as  well  as  market 
 fixed  effects.  We  obtain  positive  and  significant  coefficients  on  outlets  in  the  television 
 and  daily  newspaper  regressions.  Coefficients  are  insignificant  in  the  other  regressions. 
 One  should  not  infer  the  absence  of  relationships  from  the  absence  of  results  here.  There 
20
 20 
 simply  is  not  a  great  deal  of  variation  across  markets  in  the  change  in  the  number  of 
 outlets  over  time,  making  it  difficult  to  infer  the  relationships  in  question. 
 We  take  the  weight  of  evidence  in  this  subsection  to  indicate  that  additional 
 outlets  in  a  medium  attract  more  of  the  population  into  the  medium’s  local  audience.  In 
 addition,  the  difference  between  the  relationship  between  market  size  and  medium  use  for 
 national  and  local  media  suggests  substitution. 


 3.  Evidence  on  Substitution  across  Media  from  Cross-  Media  Relationships 
 We  seek  to  measure  whether  different  media  are  substitutes  for  one  another  using 
 an  approach  related  to  our  approach  (above)  to  whether  outlets  attract  use  within  media. 
 Here,  by  contrast,  the  question  is  whether  the  availability  of  outlets  in  other  media 
 reduces  the  use  of  this  medium.  There  are  a  variety  of  approaches  one  might  adopt  for 
 this  question  in  principle.  For  example,  one  might  examine  the  cross  section  of  markets 
 at  a  point  in  time,  asking  whether  markets  with  more,  say,  daily  newspapers,  have  lower 
 use  of  some  other  media.  The  endogeneity  problems  mentioned  above  are  severe  here. 
 Larger  markets  are  going  to  have  more  outlets  in  all  media,  giving  rise  to  positive 
 correlations  among  media  outlet  availability  as  well  as  use.  Indeed,  Table  11  reports  the 
 cross  sectional  correlations  significant  at  the  10  percent  level  or  higher  among  use  and 
 outlet  measures,  and  they  are  all  positive  (and  most  are  significant  at  the  5  percent  level). 
 In  principle  one  could  address  this  problem  with  instruments.  This  approach 
 requires  an  exogenous  source  of  variation  in  the  supply  of  each  medium’s  outlets.  Note 
 that  population,  while  it  works  for  each  individual  medium,  cannot  simultaneously  be 
21
 21 
 employed  for  multiple  media.  Since  we  lack  instruments  for  each  medium,  this  approach 
 is  not  available. 
 We  do  have  longitudinal  data,  however.  While  the  number  of  outlets  in  each 
 medium  is  endogenous  to  market  size,  the  change  in  the  number  of  outlets  is  not  so 
 clearly  endogenous.  Thus,  we  propose  to  draw  inferences  about  substitution  from  the 
 correlations  between  the  changes  in  the  availability  and  use  of  outlets  in  various  media. 
 This  approach  is  in  principle  vulnerable  to  analogous  endogeneity  concerns  (i.  e.  the  kind 
 of  market  with  a  growing  number  of  TV  stations  is  likely  to  be  growing  in  other 
 dimensions  as  well  that  simultaneously  lead  to  growth  in  outlets  in  other  media). 
 Fortunately,  we  can  ascertain  the  direction  of  likely  bias:  these  concerns  lead  to  positive 
 relationships.  Hence,  any  evidence  of  negative  correlations  is  more  strongly  suggestive 
 of  substitution. 
 Table  12  reports  the  statistically  significant  correlations  among  changes  in  supply 
 and  use  of  outlets  in  various  media.  A  striking  feature  of  the  Table,  in  contrast  to  the 
 correlations  among  levels,  is  the  presence  of  negative  –  as  well  as  positive  –  correlations. 
 The  following  correlations  are  negative  and  significant  at  the  10  percent  level  or  higher: 
 Radio  outlets  and  cable  use, 
 Radio  news  outlets  and:  Internet  use,  overall  and  for  news, 
 Daily  paper  outlets  and:  weekly  paper  outlets  and  cable  use,  and 
 Daily  paper  use  and:  and  cable  use. 
 The  largest  and  most  significant  negative  relationships  in  the  Table  are  between  daily  and 
 weekly  newspapers. 
22
 22 
 One  shortcoming  of  the  direct  correlations  is  that  they  cannot  simultaneously 
 account  for  changes  in  multiple  media.  For  example,  if  the  numbers  of  outlets  in  both 
 radio  and  television  are  changing  similarly,  then  their  correlations  with,  say,  newspaper 
 use  will  both  reflect  the  effects  of  both  the  change  in  radio  and  the  change  in  television. 
 One  way  around  this  problem,  in  principle,  is  to  employ  a  multiple  regression.  That  is, 
 we  can  regress,  say,  the  change  in  daily  newspaper  use  on  the  changes  in  the  numbers  of 
 outlets  of  each  other  medium.  The  coefficient  on  each  variable  shows  the  relationship 
 between  that  variable  and  the  dependent  variable,  holding  other  variables  constant. 
 While  appealing  in  principle,  this  approach  has  a  shortcoming  that  merits 
 discussion:  if  the  explanatory  variables  are  highly  correlated,  then  it  is  difficult  to  isolate 
 the  effect  of  each  variable  separately.  This  said,  we  proceed  with  three  sorts  of 
 regressions.  First,  we  present  regressions  of  changes  in  the  numbers  of  outlets  in  each 
 medium  on  changes  in  the  numbers  of  outlets  in  each  of  the  other  media  (“  outlets  on 
 outlets”).  This  regression  can  pick  up  substitution  in  the  sense  that  if  one  medium 
 expands  in  an  area  sufficiently  to  not  only  draw  audience  from  another  but  cause  exit  (or 
 forestall  additional  entry)  in  another,  then  we  can  detect  substitution  as  a  negative  effect. 
 Second,  we  run  regressions  of  the  change  in  the  share  of  population  using  each  medium 
 on  the  changes  in  the  numbers  if  outlets  in  each  medium  (“  use  on  outlets”).  This 
 regression  can  pick  up  substitutability  short  of  cross  effects  that  induce  exit  and  is,  a 
 priori,  the  most  appealing  approach.  However,  recall  that  we  do  not  have  outlets 
 measures  for  all  media.  For  cable  and  Internet  we  have  only  use  measures;  and  we  would 
 like  to  use  the  change  in  these  media  as  explanatory  variables.  Third,  we  run  regressions 
23
 23 
 of  the  change  in  the  share  of  population  using  each  medium  on  the  changes  in  the  share 
 of  population  using  each  other  medium  (“  use  on  “use”). 
 Bearing  all  of  our  concerns  in  mind,  we  proceed  to  Tables  13-  15.  Table  13 
 presents  regressions  of  the  changes  in  number  of  outlets  in  each  medium  on  changes  in 
 the  numbers  of  outlets  in  other  media  (for  which  outlet  data  are  available).  The  only 
 consistent  relationship  to  emerge  from  this  table  is  the  relationship  between  daily  and 
 weekly  newspapers.  Both  coefficients  are  statistically  significant.  A  one-  paper  decrease 
 in  dailies  is  associated  with  an  eight-  paper  increase  in  weeklies.  The  reverse  relationship 
 is  not  entirely  symmetric:  a  one-  paper  decrease  in  weeklies  is  accompanied  by  a  0.02 
 paper  reduction  in  dailies.  For  each  regression  we  report  a  test  of  the  hypothesis  that  all 
 cross  effects  are  zero.  We  reject  this  hypothesis  only  for  dailies  and  weeklies. 
 Table  14  proceeds  to  the  relationship  between  the  change  in  use  of  each  medium 
 and  the  change  in  the  number  of  outlets  in  the  medium  as  well  as  other  media.  The  own-effects 
 are  fairly  consistently  positive.  The  only  negative  cross-  effect  significant  at  the  5 
 percent  level  is  the  coefficient  on  radio  stations  in  the  cable  use  regression  (column  6).  If 
 one  adopts  a  more  liberal  cutoff  (a  negative  coefficient  with  a  t-  statistic  exceeding  1  in 
 absolute  value),  the  relationships  suggesting  substitution  are: 
 TV  stations  and:  cable  use,  Internet  news  use 
 Radio  stations  and:  TV  viewing,  daily  and  paper  circulation,  cable  use; 
 Radio  news  stations  and  Internet  news  use; 
 Daily  papers  and:  radio  listening,  cable  use; 
 Weekly  papers  and  radio  listening. 
24
 24 
 Based  on  results  in  Table  14,  there  is  some  reason  to  doubt  the  no  substitution  view. 
 Formally,  however,  we  reject  the  zero  cross  effects  hypothesis  only  for  the  cable  and 
 Internet  news  use  regressions. 
 Table  15  reports  the  use  on  use  regressions.  The  only  negative  coefficients  that 
 are  significant  by  conventional  standards  are  the  coefficient  on  the  change  in  cable  use  in 
 the  daily  paper  regression  and  its  inverse  (the  coefficient  on  the  change  in  daily  paper  use 
 in  the  cable  use  regression).  Other  coefficients  are  suggestive  of  substitution.  Using  the 
 liberal  cutoff  of  1  (in  absolute  value)  for  t-  statistics,  one  can  see  suggestive  evidence  of 
 substitution  in  the  following  relationships: 
 Cable  use  and  weekly  papers, 
 Radio  news  use  and  TV  use, 
 Radio  news  use  and  Internet  news  use, 
 Cable  use  and  weekly  paper  use,  and 
 Internet  news  use  and  TV  use. 
 In  spite  of  these  suggestions  of  substitution,  we  formally  reject  the  hypothesis  of  zero 
 cross  effects  for  cable  use  only. 


 VI.  Conclusion  to  Part  I 
 We  conclude  our  analysis  of  the  aggregate  data  with  the  observation  that  there  is 
 some  evidence  of  consumer  substitution  across  media.  This  raises  questions  about 
 whether  media  are  sensibly  viewed  in  isolation  for  policy  purposes.  We  delve  deeper 
 into  this  question  with  our  analysis  of  individual  data  on  media  use  in  Part  II. 
25
 25 
 Part  II:  Evidence  from  Individual  Data  and  Study  Conclusion 
 Do  consumers  view  different  media  as  substitutes  for  one  another?  This  study 
 examines  this  question  using  a  large  cross  sectional  data  based  with  information  on 
 individuals’  use  of  various  different  media.  This  study  is  a  companion  to  the  aggregate 
 data  study  in  Part  I. 
 The  advantage  of  these  micro  data  is  that  they  literally  allow  us  to  observe 
 whether  a  person  using  one  medium  also  uses  another.  The  dataset  is  large  (nearly 
 180,000  observations)  and  detailed,  containing  information  about  both  overall  use  and 
 use  for  news  in  particular.  These  data  therefore  allow  us  to  explore  substitution  by  asking 
 whether,  for  example,  individuals  using  the  newspapers  more  use  television  less.  Or 
 whether  individuals  using  more  radio  news  watch  less  television  news,  etc.  Despite  the 
 substantial  appeal  of  these  data,  consumption  data  have  the  drawback  that  individuals 
 who  use  one  medium  heavily  may  also  use  others  heavily.  Indeed,  in  the  companion 
 study  the  cross  sectional  relationship  among  the  uses  of  different  media  (across  markets) 
 revealed  strong  positive  correlations  masking  possible  substitution. 
 Our  strategies  for  overcoming  these  problems  are  threefold:  First,  we  will  control 
 for  individual  characteristics.  Second,  we  will  examine  not  only  the  use  of  each  medium 
 (and  the  use  for  news  vs.  other  motivations)  but  also  the  use  of  each  medium  for  news 
 relative  to  its  use  for  other  purposes.  Third,  we  will  also  employ  an  indirect  approach  to 
 testing  for  substitution  that  avoids  these  problems:  In  particular  we  will  ask  whether 
 persons  facing  little  local  media  content  of  interest  turn  instead  to  national  media  (cable 
 and  Internet). 
26
 26 
 We  proceed  as  follows.  First,  we  describe  the  data.  Then  we  descibe  patterns  of 
 use  of  each  medium.  Third,  we  examine  substitution.  Then  we  will  summarize  in  a  way 
 that  synthesizes  results  from  the  companion  study  using  aggregate  data. 


 I.  Data 
 Data  for  this  part  of  the  study  are  drawn  from  Scarborough  Research.  The  version 
 of  the  Scarborough  Prime  Next  dataset  employed  in  the  study  has  information  on  the 
 media  and  product  consumption  patterns  for  each  of  178,784  individuals  surveyed  in  the 
 latter  half  of  1999  and  the  first  half  of  2000  in  66  large  US  markets  (see 
 http://  www.  scarborough.  com/  primenext/).  The  Primenext  dataset  has  information  at  the 
 individual  level  on  the  use  of  multiple  media. 
 The  Scarborough  data  set  contains  a  large  number  of  questions  about  media  use. 
 Here  we  describe  the  available  measures,  by  medium. 
 1.  Newspapers.  We  have  a  binary  variable  indicating  whether  a  person  reads 
 any  daily  newspaper.  In  addition  respondents  report  which  newspaper  section 
 they  regularly  read  (news,  business,  classifieds,  comics,  entertainment,  food, 
 sports,  TV  or  radio  listings). 
 2.  Television.  Respondents  report  which  half  hours  of  television  they  watched 
 on  each  of  11  networks  during  one  week  during  the  1999-  2000  television 
 season.  6  We  aggregate  these  half  hours  in  various  ways,  including  to  total 
 half  hours.  In  addition,  by  linking  the  Scarborough  data  with  local  television 


 6  The  eleven  networks  are  ABC,  CBS,  Fox,  NBC,  PAX,  Univision,  Telemundo,  WB,  UPN,  Independent, 
 and  PBS. 
27
 27 
 schedules  for  the  markets,  we  are  able  to  construct  measures  of  local  and 
 national  news  viewing  (between  4  and  12  PM)  for  individuals  in  the  sample.  7 
 3.  Cable  and  Satellite.  Respondents  report  whether  they  have  cable.  If  they 
 have  cable,  they  report  what  sort  of  service  they  have  (basic,  expanded  basic, 
 premium).  In  addition,  respondents  indicate  whether  they  have  satellite 
 service.  Respondents  also  report  which  cable  channels  they  regularly  watch. 
 We  calculate  the  number  of  cable  channels  regularly  viewed  by  sample 
 individuals  as  a  crude  measure  of  the  intensity  of  cable  use.  Because  some  of 
 these  are  dedicated  news  channels  (CNN,  Fox  News,  MSNBC,  CNBC),  we 
 can  also  calculate  intensity  of  cable  use  for  news  as  the  number  of  dedicated 
 news  channels  regularly  viewed. 
 4.  Radio.  Respondents  report  which  of  the  radio  formats  they  listen  to.  The 
 formats,  which  are  listed  in  Table  4,  include  three  news-  related  categories 
 (“  all  news,”  “news/  talk/  information,”  and  “business”).  We  calculate  a 
 measure  of  radio  use  for  news  as  the  number  of  news  formats  regularly  used. 
 5.  Internet.  Respondents  report  whether  they  have  Internet  at  home  and  whether 
 at  work.  In  addition,  they  report  how  they  use  the  Internet  at  home  and  at 
 work  (whether  they  use  it  for  each  of  the  use  categories  in  Table  6).  A 
 number  of  these  uses  are  news-  related  (“  local  information,”  “news,” 
 “newspaper  site,”  “other  newspaper  site,”  and  “read  a  newspaper”).  This 
 allows  us  to  calculate  the  number  of  news-  related  Internet  uses  for  each 
 individual  along  with  the  total  number  of  uses.  Finally,  respondents  indicate 


 7  The  schedule  collection  process  is  fairly  time-  consuming.  Since  most  news  is  aired  between  4  and  12 
 PM,  our  restriction  allows  us  to  save  much  effort  without  losing  much  information. 
28
 28 
 how  many  hours  they  spend  on  the  Internet  per  week  (using  a  categoric 
 varable  that  takes  values  of  none,  0-  1  hour,  1-  5  hours,  5-  10,  10-  20,  and  20+, 
 with  which  we  assign  cell  midpoints  and  20  for  “20+”)  . 
 In  sum,  we  have  separate  measure  of  overall  use  and  news  use  for  television,  radio, 
 Internet,  and  cable. 


 II.  Media  Use 
 1.  Univariate  Characterizations 
 a.  Newspapers 
 Table  1  reports  that  53  percent  of  the  sample  reports  reading  a  daily  newspaper. 
 Of  those  reading  a  paper,  91  percent  report  reading  the  news  section.  Newspapers  are 
 clearly  and  important  medium  for  the  communication  of  news,  as  opposed  to  the  delivery 
 of  entertainment.  Other  popular  sections  include  entertainment  (69  percent),  business 
 (62),  comics  (61),  and  sports  (61).  Respondents  not  buying  a  paper  also  report  the 
 sections  they  regularly  read,  and  their  tendencies  to  read  each  section  are  lower. 


 b.  Television 
 The  average  number  of  weekly  half  hours  of  television  viewed  on  the  networks  is 
 35.5.  Of  these,  5.3  half  hours  are  broadcast  news  (viewed  between  4  and  12PM).  73 
 percent  of  households  have  some  sort  of  cable  television  (23  percent  have  basic,  29 
 percent  have  extended  basic,  and  22  percent  have  premium  service).  14  percent  of 
 households  have  satellite  TV.  80  percent  have  either  cable  or  satellite.  Individuals 
29
 29 
 (including  those  without  cable)  watch  an  average  of  8.4  cable  channels  regularly.  Of 
 these,  0.8  are  news  channels. 


 c.  Radio 
 In  the  sample  89  percent  of  persons  report  listening  to  at  least  one  radio  format 
 regularly.  Persons  in  the  sample  reports  listening  to  an  average  of  2.3  radio  formats.  See 
 Table  3.  The  54  formats  available  are  listed  in  Table  4.  The  most  popular  among  these 
 formats  are  news/  talk/  information  (23  percent  of  persons  listen),  country  (21),  adult 
 contemporary  (18),  oldies  (15)  contemporary  hit  radio  (14),  album  oriented  rock  (12), 
 and  classic  rock  (11).  Two  formats  that  attract  listeners  in  the  sample  are  oriented 
 toward  news:  news  (5.4  percent)  and  news/  talk/  information.  Roughly  a  quarter  of 
 persons  get  some  news  from  radio.  Average  persons  listen  to  0.28  radio  news  formats. 


 d.  Internet 
 As  Table  5  indicates,  just  over  half  of  people  in  the  sample  (55  percent)  have 
 Internet  access  at  home.  Roughly  a  third  (32  percent)  have  access  at  work.  Almost  all  of 
 the  persons  with  access  at  work  also  have  access  at  home,  so  56  percent  have  access 
 either  at  home  or  at  work.  Because  they  are  not  connected,  45  percent  of  people  spend  no 
 time  on  the  Internet  each  week;  10  percent  spend  an  hour  or  less;  22  percent  spend  1  to  5 
 hours;  12  percent  spend  5  to  10  hours,  7  percent  spend  10  to  20  hours,  and  5  percent 
 spend  over  20  hours.  If  we  apply  approximate  values  to  each  of  the  intervals  8  ,  we  can 
 develop  a  measure  of  average  time  spend  online  each  week.  Calculated  this  way,  the 
30
 30 
 average  time  online  is  3.5  hours  per  week.  Among  online  persons,  the  average  is  6.4 
 hours. 
 Table  6  shows  how  people  report  using  the  Internet  at  work  and  at  home.  Some 
 of  the  most  popular  uses  include  email  (79  percent  of  home  users),  shopping  (39),  and 
 news  (31).  9  These  data  are  important  because  they  raise  the  possibility  that  the  Internet 
 provides  a  substitute  for  other  media.  If  we  sum  together  the  number  of  uses  for  each 
 individual  (treating  home  and  work  uses  separately),  the  average  person  (including 
 nonusers)  reports  3.97  uses,  0.64  of  them  news-  related. 
 Table  8  summarizes  the  main  measures  of  media  use,  overall  and  for  news,  that 
 we  employ  below  to  study  substitution  with  individual  data.  It  is  worth  noting  that  some 
 of  these  measures  are  more  conventional  than  others  as  indicators  of  media  use.  For 
 broadcast  television  and  the  Internet  we  have  time  spent  with  the  medium  per  week.  For 
 newspapers,  by  contrast,  we  have  only  a  binary  measure  of  whether  or  not  a  person  reads 
 a  daily  paper.  For  radio  and  cable  (as  well  as  for  the  Internet)  we  have  the  number  of 
 distinct  purposes  for  which  the  individual  uses  the  medium.  While  at  first  blush  an  odd 
 measure  of  usage,  it  is  actually  a  slightly  enhanced  version  of  a  binary  “whether  I  use” 
 measure,  where  users  of  the  medium  who  use  more  than  one  type  of  service  are  coded 
 with  values  above  one.  (Individuals  using  N  types  of  Internet  sites  are  coded  with  N; 
 nonusers  are  coded  0).  In  the  Internet  context,  where  we  have  both  a  measure  of  time 
 spent  and  our  unconventional  number  of  uses  measure,  we  can  validate  the  latter.  Their 
 correlation  is  0.65,  lending  support  to  the  use  of  this  sort  of  measure.  One  should  bear  in 


 8  We  assume  persons  in  the  0-  1  category  spend  1  hour  each  week,  for  the  “1-  5”  group,  we  assume  2.5;  for 
 the  “5-  10”  group,  we  assume  7.  5;  for  the  “10-  20”  group,  we  assume  15;  and  for  the  “20+”  group  we  assume  20. 
 9  Another  8  percent  go  to  newspaper  sites,  and  15  percent  “read  a  newspaper”  when  online  at  home. 
31
 31 
 mind  that  the  virtue  of  this  data  source  is  its  inclusion  of  some  measure  of  individuals’ 
 use  of  a  variety  of  media. 


 2.  Determinants  of  Media  Use 
 The  use  of  each  medium  varies  across  persons  with  their  characteristics,  including 
 age,  education,  gender,  and  race/  Hispanic  status.  Table  7  characterizes  the  use  of  the 
 media  with  regressions  of  various  measures  of  media  use  on  age  dummies,  education 
 dummies,  a  gender  dummy,  race/  Hispanic  status  dummies.  The  excluded  age  category  is 
 18-  20;  the  excluded  education  category  is  8  th  grade  or  less,  and  the  excluded 
 race/  Hispanic  category  is  whites.  The  ten  dependent  variables  are:  broadcast  TV  half 
 hours,  TV  news  half  hours,  number  of  radio  formats  used,  the  number  of  radio  news 
 formats  used,  weekly  Internet  hours,  number  of  Internet  uses,  number  of  Internet  news 
 uses,  whether  a  person  reads  a  daily  newspaper,  the  number  of  cable  channels  used,  and 
 the  number  of  cable  news  channels  used. 
 The  use  of  TV,  overall  and  for  news,  rises  with  age  and  declines  with  education, 
 although  the  decline  with  education  is  slight  for  news.  Use  of  the  radio  overall  declines 
 with  age  and  rises  in  education,  while  the  use  of  radio  for  news  in  particular  rises  in  both 
 age  and  education.  All  measures  of  Internet  use  fall  in  age  and  rise  in  education.  Daily 
 newspaper  use,  like  use  of  radio  for  news,  rises  in  both  age  and  education.  Use  of  cable 
 channels  overall  declines  in  age  and  education,  while  use  of  cable  news  channels  rises  in 
 both  age  and  education.  With  the  exception  of  TV  (and  cable  overall  for  blacks),  blacks 
 and  Hispanics  use  these  media  less  than  whites.  With  the  exception  of  TV,  men  use  these 
 media  more  than  women. 
32
 32 
 III.  Substitution 
 1.  Direct  Evidence 
 The  Scarborough  data  are  unusual  in  that  they  allow  examination  of  how  users  of 
 one  medium  also  use  other  media.  That  is,  they  allow  researchers  to  make  statements  of 
 the  form,  “people  who  watch  a  lot  of  television  use  the  Internet  more  (or  less).”  These 
 are  potentially  interesting  descriptive  facts,  and  they  are  suggestive  about  consumer 
 substitution  across  media.  An  important  caveat  is  in  order,  however.  It  is  possible  that 
 people  differ  in  their  interest  in  information,  and  this  difference  may  manifest  itself  in  a 
 positive  correlation  among  all  sorts  of  media.  That  is,  “media  junkies”  may  read 
 newspapers  more,  watch  more  television,  and  use  the  Internet  more.  One  cannot  draw 
 firm  inferences  about  substitutability  from  these  data  directly  without  additional 
 assumptions.  Still,  because  data  of  this  sort  have  typically  been  absent  in  policy 
 discusions,  they  merit  a  look  before  we  turn  to  other  attempts  to  infer  substitutability. 
 The  first  approach  one  might  use  for  studying  substitution  is  examining 
 correlation  of  use  measures  across  media.  The  correlation  of  TV  viewing  and  time  spent 
 on  the  Internet,  for  example,  reveals  whether  persons  who  watch  more  television  also 
 spend  more  time  online.  Table  9  reports  the  correlations  among  our  basic  measures  of 
 media  use  across  persons  in  the  Scarborough  data. 
 Broadcast  TV  use  is  negatively  correlated  with  our  measures  of  radio,  Internet, 
 and  cable  use.  All  other  correlations  in  Table  9  are  positive.  The  preponderance  of 
 positive  correlations  indicates  that  persons  using  more  of  one  medium  also  use  other 
33
 33 
 media  more.  In  spite  of  this,  the  significant  negative  correlations  with  TV  use  suggest 
 some  avenues  of  substitution. 
 Table  10  reports  the  correlations  of  measures  of  the  use  of  each  medium  for  news. 
 All  of  these  correlations,  except  the  Internet-  TV  relationship,  are  positive.  Again, 
 excepting  Internet-  TV,  this  indicates  that  persons  who  use  more  news  through  one 
 medium  also  consume  more  news  though  other  media. 
 Because  we  have  separate  measures  of  news  and  other  (“  entertainment”)  uses  of 
 multiple  media,  we  can  pose  the  substitution  question  a  different  way.  We  can  calculate 
 how  much  people  use  each  medium  for  news,  relative  to  their  use  of  the  medium  for 
 other  purposes.  Term  this  “relative  news  use”  for  each  medium.  Then  we  can  ask  how 
 relative  news  use  is  correlated  across  medium.  Do  people  who  use  television  relatively 
 more  for  news  use  other  media  relatively  less  for  news?  Table  11  reports  the  correlations 
 that  answer  this  question.  The  estimated  correlations  are  negative  and  significant  for 
 TV  &  Internet,  TV  &  radio,  TV  &  cable,  TV  &  newspapers,  and  cable  &  newspapers. 
 These  negative  correlations  are  consistent  with  substitution  among  media. 
 A  shortcoming  of  this  empirical  approach,  however,  is  that  it  examines  only  two 
 media  at  a  time.  In  addition,  the  approach  does  not  account  for  any  of  the  characteristics 
 of  the  individuals  in  the  sample.  To  address  these  concerns,  we  run  three  sets  of 
 regressions  that  correspond  to  the  three  correlation  tables  above. 
 First,  Table  12  reports  regressions  of  individuals’  overall  use  of  each  medium  on 
 their  overall  use  of  the  other  media,  as  well  as  a  battery  of  individual  characteristics  (age, 
 gender,  race,  income).  For  example,  the  first  column  reports  a  regression  with  overall 
 broadcast  TV  viewing  as  the  dependent  variable  and  radio,  Internet,  newspaper,  and  cable 
34
 34 
 use  as  independent  variables.  The  only  negative  coefficients  in  the  entire  table  are  two 
 coefficients  reflecting  the  relationship  between  television  and  Internet  use. 
 Table  13  performs  an  exercise  analogous  to  the  exercise  in  Table  12,  except  that 
 the  measures  are  the  use  of  each  medium  for  news  rather  than  overall.  We  use  the  same 
 measure  of  newspaper  use  as  above.  In  this  table,  as  well,  the  only  negative  coefficients 
 are  the  TV-  Internet  coefficients. 
 Table  14  revisits  these  regressions  using  the  relative  use  of  each  medium  for 
 news.  In  this  table  there  are  a  number  of  negative  and  significant  coefficients,  indicating 
 that  persons  making  relatively  greater  use  of,  say,  the  Internet  for  news  make  relatively 
 less  use  of  television  for  news.  The  following  relationships  are  negative  and  significant: 
 Internet  &  TV,  newspapers  &  TV,  newspapers  &  radio,  newspapers  &  Internet,  cable  & 
 newspapers.  In  all  cases,  both  coefficients  corresponding  to  the  relationship  in  question 
 are  negative  and  significant.  Putting  these  results  into  words,  newspapers  seem  to  serve 
 as  substitutes  for  TV  news  on  broadcast  and  cable,  radio  news,  and  news  on  the  Internet. 
 In  addition,  the  news  on  the  Internet  appears  to  serve  as  a  substitute  for  news  on 
 broadcast  television. 


 2.  Indirect  Evidence  of  Substitution 
 Existing  research  documents  a  strong  positive  relationship  between  the  size  of 
 local  population(  s)  and  the  availability  of  local  media  (see  Waldfogel,  1999;  George  and 
 Waldfogel,  2000;  Waldfogel,  2001).  Larger  markets  have  more  radio  stations,  more 
 television  stations,  and  more  (and  more  extensive)  daily  newspapers.  Existing  studies 
35
 35 
 document  a  greater  overall  tendency  for  persons  to  consume  each  of  these  local  media  in 
 larger  media  markets. 
 Even  stronger  than  the  overall  relationship  is  the  relationship  between  particular 
 groups’  tendencies  to  consume  local  media  and  their  absolute  or  relative  size  in  the  local 
 population.  When  particular  groups  of  the  local  population  share  preferences  for  media 
 content  with  each  other,  but  not  with  the  remainder  of  the  population,  those  groups  tend 
 to  listen  to  more  radio,  watch  more  television,  and  be  more  likely  to  purchase  newspapers 
 in  markets  where  the  group  in  question  makes  up  a  larger  portion  of  local  population. 
 These  results  suggest  that  the  supply  of  local  media  appealing  to  particular  groups 
 increase  in  the  size  of  local  group  population.  We  can  use  this  variation  to  see  whether 
 the  group’s  media  consumers  are  more  likely  to  consume  nonlocal  media  when,  from 
 small  local  group  population,  we  can  infer  paucity  of  local  group-  targeted  media  supply. 
 Along  these  lines  we  propose  to  examine  how  use  of  local  and  nonlocal  media 
 among  preference  minorities  vary  with  the  size  of  their  local  populations.  If  nonlocal 
 media  (the  Internet  and  cable  television)  are  a  substitute  for  local  media,  then  we  should 
 see:  (a)  the  group  tendency  to  consume  local  media  should  increase  in  the  size  of  the 
 local  group  population  (as  has  been  documented  elsewhere);  and  (b)  the  group  tendency 
 to  consume  nonlocal  media  should  decline  in  the  size  of  the  local  group  population.  This 
 approach  is  also  employed  by  Sinai  and  Waldfogel  (2001),  which  documents  this 
 relationship  for  blacks’  Internet  use  with  the  Current  Population  Survey  data. 
 The  next  question  is  what  groups  to  examine.  Testing  these  hypotheses  requires 
 two  conditions.  First,  preferences  must  differ  across  groups.  Existing  research  shows 
 that  preferences  in  media  products  differ  sharply  by  gender,  age,  race  (black  vs.  non- 
36
 36 
 black)  and  Hispanic  status.  10  Second,  the  relative  numbers  of  persons  in  the  different 
 groups  must  differ  across  local  markets.  The  percent  black  and  Hispanic  vary 
 substantially  across  markets.  The  percent  in  each  age  category  and  gender  vary 
 substantially  less.  Consequently,  even  if  these  effects  operate  by  age,  they  will  be 
 difficult  to  detect  statistically  by  age  or  gender.  Our  strategy  here  is  to  examine  these 
 relationships  along  a  number  of  dimensions  (age,  race,  Hispanic  status)  to  see  whether  we 
 can  detect  the  operation  of  the  effects. 
 Tables  15-  17  examine  these  relationships  by  age,  race,  and  Hispanic  status.  In 
 each  table  we  regress  our  measures  of  media  use  on  CMSA  fixed  effects,  a  group 
 dummy,  and  the  group  dummy  interacted  with  the  CMSA  percent  in  the  group,  with  the 
 last  variable  generating  the  coefficient  of  interest.  Here  we  identify  the  effect  from  the 
 relationship  between  the  group  percent  in  the  CMSA  and  the  gap  between  the  group  and 
 non-  group  tendencies  to  use  the  medium  examined  in  the  regression.  Essentially,  these 
 regressions  ask  whether  the  gap  between,  say,  young  and  old  persons’  use  of  each 
 medium  is  larger  or  smaller  in  markets  with  proportionally  more  young  people.  For 
 media  with  local  content,  we  expect  young  persons’  use  of  media  content  (relative  to 
 older  persons)  to  increase  in  the  share  of  local  population  that  is  young.  To  the  extent 
 that  young  persons  have  different  preferences  from  older  persons,  we  expect  their  use  of 
 nonlocal  media  to  be  higher  in  places  where  young  persons  make  up  a  small  share  of 
 local  population.  This  would  provide  evidence,  at  least  indirectly,  that  nonlocal  media 
 provide  a  substitute  for  local  media. 


 10  See,  for  example,  Waldfogel  (1999)  for  evidence  that  blacks  and  whites,  and  Hispanics  and  non-Hispanics, 
 listen  to  different  radio  stations.  The  formats  attracting  two  thirds  of  black  listening  collectively  attract  less  than  5  percent  of  white  listening.  Waldfogel  (2001)  documents  similar  differences  between 


 black  and  white  preferences  in  television  programming. 
37
 37 
 Three  of  the  media  have  an  important  local  component  (radio,  television,  and 
 newspapers).  Since  the  supply  of  content  in  these  media  targeted  each  group  grows  in  the 
 size  of  the  local  group  population,  we  expect  the  group  use  tendency  to  increase  in  the 
 group’s  share  of  local  population.  For  radio  and  television,  this  is  what  we  find.  We  find 
 no  relationship  for  newspapers.  For  the  media  that  are  predominantly  nonlocal  (Internet 
 and  cable-  or-  satellite  connection)  we  expect  use  to  increase  in  the  group’s  local  isolation. 
 This  is  what  we  find  for  both  Internet  and  cable  television  use. 
 What  do  these  estimates  mean?  A  natural  interpretation  is  that  the  local 
 population  distribution  determines  the  supply  of  local  media  products;  hence  the  positive 
 relationships  between  local  group  share  and  the  group’s  tendencies  to  consume  local 
 media.  The  negative  relationships  between  local  group  share  and  the  tendency  to  use 
 nonlocal  media  is  naturally  interpreted  as  substitution  of  nonlocal  for  local  media. 


 IV.  Concluding  Discussion 
 We  have  now  examined  consumer  substitution  among  media  using  a  variety  of 
 data  sets  and  a  variety  of  different  approaches.  To  help  sort  out  the  results,  we 
 summarize  them  using  a  table.  Essentially,  we  have  examined  the  relationship  among 
 the  use  and  numbers  of  outlets  in  multiple  media.  Measures  of  the  following  variables 
 have  appeared  at  various  points  in  the  studies: 
 1.  TV  use  overall 
 2.  TV  outlets  overall 
 3.  TV  news  use 
 4.  Radio  use  overall 
38
 38 
 5.  Radio  outlets  overall 
 6.  Radio  use  for  news 
 7.  Radio  outlets  for  news 
 8.  Daily  newspaper  use 
 9.  Daily  outlets 
 10.  Weekly  newspaper  use 
 11.  Weekly  newspaper  outlets 
 12.  Internet  use  overall 
 13.  Internet  use  for  news 
 14.  Cable  use  overall 
 15.  Cable  use  for  news. 
 Substitution  is  reflected  by  negative  relationships  among  these  15  variables  across  media. 
 Table  18  is  a  15  x  15  matrix,  partitioned  by  medium,  summarizing  the  relationships  we 
 have  documented  from  approaches  geared  to  yield  information  about  substitution. 
 An  “A”  in  the  table  indicates  a  negative  correlation  among  changes  in  aggregate 
 measures  of  outlets  or  use  across  media  (Table  12  in  the  aggregate  study);  “B”,  “C,”  and 
 “D”  indicates  a  negative  coefficients  from  the  aggregate  changes  on  changes  regressions 
 with  t-  statistics  above  1  in  absolute  value;  they  are  rendered  bold  if  significant  at  the  5 
 percent  level  (see  Tables  13-  15  in  the  aggregate  study).  An  “E”  indicates  a  significant 
 negative  correlation  from  Table  9  of  this  study;  “F”  and  “G”  indicate  a  significant 
 negative  correlation  from  Tables10  and  11;  finally,  “H,”  “I,”  and  “J”  indicate  significant 
 negative  coefficients  in  the  regressions  in  Table  12-  14.  Note  that  the  correlations  are 
 treated  as  symmetric,  so  that  a  significant  negative  correlation  results  in  two  table  entries 
39
 39 
 symmetric  about  the  main  diagonal.  For  regression  coefficients  we  treat  the  column  as 
 the  independent  variable  and  the  row  as  the  dependent  variable. 
 Standing  back,  there  is  clearest  evidence  of  substitution  between  Internet  and 
 broadcast  TV,  both  overall  and  for  news;  between  daily  and  weekly  newspapers;  and 
 between  daily  newspapers  and  TV  news.  There  is  also  evidence  of  substitution  between 
 cable  and  daily  newspapers,  both  overall  and  for  news  consumption;  between  radio  and 
 broadcast  TV  for  news  consumption;  and  between  the  Internet  and  daily  newspapers  for 
 news  consumption.  There  is  little  or  no  evidence  of  substitution  between  weekly 
 newspapers  and  TV,  or  between  radio  and  either  Internet  or  cable.  There  is  also  some 
 indirect  evidence  of  substitution  in  the  greater  use  of  national  media  by  groups  less 
 targeted  by  local  media. 


 1.  What  does  the  Evidence  Mean? 
 What  do  the  results  in  this  study  say  about  the  questions  posed  at  the  outset  of  the 
 study?  Do  consumers  treat  different  media  as  substitutes  for  one  another?  If  there  were 
 no  relationships  among  use  or  outlets  across  media,  then  one  could  conclude  that  there  is 
 no  evidence  for  substitution.  Instead  we  find  negative  relationships  between  a  number  of 
 media.  This  at  least  suggests  some  degree  of  substitution. 
 Entertainment  is  an  end  product  in  itself,  and  while  some  persons  (“  news 
 junkies”)  find  news  entertaining,  one  might  also  view  the  demand  for  news  as  a  derived 
 demand  for  the  information  that  allows  citizens  to  carry  out  their  lives.  Weather 
 information  helps  parents  dress  their  children  –  and  themselves  –  appropriately  in  the 
40
 40 
 morning.  Political  information  makes  it  easier  for  citizens  to  know  what  is  at  stake  in 
 electoral  contests  and,  in  turn,  can  make  them  more  likely  to  vote. 
 If  substitution  is  complete,  then  the  decline  of  local  daily  newspapers  will  be 
 offset  by  increased  use  of  other  media.  We  can  define  another  sense  in  which 
 substitution  can  be  complete  or  incomplete.  With  complete  substitution,  the  civic 
 behaviors  affected  by  media  consumption  will  also  be  unaffected  by  changes  in 
 availability  or  use  of  any  particular  medium.  One  might  term  this  the  “behavioral 
 neutrality”  implication  of  complete  substitution. 
 Existing  research  on  media  consumption  and  voting  suggests  that,  even  if 
 substitution  operates,  it  is  not  complete  in  the  latter  sense.  Changes  in  availability  of 
 black-  targeted  radio  stations  and  weekly  newspapers,  1994-  1998,  affect  the  tendency  for 
 blacks  to  vote  (see  Oberholzer-  Gee  and  Waldfogel,  2001).  There  would  be  no  effect  on 
 voting  if  changes  in  use  of  other  media  offset  the  effects  of  changes  in  availability  of 
 black-  targeted  radio  stations  and  weeklies.  Although  there  is  evidence  (above,  and  Sinai 
 and  Waldfogel,  2001)  that  small  groups  substitute  nonlocal  media  (such  as  cable  and 
 Internet)  for  local  media  where  local  products  are  unavailable,  this  substitution  does  not 
 generate  behavioral  neutrality. 
 In  another  context  there  is  additional  evidence  against  behavioral  neutrality. 
 National  newspapers  serve  as  substitutes  for  local  ones,  in  the  sense  that  markets  with 
 faster  growth  in  a  major  national  paper’s  circulation  experience  greater  decline  in  local 
 daily  paper  circulation  among  readers  targeted  by  the  national  paper  (see  George  and 
 Waldfogel,  2002).  Behavioral  neutrality  fails  in  the  sense  that  in  markets  with  greater 
 increase  in  the  national  paper’s  circulation,  there  is  a  greater  decline  in  the  tendency  for 
41
 41 
 the  targeted  audience  to  vote  in  local  (non-  presidential)  elections.  Here  again,  although 
 consumers  view  the  products  are  substitutes,  they  do  not  have  the  same  effect  on  civic 
 behavior. 
 The  conception  of  each  medium  as  entirely  distinct  would  be  unduly  restrictive 
 because  there  is  evidence  (here  and  elsewhere)  that  consumers  substitute  across  media. 
 At  the  same  time,  however,  substitution  is  not  apparently  so  complete  that  the  effects  of 
 changes  in  one  medium  are  offset  by  changes  in  another  to  leave  civic  behavior 
 unchanged.  It  is  conventional  and  trite  to  conclude  a  study  with  a  call  for  more  research. 
 Nonetheless,  some  of  these  questions  will  only  be  answered  with  additional  research. 
42
 42 
 References 
 BIA,  BIA  Media  Access  Pro  –  TV,  2001.  http://  www.  bia.  com/  mediapro.  asp 
 Berry  Steven  T.  and  Joel  Waldfogel.  "Free  Entry  and  Social  Inefficiency  in  Radio  Broadcasting."  RAND  Journal  of  Economics  30,  3  (1999a):  397-  420. 


 Berry  Steven  T.  and  Joel  Waldfogel.  "Public  Radio  in  the  United  States:  Does  it  Correct  Market  Failure  or  Cannibalize  Commercial  Stations?"  Journal  of  Public  Economics  71 
 (1999b):  189-  211. 
 Berry  Steven  T.  and  Joel  Waldfogel.  “Mergers,  Station  Entry,  and  Programming  Variety  in  Radio  Broadcasting.”  Quarterly  Journal  of  Economics,  2001. 


 Burrelle’s  Media  Services.  Burrelle’s  Media  Directory,  1993  and  1999.  http://  www.  burrelles.  com/  indexmd.  html 
 Downes,  Tom;  Greenstein,  Shane.  “Universal  Access  and  Local  Internet  Markets  in  the  US.”  Research  Policy.  Vol.  31  (7).  p  1035-  52.  September  2002. 
 Duncan,  James.  Duncan’s  American  Radio,  Spring  1993  and  Spring  1997  Issues.  http://  www.  duncanradio.  com/  AmericanRadio.  html 
 Lisa  George  and  Joel  Waldfogel.  “Who  Benefits  Whom  in  Daily  Newspaper  Markets?  NBER  Working  Paper  7944,  October  2000.  Accessible  at 
 http://  rider.  wharton.  upenn.  edu/~  waldfogj/  workpap.  html 
 Lisa  George  and  Joel  Waldfogel.  “Does  the  New  York  Times  Spread  Ignorance  and  Apathy?”  The  Wharton  School,  revised  June  2002.  Accessible  at 
 http://  rider.  wharton.  upenn.  edu/~  waldfogj/  workpap.  html 
 Felix  Oberholzer  Gee  and  Joel  Waldfogel.  “Electoral  Acceleration:  The  Effect  of  Minority  Population  on  Minority  Voter  Turnout”,  NBER  Working  Paper  8252,  April 


 2001.  Accessible  at  http://  rider.  wharton.  upenn.  edu/~  waldfogj/  workpap.  html 
 Scarborough  Research,  Primenext  Multimarket  Data,  1999-  2000.  http://  www.  scarborough.  com/  prime/  index.  html 


 Peter  Siegelman  and  Joel  Waldfogel.  “Race  and  Radio:  Preference  Externalities,  Minority  Ownership,  and  the  Provision  of  Programming  to  Minorities.”  Advances  in 
 Applied  Microeconomics,  volume  10,  2001. 
 Todd  Sinai  and  Joel  Waldfogel.  “Geography  and  the  Internet:  Is  the  Internet  a  Substitute  or  a  Complement  for  Cities?”  Mimeo,  The  Wharton  School 
 Accessible  at  http://  rider.  wharton.  upenn.  edu/~  waldfogj/  workpap.  html 
43
 43 
 U.  S.  Department  of  Commerce.  Current  Population  Surveys,  (Computer  and  Internet  Use  Supplements  1997,  1998,  2000).  http://  www.  bls.  census.  gov/  cps/  cpsmain.  htm 
 Joel  Waldfogel.  “Who  Benefits  Whom  in  Local  Television  Markets?”,  mimeo,  The  Wharton  School,  December  2001.  Accessible  at 
 http://  rider.  wharton.  upenn.  edu/~  waldfogj/  workpap.  html 
 Joel  Waldfogel.  “Preference  Externalities:  An  Empirical  Study  of  Who  Benefits  Whom  in  Differentiated  Product  Markets”  NBER  Working  Paper  7391,  October  1999 
 Accessible  at  http://  rider.  wharton.  upenn.  edu/~  waldfogj/  workpap.  html 
44
 44 
 TABLES 
 (AND 
 FIGURE) 


 FOR 
 PART 


 I 
45
 45 
 Figure 
 1: 
 Media 


 Outlets 


 and 
 Market 


 Size 


 Pop 
 Total 


 1995 


 Broadcast 
 TV 
 Stations 


 2000 


 Daily 
 Papers 


 '99 


 Weekly 
 Papers 
 '99 


 Stations 
 Heard 
 in 
 Metro 


 - 
 1997 


 177 


 18844 


 1 
 824 


 Notes: 


 Log 
 scales 


 for 
 population 


 on 
 the 
 horizontal 


 axis 
 and 
 the 
 number 


 of 
 outlets 


 in 
 each 


 medium 


 on 
 the 
 vertical 


 axis. 
 See 
 text 
 for 


 explanation. 
46
 46 
 Table 
 1: 
 Television 


 Outlets, 
 1994-  2000 


 year 
 N 
 uhf 


 vhf 


 uhf+  vhf 


 network 
 indep. 
 public 


 net+  ind+ 
 pub 


 1994 
 207 


 2.29 


 2.63 
 2.46 


 1995 
 207 


 2.96 


 2.29 


 5.25 


 2.72 
 2.52 
 1.71 


 6.96 


 1996 
 207 


 3.09 


 2.34 


 5.43 


 4.20 
 1.23 
 1.71 


 7.14 


 1997 
 207 


 3.19 


 2.35 


 5.54 


 4.29 
 1.25 
 1.71 


 7.25 


 1998 
 207 


 3.37 


 2.37 


 5.74 


 4.74 
 1.40 
 1.72 


 7.87 


 1999 
 207 


 3.54 


 2.37 


 5.91 


 4.94 
 1.46 


 2000 
 207 


 3.65 


 2.39 


 6.04 


 5.02 
 1.57 
 1.73 


 8.32 


 linked 
 sample 


 year 
 N 
 uhf 


 vhf 


 uhf+  vhf 


 network 
 indep. 
 public 


 net+  ind+ 
 pub 


 1994 
 2.64 


 2.88 
 3.33 


 1995 
 143 


 3.76 


 2.63 


 6.38 


 3.01 
 3.37 
 2.04 


 8.42 


 1996 
 143 


 3.88 


 2.68 


 6.56 


 4.85 
 1.71 
 2.03 


 8.59 


 1997 
 143 


 3.98 


 2.69 


 6.67 


 4.95 
 1.72 
 2.03 


 8.71 


 1998 
 143 


 4.19 


 2.71 


 6.90 


 5.26 
 1.90 
 2.05 


 9.21 


 1999 
 143 


 4.40 


 2.71 


 7.11 


 5.43 
 1.99 


 2000 
 143 


 4.55 


 2.73 


 7.28 


 5.52 
 2.13 
 2.07 


 9.72 
47
 47 
 Table  2:  Cable  and  Television  Use,  1994-  2000  All  Markets  Linked  Sample 
 Year  N  hut  %  cable  %  N  hut  %  cable  % 
 1994  207  36.  47  63.  19  143  37.  29  62.  96  1995  207  36.  41  63.  88  143  37.  18  63.  93 


 1996  207  36.  69  66.  30  143  37.  47  66.  41  1997  207  36.  04  67.  05  143  36.  85  67.  21 
 1998  207  35.  88  67.  28  143  36.  82  67.  53  1999  207  35.  66  66.  90  143  36.  86  67.  43 
 2000  207  35.  47  67.  08  143  36.  80  67.  76 
48
 48 
 Table  3:  Quarterly  Use  of  Television,  Overall  and  News,  1999-  2001  hut 
 date  total  day  prime  time  Eve.  News  Nightly  News 
 Nov-  99  207  36.11  56.  41  50.44  41.  79  Feb-  00  207  36.67  57.  57  52.24  43.  55 


 May-  00  207  33.74  54.  44  46.36  43.  33  Jul-  00  207  33.  42  48.  60  44.  97  42.  07 
 Nov-  00  207  37.55  57.  06  51.81  43.  31  Feb-  01  207  37.22  57.  58  52.89  43.  52 
 May-  01  207  33.66  54.  07  46.03  43.  39  Jul-  01  207  33.  67  48.  47  44.  79  42.  38 
 linked  sample  date  total  day  prime  time  Eve.  News  Nightly  News 
 Nov-  99  143  37.46  57.  81  51.36  43.  31  Feb-  00  143  37.85  58.  81  52.92  44.  80 
 May-  00  143  35.07  55.  98  47.22  44.  90  Jul-  00  143  34.  83  50.  21  45.  79  43.  43 
 Nov-  00  143  38.95  58.  47  52.76  44.  78  Feb-  01  143  38.53  58.  98  53.69  44.  87 
 May-  01  143  35.15  55.  80  47.09  44.  94  Jul-  01  143  35.  06  50.  03  45.  77  43.  72 


 total  share  total  day  prime  time  Eve.  News  Nightly  News 
 Nov-  99  207  56.72  60.  36  65.49  68.  57  Feb-  00  207  55.87  61.  14  65.50  69.  01 
 May-  00  207  54.76  60.  21  63.46  66.  68  Jul-  00  207  48.92  62.  60  61.60 
 Nov-  00  207  52.40  57.  12  61.35  64.  89  Feb-  01  207  52.88  58.  27  62.78  66.  00 
 May-  01  207  52.30  57.  15  60.76  64.  00  Jul-  01  207  46.48  46.  06  59.65 


 Linked  sample  date  total  day  prime  time  Eve.  News  Nightly  News 
 Nov-  99  143  60.10  64.  54  68.30  70.  81  Feb-  00  143  59.14  65.  36  68.19  70.  97 
 May-  00  143  57.96  64.  25  65.97  68.  84  Jul-  00  143  51.71  64.  34  63.11 
 Nov-  00  143  55.78  61.  30  63.90  66.  82  Feb-  01  143  56.22  62.  51  65.35  68.  20 
 May-  01  143  55.45  61.  21  63.15  66.  17  Jul-  01  143  49.06  49.  57  61.41 
49
 49 
 Table  4:  Radio  Formats,  1993  &  1997 
 1993  1997 
 N  %  N  % 
 Adult  Contemp  (AC)  462  8.44  AC  418  6.62  AC/  Album  Oriented  Rock  (AOR)  8  0.15  AC/  AOR  4  0.06 


 AC/  Contemp.  Hit  Radio  (CHR)  10  0.18  AC/  CHR  128  2.03  AC/  Full  Service  (FS)  1  0.02  AC/  NR  42  0.67 
 AC/  Oldies  (O)  5  0.09  AC/  SAC  8  0.13  AOR  565  10.  32  AOR  375  5.94 
 AOR/  AC  5  0.09  AOR/  AC  3  0.05  Black  (B)  281  5.13  AOR/  NR  200  3.17 
 B/  AC  13  0.24  AOR/  P  80  1.27  B/  Gospel  (G)  4  0.07  B  236  3.74 
 B/  O  3  0.05  B/  AC  107  1.69  B/  Talk  (T)  4  0.07  B/  G  32  0.51 
 Big  Band/  Nostalgia  (BB)  213  3.89  B/  O  26  0.41  BB/  EZ  7  0.13  B/  T  6  0.1 
 BB/  FS  5  0.09  BB  315  4.99  BB/  O  1  0.02  BB/  EZ  3  0.05 
 Country  (C)  772  14.  1  BB/  REL  1  0.02  C/  FS  20  0.37  BB/  T  1  0.02 
 CHR  488  8.91  C  904  14.  32  CHR/  AC  15  0.27  C/  FS  14  0.22 
 CHR/  Spanish  (SP)  3  0.05  CHR  410  6.49  CHR/  Urban  (U)  9  0.16  CHR/  AC  51  0.81 
 Classical  (CL)  67  1.22  CHR/  NR  9  0.14  Classic  AOR(  CL  AOR)  151  2.76  CHR/  U  56  0.89 
 Ethnic  (E)  16  0.29  CL  70  1.11  Easy  Listening  (EZ)  2  0.04  CL  AOR  284  4.5 
 EZ/  New  Age  Contemp.  (NAC)  1  0.02  CL  HITS  83  1.31  EZ/  SAC  3  0.05  E  14  0.22 
 FS  202  3.69  EZ  13  0.21  FS/  T  41  0.75  EZ/  SAC  1  0.02 
 G  55  1  FS  96  1.52  G/  B  1  0.02  FS/  T  81  1.28 
 Jazz  (J)  59  1.08  G  62  0.98  News  (N)  37  0.68  J  118  1.87 
 News/  Talk  (N/  T)  161  2.94  KIDS  1  0.02  NAC/  AOR  2  0.04  N  75  1.19 
 O  352  6.43  N/  T  170  2.69  Religious  (REL)  200  3.65  O  394  6.24 
 Soft  Adult  Contemp.  (SAC)  379  6.92  REL  276  4.37  SP  172  3.14  SAC  253  4.01 
 Sports  (SPRTS)  70  1.28  SP  296  4.69  Talk  247  4.51  SPRTS  172  2.72 
 T/  O  4  0.07  T  408  6.46  Unknown  360  6.57  T/  CL  AOR  7  0.11 
 T/  J  5  0.08  UNK  6  0.1 
50
 50 
 Total  5476  100  6314  100 
51
 51 
 Table  5:  Radio  Outlets  and  Listening 
 Mean  Stations  News  Stations  Year  N  AQH  List.  Inside  All  Inside  All  Formats 


 1993  230  16.  83  15.  00  21.  73  0.46  0.79  11.  60  1994  230  16.  48 
 1995  230  16.  28  1996  230  16.  07 
 1997  230  15.  83  16.  65  24.  83  0.55  0.94  14.  97  1998  230  15.  65 


 Linked  Sample  Mean  Stations  News  Stations 
 N  AQH  List.  Inside  All  Inside  All  Formats 
 1993  142  16.  77  22.  76  27.  83  0.76  1.01  1994  137  16.  57 


 1995  139  16.  33  1996  139  16.  13 
 1997  143  15.  78  24.  96  31.  43  0.88  1.20  1998  139  15.  74 


 Note:  N=  143  for  all  but  AQH 
52
 52 
 Table  6:  Newspaper  Outlets  and  Circulation,  1993-  1999  Daily  Daily  Weekly  Weekly 
 Year  N  #  Outlets  Circulation  #  Outlets  Circulation 
 1993  210  8.05  309580  44.  18  531694  1999  210  7.85  295950  46.  21  728804 


 Linked  Sample  Daily  Daily  Weekly  Weekly 
 Year  N  #  Outlets  Circulation  #  Outlets  Circulation 
 1993  143  10.  15  427775  56.  98  737695  1999  143  9.92  409149  60.  20  1025342 
53
 53 
 Table  7:  Trends  in  Internet  Use,  1997-  2000  Full  Sample 
 connection  Use  for  Use  for  Use  for  Year  home  work  mail  news  info 
 1997  15.  14%  11.  95%  7.08%  12.  08%  1998  30.  25%  19.  82%  12.  04%  15.  73% 


 2000  45.  90%  33.  62%  19.  59%  23.  81% 
 Linked  Sample  connection  Use  for  Use  for  Use  for 
 Year  home  work  mail  news  info 
 1997  143  15.64%  12.05%  12.52%  7.  47%  12.  36%  1998  143  30.94%  14.13%  20.80%  13.19%  16.89% 


 2000  143  46.56%  17.23%  34.65%  20.92%  25.39% 
54
 54 
 Table 
 8: 
 Medium 


 Use 
 and 
 Outlets, 


 Cross 
 Sectional 


 Evidence 


 (1) 
 (2) 
 (3) 
 (4) 
 (5) 
 (6) 
 (7) 
 (8) 


 TV 
 HUT 


 % 
 '97 


 Radio  AQH  List. 
 '97 


 Weekly  Newspaper  Circ/  Pop.  '99 


 Daily  Newspaper  Circ/  Pop. 
 '99 


 TV 
 HUT 


 % 
 '97 


 Radio  AQH  List. 
 '97 


 Weekly  Newspaper  Circ/  Pop.  '99 


 Daily  Newspaper  Circ/  Pop. 
 '99 


 TV 
 Stations  #  uhf  &  vhf 


 stations 
 '97 


 0.7923 


 1.0346 


 (0.  1052)** 
 (0.  1278)** 


 Radio 
 Stations  #  Rec'd  '97 


 0.0199 


 0.0374 


 (0.  0049)** 
 (0.  0083)** 


 Newspapers  #  Weeklies  '99 
 0.0827 


 0.1236 


 (0.  0357)* 
 (0.  0553)* 


 Newspaper  #  Dailies  '99 
 0.4121 
 0.2089 


 (0.  0551)** 
 (0.  0750)** 


 Constant 
 31. 
 5121 


 15. 
 2567 


 31. 
 9608 


 16. 
 1550 


 29. 
 8973 


 14. 
 7054 


 29. 
 5078 


 18. 
 1791 


 (0.  7706)** 
 (0.  1751)** 
 (3.  7091)** 
 (0.  7280)** 
 (0.  9121)** 
 (0.  2742)** 
 (4.  5009)** 
 (0.  8995)** 


 Observations 
 141 
 141 
 141 
 141 
 141 
 141 
 141 
 141 


 Method 
 OLS 
 OLS 
 OLS 
 OLS 


 IV 
 IV 
 IV 
 IV 


 R- 
 squared 


 0.29 
 0.11 
 0.04 
 0.29 
 0.26 
 0.02 
 0.03 
 0.22 


 Notes: 
 Standard 


 errors 
 in 
 parentheses. 


 * 
 significant 


 at 
 5%; 


 ** 
 significant 


 at 
 1%. 


 IV 
 regressions 


 use 
 log 
 population 


 as 
 the 
 instrument 


 for 
 the 
 number 


 of 
 outlets. 
55
 55 
 Table 
 9: 
 Medium 


 Use 
 and 
 Market 


 Size 


 (1) 
 (2) 
 (3) 
 (4) 
 (5) 
 (6) 


 HUT 
 % 
 '97 


 AQH 
 List. 
 '97 


 % 
 Int 


 at 
 home 


 '98 


 Weekly  Circ/  Pop. 
 '99 
 Daily  Circ/  Pop. 


 '99 
 % 
 Cable 


 '97 


 1995 
 Pop. 
 (000) 


 0.  9505 


 0.  1701 
 0.  3851 
 2.  6237 
 0.  8302 
 -0.  1345 


 (0.  1418)** 
 (0.  0358)** 
 (0.  3252) 
 (1.  3116)* 
 (0.  2336)** 
 (0.  2774) 


 Constant 
 35. 
 1828 


 15. 
 5939 


 30. 
 2242 


 32. 
 4655 


 18. 
 8534 


 67. 
 3699 


 (0.  4075)** 
 (0.  1027)** 
 (0.  9343)** 
 (3.  7686)** 
 (0.  6713)** 
 (0.  7971)** 


 Observations 
 141 
 141 
 141 
 141 
 141 
 141 


 R- 
 squared 


 0.24 
 0.14 
 0.01 
 0.03 
 0.08 
 0.00 


 Notes: 
 Standard 


 errors 
 in 
 parentheses. 


 * 
 significant 


 at 
 5%; 


 ** 
 significant 


 at 
 1% 
56
 56 
 Table 
 10: 
 Outlets 


 and 
 Media 


 Use: 
 Fixed 


 Effects 


 Regressions 


 (1) 
 (2) 
 (3) 


 (4) 
 (5) 


 HUT 
 AQH 
 List. 


 AQH 
 News 
 List. 


 Daily 
 Circ. 
 pc 
 Weekly 


 Circ. 
 pc 


 TV 
 Stations 


 0.3133  (0.  1365)* 


 Radio 
 Stations 


 -0. 
 0156  (0.  0157) 


 Radio 
 News 
 Stations 


 0.2412  (0.  0369)** 


 # 
 Weeklies 


 0.0518  (0.  0364) 


 # 
 Dailes 


 1.2074  (0.  2201)** 


 Year=  1995 
 0.6581  (0.  2640)* 
 Year=  1996 
 0.8933  (0.  2537)** 
 Year=  1997 
 0.2337 
 -0. 
 8340 


 -0. 
 2167 


 (0.  2481) 
 (0.  1010)** 
 (0.  0373)** 


 Year=  1998 
 0.1435  (0.  2397) 
 Year=  1999 
 0.1172 


 -1. 
 2545 


 0.0904 


 (0.  2348) 
 (0.  2608)** 
 (0.  7614) 


 Constant 
 34. 
 4717 


 17. 
 2034 


 0.3859 


 9.4843 
 33. 
 7189 


 (1.  0068)** 
 (0.  4437)** 
 (0.  0451)** 


 (2.  2506)** 
 (2.  1361)** 


 Observations 
 846 
 281 
 281 


 282 
 282 


 Number 
 of 
 DMA 


 -- 
 Nielsen 


 141 


 141 
 141 


 141 
 141 


 R- 
 squared 


 0.03 
 0.45 
 0.32 


 0.32 
 0.02 


 Notes: 
 Standard 


 errors 
 in 
 parentheses. 


 * 
 significant 


 at 
 5%; 


 ** 
 significant 


 at 
 1% 
57
 57 
 TABLE 
 11: 
 Correlations 


 among 
 Cross 
 Sectional 


 Measures 


 of 
 Outlets 


 and 
 Use 


 of 
 Various 


 Media 


 | 
 TV 
 ‘99 


 Radio  (In)  ’97 


 Radio  News  (In)  ‘97 
 Dailies  ‘99 
 Weeklies  ‘99 


 TV 
 ‘98 


 Radio  ‘97 


 Radio  News  ‘97 


 Radio  News  In  ‘97 
 Dailies  ‘99 
 Weeklies  ‘99 


 Internet  ‘98 
 Internet  News  ‘98 


 Outlets:  TV  ‘99 
 1 
 Radio 
 (In) 
 ‘97 


 0.7555* 


 1 


 Radio 
 News 
 (In) 


 ‘97 


 0.4699* 
 0.5618* 


 1 


 Dailies 
 ‘99 


 0.  6505* 
 0.  6661* 


 0.  2919* 


 1 


 Weeklies 
 ‘99 
 0.5342* 


 0.6449* 


 0.3391* 
 0.8096* 


 1 


 Use:  TV  ‘98 
 0.5366* 
 0.3936* 


 0.2521* 
 0.3909* 
 0.  2921* 


 1 


 Radio 
 ‘97 


 0.3569* 
 0.3395* 


 0.3175* 
 0.2356* 
 0.  3144* 
 0.3338* 


 1 


 Radio 
 News 
 ‘97 
 0.2511* 


 0.2489* 


 0.7148* 
 0.2152* 
 0.  2553* 
 0.2091* 
 0.  3348* 


 1 


 Radio 
 News 
 In 


 ‘97 


 0.2459* 
 0.2091* 


 0.7001* 
 0.1949* 
 0.  2173* 
 0.2043* 
 0.  3167* 


 0.  9870* 


 1 


 Dailies 
 ‘99 


 0.  2080* 


 0.  5356* 
 0.  3669* 


 1 


 Weeklies 
 ‘99 


 0.1573 
 0.1959* 


 0.1925* 
 0.1511 


 0.1428 
 1 


 Internet 
 ‘98 


 0.1415 


 1 


 Internet 
 News 
 ‘98 


 0.1619 
 0.1659* 


 0.7804* 
 1 


 Cable 
 ‘98 


 0.2266* 


 0.1543 
 0.2354* 
 0.  1946* 


 Note: 
 Correlation 


 reported 
 if 
 significant 


 at 
 10% 


 level; 


 star 
 indicates 


 5% 
 significance 


 level. 
58
 58 
 TABLE 
 12: 
 Correlations 


 among 
 Changes 


 in 
 Measures 


 of 
 Outlets 


 and 
 Use 


 of 
 Various 


 Media 


 
  outlets 
 
  outlets 


 
  outlets 


 
  outlets 
 
  outlets 
 
  use 
 
  use 


 
  use 


 
  use 
 
  use 
 
  use 
 
  use 
 
  use 


 TV 
 Radio 
 Radio 
 News 


 Daily  Papers 


 Weekly  Papers 


 TV 
 Radio 


 Radio  News 


 Radio  News  (Inside) 
 Daily  Papers 
 Weekly  Papers 
 Internet 
 Internet  News 


 Change 
 in 
 Outlets: 


 TV 


 1 
 Radio 
 0.1407 
 1 


 Radio 
 News 


 0.2774* 
 1 


 Daily 
 Papers 


 1 


 Weekly 
 Papers 


 -0. 
 4411* 


 1 


 Change 
 in 
 Use: 


 TV 


 1 
 Radio 


 1 
 Radio 
 News 


 0.5306* 


 1 


 Radio 
 News 
 (Inside) 


 0.5470* 


 0.9610* 
 1 


 Daily 
 Papers 


 0.4218* 


 1 


 Weekly 
 Papers 


 1 


 Internet 
 -0.  1782* 


 1 


 Internet 
 News 


 -0. 
 1638 


 0.7991* 
 1 


 Cable 
 -0.  2319* 
 -0.  1727* 
 0.  1559 
 0.  2025* 


 -0.  1654* 


 Note: 
 Correlation 


 reported 
 if 
 significant 


 at 
 10% 


 level; 


 star 
 indicates 


 5% 
 significance 


 level. 
59
 59 
 Table 
 13: 
 The 
 Relationship 


 among 
 Changes 


 in 
 Media 


 Outlets 


 (1) 


 (2) 


 (3) 
 (4) 
 (5) 


 Chg. 
 TV 
 Stns. 


 '94- 


 '00 


 Chg. 
 Daily 


 Papers 


 '93- 


 '99 


 Chg. 
 Weekly 


 Papers 
 '93- 
 '99 


 Chg. 
 Inside 


 Radio 
 Stns 


 '93- 
 '97 


 Chg. 
 Inside 


 News 


 Radio 
 Stns 
 '93-  '97 


 Chg. 
 Inside 


 Radio 
 Stns 
 '93- 
 '97 


 0.0346 


 0.0211 


 -0. 
 1113 


 (1. 
 52) 


 (0. 
 95) 


 (0. 
 28) 


 Chg. 
 Daily 


 Papers 


 '93- 
 '99 


 0.  1354 


 -7.  9469 
 0.  3099 
 -0.  0367 


 (1. 
 55) 


 (5. 
 76)** 


 (0. 
 95) 


 (0. 
 56) 


 Chg. 
 Weekly 


 Papers 
 '93- 
 '99 


 0.0038 


 -0. 
 0245 


 -0. 
 0050 


 0.0018 


 (0. 
 78) 


 (5. 
 76)** 


 (0. 
 28) 


 (0. 
 48) 


 Chg. 
 TV 
 Stns. 


 '94- 
 '00 


 0.1276 
 1.1592 
 0.4787 
 0.0408 


 (1. 
 55) 


 (0. 
 78) 


 (1. 
 52) 


 (0. 
 65) 


 Constant 
 0.8435 
 -0. 
 3134 


 0.4616 


 1.8602 
 0.0843 


 (7. 
 93)** 


 (2. 
 57)* 


 (0. 
 21) 


 (4. 
 13)** 


 (0. 
 94) 


 H 
 0 
 : 
 Cross 


 effects= 


 0 
 (p- 


 val) 


 (0. 
 16) 


 (0. 
 00) 


 (0. 
 00) 


 (0. 
 23) 


 (0. 
 73) 


 Observations 
 141 
 141 


 141 
 141 
 141 


 R- 
 squared 


 0.04 
 0.22 


 0.20 
 0.03 
 0.01 


 Notes: 
 Absolute 


 value 
 of 
 t 
 statistics 


 in 
 parentheses. 


 * 
 significant 


 at 
 5%; 


 ** 
 significant 


 at 
 1% 
60
 60 
 Table 
 14: 
 Changes 


 in 
 Use 


 and 
 Change 


 in 
 Outlets  (1)  (2) 


 (3) 
 (4) 
 (5) 
 (6) 
 (7) 
 (8) 


 Chg. 
 HUT 


 Viewing 


 '95- 
 '00 


 Chg. 
 Daily 
 Paper 


 Circ. 


 per 
 cap 


 '93- 
 '99 


 Chg. 
 Weekly  Paper  Circ.  per  cap  '93-  '99 


 Chg. 
 AQH 
 List 
 '94- 
 '98 


 Chg. 
 AQH  News  List  '93-  '97 


 Chg. 
 Cable  Use  '94-  '00 


 Int 
 Use 


 '00 


 Int 
 News  Use  '00 


 Chg. 
 TV 
 Stns. 


 '94- 
 '00 


 0.6104 
 -0. 
 0157 


 0.1023 


 0.0779 
 -0. 
 0259 


 -0. 
 3002 


 -0. 
 7365 


 -0. 
 6838 


 (1. 
 84) 


 (0. 
 07) 


 (0. 
 14) 


 (1. 
 33) 


 (0. 
 78) 


 (1. 
 06) 


 (0. 
 91) 


 (1. 
 49) 


 Chg. 
 Inside 


 Radio 
 Stns 
 '93- 
 '97 


 -0. 
 0898 


 -0. 
 1078 


 -0. 
 1958 


 0.0091 


 -0. 
 1842 


 -0. 
 1856 


 (1. 
 01) 


 (1. 
 67) 


 (1. 
 02) 


 (0. 
 56) 


 (2. 
 43)* 


 (0. 
 85) 


 Chg. 
 Daily 


 Papers 


 '93- 
 '99 


 -0. 
 2157 


 1.4125 


 -0. 
 1983 


 -0. 
 0929 


 0.0028 


 -0. 
 2911 


 1.3533 


 0.8029 


 (0. 
 63) 


 (5. 
 72)** 


 (0. 
 27) 


 (1. 
 56) 


 (0. 
 08) 


 (1. 
 00) 


 (1. 
 62) 


 (1. 
 70) 


 Chg. 
 Weekly 


 Papers 
 '93- 
 '99 


 0.0075 
 0.0203 
 0.0444 
 -0. 
 0033 


 0.0024 


 0.0180 
 0.0504 
 0.0198 


 (0. 
 39) 


 (1. 
 48) 


 (1. 
 09) 


 (1. 
 01) 


 (1. 
 24) 


 (1. 
 11) 


 (1. 
 08) 


 (0. 
 75) 


 Chg. 
 Inside 


 News 
 Radio 
 Stns 
 '93- 


 '97 


 0.3252 
 -1. 
 1253 


 (7. 
 25)** 


 (1. 
 82) 


 Constant 
 -0. 
 9343 


 -1. 
 0182 


 0.4078 


 -0. 
 9183 


 -0. 
 1944 


 5.3499 


 47. 
 7075 


 21. 
 7674 


 (1. 
 88) 


 (2. 
 83)** 


 (0. 
 38) 


 (10.05)** 


 (4. 
 10)** 


 (12.62)** 


 (39.01)** 
 (33.21)** 


 H 
 0 
 : 
 Cross 


 effects= 


 0 
 (p- 


 val) 


 (0. 
 53) 


 (0. 
 17) 


 (0. 
 76) 


 (0. 
 28) 


 (0. 
 50) 


 (0. 
 01) 


 (0. 
 41) 


 (0. 
 09) 


 Observations 
 141 
 141 
 141 
 135 
 140 
 141 
 141 
 141 


 R- 
 squared 


 0.03 
 0.21 
 0.02 
 0.03 
 0.29 
 0.09 
 0.03 
 0.06 


 Notes: 
 Absolute 


 value 
 of 
 t 
 statistics 


 in 
 parentheses. 


 * 
 significant 


 at 
 5%; 


 ** 
 significant 


 at 
 1% 
61
 61 
 Table 
 15: 
 Changes 


 in 
 Use 


 and 
 Changes 


 in 
 Use 


 (1) 


 (2) 
 (3) 
 (4) 
 (5) 
 (6) 
 (7) 
 (8) 


 Chg. 
 HUT 


 Viewing 


 '95- 
 '00 


 Chg. 
 Daily  Paper  Circ. 


 per 
 cap 


 '93- 
 '99 


 Chg. 
 Weekly  Paper  Circ.  per  cap  '93-  '99 


 Chg. 
 AQH 
 List 
 '94- 
 '98 


 Chg. 
 AQH  News  List  '93-  '97 


 Chg. 
 Cable  Use  '94-  '00 


 Int 
 Use 


 '00 


 Int 
 News 


 Use 
 '00 


 Chg. 
 AQH 


 List 
 '94- 
 '98 


 0.3922 
 0.1176 
 -0. 
 7387 


 -0. 
 2400 


 -0. 
 8745 


 (0. 
 78) 


 (0. 
 29) 


 (0. 
 68) 


 (0. 
 61) 


 (0. 
 72) 


 Chg. 
 Daily 


 Paper 
 Circ. 
 per 
 cap 


 '93- 
 '99 


 0.0377 
 0.1188 
 0.0054 
 0.0065 
 -0. 
 1676 


 0.0271 


 -0. 
 0074 


 (0. 
 35) 


 (0. 
 51) 


 (0. 
 29) 


 (0. 
 51) 


 (2. 
 01)* 


 (0. 
 10) 


 (0. 
 05) 


 Chg. 
 Weekly 


 Paper 
 Circ. 
 per 


 cap 
 '93- 
 '99 


 0.0083 
 0.0166 


 -0. 
 0048 


 0.0010 


 -0. 
 0376 


 0.1299 


 0.0619 


 (0. 
 21) 


 (0. 
 51) 


 (0. 
 68) 


 (0. 
 21) 


 (1. 
 19) 


 (1. 
 33) 


 (1. 
 10) 


 Chg. 
 Cable 


 Use 
 '94- 
 '00 


 0.2905 
 -0. 
 1809 


 -0. 
 2903 


 -0. 
 0119 


 -0. 
 0017 


 0.2656 
 0.1563 


 (2. 
 66)** 


 (2. 
 01)* 


 (1. 
 19) 


 (0. 
 61) 


 (0. 
 13) 


 (0. 
 98) 


 (1. 
 02) 


 Int 
 Use 


 '00 


 0.0126 
 0.0030 
 0.1042 
 -0. 
 0045 


 0.0276 


 (0. 
 35) 


 (0. 
 10) 


 (1. 
 33) 


 (0. 
 72) 


 (0. 
 98) 


 Chg. 
 HUT 
 Viewing 


 '95- 
 '00 


 0.0251 
 0.0396 
 0.0121 
 -0. 
 0117 


 0.1794 


 0.0749 
 -0. 
 1268 


 (0. 
 35) 


 (0. 
 21) 


 (0. 
 78) 


 (1. 
 14) 


 (2. 
 66)** 


 (0. 
 35) 


 (1. 
 03) 


 Int 
 News 


 Use 
 '00 


 -0. 
 0073  (1.  01) 


 Chg. 
 AQH 


 Inside 


 News 
 List 


 '93- 
 '97 


 -0. 
 8342  (0.  76) 


 Constant 
 -2. 
 0030 


 -0. 
 7021 


 -3. 
 6352 


 -0. 
 5326 


 -0. 
 0025 


 3.1950 


 44. 
 3660 


 19. 
 8568 


 (1. 
 14) 


 (0. 
 49) 


 (0. 
 94) 


 (1. 
 74) 


 (0. 
 02) 


 (2. 
 35)* 


 (24.24)** 


 (21.73)** 


 H 
 0 
 : 
 Cross 


 effects= 


 0 
 (p- 


 val) 


 (0. 
 17) 


 (0. 
 45) 


 (0. 
 55) 


 (0. 
 85) 


 (0. 
 75) 


 (0. 
 02) 


 (0. 
 62) 


 (0. 
 63) 


 Observations 
 135 
 135 
 135 
 135 
 140 
 135 
 135 
 140 


 R- 
 squared 


 0.07 
 0.04 
 0.03 
 0.02 
 0.03 
 0.10 
 0.66 
 0.65 


 Notes: 
 Absolute 


 value 
 of 
 t 
 statistics 


 in 
 parentheses. 


 * 
 significant 


 at 
 5%; 


 ** 
 significant 


 at 
 1% 
62
 62 
 TABLES 
 FOR 
 PART 


 II 
63
 63 
 Table  1:  Newspaper  Use  Read  a  Daily? 
 No  Yes  Business  pages/  section  29.  00%  62.  23% 
 Classified  Advertising  36.  72%  55.  76%  Comics  33.  12%  60.  88% 
 Entertainment  41.  52%  68.  73%  Food  29.  05%  58.  77% 
 News  55.  19%  90.  77%  Sports  31.  61%  60.  77% 
 TV  or  Radio  Listings  27.  42%  55.  22% 
 N  83224  95560 


 Table  2:  Cable  and  Satellite  percent 
 Basic  22.  76%  Basic/  Extended  plus  Pay  21.  93% 
 None  26.  71%  Extended  28.  59% 


 N  178784 
 percent  No  Satellite  86.  13% 
 Have  Satellite  13.  87% 
 DirectTV  6.35%  Dish  3.02% 


 Other  Provider  3.66%  Primestar  0.  91% 
 USSB  0.47% 
 N  178784 
64
 64 
 Table  3:  Radio  Use  Number  of  Formats  Listened  to  percent 
 0  11.  39%  1  19.  83% 
 2  27.  68%  3  21.  15% 
 4  11.  44%  5  5.25% 
 6  2.10%  7  0.76% 
 8  0.30%  9  0.09% 
 10  0.02% 
 178784 
65
 65 
 Table  4:  Radio  Format  Use 
 Format 
 percent  listening 
 to  format 
 Adult  Contemporary  18.  29%  Other  0.25% 


 Adult  Standards  4.44%  Album  Adult  Alternative  2.  52% 
 Album  Oriented  Rock  11.  81% 
 All  News  5.35% 


 All  Sports  5.  19%  Alternative  7.  25% 
 Black  0.26% 
 Business  0.00% 


 Childrens  Radio  (Rel.  2  Only)  0.40%  Classic  Country  0.  79% 
 Classic  Rock  11.08%  Classical  7.62% 
 Contemporary  Christian  2.81%  Contemporary  Hit  Radio  14.  41% 
 Pop  Contemporary  Hit  Radio  6.03%  Country  21.  22% 
 Rhythmic  Contemporary  Hit  Radio  (Rel.  2  Only)  1.74%  Dance  0.00% 
 Easy  Listening  0.72%  Educational  (Rel.  2  Only)  0.86% 
 Ethnic  0.39%  Gospel  1.40% 
 Hot  AC  9.67%  Modern  AC  3.69% 
 Jazz  1.68%  New  AC/  Smooth  Jazz  4.86% 
 New  Age  0.00%  New  Country  1.43% 
 New  Rock  2.17% 
 News/  Talk/  Information  23.  02% 


 Nostalgia  1.40%  70's  Rock  0.01% 
 70's  Oldies/  Classic  Hits  2.72%  Oldies  15.  26% 
 Religious  5.97%  Rhythm  &  Blues  0.38% 
 Rhythmic  Oldies  3.27%  Soft  Adult  Contemporary  5.81% 
 Spanish  News/  Talk  (Rel.  2  Only)  0.70%  Spanish  Contemporary  2.44% 
66
 66 
 Spanish  Regional  2.17%  Spanish  Language  0.  50% 
 Spanish  Tropical  0.  91%  Southern  Gospel  0.50% 
 Talk/  Personality  3.  91%  Urban  AC  3.48% 
 Tejano  0.33%  Urban  Contemporary  6.18% 
 Urban  Oldies  0.67%  Variety  3.67% 
 Spanish  label  var  viety  0.28%  Format  Not  Provided  0.00% 


 Notes:  News-  related  formats  underlined  in  table. 
67
 67 
 Table  5:  Internet  Use  hours/  week  percent 
 0  45.  15%  0-  1  9.93% 
 1-  4  21.  76%  5-  9  11.  59% 
 10-  19  6.59%  20+  4.98% 


 None  43.  96%  Home  Only  23.  82% 
 Work  Only  1.18%  Both  31.  03% 


 N  178784 
68
 68 
 Table  6:  Internet  Uses  among  Home  and  Work  Users 
 use  %  using  at  home  for…  %  using  at  work  for… 
 Auction  site  11.  6%  4.5%  Automobile  information  19.  5%  8.8% 


 Cable  network  site  2.9%  1.7%  Chat  or  community  site  16.  8%  5.8% 
 Research/  education  79.  1%  84.  2%  E-  mail  24.  5%  20.  9% 
 Financial  information/  services  28.  6%  9.2%  Games  17.  0%  13.  2% 
 Job/  employment  search  9.8%  6.7%  Local  cable  system  site  1.2%  0.5% 


 Local  information  22.  1%  15.  2% 
 Local  television  station  site  3.7%  2.0%  Magazine  site  7.0%  5.6% 
 Medical  services/  information  6.3%  5.3% 
 News  18.  5%  10.  0% 


 Newspaper  site  31.  0%  31.  0% 
 Other  use  7.6%  7.1% 
 Other  newspaper  site  1.0%  1.1% 


 Maintain  own  home  page  17.  1%  13.  3%  Purchase  item  or  service  21.  4%  12.  0% 
 Listen  to  audio  programming/  radio  station  7.6%  5.3%  Radio  station  site  5.4%  4.2% 
 Real  estate  listings  14.  6%  13.  1%  Read  a  magazine  12.0%  5.7% 


 Read  a  newspaper  37.  7%  37.  3% 
 Shopping  39.  0%  18.  6%  Sports  scores/  updates  18.  4%  12.  4% 
 Television  network  site  5.7%  3.3%  Watch  video  programming  1.7%  1.0% 
 Work-  related  use  12.5%  49.2% 
 Users  98067  57592 
 Notes:  News-  related  uses  underlined. 
69
 69 
 Table 
 7: 
 Who 


 Uses 
 Each 
 Medium?  (1)  (2) 


 (3) 
 (4) 
 (5) 
 (6) 
 (7) 
 (8) 
 (9) 
 (10) 


 TV 
 Half 


 Hours 


 per 
 Week 


 TV 
 News  Half  Hours  (4-  12pm) 


 # 
 Radio  Formats  Used 


 # 
 Radio  News  Formats  Used 


 Weekly  Internet  Hours 


 # 
 Internet  Uses 


 # 
 Internet  News  Uses 


 Daily  Newspaper 
 # 
 Cable  Channels  Used 


 # 
 Cable  News  Channels  Used 


 age 
 21-  20 


 0.8925 
 0.0363 
 -0. 
 2160 


 -0. 
 0206 


 -0. 
 2466 


 -0. 
 2701 


 0.0258 


 -0. 
 0663 


 -0. 
 1293 


 -0. 
 0383 


 (0.  4452)* 
 (0.  0882) 
 (0.  0211)** 


 (0.  0069)** 
 (0.  0787)** 
 (0.  0725)** 


 (0.  0180) 
 (0.  0070)** 


 (0.  0994) 
 (0.  0159)* 


 age 
 25-  29 


 2.0167 
 0.2560 
 -0. 
 2765 


 0.0145 


 -0. 
 4463 


 -0. 
 4199 


 0.0157 


 -0. 
 0740 


 -0. 
 0644 


 -0. 
 0300 


 (0.  4102)** 
 (0.  0813)** 
 (0.  0194)** 


 (0.  0063)* 
 (0.  0725)** 


 (0.  0668)** 


 (0.  0166) 
 (0.  0065)** 


 (0.  0916) 
 (0.  0146)* 


 age 
 30-  34 


 2.1558 
 0.3457 
 -0. 
 2838 


 0.0688 


 -0. 
 6439 


 -0. 
 6339 


 -0. 
 0258 


 -0. 
 0289 


 0.0019 


 0.0034 


 (0.  4020)** 
 (0.  0797)** 
 (0.  0190)** 
 (0.  0062)** 
 (0.  0710)** 
 (0.  0654)** 


 (0.  0162) 
 (0.  0063)** 


 (0.  0898) 
 (0.  0143) 


 age 
 35-  39 


 3.7367 
 0.8860 
 -0. 
 3352 


 0.1219 


 -1. 
 1255 


 -1. 
 0243 


 -0. 
 0870 


 0.0367 


 -0. 
 2464 


 0.0702 


 (0.  3954)** 
 (0.  0784)** 
 (0.  0187)** 
 (0.  0061)** 
 (0.  0699)** 
 (0.  0644)** 
 (0.  0160)** 
 (0.  0062)** 
 (0.  0883)** 
 (0.  0141)** 


 age 
 40-  44 


 4.3695 
 1.1062 
 -0. 
 4043 


 0.1545 


 -1. 
 1787 


 -1. 
 2470 


 -0. 
 1260 


 0.0990 


 -0. 
 4371 


 0.0797 


 (0.  3979)** 
 (0.  0789)** 
 (0.  0188)** 
 (0.  0061)** 
 (0.  0703)** 
 (0.  0648)** 
 (0.  0161)** 
 (0.  0063)** 
 (0.  0889)** 
 (0.  0142)** 


 age 
 45-  49 


 5.7908 
 1.4608 
 -0. 
 5088 


 0.1886 


 -1. 
 1845 


 -1. 
 3372 


 -0. 
 1535 


 0.1343 


 -0. 
 3011 


 0.1193 


 (0.  4081)** 
 (0.  0809)** 
 (0.  0193)** 
 (0.  0063)** 
 (0.  0721)** 
 (0.  0664)** 
 (0.  0165)** 
 (0.  0064)** 
 (0.  0911)** 
 (0.  0146)** 


 age 
 50-  54 


 11. 
 6471 


 3.1224 


 -0. 
 7699 


 0.2179 


 -1. 
 7841 


 -2. 
 4368 


 -0. 
 3524 


 0.1679 


 -0. 
 6137 


 0.2417 


 (0.  4139)** 
 (0.  0820)** 
 (0.  0196)** 
 (0.  0064)** 
 (0.  0731)** 
 (0.  0674)** 
 (0.  0167)** 
 (0.  0065)** 
 (0.  0924)** 
 (0.  0148)** 


 age 
 55-  59 


 14. 
 4908 


 3.8996 


 -1. 
 0045 


 0.2675 


 -2. 
 0782 


 -2. 
 9276 


 -0. 
 4371 


 0.2074 


 -0. 
 8676 


 0.3164 


 (0.  4367)** 
 (0.  0866)** 
 (0.  0207)** 
 (0.  0067)** 
 (0.  0772)** 
 (0.  0711)** 
 (0.  0176)** 
 (0.  0069)** 
 (0.  0975)** 
 (0.  0156)** 


 age 
 60-  64 


 18. 
 0567 


 4.5624 


 -1. 
 1911 


 0.2889 


 -2. 
 5092 


 -3. 
 5055 


 -0. 
 5042 


 0.2502 


 -0. 
 9295 


 0.3856 


 (0.  4539)** 
 (0.  0900)** 
 (0.  0215)** 
 (0.  0070)** 
 (0.  0802)** 
 (0.  0739)** 
 (0.  0183)** 
 (0.  0072)** 
 (0.  1014)** 
 (0.  0162)** 


 age 
 65-  69 


 22. 
 3543 


 5.5710 


 -1. 
 4182 


 0.3061 


 -2. 
 9543 


 -4. 
 0423 


 -0. 
 5726 


 0.2947 


 -1. 
 0745 


 0.4448 


 (0.  4623)** 
 (0.  0916)** 
 (0.  0219)** 
 (0.  0071)** 
 (0.  0817)** 
 (0.  0752)** 
 (0.  0187)** 
 (0.  0073)** 
 (0.  1032)** 
 (0.  0165)** 


 age 
 70+ 


 26. 
 5446 


 6.5158 


 -1. 
 7677 


 0.2882 


 -3. 
 5353 


 -4. 
 5105 


 -0. 
 6388 


 0.3131 


 -1. 
 6902 


 0.4344 


 (0.  3943)** 
 (0.  0782)** 
 (0.  0187)** 
 (0.  0061)** 
 (0.  0697)** 
 (0.  0642)** 
 (0.  0159)** 
 (0.  0062)** 
 (0.  0881)** 
 (0.  0141)** 


 some 
 high 


 2.4028 


 0.5135 
 0.3248 
 0.0450 
 0.0443 
 -0. 
 1540 


 -0. 
 0280 


 0.1312 


 0.6736 
 0.0816 


 (0.  4605)** 
 (0.  0913)** 
 (0.  0218)** 
 (0.  0071)** 


 (0.  0814) 
 (0.  0750)* 


 (0.  0186) 
 (0.  0073)** 


 (0.  1028)** 
 (0.  0164)** 


 high 
 0.7760 
 0.4254 
 0.5415 
 0.1066 
 0.4578 
 0.1872 
 0.0214 
 0.2388 
 1.1956 
 0.2117 


 (0.  3919)* 
 (0.  0777)** 


 (0.  0186)** 
 (0.  0060)** 
 (0.  0692)** 
 (0.  0638)** 


 (0.  0158) 
 (0.  0062)** 


 (0.  0875)** 
 (0.  0140)** 


 some 
 coll 


 -1. 
 2629 


 0.1224 


 0.7507 
 0.1759 
 1.4572 
 1.4282 
 0.2349 
 0.2952 
 1.2157 
 0.3311 


 (0.  4052)** 
 (0.  0803) 
 (0.  0192)** 


 (0.  0062)** 
 (0.  0716)** 
 (0.  0659)** 
 (0.  0164)** 
 (0.  0064)** 
 (0.  0905)** 
 (0.  0145)** 


 college 
 -2. 
 7839 


 -0. 
 0888 


 0.8960 


 0.2711 
 1.9110 
 2.2537 
 0.4021 
 0.3429 
 0.8698 
 0.3998 


 (0.  4129)** 
 (0.  0818) 
 (0.  0195)** 


 (0.  0064)** 
 (0.  0730)** 
 (0.  0672)** 
 (0.  0167)** 
 (0.  0065)** 
 (0.  0922)** 
 (0.  0147)** 


 some 
 grad 


 -3. 
 5798 


 -0. 
 2026 


 1.0047 


 0.3400 
 2.3446 
 2.9443 
 0.5184 
 0.3710 
 0.6776 
 0.4200 


 (0.  5137)** 
 (0.  1018)* 
 (0.  0243)** 


 (0.  0079)** 
 (0.  0908)** 
 (0.  0836)** 
 (0.  0207)** 
 (0.  0081)** 
 (0.  1147)** 
 (0.  0183)** 


 grad 
 schl 


 -5. 
 1676 


 -0. 
 3416 


 0.9257 


 0.3576 
 2.5269 
 3.1556 
 0.5609 
 0.3936 
 0.5327 
 0.4503 


 (0.  4345)** 
 (0.  0861)** 
 (0.  0206)** 
 (0.  0067)** 
 (0.  0768)** 
 (0.  0707)** 
 (0.  0175)** 
 (0.  0069)** 
 (0.  0970)** 
 (0.  0155)** 


 Black 
 9.9758 
 0.7268 
 -0. 
 2975 


 -0. 
 1157 


 -0. 
 7968 


 -0. 
 8268 


 -0. 
 1161 


 -0. 
 0329 


 0.8888 


 -0. 
 0499 
70
 70 
 (0.  2393)** 
 (0.  0474)** 
 (0.  0113)** 
 (0.  0037)** 
 (0.  0423)** 
 (0.  0389)** 
 (0.  0097)** 
 (0.  0038)** 
 (0.  0534)** 
 (0.  0085)** 


 Asian 
 -1. 
 5797 


 -0. 
 2753 


 -0. 
 7174 


 -0. 
 0073 


 1.0704 


 1.0196 
 0.2998 
 -0. 
 0705 


 -0. 
 9865 


 0.0418 


 (0.  5068)** 
 (0.  1004)** 
 (0.  0240)** 


 (0.  0078) 
 (0.  0895)** 


 (0.  0825)** 
 (0.  0205)** 
 (0.  0080)** 
 (0.  1132)** 


 (0.  0181)* 


 Oth 
 Non-  Hisp 


 0.2576 
 -0. 
 0468 


 -0. 
 3040 


 -0. 
 0452 


 -0. 
 0221 


 0.0079 


 0.0094 
 -0. 
 0597 


 0.1378 


 -0. 
 0144 


 (0.  4750) 
 (0.  0941) 
 (0.  0225)** 


 (0.  0073)** 


 (0.  0839) 
 (0.  0773) 
 (0.  0192) 
 (0.  0075)** 


 (0.  1061) 
 (0.  0169) 


 Hisp 
 5.5291 
 1.1266 
 -0. 
 1834 


 -0. 
 0862 


 -1. 
 1152 


 -1. 
 1895 


 -0. 
 1600 


 -0. 
 0846 


 -1. 
 2080 


 -0. 
 1447 


 (0.  2541)** 
 (0.  0504)** 
 (0.  0120)** 
 (0.  0039)** 
 (0.  0449)** 
 (0.  0414)** 
 (0.  0103)** 
 (0.  0040)** 
 (0.  0568)** 
 (0.  0091)** 


 Male 
 -6. 
 6350 


 -0. 
 0570 


 0.1505 


 0.0972 
 1.0631 
 1.2149 
 0.2628 
 0.0656 
 1.6841 
 0.2348 


 (0.  1445)** 
 (0.  0286)* 
 (0.  0068)** 


 (0.  0022)** 
 (0.  0255)** 
 (0.  0235)** 
 (0.  0058)** 
 (0.  0023)** 
 (0.  0323)** 
 (0.  0052)** 


 Constant 
 28. 
 6164 


 2.7419 


 2.3282 
 -0. 
 0764 


 3.5133 


 4.2392 
 0.5687 
 0.1356 
 7.3119 
 0.2946 


 (0.  5149)** 
 (0.  1021)** 
 (0.  0244)** 
 (0.  0079)** 
 (0.  0910)** 
 (0.  0838)** 
 (0.  0208)** 
 (0.  0081)** 
 (0.  1150)** 
 (0.  0184)** 


 Observations 
 178784 
 178784 
 178784 
 178784 
 178784 
 178784 
 178784 
 178784 
 178784 
 178784 


 R- 
 squared 


 0.11 
 0.12 
 0.16 
 0.11 
 0.08 
 0.16 
 0.09 
 0.10 
 0.03 
 0.05 


 Notes: 
 Standard 


 errors 
 in 
 parentheses. 


 * 
 significant 


 at 
 5%; 


 ** 
 significant 


 at 
 1%. 
71
 71 
 Table 
 8: 
 Sample 


 Statistics 


 mean 


 TV 
 Half 


 Hours/  Week 


 35.47 


 News 
 Half 
 Hours 


 (4-  12 
 PM) 


 5.31 


 Whether 
 Listen 
 to 
 Radio 


 88. 
 61% 


 # 
 Radio 


 Formats 


 Listened 


 to 


 2.32 


 # 
 Radio 


 News 
 Formats 


 Used 


 0.28 


 Whether 
 Use 
 Internet 


 54. 
 85% 


 # 
 Internet 


 Uses 


 3.97 


 # 
 Internet 


 News 
 Uses 


 0.64 


 Weekly 
 Time 
 Online 


 3.  50 


 Whether 
 Cable 
 or 
 Satellite 


 79.  81% 


 # 
 Cable 


 Channels 


 Used 


 8.40 


 # 
 Cable 


 News 
 Channels 


 Used 


 0.82 


 Whether 
 Daily 
 Newspaper 


 53. 
 45% 
72
 72 
 Table 
 9: 
 Correlations 


 among 
 Overall 


 Use 
 Measures 


 shortname 
 Variable 


 view 
 radall 
 dradio 


 itime 
 iuses 


 dint 
 cab_  num 


 dcabsat 
 any_  daily 


 view 
 Television 
 Half 
 Hours 


 per 
 week 


 1 


 radall 
 Radio 
 Formats 


 Listened 


 to 


 -0. 
 1149 


 1 
 0 


 dradio 
 Whether 
 Listen 
 to 
 Radio 


 -0. 
 0937 


 0.  5387 


 1 


 0 
 0 


 itime 
 Weekly 
 Time 
 Online 


 -0.1044 
 0.  117 
 0.  0808 


 1 


 0 
 0 
 0 


 iuses 
 Number 
 of 
 Internet 


 Uses 


 -0. 
 144 


 0.1787 


 0.1148 
 0.6536 


 1 


 0 
 0 
 0 
 0 


 dint 
 Whether 
 online 


 -0.  1785 
 0.  2015 
 0.  1516 
 0.  5745 
 0.  6533 


 1 


 0 
 0 
 0 
 0 
 0 


 cab_  num 
 Number 
 of 
 Cable 


 Channels 


 Used 


 -0. 
 015 


 0.0712 


 0.0465 
 0.1196 
 0.1539 
 0.0933 


 1 


 0 
 0 
 0 
 0 
 0 
 0 


 dcabsat 
 Whether 
 Use 
 Cable 


 or 
 Satellite 


 -0. 
 0684 


 0.  0088 


 0.0121 
 0.0584 


 0.057 
 0.0806 


 0.4788 


 1 


 0 
 0.0002 


 0 
 0 
 0 
 0 


 0 


 any_  daily 
 Whether 
 Read 
 a 
 Daily 


 0.0536 
 0.0299 
 0.0208 
 0.0022 


 0.003 
 0.0066 


 0.0393 
 0.0712 


 1 


 0 
 0 
 0 
 0.3458 


 0.2071 
 0.0052 


 0 
 0 


 Note: 
 probability 


 values 
 appear 
 below 
 correlations 


 in 
 table. 
73
 73 
 Table 
 10: 
 Correlations 


 among 
 News 
 Use 
 Measures  TVnews  inews_~ 


 s 
 rnews 


 cab_  news 


 any_  da~ 
 y 


 TVnews 
 TV 
 News 


 Half 
 Hours 


 (4- 
 12PM) 


 1 


 inews_  uses 
 Number 
 of 
 Internet 


 News 
 Uses 


 -0. 
 0909 


 1 
 0 


 rnews 
 # 
 of 
 Radio 


 News 
 Formats 


 Used 


 0.0513 
 0.0542 


 1 


 0 
 0 


 cab_  news 
 # 
 of 


 Cable 


 News 
 Channels 


 Viewed 
 0.0758 
 0.145 
 0.0961 


 1 


 0 
 0 
 0 


 any_  daily 
 Whether 
 Daily 
 Newspaper 


 0.0909 
 0.0092 
 0.1556 
 0.1182 


 1 


 0 
 0.0001 


 0 
 0 


 Table 
 11: 
 Correlations 


 among 
 Gap 
 (News- 


 Ent) 
 Measures 


 TVgap 
 igap 
 radgap 


 cab_  gap 


 any_  da~ 
 y 


 TVgap 
 TV: 
 News- 


 Ent 


 1 


 Igap 
 Internet: 
 News- 
 Ent 


 -0. 
 108 


 1 
 0 


 Radgap 
 Radio: 
 News- 
 Ent. 


 -0. 
 0679 


 0.  1417 


 1 


 0 
 0 


 cab_  gap 
 Cable: 
 News- 
 Ent. 


 -0.  0083 
 0.  1264 
 0.  1119 


 1 


 0.0005 
 0 
 0 


 any_  daily 
 Whether 
 Daily 
 Newspaper 


 -0.  0218 
 0.  0021 
 0.  0675 


 -0.  0013 


 1 


 0 
 0.3706 


 0 
 0.5748 
74
 74 
 Table 
 12: 
 Do 
 People 


 Using 
 One 
 Medium 


 More 
 Use 
 Other 


 Media 


 Less? 


 (1) 
 (2) 
 (3) 
 (4) 
 (5) 


 TV 
 Half 


 Hours 


 per 
 Week 


 # 
 Radio 


 Formats 


 Used 
 Weekly 
 Internet 
 Hours 


 Daily 
 Newspaper 


 # 
 Cable 


 Channels 


 Used 


 # 
 Radio 


 Formats 


 Used 


 0.0620 


 0.0572 
 0.0271 
 0.1476 


 (0.  0501) 
 (0.  0088)** 
 (0.  0008)** 
 (0.  0111)** 


 Weekly 
 Internet 
 Hours 


 -0.  1286 


 0.  0041 


 0.  0007 
 0.  1208 


 (0.  0134)** 
 (0.  0006)** 


 (0.  0002)** 
 (0.  0030)** 


 Daily 
 Newspaper 


 1.0543 
 0.2448 
 0.0826 


 0.6196 


 (0.  1505)** 
 (0.  0071)** 
 (0.  0265)** 


 (0.  0334)** 


 # 
 Cable 


 Channels 


 Used 


 0.0937 


 0.0067 
 0.0758 
 0.0031 


 (0.  0107)** 
 (0.  0005)** 
 (0.  0019)** 
 (0.  0002)** 


 TV 
 Half 


 Hours 


 per 
 Week 


 0.0001 
 -0. 
 0040 


 0.0003 


 0.0046 


 (0.  0001) 
 (0.  0004)** 
 (0.  0000)** 
 (0.  0005)** 


 Constant 
 28. 
 0960 


 2.2279 


 2.9283 
 0.0399 
 6.3276 


 (0.  5341)** 
 (0.  0248)** 
 (0.  0944)** 
 (0.  0085)** 
 (0.  1185)** 


 Observations 
 178784 
 178784 
 178784 
 178784 
 178784 


 R- 
 squared 


 0.11 
 0.16 
 0.09 
 0.11 
 0.04 


 Notes: 
 Standard 


 errors 
 in 
 parentheses. 


 * 
 significant 


 at 
 5%; 


 ** 
 significant 


 at 
 1%. 


 Regressions 


 include 
 individual 


 controls 
 for 
 education, 


 age, 
 race, 


 and 
 gender. 
75
 75 
 Table 
 13: 
 Do 
 People 


 Using 
 News 
 from 
 One 
 Medium 


 More 
 Use 
 News 


 from 
 Other 


 Media 


 Less? 


 (1) 
 (2) 
 (3) 
 (4) 
 (5) 


 TV 
 News 


 Half 
 Hours 


 (4- 
 12pm) 


 # 
 Radio 


 News 


 Formats 
 Used 


 # 
 Internet 


 News 
 Uses 


 Daily 
 Newspaper 


 # 
 Cable 


 News 


 Channels 


 Used 


 # 
 Radio 


 News 
 Formats 


 Used 


 0.0726 


 0.0945 
 0.0611 
 0.0479 


 (0.  0304)* 
 (0.  0061)** 
 (0.  0024)** 
 (0.  0054)** 


 # 
 Internet 


 News 
 Uses 


 -0. 
 0851 


 0.0141 


 0.0048 
 0.1316 


 (0.  0117)** 
 (0.  0009)** 


 (0.  0009)** 
 (0.  0021)** 


 Daily 
 Newspaper 


 0.2909 
 0.0585 
 0.0310 


 0.1166 


 (0.  0298)** 
 (0.  0023)** 
 (0.  0060)** 


 (0.  0053)** 


 # 
 Cable 


 News 
 Channels 


 Used 


 0.2306 


 0.0092 
 0.1691 
 0.0233 


 (0.  0133)** 
 (0.  0010)** 
 (0.  0027)** 
 (0.  0011)** 


 TV 
 News 


 Half 
 Hours 


 (4- 
 12pm) 


 0.  0004 
 -0.  0035 


 0.  0018 
 0.  0073 


 (0.  0002)* 
 (0.  0005)** 


 (0.  0002)** 
 (0.  0004)** 


 Constant 
 2.6885 
 -0. 
 0962 


 0.5313 


 0.1256 
 0.1876 


 (0.  1023)** 
 (0.  0080)** 
 (0.  0206)** 
 (0.  0081)** 
 (0.  0182)** 


 Observations 
 178784 
 178784 
 178784 
 178784 
 178784 


 R- 
 squared 


 0.13 
 0.12 
 0.11 
 0.11 
 0.07 


 Notes: 
 Standard 


 errors 
 in 
 parentheses. 


 * 
 significant 


 at 
 5%; 


 ** 
 significant 


 at 
 1%. 


 Regressions 


 include 
 individual 


 controls 
 for 
 education, 


 age, 
 race, 


 and 
 gender. 
76
 76 
 Table 
 14: 
 The 
 Relationship 


 among 
 News 
 – 
 Entertainment 


 Gaps 
 across 


 Media 


 (1) 
 (2) 


 (3) 
 (4) 
 (5) 


 TV: 
 News- 


 Ent. 


 Radio: 
 News- 
 Ent. 


 Internet: 


 News- 
 Ent. 


 Daily 
 Newspaper 


 Cable: 
 News- 
 Ent. 


 Radio: 
 News- 
 Ent. 


 0.1459 


 0.0253 
 -0. 
 0140 


 0.1848 


 (0.  0408)** 
 (0.  0055)** 
 (0.  0008)** 
 (0.  0095)** 


 Internet: 
 News- 
 Ent. 


 -0. 
 1766 


 0.0046 


 -0. 
 0027 


 0.1586 


 (0.  0175)** 
 (0.  0010)** 


 (0.  0003)** 
 (0.  0041)** 


 Daily 
 Newspaper 


 -0. 
 4370 


 -0. 
 1230 


 -0. 
 1302 


 -0. 
 3693 


 (0.  1210)** 
 (0.  0070)** 


 (0.  0164)** 


 (0.  0282)** 


 Cable: 
 News- 
 Ent. 


 0.1089 


 0.0114 


 0.0535 
 -0. 
 0026 


 (0.  0101)** 
 (0.  0006)** 


 (0.  0014)** 
 (0.  0002)** 


 TV: 
 News- 


 Ent. 


 0.0005 
 -0. 
 0032 


 -0. 
 0002 


 0.0059 


 (0.  0001)** 
 (0.  0003)** 
 (0.  0000)** 
 (0.  0006)** 


 Constant 
 -22.5271 
 -2. 
 3618 


 -2. 
 7368 


 0.0710 


 -5. 
 5855 


 (0.  4341)** 
 (0.  0247)** 


 (0.  0588)** 
 (0.  0085)** 
 (0.  1011)** 


 Observations 
 178784 
 178784 


 178784 
 178784 
 178784 


 R- 
 squared 


 0.07 
 0.22 


 0.16 
 0.11 
 0.05 


 Notes: 
 Standard 


 errors 
 in 
 parentheses. 


 * 
 significant 


 at 
 5%; 


 ** 
 significant 


 at 
 1%. 


 Regressions 


 include 
 individual 


 controls 
 for 
 education, 


 age, 
 race, 


 and 
 gender. 
77
 77 
 Table 
 15: 
 Media 


 Use 
 and 
 Local 


 Isolation 


 (Youth) 


 (1) 
 (2) 
 (3) 
 (4) 
 (5) 
 (6) 
 (7) 
 (8) 
 (9) 
 (10) 


 TV 
 Half 


 Hours 


 per 
 Week 


 TV 
 News  Half  Hours  (4-  12pm) 


 # 
 Radio  Formats  Used 


 # 
 Radio  News  Formats  Used 


 Weekly  Internet  Hours 


 # 
 Internet  Uses 


 # 
 Internet  News  Uses 


 Daily  Newspaper 
 # 
 Cable  Channels  Used 


 # 
 Cable  News  Channels  Used 


 Young 
 Dummy 


 x 


 % 
 Young 


 17.7282 
 0.7965 
 1.4439 
 0.2378 
 -5.0205 
 -4.4882 
 -0.3347 
 -0.1909 
 -4.4077 
 0.3820 


 (7.7469)* 
 (1.6857) 
 (0.6156)* 
 (0.3073) 
 (2.0370)* 
 (2.0037)* 
 (0.4528) 
 (0.1195) 
 (2.1918)* 
 (0.3080) 


 Young 
 Dummy 


 -29.0308 


 -5.7107 
 0.6771 
 -0.4433 
 5.0475 
 5.8596 
 0.7056 
 -0.2223 
 3.0691 
 -0.6197 


 (3.5611)** 
 (0.8013)** 
 (0.2855)* 
 (0.1464)** 
 (0.9059)** 
 (0.9139)** 
 (0.2069)** 
 (0.0541)** 
 (0.9692)** 
 (0.1419)** 


 Constant 
 53.5560 
 8.7510 
 0.8253 
 0.2005 
 0.1374 
 -0.1971 
 -0.0515 
 0.4057 
 5.9245 
 0.6815 


 (0.7917)** 
 (0.1823)** 
 (0.0355)** 
 (0.0107)** 
 (0.1014) 
 (0.1033) 
 (0.0196)** 
 (0.0089)** 
 (0.1427)** 
 (0.0210)** 


 Observations 
 178784 
 178784 
 178784 
 178784 
 178784 
 178784 
 178784 
 178784 
 178784 
 178784 


 R- 
 squared 


 0.10 
 0.13 
 0.16 
 0.14 
 0.08 
 0.16 
 0.09 
 0.11 
 0.04 
 0.06 


 Notes: 
 Robust 
 standard 


 errors 
 in 
 parentheses 


 – 
 se’s 


 are 
 clustered 


 on 
 CMSA. 


 * 
 significant 


 at 
 5%; 


 ** 
 significant 


 at 
 1%. 


 All 
 specifications 


 include 
 age 
 dummies, 


 education 
 dummies, 


 and 
 a 
 gender 


 dummy. 
78
 78 
 Table 
 16: 
 Media 


 Use 
 and 
 Local 


 Isolation 


 (Blacks) 


 (1) 
 (2) 
 (3) 
 (4) 
 (5) 
 (6) 
 (7) 
 (8) 
 (9) 
 (10) 


 TV 
 Half 


 Hours 


 per 
 Week 


 TV 
 News  Half  Hours  (4-  12pm) 


 # 
 Radio  Formats  Used 


 # 
 Radio  News  Formats  Used 


 Weekly  Internet  Hours 


 # 
 Internet  Uses 


 # 
 Internet  News  Uses 


 Daily  Newspaper 
 # 
 Cable  Channels  Used 


 # 
 Cable  News  Channels  Used 


 Black 
 x 
 % 


 Black 


 in 
 Metro 


 Area 


 24.4042 


 0.9195 
 1.0901 
 0.0128 
 -0.4154 
 -0.9014 
 -0.2193 
 -0.0380 
 -3.0696 
 -0.3915 


 (5.0254)** 
 (0.9571) 
 (0.2897)** 


 (0.0752) 
 (0.7065) 
 (0.5800) 
 (0.1472) 
 (0.0465) 
 (1.1183)** 
 (0.1086)** 


 Black 
 4.8300 
 0.3711 
 -0.4620 
 -0.1101 
 -0.6671 
 -0.5818 
 -0.0835 
 -0.0072 
 1.3547 
 0.0057 


 (1.0996)** 
 (0.1903) 
 (0.0589)** 


 (0.0140)** 
 (0.1558)** 
 (0.1257)** 
 (0.0286)** 


 (0.0110) 
 (0.2028)** 


 (0.0238) 


 Constant 
 30.8601 
 3.3045 
 2.1940 
 -0.1175 


 3.0567 
 3.7750 
 0.5251 
 0.1002 
 6.8837 
 0.2454 


 (0.7954)** 
 (0.1807)** 
 (0.0436)** 
 (0.0197)** 
 (0.1554)** 
 (0.1498)** 
 (0.0243)** 
 (0.0084)** 
 (0.1744)** 
 (0.0219)** 


 Observations 
 178784 
 178784 
 178784 
 178784 
 178784 
 178784 
 178784 
 178784 
 178784 
 178784 


 R- 
 squared 


 0.11 
 0.13 
 0.16 
 0.15 
 0.08 
 0.16 
 0.09 
 0.11 
 0.04 
 0.06 


 Notes: 
 Robust 
 standard 


 errors 
 in 
 parentheses 


 – 
 se’s 


 are 
 clustered 


 on 
 CMSA. 


 * 
 significant 


 at 
 5%; 


 ** 
 significant 


 at 
 1%. 


 All 
 specifications 


 include 
 age 
 dummies, 


 education 
 dummies, 


 and 
 a 
 gender 


 dummy. 
79
 79 
 Table 
 17: 
 Media 


 Use 
 and 
 Local 


 Isolation 


 (Hispanics) 


 (1) 
 (2) 
 (3) 
 (4) 
 (5) 
 (6) 
 (7) 
 (8) 
 (9) 
 (10) 


 TV 
 Half 


 Hours 


 per 
 Week 


 TV 
 News  Half  Hours  (4-  12pm) 


 # 
 Radio  Formats  Used 


 # 
 Radio  News  Formats  Used 


 Weekly  Internet  Hours 


 # 
 Internet  Uses 


 # 
 Internet  News  Uses 


 Daily  Newspaper 
 # 
 Cable  Channels  Used 


 # 
 Cable  News  Channels  Used 


 Hisp. 
 Dummy 


 x 


 % 
 Hisp. 


 4.5915 
 0.7486 
 1.5035 
 -0.1005 
 -3.1912 
 -2. 
 6137 


 -0.3580 


 0.0805 
 -4.4903 
 -0.6898 


 (5.6911) 
 (1.2777) 
 (0.1913)** 


 (0.0646) 
 (0.7296)** 


 (0.9577)** 
 (0.1688)* 
 (0.0566) 
 (1.7455)* 
 (0.2320)** 


 Hispanic 
 2.7389 
 0.9868 
 -0.4575 
 -0.0700 
 -0.6133 
 -0.7520 
 -0.0787 
 -0.0845 
 -0.2282 
 -0.0054 


 (1.2216)* 
 (0.2256)** 
 (0.0364)** 
 (0.0213)** 
 (0.1490)** 
 (0.1678)** 
 (0.0324)* 
 (0.0091)** 


 (0.2548) 
 (0.0368) 


 Constant 
 30.9175 
 2.9045 
 2.2312 
 -0.0971 


 3.4554 
 4.1867 
 0.5670 
 0.1283 
 7.4749 
 0.2913 


 (0.7707)** 
 (0.1155)** 
 (0.0436)** 
 (0.0144)** 
 (0.1139)** 
 (0.1146)** 
 (0.0261)** 
 (0.0083)** 
 (0.1372)** 
 (0.0194)** 


 Observations 
 178784 
 178784 
 178784 
 178784 
 178784 
 178784 
 178784 
 178784 
 178784 
 178784 


 R- 
 squared 


 0.10 
 0.13 
 0.16 
 0.15 
 0.09 
 0.16 
 0.09 
 0.12 
 0.04 
 0.06 


 Notes: 
 Robust 
 standard 


 errors 
 in 
 parentheses 


 – 
 se’s 


 are 
 clustered 


 on 
 CMSA. 


 * 
 significant 


 at 
 5%; 


 ** 
 significant 


 at 
 1%. 


 All 
 specifications 


 include 
 age 
 dummies, 


 education 
 dummies, 


 and 
 a 
 gender 


 dummy. 
80
 80 
 Table 
 18: 
 Summary 


 of 
 Statistical 


 Results 
 on 
 Substitution 


 TV  all  use 
 TV 
 all  outlets 


 TV  news  use 


 Radio  all  use 


 Radio  all  outlets 


 Radio  news  use 
 Radio  news  outlets 


 Dailies  use 
 Dailies  outlets 
 Weeklies  use 


 Weeklies  outlets 
 Internet  use 
 Internet  news  use 
 Cable  use 
 Cable  news  use 


 TV 
 all 
 use 


 E 
 D 


 E 
 H 


 D 


 E 


 TV 
 all  Outlets 
 TV 
 news  use 


 G 
 G 
 J 


 F 
 G 


 I 
 J 


 G 


 Radio 
 all 
 use 


 E 
 Radio 
 all 
 outlets 


 C 


 C 
 C 


 A 
 C 


 Radio  news  use 
 J 
 C 


 Radio  news  outlets 
 G 


 A 
 A 


 Dailies  use 
 J 
 A 
 D 


 G 
 J 


 Dailies  outlets 
 A 
 B 


 A 
 C 


 Weeklies  use 
 D 
 Weeklies  outlets 
 C 


 A 
 B 


 Internet  use 
 E  H 


 A 


 Internet  news  use 
 F 
 G 
 H 


 I 


 D 
 A 


 J 


 Cable 
 use 


 E 


 A 
 A 
 D 


 A 


 D 


 Cable  news  use 
 G 


 G 
 J 
81
 81 
 Symbols 
 denote 
 significance 


 at 
 5% 


 level 


 (except 


 as 
 noted 


 below). 


 We 
 register 


 the 
 correlation 


 measures 
 (A, 
 E, 
 F, 
 G) 
 symmetrically. 


 For 
 the 
 measures 


 based 
 on 


 regression 
 coefficients, 


 we 
 treat 


 the 
 column 


 as 
 the 
 dependent 


 variable 
 and 
 the 
 row 


 as 
 the 
 independent 


 variable. 


 Key:  A  =  correlation 
 of 
 changes 


 on 
 changes 


 in 
 aggregate 


 data 
 (see 
 Part 
 I, 
 Table 


 12), 
 listed 


 if 
 significant 


 at 
 5% 


 level 


 B 
 = 
 regression 


 of 
 N 
 on 


 N 
 in 
 changes 


 (agg 
 data 
 – 
 see 


 Part 


 I, 
 Table 


 13), 
 listed 


 if 
 t-  stat> 


 1, 
 bold 


 if 
 significant 


 at 
 5% 


 level 


 C 
 = 
 regression 


 of 
 use 


 on 
 N 
 in 
 changes 


 (agg 
 data 
 – 
 see 


 Part 


 I, 
 Table 


 14), 
 listed 


 if 
 t-  stat> 


 1, 
 bold 


 if 
 significant 


 at 
 5% 


 level 


 D 
 = 
 regression 


 of 
 use 


 on 
 use 


 in 
 changes 


 (agg 
 data 
 – 
 see 


 Part 


 I, 
 Table 


 15), 
 listed 


 if 
 t- 
 stat> 


 1, 
 bold 


 if 
 significant 


 at 
 5% 


 level 


 E 
 = 
 correlation 


 of 
 overall 


 use 
 (micro 


 data 
 – 
 see 


 Part 


 II, 
 Table 


 9) 


 F 
 = 
 correlation 


 of 
 news 


 use 
 (micro 


 data 
 – 
 see 


 Part 


 II, 
 Table 


 10) 


 G 
 = 
 correlation 


 of 
 news- 


 ent 
 gap 
 (micro 


 data 
 – 
 see 


 Part 


 II, 
 Table 


 11) 


 H 
 = 
 micro 


 regression 


 of 
 use 


 on 
 use 


 (overall 


 media 
 use 
 measures 


 – 
 see 


 Part 


 II, 
 Table 


 12) 


 I 
 = 
 micro 


 regression 


 of 
 use 


 on 
 use 


 (news 


 measure 


 – 
 see 


 Part 


 II, 
 Table 


 13) 


 J 
 = 
 micro 


 regression 


 of 
 use 


 on 
 use 


 (gap 


 measure 


 – 
 see 


 Part 


 II, 
 Table 


 14) 
82