merger simulation in a two sided market

24
Merger Simulation in a Two-Sided Market: The Case of the Dutch Daily Newspapers Lapo Filistrucchi, TILEC and CentER, Tilburg University Tobias J. Klein, TILEC, CentER and Netspar, Tilburg University Thomas Michielsen, CentER and TSC, Tilburg University Media Economics Workshop, Moscow 28 October 2011 This paper is based on an empirical study performed for the Dutch competition authority (NMa). The views expressed in this paper are not necessarily the ones of the NMa.

Upload: new-economic-school

Post on 04-Jul-2015

367 views

Category:

Technology


3 download

DESCRIPTION

Merger Simulation in a Two-Sided Market: The Case of the Dutch Daily Newspapers. Lapo Filistrucchi, TILEC and CentER, Tilburg University. Tobias J. Klein, TILEC, CentER and Netspar, Tilburg University. Thomas Michielsen, CentER and TSC, Tilburg University. Media Economics Workshop, Moscow. 28 October 2011

TRANSCRIPT

Page 1: Merger simulation in a two sided market

Merger Simulation in a Two-Sided Market:

The Case of the Dutch Daily Newspapers

Lapo Filistrucchi, TILEC and CentER, Tilburg University

Tobias J. Klein, TILEC, CentER and Netspar, Tilburg University

Thomas Michielsen, CentER and TSC, Tilburg University

Media Economics Workshop, Moscow

28 October 2011

This paper is based on an empirical study performed for the Dutch competition

authority (NMa). The views expressed in this paper are not necessarily the ones of

the NMa.

Page 2: Merger simulation in a two sided market

Newspapers as Two-Sided Platforms Newspapers‘ publishers sell content to readers and advertising

slots to advertisers

taking into account that

advertisers care about the number of readers

and that

readers may be affected by the number of ads (or by advertising

concentration) in the newspaper.

In addition, advertisers cannot pass-through to the readers any

increase in the advertising tariff paid to the publshers because

there is no direct transaction between them.

Page 3: Merger simulation in a two sided market

Aim of the paper

Develop a structural econometric framework that allows us to simulate the effects of mergers, by

- estimating demand for differentiated products on each side of the market.

- using the estimated parameters together with a model of the supply-side to recover costs

- simulate a merger and find the new equilibrium prices and quantities

- calculate the effects of the merger on consumer welfare

Apply it to the Dutch daily newspaper market.

Page 4: Merger simulation in a two sided market

The literature-1

Merger simulation in one-sided markets

-Hausman and Leonard (1997), Nevo(2000), Ivaldi and

Verboven(2005) and many others

See also Jaffe and Weyl(2011)

Mergers in two-sided markets

-Chandra and Collard-Wexler(2009), Lionello(2010)

Anderson and MacLaren (2010), Malam(2011)…

Page 5: Merger simulation in a two sided market

The literature-2

Merger simulation in two-sided markets:

-Fan(2010)-US newspapers, mixed logit for readers, Rysman(2004)

for advertisers, no effect of ads on readers

-Van Cayseele and Vanormelingen (2010) – Belgian newspapers,

nested logit for readers, Rysman(2004) for advertisers, no

effect of ads on readers

-Jeziorski (2010) – US radio, mixed logit for listeners, log linear

demand for advertisers, no price for listeners but negative

effect of ads

-Song (2010) – German TV magazines, logit on readers side, either

logit or similar to Rysman(2004) and GLS(2002) on advertising

side, positive effect of advertising on readers

Page 6: Merger simulation in a two sided market

Indirect network effects

in the newspaper markets

Well-known and typically found in the empirical literature on media markets that demand for advertising in newspapers depends positively on their circulation.

Not generally found that readership demand depends on amount of advertising:

-Argentesi and Filistrucchi (2007), Van Cayseele and Varnomelingen (2010) and Fan (2010): no effect of advertising on the number of readers of daily newspapers in Italy, in Belgium and in the US

-Kaiser and Wright (2006) and Kaiser and Song (2009) and Song(2010): advertising increases readers demand for magazines in Germany.

-Wilbur(2008) and Jeziorski(2010): advertising affects negatively TV viewers and radio listeners in the US

Page 7: Merger simulation in a two sided market

Dutch daily newspapers 1999-2009

Page 8: Merger simulation in a two sided market

Readers demand-I

Potential market is total population above 14 years

Each reader buys at most one newspaper.

Reader i utility of reading newspaper j at time t in market m (with )

while the utility from buying the outside good is

where and are type 1 extreme value and i.i.d. across readers and newspapers and

n

iotmii

n

otm

n

iti

n

iotm vDyu 00n

iotm

n

ijtm

n

jR

n

t

a

jtii

n

jt

nn

iti

n

ijtm mjtqxpyu )(

)(and)(with, D~PDv~PvvD viviii

i

i

i

n

ijtm

mIi

Page 9: Merger simulation in a two sided market

Readers demand-II

Thus assumptions:

Page 10: Merger simulation in a two sided market

Advertisers demand

The quantity of advertising in a newspaper j at time t is

given by

Thus, assumptions:

- Price per reader is the price that matters

- No cross price or network effects (as in

Rysman(2004))

- Constant (price per reader) elasticity

a

jt

a

t

a

j

aa

jtr

jt

a

jtaa

jt Xq

pq loglog

Page 11: Merger simulation in a two sided market

Dutch Daily Newspapers Data Quarterly national level HOI data on circulation

Subscription prices from … (91% of circulation is subscriptions for the non-free newspapers).

From Lexis-Nexis data on newspaper characteristics (number of times certain keywords related to different topics have been mentioned).

NOM Print monitor data on reader characteristics (gender, age, wealth, region, percentage bread winners, percentage shopping for groceries).

CBS (Statistics Netherlands) on total population over 14 and distribution of age and gender at municipality level

Nielsen data on:

-number of advertising column millimeters and pages

-total number pages

-advertising revenues

then calculate:

-percentage advertising pages

-average price (dividing advertising revenues by advertising quantity)

Page 12: Merger simulation in a two sided market

Readers Demand Estimates

Page 13: Merger simulation in a two sided market

AD1 BAK BND BRA EIN GEL GOO HAR LEI LEW LIM NED NOO NOR NRC NRN

AD1 -2.480988 0.001473 0.015069 0.008825 0.005408 0.011402 0.00281 0.002213 0.007568 0.00842 0.00574 0.013564 0.009762 0.006527 0.105437 0.022052

BAK 0.080895 -1.719475 0 0 0 0.027805 0 0 0 0 0 0.026784 0 0 0.027291 0.007408

BND 0.059596 0 -2.082449 0.013279 0 0 0 0 0 0 0 0.001925 0 0 0.057563 0.012012

BRA 0.029287 0 0.011143 -2.213303 0.006737 0.000873 0 0 0 0 0 0.001911 0 0 0.062226 0.012432

EIN 0.021759 0 0 0.008169 -2.025203 0 0 0 0 0 0.000533 0.001603 0 0 0.064846 0.010327

GEL 0.032996 0.001466 0 0.000762 0 -2.21955 0 0 0 0 0.000963 0.004458 0 0 0.072447 0.01376

GOO 0.045426 0 0 0 0 0 -2.277802 0 0 0 0 0.008927 0 0 0.180162 0.027188

HAR 0.024741 0 0 0 0 0 0 -2.286363 0.001856 0 0 0.004585 0 0.006979 0.155682 0.028141

LEI 0.108459 0 0 0 0 0 0 0.002379 -2.367648 0 0 0.008102 0 0 0.155678 0.027646

LEW 0.042531 0 0 0 0 0 0 0 0 -1.958036 0 0.011644 0.010311 0 0.034636 0.009228

LIM 0.018478 0 0 0 0.000426 0.001071 0 0 0 0 -2.108796 0.000604 0 0 0.051237 0.009004

NED 0.176241 0.006339 0.006324 0.007482 0.005176 0.020016 0.007175 0.005328 0.007346 0.02995 0.002439 -3.155627 0.114695 0.011919 0.079783 0.016209

NOO 0.03218 0 0 0 0 0 0 0 0 0.006729 0 0.029098 -2.009618 0 0.055347 0.013059

NOR 0.021051 0 0 0 0 0 0 0.002013 0 0 0 0.002958 0 -2.069238 0.054175 0.010189

NRC 0.179735 0.000847 0.024811 0.031964 0.027471 0.042675 0.018997 0.023736 0.018518 0.011688 0.02713 0.010467 0.028621 0.028634 -3.483968 0.030703

NRN 0.149584 0.000915 0.020602 0.025412 0.017408 0.032252 0.011408 0.017073 0.013086 0.012391 0.018971 0.008462 0.026872 0.021429 0.122174 -2.283077

PAR 0.03235 0.000178 0.001495 0.001828 0.001307 0.002473 0.011824 0.017332 0.002224 0.001545 0.000998 0.003486 0.001981 0.043133 0.192984 0.04077

PZC 0.083195 0 0.082062 0 0 0 0 0 0 0 0 0.009949 0 0 0.055833 0.009793

REF 0.273722 0.019828 0.010355 0.018362 0.000365 0.031518 0.004258 0.001561 0.009444 0.004693 0.00017 0.025697 0.005611 0.003436 0.06605 0.01212

STE 0.047556 0 0 0 0 0.01135 0 0 0 0.001815 0 0.032244 0.000738 0 0.062519 0.011511

TEL 0.127816 0.003486 0.021412 0.028108 0.02031 0.038617 0.014009 0.020412 0.01291 0.027744 0.020257 0.011467 0.029802 0.056574 0.094729 0.018681

TRO 0.195305 0.003743 0.01748 0.019309 0.014964 0.03942 0.011875 0.015094 0.013555 0.027655 0.012737 0.01731 0.056237 0.03162 0.119702 0.023485

TWE 0.020654 0 0 0 0 0.014887 0 0 0 0 0 0.007836 0 0 0.038072 0.008386

VOL 0.144813 0.001071 0.023398 0.032329 0.022944 0.048248 0.013136 0.021131 0.014635 0.016847 0.027413 0.010491 0.034939 0.047032 0.130548 0.027754

Readers demand price elasticities

PAR PZC REF STE TEL TRO TWE VOL

0.006318 0.010121 0.034882 0.014408 0.178184 0.054173 0.005155 0.091211

0.001911 0 0.138728 0 0.266817 0.057005 0 0.037047

0.001155 0.039484 0.005219 0 0.118053 0.019175 0 0.058283

0.001185 0 0.007765 0 0.130034 0.017773 0 0.067573

0.001027 0 0.000187 0 0.113934 0.016702 0 0.058151

0.001398 0 0.011624 0.009951 0.155793 0.031642 0.010752 0.087943

0.037334 0 0.008771 0 0.31571 0.053248 0 0.133757

0.037849 0 0.002224 0 0.318152 0.046812 0 0.148809

0.006225 0 0.017247 0 0.257921 0.053882 0 0.132096

0.001524 0 0.003021 0.002777 0.195376 0.038749 0 0.053602

0.000628 0 6.96E-05 0 0.090915 0.011374 0 0.055586

0.008846 0.015728 0.042552 0.126931 0.20772 0.062387 0.025411 0.085863

0.001275 0 0.002357 0.000737 0.136955 0.051422 0 0.072543

0.02717 0 0.001412 0 0.254359 0.028287 0 0.095539

0.064253 0.011579 0.014349 0.032288 0.225115 0.0566 0.016197 0.140167

0.054015 0.008082 0.010477 0.023657 0.176649 0.044188 0.014196 0.118576

-2.684122 0.000829 0.002617 0.0034 0.283576 0.048574 0.00137 0.177663

0.001331 -1.99269 0.061114 0 0.15267 0.01749 0 0.046935

0.004012 0.058343 -2.821956 0.075963 0.214497 0.058608 0.024954 0.061041

0.002192 0 0.031953 -2.250627 0.200059 0.0522 0.017384 0.075619

0.03973 0.013323 0.019608 0.043477 -2.471059 0.046002 0.026373 0.099511

0.034203 0.007671 0.026926 0.057014 0.231201 -3.29951 0.020427 0.115651

0.001072 0 0.012742 0.021104 0.147317 0.022703 -1.895978 0.041247

0.055093 0.009066 0.01235 0.036373 0.22025 0.050931 0.016344 -3.023657

Page 14: Merger simulation in a two sided market

Impact of the network effects

Page 15: Merger simulation in a two sided market

Profit Maximization-I

Profits are

We cannot obtain first-order conditions in the usual way, since

and

But if it were possible to get

and … ),(~ a

t

n

t

n

t

n

t

n

t ppsMq ),(~ a

t

n

t

a

t

a

t

a

t ppsMq

Page 16: Merger simulation in a two sided market

Profit Maximization-II

So if we could solve for market-shares as functions of prices on both sides, we could rewrite as

with associated first-order conditions

Page 17: Merger simulation in a two sided market

Profit Maximization-III

We thus need where

Define

where

n

jt

a

jtna

jkp

sS

ˆˆ

a

jt

n

jtan

jkp

sS

ˆˆ

Page 18: Merger simulation in a two sided market

Profit Maximization-IV

Also define where

Page 19: Merger simulation in a two sided market

Profit maximization-V

By the implicit function theorem,

so that exists if B is non-singular…

If exists, then a solution to the first order conditions

exists (existence?)

S

S

Page 20: Merger simulation in a two sided market

Profit maximization-VI

Define an ownership matrix as in Nevo (2001)

where if product j belongs to firm r and

otherwise

Also define

where

so that the first-order conditions are (unique? maximum?)

*

t 1* jrΩ oΩ jr *

Page 21: Merger simulation in a two sided market

Recovering marginal costs

From the first-order conditions

one can obtain the mark-ups

and, by subtracting them from the observed prices,

then one obtains the marginal costs

(note that for the observed market shares and prices there is a unique

vector of marginal costs solving the f.o.c.s)

Page 22: Merger simulation in a two sided market

Full merger simulation: new equilibrium

We solve (numerically) for the new equilibrium

(stability?)

Existence, uniqueness…: our strategy

We look for restrictions on the demand functions which guarantee

existence

Weyl and White(2011) restrict the firms behaviour

Page 23: Merger simulation in a two sided market

Estimates of marginal costs

Page 24: Merger simulation in a two sided market

Summary and conclusions

• Developed a structural econometric framework to analyze hypothetical mergers in a two sided market.

• It allows to simulate quantities and prices of a post-merger equilibrium, the associated welfare changes and, in case, the productive efficiency gains necessary to counterbalance a welfare loss.

• We applied it to the Dutch market for daily newspapers and found that an hypothetical merger has some effect on prices and welfare, and that it is important to take the network effect into account. (preliminary, may change with mixed logit)

• Upcoming:

formal proofs of conditions for existence, uniqueness and stability of equilibria

a better specification for the advertising demand

a simulation using the insulating tariffs approach of Weyl and White(2011)