Académique Documents
Professionnel Documents
Culture Documents
1
K
i k ik i
k
y x
=
= + +
'&(
G
with y
i
as the dependent variable, x
ik
as - independent variables, Q
<
and Q
k
are -H& unknown parameters to be estimated, R
i
is a two.sided
independent and identically distributed error term.
!o further improve the empirical analysis, we performed a principal
component analysis. In this model, we frst executed a principal
component analysis to reduce the data to useful basic dimensions. In a
second step, a regression analysis was executed only using the basic
dimensions and still keeping the relevant information. In this way, possible
correlation problems between the amount of variables can be avoided.
Nrincipal component analysis involves a mathematical procedure that
transforms a number of possibly correlated variables into a smaller
number of uncorrelated variables called principal components. It can be
described as an orthogonal linear transformation that transforms the data
to a new coordinate system such that the greatest variance by any
pro3ection of the data comes to lie on the frst coordinate 'called the frst
principal component(, the second greatest variance on the second
coordinate, and so on. s rotation method, a varimax with -aiser
normali/ation and a eigenvalue boundary of one is used.
%. &esults
%.1 Descri!tive statistics
,e found that in )<<) only *F; '0< bars in our sample( of #elgian bars
were independent, most other bars had exclusivity agreements or were
owned by brewers or beer merchants. *0 bars '&D.F;( had agreements
with beer merchants, while &)& bars '+=.+;( had vertical agreements with
breweries. !he minimum beer price 'lager beer( was <.G+ $uro while the
maximum beer price was ).+= $uro. !he average beer price was &.&0
$uro. !he distribution of the beer price of all )D< bars in our sample is
shown in fgure ).
=
Figure )% 4istogram of the lager price
%.2 #m!irical analysis
!able * and + show the results of the diferent regression analysis. !he frst
regression shows that bar type, bar location, province and the share of
beer in the total turnover are important determinants of the beer price in
pubs. @egression & shows clearly that the price in urban, youth and club
bars are signifcant higher then in regular bars. Further, we see that the
price in city bars is higher than the price in city borders, rural centers and
the country side, and lower than in city center bars. ,e also see signifcant
price diferences between provinces. Furthermore, regression & in table *
shows that the share of beer in the total turnover has a signifcant
negative impact on the beer price, indicating higher beer prices in bars
with a lower share of beer sales. In certain bars, customers can eat snacks
and simple meals together with a higher priced beer. lso, bars that
primarily serve other beverages such as cocktails and wine, but still serve
beer, tend to set higher prices. Finally, we see that there is no signifcant
impact of the exclusivity agreements on the beer price, and a small
positive signifcant impact of ownership, indicating a higher beer price in
independent bars.
!able *% regression analysis 'part I(
"ariables &egression 1 &egression 2 &egression 3
0
'oef. t(value 'oef. t(
value
'oef. t(
value
#ar type
@egular bar . . . . . .
Mouth bar <.&<F ).G=<SSS <.&<G ).G=GSS
S
. .
Lrban bar <.&*= D.+)DSSS <.&*= D.+<+SS
S
. .
1lub bar <.<0+ *.&*0SSS <.<0* *.<=)SS
S
. .
#ar location and density
1ity centre <.<=* ).&*+SS <.<GD &.0*DS <.&<= ).F+*SS
S
1ity . . . . . .
1ity border .<.<&= .<.DDG .<.<&G .<.D&& .<.<*) .<.0&*
@ural centre .<.&)+ .*.=00SSS .<.&)) .
*.=&*SS
S
.<.&*& .
*.0+GSS
S
1ountry .<.&)G .*.)*GSSS .<.&)G .
*.)))SS
S
.<.&*= .
*.*)FSS
S
#ar density <.<&) <.G<+ <.<&= &.<** <.<<0 <.+GD
Nrovince
ntwerp . . . . . .
Flemish #rabant .<.<&+ .<.+=) .<.<<0 .<.*<) .<.<)* .<.G)+
,est Flanders .<.<G* .).+**SS .<.<G& .).*+GSS .<.<DD .&.G++S
$ast Flanders .<.<FF .).&DFSS .<.<F+ .).<0*SS .<.<+< .&.)D&
Iimburg .<.<0= .).GD<SSS .<.<0G .
).F=)SS
S
.<.<FG .&.GG0S
#ar characteristics
mount 4I lager .<.<<+ .<.*FG .<.<<+ .<.+&& .<.<<+ .<.+<G
share beer .<.<** .*.F)GSSS .<.<** .
*.FF+SS
S
.<.<+& .
+.*0FSS
S
#eer diversity .<.<<& .<.0G+ .<.<<& .<.==) <.<<< <.&)*
#ar legal situation
>wner <.<D) &.0*+S <.<D& &.=))S
$xclusivity agreement .<.<&= .<.GD& <.<<* <.&** .<.<) .<.G0<
@T <.+*& <.+)) <.*+0
Bependent variable% beer price 'lager(
S signifcant at &<;, SS signifcant at D;, SSS signifcant at &;
In @egression ) E F 'table * and +(, some variables are left out to investigate the
robustness of the impact of the diferent variables. In this way, the impact of
possible correlation between certain variables 'e.g. location and bar density( can
be studied. For example leaving out the diferent bar location variables, results in
a signifcant impact of bar density on the beer price. naly/ing the diferent
regression models, we can conclude that exclusivity agreements have no
signifcant impact on the beer price. Signifcant beer prices are found in diferent
&<
bar types, in bars with a diferent location, and in bars with a diferent share of
beer sales in their total turnover.
!able +% regression analysis 'part II(
"ariables &egression % &egression ) &egression *
'oef. t(value 'oef. t(
value
'oef. t(
value
#ar type
@egular bar . . . . . .
Mouth bar <.&&= *.<&)SSS <.&)G *.)&*SS
S
<.&<G ).=<0SS
S
Lrban bar <.&DF F.<F0SSS <.&+0 D.FGDSS
S
<.&*G D.+&<SS
S
1lub bar <.<0* *.<)=SSS <.<GG ).D<)SS <.<0* *.&&DSS
S
#ar location and density
1ity centre <.<=0 ).)*&SS . . <.<0F ).=*FSS
S
1ity . . . . . .
1ity border .<.<&& .<.*** . . .<.<)= .<.0<+
@ural centre .<.&&0 .*.F*=SSS . . .<.&*+ .
+.FD0SS
S
1ountry .<.&)= .*.&=*SSS . . .<.&+* .
+.D&GSS
S
#ar density <.<&G <.0D& . . . .
Nrovince
ntwerp . . . . . .
Flemish #rabant .<.<&= .<.DF0 .<.<&F .<.D&* .<.<&D .<.D&D
,est Flanders .<.<=F .).G00SS .<.<G* .).**0SS .<.<G+ .).+G<SS
$ast Flanders .<.<FG .).&*FSS .<.<=* .
).FD+SS
S
.<.<F= .).))DSS
Iimburg .<.<0F .).F<GSSS .<.<=0 .).+<&SS .<.<&< .
).=<)SS
S
#ar characteristics
mount 4I lager .<.<<D .<.+G* .<.<<D .<.+D0 .<.<<+ <.*FF
share beer . . .<.<** .
*.D*+SS
S
.<.<** .
*.F0+SS
S
#eer diversity .<.<<& .<.=)G .<.<<& .&.<D< <.<<& <.0)&
#ar legal situation
>wner <.<D+ &.0=0SS <.<++ &.D00 <.<D+ ).<=)SS
$xclusivity agreement .<.<)D .&.<&G .<.<<0 .<.*++ .<.<&0 .<.G0+
@T <.*00 <.*G< <.+*<
Bependent variable% beer price 'lager(
S signifcant at &<;, SS signifcant at D;, SSS signifcant at &;
&&
!o broaden the empirical analysis, a principal component analysis can be
executed to reduce the amount of variables and to structure the data. In
this way, we can construct diferent independent components and use
these components in a second step regression analysis. ?ote that we frst
redefned certain variables to reduce the amount of dummy variables. !he
variable bar location was created with a range of &.D, representing the fve
diferent locations, the variable bar type was created with a range &.+,
representing the four diferent bar types. !he variable province was
created with a range &.D, representing the fve diferent provinces. In this
way, the variables can be used to measure potential signifcant impact,
but the variables can not be used to explain the direction of the
correlations. s shown in table D, the frst principal component consists of
the variables bar location and bar density, the second principal component
consists of the variables owner and exclusivity agreement, the third
principal component consists of the variable bar type and the fourth
principal component consists of the variable amount 4I lager, indication
the beer turnover.
!able D% !he diferent principal components
"ariables 'om!onent
1 2 3 %
>wner <.&FD <.G=F <.<&< <.<*=
$xclusivity agreement <.<&+ <.=)+ .<.&*G <.<<=
mount 4I lager .<.<*= <.&<G <.&)D <.0<+
share beer .<.+G) <.))) .<.&+* .<.<+&
#ar location <.=++ <.)*+ .<.<F* <<&&
#ar density <.=F< <.)*0 .<.<&+ .<.<<D
#ar type .<.<&< .<.&*& <.GDF <.<&F
Nrovince <.*<F .<.)=0 .<.+F* <.+DD
#eer diversity <.*=+ .<.<D* <.D&D <.)++
$xtraction method% principal component analysis, rotation method% varimax with -aiser
normali/ation
!able F shows the results of the regression analysis with as dependent
variable the beer price and as independent variables the principal
components. !he results confrm the previous regression models that the
beer price is mainly determined by the bar location and bar type. ,e did
not fnd a signifcant impact of the legal and contractual relation on the
beer price.
!able F% @egression analysis of principal components
1oe5cient St. error
1onstant &.&=GS <.<&<
1omponent & 'bar location, bar density( <.&<+S <.<&<
1omponent ) 'owner, exclusivity agreement( <.<&G <.<&<
&)
1omponent * 'bar type( <.<)0S <.<&<
1omponent + 'amount 4I lager( <.<&< <.<&<
?umber of observations )D<
d3 @T <.*<
Bependent variable% beer price2 Ssignifcant at &;
). 'onclusion
!o conclude, we found no impact of existing vertical agreements between
brewers and bars on the beer price. spects such as bar location and bar type are
more important to determine the beer price than the existence of vertical
agreements. !his suggests that the beer selling price in pubs is more determined
by observing the price in competitive bars, then that it is determined by the cost.
s such, independent bars may set the same price as the ma3ority, which are
bars with an exclusivity contract 'loan agreements or lease contracts(, but earn a
higher margin. >n the other hand, it is often said that free bars have higher fxed
costs, since they have loans with external fnancing bodies such as banks.
?evertheless our research does not show that vertical agreements have no
negative market impact but only that they have no impact on the beer price.
!o investigate the relation between vertical agreements and beer price in more
detail, more specifc data is necessary 'e.g. diferent contractual relationships(.
From an econometric point of view, observations during several years 'panel
data( would also improve the soundness of the empirical analysis. !o determine
beer prices, bar managers are more considerate of competing bars8 prices than of
their own direct beer purchase cost. 4ence the inUuence of vertical integration on
the beer price should ideally be investigated using panel data. 4owever, since
data of a beer market without exclusivity contracts does not exist, such analyses
cannot be performed.
&eferences
#enson.rmor, @., Ieibowit/, V., @amachandran, B. &000, Wlobal #eer% ,hat8s on
!apX !he 9c-insey Yuaterly, & '&000(% &&&.&)&.
1##, s.d., Lnion of #elgian #rewers 'http%OOwww.beerparadise.be(
?IS, s.d., #elgian ?ational Institute for Statistics
'http%OOwww.statbel.fgov.be(
&*
Schumpeter, &0+). 1apitalism, Socialism and Bemocracy
Slade, 9.$., &00=. #eer and the tie% did divestitute of brewer.owned public
houses lead to higher beer pricesX !he $conomic Vournal &<=% DFD.F<)
&+