Vertical relations and local competition: an
empirical approach
Abstract
Analysis of the vertical relationships in industries is an important aspect of industrial
organization. In this paper we test, on the Chilean petrol market, some of the
hypotheses derived from transaction cost theory and the principal-agent problem. We
show that the local competition has an important role in the choice of a disintegrated
vertical structure, and that low levels of service investment have the same effect.
Conversely, the number of own brand outlets and a high level of investment in services
reduces the probability of disintegration. We demonstrate that the disintegrated
vertical structure has a positive effect on wholesale and retail prices, to a magnitude of
between 1.6 and 7 per cent, depending on the kind of petrol.
JEL Codes: L11, L22, L42
Keywords: Vertical structure, local competition, petrol market, services and prices
1. Introduction
Why companies buy or produce certain products and services is an important question
for the industrial organization, especially in retailing. It is important know the limits of
the firm and why it increases or reduces its number of activities1. One traditional
argument used to explain why companies produce or purchase activities, such as
retailing, was the financial problems that small firms have when quickly expanding
their businesses within the market (Caves and Murphy, 1976). Small firms do not have
full access to capital markets, and franchising is a cheaper source of finance. This
theory predicts that only small firms disintegrated some stores to monitor their
profitability, and internalize the most profitable. In this case disintegration is not a
rational decision chosen by the firm as the most efficient vertical structure, but a
financial necessity.
The fact that the researchers have noted that not only small companies franchise
outlets begins to cast doubt on the validity of this theory. Researchers have started to
observe that large companies franchise all or part of their retailing, so there must be
other reasons. Authors such as Rubin (1978) criticize this theory, because the risk of
investing in a single retail outlet is greater than in shares of different stores. Riskaverse agents would need a greater return, when investing in a single retail outlet, as
this is more expensive for the firm than offering a share of the investment to the
franchisee. So, problems of accessing capital markets are not the only factor affecting
the decision to break down part of the product chain.
In contrast to this approach others authors like Skivlas (1987) show the vertical
relationship as a rational decision to create the most efficient format. The most
popular of these theories are transaction cost analysis and the positive economic
theory, which analyze vertical relationships as a response to the principal-agent
problem2. With transaction cost theory the efficient vertical structure has to minimize
the transaction cost, the cost when a good or service is transferred across a
technologically separable interface. The principal-agent problem increases the
transaction cost, as it generates price and effort control costs from the retailer; there is
also the relationship between these two theories3.
The different theoretical models establish the relationships between the
characteristics of firms and the probability of disintegration, but in general show how
the integrated vertical relationships have limited the incentives for retailers to make
much effort. With the integrated system retailers perceive a fixed wage, and so have a
1
See Lafontaine and Slade (2007) for an extensive summary of the vertical relationships in different
industries.
2
See Katz (1989) for an extensive summary of the positive theory of vertical restraints and Williamson
(1981) or Lafontaine and Slade (1996) for transaction cost analysis.
3
Some papers that compare theoretic and empirically the restriction finance problem and the
transaction cost are: Brickley and Dark (1987), Carney and Gedajlovic (1991) and Kehoe (1996).
little incentive to make much effort. The disintegrated system increases the incentives
to make an effort; however, at the same time it increases other principal-agent
problems, such as double marginalization of prices, setting a level of services and
making an effort under the efficient firm level.
Local competition has an important role in this trade-off. Firstly, local competition
eliminates the problem of double marginalization, which is a main variable in the
principal-agent problem. Secondly, it may moderate or eliminate the costs of effort
and of service quality controls in a disintegrated channel. The competitive
environment means that the private managers offer price and quality of services at the
market equilibrium. It is not possible for private managers to take advantage of
transaction costs or incomplete contracts, because the market imposes the prices and
quality. The effect of local competition was presented by Williamson (1979) who
showed theoretically that, in the presence of competition, vertical disintegration
dominates vertical integration. The empirical literature has scarcely tested this
important theoretical relationship.
One main improvement of this paper is the introduction of local competition into the
empirical specification as an important element in explaining the choice of the vertical
relationship between firms and retailers, in contrast with other authors that analyzed
this same aspect, John and Weitz (1988), Levy (1985), Minkler (1990) or Baker and
Hubbard (2003, 2004). The other is to test the effect of different vertical relationships
on wholesale and retail prices, taking into account the endogenous character of
vertical type relationships.
We have found very little empirical evidence on the effects of local competition in the
vertical structure of the firms, and there is still less evidence about the effects of these
different structures on the wholesale and final prices. If the companies choose the
different structures in an efficient way, and we control for all the possible factors that
affect the price, then probably the type of vertical structure does not affect the price
levels. If the disintegrated outlets are situated in more competitive markets, we can
see low prices in them. However, if the local competition is not a perfect instrument to
control the price and effort, then probably the disintegrated structured has a higher
cost and consequently higher prices than the vertically integrated stores. With the
empirical specifications we have tried to estimate the effects of the different vertical
price control structures for all other factors that may affect prices.
These empirical implications are tested in the Chilean petrol market; petrol markets
are traditionally used for testing the different vertical relationships. This is because
there are clear and identifiable stages in the product channel, different vertical
relationships and a great number of different market characteristics that the different
empirical implications can be tested on; for example, services, level of competition,
investment necessities, etc.
The paper is organized as follows. In the next section we describe the main
characteristics of the Chilean petrol market, which is our reference market for testing
the empirical hypothesis. In Section 3 we develop the empirical hypothesis in the light
of the transaction costs and principal-agent problem models. In Section 4 we present
the data used to test the hypothesis, and in the Section 5 we present the empirical
specifications and results. Our conclusions are presented in Section 6.
2. The Chilean gasoline market
In the Chilean petrol market we can distinguish three segments: the refinery market,
the wholesale market and the retail market.
The refinery market is characterized by public intervention. The Government owns all
the refineries, and the price is regulated by a formula that functions in accordance with
the international price for refined petrol. The public institution that fixes the refined
petrol prices is “Empresa Nacional del Petróleo” (ENAP or the Chilean National
Petroleum Company). All companies have to pay the same price to ENAP, regardless.
The wholesale market is made up of the companies that buy the refined petrol from
ENAP; they are COPEC, ESSO, SHELL and YPF. They sell it to the petrol stations,
normally to petrol stations of the same brand. The price companies pay ENAP is
regulated and equal for all, but the wholesale prices charged to the different petrol
stations are free from restrictions. The companies can charge a different wholesale
price, not only in function of the different costs like transport costs, different labour
costs, etc, but also in function of strategic variables such as different vertical
relationships.
Finally, the retail segment presents the same companies as the wholesale market plus
the independent retailers. The different operators freely fix both the wholesale and
retail prices. However, the setting the final price is not absolutely free, and depends on
the vertical relationship between the company and the retail operator. Some petrol
stations are free to fix the retail prices, but others are set by the firms. In the next
paragraph we define the different vertical relationships, and how they affect the
freedom to fix the final price.
The vertical relationship between the companies and the retailers in the Chilean petrol
market is divided in two. It depends on whether the company owns the petrol, and if
the final price is fixed by the company or by the retail operator. If the company owns
the petrol until it is passed onto the consumer, then the retailer only works for the
company and earns a commission; this kind of retailer receives is a commission agent
and is like a company employee. On the other hand, if the company may only own the
petrol until the petrol station that resells to the consumer, receive the name of
reseller. Given this classification we can distinguish:
Resellers: Retailers own the petrol and sell the combustible to the consumers.
The company sells the petrol to the station and onto the consumers. The
volume sold and the margin together determines the profits, because the
retailer is free to fix the final price. We have denominated this a disintegrated
vertical relationship.
Commission Agent: With this kind of retailer the company maintains ownership
of the petrol until it is sold to the consumers. The commission agent is like a
company employee that sells the petrol to the consumers and receives a
commission, and whose profits depend both on the volume of petrol sold, and
on the commission fixed between the agent and the company. This commission
has traditionally a fixed component like a wage, and a variable component that
depends on the quantity sold, which increases the incentive to sell. This is
denominated an integrated vertical relationship.
In the Chilean petrol market the companies build the petrol stations and decide the
type of vertical relationship they want to have with the retailer, alternatively a
businessman may build the petrol station and decide which company to sign a contract
with. We have to take this characteristic into account, because the retailers in our
market sometimes have to decide the level of investment and assume the risk. This is
different in the United States, where the companies build the stations and then decide
on the type of vertical relationship that they have with the retailers.
So the Chilean petrol market is characterized by a public refinery sector that sells the
refined petrol at the same price to the four companies that operate in the wholesale
market. These four companies freely fix the wholesale price for the retail operators
and the independent retailers. This wholesale price can differ, not only with the
different costs, but also with other strategic variables like the different vertical
relationships4.
3. The role of local competition in vertical relationships.
As we have seen in Section One, local competition can play an important role in
electing the vertical structure. The local competition must positively affect the
disintegration for at least two reasons:
- The probability of double marginalization is eliminated. If competition exists,
then the private manager cannot fix a price level above the market equilibrium
price; so the company can disintegrate the service without the threat of the
private manager fixing a price higher than the efficient level for the company.
- Similarly, competitors make the private manager offer an efficient level of
service, both in terms of number and quality. If competition exists then the
private manager has to offer the number and the quality of services that the
market fixes. Thus, the competition eliminates the threat of underinvestment in
services and of a lower quality service than the efficient level for the company.
If stiff competition exists in the local market, then for the above reasons disintegration
vertical structure is the efficient structure for the company. As Williamson (1971, 1979)
and Anderson (1985) show, disintegrated vertical relationships are more efficient than
integrated systems and benefit competition. So Hypothesis 1 is:
H1: The level of competition in the local market positively affects the probability
of disintegrated the services.
Nonetheless, there are other characteristics that can affect the vertical structure.
These characteristics may or may not make vertical disintegration an efficient structure
for the firm.
4
For a more extensive explanation of the Chilean petrol market see Gómez-Lobo and Córdova (2006).
Concentration of Outlets.
The concentration of own brand outlets plays an important role in the decision of
whether to disintegrate. One reason to disintegrate the services is when private
managers have a greater incentive to make an effort with sales, compared to
employees who normally only receive a fixed salary. Can monitoring employee
behaviour and this control have a cost for the firm, if the company wants to assure
itself that employees make the correct level of effort? The fact that the stations are
very concentrated reduces this cost of monitoring, and increases the probability of
integrating the services; see Anderson and Schmittlein (1984), Anderson (1985) and
Brickley et al. (1991).
So the second hypothesis that we test empirically is:
H2: The concentration of outlets of the same brand has a negative effect on the
probability of disintegrating the service.
Level of Investment
The level of investment necessary for opening a new store negatively effects the
disintegration of the services, because private managers have more difficulties in
obtaining the correct financial resources. So, if the investment needed to open the
station is high, then it is less probable that the station operates under a disintegrated
structure.
H3: If the investment necessary to open a new store is substantial, then it is less
probable that the company operates under a disintegrated structure.
Level of Services
The level of services that a petrol station offers has a positive effect on the probability
of disintegration, because employees in the integrated system have no incentive to
make an effort. So, if the number of services where the firm has to control the level of
effort increases, then the probability of disintegrating this petrol station increases.
Authors such as Shepard (1993) have established that the level of services is the most
important characteristic when explaining the election of the vertical structure.
H4: The greater the number of services that the petrol station offers, then the
probability of disintegration structure increases.
We have to take into account that the level of services and the level of investment
often have a close relationship. If the petrol station offers a service that needs a great
investment, like vehicle repairs or a pharmacy, then these services have the two
conflicting effects on the vertical structure; i.e. a disintegrated system would more
probably be expected because it offers more services, conversely a less disintegrated
one due to the greater level of investment needed would more probably be expected.
We have to check the empirical analysis for the net effect of the different services,
with respect to the probability of disintegration.
There are two other traditional hypotheses tested in the literature that we do not test
in this paper. The first is the effect of transacting specific assets within the vertical
relationship. If the store has to invest in expensive specific assets, it would need a long
term contract to recover the investment. Sometimes the duration of the contracts are
limited by law, like in the European petrol market; and this means that a disintegrated
system provokes underinvestment. So, the investment in specific assets reduces the
probability of disintegration; see Williamson (1979), and check the empirical results of
Anderson and Schmittlein (1984). We can test this hypothesis because the petrol
market has no evident brand specific assets.
The second hypothesis, which is not tested in this paper, is the effect of the non-repeat
consumers. Retailers that have non-repeat consumers have no incentive to make any
effort or to provide quality services, thus they may reduce the brand value. For these
reasons the retailers situated in places with non-repeat consumers are more
probability vertically integrated. For the petrol market this hypothesis is tested on
motorway petrol stations. However, in Santiago, the Chilean capital, there are no
petrol stations on the motorways in the metropolitan area.
It is important to test the effect of the different characteristics within the vertical
structure, but it is more important check the effects of these different vertical
structures upon the wholesale and final price. For the wholesale price it is not easy
predict the possible effect of disintegration in price level. The main reason is because
with integrated firms the wholesale price is like an internal price, the same firm is both
the seller and the buyer, and possibly a real price is not reflected.
As commented in the previous section the vertical structures effect upon the retail
price can be ambiguous. If the disintegrated retailers are situated in a more
competitive environment they will probably fix lower prices. However, if local
competition is an imperfect instrument to solve the principal-agent problems, then
probably the disintegrated petrol stations will fix higher prices. From a theoretical
point of view the vertical structure can ambiguously affect the final prices, but we have
strong empirical evidence that demonstrates a positive effect between disintegration
and prices. By using different techniques authors like Barron and Umbeck (1984),
Shepard (1993), Blass and Carlton (1999) and Vita (2000) provide empirical evidence
for a positive relationship between disintegration and prices, although they didn’t take
the endogenous nature of the vertical structure into account.
The hypothesis presented in this chapter, and the effects of different vertical
relationships in prices, are tested in Section 5 using data from the Chilean petrol
market. In the next section we present the data used in the empirical specification.
4. Data
We have attempted to test these hypotheses in the Chilean petrol market using a
survey by the Chilean Government’s National Economic Prosecutor's Office (Fiscalia
Nacional Economica). We have been provided with information on the addresses,
types of contract, prices, quantities sold, services and brands. It consists of individual
daily data for 208 of the 470 petrol stations in Santiago’s metropolitan area. In the
following table we can see the statistical description of this survey. The data covers
January to September 2006.
Table 1. Statistical descriptive.
Retail price 93
Wholesale price 93
Quantity sold 93
Retail price 95
Wholesale price 95
Quantity sold 95
Retail price 97
Wholesale price 97
Quantity sold 97
Retail price Diesel
Wholesale price Diesel
Quantity sold Diesel
ENAP Price 93
ENAP Price 95
ENAP Price 97
ENAP Price Diesel
Number of rivals 0.5 miles
Number of rivals 1 mile
Number of rivals 2 miles
Own brand outlets 0.5 miles
Own brand outlets 1 mile
Own brand outlets 2 miles
Nº observations
208
109
207
170
74
163
207
106
206
206
107
205
208
208
208
208
208
208
208
208
208
208
Mean
604.29
587.96
2290.69
611.48
595.30
4787.18
624.20
605.96
1455.53
455.48
445.28
2267.30
542.40
561.55
580.62
413.58
2.01
6.16
24.34
0.75
2.01
8.48
Desv. Est.
7.92
14.89
1823.80
15.11
17.74
2781.41
10.58
13.49
1564.42
9.58
13.56
2063.74
------------1.87
4.41
14.56
0.98
1.93
6.87
In Table 1 we can see that the wholesales prices and the retail prices follow a logical
progression, the price of unleaded petrol 97 is more expensive than unleaded petrol
95, and this in turn is more expensive than unleaded petrol 93. The ENAP price follows
the same distribution pattern. This distribution is logical and reflects the increase in
the cost of obtaining high quality petrol. With respect to the number of rivals and own
brand outlets the data is rational too. The number of rivals and the number of own
brand outlets increase when the definition of the market increases. The other
characteristic that we can see is that in all the market definitions the number of rivals
is greater than the number of own brand outlets. More specific statistical relationships
among prices, number of rivals, own brand outlets and type of vertical relationship are
presented at the end of present section.
Although the general statistical descriptions do not show illogical characteristics,
another two aspects are needed to validate this database; they are the possibility of
self-selection problems in the data, and as to who we compute the number or rivals
and own brand outlets.
The fact that less than fifty per cent of the petrol stations answered the survey may
have introduced self-selection problems into the data. The possibility that only the
petrol stations with certain characteristics; for example low price, high level of
services, etc., or only some brands answered the survey introduces a possible bias into
the empirical specification. To test whether our sample is representative of the
regional petrol market, we compare our average weekly price with the price reported
by National Consumer Service (Servicio Nacional del Consumidor, SERNAC); this is a
representative survey published by the state owned company that is affiliated to the
Ministry of Economy. In the following graph we can see the two times series.
Graphic 1. Comparative evolution of prices: Data base Vs. CERNAC. (Unleaded gasoline
93)
We can see that the two series of data are very similar, with a correlation of 0.99 for
unleaded petrol 935.
The percentage of different types of vertical relationship for the whole market in our
data is similar. The percentage of resellers in the Chilean petrol market is 40.9 per
cent, and in our database this percentage is 40.5 per cent. So, in general terms, the
percentage of petrol stations in the two types of vertical relationship is representative
for the whole petrol market.
In conclusion, the prices and the two types of vertical relationship are representative
of the market, and for these reasons we are able to accept that our data is
representative of the petrol market in Santiago’s metropolitan area.
The second question is how we compute the number of rivals and own brand pumps in
the different geographical markets; i.e. half a mile, one mile and two miles. To resolve
this question we sought the longitudinal and latitudinal points using “Google Earth”
5
This result is very similar for the other kinds of petrol; i.e. unleaded petrol 95, unleaded petrol 97 and
diesel.
software. As we have the addresses for the 470 petrol stations, we consequently
located the geographical points for all the petrol stations in the market. It is important
to have the complete information, in order to take into account all the possible rivals
and own brand outlets for the pumps. With the geographical points we computed a
470 by 470 matrix with 220,900 distances that shows every petrol station with respect
to the other 469 pumps. These distances were computed using “Matlab” software.
With this same software we designed two programs to compute the number of rivals
and the number of own brand petrol stations in a half mile, one mile and two mile
radius.
After explaining these two factors we present some descriptive statistics that shed
light on the result of our hypothesis.
Table 2. Number of rivals and own brand outlets for different types of vertical relations.
Rivals 0.5 miles
Rivals 1 mile
Rivals 2 miles
Own brand outlets 0.5 mile
Own brand outlets 1 mile
Own brand outlets 2 miles
Resellers Commission Agent
2.24
1.93
6.26
6.25
24.99
24.35
0.63
0.84
1.75
2.29
7.10
9.83
We can see that on average, the reseller’s type of vertical relationship has on average
more rivals than the commission agent’s for the different distances. On the other
hand, we can see that commission agents have a greater average number of own
brand outlets compared to retailers. These results are compatible with hypotheses H1
and H2, a greater number of rivals increases the competition and so increases the
probability of a disintegrated type of vertical relationship. In the same way, a greater
number of own brand outlets reduces the control costs and increases the probability
of an integrated vertical relationship.
Table 3. Level of services for different types of vertical relations.
Resellers
Wash-Machine
Lubrication
Food Shop
Pharmacy
Toilets
Cash-Machine
61%
60%
51%
1%
48%
50%
Commission
Agent
71%
68%
39%
2%
32%
35%
In Table 3 we show that the resellers have a greater percentage of outlets with food
shops, public toilets and cash dispensers, which are the services requiring a lower level
of investment. On the other hand, the commission agents have a higher percentage of
lubrication services, pharmacies and automatic car washes. These latter services
involve a greater level of investment and resellers may encounter financial problems
when trying offering these kinds of services. In Hypothesis Three (H3) we would expect
to find a negative relationship between disintegration, the retailer type of vertical
relationship, and the existence of high investment services such as lubrication,
pharmacies and automatic car washes. The existence of low investment services is
related to a greater probability of disintegration, because the commission agents have
low incentives to make an effort and controlling a greater number of services implies a
greater cost of control. So we would expect a positive relationship between the low
investment services, such as food shop, toilets and cash dispensers, and the probability
of a disintegrated type of vertical relationship; this is our fourth hypothesis.
Table 4. Retail price for different types of vertical relations.
Retail price 93
Wholesale price 93
Margin 93
Retail price 95
Wholesale price 95
Margin 95
Retail price 97
Wholesale price 97
Margin 97
Retail price Diesel
Wholesale price Diesel
Margin Diesel
Resellers
604.67
(7.40)
586.46
(9.89)
18.21
614.56
(7.84)
594.42
(9.95)
20.14
624.22
(8.64)
604.60
(10.16)
19.62
455.89
(8.58)
444.21
(10.09)
11.68
Commission Agent
604.04
(8.27)
590.87
(21.42)
13.17
608.73
(19.05)
598.76
(34.85)
9.97
624.20
(11.73)
608.60
(18.22)
15.60
455.20
(10.22)
447.32
(18.44)
7.88
*Standard deviations in brackets
We can see that resellers’ prices are higher than commission agents’ retail prices for all
kinds of petrol. In the previous section we have discussed the two effects that can
affect retail prices, but this data shows that the local competition can be an imperfect
instrument to eliminate the principal-agent problem. We have to check the
econometric specification to see if the type of vertical relationship is the reason for
this price differences, or whether they are due to other factors. The wholesale price
for the reseller is lower than for the commission agent, the reason is that for the
commission agent it is an internal transfer not a real price; so the wholesale prices
declared in the survey may be below the real price.
Given higher retail prices and lower wholesale prices the reseller’s type of vertical
relationship presents a greater margin than the commission agent’s. In fact the
reseller’s margin is 1.3 times that of the commission agent for unleaded petrol 93 and
97, 1.5 times for diesel, and two times for unleaded petrol 95. Nevertheless, we must
state that the commission agent’s is not a real retail margin, for the same reason that
the wholesale price cannot reflect a real value. We have to take into account this
possible bias when estimating the effect of the different vertical relationships on the
wholesale price.
5. Empirical specification and results.
In this section we define the empirical specification and the main econometric results
in two stages. In the first part we define the choice between the different vertical
relationships. In the second part we show the effect of this vertical relationship on the
wholesale and retail prices.
To arrive at the choice between the two different vertical relationships we estimated
the following logit specification, where the dependent variable takes the value one for
the disintegrated structure and zero for the integrated structure:
tvr   0   1nrivalsi   2nownbrandi   3cashmachi   4toiletsi  5 foodi
 6 pharmacyi  7 lub ricationi  8 washmachi   i
The two different types of vertical relationship depend on the number of rivals petrol
station i has, the number of own brand outlets, and a set of dummy variables that take
the value one if the pump offers this service and zero if not. We estimate the
specification with the half mile, one mile and two mile definitions of the variable,
number of rivals and number of own brand outlets; this is in order to contrast the
sensitivity of results with the definition of these variables. The anticipated signs for the
variables in this specification are shown in the following table:
Table 5. Expected signs in the equation of vertical relationships
Number of rivals
Number of own brand outlets
Wash machine
Pharmacy
Lubrication
Food shop
Toilets
Cash machine
Expected sign
+
+
+
+
To estimate the logit specification we collapse the database into the 208 petrol
stations average, because the characteristics that determine the vertical relationship
do not change over the period of reference. The econometric results obtained for the
logit specification are presented in Table 6:
Table 6. Econometric results of logit specification, choose of vertical relationships.
0.5 miles
1 mile
2 miles
Nº Rivals
0.139*
0.031
0.019*
(0.080) (0.035)
(0.011)
Nº own brand outlets
-0.268*
-0.126 -0.063**
(0.167) (0.089)
(0.026)
Cash-Machine
0.277
0.243
0.178
(0.456) (0.459)
(0.462)
Toilets
0.457
0.501
0.396
(0.331) (0.333)
(0.339)
Food
0.052
0.038
0.140
(0.462) (0.460)
(0.460)
Pharmacy
-0.036
-0.170
-0.142
(1.183) (1.222)
(1.275)
Lubrication
-0.061
-0.064
-0.030
(0.395) (0.381)
(0.380)
Wash-Machine
-0.433
-0.385
-0.339
(0.396) (0.386)
(0.390)
No observations
208
208
208
Wald Chi2
11.77
10.17
13.55*
(0.1619) (0.2536) (0.0943)
2
Pseudo R
0.0473
0.0392
0.0532
Robust Standard Deviation in brackets. P-value: ***(1%), **(5%), *(10%)
We can see in the previous table that with the three geographical market definitions
the logit estimation coefficients present the expected signs for all the variables. The
signs of the coefficients confirm the four hypotheses, although only the coefficient of
rivals and own brand outlets are statistically significance different from zero at a five
per cent of confidence level.
The logit specification shows that the level of competition positively affects the
probability of disintegration, and confirms our first hypothesis. The coefficient of the
number of own brand outlets is negative, which confirms that a reduction in the
control cost reduces the probability of disintegration; this is our second hypothesis.
We have two clear, different effects for the services offered by petrol stations. The
high investment services, lubrication, automatic car washes and pharmacies, have a
negative effect on the probability of disintegration. The fact that individual retailers
have financial constraints that do not permit this kind of investment explains this
relationship, and confirms our third hypothesis. On the other hand, the services that
not require high level of investment, like food shops, toilets and cash dispensers, have
positive effects on the probability of disintegration. These results confirm our fourth
hypothesis, that as the level of service increases so does the probability of
disintegration; this is because they increase the control costs in an integrated vertical
structure.
After analyzing the vertical structure election we now analyze the effect of these
vertical structures on the wholesale and final price. To estimate the effect of the
different types of vertical relationship on the wholesale price, we propose the
following specification:
wholesalepricei   0   1refinepricei   2trvi   3quantityi   i
tvr   0   1nrivalsi   2 nownbrandi   3washmachi   4 pharmacyi 
  5 lub bricationi   6 foodi   7toiletsi   8cashmachi   i
The wholesale price for petrol station i depends on the refined price which is fixed by
the public firm ENAP and is equal for all companies, the different types of vertical
relationship, and the quantity sold by the pump as the quantity bought may attract a
discount. The fact that the refinery prices fixed by ENAP are equal for all companies
means that these variables are constant, when we collapse the database. For this
reason we eliminate the constant term in the regression. We assume this type of
vertical relationship to be endogenous, so we use the same logit specification variables
as instruments.
The results using a Two Least Squares estimator with Instrumental Variables are shown
in the following tables:
Table 7a. Effect of the vertical relations in the wholesale prices, unleaded gasoline 93 (Two
Least Squares)
Type of vertical relationship
Quantity sell
Refined price
No observations
Hansen J Statistic
F-Statistic
0.5 miles
1 mile
2 miles
16.080*
13.709
10.955
(9.301)
(9.630)
(9.459)
0.004**
0.003**
0.003**
(0.002)
(0.002)
(0.002)
0.943***
0.947***
0.951***
(0.016)
(0.017)
(0.017)
109
109
109
8.446
7.435
8.237
(0.2949)
(0.3850)
(0.3121)
58925.83*** 56803.26*** 62532.17***
(0.0000)
(0.0000)
(0.0000)
Robust Standard Deviation in brackets. P-value: ***(1%), **(5%), *(10%)
Table 7b. Effect of the vertical relations in the wholesale prices, unleaded gasoline 95 (Two
Least Squares)
Type of vertical relationship
Quantity sell
Refined price
No observations
Hansen J Statistic
F-Statistic
0.5 miles
1 mile
2 miles
21.396*
22.608
24.816
(12.694)
(18.367)
(23.284)
0.002**
0.002*
0.002*
(0.001)
(0.001)
(0.001)
0.926***
0.925***
0.921***
(0.022)
(0.031)
(0.038)
72
72
72
4.856
4.205
4.210
(0.5624)
(0.6489)
(0.6483)
51830.53*** 32951.90*** 30957.60***
(0.0000)
(0.0000)
(0.0000)
Robust Standard Deviation in brackets. P-value: ***(1%), **(5%), *(10%)
Table7c. Effect of the vertical relations in the wholesale prices, unleaded gasoline 97 (Two
Least Squares)
0.5 miles
1 mile
2 miles
Type of vertical relationship
7.659
4.604
3.115
(6.736)
(7.330)
(7.918)
Quantity sell
0.004***
0.004***
0.004***
(0.001)
(0.001)
(0.001)
Refined price
0.957***
0.961***
0.963***
(0.010)
(0.011)
(0.011)
No observations
106
106
106
Hansen J Statistic
12.209*
7.997
6.249
(0.0939)
(0.3329)
(0.5110)
F-Statistic
84510.38*** 81748.74*** 83100.50***
(0.0000)
(0.0000)
(0.0000)
Robust Standard Deviation in brackets. P-value: ***(1%), **(5%), *(10%)
Table 7d. Effect of the vertical relations in the wholesale prices, Diesel gasoline (Two Least
Squares)
Type of vertical relationship
Quantity sell
Refined price
No observations
Hansen J Statistic
F-Statistic
0.5 miles
1 mile
2 miles
14.404**
13.850*
11.921
(6.982)
(7.973)
(8.480)
0.001
0.001
0.001
(0.001)
(0.001)
(0.001)
0.958***
0.959***
0.962***
(0.015)
(0.017)
(0.018)
107
107
107
8.469
8.373
9.222
(0.2931)
(0.3008)
(0.2371)
33018.55*** 30486.40*** 33975.97***
(0.0000)
(0.0000)
(0.0000)
Robust Standard Deviation in brackets. P-value: ***(1%), **(5%), *(10%)
In all the specifications the F-Statistic shows that we can reject that all the variables
that are equal to zero. The Hansen J Statistic shows that the vertical relationship
variable instruments are valid. We must warn you that our instruments are not strong,
and can be biased towards zero in the variable coefficient for the vertical relationship;
this concept is known in economic literature as the attenuation bias.
We can see in the previous tables that disintegration positively affects the wholesale
price, although it is not very statistically different from zero. The attenuation bias
mentioned in the previous paragraph can help this result. The results show that the
principal variable that affects wholesale price is the refinery price, and the value of the
coefficient is very close to one; so changes in the refinery price are passed on to the
wholesale price. The last empirical implication that the results show is the positive
relationship between the quantity sold by the petrol station and the wholesale price.
This variable has the opposite sign to what we had expected, and shows that the
stations with a greater volume of sales do not receive reductions in wholesale prices.
One possible explanation for this is price discrimination. The firm knows which petrol
stations have larger volumes of sales and consumers with a high disposition to pay
high prices, and subsequently apply higher prices to these petrol stations. So, for
wholesale prices we demonstrate a positive relationship between disintegration and
wholesale prices, although this relationship is not strong. Probably what affects the
results are that the integrated vertical structure of this wholesale price is only
transferred within the same company affects these results.
To arrive at the effect of the different vertical relationships on the final retail price, we
propose the following equation:
retailpricei   0   1wholesalepricei   2trvi   3quantityi   4nrivalsi 
4
 5nownbrandi   j  brandi   i
j 1
tvri   0  1washmachi   2 pharmacyi   3 lub ricationi   4 food i 
 5toiletsi   6cashmachi  i
25
quantityi   0   1vehiclesh   2 populationh   h  regioni
h 1
The retail price of the petrol station i depends on the wholesale price, the type of
vertical relationship, the quantity sold, the number of rivals in the predefined market,
the number of own brand outlets in the same markets, and the dummies variables that
bring together the brand effect on prices. The instruments for this type of vertical
relationship are dummies variables of the different services offered by the petrol
station, number of vehicles and population of the region.
Table 8a. Effect of the vertical relations in the retail prices, unleaded gasoline 93.
Type of vertical
relation
Quantity sold
Wholesale
price
Nº of rivals
Nº own brand
outlets
Esso
Shell
YPF
Constant
Nº obs.
Centered R2
Hansen J
Statistic
F(8,100)
0.5 miles
1 mile
2 miles
10.846***
9.708*** 10.324***
(3.309)
(3.023)
(3.094)
-0.003***
-0.003***
-0.003***
(0.001)
(0.001)
(0.001)
0.198***
0.203***
0.202***
(0.070)
(0.071)
(0.069)
-0.513*
-0.514***
-0.163***
(0.273)
(0.126)
(0.041)
-0.069
0.203
0.161
(0.589)
(0.388)
(0.115)
-14.382*** -14.530*** -15.854***
(4.446)
(4.004)
(4.507)
-13.276*** -12.429*** -13.275***
(2.833)
(2.696)
(2.696)
-12.254*** -11.181*** -11.550***
(3.160)
(2.918)
(2.885)
494.679*** 494.172*** 494.838***
(40.356)
(40.381)
(39.298)
109
109
109
0.20
0.25
0.25
44.658
46.550
47.375
(0.1810)
(0.1350)
(0.1181)
4.62***
6.13***
6.00***
(0.0001)
(0.0000)
(0.0000)
Robust Standard Deviation in brackets. P-value: ***(1%), **(5%), *(10%)
Table 8b. Effect of the vertical relations in the retail prices, unleaded gasoline 95.
Type of vertical
relation
Quantity sold
Wholesale price
Nº of rivals
Nº own brand
outlets
Esso
Shell
YPF
Constant
Nº obs.
Centered R2
Hansen J
Statistic
F(8,63)
0.5 miles
1 mile
2 miles
42.298*
38.809**
47.712**
(22.574)
(18.458)
(21.428)
-0.002*
-0.001*
-0.002**
(0.001)
(0.001)
(0.001)
0.323**
0.307**
0.337**
(0.160)
(0.153)
(0.145)
-0.262
-0.710**
-0.256***
(0.404)
(0.305)
(0.082)
0.486
0.281
0.462
(1.630)
(1.140)
(0.318)
-29.155
-26.672
-37.417*
(21.579)
(17.994)
(21.511)
-30.048
-25.864
-35.133*
(20.840)
(16.973)
(20.071)
-27.039
-23.175
-30.685
(20.879)
(16.706)
(19.856)
415.403*** 426.712*** 410.407***
(95.253)
(90.442)
(86.167)
72
72
72
0.23
0.31
0.26
35.462
37.263
35.156
(0.3992)
(0.3213)
(0.4132)
1.38
2.18**
2.58**
(0.2233)
(0.0411)
(0.0168)
Robust Standard Deviation in brackets. P-value: ***(1%), **(5%), *(10%)
Table 8c. Effect of the vertical relations in the retail prices, unleaded gasoline 97.
Type of vertical
relation
Quantity sold
Wholesale price
Nº of rivals
Nº own brand
outlets
Esso
Shell
YPF
Constant
Nº obs.
Centered R2
Hansen J Statistic
F(8,97)
0.5 miles
1 mile
2 miles
9.933**
9.133**
9.363**
(4.778)
(4.530)
(4.659)
-0.003***
-0.003***
-0.003***
(0.001)
(0.001)
(0.001)
0.305***
0.299***
0.294***
(0.112)
(0.113)
(0.110)
-0.845**
-0.529***
-0.166***
(0.351)
(0.162)
(0.048)
0.297
-0.208
0.057
(0.770)
(0.507)
(0.166)
-10.007*
-9.739*
-10.511*
(6.003)
(5.629)
(6.535)
-11.314***
-10.854**
-11.186**
(4.400)
(4.374)
(4.598)
-8.800*
-8.336*
-8.218*
(4.948)
(4.683)
(4.836)
443.571*** 449.156*** 452.106***
(66.876)
(67.281)
(65.022)
106
106
106
0.15
0.19
0.20
42.632
40.064
42.546
(0.2417)
(0.3358)
(0.2445)
2.60**
3.38***
3.92***
(0.0129)
(0.0018)
(0.0005)
Robust Standard Deviation in brackets. P-value: ***(1%), **(5%), *(10%)
Table 8d. Effect of the vertical relations in the retail prices, Diesel gasoline.
Type of vertical
relation
Quantity sold
Wholesale price
Nº of rivals
Nº own brand
outlets
Esso
Shell
YPF
Constant
Nº obs.
Centered R2
Hansen J
Statistic
F(8,98)
0.5 miles
1 mile
2 miles
10.218*** 10.317*** 10.389***
(3.153)
(2.872)
(3.138)
-0.002***
-0.002***
-0.002***
(0.001)
(0.001)
(0.001)
0.186*
0.181*
0.190*
(0.108)
(0.108)
(0.106)
-0.489*
-0.396***
-0.106**
(0.303)
(0.151)
(0.053)
-0.889
-0.728
-0.116
(0.863)
(0.551)
(0.160)
-6.251*
-6.980**
-6.960*
(3.797)
(3.346)
(4.283)
-7.127**
-7.299**
-7.306**
(2.931)
(2.907)
(3.004)
-6.741**
-7.265**
-7.087**
(3.228)
(3.049)
(3.193)
374.394*** 379.420*** 374.612***
(48.460)
(47.711)
(47.185)
107
107
107
0.23
0.26
0.25
40.144
37.923
42.367
(0.3326)
(0.4271)
(0.2506)
4.92***
5.51***
4.94***
(0.0000)
(0.0000)
(0.0000)
Robust Standard Deviation in brackets. P-value: ***(1%), **(5%), *(10%)
Like the wholesale price specifications for retail prices the F Statistic shows that we can
reject that all variables are equal to zero. The Hansen J Statistic shows that our
instruments are valid, but that they present the same attenuation bias as with the
wholesale price. So we have to take into account that the coefficient of the type of
vertical relationship can be biased toward zero.
In all cases the empirical results show a positive and significant relationship between
disintegration and the retail prices. The magnitude of this coefficient is between 1.6
and 2 per cent, or around 10 pesos, except for unleaded petrol 95 which is 7 per cent.
This is the effect of this type of vertical relationship, at least for the retail price; if we
consider the attenuation bias, this effect may be greater. This result is in line with
those of Barron and Umbeck (1984), Shepard (1993), Blass and Carlton (1999) and Vita
(2000).
The other variables in the specification present the expected sign, and are statistically
significant in the majority of the cases. The quantity sold presents a negative
coefficient indicating the possibility of small economies of scale in petrol retailing.
Similar results are shown by Jiménez and Perdiguero (2010) for the Canary Islands
petrol market. The wholesale price is positive and statistically significant as expected.
The number of rivals in the different predefined markets has a negative effect on the
retail price. This variable reaches the level of competition in the market, so we would
expect a negative relationship between competition and price levels. Similarly, the
number of own brand outlets is a proxy for market concentration, and increases the
possibility of market power; we would expect a positive relationship between number
of own brand outlets and the level of prices.
Finally, the brand dummies variables show that all the brands fix their prices below
those of COPEC, the biggest company in the market. The company with the highest
market share fixes the highest retail price, so we can deduce that imperfect
competition exists in the Chilean petrol market.
To summarize the empirical results obtained by the different specifications we can say
that:
- The disintegration type of vertical relationship is more probable in markets
with high competition and less high level service investment. On the other
hand, integration is more likely for petrol stations with high investment services
and own brand outlets concentration.
- The disintegration type of vertical relationship has a positive effect on the
wholesale price and on the retail price, although we take into account the
attenuation bias into our estimations.
6. Conclusions
The analysis of the vertical relationships within industries has been the focus of
attention for the industrial organization for a long time. Using the hypothesis derived
for the transaction-cost theory, and the principal-agent problem, we have tried to
empirically test the Chilean petrol market. Our main hypothesis is that local
competition plays an important role in the choice of vertical relationship. First,
because it eliminates or reduces the problem of double marginalization, which is one
of the main principal-agent problems. Second, because it reduces the control costs
both for the level of effort and services that the retailer offers, and that can be below
the efficient firm level. So, local competition has a positive effect on the disintegration
decisions. Other empirical hypotheses like the reduction of cost control of employees
combined with the concentration of the outlets and the effect of different services are
tested too.
The empirical results show that the local competition positively affects the
disintegration decision. In the same way the results confirm the rest of the derived
hypothesis; i.e. the concentration of own brand outlets increase the probability of
outlet integration, heavy investment services mean that the firms decide to integrate,
and conversely low level investment services increase the likelihood of disintegration.
However, we did not only check the variables that affect vertical relationship decisions,
we also checked the effect of different vertical relationships on the wholesale and
retail price. Both in the wholesale and in the retail market we observe a positive
relationship between the disintegrated vertical structure and prices. This result was
also obtained using different approaches by authors such as Barron and Umbeck
(1984) and Shepard (1993). The increase in the retail prices for the different vertical
structure is between 1.6 and 7 per cent. So, local competition is an imperfect
instrument for eliminating the transaction cost and the principal-agent problems.
These results suggest that firms choose the vertical relationships as a logical decision,
in order to have the most efficient vertical structure that minimizes the transaction
cost and principal-agent problems. Although local competition is not a perfect
instrument, the divorcement policies applied in the past in the United States introduce
more inefficiency into vertical structures and increase the final price, as is shown by
Blass and Carlton (1999) and Vita (2000).
Bibliography
- Anderson, E. (1985) “The salesperson as an outside agent or employee: A transaction
cost analysis” Marketing Science, Vol. 4(3), pp. 234-254.
- Anderson, E., and Schmittlein, D. C. (1984) “Integration of the sales force: An
empirical examination” The RAND Journal of Economics, Vol. 15(3), pp. 385-395.
- Baker, G. P., and Hubbard, T. N. (2003) “Make versus buy in trucking: Asset
ownership, job design, and information” The American Economic Review, Vol. 93(3),
pp. 551-572.
- Baker, G. P., and Hubbard, T. N. (2004) “Contractibility and asset ownership: Onboard computers and governance in U.S. trucking” The Quarterly Journal of Economics,
Vol. 119(4), pp. 1443-1479.
- Barron, J.M., and Umbeck, J.R. (1984) “The effects of different contractual
arrangements: the case of retail gasoline market” Journal of Law and Economics, Vol.
27, pp. 313-328.
- Blass, A.A., and Carlton, D.W. (1999) “The choice of organizational form in gasoline
retailing and the costs of laws limiting that choice” NBER Working Paper, Nº 7435.
- Brickley, J. A., and Dark, F. H. (1987) “The choice of organizational form: The case of
franchising” Journal of Financial Economics, Vol. 18, pp. 401-420.
- Brickley, J. A., Dark, F. H., and Weisbach, M. S. (1991) “An agency perspective on
franchising” Financial Management, Vol. 20(1), pp. 27-35.
- Carney, M., and Gedajlovic, E. (1991) “Vertical integration in franchise systems:
Agency theory and resources explanations” Strategic Management Journal, Vol. 12(8),
pp. 607-629.
- Caves, R. A., and Murphy, W. F. (1976) “Franchising: firms, markets and intangibles
assets” Southern Economic Journal, Vol. 42, pp. 572-586.
- Goméz-Lobo, A., and Córdova, F. (2006) “Los precios de los combustibles líquidos a lo
largo del país: posibles explicaciones” in Arellano, S., and Valdés, S. “La industria de los
combustibles líquidos. El caso de Chile” pp. 203-247. Ed. Univerdidad Católica de Chile
(Santiago de Chile).
- Jiménez, J.L., and Perdiguero, J. (2010) “Regional finance and competition policy: The
Canary Islands petrol market” Applied Economics, Vol. 42(10), pp. 1245-1255.
- John, G., and Weitz, B. A. (1988) “Forward integration into distribution: An empirical
test of transaction cost analysis” Journal of Law, Economics, and Organization, Vol.
4(2), pp. 337-355.
- Katz, M. L. (1989) Vertical contractual relations, in R. Schmalensee and R. D. Willig,
eds., Handbook of Industrial Organization. New York: North-Holland.
- Kehoe, M. R. (1996) “Franchising, agency problems, and the cost of capital” Applied
Economics, Vol. 28, pp. 1485-1493.
- Lafontaine, F., and Slade, M. (1996) “Retail contracting and costly monitoring: Theory
and evidence” European Economic Review, Vol. 40, pp. 923-932.
- Lafontaine, F., and Slade, M. (2007) “Vertical integration and firm boundaries: The
evidence” Journal of Economic Literature, Vol. 45, pp. 629-685.
- Levy, D. T. (1985) “The transaction cost approach to vertical integration: an empirical
examination” The Review of Economics and Statistics, Vol. 76(3), pp. 438-445.
- Minkler, A. P. (1990) “An empirical analysis of a firm’s decision to franchise”
Economics Letters, Vol. 34, pp. 77-82.
- Rubin, P. H. (1978) “The theory of the firm and the structure of the franchise
contract” The Journal of Law and Economics, Vol. 21(1). pp. 223-233.
- Shepard, A. (1993) “Contractual form, retail price, and asset characteristics in
gasoline retailing” The RAND Journal of Economics, Vol. 24(1), pp. 58-77.
- Sklivas, D. S. (1987) “The strategic choice of managerial incentives” The RAND Journal
of Economics, Vol. 18(3), pp. 452-258.
- Vita, M.G. (2000) “Regulatory restrictions on vertical integration and control: the
competitive impact of gasoline divorcement policies” Journal of Regulatory Economics,
Vol. 18(3), pp. 217- 233.
- Williamson, O. E. (1971) “The vertical integration of production: market failure
considerations” The American Economic Review, Vol. 61(2), pp. 112-123.
- Williamson, O. E. (1979) “Transaction-cost economics: The governance of contractual
relations” The Journal of Law and Economics, Vol. 22(2), pp. 233-261.
- Williamson, O. E. (1981) “The economics of organization: The transaction cost
approach” The American Journal of Sociology, Vol. 87(3), pp. 548-577.