The influence of market competition on ownership form
Academic Year 2016/2017
Master Thesis
Master of Science in Business Economics
Specialization Managerial Economics and Strategy
Faculty of Economics and Business
University of Amsterdam
Student: Roderick Vroom
Student number: 0113492
Date: December, 2016
First supervisor: prof. dr. R. Sloof Second supervisor: dr. J. van de Ven Number of ECTS: 15
INDEX
Abstract ... 1
Introduction ... 2
Theory development ... 2
Factors influencing ownership form ... 3
Factors in favor of franchising ... 3
Resource scarcity ... 3
Knowledge ... 4
Agency problems ... 4
Factors in favor of company owned units ... 5
Free riding ... 5
Hold-up problem ... 6
High powered incentives ... 7
Monitoring ... 7 Rivalry ... 9 Hypothesis ... 11 Methods ... 12 Data ... 12 Variables ... 12 Dependent variable ... 12 Independent variables ... 13 Control variables ... 15 Estimation ... 17
Remarks about the data ... 18
Sample ... 19
Results ... 19
Descriptive statistics ... 19
Regressions ... 27
Regressions with the segmented intensity of rivalry ... 32
Robustness checks ... 33
Conclusion ... 35
Discussion ... 35
Abstract
There is a large amount of literature considering the tradeoffs between franchising and
company ownership as two alternative ownership arrangements. The literature has identified a series of factors that favor one form over the other. This literature considers arguments such as financial constraints, agency problems, knowledge advantages, and free riding. However, the prior literature did not consider the effect of rivalry on the tradeoff between these two ownership forms. In this paper, I argue that the intensity of rivalry in a local market affects this tradeoff. More precisely, I argue that the intensity of rivalry will serve as a natural monitoring device, reducing the need for costly monitoring by the firm. As monitoring costs are a factor limiting the benefits of company-ownership, a reduction in monitoring costs will favor the use of company ownership relative to franchising. Using data from the Texas hotel industry, I provide evidence consistent with the notion that a high intensity of rivalry leads to an increase in the proportion of company ownership over franchising.
Introduction
Many firms with chains, e.g., hotels and fast food chains, have mixed ownership forms. That is, some establishments are franchised and some are company owned. In the franchise ownership form, the firm, i.e., the franchisor, enters into a contract with the franchisee to manage an individual unit of the firm. The franchisee becomes the residual claimant after paying a franchise fee to the franchisor. This franchise ownership form diminishes the
principal-agent problem, which is an important subject in the economic literature, as it makes the franchisee residual claimant. Alternatively, the firm could also decide to hire a
professional manager who will run the establishment for the firm. In this case, the ownership form is company owned. The reasons why a firm decides for franchise or company ownership can be influenced by different factors: capital constraints, risk, incentives, market
environment, complexity of the establishment, and many others.
While the prior literature examines these factors, for example the influence of the distance of an establishment to their headquarters (Minkler, 1990), this paper focuses on a specific market environment factor influencing the ownership form decision: the intensity of rivalry. More specifically, this paper concentrates on how rivalry in a local market, through the cost of monitoring, affects the ownership form. The current global trend that more establishments are part of chains, and thus firms have to make the decision between franchising or keeping the unit company owned, makes this an interesting topic.
This paper will use a dataset of hotels in Texas, USA, to investigate the relationship between rivalry and ownership form. First, this paper describes how the intensity of rivalry reduces the cost of monitoring, and how the cost of monitoring influences ownership form. From this theory, this paper presents a framework and derives a hypothesis. The next chapter describes the research method to test the hypothesis, followed by the empirical results. At the end the paper presents a discussion.
Theory development
This chapter describes how rivalry influences the decision of a company to franchise a unit or keep a unit company owned, i.e., how rivalry influences the decision of ownership form. First, I explain the factors influencing the choice of ownership form. Then I describe in more detail
the role of monitoring in the choice of ownership form. This is followed by an explanation of how rivalry can work as a monitoring device. At the end of this chapter, I will formulate a hypothesis based on the previous theory described.
Factors influencing ownership form
The decision by a company to franchise a unit or to keep the unit company owned has been extensively studied in the economic literature. This paragraph describes factors influencing the tradeoff in the management decision to franchise a unit or keep it under company ownership. All factors favor the tradeoff in a certain direction, for example into franchising, but all factors also have characteristics which pushes the tradeoff in the other direction. This section starts with the factors which in the tradeoff tilt toward franchising.
Factors in favor of franchising Resource scarcity
The traditional franchise literature emphasizes that the decision to franchise a unit is often driven by financial constraints (Oxenfeldt and Kelly, 1969; Oxenfeldt and Thompson, 1969). A financial constraint of a firm to open a new unit could be alleviated by getting a franchisee to invest in his or her own unit. Especially new companies who do not have the funds to start new units will choose to franchise. Further in the lifecycle of the firm, Oxenfeldt and Kelly expect firms to repurchase units from franchisees1 to focus more on “operating efficiencies
and market development” (1969, p. 74). In other words, when a new firm wants to grow fast, the ownership tradeoff tilts in favor of franchising to expand rapidly. When the firm matures, it wants more control over the units and therefore the tradeoff tilts in favor of company owned units.
However, the investment by a franchisee in a unit implies a significant undiversified risk for the franchisee. As a consequence, to be compensated for this risk, the franchisee needs to be remunerated and will require a higher rate of return (Rubin, 1978). These two opposing factors imply that there is no simple, general optimum between the different ownership forms
1 Franchisee is the person who manages a unit. Franchisor is the firm which allows a franchisee to run a unit of
and that different ratios of company-ownership versus franchising may be observed depending on the specifics of a situation (Brickley and Dark, 1987).
Knowledge
An alternative reason to franchise a unit could be to attract a franchisee who has profound knowledge of a local market in areas where this local knowledge is essential (Minkler, 1990). Minkler states that according to the search cost theory, it can be very costly for a firm to acquire knowledge of a local market. So instead of acquiring this costly knowledge, it could be cheaper to enter into a franchise contract with a franchisee who already possesses this knowledge about the local market. In this case, the tradeoff of ownership form would lean more towards franchising than towards company ownership.
Agency problems
When a firm hires a manager to perform certain tasks for the firm, agency problems may arise. The manager, i.e., the agent, may in the case of a conflict of interest act in self-interest and not in the best interest of the firm, i.e., the principal. There are many solutions to the principal agent problem, such as monitoring, pay for performance, identification with a
corporate culture, etc. None of the solutions is perfect, each has its benefits and its drawbacks, but they all try to align the incentives between the principal and the agent. Franchising is one of the solutions to the principal-agent problem.
An important difference between a manager of a company owned unit and a franchisee is a different payoff structure, which results in different incentives. The manager of a company owned unit receives a fixed salary, and maybe a bonus depending on the performance of the unit, while the franchisee is the residual claimant. The residual claimant gets all that is left over after paying costs and royalties. The effect is that the manager of a company owned unit has low-powered incentives, in contrast to the franchisee who has high-powered incentives. As a consequence of this compensation scheme, the professional manager2 has the incentive
to shirk. The solution to aligning the incentives between the firm and the professional manager and to provide strong incentives to the professional manager therefore would be to
2 In this paper, the term manager of a company owned unit and a professional manager are used interchangeably.
franchise the unit, and make the manager residual claimant. In summary, franchising can solve the shirking problem, which arises from the principal-agent relation.
Shirking
As explained above, the manager of a company owned unit has the incentive to shirk, i.e., to work less hard, because of the payoff structure. Since the professional manager will receive, at least partly, a fixed salary, the manager does not decrease his monetary benefit, at least in the short term, when he does not provide effort. Besides shirking behavior, the professional manager could enter in perquisite taking behavior, for example, by consuming expensive dinners at cost of the firm, or by purchasing luxurious office furniture or deluxe automobiles. The professional manager is tempted in the direction of this behavior because he will get all the benefits of his behavior, e.g., a deluxe car, but he does not bear the full cost: the firm does. A solution to shirking/perquisite taking behavior could be to franchise a unit: the manager becomes residual claimant, and therefore has less incentive to shirk, because he bears the full cost of his behavior. Consequently, the stronger the agency problems, the higher proportion of franchising relative to company ownership is expected. Another solution to the shirking problem is monitoring, which is described in more detail in the next section of this paper.
Factors in favor of company owned units
Apart from arguments linked with resource scarcity, knowledge and agency problems, the franchise literature has identified other factors, specific to the franchising context, that
influence the tradeoffs between franchising and company ownership. The main characteristics of the franchisee which influences these factors are:
Residual claimant, and therefore high powered incentives.
No hierarchical relationship with the firm, and therefore difficult adjusting behavior of the franchisee by the firm.
The high powered incentives and lack of hierarchical relationship of the franchisee can result in problems such as free riding, hold-up problems, and inefficient risk bearing (Brickley and Dark, 1986).
Free riding
Free riding can be defined as enjoying certain benefits without contributing any effort or only a small amount of effort. When a franchisee tries to free ride on the trademark, for example by
opening a McDonald’s franchise but serving food that does not adhere to the normal McDonald’s quality, this bad behavior comes with a cost for the company as a whole, not only the franchised unit itself (Brickley, Dark and Weisbach, 1991). The cost of serving inferior food will be borne by other units and by the franchisor whose brand is less valuable in the future. This behavior is more prevalent when they are less returning customers to a unit, for example for a fast-food restaurant at the highway (Klein and Leffler, 1981; Hooiveld, 2005). Apart from the franchisee, also the franchisor can exhibit free riding behavior. The franchisor, who is often responsible for national advertising, could reduce advertising to have less expenses. Solutions to free riding behavior of the franchisor could be to have the
franchisor own several units, receive a percentage of the revenue of franchised units, or when the franchisor intends to start future units (Brickley and Dark, 1986, p. 406). To avoid free riding behavior a firm could decide to favor company ownership because within this form the firm can use the hierarchical relation with the professional manager to avoid free riding behavior.
Hold-up problem
When there are relation-specific assets, such as a building which can only be used within the boundaries of the franchise contract, both parties can show opportunistic behavior. The idea behind the relation-specific assets is that after entering the contract, i.e., after initial
investments are made, one of the parties involved can try to renegotiate the initial contract. This is also known as the hold-up problem. An example of a franchisor who enters in post-contractual opportunistic behavior is when the franchisee has invested in a building, which can only be used within the franchise contract, and tries to renegotiate the franchise fees paid to the franchisor and tries to get a higher revenue from the franchisee. An example of such a building could be a building used for indoor wall climbing. These building are so specific that using them more generally, e.g., as an office, would require significant investments. Also the franchisee can enter in post-contractual opportunistic behavior: If the indoor wall climbing building would be owned by the franchisor, the franchisee could try to lower the lease since the building cannot be used easily for other purposes. Because a professional manager has less opportunity due to the hierarchic relation to enter in post-contractual opportunistic behavior a firm would favor company ownership to avoid the hold-up problem.
High powered incentives
Generally high-powered incentives are seen as positive, because they induce greater
efficiency and effort (Larkin, 2007; Frant, 1996). For example, people work harder to get the job finished. However, high powered incentives could intensify the franchise problems mentioned above: hold-up and free riding. The stronger the incentive, the more inclined a manager will be to engage in bad behavior, e.g., dishonest behavior (Williamson, 1985).
These specific franchise related problems mentioned above, i.e., hold-up problem, free riding and high powered incentives, will influence the ownership form in the direction of company ownership, because these problems are less severe under the company ownership form.
In summary, agency theory and franchise theory offer explanations of how certain factors influence the choice of ownership. Some factors tend towards franchising, for example shirking. Other factors influence the tradeoff in favor of company owned units, like free riding and the hold-up problem. The resource scarcity theory could work in both ways. Interestingly, there are also other solutions to the agency problem, like monitoring, which is the most studied solution to agency problems. The next paragraph shows that previous empirical research confirms that different proxies for monitoring costs, such as distance to headquarters, influence ownership form, i.e., lower monitoring cost lead to more company owned units.
Monitoring
Monitoring is the systematic process of observing, tracking and gathering data or activities with the purpose of measuring targets and goals. The data gathered is, according to
Investorwords.com3, used to analyze and evaluate a unit to assess its effectiveness and where
needed to adjust its inputs.
Firms engage in different kind of monitoring, for example employee monitoring, business monitoring, financial monitoring, etc. The main reason, according to Zimmerman (2009), is to make sure that plans are implemented as planned. Typically, the management of a firm will decide to monitor certain data which will result in a periodical report. This report will be used
to change the strategy of a firm, determine the bonus payments to employees, determine investment opportunities, etc. In summary, monitoring leads to information which will be used for future decisions by the management.
Monitoring should lead to better behavior, but sometimes the effects are negative. The result can be negative when for example a manager has multiple tasks, but only one task is
monitored. Because of the monitoring system the manager will put more effort in the
monitored task and not in the other, which maybe is not in the best interest of the firm (see for example Milgrom and Roberts (1992) where the Equal Compensation Principle is linked to multiple tasks).
Milgrom and Roberts (1992) also discuss the Monitoring Intensity Principal, which suggests that more resources should be spent on measurement when the agent’s piece rate is higher, making careful performance measurement and piece rate complementary. The result of the Monitoring Intensity Principal is that if a company decides to pay a manager more on
performance it also has to improve the measurement of performance, and therefore increases monitoring costs.
To mitigate shirking behavior, besides franchising as an ownership form, monitoring is often used. Monitoring influences shirking behavior, and consequently the choice of ownership form. If there is less shirking behavior, there is less need to franchise a unit. Monitoring is, as a result of this behavior, extensively studied in the franchise literature (Brickley and Dark 1987; Carney and Gedajlovic 1991; Lafontaine 1992; Kehoe 1996).
Monitoring as solution to the shirking problem
When an agent can be monitored, he can be paid and punished to make his incentives stronger and aligned with the firm. Therefore, there is a direct link between the cost of monitoring and the severity of agency problems. Since the severity of agency problems influences the choice of ownership form, the cost of monitoring also influences the ownership form decision made by a firm. Brickley and Dark (1987) investigate the distance of a unit to the headquarters of the firm, which is a proxy for the cost of monitoring. They argue that the cost of monitoring increases with the distance between a unit and the headquarter. Minkler (1990), Brickley, Dark and Weisbach (1991), and Carney and Gedajlovic (1991) investigate the outlet density as a determinant of monitoring costs. Carney and Gedajlovic argue that if units are more
dispersed, monitoring costs will be higher, due to an increase in travelling cost. Baker and Hubbard (2003) find that the adoption of on-board computers in a trucking company lead to lower monitoring costs. Also the return rate of customers, as described by Klein and Leffner (1981), has an influence on monitoring costs: when the return rate of customers is high, there is less need to monitoring, since the returning customers rate the unit. If the customer would not like the unit, he or she would not return. Therefore, there is a negative relationship between return rate and monitoring costs.
All studies above confirm empirically the theory that lower monitoring costs mitigate shirking behavior. Less agency problems, through monitoring, influences the choice of ownership form, since high monitoring costs would be one of the main factors why management would choose for franchising.
Of course monitoring is also applied in the relationship of franchise contracts. The franchisor would for example like to know if the franchisee serves the coffee in the right temperature. However, the amount of monitoring is less than for a manager under the principal-agent problem, because with the franchisee the right incentives are already there. For that reason, the monitoring cost will be lower for a franchised unit than for a company owned unit.
Rivalry
Interestingly, market conditions have not been extensively studied as an influencing factor for the management decision to franchise or to keep a unit company owned. Particularly rivalry, as a market condition, has not been studied.
Competitive rivals are units or organizations with comparable products or services who target the same customer group (Scholes and Johnson, 2002). Porter (1980) states that the intensity of rivalry within an industry is a main determinant of the competitive structure of that industry. When the intensity of rivalry is strong, the different rivals will compete heavily on price and therefore diminish the possibility of profit. This is opposite to a market where the intensity of rivalry is weak. In this market the potential to make a profit is larger.
Markets with high intensity of rivalry are characterized by a large number of suppliers, a slow market growth rate, a low number of customers per rival, low brand loyalty, when suppliers
have an equal size, when fixed costs are high, when the capacity is larger than demand, and when the cost of switching supplier is low (Scholes and Johnson, 2002).
In the absence of monitoring, as described above, the agent may be tempted to shirk. Shirking will lead to lower profitability. In an environment of munificence, this will only slightly reduce profitability, perhaps even without being noticed by the firm’s headquarters. In
contrast, in an environment of high intensity of rivalry, profits may significantly decrease and the unit’s mere existence may be challenged. In other words, the high intensity of rivalry will make non-optimal behavior by the professional manager more visible. As a result, this will lead to a reduced incentive to shirk for the professional manager. Hence, the market acts as a natural monitoring device, i.e., both monitoring by a firm and high intensity of rivalry have the same outcome, a decrease in shirking behavior by the professional manager.
Because monitoring influences the ownership form, the logically leads to the following research question:
How does the intensity of rivalry influence the choice of ownership form?
Although I argue above that the result on behavior is similar, i.e., less shirking behavior, there are big differences between monitoring by the firm and the intensity of rivalry as natural monitoring device. Where the firm incurs cost of implementing a monitoring device, the market does it for free, there are no direct costs for the firm4. The term natural is introduced
because the monitoring happens at zero cost for the firm. Another difference between
monitoring by the firm and natural monitoring by the market is the timing. Monitoring by the firm will provide information ex ante. The management can adjust the strategy of the firm to for example maximize profit or avoid bankruptcy. By the time the natural monitoring device gives information, i.e., punishes the unit to go bankrupt, it could be too late to adjust. Perhaps the natural monitoring device is more like a possible punishment, which should give the manager the right incentives.
Of course the two different forms of monitoring can co-exist. Besides the natural monitoring by the intensity of rivalry, the firm can observe sales data of the unit and make surprise visits
to check opening hours. This paper argues that when there is natural monitoring, there is less need for firm monitoring, because they act as substitutes. Both monitoring devices influence the shirking behavior of the professional manager. As a consequence, the monitoring cost of a unit within a high intensity of rivalry market will be lower than the monitoring cost of a unit with low intensity of rivalry. In summary, with lower cost of monitoring, there are less agency problems and therefore I expect the proportion of company owned units to be larger
In conclusion, I argue in this paper that rivalry can serve as a natural monitoring device for companies, which will lead to more company owned units as ownership form. In other words, when companies can benefit from a natural monitoring device, e.g., rivalry, instead of
investing effort and resources in building a monitoring system, firms will use this device and choose more often for company owned units.
Hypothesis
While the prior literature, as explained above, has studied the tradeoff between franchising and ownership from different perspectives, no prior study has explored how rivalry may influence the choice between franchising and company ownership. This paper wants to add rivalry in a local market to the discussion about factors influencing ownership form. Figure 1 shows the effect of the intensity of rivalry on monitoring. High intensity of rivalry leads to lower monitoring cost for the firm, and lower monitoring cost lead to more company ownership.
Figure 1: Effect of rivalry through monitoring cost on ownership form
From Figure 1 the following hypothesis can be derived:
H1: The intensity of rivalry is positively related with the proportion of company-ownership relative to franchising.
In the next chapter this paper describes the data used to provide empirical evidence for the hypothesis.
Methods
Data
The data used in this study is from the hotel industry in Texas (US). The hotel industry competes in a local market where customers perceive higher value through for example quality segments and brands. The variable zip code is used as one of the principal factors to determine the boundary of the local market. The data set is combined from two different sources: Source Strategy Inc., a consultancy firm, and the State of Texas Comptroller’s Office. Because the data contains ownership form, location and segment, it is very suitable to investigate the influence of rivalry, i.e., competition, on ownership form. The data of the State of Texas Comptroller’s office is used before by Chung and Kalnins (2001) who investigate the effects of externalities because of physical proximity, such as higher demand or
production efficiencies. The data from Source strategy is used by McCann and Vroom (2014) who use the data to explore how objectives of a firm, especially nonfinancial objectives, influence competitive actions such as entry, exit and pricing decisions.
The main characteristics of the data are:
- The data is panel data, also called cross-sectional time series data, because for each hotel there are up to 34 quarterly observations.
- Over 115,000 observations from more than 4,500 hotels in 900 zip codes in a period from 1997 until 2005 of 34 quarters (unbalanced).
- The data includes, but is not limited to, the brand of the hotel, the quality of the hotel (6-scale ranking from Smith Travel Research Market Price segments with the values Luxury, Upper Upscale, Upscale, Upper Midscale, Midscale and Economy), the capacity of the hotel (number of beds) and the ownership form (company owned, franchised, membership chain and independent).
Variables
Dependent variable
As stated in the research question, the main interest of this paper is how the intensity of rivalry influence the choice of ownership form. Therefore, ownership form is the dependent variable of interest.
Ownership form
Four different ownership forms are registered in the dataset: independent owners, franchisers, company owned and membership chain. Since we are interested in the choice between
franchising a unit or keeping it company owned, independent owners are excluded from the regressions (approx. 54,000 observations), because independent owners are never franchised or company owned. Note however that these observations are included when calculating certain variables such as market concentration. Franchise and membership chain have in common that they are both the residual claimant. While a member of a membership chain pays a fixed annual fee to the franchisor, a franchisee pays a percentage of the revenue to the franchisor. Given that both categories are equal for the purposes of my theoretical
development, I treat them also equally in the regressions. (For simplicity, when I refer to franchising, I mean either pure franchising or membership chain.)
To measure ownership form, i.e., franchised or company owned, I use the dummy variable company owned (CO). Company owned (CO) is set to 1 if the hotel is company owned, and therefore run by a professional manager who is the agent in a principal-agent relationship, and 0 if the hotel is franchised or part of a membership chain. In the dataset, the ratio franchised and company owned is approximately 3 to 1. Because CO is a binary variable, I will use logit regressions (Stock and Watson, 2011), and since the ownership forms company owned (CO) and franchise (FR) are complementary, I will use company owned (CO) as dependent variable.
Independent variables
From the theory formed in the chapter above, the intensity of rivalry should influence the choice of a company to franchise a unit or keep a unit company owned. Therefore, the intensity of rivalry is the main independent variable in the regressions used to provide evidence for the theory.
Intensity of rivalry
To measure the intensity of rivalry in the hotel market, this paper uses the
Herfindahl-Hirschman Index (HHI). According to Pan (2005), the HHI is the most used index in research to measure local rivalry since it includes all firms in the local market, not only the largest. The HHI is calculated as the sum of squared market shares of all hotels in a local market. The
local market, i.e., the market where rivalry takes place, is defined in this paper as all hotels within the same zip code. The dataset also contains information about counties, but with only 226 counties, compared to 919 zip codes, the local market defined per county would on average have the size of 2,664 km2, compared to 556 km2 per zip code5. Because the hotel
industry typically competes with hotels in the same area, I choose to define zip code as the local market. The market share of a hotel is calculated as the number of rooms of that individual hotel divided by the total rooms available in the local market. In contrast to the definition of the dependent variable ownership form, all ownership forms are taken into account to calculate the Herfindahl-Hirschman Index (HHI), also independent owners. That is because it typically does not matter for the local competition what the ownership form of a single hotel is. They all compete in that local market.
Because the Herfindahl-Hirschman Index is an inverse measure of competition, I create a new variable RIVALRY, which is calculated as 1 minus the HHI. The result is that the variable RIVALRY increases, to a maximum of nearly 1, when the rivalry on the local market increases.
Apart from the above-mentioned variable RIVALRY, I also calculate a segmented variable rivalry (RIVALRY_SEGM). Instead of grouping all hotels together in one zip code, I divide hotels in one zip code into two segments: economy and luxury. This segmentation on quality could be important because it affects rivalry in that local market. In other words, although hotels are in the same area, economy hotels might compete more directly with other economy hotels and luxury hotels might compete more directly with other luxury hotels. The
segmented rivalry (RIVALRY_SEGM) is calculated with local markets defined as hotels in the same zip code and in the same segment. Because the intensity of rivalry is not only determined on location but also on segmentation, I expect that the RIVALRY_SEGM will give more accurate empirical findings.
To determine the two segments, economy (LOW) and luxury (HIGH), within a zip code, I use the average daily rate (ADR) of a hotel, which is available in the dataset. The average daily rate is the average room price of a hotel during one period.
As a cutoff point for the low/high segment I choose the 75th percentile of the average daily rate (ADR) in a period. To avoid the influence of inflation or other time factors, the cutoff point is calculated per period. All observations in a period with an ADR below the 75th
percentile belong to the LOW segment, and all observations with an ADR equal or above the 75th percentile belong to the HIGH segment. With this segmentation, I argue that hotel
customers will typically differentiate between the two segments (see Table 4 for descriptive statistics). A customer will choose for example to stay in a hotel in the LOW segment, and determine his choice of hotel within this segment.
Table 1 below shows descriptive statistics for both measures RIVALRY and
RIVALRY_SEGM of the independent variable intensity of rivalry. Both variables range from 0 (monopoly) to close to 1 (perfect competition).
TABLE 1
Descriptive statistics for intensity of rivalry
Rivalry Segmented Rivalry
Variable in dataset RIVALRY RIVALRY_SEGM
Definition of local market zip code zip code per segment (LOW/HIGH)
Number of local markets 27,241 34,905
Average number of hotels on a local market 4.2 3.3 Mean 0.72 0.65 Standard deviation 0.25 0.28 Minimum 0.00 0.00 Maximum 0.97 0.96 Control variables
The following variables are used to control for a potential source of influence on the dependent variable, ownership form:
Size
The capacity of a hotel, i.e., the number of rooms, is used as a proxy for size. It could be that larger hotels require more capital and therefore larger hotels are more likely to be franchised. This would imply a negative relationship between size and the dependent variable company owned. However, it could also be that larger hotels are more complex to manage and thus require being company owned. Another influence could be that a large hotel is regarded as a
flagship hotel and thus also managed by the company itself. Therefore, the influence of size on the dependent variable company owned could be positive or negative.
Royalty fee
When the royalty fee of a franchisee, i.e., the percentage fee a franchisee has to pay to the franchisor, becomes large it becomes less attractive to franchise. Therefore, the royalty fee could influence the ownership form and cause a positive relationship between royalty fee and the dependent variable company owned.
Income
Income is measured as the mean income per capita in a zip code. Income could influence ownership form because in rich areas maybe luxury hotels are built to comply with the surrounding area. An example of such a high-class market could be Nice in France, which would lead to a high upfront investment.6 Similar as the control variable size, the relationship
between income and the dependent variable company owned (CO) could be positive or negative.
Segment
It could be that in the LOW segment, i.e., economy hotels, less capital is required. This could lead to a positive relationship between the low segment and the dependent variable company owned.
Density
As Minkler (1990) describes, the density of outlets has an influence on the ownership form. Minkler states that with a high density of units from the same corporation, travelling cost will be less, resulting in lower monitoring costs, and thus units will be more likely to be company owned. Therefore, a positive relation is expected between density and company owned (CO). Attractiveness
When a local market is very attractive, e.g., when there is less competition, a hotel could decide to profit from this situation and keep the unit company owned. In this case,
competition influences the ownership form through attractiveness and not through monitoring costs. As a proxy for attractiveness I will use the occupancy of hotels in a zip code. I expect that when the attractiveness, i.e., occupancy, of a market increases, the ownership form of company owned becomes more prevalent, resulting in a positive relationship between attractiveness and company owned.
6 Of course a very large economic hotel with 500 rooms requires typically a larger investment than a very
Estimation
To estimate the relationship between the intensity of rivalry and ownership form the
hypothesis H1: The intensity of rivalry is positively related with the proportion of company-ownership relative to franchising can be rewritten for estimations. Using the dependent, independent and control variables stated in the paragraph above, the hypothesis can be estimated as:
(M1)
where:
i = 1,….,4561 represents the individual hotels,
t = 1,….,34 represent the quarterly periods in the sample period, CO = Company owned: Taking the value 1 if ownership form is a
company owned hotel, and 0 if franchised,
RIVALRY = Intensity of rivalry: Calculated as 1 minus the HHI by squaring the market share within a zip code for each firm and then summing the squares,
FEE = Royalty fee to be paid to the franchisor,
SIZE = The hotel’s size, which is the number of available rooms in a hotel, INCOME = Mean income per capita in a zip code,
SEGM = Segment: If segment is LOW it takes the value 1. If segment is HIGH it takes the value 0,
DENSITY = Density of units. Calculated as the total of units belonging to the same corporate headquarters in the state Texas in a period,
ATTRACT = Attractiveness of a market. Calculated as the average occupancy of hotels in a zip code in a period,
For the segmented market, where the intensity of rivalry is calculated per zip code and per segment (LOW/HIGH), the intensity of rivalry can be estimated as:
_
(M2)
where:
RIVALRY_SEGM = Intensity of rivalry: Calculated as 1 minus the HHI by squaring the market share for each unit in zip code in the same segment
(LOW/HIGH) and then summing the squares. For an explanation of the other variables see estimation 1 (M1).
The difference between estimation M2 and estimation M1 is that instead of RIVALRY as independent variable, RIVALRY_SEGM is used as the independent variable.
In both estimations (M1 and M2) the theory presented in this paper expects the coefficient of RIVALRY (or RIVALRY_SEGM) to be positive, i.e., H1: β > 0.
Remarks about the data
The proportion of franchised units to company owned units is almost three to one. This could indicate that the hotel market in Texas is relative new, with many firms still in the beginning of their lifecycle (Oxenfeldt and Kelly, 1969). Another reason could be that the size of Texas, which is, according to Wikipedia pages7, larger than France, makes monitoring more difficult
and expensive and therefore pushing firms into franchising as ownership form (Milgrom and Roberts, 1992).
Sample
Some of the observations are removed from the data for the results presented in the next chapter:
Observations where the average daily rate is 0, the zip code is unknown or the capacity of a hotel is unknown. In these observations, the data for a hotel is not complete (89 observations).
All observations where the owner is an independent owner. Independent owners do not have the choice between the ownership form company owned and franchise (54,232 observations).
Results
This chapter describes the results of the regressions to test the hypothesis under investigation. First, this chapter describes the main features of the dataset used.
Descriptive statistics
As mentioned above, the dataset consist of panel data, i.e., each hotel has multiple
observations over time. In the dataset, each individual hotel (HOTELID) is the panel variable and the quarterly period (PERIOD) is the time variable. An example of the data is presented in Table 2 below:
TABLE 2 Example of dataset
HOTELID PERIOD CO RIVALRY SIZE FEE
3842 1 0 0.63 70 4 3842 2 0 0.63 70 4 3842 3 0 0.63 70 4 3842 4-33a 0 0.63 70 4 3842 34 0 0.63 70 4 4126 1 1 0.56 122 4b 4126 2-5a 1 0.56 122 4b 4126 6 1 0.56 122 4b 4126 7 1 0.59 124 4b 4126 8 1 0.59 124 4b 4126 9 1 0.69 124 4b 4126 10-34a 1 0.69 124 4b
HOTELID = Id of each individual hotel
PERIOD = Quarterly periods in the sample period
CO = Company owned. Value of 1 if hotel is company owned in that period RIVALRY = Intensity of rivalry calculated per period and zip code
SIZE = Capacity of the hotel in that period
FEE = Royalty fee to be paid to franchisor in that period
a 4-33 means that for period 4 till period 33 the results are the same. That is the same CO, RIVALRY, SIZE and
FEE.
b. Of course the company owned unit does not pay a royalty fee. The value here is derived from the brand of the
hotel and only paid if the unit is franchised, and not if the unit is company owned.
For the following reasons, I have decided not to perform a panel regression although the data has clearly the structure of panel data:
1. The dependent variable, CO, must at least be measured twice for each hotel. This causes 531 hotels (out of 4562 hotels) to be dropped from the sample.
2. Changes of the independent variable, RIVALRY, are zero or rather small within the observations of a hotel. Although it is difficult to define when the RIVALRY changes substantially, in the data example above (in Table 2), the changes in RIVALRY for hotel with HOTELID 4126 from 0.59 to 0.69 looks large, but is not with the
RIVALRY in a scale from 0 (monopoly) to close to 1 (perfect competition). Only 4126 observations are left when all observations are dropped where the RIVALRY changes less than two standard deviations within the 34 periods.
3. Changes of the dependent variable, CO, are not often observed within the observations of a hotel. Only in approximate 10% of the hotels the ownership form changed.
Allison (2005) notes that when the independent variable does not change over time, panel regression is not the right specification, which leads to the conclusion that it could be better to do a straightforward logistic regression on the data. Therefore, I consider the data to be cross-sectional. However, this leads to the problem that there are multiple observations, in period 1 to 34, with the same (or similar) values. Therefore, I will create four different subsets which I will test separately:
Subset I - Period
For each period, 1 to 34, I will regress models M1 and M2 separately to obtain the effect of competition on ownership form. By splitting the periods, I prevent distorted results because of multiple observations with similar values. In other words, I want to prevent weighting, where certain hotels are more weighted because they have more (similar) observations.
Subset II - 1st observation
The second subset consists of all first observations in the dataset of a hotel. For a hotel that has observations in all time periods, this would be the observation in period 1. For a hotel that entered the Texas hotel market during the years the dataset spans, this would be the first quarter the hotel is included in the dataset.
Subset III - Entry
The third subset consists of observations when a hotel enters the market. Of course the owner of a hotel, when it is not independent, has to make the decision on entry if a hotel should be company owned or franchised which is the dependent variable in the regressions. Because I cannot determine if in the first period the hotel is new in the market, period 1 is dropped.
Subset IV - Change in ownership form
The fourth subset are hotels where the ownership form changed from franchise to company owned or vice versa. Only observations where the ownership form actually changed between two successive periods are taken into account. These observations are very interesting because they signal when a company changed its ownership form and at the same time give
information about the intensity of rivalry. This subset is limited to ownership change towards franchise or company owned units and not towards independent owners.
Before looking at the different subsets, it is important to note that the expected sign in the relationship between competition and ownership form should be positive. In absolute numbers that means that the average RIVALRY of company owned units should be larger than the
average RIVALRY for franchised units. Table 3 provides a comparison of the four subsets regarding company owned and franchised units. The intensity of rivalry is, as expected from the theory developed in this paper, for all subsets higher for company owned units than for franchised units. This provides preliminary evidence consistent with my hypothesis.
TABLE 3
Descriptive statistics of hotels, market conditions and ownership forms
Comparison between ownership forms (Standard deviation in parentheses) Subset I: PERIOD (AVGa) Subset II: 1ST OBSERVATION Subset III: ENTRY Subset IV: CHANGE IN OWNERSHIP FORM
Variable Overall CO FR Overall CO FR Overall CO FR Overall CO FR
Obs. 1798 490 1308 2269 562 1707 966 204 762 355 68 287 RIVALRY 0.76 (0.21) (0.20) 0.79 (0.22)0.75 (0.24) 0.72 (0.23) 0.74 (0.24)0.71 (0.26)0.71 (0.28)0.72 (0.25)0.71 (0.18)0.78 (0.14)0.79 (0.19) 0.78 FEE 3.89 (1.95) 3.17b (1.94) 4.16 (1.88) 3.84 (1.97) 2.98b (2.02) 4.12 (1.87) 3.69 (2.07) 2.08b (2.30) 4.12 (1.77) 4.04 (1.74) 4.16b (1.41) 4.01 (1.81) SIZE 119.93 (107.84) 162.80 (134.54) 103.84 (90.88) 115.27 (107.06) 157.13 (135.22) 101.48 (91.97) 95.45 (88.22) 151.11 (122.20) 80.54 (69.52) 134.89 (102.11) 216.21 (116.31) 115.62 (88.26) INCOME 21.81 (12.92) 25.50 (17.36) 20.41 (10.38) 21.92 (13.50) 25.46 (16.86) 20.75 (11.97) 22.95 (13.85) 28.93 (13.02) 21.34 (13.62) 22.16 (10.85) 26.90 (11.86) 21.03 (10.30) LOW 0.73 (0.44) 0.69 (0.46) 0.75 (0.43) 0.70 (0.46) 0.67 (0.47) 0.71 (0.46) 0.70 (0.46) 0.72 (0.45) 0.69 (0.46) 0.74 (0.44) 0.43 (0.50) 0.82 (0.39) HIGH 0.27 (0.44) (0.46) 0.31 (0.43)0.25 (0.46) 0.30 (0.47) 0.33 (0.46)0.29 (0.46)0.30 (0.45)0.28 (0.46)0.31 (0.44)0.26 (0.50)0.57 (0.39) 0.18 DENSITY 186.06 (116.97) 88.42 (57.52) 223.10 (112.45) 157.46 (101.60) 67.48 (49.66) 187.08 (96.78) 160.33 (112.14) 41.89 (50.92) 192.04 (102.41) 168.54 (135.01) 104.99 (61.67) 183.60 (143.12) ATTRACT 57.31 (8.92) 59.55 (8.21) 56.45 (9.02) 56.13 (10.95) 59.67 (10.49) 54.96 (10.85) 56.00 (10.98) 60.22 (11.36) 54.87 (10.60) 55.73 (9.23) 58.31 (8.85) 55.11 (9.23)
CO = Company owned units INCOME = Mean income per capita in zip code divided by 1,000
FR = Franchised units LOW = Hotel is within the low (budget) segment
RIVALRY = Intensity of rivalry calculated per period and zip code HIGH = Hotel is within the high (luxury) segment FEE = Royalty fee to be paid by franchisee to franchisor per
period
DENSITY = Number of units belonging to the same corporate headquarters in the state (TX) per period for a hotel. Example: Individual = 1 SIZE = Capacity of a hotel per period ATTRACT = Attractiveness of a market per period and zip code
a 34 values are calculated (for each period) from which the mean (average) is presented.
The capacity of company owned units is on average larger than for franchised units. A possible reason is that larger hotels are more complex to run and therefore managed by a professional. Furthermore, Table 3 shows that the average income per capita is higher for company owned hotels than for franchised hotels. This could be explained by the fact that more luxury hotels are built in more expensive areas, which are more complex to run, and therefore company owned.
Table 4 shows some characteristics of the two segments LOW and HIGH. These segments, LOW for economy hotels and HIGH for luxury hotels, are used to calculate the intensity of rivalry not only based on zip code but also on segment, the RIVALRY_SEGM. The data shows that the LOW segment hotels are smaller and are located in zip codes with less income than hotels in the HIGH segment.
TABLE 4
Descriptive statistics for the segments LOW and HIGH for all periods
(Standard deviation in parentheses)
LOW HIGH Number of observations 86428 28780 Number of competitors 3.45 (3.18) 2.94 (4.25)
Minimum price (ADR) 4.74 58.80
Maximum price (ADR) 78.51 994.83
ADR 41.98 (13.38) 100.23 (35.60) RIVALRY_SEGM 0.67 (0.27) 0.60 (0.32) FEE 1.94 (2.39) (2.42) 2.43 SIZE 72.42 (55.42) 127.75 (152.99) INCOME 18,267.27 (8,018.29) 25,500.35 (15,755.75) CO 0.13 (0.34) 0.18 (0.38) FR 0.39 (0.49) 0.39 (0.49) IND 0.48 (0.50) (0.50) 0.43
ADR = Average Daily Rate of a hotel
IND = Independently owned units
RIVALRY_SEGM = Intensity of rivalry calculated per period and zip code and per segment See Table 3 for an explanation of the other variables.
The next page shows a table with correlations between the variables used for the regressions with observations from all periods. The correlations confirm that large hotels tend to be professional managed (ρ=0.4114). Furthermore, it shows that company owned hotels operate in a more competitive market than franchised hotels (ρ=0.0878). The correlation between the size of hotel and the intensity of rivalry is strong and positive: ρ=0.2197. This indicates that large hotels compete in a more competitive market than small hotels. As both the Pearson and Spearman correlations of CO and RIVALRY are positive, these correlations provide more preliminary evidence in line with my hypothesis.
TABLE 5 Correlation Matrix(a)
(Spearman coefficients in the upper triangle; Pearson coefficients in the lower triangle) Approximately 61.000 observations are used to calculate every correlation coefficient
CO RIVALRY FEE SIZE INCOME LOW DENSITY ATTRACT
CO 1.0000 0.0878 -0.2960 0.4114 0.1815 -0.0562 -0.5741 0.1475 RIVALRY 0.0696 1.0000 0.0314 0.2197 0.0263 -0.1633 -0.0916 -0.0009 FEE -0.2284 0.0598 1.0000 -0.0693 -0.0334 -0.1277 0.3724 -0.0223 SIZE 0.2486 0.0951 -0.0374 1.0000 0.1834 -0.3614 -0.4227 0.1719 INCOME 0.1767 -0.0291 -0.0492 0.2276 1.0000 -0.2265 -0.1924 0.2187 LOW -0.0562 -0.1062 -0.1271 -0.3819 -0.2362 1.0000 0.2349 -0.2374 DENSITY -0.5115 -0.0760 0.3929 -0.2750 -0.1872 0.2568 1.0000 -0.2114 ATTRACT 0.1465 -0.0355 -0.0173 0.1186 0.1940 -0.2251 -0.1950 1.0000
(a) All correlations are significantly different from zero at 0.01 significance level of correlation coefficients.
27
Before presenting the regressions on the different subsets, I show in Table 6 the mean
comparison of the independent variable RIVALRY, for the two ownership forms. As expected from the theory, the variable RIVALRY is larger for company owned units, confirming the theory that with a higher intensity of rivalry, the market works as a natural monitoring
instrument. However, only for subset I: Period and subset II: 1st observation, the mean of the two
ownership forms is significantly different at the 5 percent level.
TABLE 6
Mean comparison of intensity of rivalry (RIVALRY)
Ownership form Subset Number of observations CO FR Significance of difference
I: PERIOD (AVG) Between 1303 and 2126 0.79 0.75 p < 0.05 II: 1ST OBSERVATION 2269 0.74 0.71 p < 0.05
III: ENTRY 966 0.72 0.71 n.s.
IV: CHANGE IN OWNERSHIP FORM
355 0.79 0.78 n.s.
For robustness of the results of the mean comparison test, I have verified with graphs of the data that there are no significant outliers and performed a Wilcoxon rank-sum (Mann-Whitney) test, which does not assume that RIVALRY is normally distributed, which showed that the groups company owned and franchised are significantly different except for subset III: Entry and subset IV: change in ownership form (see last column of Table 6).
Regressions
The previous paragraph presents a mean comparison of the intensity of rivalry between the two ownership forms: company owned and franchised. This paragraph will show regressions in the expectation that removing the effect of control variables on the dependent variable, these
regressions will more clearly indicate the relationship between rivalry and ownership form. Every regression, with RIVALRY as independent variable, is tested twice: once with only RIVALRY and once with RIVALRY and six control variables.
28
The dataset contains an extra control variable available: Brand. This is a dummy variable which distinguishes the 86 brands of hotels in the dataset. This control variable brand could be used to control for brand specific effects. From the 86 brands, 33 brands have only franchised as ownership form, 22 brands are strictly company owned, and 31 brands have units with mixed ownership forms: company owned and franchised. As a consequence, many brands (55 out of 86) have only one single form of ownership which could explain why, when I add brand to the regression as a control variable, the coefficient of rivalry becomes insignificant: there is not enough variance of ownership form within a brand. This could be explained by corporations, like ACCOR, that own 4 brands. Maybe ACCOR chooses a certain brand, and corresponding
ownership form, to suite the local market. In other words, ACCOR studies the local market, decides which ownership form would be appropriate, and uses their own brand that corresponds with the chosen ownership form. Future research could provide more insight if corporations like ACCOR link brands to ownership form and if they use their brands based on the characteristics of the local market. For the reason stated above, I have not used brand as a control variable in this paper.
Table 7 presents a logistic regression of Period 34. This period is chosen as example because it has the most observations (2126). It shows that hotels in competitive markets, i.e., with a higher RIVALRY, tend to be more often company owned; the coefficient of RIVALRY is 0.640, which supports the theory developed in this paper. The RIVALRY coefficient of 0.640 indicates that the log odds of company owned units (versus franchised) increases by 0.640 for a one unit increase in the independent variable RIVALRY or in other words that the odds of being a company owned unit (versus franchised) increased by a factor of 0.640 for an increase of one unit of the
RIVALRY. Controlling for other variables, such as size of a hotel, the royalty fee to be paid if the unit is franchised or which segment the hotel belongs to, alters the results of the 1st estimation
(RIVALRY coefficient from 0.640 to 0.994). As expected, the adjusted R2 increases from 0.002,
29
TABLE 7
Ownership as a function of the intensity of rivalry (Period 34)
(standard error in parentheses)(a)
Subset I: PERIOD 34 (1) (2) VARIABLES CO CO RIVALRY 0.640** 0.994*** (0.283) (0.352) SIZE 0.006*** (0.001) FEE -0.137*** (0.032) SEGMENT 0.767*** (0.175) INCOME 0.000 (0.000) DENSITY -0.011*** (0.001) ATTRACT 0.040*** (0.009) Constant -1.741*** -3.592*** (0.231) (0.709) Observations 2,126 2,126 Pseudo R2 0.002 0.291 a † p < 0.10, * p<0.05, ** p < 0.01, *** p < 0.001
To see if the presented evidence is consistent for all periods I run regressions for all 34 periods. The result is presented in Figure 2. It is clear from Figure 2 that for all periods separately the coefficient for RIVALRY is significantly different from zero (at less than 5 percent) and that the coefficient for RIVALRY is positive for all periods. This provides further evidence for the hypothesis in this paper that a high intensity of rivalry is related with a higher proportion of company owned units relative to franchised units.
30
Figure 2: RIVALRY coefficient and p-value for all 34 periods
Table 8 presents the same regressions but now on subset II: 1st observation, subset III: Entry on
the market of a hotel and subset IV: Ownership change. The results are similar for subset II: 1st
observation as the regression above on subset I: Period. The main difference is that the
RIVALRY coefficient in subset II is no longer significant different when control variables are added to the regression.
0.00 0.10 0.05 P-va lu e .5 .7 .9 1.1 1.3 R IVAL R Y -co e f. 1 8 17 26 34 Period
RIVALRY coef. p-value
31
TABLE 8
Ownership as a function of the intensity of rivalry
(1st observation, entry, ownership change) (a)
Subset II:
1st Observation Subset III: Entry Ownership changeSubset IV:
(1) (2) (1) (2) (1) (2) VARIABLES CO CO CO CO CO CO RIVALRY 0.439** 0.277 0.107 -0.244 0.413 -0.592 (0.213) (0.279) (0.308) (0.505) (0.785) (0.918) SIZE 0.003*** 0.010*** 0.006*** (0.001) (0.002) (0.002) FEE -0.122*** -0.382*** -0.045 (0.033) (0.060) (0.102) SEGMENT 0.725*** -0.344 -0.674* (0.152) (0.302) (0.398) INCOME 0.000 0.000 0.000 (0.000) (0.000) (0.000) DENSITY -0.021*** -0.023*** -0.005*** (0.001) (0.002) (0.002) ATTRACT 0.022*** 0.039*** 0.001 (0.006) (0.012) (0.019) Constant -1.430*** -0.668 -1.394*** -0.921 -1.743*** -1.299 (0.164) (0.489) (0.235) (0.906) (0.634) (1.381) Observations 2,269 2,269 966 966 359 355 Pseudo R2 0.002 0.352 0.000 0.538 0.001 0.199
a † p < 0.10, * p<0.05, ** p < 0.01, *** p < 0.001. (standard error in parentheses)
However, for subset III: Entry and subset IV: Ownership change the RIVALRY coefficient is not anymore significantly different from zero. This could be caused by the low amount of
observations (966 and 359). Also, there might be other reasons to change ownership than local competition. For example, in the dataset there are some hotels that frequently change ownership, i.e., more than four times during the 34 periods of time. For these hotels, the intensity of rivalry does not change significantly indicating that there must be other reasons why hotels change
32
ownership, for example when a franchisee retires and the franchisor needs time to find a new franchisee.
From the regressions in Table 7 and Table 8, it is clear that the size of a hotel has a significant positive influence on the proportion of company owned hotels compared to franchised hotels. This would support the theory that complex hotels or flagship hotels are managed by professional managers. Surprisingly, the royalty fee paid by a franchisee has a negative effect on the
proportion company owned compared to franchised. In other words, when the fee is higher the proportion company owned compared to franchised gets lower. Typically, one would expect that when the fee increases the interest to franchise decreases. A possible explanation for this negative effect could be that the royalty fee has a spurious relationship with brand, which as stated above, I do not add to the regressions because of a lack of variance within brands. Similarly as size, segment has a positive influence on company ownership. Income per capita in a zip code has no effect on the ownership form. Density, i.e., the number of units belonging to the same
corporation, has a negative relation with the proportion of company owned hotels. This result is not expected from the theory as a higher density should lead to lower monitoring costs, and therefore more company owned units. A possible explanation is that the state of Texas is too large to benefit from more units in the same area. The last control variable, attractiveness of the market, confirms with a positive coefficient the theory that corporations will keep more units company owned when the local market is more attractive.
Regressions with the segmented intensity of rivalry
This paper is interested in whether the intensity of competition in the market influences the choice of ownership. How the market is defined influences the measures of competition, which in this paper is the way the intensity of rivalry, RIVALRY, is calculated. Because competition possibly takes place not only within a zip code, but also within a segment (LOW/HIGH), this paragraph presents regressions with the independent variable RIVALRY_SEGM. Table 9 shows the difference between RIVALRY, calculated only on zip code and RIVALRY_SEGM,
calculated on zip code and within a segment LOW or HIGH. The results presented in the table are for all four subsets weaker for RIVALRY_SEGM than for RIVALRY. A possible
33
explanation could be that there are not enough hotels within a zip code to justify dividing the hotels into two segments.
TABLE 9
Comparison(a) of the intensity of rivalry (RIVALRY)
and
the segmented intensity of rivalry (RIVALRY_SEGM)
Subset RIVALRY RIVALRY_SEGM
I: Period (period 34) 0.64* 0.46*
II: 1ST Observation 0.44* 0.13
III: Entry 0.11 -0.21
IV: Change in ownership form 0.41 -0.81
a Result of a logit-regression test on the two groups company owned and franchised without control variables.
The numbers presented are the regression coefficients of RIVALRY and RIVALRY_SEGM respectively. † p < 0.10, * p<0.05, ** p < 0.01, *** p < 0.001
On average, see Table 1, there are only 3 hotels competing in a zip code and segment. Maybe this is not enough to calculate a meaningful RIVALRY_SEGM.
Robustness checks
The first assumption of logistic regressions is that the dependent variable consists of two categorical, independent (unrelated) groups, which in this dataset are the ownership form:
company owned and franchised. The second assumption is that the observations are independent; observations between groups or within a group should be independent. This second assumption is the main reason I have created four different subsets. Furthermore, I checked for multicollinearity between the independent variables and found no cause for concern.
Other concerns regarding the regressions are for example, but not limited to, reverse causality, endogeneity and missing control variables. Reverse causality in this paper is where competition does not influence ownership form, but vice versa, ownership form influences competition. To investigate if reverse causality is causing the results shown in this paper, I run the same
regressions with a lagged variable of intensity of rivalry. The idea is that it is impossible, or at a minimum less likely, that the current ownership form influences the rivalry of one or two years ago. Figure 3 shows the result of 34 periods for three different regressions: once with the
34
intensity of rivalry in the same period, once with a lagged rivalry of one year, and once with a lagged rivalry of two years.
Figure 3: Comparison of regressions with rivalry and lagged rivalry
With similar results for the different regressions, reverse causality for the model presented in this paper does not seem to pose a serious problem.
Endogeneity problems occur when the error term correlates to an explanatory variable. Sources of endogeneity are missing variables, simultaneous and / or reverse causality, measurement errors and sample selection. The endogeneity problem can for example be solved by adding an
appropriate instrumental variable. An example of an instrumental variable could be if the state of Texas changed the law, and by doing so influence competition, during the observed period. Unfortunately, I do not have such an instrumental variable.
Control variables are variables which influence the dependent variable, but that particular influence is not the main topic of this paper. An example of a control variable is the amount
0.00 0.10 0.05 P-va lu e .5 .7 .9 1.1 1.3 RI V A LR Y -c o ef . 1 8 17 26 34 Period coefficient p-value
Lagged one year Lagged two years
35
needed for the initial investment of a new unit. Although I tried to control for influences on the dependent variable, for example to add the size of a hotel, there are many more possible influences I could not control for.
Conclusion
For the subsets I and II, all regressions show a positive relationship between RIVALRY and company owned units, confirming the theory stated in this paper. However, for the subsets III and IV these findings are not significant. The independent variable RIVALRY_SEGM, calculated on zip code and segment, only partially supports these findings.
Discussion
This paper shows evidence that market competition influences ownership form. More
specifically, this paper argues that more intense rivalry lowers monitoring cost and thus makes company ownership more preferable compared to franchising units. The results provide preliminary support for this argument.
In the theory presented in this paper, some factors are discussed that influence the ownership form of a company. While resource scarcity, knowledge and agency problems favor towards franchising, free riding, hold-up problems and high-powered incentives favor company
ownership. This paper adds natural monitoring, through competition, as a factor influencing the ownership form.
Previous literature has demonstrated that several factors influence ownership form through monitoring costs: distance to headquarters, outlet density, improved technology and returning customers. This paper adds one factor: competition.
However, this paper cannot determine indisputable if competition influences the ownership form through monitoring costs. An alternative explanation could be that competition relates to the profitability of a market which influences the chosen ownership form. I tried in this paper to
36
control for this different route with the control variable attractiveness of the market, but that does not fully eliminate the possibility that there are other mechanisms than monitoring costs
influencing the ownership form. For example, knowledge could be an alternative route for competition to influence ownership form. When a market has less competition, i.e., less rivals, the knowledge of that market is also less. Therefore, according to Minkler (1990), it is better to franchise. In this case, less competition influences, through knowledge, the ownership form. Other mechanisms could be local demand and/or lifecycle of the market.
These different mechanisms, such as knowledge, local demand and lifecycle are not investigated in this paper, and are therefore limitations on the results found. These different mechanisms could play a role on the choice of ownership form and would therefore be a good subject for future research.
Another limitation of this paper is that the intensity of rivalry on a market in equilibrium not only depends on the supply side, i.e., the number of available rooms in a market, but also depends on the demand side. To mitigate this problem, I added attractiveness of a market as proxy for demand, but it would be better to measure demand directly, by for example questionnaires with hotels.
To improve the results it would be interesting to do the same study but now on data in a European country, such as France. With more than 65,000 hotels, according to Hotelscombined.com8, there
is more competition which will lead to a more accurate measure of the intensity of rivalry compared to Texas with 4500 hotels in an area larger than France. While this paper did not find significant results for a segmented RIVALRY, an analysis using data from France could. It would also be interesting to replicate this study in other franchise sectors than hotels: for example, supermarkets, fast food and normal restaurants, and real estate agents.
The results suggest that when deciding ownership form companies take the intensity of rivalry as one of the factors before deciding if a unit should be franchised as more competition may
37
diminish the need to monitor, favoring company ownership. An implication of the result of this paper could be that companies study more carefully the local market competition before deciding on the ownership form.
While this study has controlled, to the extent possible, for other factors influencing the choice of ownership form, future research could develop a more comprehensive framework of all
determinants of ownership form as found in previous literature, allowing for the investigation of the interaction between these different determinants, including distance to headquarters and the intensity of rivalry.
Given the global trend that independently owned retail outlets and restaurants are replaced by outlets that are member of a chain, the question of ownership form is becoming increasingly relevant. As a consequence, more research should explore the antecedents of ownership form choice.
In many decades if not centuries of research on the consequences of competition, to my
knowledge the effect of competition on the choice of ownership has never been investigated. This study is a first step in understanding how competition affects ownership choice, in the process shedding more light on the underlying causal mechanism of monitoring costs. More research is needed to further enlarge our understanding of competition on other factors of importance in the economy.
