1
Possible drivers in geographical expansion of a service firm - a case study
Academic Year 2017/2018 Master Thesis
Master of Science in Business Economics Specialization Managerial Economics and Strategy
Faculty of Economics and Business University of Amsterdam
Student: David Bonanno Student number: 11805250 Date: August 2018
First supervisor: prof. dr. R. Sloof Number of ECTS: 15
2 Summary
Abstract ... 3
1. Introduction ... 4
2. Theoretical background and literature review ... 6
3. Dataset ... 12 4. Results ... 19 5. Limitations ... 25 6. Conclusions ... 27 References ... 28 Statement of Originality
This document is written by Student David Bonanno, who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document are original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
3 Abstract
There are plenty of studies regarding the integration of business processes. In the past, the major part was related to the manufacturing of physical objects. More recently, another strand of studies focused on the integration decision for companies operating in franchising. This thesis combines theories from multiple sources and aims at testing whether frequency, costs of control and size of business unit to internalise influence integration decision. To do so, a dataset from an Italian company that recently underwent this kind of choices is analysed. Results support only costs of control as variable taken into consideration when deciding to integrate or not by the firm in analysis.
4 1. Introduction
When a firm is planning to expand its area of operation, the first question that should be answered is whether to do so by investing directly in the opening of a new company owned branch or to enter into a business relationship with a local company. In fact, at the base of this choice lies a trade-off between costs and control: the former option would result in higher control of the operations, but high costs, on the other hand the latter would allow lower investments, but also imply lower control and more exposition to hold up problems. According to the market firms intend to expand, one or the other option may be more profitable, as maybe the investment costs are limited or the risk of hold up is low.
Prior literature focused mainly on the vertical integration of each single point of sales given its characteristics, according to some theories such as Moral Hazard, Transaction Cost Economics and Property Rights Theory. Some variables taken into consideration for this kind of studies are revenues (Martin (1988), Norton (1988)), distance from headquarters (Minkler (1990)), or risk (Woodruff (2002)). However, it may happen that firms decide the go-to-market plan on a regional basis. This may be the case especially if the firms in consideration render services to groups of customers located in specified areas and points of sale or premises are not needed to be directly accessible for clients. As multiple customers will be served based on geographical grouping, we can treat variables at a regional level, as the go-to-market strategy will be planned by area and not by specific client. Specifically, we want to test whether monitoring costs, frequency of the transaction, potential customer base, together with characteristics of possible third companies operating locally, have influence on the different strategy a company may adopt. According to what the company subject of the case study does, we also consider two ways of disintegration, further explained during the analysis.
This thesis will use a dataset from an Italian company that supplies and support supermarkets and restaurants with their ICT instruments (both hardware and software), such as cash registers and tablets for orders, in order to perform a case study and validate the theories treated. Italian provinces will give the size of each area for which the decision should be taken. The main focus will be on the expansion on the Italian territory and how they cover their clients throughout the nation depending on the density of clients (acquired and potential), the distance from the headquarters and the size and shared areas of possible partners or outsourcers. A priori, we could expect that the number of clients in a certain area will induce the firm to a direct investment, as more control is desired and the investment cost can be better distributed across all clients. On the other hand, the further away
5
the new target area is, the more difficult it is to control the branch, and maybe it would be easier to rely on a company that is already present in the market. We expect also companies to be reluctant to disintegrate provinces that constitute a high share of their revenues.
Results show that just the costs of control play a major role in the decision of the firm in analysis, with a slight influence of frequency on the way to disintegrate.
The remainder of the thesis is structured as follows: in the next section a theoretical background as well as a brief literature review are drawn before stating the research question and the hypotheses, while in section 3 the data and the model are presented. Then, the results of the model are discussed in section 4. In section 5, possible limitations are taken into consideration. Section 6 concludes the thesis.
6 2. Theoretical background and literature review
In this chapter some of the main theories on vertical integration regarding potential factors in analysis are briefly reviewed. We start by mentioning some of the most prominent theories in Organisational Economics which attempt to define the boundaries of firms based on frequency of transactions: Transaction Cost Economics and Decision Rights Model. Subsequently, we take into consideration the effect of Monitoring Costs, explained by theories based on the basic Moral Hazard model. Then, the size of the unit to potentially integrate is considered in light of previous literature regarding this variable. We conclude the chapter by stating the theoretical hypothesis that will be tested in the case study.
Frequency
The first theory in analysis that aims to explain the propensity for integration based on frequency is the Transaction Cost Economics theory1: according to it, production and activities should be internalised when serious contractual incompleteness may undermine the effectiveness of a business relationship. However, a contract regulating the relationship between two businesses would be incomplete by definition: in fact, according to Hart (1995), it is possible to trace down three main sources of contract incompleteness. The first is that it is not possible to foresee the future, and even to think very far ahead, so some scenarios may not be regulated in the contract. Parties will just focus on the most plausible contingencies. A second source for incompleteness is the fact that, provided that some plans can be made, parties must find common wording and agree on them. The last source of contractual incompleteness is that, once parties have agreed on the contractual terms, it must be possible to verify and enforce the agreement in a court of law, if needed. In addition, some of aspects of a relationship, such as the effort level of the agent in a typical principal-agent model, are difficult to measure, contract on and prove in front of a judge, as there is no way to measure. For these reasons, all contracts are somewhat incomplete.
To cause its undesired effect however, contractual incompleteness must be combined with transaction characteristic of complexity, specificity and, indeed, frequency. Complexity measures how well a solution can be defined and contracted on by the two parties and specificity indicates the degree to which the investments undertaken by the parties are relationship specific, and therefore how much surplus is lost if the relationship terminates. Frequency, on the other hand, simply indicates how many times the transaction occurs. A high number of transactions will result in a high chance for one of the firm to hold up the other.
7
In order to minimise the problems that may arise from this contractual incompleteness, a firm may want to internalise part of the value creation: if there is a conflict between the contractors, a (possibly) long and troublesome legal dispute may originate. On the other hand, if both processes are carried out internally, a senior manager could resolve the conflict in a fast and inexpensive way, as the employees will have to comply with his/her decision, keeping also in mind the welfare of the company as a whole.
In the 80s this theory has been extensively researched. Both Montverde and Teece (1982) and Masten, Meehan and Snyder (1989) researched the car manufacturing industry to verify what led to vertical integration. They based their papers on two of the biggest car manufacturer at the time in the USA, Ford and General Motors. Their results point toward the validity of Transaction Economics Theories: components which required extensive engineering efforts and/or are very specific to the car, tend to be produced in house. On the other hand, inputs with low degree of engineering or common across different vehicles, or even producers, are more likely to be outsourced. The same results have been obtained by Masten (1984), researching the vertical integration of the aerospace industry. It has to be noted however, that neither of the two studies focus on frequency of transactions.
As regard our hypothesis for this thesis, we assume, accordingly to the basics of Transaction Cost Economics2 that frequent transactions may increase the hold-up chances. Therefore, we expect that the higher the number the more a firm will internalise processes, as not to leave room to haggling behaviours.
Another approach that helps relating frequency of transactions to integration decision is the adaptation theory. According to it, the final owner of the company detains the residual control rights, which are “the rights to decide all usages of the asset in any way not inconsistent with any prior contract custom, or law.”3
Therefore, in case of any disruption or occurrence of any even to foreseen in the contract, each company may act in its own best interest, even if this means causing harm to the other.
Adaptation theory is part of the broader Decision Rights Model: a party in a contract may want to purchase the decision rights over the other party if the price for doing so is lower than the expected loss (or foregone revenues) caused by the choice of the latter in a disruptive event.
2
See Williamson (1981)
8
Forbes and Lederman (2009) focus on adaptation as motivation for integration. They take into consideration the airline industry and analyse whether flights are operated by major airlines or regional carriers, distinguishing the latter between regional ones owned by major or independent ones. The unforeseeable disruptions to the contract usually come from precipitations, as bad weather increases the chances of delay or cancellation. It must be noted that interests of major and regional airlines usually don’t match: majors focus on the overall benefit, taking into consideration also other connecting flights within their network, while regional carriers are interested in maximising profits just for their routes. Results show that in more rainy or snowy routes, and therefore more prone to disruptions, major airlines prefer to operate themselves or to use owned regional carrier rather than independent ones.
Again, this theory supports the prediction that frequency is positively correlated with integration. In this thesis’ framework, the higher is the number of clients in an area, the more are the transaction susceptible of hold-ups. By internalising a company or a process, the number of transactions is cut off, and chances of holding-up together with them.
Monitoring Costs
The second measure of interest for this thesis is Monitoring Costs. The analysis of the influence of Monitoring Costs on integration derives from the Moral Hazard model and from the Incentive Pay literature. According to the theoretical model, a principal delegates certain activities to an agent. The profitability of these activities depends on the effort that the agent exerts. As the effort is typically not contractible4, the agent would have the incentive to not exert effort and to cash out just a fixed remuneration. To avoid this, the principal can tie the agent’s remuneration to the outcome as to align their interests. If the principal does so, the agent is incentivised to exert effort, as it directly impacts the productivity (and therefore his payoff). A further step of this model introduces the possibility for the principal to control the agent by monitoring it. However, the monitoring activity has a cost that reduces the principal payoff. In order to monitor a branch, there will be travelling costs to reach it. Furthermore, a manager has to forego some of his activities to carry out the inspections, or even a new employee must be hired with the specific task of controlling the separate branches. If monitoring is too expensive, a principal may be better off just using a high-powered incentive scheme and not directly controlling agent’s activity. In this way the agent is incentivised
4
The effort level is assumed not contractible as it is not possible to prove it in front of a judge/third party. See Hart (1995)
9
even more to exert effort and his interest are much more tied to principal’s ones. In fact, as Holmstrom and Milgrom (1994) find, usually higher degrees of freedom correspond to high-powered incentives (often the ownership of the assets as in franchising), while low-high-powered incentives are commonly used combined with stricter control.
This model has been extensively researched in the Franchising literature. When opening a new owned branch of a chain (e.g. fast food, car dealers, etc.), a firm may decide between two different options: opening a new branch or find somebody willing to franchise the name. If the company chooses to open a new branch, it must ensure itself a complete alignment of it to the company policy, besides dealing with the entire bureaucratic burden. On the other hand, franchising to somebody else may reduce these costs since less control is required. In fact, agent will be more aligned to principal objectives through the higher share of variable remuneration computed on the outcome. Brickley and Dark (1987) research the ownership form of companies operating in several industries also based on distance to the closest headquarters. As they assume that “owned units require additional monitoring visits”, their hypothesis is that “owned units will be observed in location where cost of monitoring is low and franchised units where the cost of monitoring is high”5
. Distance increases indeed travelling costs for monitoring, and for this reason, units far away from headquarters are expected to be franchised. Their results support their hypothesis, as they find a statistically significant positive relationship between distance and the probability of franchising a unit. A similar study has been conducted by Minkler (1990), which study the ownership form of Taco Bell restaurants located around Sacramento, where headquarters are located. Again, results show that closer restaurants have more chances to be integrated, while further restaurants are more likely to be franchised.
Size of the business unit
Another measure that may influence the integration decision is the size of the business unit. This argument is derived again from the Moral Hazard model and assumes that outlet size does not influence the optimal decision per se; however, it recognises that the size of the unit is usually interacting with the uncertainty on the market by amplifying the risk component of the agent’s remuneration. As agents are usually risk averse, they will be less likely to accept a contract, ceteris paribus. For this reason, we expect to find that the biggest units will be less likely to be integrated.
5Brickley and Dark (1987), p. 408.
10
One of the studies that takes into consideration size of units as factor for integration is Martin (1988). He uses the risk influencing framework discussed above together with a monitoring cost analysis. He finds that almost across all industries included in the specification, outlet sales (taken as measure of outlet size) were higher in company owned outlet than franchised ones. Another study on this topic is from Lafontaine and Shaw (2005). They make use of a panel data to monitor the development of franchising across several businesses operating in different industries in the U.S.A. and Canada. Again, they find that the number of employee (used as proxy for outlet size) is higher in company owned units. They motivate this finding by stating that firms have more at stake in larger outlets, since operation are more complex, the turnover is higher and can generate higher externalities. Furthermore, by running the biggest outlets directly, firms aim also at building and protecting their brands.6 According to this view, the risk approach does not limit itself to the risk propensity of the agent, but also to the one of the principal, as he may have incentive to keep control of the flagship outlets.
It is interesting to note how the measurement of outlet size has been proxied by several different variables. Lafontaine and Slade (2007) summarise the existing literature on vertical integration and firm boundaries based on the explanatory variables used. In the past, variable such as initial investment for franchisee, average sales, paper capacity for pulp and paper mills, number of rooms in hotels, employees per outlet and value of plant shipments for manufacturing plants have been used. Despite all the variety, in all but one studies, higher size of units implies that integration is more likely with high statistical significance.7 Therefore, literature points quite strongly to the direction where larger business units are more likely to be integrated.
Hypothesis
The research question that stems from the literature reviewed is:
How does frequency, costs of monitoring and size of business units influence integration decisions taken not at a single shop level, but at provincial level?
Here are summarised the hypothesis that follow from the past literature about integration. The theories mentioned above will be assessed in a framework where the decision to integrate is taken for a predetermined area and all clients of a firm in that area will be served according to the same strategy.
6
See Lafontaine and Shaw (2005), p. 145.
11
Hypothesis 1: The higher the number of clients in an area, and therefore the higher the number of transactions, the more likely a business will operate directly in that area.
Hypothesis 2: The further away the area is from the headquarters, and therefore the higher the monitoring costs, the less likely the area will be integrated.
Hypothesis 3: The higher the revenue share for a given area (as measure of size on the overall business revenue), the higher the chance a firm will operate directly.
In the following chapter, these hypotheses will be re-examined based on the actual variables available in the case study.
12 3. Dataset
In order to verify the theories explained above and possibly apply them to geographical expansion, I wanted to analyse a firm that recently had to take these decisions. The firm analysed (called firm A due to privacy requests) is a Small-Medium Enterprise with 80 employees and a turnover of 10 million Euros. It operates in the Information and Telecommunication sector and the core business is providing and maintaining PDA computer, cash registers and computer networks (both hardware and software) to shops, supermarkets and restaurants.
It started operating in the 90s in the north of Italy and at that time the main focus were slicers and low-end cash registers for small shops. During the year, to meet the market demand and shift towards new technologies, it started specialising also in software platforms for Mass Market Retailers (MMRs) and restaurants.
Firm A has recently won some contracts with supermarkets and restaurants chains for which it had to guarantee support coverage all over the Italian territory. This situation let the firm face the choice whether to invest directly (by opening a new branch or acquiring an already existing company) or to set up a binding contract to a local firm already operating there. The latter option includes two different perspectives: finding an outsourcer or a partner. Usually, the decision wheter or not to integrate is taken first. Then, if the choice has been not to integrate, based on the best candidate available, partnership or outsourcing contracts are proposed. In the following paragraph these alternatives are examined.
Direct investment
This solution is the one with the lowest hold up costs, as they are non existing. Firm A operates directly in an area or buys a firm already operating there. As all the transaction are carried out internally, there will not be any chances of haggling behaviours as any dispute can also be solved internally. However, this solution implies high cost of control, as it will be required the presence of a business representative to control the operations of the branch (section Monitoring Costs in previous chapter).
Outsourcing
Firm A assigns to a local firm the support of its client in the chosen areas. Typically, outsourcing entails high powered incentives, as a flat rate is paid from Firm A to the local firm and whatever the latter can save makes up its profit. It is also the easiest relationship to break: if a local firm does not
13
keep up with the standards or acts against the interest of Firm A, it is quite easy to terminate a contract and move on to another local firm, as usually little relationship specific investments are made, if any (the major item may be training courses). The incentive for local firms not to engage in haggling behaviour and to satisfy Firm A request is indeed given by the fact that they can easily lose the clients passed down from Firm A.
Partnership
If the relationship with an outsourcer is straightforward, as it operates simply as a sub-contractor, the relationship between the firm and a partner is more articulated. They are similar in the fact that no big initial investments are required and the relationship specific investments would consist just in training courses. On the other hand, this form of interaction usually takes place when not only Firm A has some part of its clients in the partner area, but also when some of the partner’s customers are on Firm A covered area. In this way, they would serve each other in the difference areas of operations. One of the positive aspects of this solution is to mitigate the hold-up problem, since firm A is able to reciprocate the behaviour of the partner firm. In fact, if a partner does not serve well Firm A’s client in its area, Firm A may decide not to put effort in serving their partner’s clients either. Also the contrary is true: if a local firm puts much effort to serve well Firm A’s clients, it would correctly expect their clients to be served at best as well.8 We expect this solution to be the preferred one when possible, as it allows having more control over the area than with a simple outsourcer without the hefty initial investments and slow adaptation to market changes that would derive from that. Compensation for a partner is usually computed on the basis of the actual interventions it did for Firm’s A customers and the interventions of Firm A’s in support of outsourcer’s clients.
In order to verify what may influence this decision, I collected some objective variable that, according to the previews literature examined in the last chapter, may have explanatory power. The decision is taken to provincial level, so all the data refer to the go-to-market strategy for the given province. Only data for supermarkets and restaurants were available, as supply to small shops is now limited and not meticulously recorded.
Number of supermarkets and restaurants (Shops)
8 In this frame work reciprocity and inequity aversion is assumed to hold, according to Fehr and Schmidt (1999).
14
This is the number in units of supermarkets and shops within the provincial borders. We expect, according to the theoretical analysis carried out in the previous chapter, that the higher the number of clients, the more likely the firm is willing to invest, as it a higher number of clients implies a higher frequency of transactions. Furthermore, a higher number of clients allows spreading the initial investments, lowering the investment per client ratio and therefore allowing for lower prices.
Revenue for each province
This is the contribution in percentage of each province to the total revenues of the business. It is shown as percentage amount as requested due to privacy by the firm itself and it is another measure for the volume of the business. We expect that the higher is the turnover, the more the firm has incentives to invest, as the given area has more strategic importance. Furthermore, if another company (partner or outsourcer) operates under Firm A’s name, they may not have incentives to exert as much effort as if they were operating under its name. This would lead to potential ruining of image for Firm and not having the contracts confirmed in the long run. Another expectation is that, given the firm decides not to invest directly in a highly worth province, it will prefer a partner, as their interests are more aligned due to the opportunity to reciprocate.
Distance from Firm A’s headquarter
This metric is a proxy for monitoring costs. It is computed as the distance in kilometres of a one way road trip from the headquarters of Firm A to any provincial administrative centre (in Italian “capoluogo di provincia”, the city where administrative institutions at provincial level seat). The distance has been recording within the 8th and the 12th May 2018 and within the 12th and the 16th June 2018 using Google Maps as reference, set to find the shortest route possible. The registration took place twice as to make sure to get the shortest route without the influence of road works and repairs, which would have made the distance bigger. Provincial administrative centres, differently from regional administrative centres, are usually located in a central position within the area they are capital city of. Even if a small number is not, the limited size of provinces will not let them influence the results. From our previous analysis, we expect that the closer the province is to the headquarters, the cheaper is for the firm to exercise control over the team working there. As we go further away, the more difficult it is and more effort will be required to communicate with the headquarters.
15
Firms will prefer not being held up in huge investments with team located far away and difficult to control, maybe even in a non familiar market and therefore they will be more likely to choose to turn to a company already operating in the area.
Number of inhabitants
By accounting for the number of inhabitants, we want to account for the potential business to develop in a certain area. A high amount of inhabitants in a certain area would mean that there is a large number of shops and services that may become clients. According to the theories reviewed in the previous section, we might expect that the number of inhabitants in a certain area is positively correlated to the probability of integration.
Overlap
Overlap measures the number of shops that a firm has within Firm A’s integrated areas. It is expressed in tens and the number is rounded to the nearest ten. This is due to the fact that this number may vary and keeping track of the single shops may be an inefficient amount of work. A higher number of clients in each other range gives more chance to reciprocate (and therefore align the interests). For this reason it impacts the costs of monitoring: the more others’ clients within Firm A’s region, the more chances the latter has to reciprocate and therefore the less it has to spend on direct monitoring. We expect that firms with high numbers of overlap will be more likely to be chosen as partners.
Size
Size measure the size of a firm in terms of employees. The effect is not immediately clear: a small firm may be preferred to establish a partnership as it will be easier to bargain with; on the other hand, a big firm already known in the local market may signal better service quality.
Distance between HQ
While we measure the distance between Firm A’s headquarters and each provincial administrative centre for the integration decision, we allow for each province to be served by a firm that may not be located there if the decision has been not to integrate. Hence, the variable Distance to HQ measures the distance from the headquarters of Firm A to the headquarters of each of the firm operating in a given province, using the same methodology
16
as explained for Distance. In this case, we expect Firm A to be more likely to choose a partnership as relationship if the other firm is located far away: in partnership, incentives to deceive are less and therefore less monitoring is required. On the other hand, when Firm A just outsources an area, the amount of monitoring required is higher, despite being always lower than what required by direct investments. Given a certain amount of monitoring, the cost of doing so will be lower in partnership than in outsourcing.
The model
The decision model of the firm is articulated in two different phases: in phase 1, the area is analysed and the decision whether to invest directly or turn to a local business is made. In stage 2, if the latter choice has been made, the business assess the local firm and decide if it is better for them to set up a partnership or a sub-contractor relationship. In order to capture this, I am going to use two separate logit regressions. The first one will assess the probability of internalising a province, while the second will determine the probability of having a partner, rather than an outsourcer, running the area after the decision of non integration has been taken. The specifications are as follows:
Regression (1)
where:
i =1,...,101 represents each province numerical ID.
INTEGRATION =1 if Firm A decided to invest directly, 0 otherwise REVENUES is the measure in percentage of the revenue from the area
DISTANCE is the distance in km from Firm A’s HQ to the provincial administrative centre INHABITANTS is the number of people residents in a province
is the random error term.
This regression analyses the probability P of Firm A to integrate given the variable cited above. It uses the distribution function , which is the distribution function of a logistic variable. For a given variable t, the distribution function will look like:
17
and t will be computed like:
Regression (2)
where:
j =1,...,82 represents each province numerical ID for which integration has not been chosen..
PARTNERSHIP =1 if Firm A decided to set up a partnership, 0 otherwise
DISTANCEHQ is the distance in km from Firm A’s HQ to the outsourcer’s or partner’s HQ
SIZE is the amount of employees of the outsourcer or partner
OVERLAP is the tens of partners/outsourcer clients within Firm A area of business
SHOPS is the sum of supermarkets and restaurants in the province is the random error term.
This regression analyses the probability P of Firm A to set up a partnership according to the data included in the regression. It uses again the distribution function , but this time variable v is defined differently:
and:
As high correlation will be found between number of shops (sum of restaurants and supermarkets) and revenues, multiple separate regressions will be run (containing one variable among
REVENUES, SUPERMARKET, RESTAURANTS and SHOPS per specification). Due to the close
positive relationship and meaning of these variables, we are able to extend the expectations of the coefficient of one variable also to the others.
18
Now that the regressions are defined, it is possible to list the predictions over the sign of each coefficient according to the literature reviewed in section 2.
As our hypothesis is that revenues are positively related to integration, we expect this coefficient to be positive
As cost of control increase with distance and makes integration less efficient, we expect this coefficient to be negative.
Higher population let us presume more potential clients and more business opportunities. For this reasons, we expect this coefficient to be positive
As discussed above, we expect Firm A to be more willing to commit into a partnership when the costs of control (and therefore the distance between headquarters) decreases. For this reasons, we expect this coefficient to be negative
As already stated this coefficient is somewhat ambiguous and may capture subjective preferences of Firm A’s management. Overall, we expect it to be positive, as big firms are usually more famous and more reliable.
This coefficient is expected to be positive as the higher is the number of clients in Firm A’s area, its interests and its partners are aligned.
This coefficient is expected to be positive as, the higher number of transactions and revenue share, the more Firm A will prefer to rely on partner lowly incentivized to engage in opportunistic behaviours.
In the next section, after a review on the statistical highlights on the data, the results of the regressions are presented and discussed.
19 4. Results
This chapter contains the analysis of the data and of the regression introduced in section 3, as well as a brief discussion on these results.
We start by a simple review of the summary statistics of the data, contained in Table 1. Table 1: Summary Statistics
Variables N mean sd
Strategy : Direct Investment
Supermarkets 19 9.947 4.961 Restaurants 19 5.053 3.643 Shops 19 15 7.86 Revenues 19 .017 .008 Distance 19 193.737 97.135 Inhabitants 19 633000 343000 Strategy: Outsourcer Supermarkets 38 3.763 2.353 Restaurants 38 2.026 3.158 Shops 38 5.789 4.709 Revenues 38 .006 .004 Distance 38 584.184 227.902 Inhabitants 38 466000 501000 Overlap 38 .763 .883 Size 38 42.921 9.095 DistanceHQ 38 542.974 235.105 Strategy: Partner Supermarkets 44 6.25 7.555 Restaurants 44 2.545 6.673 Shops 44 8.795 13.711 Revenues 44 .01 .013 Distance 44 823.864 412.989 Inhabitants 44 672000 801000 Overlap 44 4.614 1.603 Size 44 44.909 8.741 DistanceHQ 44 689.205 308.991
The summary is divided into the three types of strategy that Firm A can adopt to better appreciate possible differences. Out of 101 provinces, 19 are directly served by Firm A’s employees, 38 are
20
outsourced and 44 are served by partners. It can be noticed that the mean of supermarket and restaurants (and therefore the mean of the overall number of shops) is higher in provinces served directly from Firm A than in the one disintegrated at least at a 5% significance level. Revenues are higher and distance is lower even at a 1% confidence level. On the other hand, there is no statistical difference between the mean numbers of inhabitants.
Comparing the two way of operating under disintegration, we find that the difference in the average amount of shop is not statistically significant due to the non significant difference in the average number of restaurants. Taken alone, the sole mean of supermarket is higher in Partnership at 10% significance level. Also the average size of the firms chosen for one or the other strategy appears not to be statistically different, as well as the average province population. On the other hand, average distance to the provincial administrative centre and to other firms’ headquarters distance is statistically higher at least at 5% level, same level of significance of the difference between the averages of overlap and revenues, again higher in Partnership.9
Before analysing the results from the regressions, we consider the correlation coefficients reported in Table 2.
Table 2: Correlation Table
Supermarkets Restaurants Shops Revenues Distance Inhabitants Overlap Size DistanceHQ
Supermarkets 1 Restaurants 0.788 1 Shops 0.954 0.937 1 Revenues 0.998 0.791 0.954 1 Distance -0.198 -0.252 -0.236 -0.197 1 Inhabitants 0.727 0.807 0.808 0.721 0.021 1 Overlap -0.063 -0.143 -0.105 -0.062 0.453 0.042 1 Size -0.339 -0.221 -0.300 -0.343 0.511 -0.063 0.456 1 DistanceHQ -0.245 -0.250 -0.262 -0.245 0.933 -0.005 0.488 0.649 1
As expected, the correlation among Supermarkets, Restaurants, Shops (the sum of the previous two) and Inhabitants is high: more facilities will be opened where there are more people to serve. For this
9 Differences between provinces where Firm A has directly invested and provinces where it has not, has been computed
using the data from Partnership and Outsourcing as a single set representing disintegration. The levels of significance have been computed using the Student t-test.
21
reason, several specifications of Regression 1 are computed to avoid multicollinearity. Also Distance and Distance between headquarters are highly correlated. This is due to the fact that, if Firm A chooses not to integrate, it will look for a partner operating in the nearby of the area. Regression 2 will be dependent only on Distance between HQs.
We will now analyse the integration decision of Firm A based on the results of the regression. In Table 3 the regression coefficients for Regression 1 are shown.
Table 3
(1) (2) (3) (4)
VARIABLES Integration Integration Integration Integration
Distance -0.052** -0.051** -0.052*** -0.057*** (0.023) (0.023) (0.019) (0.022) Inhabitants 0.000 0.000 0.000 0.000 (0.000) (0.000) (0.000) (0.000) Revenues 19.098 (192.907) Supermarkets 0.060 (0.308) Restaurants -0.160 (0.223) Shops -0.087 (0.168) Constant 15.193** 14.776** 14.788** 16.619** (7.177) (7.012) (5.775) (6.533) Observations 101 101 101 101 Pseudo R2 0.861 0.861 0.866 0.863
Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
From the table it is immediately clear that Distance is highly correlated with the decision to integrate: in the first two specifications, with Revenues and Supermarket, Distance coefficient is significant at 5% level, rising to 1% level when using Restaurants or the total amount of shops. As this coefficient in negative, it supports our hypothesis: the further away the province is from Firm A’s headquarters, the less likely it is to be integrated in its area of direct operations. Inhabitants coefficients are never significant. This means that apparently the choice whether to integrate is not related to the amount of potential business that may develop in a certain province. Considering
22
customers already served by Firm A, we find mixed results, but never significant at any confidence level. Revenues coefficient seems to point in the direction of our hypothesis: being positive it suggests that area of great value for Firm A will be more likely to be covered directly. However we are unable to draw this conclusion due to the high insignificance of this coefficient.
It is however interesting to notice the high value of Pseudo-R2: all specifications of Regression 1 capture more than 86% of the total variance. Despite being Distance the only variable for which we found a strong correlation, it is able to explain the major part of the integration choices.
Now, we are going to analyse coefficients from Regression 2, as shown in Table 4. Again, this logit regression expresses the probability of setting up a partnership with the local firm selected, after having renounced to invest directly in the area.
Table 4
(1) (2) (3) (4)
VARIABLES Partner Partner Partner Partner
Overlap 2.464*** 2.437*** 2.284*** 2.358*** (0.688) (0.675) (0.604) (0.638) Size -0.040 -0.039 -0.024 -0.032 (0.065) (0.065) (0.060) (0.062) DistanceHQ -0.001 -0.001 -0.001 -0.001 (0.002) (0.002) (0.002) (0.002) Revenues 161.586 (146.629) Supermarkets 0.227 (0.174) Restaurants 0.094 (0.063) Shops 0.069* (0.040) Constant -4.753* -4.501* -3.912 -4.042 (2.868) (2.700) (2.496) (2.537) Observations 82 82 82 82 Pseudo R2 0.799 0.794 0.757 0.773
Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
From this table we can see that the coefficient on overlap is positive and statistically significant. This is in line with what we supposed in section 3. As there are more chances of being reciprocal
23
and therefore reducing controlling costs, Firm A will want to take advantage of that and set up partnerships. As regard size of the candidate firm and its physical distance, we see that coefficients suggest that smaller and more distant firms are more likely to be chosen as partners, even though the coefficients are not statistically different from 0. Moving on to Revenue coefficient, a positive value suggests that more profitable areas are more likely to be assigned under partnership, but again, the statistical significance is low here. Lastly, if we consider the nominal amount of shops, we see that the coefficient is also positive: firm would prefer to settle up partnerships in high client density areas. Even though only the coefficient for Shops (the sum of Restaurants and Supermarkets) is statistically different from 0 in more than 90% of the cases, the confidence level for the coefficient of Restaurants and Supermarkets are at 13.5% and 19.3% respectively. The lack of significance in this case may be due to the limited amount of data: only 82 observations are available for non integrated provinces.
Discussion
Only few of our hypotheses in Section 2 and 3 have been confirmed by a proper confidence level by the results of the experiment. Apparently, for the firm in analysis, the frequency of the transaction, measured by the amount of restaurants, supermarkets and the sum of the two, does not influence the integration decision process. Coefficients do not support the theory stating that a higher frequency leads to higher probability of integration even if we discard significance: some of them are positive, while some others are negative.
The coefficient for revenues, on the other hand, behaves like expected, even with a no statistical significance. The higher the profitability of the province, the less likely it will be disintegrated, as predicted by literature on size of franchising outlets reviewed in section 2.
The coefficient for cost of control, here measured by the distance of the province to the headquarters, is on a complete different page. Not only it is positive, entailing that integration will occurs more often for areas closer to the HQs, but it is also statistical significant. This result confirms the hypothesis previously made and aligns this study with the main literature regarding costs of monitoring.
As regard the form of disintegration, in this case subjective just for the firm in analysis, we find once again that the cost of monitoring plays a role, even if in a different way. It is much more important for the firm to be able to reciprocate partners’ behaviour than where they are located, as the latter coefficient is not significant. This probably means that having the chance to be reciprocal
24
takes away from the partners a lot of incentives to behave unfairly. In light of what we found, to get rid of the same amount of moral hazard, we expect a lot more has to be spent on traditional ways of monitoring.
As concern the number of restaurants or supermarkets, results seem in line with the predictions: Firm A will trust more partners, which are easier to control, using them in provinces with higher revenue shares.
Size coefficient seems to suggest that smaller firm will be preferred as partner, maybe because it is easier to deal with them. If a big firm is taking care of an area, Firm A may prefer to just outsource to them, as it may be too complicated and expensive to bargain with it. However, results for this specific study do not show any level of significance.
From this study we can notice how the main driver for Firm A’s choices is strictly cost of monitoring.
25 5. Limitations
In this section limitations are acknowledged. Despite taking any possible measure to prevent low validity, the structure of the research itself carries intrinsic biases that might affect it.
Internal validity
This case study is based on field data. The data has been collected directly from a firm; however there might not be any actual logical correlation among the measures. Even data that seems supportive of theoretical hypotheses may be due to chance or other hypotheses not mentioned in this thesis. This is due to the intrinsic nature of field data, which entails the lack of control on causality that an experiment would grant.
External validity
Field data is usually regarded as a way to reach high standards of external validity. On the other hand, this thesis uses a dataset coming from just one firm with a limited number of observations (101 provinces). Despite this amount being enough to retain a statistical significance, these results may apply differently in other sectors or in other geographical areas, where market situations and culture and traditions may lead to different outcomes.
Robustness
Some of the data that may influence the integration decision may have not been taken into consideration as not available. This may cause endogeneity due to lack of important regressors. E.g., income per capita at provincial level may have indicated which provinces are the richest and which are the poorest, and therefore showing that firms have more incentive to invest in a richer area as business opportunities will outnumber the ones in a poorer area. However, such piece of information was not available at provincial level and therefore has not been taken into consideration.
Endogeneity may also be caused by reverse causality, especially in Regression 2. It may be the case that the number of overlaps has grown after a business became partner. I.e., the business acquired clients in Firm A’s territory after assuring itself a partnership contract. In order to assess for this, we would need a panel data with the development of overlaps and partners through the years.
Unfortunately these documents were not available.
26
It was thought initially to use a nested logit model or a multiple choice logit model. Unfortunately these two types of regression were not performing properly, as the complexity of the data would make them encounter often flat region in the maximisation of the likelihood when entering more than a couple regressors in the model. In order to avoid that, we used two separate logit, even if the condition of independence from irrelevant alternatives (IIA) may not be satisfied and thus alter the standard errors.
27 6. Conclusions
This thesis reviews the previous literature based on drivers of integration. More specifically, it focuses on frequency, costs of monitoring and size of business unit to integrate as variables taken into consideration in the decision process. The approach is not the traditional one used in franchising studies, where the decision is made on a single shop, but it takes to a geographical perspective. This means that the strategy is set not for one shop, but for an entire area. The decision taken into this new framework are essential for all the service companies that may not have shops open to public and just render services door to door.
The theories reviewed are then applied to a case study. This case study is based on an Italian small-medium enterprise operating in the ITC market, which had to make this decision recently. Specifically, it had to decide how grant support over the entire Italian territory after winning some contracts. The decision of the go to market strategy was made up of two phases. In the first one, it was decided whether the province should be integrated in the areas covered directly or not. If not, a type of relationship between partnership and outsourcing was chosen for the firm operating locally. Despite fully supporting theories regarding costs of monitoring, there is no evidence that frequency and size of business unit have been taken into consideration or played a relevant role in the first stage of process. Similar results were obtained for the second stage, where frequency (proxied with the total number of shops) may have a role, beside the strong importance of overlaps (used as proxy for lower monitoring costs). Lack of statistical significance may be also due to a low number of observations or be related to specific action taken by the firm in consideration.
In order to obtain better results and with more statistical meaning, other firm that faced the same problem of the one taken as reference in the experiment should be surveyed. A larger amount of data with possible differences in sectors and practices may enhance the significance of the results.
28 References
Brickley, J. & Dark, F. (1987).The choice of organizational form The case of franchising, Journal
of Financial Economics, 18(2), 401-420.
Fehr, E., & Schmidt, K. (1999). A Theory of Fairness, Competition, and Cooperation. The
Quarterly Journal of Economics, 114(3), 817-868.
Fehr, E., Kirchsteiger, G. & Riedl, A. (1998). Gift exchange and reciprocity in competitive experimental markets, European Economic Review, 42(1), 1-34.
Fehr, E. & Schmidt, K. (2005). The Economics of Fairness, Reciprocity and Altruism - Experimental Evidence and New Theories, Munich Discussion Paper No. 2005-20.
Forbes, S., & Lederman, M. (2010). Does vertical integration affect firm performance? Evidence from the airline industry. The RAND Journal of Economics, 41(4), 765-790.
Goldberg, V. (1976). Regulation and Administered Contracts. The Bell Journal of Economics, 7(2), 426-448.
Hart, O. (1995). Firms, Contracts, and Financial Structure. Oxford University Press
Holmstrom, B., & Milgrom, P. (1994). The Firm as an Incentive System. The American Economic
Review, 84(4), 972-991.
Joskow, P. (1985). Long Term Vertical Relationships and the Study of Industrial Organization and Government Regulation. Zeitschrift Für Die Gesamte Staatswissenschaft / Journal of Institutional
and Theoretical Economics, 141(4), 586-593.
Lafontaine, F., & Shaw, K. (2005). Targeting Managerial Control: Evidence from Franchising. The
RAND Journal of Economics,36(1), 131-150.
Lafontaine, F., & Slade, M. (2007). Vertical Integration and Firm Boundaries: The Evidence. Journal of Economic Literature, 45(3), 629-685.
Martin, R. (1988). Franchising and Risk Management. The American Economic Review, 78(5), 954-968.
Masten, S. (1984). The Organization of Production: Evidence from the Aerospace Industry. The
Journal of Law & Economics, 27(2), 403-417.
Masten, S., Meehan, J. & Snyder,. (1989). Vertical integration in the U.S. auto industry : A note on the influence of transaction specific assets, Journal of Economic Behavior & Organization, 12(2), 265-273.
Minkler, A. (1990). An empirical analysis of a firm's decision to franchise, Economics
29
Monteverde, K., & Teece, D. (1982). Supplier Switching Costs and Vertical Integration in the Automobile Industry. The Bell Journal of Economics, 13(1), 206-213.
Norton, S. (1988). An Empirical Look at Franchising as an Organizational Form. The Journal of
Business, 61(2), 197-218.
Williamson, O. (1971). The Vertical Integration of Production: Market Failure Considerations. The
American Economic Review,61(2), 112-123.
Williamson, O. (1975), Markets and Hierarchies: Analysis and Antitrust Implications, Free Press, New York.
Williamson, O. (1981). The Economics of Organization: The Transaction Cost Approach. American
Journal of Sociology, 87(3), 548-577.
Williamson, O. (1985), The Economic Institutions of Capitalism, Free Press, New York.
Woodruff, C. (2002). Non-contractible investments and vertical integration in the Mexican footwear industry, International Journal of Industrial Organization, 20(8), 1197-1224.
