The main purpose of the regulation aims to deter any franchiser without experience in business management from indiscreetly recruiting franchisees. If there is no restriction such as the regulation, any franchiser unskilled in supporting franchisees could even recruit them. In this case, franchisees are highly likely to be in difficulty of operating their stores and be faced with financial bankruptcy. Accordingly, the regulation imposes any franchiser on the mandatory obligation to operate business management for a certain period of time. Once a franchiser experiences the business management from operating any direct retail store, the franchiser also could build up the ability to support franchisees, which reduces the possibility that the franchisees would be in trouble with the financial bankruptcy.

Moreover, the franchise market seems to be considered highly saturated in the South Korea. According to the market analysis2 about the domestic franchise business done by Ministry of Trade, Industry and Energy of South Korea in 2018, the number of franchisers registered in South Korea was 3.5 times as many as Japan, while the population in South Korea was less than the half of the whole Japanese people. Therefore, franchisees confronting fierce competition were expected to gain nothing but negative profits due to market saturation in the franchise market of South Korea.

In spite of the severe market saturation, there have still been new entries in the franchise market of South Korea. The reasonable explanation behind this phenomenon is that a considerable portion of workforce has no choice but to run a business in the franchise market since they can not achieve new another jobs after losing jobs. More specifically, the Statistics Korea3 demonstrates that 54 percent of the self-employed workers were employees before becoming self-employed in Korea. In other words, once workers experience unemployment, more than half of those job seekers hardly succeed to get another jobs and some of them are forced to run their own business in the franchise market. Consequently, the self-employment rate in South Korea turns out to be the 7th highest of all OECD countries from 2015 to 2020 according to the investigation of OECD.4

Furthermore, when many jobless people are compelled to start a franchise business, they are forced to invest all their own money into initiating the franchise business since they can seldom borrow money from banks. According to a survey5 conducted by Hana Institute of Finance in 2018, located in South Korea, it was reported that around 69 percent of respondents surveyed invested their own money into starting a franchise business. Consequently, they are highly exposed to negative external shock. For instance, when a franchiser suffers from default risk or bankruptcy, franchisees of the franchiser are also likely to be easily broke since they scarcely have financial alternatives to handle the liquidity crises of the franchiser. In spite of that, franchisees with high default risk keep coming into the franchise market. Accordingly, the competition in the franchise market is not able to self-regulate this issue, and the KFTC insisted that the new regulation needs to be introduced to the Fair Transactions in the Franchise Business Act.

A literature provides the rationale behind the reguation as previously indicated. In other words, Lee [3] demonstrates that almost 60 percent of trademarks in the franchise market of South Korea depended soley on franchisees’ stores in 2016. He also points out the likelihood that risks such as the decrease in sales might be transferred to franchisees due to lack of franchisers’ experience or insight about the competitiveness of their products. At this respect, the new regulation appears to be effective in alleviating the issue brought about by the market saturation in the franchise market of South Korea. Particularly, the regulation is expected to refrain franchisers from imprudently recruiting franchisees and deter from new entries to the franchise market, reducing the problem relevant to the severe market saturation. Furthermore, franchisees would be well trained by franchisers with abundant experience to operate the franchise business. Accordingly, profits of franchisees would be enhanced and they are less likely to be in the financial difficulties. Based on this hypothesis, this paper empirically detects whether franchisers to operate their own stores are different from those without any their own stores. If they are distinctive, franchisees of the former are likely more profitable than those of the latter. Then, the regulation of the KFTC is justified by this intrinsic distinction.

Meanwhile, there are two categories of economic literatures relevant to the franchise market; territorial encroachment and franchiser-franchisee relationship. First of all, as regards territorial encroachment, Kalnins [4] empirically proves that the territorial encroachment occurred from the Texas lodging industry in the 1990s. He concludes that a franchisee’s new entry nearby incumbents cannibalizes revenues of incumbents with the same brand. His findings are the first study to detect the existence of systematic cannibalization in the franchise industry specific to the lodging industry requiring lots of initial investment. Unlike his study, this study focuses on the business sectors (such as pizza, convenient stores, coffee, and confectionery and bakery), requesting relatively small investment, compared to the lodging industry.

Next, as regards franchiser-franchisee relationship, the economic literature provides perspective on principal-agent. For instance, Blair and Kaserman [5] investigate the relative importance of entry (or franchise) fee and royalty that a franchiser imposes on a franchisee. According to their research, when a franchiser is faced with high demand uncertainty, the franchiser puts more weight on the royalty. However, the entry fee is placed more emphasis on in the case of low demand uncertainty since a franchiser likely harvests rents from a franchisee at the initial start-up period.

Besides, Mathewson and Winter [6] research the effect of contractual elements on franchisers and franchisees. They theoretically show that the royalty fee is more standardized when the low variances of local quality that franchisees supply are observed. They also demonstrate that a franchiser imposes the standardized retail prices and store hours on franchisees since the franchiser needs to prevent franchisees from misreporting local demand as low and reducing efforts to keep the high level of local quality.

Additionally, Michael [7] also holds an enquiry about the inherent problem of organizational form in franchising. His empirical findings articulate that the franchise contract gives rise to free-riding and decreases quality in the nature of franchising. As previously mentioned, based on the regulation by the lawmakers and the KFTC, it is inferred that franchisers with their own stores (direct retail stores) are better at operating the franchise business than franchisers only recruiting franchisees. For instance, once the former operates the direct retail stores over a certain period of time, tacit knowledge could be built up to save cost in an efficient way. Therefore, the former hands down the tacit knowledge to franchisees and the likelihood to make profit also increases.

This study also deals with the franchise market. However, unlike the previous literature mentioned earlier, this study focuses on the regulation of a competition authority when there are excess entries (almost surely causing the market saturation) and fierce competition in the franchise market. There exist two effects due to the new regulation. The one is that the new entries would reduce since franchisers are not able to recruit franchisees as fast as in the past ahead of the regulation. The other is the likelihood that franchisees go bankrupt would also diminish. Especially, it is inferred from the second effect that franchisers with their own stores (direct retail stores) are better at operating the franchise business than franchisers only recruiting franchisees. For instance, once the former operates the direct retail stores over a certain period of time, tacit knowledge could be built up to save cost in an efficient way. Therefore, the former hands down the tacit knowledge to franchisees and the likelihood to make profit also increases.

Accordingly, the regulation of the KFTC implies that there could be distinction between franchisers with or without their direct retail stores. The former is more likely to accumulate either tangible or intangible assets to save the production cost and reduce financial risk for franchisees, compared to the latter. Thus, franchisees of the former are more likely to be profitable rather than the latter. This study does not try to estimate the effect and validity of the new regulation. Instead, it tries to test if there exists the distinction between franchisers with or without, providing the rationale behind introduction of the regulation. That is, it is to detect the reason to introduce the new regulation. If it is not confirmed that there exists the distinction between the two groups, the introduction of the regulation would be questionable and doubtful. Therefore, once it is revealed that this hypothesis is valid, the introduction of the regulation by the KFTC is considered to be justfiable. In this regard, the former would be compared with the latter in terms of profit per franchisee.

Additionally, this study contributes to the economic literature by adding the simple method to estimate cost of a franchisee. More specifically, this study provides a simple instance to use the lump-sum (franchise) fees and operating cost per franchisee across all trademarks to estimate the average operating cost per franchisee of each trademark for each franchise business sector. This simple method is expected to be of practical use since it does not require heavy amount of data relevant to characteristics of cost.

This study is organized as follows. In Section 2, a simple model is introduced to estimate the operating cost per franchisee for each trademark of a franchiser. Then, profit is calculated from the estimated operating cost per franchisee and the sales amount per franchisee for each trademark. Finally, the distributions of the estimated profit are derived according to two groups, franchisers with franchiser owned stores and without them, and it is examined if there exists the difference between the mean values of those distributions. In Section 3, it is estimated how the operation of franchiser owned stores (direct retail stores) impacts sales per franchisee through cost savings. Finally, a conclusion is made in Section 4.

A two-period model6 is depicted as follows. In a small territory, there are two franchisers and two applicants to be franchiees. The two franchisers 1, 2 at the upstream stage occupy two different trademarks for an identical product in the market, respectively. At the downstream stage, each applicant buys a right from one franchiser in the period one and becomes a franchisee. The right indicates that the franchisee is able to produce and sell the identical product in the period two. In compensation for the right, the franchisee pays a lump-sum (franchise) fee7 a franchiser in the period one. The lump-sum fee8 equals $F_i$, $i=1,2$ for each franchiser.

Besides, franchisees own distinctively different preferences over trademarks. Thus, each franchisee contracts with a different franchiser. Moreover, a franchisee pays a per-unit output royalty9 which equals $R_i$, $i=1,2$ when the franchisee produces and sells the product in the period two. Each per-unit output cost equals $c_i$, $i=1,2$, and $c_1>c_2$.10 $c_i$, $i=1,2$ is the common knowledge to all participants.

$e^u$ and $e^d$ are implicit discount factors of franchisers and franchisees, respectively. Since a franchisee is assumed to be financially more conservative than a franchiser, $1>e^u>e^d$.11

In the period one, a franchiser would expect the present value of his profit as follows.

${\pi }_i^u=F_i+e^u{R}_i{Q}_i\left(R_i\right)$, $i=1,2$12

where $Q_i\left(R_i\right)$ is the output of a franchisee and the function of the output royalty, $R_i$.

The present value of a franchisee’s profit would be expected as follows.

${\pi }_i^d=e^d[P\left[Q_m\right]Q_i\left(R_i\right)-c_i{Q}_i\left(R_i\right)-R_i{Q}_i\left(R_i\right)]-F_i$, $i=1,2$13

where $c_i{Q}_i\left(R_i\right)$ is the total production cost for a franchisee, and $P\left[Q_m\right]$ is an inverse demand.

where $Q_m$ is the total production output in the whole market.

where $P\left[Q_m\right]$ is assumed to be a linear inverse demand for the product, and $a>0$, $b>0$.

Then, a franchisee’s profit at the time of contract is written as follows.

Since $c_i$ is the common knowledge, franchisers determine the optimal $R_i$ and $F_i$ on the condition that a franchisee earns at most normal profit14 at the time of the contract.

Thus, a franchisee would accept $R_i$ and $F_i$ taken as given and maximize ${\pi }_i^d$ over $Q_i$.

$\frac{\partial {\pi }_i^d}{\partial Q_i}=a-2b{Q}_i-b{Q}_j-c_i-R_i=0$, $i\ne j,i=1,2,j=1,2$

Then, the optimal output of the product is designated as the following manner.

$Q_i^{*}=\frac{1}{3b}(a+c_j+R_j-2c_i-2R_i)$15

Finally, a franchiser’s optimization problem is characterized as the follows.

$Ma{x}_{F_i,R_i}{\pi }_i^{u}s.t.{\pi }_i^d\left(Q_i^{*}\right)=0$16

where ${\pi }_i^d\left(Q_i^{*}\right)=0$ implies the normal profit of a franchisee at the optimal output.

Therefore, the optimal franchise fee, $F_i$ and output royalty, $R_i$ are solved as follow way.

The assumption, $c_1>c_2$, implies that a franchisee of a franchiser 2 produces the final product in relatively more cost-saving manner, compared to a franchisee of a franchiser 1. Thus, the franchisee of the franchiser 2 could make excess profit. In response, the franchiser 2 likely reaps the excess profit through the higher franchise fee, $F_2$ and output royalty $R_2$, which makes it only possible for the franchisee to gain normal profit.

Meanwhile, the franchise fee, $F_i$ is rewritten as follows.

where ${\beta }_0=(10e^{2u}-6e^{u+d}){\left(\frac{e^u}{9bH}\right)}^2$, ${\beta }_1=(14e^{2u}-6e^{u+d}){\left(\frac{e^u}{9bH}\right)}^2$, ${\beta }_2=4e^{2u}{\left(\frac{e^u}{9bH}\right)}^2$

Taking the square root of both sides, the equation is rearranged as follows.

When $F_i>F_j$, the difference between $\sqrt{F_i}$ and $\sqrt{F_j}$ is taken as the following way.

$\sqrt{F_i}-\sqrt{F_j}=({\beta }_1+{\beta }_2)(c_j-c_i)$, where $F_i>F_j$ and $c_i<c_j$

Then, the equation is provided to derive the cost, $c_j$ of a franchisee, j from the assumed relationship between two franchise fees, $F_i$, $F_j$ as.;

In the equation described earlier, once $\sqrt{F_i}$, $\sqrt{F_j}$, $\beta ={\beta }_1+{\beta }_2$, and $c_i$ are determined, $c_j$ is derived. In a small territory, two trademarks of franchises are supposed to be randomly drawn from the pool, $1,2,\cdots ,i,j,\cdots ,k$ in the same business sector (business category). Thus, they are assumed to compete against each other between adjacent points. Furthermore, it is assumed that $c_{min}=c_1<\cdots <c_i<c_j<\cdots <c_k$. Then, it is valid that $F_{max}=F_1>\cdots >F_i>F_j>\cdots >F_k$ from the relationship17 $F_i>F_j$ and $c_i<c_j$ in the same business sector. Accordingly, $F_{max}=F_1>\cdots >F_i>F_j>\cdots >F_k$ corresponds to $c_{min}=c_1<\cdots <c_i<c_j<\cdots <c_k$. Moreover, if $F_{max}$ and $F_i$ are observed, and $c_{min}$ and $\beta$ are estimated, the corresponding $c_i$ is finally derived. Thus, above all, the information on $F_i$ needs to be collected .

Fortunately, the information on each $F_i$ is available from the Franchise Disclosure Document18 recorded and managed by the Korea Fair Trade Mediation Agency, called the KOFAIR. The KOFAIR announces the Franchise Disclosure Document every year. More specifically, either a current franchiser or any applicant who wants to be a franchiser and start the franchise business fills out the Franchise Disclosure Document and submits it to the KOFAIR. The Franchise Disclosure Document contains various information on trademarks, the number of franchisees’s stores per trademark, sales amount per franchisee for each trademark19, and franchise fee (entry fee, $F_i$) per franchisee for each trademark20, etc. In other words, the Franchise Disclosure Document shows the business performances of both franchiser and franchisee. Whenever anyone wants to be a franchisee for certain trademark, the applicant needs information on the business performance of a franchise chain store on average. Based on the information, average business performances would be compared across all trademarks. Then, when applying for the specific trademark and confirming the information, the applicant expects on average how much earnings would be in the future. Finally, the applicant chooses to operate a franchise chain store for the specific trademark. Therefore, the Franchise Disclosure Document provides the information on the average business performance of a franchise chain store, corresponding to a specific trademark.

Once the information on $F_i$ is available, a few more steps to estimate $c_{min}$ and $\beta$ are taken as follows. In order to do this, $F_{max}$ and $F_{median}$ are taken, where the subscript, ’median’ means the median value of $F_1,\cdots ,F_i,F_j,\cdots ,F_k$. Then, $F_{median}$ corresponds to $c_{median}$ since it is assumed that $F_{max}=F_1>\cdots >F_i>F_j>\cdots >F_k$ corresponds to $c_{min}=c_1<\cdots <c_i<c_j<\cdots <c_k$.

However, $c_{median}$ is not directly available from the Franchise Disclosure Document in each business sector. Instead, the information about the operating cost per franchisee for each business sector is available from the Korean Statistical Information Service, called the KSIS. Thus, $c_{median}$ is assumed to be the operating cost per franchisee21 across all trademarks for a business sector.

Once $F_{max}$, $F_{median}$ and $c_{median}$ are collected, $c_{min}$ and $\beta$ are estimated as the follows.

At first, let $({\beta }_1+{\beta }_2)=\beta$, where $\beta >0$.

Then, it is kept that $\frac{{\sqrt{F}}_{max}-{\sqrt{F}}_{median}}{{\beta }_1+{\beta }_2(=\beta )}+c_{min}=c_{median}$.

$c_{min}$ is initialized at a value, $c_{min}^0$.

${\beta }^0$ is estimated in the way through the regression of ${\sqrt{F}}_{max}-{\sqrt{F}}_{median}$ on $c_{median}-c_{min}^0$, given $F_{max}$, $F_{median}$, $c_{median}$, and $c_{min}^0$.

Next, $c_{min}^1$ is calculated, given $F_{max}$, $F_{median}$, $c_{median}$, and ${\beta }^0$.

Then, ${\beta }^1$ is calculated, based on $F_{max}$, $F_{median}$, $c_{median}$, and $c_{min}^1$.

In the same way, ${\beta }^0,{\beta }^1\cdots {\beta }^n$ are estimated until ${\beta }^n$ converges to $\beta$.22

Once $c_{min}$ and $\beta$ are estimated, ${\sqrt{F}}_{median}$ and $c_{median}$ are replaced by ${\sqrt{F}}_i$ and $c_i$, which results in the equation, $\frac{{\sqrt{F}}_{max}-{\sqrt{F}}_i}{\beta }+c_{min}=c_i$. Now, since ${\sqrt{F}}_{max}$, $c_{min}$, $\beta$ and ${\sqrt{F}}_i$ are known, $c_i$ is derived.

As mentioned above, $\left(F_{max},c_{min}\right)$ and $\left(F_i,c_i\right)$ are inserted into the equation as follows.

Besides, $c_{min}$ and $c_i$ indicate the marginal costs in the equation. Since the marginal cost could be derived from the total cost, the equation (2.1) is transformed into the equation (2.2) as follows.

Once a few steps taken below, I reach the equation (2.3), where $\left(F_{max},c_{min},Q_{min},p\right)$ and $\left(F_i,c_i,Q_i,p\right)$.

${\sqrt{F}}_{max}-{\sqrt{F}}_i=\frac{\beta }{Q_{min}}(\frac{p{Q}_{min}}{p{Q}_i}T{C}_i-T{C}_{min})$

In the equation (2.3), p implies the market price for a product in the equilibrium. In order to analyze the equation (2.3), relevant data is collected from the Franchise Disclosure Document as mentioned earlier. Some business categories (business sectors), where the similarity of the product seems to be high, are selected. Before launching the analysis, the sales, $\left(S=pQ\right)$ data is adopted except the case such that the sales amount per franchisee is recorded as either zero or a missing value. The business categories chosen are articulated as in Table 1. The data from 2018 till 2021 consists of business sectors, trademarks in each business sector, total number of franchisees by trademark, and sales amount per franchisee by trademark, etc.

As previously stated, the data is also gathered from the KSIS, which is about the operating cost per franchisee according to each business category from 2018 to 2021. Although the operating cost per franchisee means the total cost, $TC$, the equation (2.3) is still available. In the next section, the operating cost per franchisee would be transformed into the marginal cost.

Moreover, as metioned before, the operating cost per franchisee from the KSIS needs to correspond to the marginal cost, $c_{median}$. For instance, if “A” trademark is a median within “Chicken” franchise in terms of the franchise fee, $F_i$ in 2018, $c_{median}$ for “Chicken” franchise is supposed to correspond to the operating cost per franchisee, $T{C}_{median}$ for “Chicken” franchise from the KSIS. Thus, it is also assumed that the “A” trademark takes the operating cost per franchisee, $T{C}_{median}$ for “Chicken” franchise from the KSIS. All estimated costs are not economic costs but accounting costs. The relevant information is articulated as shown in Table 2.

Then, the algorithm23 mentioned above is applied in order to derive $\tilde{\beta }$ of the business categories selected. Since the 6 business categories (business sectors) are chosen from 2018 to 2021, 24 data about $c_{median}$ corresponding to the 6 categories are available as in Table 2, and those 24 data are used to estimate $\tilde{\beta }$ and $c_{min}$. According to the application of the algorithm earlier, the result is shown in Table 3 and Table 4.

As previously stated, the sales per franchisee, $s_1,\cdots ,s_i,s_j,\cdots ,s_k$ corresponding to $F_1,\cdots ,F_i,F_j,\cdots ,F_k$ are also achieved from the Franchise Disclosure Document. Accordingly, the estimated profit per franchisee would be derived since $s_i$ is observed, and $T{C}_i$, responding to $c_i$, is estimated. Finally, the estimated profit per franchisee24 according to $F_1,\cdots ,F_i,F_j,\cdots ,F_k$ is also derived, where each $F_1,\cdots ,F_i,F_j,\cdots ,F_k$ responds to franchise fees of different trademarks within the same business sector. Thus, the mean values are shown in Table 5, which are about the estimated profit per franchisee for each trademark in the business categories. The mean values in “Convenient store” seem to be negative in 2020 and 2021. Since all the mean values seem to be positive except “Convenient store” franchise, performances of stores would not be in extreme situation on the whole.

Table 6, Table 7 and Table 8 show t-test results. For each year from 2018 to 2021, the t-test is used to check the differences in mean values between two groups with respect to the estimated profit, estimated operating cost, and sales amount. “Without” indicates “group without franchisers’ stores”, which implies that franchisers do not operate their own stores, and the counterparts, franchisees do not need to compete against their franchisers’ stores. Besides, “With” stands for “group with franchisers’ stores”, which means that franchisers run their own stores, and the counterparts, franchisees have no choice but to compete against either their franchisers’ stores.

According to Tables 6 and 7, the results state statistically sigificant mean differences across two groups, “Without” versus “With” in terms of the sales and operating cost.25 That is, the sales and estimated operating cost per franchisee in the group “With” tend to be higher than the group “Without” on average. Particularly, the estimated profit per franchisee in the group “With” has likely inclination to exceed the profit per franchisee in the group “Without” in spite of higher operating cost of the group “With”.

The subscript, i indicates a trademark, and the t implies year, 2018, 2019, 2020 and 2021. The data applied to Table 1 is used to esimate the equation (3.1), and it is ruled out that the sales value per franchisee according to each trademark is either zero or missing value. The dependent variable, $sales\_adj$ indicates the sales per franchisee for each trademark adjusted by CPI (Consumer Price Index). An independent variable, $mc$ implies the marginal cost per franchisee for each trademark. A few steps are taken in order to derive $sales\_adj$ and $mc$. First of all, since sales per franchisee is nominal value, it is switched to real value after the sales per franchisee is divided by CPI according to each year. Then, the estimated operating cost per franchisee, total cost, is also nominal value. Thus, the marginal cost is assumed to be the value in which the estimated operating cost per franchisee is divided by real value, the adjusted sales per franchisee.

Other independent variable, $du{m}_{it}$ is the dummy variable to equal one if franchiser owned stores (direct retail stores) for each trademark exist and zero, otherwise. Lastly, $log\left(mc\right)\times dum$ implies the interaction between the estimated marginal cost per a franchisee and the dummy variable. This variable is the main interest of this study since it measures the effect of a franchiser’s tangible or intangible assets on a franchisee’s sales amount through the operating marginal cost. Finally, the log transformation of some variables is taken. Table 11 shows the descriptive statistics about the sales and marginal operating cost, variables employed in the equation (3.1). As previously stated, the sales indicated in Table 10 is not nominal value but real value.

As expected earlier,28 ${\alpha }_1$ is expected to be negative since more cost-efficient franchisees are plausibly more productive in terms of the marginal cost. ${\alpha }_2$ is also likely expected to be negative. More specifically, any direct retail stores (franchiser owned stores) directly compete against stores of franchisees, if consumers are not disturbed to get access to either direct retail store or franchise chain at the same time. In this context, it is possible that the former directly and negatively influences the latter. Thus, it is concluded that ${\alpha }_2<0$. Moreover, ${\alpha }_3$ is expected to be positive. As mentioned earlier, either tangible or intangible assets of a franchiser with direct retail stores might induce cost savings of franchisees. At this respect, any direct retail stores might indirectly and positively affect franchise chains. Thus, the interaction might lead to cost savings of a franchisee, and the improved cost-efficient also expects the more productive, increasing the sales of a franchisee. Lastly. the random effect model is employed in order to estimate the equation (3.1) instead of the fixed effect model since the variation across all trademarks is assumed to be uncorrelated with the regressors included in the equation (3.1). Particularly, one regressor, marginal cost is more likely to be affected by efficiency of business management through franchisers and franchisees than characteristics of trademarks. Furthermore, another regressor, existence of direct retail stores is likely impacted by business size of franchisers. In sum, characteristics of franchisers and franchisees would have more effect on the two regressors than those of trademarks. Meanwhile, the result of the random effect model is compared to that of the regression model to show that the former model is addressing the endogeneity issue.

Table 12 shows the estimation result. The two regression models seem to show the endogeneity issue. Particularly, the coefficient of the marginal cost is not statistically significant in the regression model. Considering Table 6, Table 7 and Table 11, the more operating cost likely implies the more sales on average since the group, “With” is inclined to be above the group, “Without” with respect to the sales, operating cost, and marginal cost per franchisee. Thus, the sales and marginal cost per franchise would have positive relationship under the endogeity issue unsolved. Thus, once the endogeneity issue removed, the marginal cost negatively impacts the sales. This is the reason why the random effect model is employed. In the random effect models 1, 2, 1 percent increase in the marginal cost per franchisee leads to 1.25 ∼ 1.31 percent decrease in sales per franchisee, which demonstrates the negative causal relationship between the two variables. Accordingly, the more cost efficient a franchisee is, the more productive the franchisee would be.

Furthermore, one variable $log\left(\#rivalstores\right)$ is included in the random effect model 2 and regression model 2. It indicates the log transformation in the number of rival stores which is derived from the method to subtract the number of franchisees and franchiser owned stores by trademark from all franchise stores29 summed across trademarks within the same category of business. For instance, suppose that there are three trademarks A, B, and C in the case of Pizza, and there are only franchisees’ stores, 10 for A, 20 for B, 30 for C, respectively. Then, the number of rival franchise stores is 5030 for A, 40 for B, and 30 for C. As indicated in Table 12, 1 percent increase in the number of rival stores induces 0.05 percent decrease in sales per franchisee since each franchisee competes against other rival franchise stores depriving of consumers.

As expected earlier, the existence of franchisers’ stores negatively impacts the sales of franchisees. Despite such negativity, the result shows that the coefficient of the variable, interaction is positive, where the variable, interaction, implies the impact of either tangible or intangible assets of a franchiser on the sales per franchisee through the operating marginal cost per franchisee. Therefore, there might be a linked structure that those assets likely have impact on the estimated marginal operating cost per franchisee, which affects the sales of a franchisee in a cost-saving manner. Along the line, franchisers to run his own stores are more likely to help franchisees operate their stores in a cost-efficient manner. Accordingly, there could exist distinction between franchisers with or without their direct retail stores, which grants the rationale behind the introduction of the new regualtion by the KFTC.

However, since any direct retail stores of franchisers compete against franchisees’ stores, such a competitive relation put franchisees in an unfavorable position, plausibly reducing their sales amount. Even more, any direct retail stores appear to more deprive their franchisees of consumers than rival franchise stores, although the two variables have different units and coefficient of one variable is not easily compared with that of another variable. Furthermore, when the two variables, dummy variable and interaction considered, the effect of the dummy variable seems to exceed that of the interaction. Thus, the negative effect that the existence of direct retail stores from franchisers brings about, competing against franchisees and acting as a substitute for the franchisees, might be more dominant, compared to the positive influence that franchisers help franchisees through their tangible or intangible assets, which enhances sales of franchisees.