國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
22
VI. Results
Briefly speaking, the “demographic factors” model is almost consistent with the past thesis. Only Pct_Men shows a different sign compared to the literature. As for the
“distance to the nearest competitor” model, this paper finds an interesting result about the percentage of children, which is less mentioned by past literature. Finally, in the
“franchising” model, service variables are not significant enough to explain the ownership of outlets.
Ⅵ.1 Demographic factors
Results obtained under zero-inflated Poisson regression are given in Table 6. The Vuong statistic (4.57) indicates that zero-inflated model is quite suited. The significantly positive coefficient of Density variable indicates that fast-food outlets tend to be set up in densely populated areas. Also, the signs of the coefficient of TRA_rank and HRS_rank are positive and significant, which shows that transportation hubs have great effects on the entry decision of fast-food firms. The Pct_Men variable is significantly negative, which means an area with higher male population has relatively less fast-food outlets. This result is quite different from that of the literature. It might because of different preferences of people in Taiwan and the US. The Avg_Age variable is significant, showing a negative sign.
It indicates that if the area is averagely younger, there would be more fast-food outlets. The result is consistent with the literature, showing that fast-food is more popular with children and youths.
In addition, the Edu_year variable is significantly negative. Since the model has controlled the age and income factor, this variable can be considered the independent effect of educational level. The negative sign means that if people of specific district have had little schooling, there are more fast-food outlets in that district. Avg_Income and
‧
Attractions are control variables, making other independent variables explained better.
The Population in the inflation logit model is significantly negative, meaning that the population can explain the inflated zero observations well. In addition, the negative sign means that in those sparsely populated district, there are more likely no fast-food outlets.
Table 6 Zero-Inflated Poisson Estimation Results of Demographic Factors
Dependent Variable
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
24
Ⅵ.2 Distance to nearest competitors
The Results of the robust OLS estimation are shown in Table 7. By using Diff_Dis and Dis_kn dummies to control the effect of those outlets far from their competitors, the
coefficient of the market size variable Area_km2 is significantly positive. The result is consistent with the theory, proving that if the market is geographically large enough, MOS would prefer to keep away from McDonald’s to avoid competition. Also, locating at two opposite sides would create an environment similar to local monopoly, allowing both firms to gain higher profits.
The Dis_kn variable is used to control the fixed effects of competitors which are far from McDonald’s outlets. As the controlled distance becomes greater, the coefficient becomes greater. This means that when higher distance is controlled, the explanatory power of the Dis_kn variable becomes greater.
The HHI variable shows a positive sign but is less significant if the controlled distance is too long. HHI is considered an index measuring the degree of market concentration and competition. As a weaker firm, MOS tends to move further away from McDonald’s in the district with high HHI.
Finally, by controlling the educational variable, the Pct_Child and Pct_Elder variable can better explain the effect of age factor. The Pct_Child variable shows a significantly positive sign regardless of the controlled distance, meaning that in those districts with more children, MOS tend to keep away from McDonald’s.
‧
Table 7 Robust OLS Estimation Results of Distance to Nearest Competitor
MOS_Dist
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
26
Ⅵ.3 Franchising
Table 8 shows the results of the probit estimation of the franchising model. The coefficient of the demographic variable, such as Population, is significantly negative, which means that if the location of an outlet is sparsely populated, it costs much more for firms to monitor the agents, making the outlet more likely to be franchised.
As for the service dummies, the coefficients on the auxiliary service variables needs more unobservable effort to be significantly positive, which means that the more auxiliary services an outlet provides, the more likely it is to be franchised. On the other hand, those coefficients on other services that need more observable efforts should be negative.
However, according to my results, both variables are not significant. In fact, since over 90 percent (90.37%) of McDonald’s outlets provide wireless internet services, it may be the reason that the WIFI variable cannot be a good proxy for observable service.
The Drivethru variable is significantly positive, which is consistent with the past literature. The positive sign means that if an outlet provides drive-through service, it is likely that it would be franchised. Moreover, the marginal effect shows that if an outlet provides drive-through service, the possibility of this outlet to be franchised increases about 6 percent.
‧
Variable Coefficients Marginal Effects
Population －0.0025***
The final part will conclude and give some extensions and suggestions that can improve this research.