The Starbucks Impact on the Real Estate Market is Being Modeled

The aim of this research paper is to determine if Starbucks coffee shops have an impact on the selling price of condominiums in North York, Toronto. The issue is that as Starbucks expands, land and even houses gain in value as long as they are in the same area as the coffee shop. The Starbucks influence is what it’s called. According to the Starbucks impact principle, the mere sight of the distinctive green Starbucks sign is enough to guarantee that your local prices will rise! The concern is whether this is justifiable, and if not, how far has it gone. In this paper, we are going to use regression analysis models as our main analysis tool. Stepwise multiple regressions may be used together with residual analysis to get the impact of the models as well as their complexities. The impact of Each Variable will be studied using correlation analysis.

Introduction

Starbucks Corporation is an American coffee company with a chain of coffee shops all over America. Starbucks was founded in Seattle, Washington DC in the year 1971. It has expanded to other regions of the world; almost 23,768 locations. Starbucks is able to distinguish itself from other big coffee companies through their quality, taste and customer experience. They use up to date coffee making machines which ensure the good quality of the coffee is consistent. They have top-notch customer service and provide coffee both hot and cold. It has gone ahead to introduce fresh juices and even pastries to broaden their customer network (Garthwaite et al., 2017).

In 1984, the original owner of Starbucks decided to buy Peet’s. Coffee business in America at that time was not doing so well but coffee specialty was interestingly doing well. They made profit and decided to expand while still in Seattle. They opened other five shops totaling up to 6 shops in Seattle. Jerry Baldwin, the original owner of Starbucks sold the enterprise to Howard Schultz who rebranded his coffee outlets to Starbucks. He began to expand too till he was able to open a new shop outside Seattle. He opened the new outlet in Waterfront Station in Vancouver. By 1989, Howard had opened a total of 46 stores in the North and Mid-West of America. Starbucks was able to make it to the stock markets in 1992. The company’s market value rose making revenue of up to 3.5 million US dollars. Starbucks management was confident due to how their value was rising and was able to open more outlets. This subsequently led to the rise of their share, 100 times than it was. The company expanded keeping technology at its bay. They even developed a Starbucks app which was successful in getting more in-store purchases.

Starbucks has had its share of successes and failures too. Recently, Starbucks celebrated its fifteenth anniversary of Frappuccino line-up. Starbucks want to introduce blended yoghurt. Starbucks introducing yoghurt was a twist for the existing market. The yoghurt to be introduced will come in two flavors, banana and red berry. Starbucks chose a good timing for the launch since it was during summer. The success of Starbucks restaurants has been the good coffee and the efficient distribution system all over the United States. Its good business for coffee distributors since Starbucks has given coffee a new sachet. Starbucks has done this for all it’s coffee in all it’s shop and also in 80% of the coffee sold in supermarkets. Starbucks creativity is a gold mine in the dormant industry. Kraft can now sell new coffee brands Espresso, Master Blend, Colombian Extreme and Rich French Toast. This has made the coffee house to prosper and upgrade mail-order business. At the end of it all, the coffee sachet has had a big effect. Ten years ago, 3% of all coffee sold in the United States was at a premium price. Today, 40% of the coffee is sold at premium prices. The Starbuck effect is real and has been tracked on 39 categories of fast moving goods that have been measured as the percentage of products that have been sold at a premium. There is a lot of evidence on the Starbuck effect. This has been happening since whenever a company increases premiums based on a particular product, or to its delivery system, the entire category amounts to huge profits and high prices.

Yoghurt is another example. In the 1980’s, yoghurt lost its fame of being a healthy product. Dannon established different innovations on yoghurt making and even packaging, raising the price and the market share too. The profit margin has been rising by 5% from the year 1990 to 1997. The creativity was a huge investment for the yoghurt company and it paid off.

According to research on the Starbucks Effect, between 1997 and 2014, homes within a quarter mile from a Starbuck outlet has increased in value by 96% (Garthwaite et al., 2017).. This means that there is an undeniable correlation between Starbucks café location and the neighboring home appreciation. True properties tend to start being expensive once they are located near any Starbuck outlet. The properties are appreciating at a very fast pace than the U.S. housing laws allow. An average American home has now appreciated by 65% but a house next to a Starbuck store has appreciating by almost 96%. Homes and properties appreciate every day, so how do we know it has everything to do with Starbucks? Take a look at Dunkin’ Donuts, which is another prominent coffee outlet. Homes near Dunkin Donuts have had the same historical trend. The houses near Dunkin appreciate faster than United States housing laws but not as fast as Starbuck’s. According to Zullow, between 1997and 2012, homes near Dunkin’ Donuts appreciated by 80% while those near Starbucks appreciated by 96% almost doubling their value. The basic reason for this discovery was that people genuinely liked drinking coffee and saw Starbucks as a proxy for their gentrification, hence people paid huge premiums for their homes near Starbucks.

The major modeling problem is to check whether there is a Starbucks effect in North York, Toronto. There is a likelihood that the reason for hiking of land and houses prices is because of the expansion of Starbucks to the area. Another objective is to ensure all variables in our data are statistically significant. If they will be statistically significant, then it will mean that the regression model results will be accurate. Regression analysis is our main data analysis method, hence will want clear results from them. Another objective it to prove that Starbucks effect has had an influence to the real estate pricing. Lastly, another objective in this study is to prove that all the other variables are statistically significant.

This type of information is very important. Real estate is a major contributor of revenue to the country. Thus, the results can help in implementation of better guidelines for Starbucks and the real estate sector. The findings from this report can be used by the government for planning and budgeting purposes. This is why the data needs to be valid and clean.

We are going to use an economic model for this paper. This type of model will mainly concentrate on the economic value of both sectors; real estate and Starbucks coffee company. Concerns about the financial forms will be included. Natural resource productions will be considered as well as the economic sustainability of the two ventures. We will define competitive advantage and try find out how each of the sectors compete with each other concerning the advantageous inputs they give (Conroy, Narwold, & Sandy, 2013).

Data Collection

In this study I am going to use secondary data. I chose secondary data since I could get accurate information. I chose data from realtor’s website because it is a top real estate company which is very competitive in the real estate market. Secondary data is supposed to be undertaken carefully and with due diligence. The method is cost effective since one does not have to go to the field. I got the data from their main website, hence was sure I got the right information. I also got to check the initial purpose of the data. The data was used in laying down strategies for their company hence was useful to them. If it was useful to the company, then I was sure it was credible and accurate. I also checked the date when the data was collected. I was looking for something collected over the last three months (Conroy, Narwold, & Sandy, 2013). This is because, determining the Starbuck’s effect, it was good to have recent and credible data. I also checked the numerical in the data. I ensured the aggregate data described a group of observations on a given criterion while a disaggregated data gave details on individuals or single entities.

Data limitations

One of the major limitations of using the secondary data was the fact that the data was not specifically designed for my research. Some variables I wanted in my research were missing and some variables which I did not need were in the data set. One is never 100% sure of the reliability and validity of the data set. It may not have been collected from the right sub-groups or required persons of interest. The data set I used was open, hence publicly available. This was a limitation since for the real estate company to maintain confidentiality; they had to delete identifying variables about the respondents, the names, location, specific age and maybe ethnicity. Another major limitation I encountered was I was unaware of the glitches that were there during data collection. A researcher understands his or her research better when he takes part in the data collection, since one can ask or want to know more on the issue being researched on.

Data splitting

Data splitting is the act of partitioning available data into two portions; in this case, I divided my data into data for the predictive model and the other to evaluate the model’s performance. This was important since I was creating models to be analyzed through regression analysis. This was important for the regression models I was going to create. One model was to implement regression analysis, the other was to summarize the estimates and be able to check the significance of the estimates. Splitting the data will enable me to make precise predictions.

Modeling techniques

I decided to use regression analysis as it is used to estimate relationships. It consists of techniques that are used for modeling and analyzing several variables. It is also used when one has focus on the dependent variable and one or more independent variables (Cheng & Phillips, 2014). Regression analysis is used as a statistical modeling tool to be able to decide which factors or variables matter most and which to ignore. It also helps us see the relationship between the variables and how well they can interact with each other. In regression analysis there is a regression line that is used to show if the data is linearly fit. The line best explains the relationship between the dependent and the independent variable. It also contains an error term. The error term helps us has to be there for the independent variable is never a perfect predictor of dependent variables. The error term is able to show us if the regression line is fit. If it is too big, then the line is not fit for the data.

In this study, the dependent variable is the price list while the independent variables are distance to nearest Starbuck, floor size, number of bedrooms, number of washrooms, distance to subway and maintenance cost. My aim is to find out the relationship of the dependent variable with each one of the independent variables. Even though there can be dangers including too many variables in regression analysis, the impact of multiple variables can be assessed at once. This method will help me understand if the distance from Starbucks affects the price of condos in North Toronto. The regression analysis will also help in determining the correlation between various variables and not the causation (Cheng & Phillips, 2014).

Model limitations

One of the limitations I encountered was getting a wrong analysis. This could be an outcome of using secondary data since regression analysis is quite sensitive. When the data is not correct, the results will not be correct hence the predictions. I had to ensure that my results were accurate by working on the regression to explain 90% of the relationship, which was quite high, considering this, was secondary data. Lastly, I made a mistake of my intuition lie above the data. At one time I wanted the data to fit my understanding and the facts that I knew.

Model assumptions

Normally in regression there are five major assumptions, which I was keen enough to apply in this study. The variables had a linear relationship, which means that the relationship between the dependent variable had to be linear with that of the independent variable. I was also keen to check on the outliers since regression analysis is sensitive to outliers.

Secondly, all the variables had to be multivariate normal. I checked this by checking the goodness of fit test of the whole data. Next I assumed that there was little or no multicollinearity. Multicollinearity occurs when the independent variables are no longer independent from each other. The data correlation matrix had to be 1. To achieve multicollinearity, the tolerance level of one independent variable would or could influence the other independent variables. The variance inflation factor had to be well defined for the multicollinearity to be achieved. Lastly, the condition index value indicated a value> 30 which meant a strong multicollinearity.

Fourthly I assumed that there was no or little autocorrelation in the data. My last assumption was that the linear regression was homoscedasticity. This in a simpler term was to check if the error terms along the regression were equal.

Data analysis and results

I chose excel since it is more flexible with any type of data. It is also flexible and one can use multiple of variables.

SUMMARY OUTPUT

Regression Statistics

Multiple R

0.771834

R Square

0.595727

Adjusted R Square

0.588757

Standard Error

159928.7

Observations

60

ANOVA

df

SS

MS

F

Significance F

Regression

1

2.19E+12

2.19E+12

85.46755

5.24E-13

Residual

58

1.48E+12

2.56E+10

Total

59

3.67E+12

Coefficients

Standard Error

t Stat

P-value

Lower 95%

Upper 95%

Lower 95.0%

Upper 95.0%

Intercept

16804.64

59278.38

0.283487

0.777813

-101854

135463.2

-101854

135463.2

Sq. Feet

694.1024

75.07977

9.244866

5.24E-13

543.8139

844.3909

543.8139

844.3909

SUMMARY OUTPUT

R

egression Statistics

Multiple R

0.583032

R Square

0.339927

Adjusted R Square

0.328546

Standard Error

204355

Observations

60

ANOVA

df

SS

MS

F

Significance F

Regression

1

1.25E+12

1.25E+12

29.86901

1.02E-06

Residual

58

2.42E+12

4.18E+10

Total

59

3.67E+12

Coefficients

Standard Error

t Stat

P-value

Lower 95%

Upper 95%

Lower 95.0%

Upper 95.0%

Intercept

81532.02

86283.56

0.944931

0.348614

-91183.3

254247.3

-91183.3

254247.3

Bedrooms

228293.8

41771.85

5.465255

1.02E-06

144678.4

311909.2

144678.4

311909.2

SUMMARY OUTPUT

Regression Statistics

Multiple R

0.537679

R Square

0.289099

Adjusted R Square

0.276842

Standard Error

212077.1

Observations

60

ANOVA

df

SS

MS

F

Significance F

Regression

1

1.06E+12

1.06E+12

23.58659

9.42E-06

Residual

58

2.61E+12

4.5E+10

Total

59

3.67E+12

Coefficients

Standard Error

t Stat

P-value

Lower 95%

Upper 95%

Lower 95.0%

Upper 95.0%

Intercept

177465.2

77678.79

2.284603

0.026013

21974.17

332956.1

21974.17

332956.1

Maintenance ($)

592.9702

122.0957

4.856603

9.42E-06

348.5691

837.3713

348.5691

837.3713

SUMMARY OUTPUT

Regression Statistics

Multiple R

0.554193

R Square

0.307129

Adjusted R Square

0.295183

Standard Error

209370.4

Observations

60

ANOVA

df

SS

MS

F

Significance F

Regression

1

1.13E+12

1.13E+12

25.70972

4.35E-06

Residual

58

2.54E+12

4.38E+10

Total

59

3.67E+12

Coefficients

Standard Error

t Stat

P-value

Lower 95%

Upper 95%

Lower 95.0%

Upper 95.0%

Intercept

160267.2

77861.53

2.058362

0.044057

4410.43

316124

4410.43

316124

Washrooms

226679.1

44705.7

5.070475

4.35E-06

137191

316167.3

137191

316167.3

I decided on concentrating on the numerical values of the data first. My dependent variable was the price while the other variables were independent.

The linear regression equation for the sq. feet is

Y=16804.64X + 694.1024

This is read as the price of the condos is equal to variable cost of $16804.64 times the number of square feet of the condos plus the fixed cost of $694. The coefficient of correlation of the variable is 0.77. This is a positive relatively strong relationship between the variable and the dependent variable. The R square tells the importance of the output. In our case the R square is 0.60. This means that the output variable’s variance is explained by the input variable variance by 60%. The adjusted R of this variable is 0.588 meaning it explains 59% accuracy of the regression equation. The F value of this variable on the regression equation is 0.000, meaning there is a 0.000% chance that the regression output was merely by chance occurrence. This means that the variable is suitable to be in the model.

The linear regression equation for the no. of bedrooms

Y=81532.02 X + 228293.8

This is read as the price of the condos is equal to variable cost of $81532.02 times the number of square feet of the condos plus the fixed cost of $22830. The coefficient of correlation of the variable is 0.58. This is a fair positive relationship between the variable and the dependent variable. The R square tells the importance of the output. In our case the R square is 0.33. This means that the output variable’s variance is explained by the input variable variance by 30%. The adjusted R of this variable is 0.32 meaning it explains 32% accuracy of the regression equation. The F value of this variable on the regression equation is 0.000, meaning there is a 0.000% chance that the regression output was merely by chance occurrence. This means that the variable is suitable to be in the model.

The linear regression equation for the no. of washrooms

Y=160267.2 X + 226679.1

This is read as the price of the condos is equal to variable cost of $160267.2 times the number of square feet of the condos plus the fixed cost of $226680. The coefficient of correlation of the variable is 0.55. This is a positive fair relationship between the variable and the dependent variable. The R square tells the importance of the output. In our case the R square is 0.30. This means that the output variable’s variance is explained by the input variable variance by 30%. The adjusted R of this variable is 0.29 meaning it explains 29% accuracy of the regression equation. The F value of this variable on the regression equation is 0.000, meaning there is a 0.000% chance that the regression output was merely by chance occurrence. This means that the variable is suitable to be in the model.

The linear regression equation for the maintenance cost

Y=177465.2 X+ 592.9702

This is read as the price of the condos is equal to variable cost of $177465.2 times the number of square feet of the condos plus the fixed cost of $592.9702. The coefficient of correlation of the variable is 0.53. This is a positive fair relationship between the variable and the dependent variable. The R square tells the importance of the output. In our case the R square is 0.28. This means that the output variable’s variance is explained by the input variable variance by 28%. The adjusted R of this variable is 0.27 meaning it explains 27% accuracy of the regression equation. The F value of this variable on the regression equation is 0.000, meaning there is a 0.000% chance that the regression output was merely by chance occurrence. This means that the variable is suitable to be in the model.

Conclusion

From the results, price is the only variable with a higher correlation, and a greater variability in terms of the values of R square and adjusted R. The other variables do not have strong correlation and hence cannot explain the variation in the models created. This means that price have a good correlation with the square feet of the condos, establishing a relationship between the price of the condos and Starbucks outlets. Starbucks indeed has an effect on housing in its neighboring environment.

References

Cheng, H. G., & Phillips, M. R. (2014). Secondary analysis of existing data: opportunities and implementation. Shanghai archives of psychiatry, 26(6), 371.

Conroy, S., Narwold, A., & Sandy, J. (2013). The value of a floor: valuing floor level in high-rise condominiums in San Diego. International Journal of Housing Markets and Analysis, 6(2), 197-208.

Garthwaite, C., Garthwaite, C., Busse, M., Busse, M., Brown, J., Brown, J., ... & Merkley, G. (2017). Starbucks: A Story of Growth. Kellogg School of Management Cases, 1-20.