Is the season of the year

relative to the retail sales

percentage of fast food
restaurants?
Math Studies Internal Assessment
Elaura Ligon

March 17, 2014

Introduction
Statement of Task
In the everyday lives of American society, the purchase of fast food has become a
common occurrence. This occurrence can be seen in many countries around the world as
well, however the true question is whether or not the success of these fast food restaurants is
the same all year round. Is there a season that is less successful? Or is there a season that
dominates over all of the rest? As a student with an interest in business and culinary arts, I
had personally grown curious of these very questions myself. Being that fast food and
consumerism plays such a major role in society today, the results of this analysis may
influence the sales of the food industry in the future. So predominantly, I would like to
answer the following questions:

1) Is there a relationship between the season of the year and the retail sales percentages
in certain fast food restaurants?
2) Do all three of the analyzed restaurants have similar data? Or does one reign over the
others?
In order to collect my data I emailed representatives from McDonalds, Wendys,
Burger Kings, and Arbys corporations and I gathered sets of data for each month of the year
dating back from December of 2007 to November of 2012 that told me the sales percentages
reached in each month and if the percentages were positive or negative. Something specific
that should be recognized about retail sales is that the percentages may or may not add up to
equal 100%; retail percentages represent the increase or decrease in sales over time that is
presented in the form a percentage value that is based on the initial set sales goal. Once the
data was gathered, I grouped each set into the four seasons of the year: Winter (December
through February), Spring (March through May), Summer (June through August), and
Autumn (September through November). Since there were three different sets of percentages
for each season of each year, I added together the percentages for each season of each year
from December of 2007 to November of 2012 which was a total of five years of data. Once
all three of the percentages were added together for each season, I in turn averaged together
those totals with the totals of the corresponding seasons for the five years of data collected.
Because I will be performing a Chi Squared test, I made sure that each average came out as a
whole number based on the information given and that none of the percentages were skewed
to get the results I was looking for. This process will be repeated three additional times, once
for each fast food restaurant named above.
Once all four of my sets of data are presentable and averaged together in a chart and
in a graph, they will be put into a matrix where I will begin my Chi Squared test. The nature
of a Chi Squared test is to be used with categories; in this case my categories are the seasons
of the year and the fast food restaurant being analyzed. It contains observed values and
expected values, which will be presented in my Mathematical Processes section during the
1

Chi Squared test, and it is also a test of independence rather than a test determining the
dependency of certain sets of data. However, based on the gathered data I hypothesize that
the retail sales percentage of fast food restaurants is dependent on the season of the year.

Information and Measurement

Below is a chart with the averaged percentages for each of the four fast food
restaurants being analyzed in this paper, as well as a stacked bar graph presenting the data
visually. The data presented in these charts and graphs will help me to determine how
significant the results of my Chi Squared test will be and will also assist me in answering my
second question about the similarities of sales percentages in these four fast food restaurants.
McDonalds
Wendys
Arbys
Burger Kings

Winter
23%
20%
18%
21%

Spring
17%
14%
13%
15%

Summer
20%
15%
16%
13%

Autumn
13%
14%
9%
12%

Average Retail Percentages

80%
70%
60%
50%
40%
Autumn
Percentages
30%

13%
14%

20%

12%
13%
Winter

Summer
17%

15%

Spring 16%

14%

13%

23%

20%

18%

21%

McDonald's

Wendy's

Arby's

Burger King

20%
10%

9%

15%

0%
Restaurant

Mathematical Processes and Calculations

In order to perform a Chi Squared test, you need a matrix of observed values which is
presented below. Column totals, row totals, and a table total can be found by adding up the
numbers in each row/column and then adding up those totals to find the table total. When you
compare the sum of the column totals and the sum of the row totals, they should both add up
to the same table total. Otherwise, there is an issue with the data within your matrix of
observed values.

Observed Values:

23
20
18
21

17
14
13
15

20 13
15 14
16 9
13 12

The row totals, column totals, and the table total have been presented in a table below
to complete the matrix for the expected values.

McDonalds
Wendys
Arbys
Burger King
Total:

Winter
23
20
18
21
82

Spring
17
14
13
15
59

Summer
20
15
16
13
64

Autumn
13
14
9
12
48

Total:
73
63
56
61
253

In order to find the expected values, you have to multiply the row total with the
column total and then divide that by the table total of the above matrix of observed values.
Once that is done, your matrix of expected values should have values that are greater than or
equal to 1 and contain no more than 20% that are less than 5. My expected values meet this
criterion, having none less than 5 and all expected values being greater than or equal to 1.

Expected Values:

7382
=23.66
253
6382
=20.42
253
5682
=18.15
253
6182
=19.77
253

7359
=17.02
253
6359
=14.69
253
5659
=13.06
253
6159
=14.23
253

7364
=18.47
253
6364
=15.94
253
5664
=14.17
253
6164
=15.43
253

7348
=13.85
253
6348
=11.95
253
5648
=10.62
253
6148
=11.57
253

Once the expected values are found, you need to determine Chi Squared, which is
determined by finding the sum of the observed value minus the expected value squared and
then it is divided by the expected value. Chi Squared is frequently used to either accept or
reject the null hypothesis which states that the factors are independent, and is represented by
the symbol 2. The formula used to express Chi Squared is below, in which fo represents the
observed value and fe represents the expected value.

( f of e )
fe

For this particular Chi Squared test, I will be testing at a 95% accuracy level with 9
degrees of freedom presented. In statistics, degrees of freedom represent the number of
values in the final statistical calculations that are able to be varied. In order to find the value
for 2 in this analysis, we would put together all of the values seen in the previous formula
that were found for the observed and expected values and find the sum.

2=

( 2323.66 )2 ( 1717.02 )2 ( 2018.47 )2 ( 1313.85 )2 ( 2020.42 )2 ( 1414.69 )2 ( 1515.94 )2 ( 1

+
+
+
+
+
+
+
23.66
17.02
18.47
13.85
20.42
14.69
15.94

Now that I have found 2, I need to determine how likely a 2 value of 1.6495 if the
difference in retail sales during each season is due to chance. The variable that will determine
that is defined as a p-value. In this 2 test, the p-value is .9959 which is a very high p-value in
reference to data significance.
Compared to a predetermined level of statistical significance called an alpha level ()
that is set at .05, you can determine that the p value is greater than . This proves the null
hypothesis to be true and concludes that the proportions presented are independent of each
other.

Interpretation of Results
According to my gathered data and the outcome of the averages and the stacked
column graph presented in my Information and Measurement section, there are visible
results that answer my previously stated questions. At a glance, one can see that the increase
in retail percentages often occurs in the winter months of December, January, and February
where in the autumn months of September, October, and November the retail sales percentage
increases are often the lowest. This indicates that the most sales occur in the winter compared
to every other season. Also, McDonalds has the highest overall increase compared to the
other 3 fast food restaurants selected, and Arbys has the lowest increase in this set of data as
well. Keeping in mind that the presented retail percentages for each season were added and
averaged over 5 years, the data may not be truly representative of the success of each
individual fast food restaurant. However, after completing the extensive analysis of my
gathered data and processing it through multiple mathematical processes, it can be concluded
that the retail sales percentage of fast food restaurants is in fact independent on the season of
the year. Though this is not the outcome I was expecting, it is at a 95% accuracy level and my
2 test of independence had presented a p-value too high to prove any significance of data.
4

Unfortunately, I do not know the outcome I would have gotten if I had done a 2 test on retail
sales in full service restaurants in comparison to the results I had gotten with the selected fast
food restaurants. Because of this I cannot fully determine if retail sales is completely
independent of the season of the year in the food industry overall.

Validity
Despite how accurate my collected data was and how thorough I was in completing
my mathematical processes to their utmost potential, I was unable to reach the conclusion I
was hoping for that stated that the retail sales percentages were dependent on the season of
the year. I had used a chart to add and average together the seasonal percentages over 5 years
of collected data between December 2007 to November of 2012 in order to have whole
percentages that were able to be processed in a 2 test of independence, which was then
presented in a stacked column graph to show a visual representation of my data not only
presenting the seasonal percentages independently but as an overall annual average for each
fast food restaurant. This type of graph was appropriate because it is important to view the
overall change in order to determine of one restaurant was more successful than another in
certain seasons or generally in the 5-year average. I had also used a 2 test of independence
because I was working with a set of categories and I needed it to show whether my two
factors, fast food restaurants and the season of the year, were at all dependent or independent
of each other and to what extent that information is significant. This was appropriate because
if the p-value of the 2 test had been less than the level, my data would have declined the
null hypothesis and proved that my two factors were actually dependent of each other.
However, as it stands my p-value of .9959 is far higher than the set of .05 and unfortunately
a p-value that high indicates little to no significance in the data collected.
Though my data was collected from the corporate base of each individual fast food
restaurant, there could be many reasons why I did not reach my hypothesized conclusion. For
example, the data could have been skewed by the data processing department of the fast food
restaurant itself in order to present higher retail sales for marketing purposes. And even if that
is not the case, it is possible that my data is simply just not significant; I can have an opinion
on a relationship between two factors, however that does not mean they have to correlate.
Just because data looks significant, my analysis has proved that what you see is not always
what it seems to be. And for me that is the true value of analysis; finding flaws in the data and
using it to see things that had previously gone unseen.
As unfortunate as it is that my data did not give me the results I was hoping for,
analysis is never fully complete until all factors have been evaluated. In the future when I
think of this topic again, I will go to greater depths to evaluate any and all factors involved
and I will complete my analysis with a more in-depth conclusion for my data. Though this
portion of the analysis is complete, there is still much work to be done in the future.