1. Your company is considering buying a coffee shop. You are curious about this store’s sales on a typical week-day. You should:

a. Look at the mean value of sales in week-days, because this is the value that happens most of the time

b. Look at the mean value of sales in week-days, if you worry about outliers

c. Look at the median value of sales in week-days, because it captures the most features of this variable’s distribution

d. Look at the standard deviation of the value of sales in week-days, if the mean of this variable is smaller than the mode

e. None of the above

Problem #1

With all the attention that “big data” has received recently, it is natural to ask whether there is evidence that using data improves an organization’s performance. Professor Kristina McElheran (from Rotman) and Professor Erik Brynjolfsson (from the Sloan School of Management at MIT) have recently begun a large-scale study of exactly this question. Their study is based on new data on management practices collected by the U.S. Census Bureau. The data come from a survey sent to approximately 50,000 manufacturing establishments in the U.S. in 2010.

The survey contains 36 multiple choice questions about the establishment. Two of the questions on the survey ask the establishments about their use of data in 2010.

Q27: “What best describes the availability of data to support decision making at this establishment?”

1 – Data to support decision making are not available

2 – A small amount of data to support decision making is available

3 – A moderate amount of data to support decision making is available

4 – A great deal of data to support decision making is available

5 – All the data we need to support decision making is available

Q28: “What best describes the use of data to support decision making at this establishment?”

1 – Decision making does not use data

2 – Decision making relies slightly on data

3 – Decision making relies moderately on data

4 – Decision making relies heavily on data

5 – Decision making relies entirely on data

In addition to these questions, the survey asks establishments about a range of management practices - such as, the use of key performance indicators, production targets and performance bonuses as well as the establishment’s practices for promotion and dismissal of managers.

The survey is combined with data on each establishment’s annual revenues, costs of materials and total employment. The researchers construct a variable called Value_Added which equals an establishment’s revenues minus its cost of materials and use this as their main measure of establishment performance.

The response rate for the survey was approximately 70% and the final sample of data contained responses from 30,000 establishments.

(a) The first thing the researchers wanted to know was what fraction of the establishments gave each of the different possible answers to the question about data availability (Q27). What type of table would they construct to determine this?

(b) The researchers then constructed a dummy variable called DDDM that equals one for establishments that gave a “4” or “5” to both Q27 and Q28 and zero otherwise. They used this variable to identify establishments that could be considered to practice “data-driven decision-making”. Using this variable, they made the following table to explore the relationship between data-driven decision-making and establishment performance:

DDDM                 Mean of Value_Added ($)

0                                     4,235,000

1                                     4,778,00

The values in the table are statistically different.

What can you conclude from this table (or what do the data say?)?

(c) Is the above table consistent with the claim that data-driven decision-making improves an organization’s performance? Explain.

(d) Offer an alternative explanation of the above table based on an omitted variables story.

(e) Based on the variables available to the researchers, explain how they could control for establishment size (in terms of number employees) to determine whether data-driven decision-making improves performance even for establishments of similar sizes (number of employees). Describe in words what they could do and produce an example of the type of table they could create (you can make up the numbers in the table).