Today I graduated from Western Michigan University with a Ph.D. in Applied Economics. Here is the introduction to my dissertation which earned me that degree.

---

Adam Smith published An Inquiry into the Nature and Causes of the Wealth of Nations in 1776 at the dawn of what would become known as the Industrial Revolution. The Scottish philosopher was privy to the advances being made in various industries in the United Kingdom, such as steam-powered engines, textiles, crop rotation, and coal mining, among others. New technologies initiated, and then accelerated, rapid innovation across multiple sectors. Historians Lawrence McCaffrey and T. S. Ashton describe the Industrial Revolution beginning in about 1760 when “a wave of gadgets swept over England.” They continue,

"It was not only gadgets, however, but innovations of various kinds—in agriculture, transport, manufacture, trade, and finance—that surged up with a suddenness for which it is difficult to find a parallel at any other time or place." [1]

Indeed, the birth of economics as a discipline coincided with a new era of massive economic growth brought about by technological advances. Smith even used the backdrop of contemporary manufacturing to articulate many of his ideas, such as describing specialization of labor with a pin factory. He answers the titular question of his book by concluding that wealth does not lie in piles of gold but in the production and exchange of physical goods.

Today, we are in the midst of a similar economic revolution brought about by new technologies. It is now the Digital Age, which is characterized by the process of converting information into digital “ones” and “zeroes.” This has led to fundamental changes to the economy, and with that, changes to the context in which economic agents make decisions. The transition from the Industrial Age to the Digital Age began with advances in three key areas: big data, low-cost computing, and modern analytical methods. Eventually, these three areas converged to create an inflection point where development of digital technologies accelerated, having profound impact on the economy and the world. The characteristics, abilities, and implications of these advancements also provide through lines into the research discussed below. I consider the data, computing, and methods of the Digital Age in turn.

The first building block of the Digital Age is big data. Data in digital format came about by applying mathematical logic to the functioning of electrical relays. The process of digitization dramatically reduces the costs associated with processing, storing, and transmitting information. The economic theorist, Peter Drucker, in 1969 described data as becoming the primary resource for businesses and that the world was shifting toward a “knowledge economy” [2]. The prefix “big” in big data usually refers to data large in volume, variety, or velocity. High-volume data, meaning there are many observations, increased with the commercialization of database systems in the 1960’s. Now organizations around the globe store raw data in immense volumes. According to Statistica, there will be an estimated 74,000 exabytes of data created, captured, copied, or consumed in 2021 alone [3]. For reference, all human language ever spoken could fit within five exabytes of data [4]. Next, is the high variety of data. This can represent the number of variables within a dataset, the dimensions of a data model (e.g., an Uber ride that links to a data dimension about the driver profile), the variety of data types (e.g., tweets that include multimedia), or even metadata about the data collected (e.g., a mobile phone message that includes a timestamp and geographic location from when and where messages were sent.). Finally, high-velocity information is considered big data when the data is being collected, transmitted, or consumed at a high frequency [5]. Big data has even become so ubiquitous that the term is falling in popularity since its peak in 2014 [6] and more often is being referred to simply as “data.”

The second building block of the Digital Age is rise of computing power. This is best exemplified by Moore’s Law—the observation that every 18-24 months the number of transistors on a dense integrated circuit approximately doubles. The exponential growth of computing power means that costs for computing resources have likewise led to exponentially decreasing costs. The primary benefit of low-cost computing is how as a general-purpose technology it can build not just new inventions but new industries [7]. For example, big data collection and storage became economically feasible at a large scale because of innovations in computing. Mass-market mobile devices are also only possible because of low computing costs.

The third building block of the Digital Age is modern analytical methods. This includes modern statistical methods like machine learning and artificial intelligence as well as the means for handling and analyzing large datasets like parallel computation and big analytics engines such as Apache Spark [8]. In many use cases, machine learning approaches are improvements to traditional statistical and econometric analysis; however, many also require large amounts of data and computing resources to become useful. This reveals the complementarity with big data and low-cost computing power where large datasets become more valuable if they can be used to gain new insights, and machine learning methods are only feasible with the data, software, and hardware that can support computing-intensive algorithms. Although these modern methods are not expected to replace all areas of econometrics, like identification and causal inference, there is much to gain from incorporating machine learning and artificial intelligence to address economic problems of prediction or high-dimensional data [9]. Sundar Pichai, the CEO of Google, affirms the fundamental influence artificial intelligence can have on the world. In 2018 he said, “AI is probably the most important thing humanity has ever worked on. I think of it as something more profound than electricity or fire.” [10]

The three chapters in this dissertation apply microeconomic thought in the Digital Age. Each chapter leverages the challenges and opportunities that come from advances in big data, low-cost computing, or modern analytics to answer questions of economic relevance.

Chapter One [11] investigates how the new, digital technology behind Transportation Network Companies (TNCs) like Uber and Lyft impact the relationship between the quantity of car service trips and fuel prices in New York City. Specifically, I find that a single day 1% increase in fuel prices is associated with a 0.367 to 0.486% decrease in the number of TNC trips and a 0.033 to 0.088% increase in the number of taxi trips, even though the drivers under both systems pay for their own gasoline. Put in average terms, a one-day 1.64 cent increase in fuel prices will in New York City decrease the number of TNC trips by 1,266 to 1,677 and increase the number of taxi trips by 125 to 333. I attribute this difference to the flexible nature of TNCs and the fixed behavior of taxicabs. TNCs allow drivers to reduce their supply when operating costs are higher while taxi drivers are often restricted within a regularly scheduled shift due to higher city regulations during the period of study. Uber drivers who are unwilling to pay higher gasoline costs will leave the market, thus reducing competition for the taxi drivers who are stuck paying for the higher gasoline costs but benefit from more trips, in what I call a “rigidity dividend.” There are also nonlinear effects where TNCs, but not taxis, respond more to the first percent change in fuel prices than subsequent percent changes in fuel prices, indicating a diminishing marginal elasticity.

Chapter Two considers how social media influences the information sharing and decision-making processes for customers deciding to visit monopolistic competitive brick and mortar stores. This reflects the ways in which new media can have measurable, economic effects in the Digital Age. In the chapter, I use hierarchical linear regression to account for the random effects of brand- and store-heterogeneity and demonstrate this method is an improvement to ordinary linear regression. I find that online behavior influences offline store visits, especially for changes in the popularity, sentiment, and disagreement around a brand on social media. For example, when social media mentions of a brand increase one standard deviation either in per-like popularity, sentiment, or disagreement, then next-day foot traffic to stores of that monopolistic competitive brand will increase by 0.02 standards deviations. The manner social media activity precedes retail foot traffic is consistent with a causal pathway. This modest but meaningful effect, however, fully dissipates within one week. Additionally, the results are somewhat inconsistent when considering different weights of the social media variables.

Chapter Three considers how to incorporate a large dataset of demographic and economic geographic variables into business decision-making using mobile phone data and machine learning methods. The case study demonstrates how including zip code-level tax data reduces prediction error in models of customer visits to retail stores. I consider three approaches that also address the effects of pervasive multicollinearity. OLS with variable reduction though a model-dependent variance inflation factor threshold proves to be highly interpretable with poor prediction. Ridge and lasso regression show poor interpretability and good prediction. Random forest regression offers moderate interpretation through ranked feature importance but the best prediction overall. I find store-specific variables to be the most predictive with additional, valuable insights from economic and demographic variables relating to the population around a store. The results demonstrate the importance of economic geography in the store location decision for multiunit retail companies.

The Digital Age is not only defined by advances in big data, low-cost computing, and modern analytical methods; moreover, the convergence of these three areas accelerates the Digital Age and—like the Industrial Age before—pivots the economy into a new direction. The composition of the economy is also transformed. As of January 28, 2021, 27.9% of the S&P 500 is composed of firms classified primarily as “Information Technology,” [12] not to mention the many other firms that have heavily incorporated information technology. [13] Some markets have been highly disrupted by digital technologies, such as transportation and retail, while other markets only recently came into existence, such as social media or mobile smart phones. In one poignant example about the transition from an Industrial Age to a Digital Age, the author Andrew McAfee [14] reports how the United States economy has been growing in Gross Domestic Product, while at the same time decreasing the amount of raw materials used in production. Smith’s lesson from The Wealth of Nations is how there is merit in applying economic thought during times of broad economic change. This dissertation sheds some light into how the technologies of the Digital Age impact real-world business and consumer decision making.

References

1. McCaffrey, Lawrence J., and T. S. Ashton. 1964. “The Industrial Revolution, 1760–1830.” The Western Political Quarterly. https://doi.org/10.2307/445802 . Quote from page 42.
2. Issitt, Micah L. 2015. The Digital Age. Reference Shelf; v. 87, No. 4. Grey House Publishing.
3. Source: https://www.statista.com/statistics/871513/worldwide-data-created/ (accessed 1/31/21).
4. There are over one billion gigabytes in each exabyte, which is similar to the size difference between the Earth and the Sun. Source: https://www.backblaze.com/blog/what-is-an-exabyte/ (accessed 2/11/21).
5. Mendelevitch, Ofer, Casey Stella, and Douglas Eadline. 2016. Practical Data Science with Hadoop® and Spark: Designing and Building Effective Analytics at Scale. 1st ed. The Addison-Wesley Data and Analytics Series. Addison-Wesley Professional.
6. Source: https://trends.google.com/trends/explore?date=all&geo=US&q=big%20data (accessed 1/31/21).
7. Brynjolfsson, Erik, and Andrew McAfee. 2014. The Second Machine Age: Work, Progress, and Prosperity in a Time of Brilliant Technologies. W. W. Norton & Company.
8. Apache Spark is an open-source project in big data analytics that completes tasks across clusters of computers. It is the largest and most widely used interface of its kind.
9. Athey, Susan. 2019. “The Impact of Machine Learning on Economics.” In The Economics of Artificial Intelligence: An Agenda, 507–547. University of Chicago Press. http://www.nber.org/chapters/c14009 .
10. Source: https://www.weforum.org/agenda/2018/01/google-ceo-ai-will-be-bigger-than-electricity-or-fire (accessed 3/7/21).
11. To be fully transparent, a shortened version of this chapter was published in the Journal of Industry, Competition and Trade (Weinandy and Ryan 2021). https://rdcu.be/cc8eC .
12. Source: https://www.ssga.com/us/en/institutional/etfs/funds/spdr-sp-500-etf-trust-spy (accessed 1/31/21).
13. The brick-and-mortar retailer profiled in Chapter Three considers itself a “technology company.” Source: author interview on January 27, 2021 with a former corporate officer of realty strategy and innovation for the company.
14. McAfee, Andrew. 2019. More from Less: The Surprising Story of How We Learned to Prosper Using Fewer Resources—and What Happens Next. Scribner.

• Thomas J. Weinandy, Ph.D.
• 2021-05-01
• ~20 min read
• Economics, Retail, Fuel, AI