Modern economics has masqueraded as a “science” for many decades now, but at its heart the discipline of economics is based on words and theories, not numbers. In fact, a strong argument can be made that economics is the study of the intersection of finance and culture rather than “the science that deals with the production, distribution, and consumption of goods and services.”
Economics as it is generally understood today began as a branch of philosophy in the 17th and 18th century, with famous political philosophers and theorists such as John Locke and David Hume laying the groundwork that Adam Smith and others would systematize and rigorize into classical economics. However you want to categorize economics as an academic discipline, make no mistake about it, unlike in physics, chemistry or biology, research in economics is not based on the scientific method.
2020 Letter: Kerrisdale Outlines Long Thesis For This ESG Tech Stock
Sahm Adrangi's Kerrisdale Capital was up 6.5% for the fourth quarter, including a decline of 0.3% in October and gains of 3.2% and 3.5% for November and December, respectively. For comparison, the S&P 500 gained 12.2% during the fourth quarter, while the Barclay Hedge Fund Index was up 9.3%. Q4 2020 hedge fund letters, conferences Read More
Of interest, econometrics is branch of economics that applies mathematical methods (statistics) to describe economic systems.
Economics’ math problem
Noah Smith of Bloomberg View says that the discipline of economics has a problem in the way it uses math. In his September 1st article, Smith argues it appears that modern economics has been trying to evolve in to a subfield of applied math, but also notes that “applied math disciplines — computational biology, fluid dynamics, quantitative finance — mathematical theories are always tied to the evidence. If a theory hasn’t been tested, it’s treated as pure conjecture.”
Smith then goes on to point out that the theoretical framework of economics puts the cart in front of the horse in terms of applying the scientific method. “Traditionally, economists have put the facts in a subordinate role and theory in the driver’s seat. Plausible-sounding theories are believed to be true unless proven false, while empirical facts are often dismissed if they don’t make sense in the context of leading theories.”
He goes on to note that nothing has really changed in the theoretical approach of economics over the intervening centuries, and simply adding statistics to an existing theoretical framework does not make it “scientific”. . Smith writes: “…it was just as true back when economics theories were written out in long literary volumes. Econ developed as a form of philosophy and then added math later, becoming basically a form of mathematical philosophy.”
Push towards machine learning reflects the new direction of economics
The surge of interest in machine learning in the field of economics is certainly related to the trend towards empiricism. Machine learning can be generally defined as a set of statistical data analysis techniques to identify key features of the data not using a specific theory. As Smith says: “…machine learning “lets the data speak.” He also highlights that with Big Data moving to the fore today, machine learning is suddenly a hot field, and has become a key tool in the exploding field of data science.
In his piece, Smith also highlights the ongoing machine learning research of Stanford economists Susan Athey and Guido Imbens. He notes the two academics discussed machine learning techniques to at a recent meeting of the National Bureau of Economic Research. One key point they made was that machine learning techniques emphasized causality notably less than traditional economic statistical techniques (ie, econometrics).
This is a big deal, for what it really means is that machine learning in economics is more focused on forecasting than trying to comprehend the specific impacts of policy.
While this focus may make machine learning less interesting to some economists, given they are more focused on giving policy recommendations than making forecasts. That said, Athey and Imbens have been developing machine learning techniques that could be used to isolate causal effects. This, of course, would eventually permit economists to make conclusions about policy implications.
Smith describes Athey and Imbens approach as wrestling with the problem of how to identify treatment effects. He offers a brief explanation of the idea of a treatment effect, making it obvious why the new method would be of interest to many economists: “A treatment effect is the difference between what would happen if you administer some “treatment” — say, raising the minimum wage — and what would happen without the treatment.”