I originally intended to write an article of this sort two years ago. However, I recognised that the Dunning-Kruger effect might come into play with the quality of an article of a first-year student. After procrastinating I guess it is prime time in which I take this beloved discipline to judgement; to provide opinions on this discipline’s relevance relative to other disciplines and how it can be improved to better fit the needs of our broader economic systems. Econometrics is arguably the most divisive commerce-based discipline in Monash’s business school. Sadly, many who do get to do econometrics, those who love it
or want to burn it to the ground, do not understand what it exactly is, and what problems the discipline itself is interested in solving.
The soul of this discipline does not come from theoretical economics or the social sciences, but rather, from statistics. To put it simply, econometrics is a bag of statistical methods used to empirically analyse non-experimental data, which we are so fixated on separating with other, nice data for good reason. Non-experimental data is like an ambiguous love interest. They are unreliable, noisy, and often you might gauge that they’re fond of you but are you really sure? or is it because they have hidden intentions? As a result, the discipline has geared its focus towards dealing with the intricacies of economic and financial data where you don’t get to do experiments, have limited or low-quality input and want to exactly identify causal relationships between certain phenomena of this definition. To empirically, and academically test economic relationships and opinions in a sea of subjective theories and politics is what we’re looking to do. Though, we don’t often handle that “academic” and “empirical” spirit that well, especially when conveying information of this sort to undergraduate students.
Common criticisms and mishaps of econometrics
There are quite a few (if not many) criticisms hurled towards the discipline and profession of an econometrician. Many of these would come from esteemed quantitative fields, even from econometricians themselves. Though many criticisms or insults can be classified as redonkulous outcries, some are justified (according to yours truly, me) to a certain degree, which I’ve divided into 3 broad categories:
- Model specification and theory
The essence of many procedures in econometrics starts from specifying an aim, a response variable, several variables that are assumed to constitute towards the “data-generating process”, and performing some tests and inference based on the “model” they’ve created. For this, the principal point of criticism is that underlying variables are usually chosen by the researcher through preconceived theory and assumptions, which may or may not defeat the purpose of statistical inference as a tool to “empirically” examine economic relationships.
To explain this better I want to bring your attention to Tatu Westling’s article “Male Organ and Economic Growth: Does Size Matter?”, in which he found “robust statistical links” to show that the size of the male penis is inversely-related to economic growth, due to “risk-taking”, “testosterone” and “self-esteem”.[i] You’d be absolutely right if this paper’s claim seems like absolute bonkers, without careful undertakings of assumption and theory, econometricians are perfectly capable of “torturing” data until it justifies weird economic propositions. With our confirmation biases and the imperfection of the human intuition, are we also perfectly capable of pointing out illogical implications of econometric research before it’s too late?
- Limitation in scope and practice
Being a standalone department, econometrics and its researchers usually experience silos in communication with the broader statistics community. As a result, advancements and revolutions in statistical methodologies in fields such as machine learning and experimental statistics do not usually roll down to the field of econometrics, and vice versa.
Examples of this would include how econometricians have not utilised non-parametric (model-based) methods extensively in their research undertakings (until now). This further heightens the learning curve when it comes to adopting big data methodologies developed by artificial intelligence researchers that are non-parametric in nature. Conversely, useful concepts that are taught to analyse time series in econometrics (ETC3450/3460 in Monash if you’re interested) such as cointegration and ARCH frameworks that are useful in analysing long-term relationships between economic variables have also not trickled-down to the broader statistics community.
Perhaps this criticism shouldn’t be directed solely at econometrics, but it would be helpful if researchers from multiple disciplines addressed gaps in communication such as these.
- Teaching econometrics
It is obvious that the future of econometrics would lie on the shoulders of aspiring researchers and econometricians, who are going through their undergraduate education. Often, it is easy for educators to overlook the setbacks of econometrics to teach them in traditional approaches, even when taught models are under repeated scrutiny.
With what we know about economic data and appropriate treatment of frequentist and Bayesian statistics, it is unfortunate that many institutions lag behind when it comes to revamping their curriculum in a favourable way for the discipline by sticking to black-box and “theory is gospel” approaches to education in economics, let alone econometrics itself.
The value of education in econometrics would hence not be conducive for those who are practising in the industry, since they would find it hard to anticipate the depth of complexity of economic data and variables. It would also not be favourable for aspiring academics, as they would not develop the intuition and drive to question already obsolete methods they’ve learnt as an undergraduate early on.
Significant innovative (almost entrepreneurial) development in econometrics needs to take place in how it’s investigated, improved and taught to ensure its own relevance and survival relative to the myriad complex information bestowed upon econometricians.
Econometrics’ survival is especially important and essential to the broader field of economics and finance, as our way of structuring empirical arguments are optimised for the industry, relative to many other disciplines that are trying to solve similar problems. Fortunately, coming from one of the best econometrics departments in the world, we can already see innovative developments and interdisciplinary communication to improve the current state of research and education in econometrics taking place. From adapting non-parametric methods and methods of handling big data used in other disciplines, to dismantling the black-box method of conveying information to the future generation of econometrics students, I am confident that econometrics is in good hands, and will continue to flourish, and be relevant.
[i] Westling, T. (2011, July). Male organ and economic growth, does size matter? Retrieved from https://helda.helsinki.fi/bitstream/handle/10138/27239/maleorga.pdf