National open university of nigeria introduction to econometrics I eco 355

 
 
 

---

28 
3.4. THE DIFFERENCE BETWEEN ECONOMETRICS MODELING AND 
MACHINE LEARNING 
Econometric models are statistical models used in econometrics. An econometric 
model specifies the statistical relationship that is believed to be held between the various 
economic quantities pertaining to a particular economic phenomenon under study. 
On the other hand- Machine learning is a scientific discipline that explores the 
construction and study of algorithms that can learn from data. So that makes a clear 
distinction right? If it learns on its own from data it is machine learning. If it is used for 
economic phenomenon it is an econometric model. However the confusion arises in the 
way these two paradigms are championed. The computer science major will always say 
machine learning and the statistical major will always emphasize modeling. Since 
computer science majors now rule at face book, Google and almost every technology 
company, you would think that machine learning is dominating the field and beating poor 
old econometric modeling. 
But what if you can make econometric models learn from data? 
Lets dig more into these algorithms. The way machine learning works is to optimize 
some particular quantity, say cost. A loss function or cost function is a function that maps 
a value(s) of one or more variables intuitively representing some ―cost‖ associated with 
the event. An optimization problem seeks to minimize a loss function. Machine learning 
frequently seek optimization to get the best of many alternatives. 
Now, cost or loss holds different meanings in econometric modeling. In econometric 
modeling we are trying to minimize the error (or root mean squared error). Root mean 
squared error means root of the sum of squares of errors. An error is defined as the 
difference between actual and predicted value by the model for previous data. 
The difference in the jargon is solely in the way statisticians and computer scientists are 
trained. Computer scientists try to compensate for both actual error as well as 
computational cost – that is the time taken to run a particular algorithm. On the other 
hand statisticians are trained primarily to think in terms of confidence levels or error in 
terms or predicted and actual without caring for the time taken to run for the model. 
That is why data science is defined often as an intersection between hacking skills (in 
computer science) and statistical knowledge (and math). Something like K Means 
clustering can be taught in two different ways just like regression can be based on these 
two approaches. I wrote back to my colleague in Marketing – we have data scientists. 
They are trained in both econometric modeling and machine learning. I looked back and 
had a beer. If university professors don‘t shed their departmental attitudes towards data 

---

29 
science, we will have a very confused set of students very shortly arguing without 
knowing how close they actually are.