Think of any specialised profession: tailor, baker, accountant or teacher. How accurately can you describe their work? In fact, if you're among the working stiffs, how would you describe your work?
You might say a tailor sews and mends clothes, a baker bakes, a banker works in the finance sector and a teacher obviously teaches. And what kind of question is that, anyway?
It's a very apt one, thanks for asking. A baker doesn't just bake; s/he measures and mixes ingredients, buys and maintains an inventory of said ingredients, packages and arranges finished products for sale and, finally, interacts with customers and conducts financial transactions.
None of that even touches on cleaning up during the process, maintaining health and safety standards, complying with local laws - everything from keeping the shop clean to how ingredients are stored.
The short of it is: in every profession, a lot goes on behind the scenes.
Superprof now takes you behind the scenes of one particular profession; one so impactful our way of life depends on those workers being right - or as close to right as possible, as often as possible.
We're talking about econometrics, of course. Specifically, applied econometrics - as opposed to theoretical econometrics, which we discuss in a related article.
An Overview of Econometrics
For millennia, anyone with a head for higher-order thinking inevitably turned their thoughts to economics.
From Plato and his student Aristotle to Rene Descartes and John Locke - and, of course, the father of economics as we know it, Adam Smith... all of these great minds and others devised various economic theories to explain human experiences such as power, ownership and inequality.
Unfortunately, just as we regularly debate an idea presented without proof, many of their ideas were either debated or outright contradicted, sometimes by themselves, through their own reasoning. Or, upon discovering that no overarching theory represented every economic phenomenon, they went on to advance different, more elaborate theories.
In God we trust. All others must bring data. - W. Edwards Deming
Therein lay the problem for those early philosopher-economists. Their datasets must have been limited to only what they could observe (or what was reported to them) as the maths that underpin statistics didn't exist until the mid-17th Century. And then, it took another 260 years for anyone to wed the practice of scientific study to economic theory.
Econometrics = economics + metrics; literally: economic measurement
We have Polish economist Pawel Ciompa to thank for being the first to use that word even if Ragnar Frisch usually gets the credit for coining it. It might be logical to credit him with it as he also formulated the terms macroeconomic and microeconomic, and he's renowned for formalising production theory, among his other major accomplishments in the field of economics.
What applied econometrics all boil down to is statistical methods applied to economic data to provide provable, observable relationships of economic phenomena.
To conduct this empirical study, econometrists follow exacting steps. Let's look at them now.
Selecting a Premise
The purpose of analysing economic phenomena is the reduction of uncertainty.
Right now, we're living in fraught economic times. In its early days, this pandemic stopped the production of goods and interrupted supply chains. For the past year, it has stalled economic growth - nobody is shopping, travelling, dining out or buying big-ticket items. And it has disproportionately impacted the poor.
That is far too much uncertainty to pack into a single premise.
When econometrists 'select' their hypothesis, they deliberately narrow their focus to just a few variables at play. In fact, it's not so much the econometrists that select which variables to test; usually, it's the policymakers. They're the ones who want to stay ahead of any economic calamity so they need the most accurate picture of what's going on and what might happen because of it.
To select a hypothesis or examine a phenomenon more closely, the theory that needs to be tested must be defined. Once a definition is formulated, the econometrist (or policymaker) chooses the variables that best illustrate the phenomenon's cause and effect.
For our example, we'll look at the number of restaurants going out of business because they're not earning enough to stay viable during the shutdowns.
That makes the restaurants the dependent variable - they are dependent on people ordering food. The independent variable is consumer spending across the restaurant industry.
Of course, there are other variables to consider, such as the business' overhead, paying staff and buying ingredients. We'll show you how they fit into the equation in just a mo.
Now that a qualitative relationship between consumer spending on restaurant food and restaurants not earning enough has been established, we need to figure out which questions we want answers to.
You should never do anything without a clear idea of what you hope to achieve. Just as we shouldn't invest any money, effort or time into anything that has no clear goal, econometrists lay out which answers they're looking for before they begin their search.
That sounds a bit suspicious, but it's only a bit like saying "Let's drive somewhere!", as opposed to "Let's drive to the beach!". The first one is rather wild and carefree; it promises no certain outcome. The second one provides a specific destination and even says how you'll get there.
Bottom line: even with only two or three variables to consider, econometrists begin with the end in mind by laying out their objectives before they do any calculations.
Doing so helps them stay focused on the three goals of econometrics.
Developing a Model
With goals firmly established, econometrists develop their models. The first one is an economic model, and a rather simple one, at that. At its most basic, it might just be:
dependent variable = f(independent variables)
with f = function of the independent variables
Plugging our variables in, our equation looks like this: restaurant income = f(customer orders, cost of food, employee wages)
Remember, the purpose of econometrics is to conduct a quantitative analysis and/or to reveal economic relationships. That's why one must first develop an economic model - but, again: it needn't be as elaborate as a standard model.
Developing an Econometric Model
Now you need an econometric model to these the model you just developed. This one is going to be much more detailed because it will be a mathematical representation of the qualifiers listed in the economic model.
Invoking again our restaurant scenario, our econometric model - a linear equation will look like this:
restaurant income = ß1 + ß2 (customer orders) + ß3 (cost of food) + ß4 (employee wages) + u
In our econometric model:
- ß1 = other factors affecting restaurant income (increase in takeout boxes and plastic cutlery, for example)
- ß2 = coefficient of customer orders
- ß3 = coefficient of food costs
- ß4 = coefficient of employee wages
- u = random error
This model is much more detailed, with variables to identify various coefficients that may impact the restaurant's bottom line.
Are you wondering about the U?
It's standard practice to include room for error in any statistical calculation. You've likely seen that U if you've ever read any survey results; it's highlighted with 'a +/- margin of error' percentage, usually presented toward the end of the survey results.
You didn't know you'd already had an introduction to econometrics when you read all of those surveys, did you?
Estimating Coefficient Values
At this point, applied econometrics gets a bit more complicated because you have to obtain a value for each coefficient and, for that, you need data.
That information should be collected on the ground; those numbers should be 'real' as opposed to experimental data collected from a controlled environment. Also, the type of data you select matters.
Will it come from a cross-section of the restaurant industry, with some numbers coming from fine dining and others from fast food? Will the econometrist use panel data or chronologically graphed data points?
Several factors go into choosing the right type of data for these calculations, among them the type of study being carried out, the sample size and, always, the time limits econometrists always labour under.
Once the values for all coefficients have been estimated, the econometrist will choose the right statistical method to meet the purpose. And then, it's just a matter of plugging all of the numerical data into the linear equation.
Analysing and Validating Data
With all of the numbers properly placed within the model, it's time for assessment. Do the study's estimates meet expectations? Are the results acceptable?
If the coefficients reflect their anticipated significance, the estimates deliver their expected value and the results mirror the assumptions that were previously established, then you may say that the hypothesis has been validated.
Before tucking that feather in our cap, however, we must affirm the coefficient of determination.
Typically designated R2, it shows the extent to which independent variables impact the dependent variable. This number lies between 0 and 1, and the higher it is, the more fitting the hypothesis.
Of course, this is an overly simplified view of applied econometrics; just a fraction of the overall discipline. To get a better understanding of econometrics and its importance, we'd have to go much deeper into the topic.